diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,60025 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.5032620787108024,
+  "global_step": 100000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 2.3859,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 2.4248,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5e-06,
+      "loss": 2.3962,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 2.293,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5e-06,
+      "loss": 2.3082,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3e-06,
+      "loss": 2.2345,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.5000000000000004e-06,
+      "loss": 2.2395,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 2.1728,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.5e-06,
+      "loss": 2.1249,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5e-06,
+      "loss": 2.0972,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 2.0611,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6e-06,
+      "loss": 2.0679,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.5000000000000004e-06,
+      "loss": 2.0544,
+      "step": 130
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.000000000000001e-06,
+      "loss": 2.0341,
+      "step": 140
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.5e-06,
+      "loss": 2.0085,
+      "step": 150
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 1.9927,
+      "step": 160
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.500000000000002e-06,
+      "loss": 1.9885,
+      "step": 170
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9e-06,
+      "loss": 1.9823,
+      "step": 180
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.5e-06,
+      "loss": 1.995,
+      "step": 190
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1e-05,
+      "loss": 1.9417,
+      "step": 200
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.05e-05,
+      "loss": 1.9702,
+      "step": 210
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 1.922,
+      "step": 220
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1500000000000002e-05,
+      "loss": 1.9387,
+      "step": 230
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2e-05,
+      "loss": 1.9228,
+      "step": 240
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.25e-05,
+      "loss": 1.9337,
+      "step": 250
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 1.9142,
+      "step": 260
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3500000000000001e-05,
+      "loss": 1.8961,
+      "step": 270
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4000000000000001e-05,
+      "loss": 1.9082,
+      "step": 280
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.45e-05,
+      "loss": 1.8773,
+      "step": 290
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5e-05,
+      "loss": 1.8883,
+      "step": 300
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.55e-05,
+      "loss": 1.868,
+      "step": 310
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 1.8533,
+      "step": 320
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.65e-05,
+      "loss": 1.8577,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 1.8503,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.75e-05,
+      "loss": 1.8663,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8e-05,
+      "loss": 1.8645,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.85e-05,
+      "loss": 1.8565,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9e-05,
+      "loss": 1.8337,
+      "step": 380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9500000000000003e-05,
+      "loss": 1.8413,
+      "step": 390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2e-05,
+      "loss": 1.8264,
+      "step": 400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.05e-05,
+      "loss": 1.8526,
+      "step": 410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1e-05,
+      "loss": 1.8355,
+      "step": 420
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.15e-05,
+      "loss": 1.8277,
+      "step": 430
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 1.8201,
+      "step": 440
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.25e-05,
+      "loss": 1.811,
+      "step": 450
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3000000000000003e-05,
+      "loss": 1.8282,
+      "step": 460
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.35e-05,
+      "loss": 1.8144,
+      "step": 470
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4e-05,
+      "loss": 1.8357,
+      "step": 480
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.45e-05,
+      "loss": 1.7797,
+      "step": 490
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5e-05,
+      "loss": 1.8027,
+      "step": 500
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5500000000000003e-05,
+      "loss": 1.8005,
+      "step": 510
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6000000000000002e-05,
+      "loss": 1.7927,
+      "step": 520
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6500000000000004e-05,
+      "loss": 1.805,
+      "step": 530
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7000000000000002e-05,
+      "loss": 1.8162,
+      "step": 540
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7500000000000004e-05,
+      "loss": 1.8254,
+      "step": 550
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 1.7969,
+      "step": 560
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8499999999999998e-05,
+      "loss": 1.7829,
+      "step": 570
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9e-05,
+      "loss": 1.7998,
+      "step": 580
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.95e-05,
+      "loss": 1.8069,
+      "step": 590
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3e-05,
+      "loss": 1.7736,
+      "step": 600
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.05e-05,
+      "loss": 1.7919,
+      "step": 610
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1e-05,
+      "loss": 1.7858,
+      "step": 620
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.15e-05,
+      "loss": 1.7671,
+      "step": 630
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 1.799,
+      "step": 640
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2500000000000004e-05,
+      "loss": 1.7759,
+      "step": 650
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3e-05,
+      "loss": 1.7733,
+      "step": 660
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.35e-05,
+      "loss": 1.8027,
+      "step": 670
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 1.7611,
+      "step": 680
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.45e-05,
+      "loss": 1.7862,
+      "step": 690
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5e-05,
+      "loss": 1.7479,
+      "step": 700
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.55e-05,
+      "loss": 1.7596,
+      "step": 710
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6e-05,
+      "loss": 1.759,
+      "step": 720
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.65e-05,
+      "loss": 1.7892,
+      "step": 730
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7e-05,
+      "loss": 1.7711,
+      "step": 740
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7500000000000003e-05,
+      "loss": 1.7686,
+      "step": 750
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8e-05,
+      "loss": 1.7678,
+      "step": 760
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.85e-05,
+      "loss": 1.7617,
+      "step": 770
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9000000000000006e-05,
+      "loss": 1.7645,
+      "step": 780
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9500000000000005e-05,
+      "loss": 1.7407,
+      "step": 790
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4e-05,
+      "loss": 1.7896,
+      "step": 800
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.05e-05,
+      "loss": 1.7757,
+      "step": 810
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.1e-05,
+      "loss": 1.7546,
+      "step": 820
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.15e-05,
+      "loss": 1.7701,
+      "step": 830
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.2e-05,
+      "loss": 1.758,
+      "step": 840
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.25e-05,
+      "loss": 1.7491,
+      "step": 850
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3e-05,
+      "loss": 1.7431,
+      "step": 860
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.35e-05,
+      "loss": 1.7389,
+      "step": 870
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 1.7661,
+      "step": 880
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4500000000000004e-05,
+      "loss": 1.7566,
+      "step": 890
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.5e-05,
+      "loss": 1.752,
+      "step": 900
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.55e-05,
+      "loss": 1.7374,
+      "step": 910
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 1.7381,
+      "step": 920
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6500000000000005e-05,
+      "loss": 1.7635,
+      "step": 930
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.7e-05,
+      "loss": 1.7354,
+      "step": 940
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.75e-05,
+      "loss": 1.7467,
+      "step": 950
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8e-05,
+      "loss": 1.7591,
+      "step": 960
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.85e-05,
+      "loss": 1.7569,
+      "step": 970
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9e-05,
+      "loss": 1.772,
+      "step": 980
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9500000000000004e-05,
+      "loss": 1.7421,
+      "step": 990
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5e-05,
+      "loss": 1.714,
+      "step": 1000
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.99949494949495e-05,
+      "loss": 1.7284,
+      "step": 1010
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9989898989898994e-05,
+      "loss": 1.7509,
+      "step": 1020
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.998484848484849e-05,
+      "loss": 1.755,
+      "step": 1030
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.997979797979798e-05,
+      "loss": 1.7603,
+      "step": 1040
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.997474747474748e-05,
+      "loss": 1.7558,
+      "step": 1050
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.996969696969697e-05,
+      "loss": 1.7417,
+      "step": 1060
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.996464646464647e-05,
+      "loss": 1.7464,
+      "step": 1070
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9959595959595964e-05,
+      "loss": 1.7629,
+      "step": 1080
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.995454545454546e-05,
+      "loss": 1.7347,
+      "step": 1090
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.994949494949495e-05,
+      "loss": 1.7364,
+      "step": 1100
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.994444444444445e-05,
+      "loss": 1.7307,
+      "step": 1110
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.993939393939394e-05,
+      "loss": 1.7203,
+      "step": 1120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.993434343434344e-05,
+      "loss": 1.7295,
+      "step": 1130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.992929292929293e-05,
+      "loss": 1.7422,
+      "step": 1140
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.992424242424243e-05,
+      "loss": 1.7292,
+      "step": 1150
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.991919191919192e-05,
+      "loss": 1.7553,
+      "step": 1160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.991414141414142e-05,
+      "loss": 1.7143,
+      "step": 1170
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.990909090909091e-05,
+      "loss": 1.7428,
+      "step": 1180
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9904040404040406e-05,
+      "loss": 1.7412,
+      "step": 1190
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.98989898989899e-05,
+      "loss": 1.737,
+      "step": 1200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.98939393939394e-05,
+      "loss": 1.7611,
+      "step": 1210
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9888888888888894e-05,
+      "loss": 1.7197,
+      "step": 1220
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.988383838383839e-05,
+      "loss": 1.7382,
+      "step": 1230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.987878787878788e-05,
+      "loss": 1.7258,
+      "step": 1240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9873737373737375e-05,
+      "loss": 1.7281,
+      "step": 1250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.986868686868687e-05,
+      "loss": 1.7311,
+      "step": 1260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.986363636363637e-05,
+      "loss": 1.7248,
+      "step": 1270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.985858585858586e-05,
+      "loss": 1.7318,
+      "step": 1280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.985353535353536e-05,
+      "loss": 1.7184,
+      "step": 1290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.984848484848485e-05,
+      "loss": 1.738,
+      "step": 1300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9843434343434344e-05,
+      "loss": 1.7415,
+      "step": 1310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.983838383838384e-05,
+      "loss": 1.7202,
+      "step": 1320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9833333333333336e-05,
+      "loss": 1.7316,
+      "step": 1330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.982828282828283e-05,
+      "loss": 1.7342,
+      "step": 1340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.982323232323233e-05,
+      "loss": 1.7294,
+      "step": 1350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.981818181818182e-05,
+      "loss": 1.7419,
+      "step": 1360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.981313131313132e-05,
+      "loss": 1.7253,
+      "step": 1370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.980808080808081e-05,
+      "loss": 1.7021,
+      "step": 1380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9803030303030305e-05,
+      "loss": 1.7091,
+      "step": 1390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.97979797979798e-05,
+      "loss": 1.7341,
+      "step": 1400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.97929292929293e-05,
+      "loss": 1.7301,
+      "step": 1410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9787878787878787e-05,
+      "loss": 1.7223,
+      "step": 1420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.978282828282829e-05,
+      "loss": 1.7146,
+      "step": 1430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.977777777777778e-05,
+      "loss": 1.7109,
+      "step": 1440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9772727272727275e-05,
+      "loss": 1.7211,
+      "step": 1450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.976767676767677e-05,
+      "loss": 1.7566,
+      "step": 1460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9762626262626267e-05,
+      "loss": 1.7203,
+      "step": 1470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9757575757575756e-05,
+      "loss": 1.7357,
+      "step": 1480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.975252525252526e-05,
+      "loss": 1.7236,
+      "step": 1490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.974747474747475e-05,
+      "loss": 1.7334,
+      "step": 1500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9742424242424244e-05,
+      "loss": 1.7181,
+      "step": 1510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.973737373737374e-05,
+      "loss": 1.7511,
+      "step": 1520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9732323232323236e-05,
+      "loss": 1.735,
+      "step": 1530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9727272727272725e-05,
+      "loss": 1.7399,
+      "step": 1540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.972222222222223e-05,
+      "loss": 1.728,
+      "step": 1550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.971717171717172e-05,
+      "loss": 1.7038,
+      "step": 1560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.971212121212121e-05,
+      "loss": 1.7156,
+      "step": 1570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.970707070707071e-05,
+      "loss": 1.7087,
+      "step": 1580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9702020202020205e-05,
+      "loss": 1.7105,
+      "step": 1590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9696969696969694e-05,
+      "loss": 1.7299,
+      "step": 1600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.96919191919192e-05,
+      "loss": 1.7026,
+      "step": 1610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9686868686868686e-05,
+      "loss": 1.6942,
+      "step": 1620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.968181818181818e-05,
+      "loss": 1.7014,
+      "step": 1630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.967676767676768e-05,
+      "loss": 1.7221,
+      "step": 1640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.9671717171717174e-05,
+      "loss": 1.6948,
+      "step": 1650
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 4.966666666666667e-05,
+      "loss": 1.7078,
+      "step": 1660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9661616161616166e-05,
+      "loss": 1.7111,
+      "step": 1670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9656565656565655e-05,
+      "loss": 1.6857,
+      "step": 1680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.965151515151515e-05,
+      "loss": 1.7298,
+      "step": 1690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.964646464646465e-05,
+      "loss": 1.7265,
+      "step": 1700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.964141414141414e-05,
+      "loss": 1.7249,
+      "step": 1710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.963636363636364e-05,
+      "loss": 1.6685,
+      "step": 1720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9631313131313135e-05,
+      "loss": 1.7268,
+      "step": 1730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9626262626262624e-05,
+      "loss": 1.713,
+      "step": 1740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.962121212121213e-05,
+      "loss": 1.7104,
+      "step": 1750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9616161616161616e-05,
+      "loss": 1.6992,
+      "step": 1760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.961111111111111e-05,
+      "loss": 1.6948,
+      "step": 1770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.960606060606061e-05,
+      "loss": 1.7052,
+      "step": 1780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9601010101010104e-05,
+      "loss": 1.685,
+      "step": 1790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9595959595959594e-05,
+      "loss": 1.7155,
+      "step": 1800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9590909090909096e-05,
+      "loss": 1.7429,
+      "step": 1810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9585858585858586e-05,
+      "loss": 1.6912,
+      "step": 1820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.958080808080808e-05,
+      "loss": 1.707,
+      "step": 1830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.957575757575758e-05,
+      "loss": 1.7047,
+      "step": 1840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9570707070707073e-05,
+      "loss": 1.7287,
+      "step": 1850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.956565656565657e-05,
+      "loss": 1.706,
+      "step": 1860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9560606060606065e-05,
+      "loss": 1.6727,
+      "step": 1870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.955555555555556e-05,
+      "loss": 1.7009,
+      "step": 1880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.955050505050505e-05,
+      "loss": 1.6847,
+      "step": 1890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9545454545454553e-05,
+      "loss": 1.7243,
+      "step": 1900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.954040404040404e-05,
+      "loss": 1.7253,
+      "step": 1910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.953535353535354e-05,
+      "loss": 1.6957,
+      "step": 1920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9530303030303035e-05,
+      "loss": 1.6947,
+      "step": 1930
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.952525252525253e-05,
+      "loss": 1.7193,
+      "step": 1940
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.952020202020202e-05,
+      "loss": 1.7126,
+      "step": 1950
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.951515151515152e-05,
+      "loss": 1.7098,
+      "step": 1960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.951010101010101e-05,
+      "loss": 1.7156,
+      "step": 1970
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.950505050505051e-05,
+      "loss": 1.705,
+      "step": 1980
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9500000000000004e-05,
+      "loss": 1.6952,
+      "step": 1990
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.94949494949495e-05,
+      "loss": 1.7037,
+      "step": 2000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.948989898989899e-05,
+      "loss": 1.6809,
+      "step": 2010
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.948484848484849e-05,
+      "loss": 1.6807,
+      "step": 2020
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.947979797979798e-05,
+      "loss": 1.6734,
+      "step": 2030
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.947474747474748e-05,
+      "loss": 1.724,
+      "step": 2040
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.946969696969697e-05,
+      "loss": 1.682,
+      "step": 2050
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.946464646464647e-05,
+      "loss": 1.7049,
+      "step": 2060
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.945959595959596e-05,
+      "loss": 1.6993,
+      "step": 2070
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.945454545454546e-05,
+      "loss": 1.7018,
+      "step": 2080
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.944949494949495e-05,
+      "loss": 1.6812,
+      "step": 2090
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9444444444444446e-05,
+      "loss": 1.6906,
+      "step": 2100
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.943939393939394e-05,
+      "loss": 1.6737,
+      "step": 2110
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.943434343434344e-05,
+      "loss": 1.7029,
+      "step": 2120
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9429292929292934e-05,
+      "loss": 1.6934,
+      "step": 2130
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.942424242424243e-05,
+      "loss": 1.702,
+      "step": 2140
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.941919191919192e-05,
+      "loss": 1.7084,
+      "step": 2150
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9414141414141415e-05,
+      "loss": 1.7036,
+      "step": 2160
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.940909090909091e-05,
+      "loss": 1.7063,
+      "step": 2170
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.940404040404041e-05,
+      "loss": 1.6892,
+      "step": 2180
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.93989898989899e-05,
+      "loss": 1.6794,
+      "step": 2190
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.93939393939394e-05,
+      "loss": 1.6871,
+      "step": 2200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.938888888888889e-05,
+      "loss": 1.6869,
+      "step": 2210
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9383838383838384e-05,
+      "loss": 1.7052,
+      "step": 2220
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.937878787878788e-05,
+      "loss": 1.7086,
+      "step": 2230
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9373737373737376e-05,
+      "loss": 1.6953,
+      "step": 2240
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.936868686868687e-05,
+      "loss": 1.7103,
+      "step": 2250
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.936363636363637e-05,
+      "loss": 1.7079,
+      "step": 2260
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.935858585858586e-05,
+      "loss": 1.6991,
+      "step": 2270
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.935353535353536e-05,
+      "loss": 1.7019,
+      "step": 2280
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.934848484848485e-05,
+      "loss": 1.6868,
+      "step": 2290
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.9343434343434346e-05,
+      "loss": 1.6926,
+      "step": 2300
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.933838383838384e-05,
+      "loss": 1.6824,
+      "step": 2310
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.933333333333334e-05,
+      "loss": 1.6867,
+      "step": 2320
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.932828282828283e-05,
+      "loss": 1.6972,
+      "step": 2330
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.932323232323233e-05,
+      "loss": 1.6731,
+      "step": 2340
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.931818181818182e-05,
+      "loss": 1.7111,
+      "step": 2350
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9313131313131315e-05,
+      "loss": 1.6529,
+      "step": 2360
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.930808080808081e-05,
+      "loss": 1.6785,
+      "step": 2370
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.930303030303031e-05,
+      "loss": 1.6767,
+      "step": 2380
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9297979797979796e-05,
+      "loss": 1.7015,
+      "step": 2390
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.92929292929293e-05,
+      "loss": 1.7027,
+      "step": 2400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.928787878787879e-05,
+      "loss": 1.6933,
+      "step": 2410
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9282828282828284e-05,
+      "loss": 1.6941,
+      "step": 2420
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.927777777777778e-05,
+      "loss": 1.6838,
+      "step": 2430
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9272727272727276e-05,
+      "loss": 1.677,
+      "step": 2440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9267676767676765e-05,
+      "loss": 1.6852,
+      "step": 2450
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.926262626262627e-05,
+      "loss": 1.6961,
+      "step": 2460
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.925757575757576e-05,
+      "loss": 1.6789,
+      "step": 2470
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.925252525252525e-05,
+      "loss": 1.6558,
+      "step": 2480
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.924747474747475e-05,
+      "loss": 1.6967,
+      "step": 2490
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9242424242424245e-05,
+      "loss": 1.7008,
+      "step": 2500
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9237373737373734e-05,
+      "loss": 1.6772,
+      "step": 2510
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.923232323232324e-05,
+      "loss": 1.6842,
+      "step": 2520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9227272727272726e-05,
+      "loss": 1.6625,
+      "step": 2530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.922222222222222e-05,
+      "loss": 1.6813,
+      "step": 2540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.921717171717172e-05,
+      "loss": 1.6693,
+      "step": 2550
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9212121212121214e-05,
+      "loss": 1.6876,
+      "step": 2560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.920707070707071e-05,
+      "loss": 1.681,
+      "step": 2570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9202020202020206e-05,
+      "loss": 1.6758,
+      "step": 2580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9196969696969695e-05,
+      "loss": 1.6962,
+      "step": 2590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.919191919191919e-05,
+      "loss": 1.7019,
+      "step": 2600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.918686868686869e-05,
+      "loss": 1.7146,
+      "step": 2610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9181818181818183e-05,
+      "loss": 1.6702,
+      "step": 2620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.917676767676768e-05,
+      "loss": 1.6831,
+      "step": 2630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9171717171717175e-05,
+      "loss": 1.6871,
+      "step": 2640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9166666666666665e-05,
+      "loss": 1.6771,
+      "step": 2650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.916161616161617e-05,
+      "loss": 1.6857,
+      "step": 2660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9156565656565657e-05,
+      "loss": 1.6757,
+      "step": 2670
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.915151515151515e-05,
+      "loss": 1.7256,
+      "step": 2680
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.914646464646465e-05,
+      "loss": 1.6961,
+      "step": 2690
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9141414141414145e-05,
+      "loss": 1.6832,
+      "step": 2700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.913636363636364e-05,
+      "loss": 1.6756,
+      "step": 2710
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9131313131313137e-05,
+      "loss": 1.6671,
+      "step": 2720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.912626262626263e-05,
+      "loss": 1.6918,
+      "step": 2730
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.912121212121212e-05,
+      "loss": 1.6845,
+      "step": 2740
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9116161616161625e-05,
+      "loss": 1.6701,
+      "step": 2750
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9111111111111114e-05,
+      "loss": 1.6746,
+      "step": 2760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.910606060606061e-05,
+      "loss": 1.6643,
+      "step": 2770
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9101010101010106e-05,
+      "loss": 1.6962,
+      "step": 2780
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.90959595959596e-05,
+      "loss": 1.6791,
+      "step": 2790
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.909090909090909e-05,
+      "loss": 1.6711,
+      "step": 2800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9085858585858594e-05,
+      "loss": 1.6633,
+      "step": 2810
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.908080808080808e-05,
+      "loss": 1.6704,
+      "step": 2820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.907575757575758e-05,
+      "loss": 1.6868,
+      "step": 2830
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9070707070707075e-05,
+      "loss": 1.6727,
+      "step": 2840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.906565656565657e-05,
+      "loss": 1.6447,
+      "step": 2850
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.906060606060606e-05,
+      "loss": 1.6809,
+      "step": 2860
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.905555555555556e-05,
+      "loss": 1.6669,
+      "step": 2870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.905050505050505e-05,
+      "loss": 1.6816,
+      "step": 2880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.904545454545455e-05,
+      "loss": 1.6692,
+      "step": 2890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9040404040404044e-05,
+      "loss": 1.6679,
+      "step": 2900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.903535353535354e-05,
+      "loss": 1.6839,
+      "step": 2910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.903030303030303e-05,
+      "loss": 1.6898,
+      "step": 2920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.902525252525253e-05,
+      "loss": 1.6991,
+      "step": 2930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.902020202020202e-05,
+      "loss": 1.6943,
+      "step": 2940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.901515151515152e-05,
+      "loss": 1.6597,
+      "step": 2950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.901010101010101e-05,
+      "loss": 1.6634,
+      "step": 2960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.900505050505051e-05,
+      "loss": 1.6687,
+      "step": 2970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.9e-05,
+      "loss": 1.6844,
+      "step": 2980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.89949494949495e-05,
+      "loss": 1.7057,
+      "step": 2990
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.898989898989899e-05,
+      "loss": 1.682,
+      "step": 3000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8984848484848486e-05,
+      "loss": 1.6731,
+      "step": 3010
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.897979797979798e-05,
+      "loss": 1.6619,
+      "step": 3020
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.897474747474748e-05,
+      "loss": 1.6682,
+      "step": 3030
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.896969696969697e-05,
+      "loss": 1.6631,
+      "step": 3040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.896464646464647e-05,
+      "loss": 1.6799,
+      "step": 3050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.895959595959596e-05,
+      "loss": 1.6814,
+      "step": 3060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8954545454545456e-05,
+      "loss": 1.6771,
+      "step": 3070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.894949494949495e-05,
+      "loss": 1.6785,
+      "step": 3080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.894444444444445e-05,
+      "loss": 1.6541,
+      "step": 3090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8939393939393944e-05,
+      "loss": 1.6602,
+      "step": 3100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.893434343434344e-05,
+      "loss": 1.6627,
+      "step": 3110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.892929292929293e-05,
+      "loss": 1.6866,
+      "step": 3120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8924242424242425e-05,
+      "loss": 1.6946,
+      "step": 3130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.891919191919192e-05,
+      "loss": 1.6798,
+      "step": 3140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.891414141414142e-05,
+      "loss": 1.6546,
+      "step": 3150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.890909090909091e-05,
+      "loss": 1.6715,
+      "step": 3160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.890404040404041e-05,
+      "loss": 1.6732,
+      "step": 3170
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.88989898989899e-05,
+      "loss": 1.6616,
+      "step": 3180
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.88939393939394e-05,
+      "loss": 1.6716,
+      "step": 3190
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.888888888888889e-05,
+      "loss": 1.6543,
+      "step": 3200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8883838383838386e-05,
+      "loss": 1.6651,
+      "step": 3210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.887878787878788e-05,
+      "loss": 1.6608,
+      "step": 3220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.887373737373738e-05,
+      "loss": 1.6736,
+      "step": 3230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.886868686868687e-05,
+      "loss": 1.6648,
+      "step": 3240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.886363636363637e-05,
+      "loss": 1.6777,
+      "step": 3250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.885858585858586e-05,
+      "loss": 1.6731,
+      "step": 3260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8853535353535355e-05,
+      "loss": 1.6786,
+      "step": 3270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.884848484848485e-05,
+      "loss": 1.6688,
+      "step": 3280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.884343434343435e-05,
+      "loss": 1.6511,
+      "step": 3290
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8838383838383836e-05,
+      "loss": 1.6323,
+      "step": 3300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.883333333333334e-05,
+      "loss": 1.6872,
+      "step": 3310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.882828282828283e-05,
+      "loss": 1.6557,
+      "step": 3320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8823232323232324e-05,
+      "loss": 1.6818,
+      "step": 3330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.881818181818182e-05,
+      "loss": 1.6654,
+      "step": 3340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8813131313131316e-05,
+      "loss": 1.6855,
+      "step": 3350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8808080808080805e-05,
+      "loss": 1.6832,
+      "step": 3360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.880303030303031e-05,
+      "loss": 1.663,
+      "step": 3370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.87979797979798e-05,
+      "loss": 1.6512,
+      "step": 3380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.879292929292929e-05,
+      "loss": 1.6717,
+      "step": 3390
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.878787878787879e-05,
+      "loss": 1.6528,
+      "step": 3400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8782828282828285e-05,
+      "loss": 1.6782,
+      "step": 3410
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8777777777777775e-05,
+      "loss": 1.6654,
+      "step": 3420
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.877272727272728e-05,
+      "loss": 1.6833,
+      "step": 3430
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8767676767676767e-05,
+      "loss": 1.6662,
+      "step": 3440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.876262626262626e-05,
+      "loss": 1.6677,
+      "step": 3450
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.875757575757576e-05,
+      "loss": 1.6683,
+      "step": 3460
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8752525252525255e-05,
+      "loss": 1.6527,
+      "step": 3470
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.874747474747475e-05,
+      "loss": 1.6498,
+      "step": 3480
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8742424242424246e-05,
+      "loss": 1.6842,
+      "step": 3490
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8737373737373736e-05,
+      "loss": 1.6794,
+      "step": 3500
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.873232323232323e-05,
+      "loss": 1.6509,
+      "step": 3510
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.872727272727273e-05,
+      "loss": 1.6545,
+      "step": 3520
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8722222222222224e-05,
+      "loss": 1.6478,
+      "step": 3530
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.871717171717172e-05,
+      "loss": 1.6675,
+      "step": 3540
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8712121212121216e-05,
+      "loss": 1.6562,
+      "step": 3550
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8707070707070705e-05,
+      "loss": 1.6731,
+      "step": 3560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.870202020202021e-05,
+      "loss": 1.6605,
+      "step": 3570
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8696969696969704e-05,
+      "loss": 1.678,
+      "step": 3580
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.869191919191919e-05,
+      "loss": 1.6822,
+      "step": 3590
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.868686868686869e-05,
+      "loss": 1.6447,
+      "step": 3600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.8681818181818185e-05,
+      "loss": 1.6933,
+      "step": 3610
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.867676767676768e-05,
+      "loss": 1.6572,
+      "step": 3620
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.867171717171718e-05,
+      "loss": 1.6578,
+      "step": 3630
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.866666666666667e-05,
+      "loss": 1.6721,
+      "step": 3640
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 4.866161616161616e-05,
+      "loss": 1.6798,
+      "step": 3650
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.865656565656566e-05,
+      "loss": 1.676,
+      "step": 3660
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8651515151515154e-05,
+      "loss": 1.6605,
+      "step": 3670
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.864646464646465e-05,
+      "loss": 1.6575,
+      "step": 3680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8641414141414146e-05,
+      "loss": 1.6909,
+      "step": 3690
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.863636363636364e-05,
+      "loss": 1.692,
+      "step": 3700
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.863131313131313e-05,
+      "loss": 1.6913,
+      "step": 3710
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8626262626262634e-05,
+      "loss": 1.6853,
+      "step": 3720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.862121212121212e-05,
+      "loss": 1.6743,
+      "step": 3730
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.861616161616162e-05,
+      "loss": 1.6654,
+      "step": 3740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8611111111111115e-05,
+      "loss": 1.654,
+      "step": 3750
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.860606060606061e-05,
+      "loss": 1.6586,
+      "step": 3760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.86010101010101e-05,
+      "loss": 1.6569,
+      "step": 3770
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.85959595959596e-05,
+      "loss": 1.6674,
+      "step": 3780
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.859090909090909e-05,
+      "loss": 1.6614,
+      "step": 3790
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.858585858585859e-05,
+      "loss": 1.67,
+      "step": 3800
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8580808080808084e-05,
+      "loss": 1.6461,
+      "step": 3810
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.857575757575758e-05,
+      "loss": 1.6811,
+      "step": 3820
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.857070707070707e-05,
+      "loss": 1.6756,
+      "step": 3830
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.856565656565657e-05,
+      "loss": 1.6588,
+      "step": 3840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.856060606060606e-05,
+      "loss": 1.6565,
+      "step": 3850
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.855555555555556e-05,
+      "loss": 1.671,
+      "step": 3860
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8550505050505053e-05,
+      "loss": 1.6845,
+      "step": 3870
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.854545454545455e-05,
+      "loss": 1.6652,
+      "step": 3880
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.854040404040404e-05,
+      "loss": 1.6908,
+      "step": 3890
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.853535353535354e-05,
+      "loss": 1.6584,
+      "step": 3900
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.853030303030303e-05,
+      "loss": 1.6824,
+      "step": 3910
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.852525252525253e-05,
+      "loss": 1.6293,
+      "step": 3920
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.852020202020202e-05,
+      "loss": 1.681,
+      "step": 3930
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.851515151515152e-05,
+      "loss": 1.6696,
+      "step": 3940
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.851010101010101e-05,
+      "loss": 1.6598,
+      "step": 3950
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.850505050505051e-05,
+      "loss": 1.6509,
+      "step": 3960
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.85e-05,
+      "loss": 1.672,
+      "step": 3970
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8494949494949496e-05,
+      "loss": 1.6688,
+      "step": 3980
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.848989898989899e-05,
+      "loss": 1.6573,
+      "step": 3990
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.848484848484849e-05,
+      "loss": 1.66,
+      "step": 4000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8479797979797984e-05,
+      "loss": 1.6772,
+      "step": 4010
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.847474747474748e-05,
+      "loss": 1.6472,
+      "step": 4020
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.846969696969697e-05,
+      "loss": 1.659,
+      "step": 4030
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8464646464646465e-05,
+      "loss": 1.652,
+      "step": 4040
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.845959595959596e-05,
+      "loss": 1.6613,
+      "step": 4050
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.845454545454546e-05,
+      "loss": 1.6592,
+      "step": 4060
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.844949494949495e-05,
+      "loss": 1.6561,
+      "step": 4070
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.844444444444445e-05,
+      "loss": 1.6666,
+      "step": 4080
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.843939393939394e-05,
+      "loss": 1.6457,
+      "step": 4090
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.843434343434344e-05,
+      "loss": 1.6444,
+      "step": 4100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.842929292929293e-05,
+      "loss": 1.6396,
+      "step": 4110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8424242424242426e-05,
+      "loss": 1.6522,
+      "step": 4120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.841919191919192e-05,
+      "loss": 1.6732,
+      "step": 4130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.841414141414142e-05,
+      "loss": 1.6499,
+      "step": 4140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.840909090909091e-05,
+      "loss": 1.6533,
+      "step": 4150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.840404040404041e-05,
+      "loss": 1.6647,
+      "step": 4160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.83989898989899e-05,
+      "loss": 1.65,
+      "step": 4170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8393939393939395e-05,
+      "loss": 1.6571,
+      "step": 4180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.838888888888889e-05,
+      "loss": 1.6659,
+      "step": 4190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.838383838383839e-05,
+      "loss": 1.6482,
+      "step": 4200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8378787878787876e-05,
+      "loss": 1.6335,
+      "step": 4210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.837373737373738e-05,
+      "loss": 1.6609,
+      "step": 4220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.836868686868687e-05,
+      "loss": 1.678,
+      "step": 4230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8363636363636364e-05,
+      "loss": 1.6559,
+      "step": 4240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.835858585858586e-05,
+      "loss": 1.6838,
+      "step": 4250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8353535353535356e-05,
+      "loss": 1.6733,
+      "step": 4260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8348484848484846e-05,
+      "loss": 1.6604,
+      "step": 4270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.834343434343435e-05,
+      "loss": 1.6719,
+      "step": 4280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.833838383838384e-05,
+      "loss": 1.6519,
+      "step": 4290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8333333333333334e-05,
+      "loss": 1.6679,
+      "step": 4300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.832828282828283e-05,
+      "loss": 1.6534,
+      "step": 4310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 4.8323232323232326e-05,
+      "loss": 1.663,
+      "step": 4320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8318181818181815e-05,
+      "loss": 1.6658,
+      "step": 4330
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.831313131313132e-05,
+      "loss": 1.6692,
+      "step": 4340
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.830808080808081e-05,
+      "loss": 1.6647,
+      "step": 4350
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.83030303030303e-05,
+      "loss": 1.6696,
+      "step": 4360
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.82979797979798e-05,
+      "loss": 1.6548,
+      "step": 4370
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8292929292929295e-05,
+      "loss": 1.6537,
+      "step": 4380
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.828787878787879e-05,
+      "loss": 1.6533,
+      "step": 4390
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.828282828282829e-05,
+      "loss": 1.6593,
+      "step": 4400
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8277777777777776e-05,
+      "loss": 1.6534,
+      "step": 4410
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.827272727272727e-05,
+      "loss": 1.6509,
+      "step": 4420
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8267676767676775e-05,
+      "loss": 1.6817,
+      "step": 4430
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8262626262626264e-05,
+      "loss": 1.6624,
+      "step": 4440
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.825757575757576e-05,
+      "loss": 1.6419,
+      "step": 4450
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8252525252525256e-05,
+      "loss": 1.6385,
+      "step": 4460
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.824747474747475e-05,
+      "loss": 1.6499,
+      "step": 4470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.824242424242425e-05,
+      "loss": 1.6479,
+      "step": 4480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8237373737373744e-05,
+      "loss": 1.6519,
+      "step": 4490
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.823232323232323e-05,
+      "loss": 1.6616,
+      "step": 4500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.822727272727273e-05,
+      "loss": 1.6671,
+      "step": 4510
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8222222222222225e-05,
+      "loss": 1.6562,
+      "step": 4520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.821717171717172e-05,
+      "loss": 1.6591,
+      "step": 4530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.821212121212122e-05,
+      "loss": 1.6547,
+      "step": 4540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.820707070707071e-05,
+      "loss": 1.6729,
+      "step": 4550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.82020202020202e-05,
+      "loss": 1.646,
+      "step": 4560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.81969696969697e-05,
+      "loss": 1.655,
+      "step": 4570
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8191919191919194e-05,
+      "loss": 1.6563,
+      "step": 4580
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.818686868686869e-05,
+      "loss": 1.6547,
+      "step": 4590
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8181818181818186e-05,
+      "loss": 1.6581,
+      "step": 4600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.817676767676768e-05,
+      "loss": 1.6554,
+      "step": 4610
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.817171717171717e-05,
+      "loss": 1.6712,
+      "step": 4620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8166666666666674e-05,
+      "loss": 1.6391,
+      "step": 4630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8161616161616163e-05,
+      "loss": 1.6579,
+      "step": 4640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.815656565656566e-05,
+      "loss": 1.6322,
+      "step": 4650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8151515151515155e-05,
+      "loss": 1.6656,
+      "step": 4660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.814646464646465e-05,
+      "loss": 1.6635,
+      "step": 4670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.814141414141414e-05,
+      "loss": 1.6368,
+      "step": 4680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.813636363636364e-05,
+      "loss": 1.6736,
+      "step": 4690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.813131313131313e-05,
+      "loss": 1.6629,
+      "step": 4700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.812626262626263e-05,
+      "loss": 1.6663,
+      "step": 4710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8121212121212125e-05,
+      "loss": 1.6561,
+      "step": 4720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.811616161616162e-05,
+      "loss": 1.6512,
+      "step": 4730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.811111111111111e-05,
+      "loss": 1.6525,
+      "step": 4740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.810606060606061e-05,
+      "loss": 1.652,
+      "step": 4750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.81010101010101e-05,
+      "loss": 1.6346,
+      "step": 4760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.80959595959596e-05,
+      "loss": 1.6402,
+      "step": 4770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8090909090909094e-05,
+      "loss": 1.6479,
+      "step": 4780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.808585858585859e-05,
+      "loss": 1.644,
+      "step": 4790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.808080808080808e-05,
+      "loss": 1.6595,
+      "step": 4800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.807575757575758e-05,
+      "loss": 1.6323,
+      "step": 4810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.807070707070707e-05,
+      "loss": 1.6258,
+      "step": 4820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.806565656565657e-05,
+      "loss": 1.6544,
+      "step": 4830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.806060606060606e-05,
+      "loss": 1.6664,
+      "step": 4840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.805555555555556e-05,
+      "loss": 1.6687,
+      "step": 4850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.805050505050505e-05,
+      "loss": 1.6529,
+      "step": 4860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.804545454545455e-05,
+      "loss": 1.6368,
+      "step": 4870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.804040404040404e-05,
+      "loss": 1.6285,
+      "step": 4880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8035353535353536e-05,
+      "loss": 1.6472,
+      "step": 4890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.803030303030303e-05,
+      "loss": 1.6621,
+      "step": 4900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.802525252525253e-05,
+      "loss": 1.6545,
+      "step": 4910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8020202020202024e-05,
+      "loss": 1.6269,
+      "step": 4920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.801515151515152e-05,
+      "loss": 1.6664,
+      "step": 4930
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.801010101010101e-05,
+      "loss": 1.6418,
+      "step": 4940
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8005050505050505e-05,
+      "loss": 1.6614,
+      "step": 4950
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.8e-05,
+      "loss": 1.6551,
+      "step": 4960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.79949494949495e-05,
+      "loss": 1.6361,
+      "step": 4970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 4.798989898989899e-05,
+      "loss": 1.6605,
+      "step": 4980
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.798484848484849e-05,
+      "loss": 1.6622,
+      "step": 4990
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.797979797979798e-05,
+      "loss": 1.6434,
+      "step": 5000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.797474747474748e-05,
+      "loss": 1.6453,
+      "step": 5010
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.796969696969697e-05,
+      "loss": 1.6518,
+      "step": 5020
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7964646464646466e-05,
+      "loss": 1.661,
+      "step": 5030
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.795959595959596e-05,
+      "loss": 1.6522,
+      "step": 5040
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.795454545454546e-05,
+      "loss": 1.6567,
+      "step": 5050
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.794949494949495e-05,
+      "loss": 1.6263,
+      "step": 5060
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.794444444444445e-05,
+      "loss": 1.6333,
+      "step": 5070
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.793939393939394e-05,
+      "loss": 1.6576,
+      "step": 5080
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7934343434343436e-05,
+      "loss": 1.6412,
+      "step": 5090
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.792929292929293e-05,
+      "loss": 1.6563,
+      "step": 5100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.792424242424243e-05,
+      "loss": 1.6484,
+      "step": 5110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.791919191919192e-05,
+      "loss": 1.6755,
+      "step": 5120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.791414141414142e-05,
+      "loss": 1.6511,
+      "step": 5130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.790909090909091e-05,
+      "loss": 1.641,
+      "step": 5140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7904040404040405e-05,
+      "loss": 1.6595,
+      "step": 5150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.78989898989899e-05,
+      "loss": 1.6463,
+      "step": 5160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.78939393939394e-05,
+      "loss": 1.6258,
+      "step": 5170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7888888888888886e-05,
+      "loss": 1.6357,
+      "step": 5180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.788383838383839e-05,
+      "loss": 1.6388,
+      "step": 5190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.787878787878788e-05,
+      "loss": 1.6379,
+      "step": 5200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7873737373737374e-05,
+      "loss": 1.6626,
+      "step": 5210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.786868686868687e-05,
+      "loss": 1.6658,
+      "step": 5220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7863636363636366e-05,
+      "loss": 1.657,
+      "step": 5230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7858585858585855e-05,
+      "loss": 1.6322,
+      "step": 5240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.785353535353536e-05,
+      "loss": 1.6558,
+      "step": 5250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.784848484848485e-05,
+      "loss": 1.6422,
+      "step": 5260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.784343434343434e-05,
+      "loss": 1.6475,
+      "step": 5270
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7838383838383846e-05,
+      "loss": 1.6469,
+      "step": 5280
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7833333333333335e-05,
+      "loss": 1.6315,
+      "step": 5290
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.782828282828283e-05,
+      "loss": 1.6421,
+      "step": 5300
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.782323232323233e-05,
+      "loss": 1.6364,
+      "step": 5310
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.781818181818182e-05,
+      "loss": 1.6208,
+      "step": 5320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.781313131313131e-05,
+      "loss": 1.6288,
+      "step": 5330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7808080808080815e-05,
+      "loss": 1.6577,
+      "step": 5340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7803030303030304e-05,
+      "loss": 1.6569,
+      "step": 5350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.77979797979798e-05,
+      "loss": 1.631,
+      "step": 5360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7792929292929296e-05,
+      "loss": 1.6544,
+      "step": 5370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.778787878787879e-05,
+      "loss": 1.6715,
+      "step": 5380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.778282828282828e-05,
+      "loss": 1.6374,
+      "step": 5390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7777777777777784e-05,
+      "loss": 1.634,
+      "step": 5400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.777272727272727e-05,
+      "loss": 1.6366,
+      "step": 5410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.776767676767677e-05,
+      "loss": 1.6491,
+      "step": 5420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7762626262626265e-05,
+      "loss": 1.6384,
+      "step": 5430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.775757575757576e-05,
+      "loss": 1.6525,
+      "step": 5440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.775252525252526e-05,
+      "loss": 1.6514,
+      "step": 5450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.774747474747475e-05,
+      "loss": 1.6547,
+      "step": 5460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.774242424242424e-05,
+      "loss": 1.6461,
+      "step": 5470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.773737373737374e-05,
+      "loss": 1.6307,
+      "step": 5480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7732323232323234e-05,
+      "loss": 1.6441,
+      "step": 5490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.772727272727273e-05,
+      "loss": 1.6234,
+      "step": 5500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7722222222222226e-05,
+      "loss": 1.6683,
+      "step": 5510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.771717171717172e-05,
+      "loss": 1.6375,
+      "step": 5520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.771212121212121e-05,
+      "loss": 1.6559,
+      "step": 5530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7707070707070714e-05,
+      "loss": 1.6585,
+      "step": 5540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7702020202020204e-05,
+      "loss": 1.6559,
+      "step": 5550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.76969696969697e-05,
+      "loss": 1.6368,
+      "step": 5560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7691919191919196e-05,
+      "loss": 1.6391,
+      "step": 5570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.768686868686869e-05,
+      "loss": 1.6462,
+      "step": 5580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.768181818181818e-05,
+      "loss": 1.6432,
+      "step": 5590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7676767676767684e-05,
+      "loss": 1.6421,
+      "step": 5600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.767171717171717e-05,
+      "loss": 1.656,
+      "step": 5610
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.766666666666667e-05,
+      "loss": 1.6433,
+      "step": 5620
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.7661616161616165e-05,
+      "loss": 1.6664,
+      "step": 5630
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.765656565656566e-05,
+      "loss": 1.6503,
+      "step": 5640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.765151515151515e-05,
+      "loss": 1.6602,
+      "step": 5650
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.764646464646465e-05,
+      "loss": 1.6161,
+      "step": 5660
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.764141414141414e-05,
+      "loss": 1.6561,
+      "step": 5670
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.763636363636364e-05,
+      "loss": 1.6367,
+      "step": 5680
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7631313131313134e-05,
+      "loss": 1.6431,
+      "step": 5690
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.762626262626263e-05,
+      "loss": 1.626,
+      "step": 5700
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.762121212121212e-05,
+      "loss": 1.6528,
+      "step": 5710
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.761616161616162e-05,
+      "loss": 1.6328,
+      "step": 5720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.761111111111111e-05,
+      "loss": 1.6226,
+      "step": 5730
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.760606060606061e-05,
+      "loss": 1.6435,
+      "step": 5740
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.76010101010101e-05,
+      "loss": 1.628,
+      "step": 5750
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.75959595959596e-05,
+      "loss": 1.6341,
+      "step": 5760
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.759090909090909e-05,
+      "loss": 1.6383,
+      "step": 5770
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.758585858585859e-05,
+      "loss": 1.6399,
+      "step": 5780
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.758080808080808e-05,
+      "loss": 1.6445,
+      "step": 5790
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7575757575757576e-05,
+      "loss": 1.6424,
+      "step": 5800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.757070707070707e-05,
+      "loss": 1.6491,
+      "step": 5810
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.756565656565657e-05,
+      "loss": 1.6478,
+      "step": 5820
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7560606060606064e-05,
+      "loss": 1.6372,
+      "step": 5830
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.755555555555556e-05,
+      "loss": 1.6372,
+      "step": 5840
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.755050505050505e-05,
+      "loss": 1.6512,
+      "step": 5850
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7545454545454545e-05,
+      "loss": 1.6321,
+      "step": 5860
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.754040404040404e-05,
+      "loss": 1.6276,
+      "step": 5870
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.753535353535354e-05,
+      "loss": 1.64,
+      "step": 5880
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7530303030303033e-05,
+      "loss": 1.6412,
+      "step": 5890
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.752525252525253e-05,
+      "loss": 1.6662,
+      "step": 5900
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.752020202020202e-05,
+      "loss": 1.6436,
+      "step": 5910
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.751515151515152e-05,
+      "loss": 1.6507,
+      "step": 5920
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.751010101010101e-05,
+      "loss": 1.6629,
+      "step": 5930
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.750505050505051e-05,
+      "loss": 1.6341,
+      "step": 5940
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.75e-05,
+      "loss": 1.6244,
+      "step": 5950
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.74949494949495e-05,
+      "loss": 1.6258,
+      "step": 5960
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.748989898989899e-05,
+      "loss": 1.6411,
+      "step": 5970
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.748484848484849e-05,
+      "loss": 1.6303,
+      "step": 5980
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.747979797979798e-05,
+      "loss": 1.6269,
+      "step": 5990
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7474747474747476e-05,
+      "loss": 1.631,
+      "step": 6000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.746969696969697e-05,
+      "loss": 1.6192,
+      "step": 6010
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.746464646464647e-05,
+      "loss": 1.6297,
+      "step": 6020
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.745959595959596e-05,
+      "loss": 1.6502,
+      "step": 6030
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.745454545454546e-05,
+      "loss": 1.6823,
+      "step": 6040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.744949494949495e-05,
+      "loss": 1.6216,
+      "step": 6050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7444444444444445e-05,
+      "loss": 1.6328,
+      "step": 6060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.743939393939394e-05,
+      "loss": 1.6319,
+      "step": 6070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.743434343434344e-05,
+      "loss": 1.6162,
+      "step": 6080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7429292929292926e-05,
+      "loss": 1.6407,
+      "step": 6090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.742424242424243e-05,
+      "loss": 1.6339,
+      "step": 6100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.741919191919192e-05,
+      "loss": 1.6267,
+      "step": 6110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7414141414141414e-05,
+      "loss": 1.6332,
+      "step": 6120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.740909090909092e-05,
+      "loss": 1.6354,
+      "step": 6130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7404040404040406e-05,
+      "loss": 1.6322,
+      "step": 6140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.73989898989899e-05,
+      "loss": 1.6402,
+      "step": 6150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.73939393939394e-05,
+      "loss": 1.6482,
+      "step": 6160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7388888888888894e-05,
+      "loss": 1.6326,
+      "step": 6170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.738383838383838e-05,
+      "loss": 1.6442,
+      "step": 6180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7378787878787886e-05,
+      "loss": 1.6376,
+      "step": 6190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7373737373737375e-05,
+      "loss": 1.6408,
+      "step": 6200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.736868686868687e-05,
+      "loss": 1.6244,
+      "step": 6210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.736363636363637e-05,
+      "loss": 1.6361,
+      "step": 6220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.735858585858586e-05,
+      "loss": 1.6419,
+      "step": 6230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.735353535353535e-05,
+      "loss": 1.6455,
+      "step": 6240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7348484848484855e-05,
+      "loss": 1.6282,
+      "step": 6250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7343434343434344e-05,
+      "loss": 1.622,
+      "step": 6260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.733838383838384e-05,
+      "loss": 1.6374,
+      "step": 6270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7333333333333336e-05,
+      "loss": 1.6466,
+      "step": 6280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.732828282828283e-05,
+      "loss": 1.6307,
+      "step": 6290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.732323232323232e-05,
+      "loss": 1.6362,
+      "step": 6300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.7318181818181824e-05,
+      "loss": 1.6357,
+      "step": 6310
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7313131313131314e-05,
+      "loss": 1.6337,
+      "step": 6320
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.730808080808081e-05,
+      "loss": 1.632,
+      "step": 6330
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7303030303030306e-05,
+      "loss": 1.6418,
+      "step": 6340
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.72979797979798e-05,
+      "loss": 1.6261,
+      "step": 6350
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.72929292929293e-05,
+      "loss": 1.6333,
+      "step": 6360
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7287878787878794e-05,
+      "loss": 1.6278,
+      "step": 6370
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.728282828282828e-05,
+      "loss": 1.6098,
+      "step": 6380
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.727777777777778e-05,
+      "loss": 1.6289,
+      "step": 6390
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7272727272727275e-05,
+      "loss": 1.646,
+      "step": 6400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.726767676767677e-05,
+      "loss": 1.6265,
+      "step": 6410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.726262626262627e-05,
+      "loss": 1.6352,
+      "step": 6420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.725757575757576e-05,
+      "loss": 1.6333,
+      "step": 6430
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.725252525252525e-05,
+      "loss": 1.6404,
+      "step": 6440
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7247474747474755e-05,
+      "loss": 1.6207,
+      "step": 6450
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7242424242424244e-05,
+      "loss": 1.635,
+      "step": 6460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.723737373737374e-05,
+      "loss": 1.6287,
+      "step": 6470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7232323232323236e-05,
+      "loss": 1.6584,
+      "step": 6480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.722727272727273e-05,
+      "loss": 1.6239,
+      "step": 6490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.722222222222222e-05,
+      "loss": 1.6374,
+      "step": 6500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7217171717171724e-05,
+      "loss": 1.6269,
+      "step": 6510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.721212121212121e-05,
+      "loss": 1.6503,
+      "step": 6520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.720707070707071e-05,
+      "loss": 1.6359,
+      "step": 6530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7202020202020205e-05,
+      "loss": 1.6359,
+      "step": 6540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.71969696969697e-05,
+      "loss": 1.6421,
+      "step": 6550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.719191919191919e-05,
+      "loss": 1.6447,
+      "step": 6560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.718686868686869e-05,
+      "loss": 1.6462,
+      "step": 6570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.718181818181818e-05,
+      "loss": 1.6224,
+      "step": 6580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.717676767676768e-05,
+      "loss": 1.6544,
+      "step": 6590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7171717171717174e-05,
+      "loss": 1.6191,
+      "step": 6600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.716666666666667e-05,
+      "loss": 1.6295,
+      "step": 6610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.716161616161616e-05,
+      "loss": 1.6229,
+      "step": 6620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.715656565656566e-05,
+      "loss": 1.5952,
+      "step": 6630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.715151515151515e-05,
+      "loss": 1.6461,
+      "step": 6640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.714646464646465e-05,
+      "loss": 1.634,
+      "step": 6650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7141414141414143e-05,
+      "loss": 1.6204,
+      "step": 6660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.713636363636364e-05,
+      "loss": 1.6346,
+      "step": 6670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.713131313131313e-05,
+      "loss": 1.6438,
+      "step": 6680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.712626262626263e-05,
+      "loss": 1.629,
+      "step": 6690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.712121212121212e-05,
+      "loss": 1.6358,
+      "step": 6700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7116161616161617e-05,
+      "loss": 1.6119,
+      "step": 6710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.711111111111111e-05,
+      "loss": 1.6454,
+      "step": 6720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.710606060606061e-05,
+      "loss": 1.6354,
+      "step": 6730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7101010101010105e-05,
+      "loss": 1.6471,
+      "step": 6740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.70959595959596e-05,
+      "loss": 1.6358,
+      "step": 6750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.709090909090909e-05,
+      "loss": 1.6478,
+      "step": 6760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7085858585858586e-05,
+      "loss": 1.618,
+      "step": 6770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.708080808080808e-05,
+      "loss": 1.6376,
+      "step": 6780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.707575757575758e-05,
+      "loss": 1.6227,
+      "step": 6790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7070707070707074e-05,
+      "loss": 1.619,
+      "step": 6800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.706565656565657e-05,
+      "loss": 1.6213,
+      "step": 6810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.706060606060606e-05,
+      "loss": 1.6383,
+      "step": 6820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7055555555555555e-05,
+      "loss": 1.6227,
+      "step": 6830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.705050505050505e-05,
+      "loss": 1.6508,
+      "step": 6840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.704545454545455e-05,
+      "loss": 1.6319,
+      "step": 6850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.704040404040404e-05,
+      "loss": 1.6438,
+      "step": 6860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.703535353535354e-05,
+      "loss": 1.6395,
+      "step": 6870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.703030303030303e-05,
+      "loss": 1.6408,
+      "step": 6880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.702525252525253e-05,
+      "loss": 1.6314,
+      "step": 6890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.702020202020202e-05,
+      "loss": 1.619,
+      "step": 6900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7015151515151516e-05,
+      "loss": 1.615,
+      "step": 6910
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.701010101010101e-05,
+      "loss": 1.6187,
+      "step": 6920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.700505050505051e-05,
+      "loss": 1.6507,
+      "step": 6930
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.7e-05,
+      "loss": 1.6427,
+      "step": 6940
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.69949494949495e-05,
+      "loss": 1.6277,
+      "step": 6950
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.698989898989899e-05,
+      "loss": 1.6412,
+      "step": 6960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.6984848484848485e-05,
+      "loss": 1.6336,
+      "step": 6970
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.697979797979799e-05,
+      "loss": 1.6298,
+      "step": 6980
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.697474747474748e-05,
+      "loss": 1.6299,
+      "step": 6990
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.696969696969697e-05,
+      "loss": 1.6271,
+      "step": 7000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.696464646464647e-05,
+      "loss": 1.6414,
+      "step": 7010
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6959595959595965e-05,
+      "loss": 1.6157,
+      "step": 7020
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6954545454545454e-05,
+      "loss": 1.6368,
+      "step": 7030
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.694949494949496e-05,
+      "loss": 1.6445,
+      "step": 7040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6944444444444446e-05,
+      "loss": 1.6481,
+      "step": 7050
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.693939393939394e-05,
+      "loss": 1.6363,
+      "step": 7060
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.693434343434344e-05,
+      "loss": 1.6434,
+      "step": 7070
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6929292929292934e-05,
+      "loss": 1.6371,
+      "step": 7080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6924242424242424e-05,
+      "loss": 1.6378,
+      "step": 7090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6919191919191926e-05,
+      "loss": 1.6303,
+      "step": 7100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6914141414141416e-05,
+      "loss": 1.63,
+      "step": 7110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.690909090909091e-05,
+      "loss": 1.6254,
+      "step": 7120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.690404040404041e-05,
+      "loss": 1.6132,
+      "step": 7130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6898989898989903e-05,
+      "loss": 1.6109,
+      "step": 7140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.689393939393939e-05,
+      "loss": 1.6162,
+      "step": 7150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6888888888888895e-05,
+      "loss": 1.6365,
+      "step": 7160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6883838383838385e-05,
+      "loss": 1.6352,
+      "step": 7170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.687878787878788e-05,
+      "loss": 1.6418,
+      "step": 7180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.687373737373738e-05,
+      "loss": 1.6308,
+      "step": 7190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.686868686868687e-05,
+      "loss": 1.6116,
+      "step": 7200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.686363636363636e-05,
+      "loss": 1.623,
+      "step": 7210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6858585858585865e-05,
+      "loss": 1.6409,
+      "step": 7220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6853535353535354e-05,
+      "loss": 1.6029,
+      "step": 7230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.684848484848485e-05,
+      "loss": 1.6205,
+      "step": 7240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6843434343434346e-05,
+      "loss": 1.6216,
+      "step": 7250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.683838383838384e-05,
+      "loss": 1.6374,
+      "step": 7260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.683333333333334e-05,
+      "loss": 1.6468,
+      "step": 7270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6828282828282834e-05,
+      "loss": 1.6215,
+      "step": 7280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.682323232323232e-05,
+      "loss": 1.6323,
+      "step": 7290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.681818181818182e-05,
+      "loss": 1.6295,
+      "step": 7300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6813131313131315e-05,
+      "loss": 1.6446,
+      "step": 7310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.680808080808081e-05,
+      "loss": 1.6385,
+      "step": 7320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.680303030303031e-05,
+      "loss": 1.6279,
+      "step": 7330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.67979797979798e-05,
+      "loss": 1.6262,
+      "step": 7340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.679292929292929e-05,
+      "loss": 1.6345,
+      "step": 7350
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6787878787878795e-05,
+      "loss": 1.6164,
+      "step": 7360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6782828282828284e-05,
+      "loss": 1.669,
+      "step": 7370
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.677777777777778e-05,
+      "loss": 1.63,
+      "step": 7380
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6772727272727276e-05,
+      "loss": 1.6254,
+      "step": 7390
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.676767676767677e-05,
+      "loss": 1.635,
+      "step": 7400
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.676262626262626e-05,
+      "loss": 1.6012,
+      "step": 7410
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6757575757575764e-05,
+      "loss": 1.6146,
+      "step": 7420
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.675252525252525e-05,
+      "loss": 1.6419,
+      "step": 7430
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.674747474747475e-05,
+      "loss": 1.6159,
+      "step": 7440
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6742424242424245e-05,
+      "loss": 1.6327,
+      "step": 7450
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.673737373737374e-05,
+      "loss": 1.6441,
+      "step": 7460
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.673232323232323e-05,
+      "loss": 1.6236,
+      "step": 7470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.672727272727273e-05,
+      "loss": 1.6448,
+      "step": 7480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.672222222222222e-05,
+      "loss": 1.625,
+      "step": 7490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.671717171717172e-05,
+      "loss": 1.6218,
+      "step": 7500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6712121212121214e-05,
+      "loss": 1.5983,
+      "step": 7510
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.670707070707071e-05,
+      "loss": 1.6189,
+      "step": 7520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.67020202020202e-05,
+      "loss": 1.6175,
+      "step": 7530
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.66969696969697e-05,
+      "loss": 1.6234,
+      "step": 7540
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.669191919191919e-05,
+      "loss": 1.6074,
+      "step": 7550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.668686868686869e-05,
+      "loss": 1.6238,
+      "step": 7560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6681818181818184e-05,
+      "loss": 1.6086,
+      "step": 7570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.667676767676768e-05,
+      "loss": 1.6241,
+      "step": 7580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.667171717171717e-05,
+      "loss": 1.6205,
+      "step": 7590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.666666666666667e-05,
+      "loss": 1.6353,
+      "step": 7600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.666161616161616e-05,
+      "loss": 1.6227,
+      "step": 7610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.665656565656566e-05,
+      "loss": 1.6313,
+      "step": 7620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.665151515151515e-05,
+      "loss": 1.6151,
+      "step": 7630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.664646464646465e-05,
+      "loss": 1.5961,
+      "step": 7640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.6641414141414145e-05,
+      "loss": 1.6193,
+      "step": 7650
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.663636363636364e-05,
+      "loss": 1.6309,
+      "step": 7660
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.663131313131313e-05,
+      "loss": 1.6068,
+      "step": 7670
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6626262626262626e-05,
+      "loss": 1.6099,
+      "step": 7680
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.662121212121212e-05,
+      "loss": 1.6093,
+      "step": 7690
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.661616161616162e-05,
+      "loss": 1.6228,
+      "step": 7700
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6611111111111114e-05,
+      "loss": 1.6068,
+      "step": 7710
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.660606060606061e-05,
+      "loss": 1.6207,
+      "step": 7720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.66010101010101e-05,
+      "loss": 1.6028,
+      "step": 7730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6595959595959595e-05,
+      "loss": 1.6016,
+      "step": 7740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.659090909090909e-05,
+      "loss": 1.6192,
+      "step": 7750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.658585858585859e-05,
+      "loss": 1.6177,
+      "step": 7760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.658080808080808e-05,
+      "loss": 1.6297,
+      "step": 7770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.657575757575758e-05,
+      "loss": 1.6011,
+      "step": 7780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.657070707070707e-05,
+      "loss": 1.6135,
+      "step": 7790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.656565656565657e-05,
+      "loss": 1.6415,
+      "step": 7800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.656060606060606e-05,
+      "loss": 1.6171,
+      "step": 7810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6555555555555556e-05,
+      "loss": 1.6123,
+      "step": 7820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.655050505050505e-05,
+      "loss": 1.6264,
+      "step": 7830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.654545454545455e-05,
+      "loss": 1.6302,
+      "step": 7840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6540404040404044e-05,
+      "loss": 1.6286,
+      "step": 7850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.653535353535354e-05,
+      "loss": 1.6341,
+      "step": 7860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6530303030303036e-05,
+      "loss": 1.6319,
+      "step": 7870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6525252525252525e-05,
+      "loss": 1.6063,
+      "step": 7880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.652020202020203e-05,
+      "loss": 1.6545,
+      "step": 7890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.651515151515152e-05,
+      "loss": 1.6102,
+      "step": 7900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6510101010101013e-05,
+      "loss": 1.6056,
+      "step": 7910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.650505050505051e-05,
+      "loss": 1.6094,
+      "step": 7920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6500000000000005e-05,
+      "loss": 1.6297,
+      "step": 7930
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6494949494949495e-05,
+      "loss": 1.6293,
+      "step": 7940
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6489898989899e-05,
+      "loss": 1.6279,
+      "step": 7950
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6484848484848487e-05,
+      "loss": 1.6213,
+      "step": 7960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.647979797979798e-05,
+      "loss": 1.6082,
+      "step": 7970
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.647474747474748e-05,
+      "loss": 1.631,
+      "step": 7980
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6469696969696975e-05,
+      "loss": 1.6268,
+      "step": 7990
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6464646464646464e-05,
+      "loss": 1.5962,
+      "step": 8000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6459595959595967e-05,
+      "loss": 1.6331,
+      "step": 8010
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6454545454545456e-05,
+      "loss": 1.6319,
+      "step": 8020
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.644949494949495e-05,
+      "loss": 1.6111,
+      "step": 8030
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.644444444444445e-05,
+      "loss": 1.6227,
+      "step": 8040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6439393939393944e-05,
+      "loss": 1.6104,
+      "step": 8050
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.643434343434343e-05,
+      "loss": 1.6196,
+      "step": 8060
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6429292929292936e-05,
+      "loss": 1.6176,
+      "step": 8070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6424242424242425e-05,
+      "loss": 1.5878,
+      "step": 8080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.641919191919192e-05,
+      "loss": 1.6234,
+      "step": 8090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.641414141414142e-05,
+      "loss": 1.6236,
+      "step": 8100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.640909090909091e-05,
+      "loss": 1.6386,
+      "step": 8110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.64040404040404e-05,
+      "loss": 1.6122,
+      "step": 8120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6398989898989905e-05,
+      "loss": 1.602,
+      "step": 8130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6393939393939394e-05,
+      "loss": 1.6049,
+      "step": 8140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.638888888888889e-05,
+      "loss": 1.6375,
+      "step": 8150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6383838383838386e-05,
+      "loss": 1.6318,
+      "step": 8160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.637878787878788e-05,
+      "loss": 1.6079,
+      "step": 8170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.637373737373738e-05,
+      "loss": 1.6166,
+      "step": 8180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6368686868686874e-05,
+      "loss": 1.6308,
+      "step": 8190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.636363636363636e-05,
+      "loss": 1.6029,
+      "step": 8200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.635858585858586e-05,
+      "loss": 1.6505,
+      "step": 8210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6353535353535355e-05,
+      "loss": 1.6272,
+      "step": 8220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.634848484848485e-05,
+      "loss": 1.6009,
+      "step": 8230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.634343434343435e-05,
+      "loss": 1.6169,
+      "step": 8240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.633838383838384e-05,
+      "loss": 1.625,
+      "step": 8250
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.633333333333333e-05,
+      "loss": 1.5903,
+      "step": 8260
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6328282828282835e-05,
+      "loss": 1.6198,
+      "step": 8270
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6323232323232324e-05,
+      "loss": 1.5909,
+      "step": 8280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.631818181818182e-05,
+      "loss": 1.6033,
+      "step": 8290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.6313131313131316e-05,
+      "loss": 1.5955,
+      "step": 8300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 4.630808080808081e-05,
+      "loss": 1.6278,
+      "step": 8310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.63030303030303e-05,
+      "loss": 1.6227,
+      "step": 8320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6297979797979804e-05,
+      "loss": 1.6084,
+      "step": 8330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6292929292929294e-05,
+      "loss": 1.6097,
+      "step": 8340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.628787878787879e-05,
+      "loss": 1.621,
+      "step": 8350
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6282828282828286e-05,
+      "loss": 1.6295,
+      "step": 8360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.627777777777778e-05,
+      "loss": 1.6184,
+      "step": 8370
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.627272727272727e-05,
+      "loss": 1.6265,
+      "step": 8380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6267676767676774e-05,
+      "loss": 1.6213,
+      "step": 8390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.626262626262626e-05,
+      "loss": 1.6118,
+      "step": 8400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.625757575757576e-05,
+      "loss": 1.5938,
+      "step": 8410
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6252525252525255e-05,
+      "loss": 1.6498,
+      "step": 8420
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.624747474747475e-05,
+      "loss": 1.6106,
+      "step": 8430
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.624242424242424e-05,
+      "loss": 1.6325,
+      "step": 8440
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.623737373737374e-05,
+      "loss": 1.6129,
+      "step": 8450
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.623232323232323e-05,
+      "loss": 1.5875,
+      "step": 8460
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.622727272727273e-05,
+      "loss": 1.6323,
+      "step": 8470
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6222222222222224e-05,
+      "loss": 1.6069,
+      "step": 8480
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.621717171717172e-05,
+      "loss": 1.6306,
+      "step": 8490
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.621212121212121e-05,
+      "loss": 1.6234,
+      "step": 8500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.620707070707071e-05,
+      "loss": 1.6481,
+      "step": 8510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.62020202020202e-05,
+      "loss": 1.6252,
+      "step": 8520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.61969696969697e-05,
+      "loss": 1.5986,
+      "step": 8530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.619191919191919e-05,
+      "loss": 1.6187,
+      "step": 8540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.618686868686869e-05,
+      "loss": 1.5918,
+      "step": 8550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.618181818181818e-05,
+      "loss": 1.6391,
+      "step": 8560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.617676767676768e-05,
+      "loss": 1.6291,
+      "step": 8570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.617171717171717e-05,
+      "loss": 1.6319,
+      "step": 8580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6166666666666666e-05,
+      "loss": 1.6048,
+      "step": 8590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.616161616161616e-05,
+      "loss": 1.6138,
+      "step": 8600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.615656565656566e-05,
+      "loss": 1.59,
+      "step": 8610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6151515151515154e-05,
+      "loss": 1.6271,
+      "step": 8620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.614646464646465e-05,
+      "loss": 1.6116,
+      "step": 8630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.614141414141414e-05,
+      "loss": 1.6188,
+      "step": 8640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6136363636363635e-05,
+      "loss": 1.636,
+      "step": 8650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.613131313131313e-05,
+      "loss": 1.6178,
+      "step": 8660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.612626262626263e-05,
+      "loss": 1.6122,
+      "step": 8670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.612121212121212e-05,
+      "loss": 1.6326,
+      "step": 8680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.611616161616162e-05,
+      "loss": 1.6231,
+      "step": 8690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6111111111111115e-05,
+      "loss": 1.6315,
+      "step": 8700
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.610606060606061e-05,
+      "loss": 1.6219,
+      "step": 8710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.610101010101011e-05,
+      "loss": 1.6359,
+      "step": 8720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6095959595959597e-05,
+      "loss": 1.622,
+      "step": 8730
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.609090909090909e-05,
+      "loss": 1.6263,
+      "step": 8740
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.608585858585859e-05,
+      "loss": 1.6318,
+      "step": 8750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6080808080808085e-05,
+      "loss": 1.6253,
+      "step": 8760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.607575757575758e-05,
+      "loss": 1.6243,
+      "step": 8770
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6070707070707077e-05,
+      "loss": 1.6245,
+      "step": 8780
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6065656565656566e-05,
+      "loss": 1.6013,
+      "step": 8790
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.606060606060607e-05,
+      "loss": 1.6251,
+      "step": 8800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.605555555555556e-05,
+      "loss": 1.6072,
+      "step": 8810
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6050505050505054e-05,
+      "loss": 1.6027,
+      "step": 8820
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.604545454545455e-05,
+      "loss": 1.6429,
+      "step": 8830
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6040404040404046e-05,
+      "loss": 1.5988,
+      "step": 8840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6035353535353535e-05,
+      "loss": 1.6269,
+      "step": 8850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.603030303030304e-05,
+      "loss": 1.609,
+      "step": 8860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.602525252525253e-05,
+      "loss": 1.59,
+      "step": 8870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.602020202020202e-05,
+      "loss": 1.6119,
+      "step": 8880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.601515151515152e-05,
+      "loss": 1.6465,
+      "step": 8890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6010101010101015e-05,
+      "loss": 1.6351,
+      "step": 8900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.6005050505050504e-05,
+      "loss": 1.6245,
+      "step": 8910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 1.6176,
+      "step": 8920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.5994949494949496e-05,
+      "loss": 1.5979,
+      "step": 8930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.598989898989899e-05,
+      "loss": 1.6051,
+      "step": 8940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.598484848484849e-05,
+      "loss": 1.6245,
+      "step": 8950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.5979797979797984e-05,
+      "loss": 1.6314,
+      "step": 8960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.597474747474747e-05,
+      "loss": 1.6221,
+      "step": 8970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.5969696969696976e-05,
+      "loss": 1.6008,
+      "step": 8980
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5964646464646465e-05,
+      "loss": 1.6254,
+      "step": 8990
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.595959595959596e-05,
+      "loss": 1.6279,
+      "step": 9000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.595454545454546e-05,
+      "loss": 1.6121,
+      "step": 9010
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.594949494949495e-05,
+      "loss": 1.6228,
+      "step": 9020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.594444444444444e-05,
+      "loss": 1.6111,
+      "step": 9030
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5939393939393945e-05,
+      "loss": 1.5937,
+      "step": 9040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5934343434343434e-05,
+      "loss": 1.624,
+      "step": 9050
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.592929292929293e-05,
+      "loss": 1.6169,
+      "step": 9060
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5924242424242426e-05,
+      "loss": 1.6331,
+      "step": 9070
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.591919191919192e-05,
+      "loss": 1.6341,
+      "step": 9080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.591414141414142e-05,
+      "loss": 1.6423,
+      "step": 9090
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5909090909090914e-05,
+      "loss": 1.6446,
+      "step": 9100
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5904040404040404e-05,
+      "loss": 1.6269,
+      "step": 9110
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.58989898989899e-05,
+      "loss": 1.5896,
+      "step": 9120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5893939393939396e-05,
+      "loss": 1.6127,
+      "step": 9130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.588888888888889e-05,
+      "loss": 1.6389,
+      "step": 9140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.588383838383839e-05,
+      "loss": 1.6259,
+      "step": 9150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5878787878787883e-05,
+      "loss": 1.6083,
+      "step": 9160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.587373737373737e-05,
+      "loss": 1.6153,
+      "step": 9170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.586868686868687e-05,
+      "loss": 1.6211,
+      "step": 9180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5863636363636365e-05,
+      "loss": 1.6003,
+      "step": 9190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.585858585858586e-05,
+      "loss": 1.5982,
+      "step": 9200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.585353535353536e-05,
+      "loss": 1.5977,
+      "step": 9210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.584848484848485e-05,
+      "loss": 1.6198,
+      "step": 9220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.584343434343434e-05,
+      "loss": 1.5837,
+      "step": 9230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5838383838383845e-05,
+      "loss": 1.6126,
+      "step": 9240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5833333333333334e-05,
+      "loss": 1.606,
+      "step": 9250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.582828282828283e-05,
+      "loss": 1.6145,
+      "step": 9260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5823232323232326e-05,
+      "loss": 1.6261,
+      "step": 9270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.581818181818182e-05,
+      "loss": 1.5983,
+      "step": 9280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.581313131313131e-05,
+      "loss": 1.6122,
+      "step": 9290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5808080808080814e-05,
+      "loss": 1.6162,
+      "step": 9300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.58030303030303e-05,
+      "loss": 1.6217,
+      "step": 9310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.57979797979798e-05,
+      "loss": 1.6389,
+      "step": 9320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5792929292929295e-05,
+      "loss": 1.6148,
+      "step": 9330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.578787878787879e-05,
+      "loss": 1.5894,
+      "step": 9340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.578282828282828e-05,
+      "loss": 1.6242,
+      "step": 9350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.577777777777778e-05,
+      "loss": 1.6066,
+      "step": 9360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.577272727272727e-05,
+      "loss": 1.6111,
+      "step": 9370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.576767676767677e-05,
+      "loss": 1.597,
+      "step": 9380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5762626262626264e-05,
+      "loss": 1.6258,
+      "step": 9390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.575757575757576e-05,
+      "loss": 1.6097,
+      "step": 9400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.575252525252525e-05,
+      "loss": 1.6063,
+      "step": 9410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.574747474747475e-05,
+      "loss": 1.6093,
+      "step": 9420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.574242424242424e-05,
+      "loss": 1.602,
+      "step": 9430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.573737373737374e-05,
+      "loss": 1.5992,
+      "step": 9440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.573232323232323e-05,
+      "loss": 1.6231,
+      "step": 9450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.572727272727273e-05,
+      "loss": 1.6313,
+      "step": 9460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.572222222222222e-05,
+      "loss": 1.6246,
+      "step": 9470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.571717171717172e-05,
+      "loss": 1.6203,
+      "step": 9480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.571212121212121e-05,
+      "loss": 1.603,
+      "step": 9490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5707070707070706e-05,
+      "loss": 1.6215,
+      "step": 9500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.57020202020202e-05,
+      "loss": 1.6057,
+      "step": 9510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.56969696969697e-05,
+      "loss": 1.5958,
+      "step": 9520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5691919191919194e-05,
+      "loss": 1.6193,
+      "step": 9530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.568686868686869e-05,
+      "loss": 1.6331,
+      "step": 9540
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5681818181818186e-05,
+      "loss": 1.5953,
+      "step": 9550
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5676767676767676e-05,
+      "loss": 1.6211,
+      "step": 9560
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.567171717171718e-05,
+      "loss": 1.5924,
+      "step": 9570
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.566666666666667e-05,
+      "loss": 1.6128,
+      "step": 9580
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5661616161616164e-05,
+      "loss": 1.6009,
+      "step": 9590
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.565656565656566e-05,
+      "loss": 1.6062,
+      "step": 9600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.5651515151515156e-05,
+      "loss": 1.611,
+      "step": 9610
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.564646464646465e-05,
+      "loss": 1.6169,
+      "step": 9620
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.564141414141415e-05,
+      "loss": 1.6153,
+      "step": 9630
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.563636363636364e-05,
+      "loss": 1.6123,
+      "step": 9640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.563131313131313e-05,
+      "loss": 1.6036,
+      "step": 9650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.562626262626263e-05,
+      "loss": 1.6326,
+      "step": 9660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5621212121212125e-05,
+      "loss": 1.6479,
+      "step": 9670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.561616161616162e-05,
+      "loss": 1.619,
+      "step": 9680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.561111111111112e-05,
+      "loss": 1.6194,
+      "step": 9690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5606060606060606e-05,
+      "loss": 1.5868,
+      "step": 9700
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.560101010101011e-05,
+      "loss": 1.6207,
+      "step": 9710
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.55959595959596e-05,
+      "loss": 1.6068,
+      "step": 9720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5590909090909094e-05,
+      "loss": 1.6219,
+      "step": 9730
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.558585858585859e-05,
+      "loss": 1.6189,
+      "step": 9740
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5580808080808086e-05,
+      "loss": 1.6043,
+      "step": 9750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5575757575757575e-05,
+      "loss": 1.6052,
+      "step": 9760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.557070707070708e-05,
+      "loss": 1.5894,
+      "step": 9770
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.556565656565657e-05,
+      "loss": 1.6034,
+      "step": 9780
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.556060606060606e-05,
+      "loss": 1.6332,
+      "step": 9790
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.555555555555556e-05,
+      "loss": 1.6161,
+      "step": 9800
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5550505050505055e-05,
+      "loss": 1.5887,
+      "step": 9810
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5545454545454544e-05,
+      "loss": 1.6113,
+      "step": 9820
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.554040404040405e-05,
+      "loss": 1.6285,
+      "step": 9830
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5535353535353536e-05,
+      "loss": 1.6176,
+      "step": 9840
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.553030303030303e-05,
+      "loss": 1.6223,
+      "step": 9850
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.552525252525253e-05,
+      "loss": 1.6006,
+      "step": 9860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5520202020202024e-05,
+      "loss": 1.6102,
+      "step": 9870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5515151515151513e-05,
+      "loss": 1.6162,
+      "step": 9880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5510101010101016e-05,
+      "loss": 1.6136,
+      "step": 9890
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5505050505050505e-05,
+      "loss": 1.6059,
+      "step": 9900
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.55e-05,
+      "loss": 1.5856,
+      "step": 9910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.54949494949495e-05,
+      "loss": 1.6243,
+      "step": 9920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5489898989898993e-05,
+      "loss": 1.5928,
+      "step": 9930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.548484848484848e-05,
+      "loss": 1.6409,
+      "step": 9940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5479797979797985e-05,
+      "loss": 1.6176,
+      "step": 9950
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5474747474747475e-05,
+      "loss": 1.6133,
+      "step": 9960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.546969696969697e-05,
+      "loss": 1.5997,
+      "step": 9970
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5464646464646467e-05,
+      "loss": 1.6179,
+      "step": 9980
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.545959595959596e-05,
+      "loss": 1.6067,
+      "step": 9990
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.545454545454546e-05,
+      "loss": 1.6228,
+      "step": 10000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5449494949494955e-05,
+      "loss": 1.6093,
+      "step": 10010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5444444444444444e-05,
+      "loss": 1.5937,
+      "step": 10020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.543939393939394e-05,
+      "loss": 1.5927,
+      "step": 10030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5434343434343436e-05,
+      "loss": 1.6113,
+      "step": 10040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.542929292929293e-05,
+      "loss": 1.6015,
+      "step": 10050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.542424242424243e-05,
+      "loss": 1.6259,
+      "step": 10060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5419191919191924e-05,
+      "loss": 1.5966,
+      "step": 10070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.541414141414141e-05,
+      "loss": 1.6245,
+      "step": 10080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.540909090909091e-05,
+      "loss": 1.5845,
+      "step": 10090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5404040404040405e-05,
+      "loss": 1.5985,
+      "step": 10100
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.53989898989899e-05,
+      "loss": 1.607,
+      "step": 10110
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.53939393939394e-05,
+      "loss": 1.6198,
+      "step": 10120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.538888888888889e-05,
+      "loss": 1.6153,
+      "step": 10130
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.538383838383838e-05,
+      "loss": 1.6051,
+      "step": 10140
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5378787878787885e-05,
+      "loss": 1.6451,
+      "step": 10150
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5373737373737374e-05,
+      "loss": 1.6066,
+      "step": 10160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.536868686868687e-05,
+      "loss": 1.6312,
+      "step": 10170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5363636363636366e-05,
+      "loss": 1.6036,
+      "step": 10180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.535858585858586e-05,
+      "loss": 1.61,
+      "step": 10190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.535353535353535e-05,
+      "loss": 1.6037,
+      "step": 10200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5348484848484854e-05,
+      "loss": 1.6068,
+      "step": 10210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.534343434343434e-05,
+      "loss": 1.6077,
+      "step": 10220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.533838383838384e-05,
+      "loss": 1.5989,
+      "step": 10230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5333333333333335e-05,
+      "loss": 1.6129,
+      "step": 10240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.532828282828283e-05,
+      "loss": 1.5982,
+      "step": 10250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.532323232323232e-05,
+      "loss": 1.5904,
+      "step": 10260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.531818181818182e-05,
+      "loss": 1.602,
+      "step": 10270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.531313131313131e-05,
+      "loss": 1.613,
+      "step": 10280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.530808080808081e-05,
+      "loss": 1.6127,
+      "step": 10290
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5303030303030304e-05,
+      "loss": 1.6085,
+      "step": 10300
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.52979797979798e-05,
+      "loss": 1.5895,
+      "step": 10310
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.529292929292929e-05,
+      "loss": 1.5978,
+      "step": 10320
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.528787878787879e-05,
+      "loss": 1.6058,
+      "step": 10330
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.528282828282828e-05,
+      "loss": 1.5926,
+      "step": 10340
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.527777777777778e-05,
+      "loss": 1.605,
+      "step": 10350
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5272727272727274e-05,
+      "loss": 1.6228,
+      "step": 10360
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.526767676767677e-05,
+      "loss": 1.5985,
+      "step": 10370
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.526262626262626e-05,
+      "loss": 1.6032,
+      "step": 10380
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.525757575757576e-05,
+      "loss": 1.6002,
+      "step": 10390
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.525252525252526e-05,
+      "loss": 1.5882,
+      "step": 10400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.524747474747475e-05,
+      "loss": 1.6078,
+      "step": 10410
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.524242424242425e-05,
+      "loss": 1.6181,
+      "step": 10420
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.523737373737374e-05,
+      "loss": 1.6047,
+      "step": 10430
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5232323232323235e-05,
+      "loss": 1.6257,
+      "step": 10440
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.522727272727273e-05,
+      "loss": 1.5763,
+      "step": 10450
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.522222222222223e-05,
+      "loss": 1.5992,
+      "step": 10460
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5217171717171716e-05,
+      "loss": 1.6027,
+      "step": 10470
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.521212121212122e-05,
+      "loss": 1.6079,
+      "step": 10480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.520707070707071e-05,
+      "loss": 1.5981,
+      "step": 10490
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5202020202020204e-05,
+      "loss": 1.594,
+      "step": 10500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.51969696969697e-05,
+      "loss": 1.5932,
+      "step": 10510
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5191919191919196e-05,
+      "loss": 1.5794,
+      "step": 10520
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.518686868686869e-05,
+      "loss": 1.6144,
+      "step": 10530
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.518181818181819e-05,
+      "loss": 1.6154,
+      "step": 10540
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.517676767676768e-05,
+      "loss": 1.5988,
+      "step": 10550
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.517171717171717e-05,
+      "loss": 1.6048,
+      "step": 10560
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.516666666666667e-05,
+      "loss": 1.6077,
+      "step": 10570
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5161616161616165e-05,
+      "loss": 1.5942,
+      "step": 10580
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.515656565656566e-05,
+      "loss": 1.6203,
+      "step": 10590
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.515151515151516e-05,
+      "loss": 1.5987,
+      "step": 10600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5146464646464646e-05,
+      "loss": 1.5935,
+      "step": 10610
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.514141414141415e-05,
+      "loss": 1.6134,
+      "step": 10620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.513636363636364e-05,
+      "loss": 1.5969,
+      "step": 10630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5131313131313134e-05,
+      "loss": 1.5902,
+      "step": 10640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.512626262626263e-05,
+      "loss": 1.62,
+      "step": 10650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5121212121212126e-05,
+      "loss": 1.5724,
+      "step": 10660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5116161616161615e-05,
+      "loss": 1.6029,
+      "step": 10670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.511111111111112e-05,
+      "loss": 1.6126,
+      "step": 10680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.510606060606061e-05,
+      "loss": 1.6037,
+      "step": 10690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.51010101010101e-05,
+      "loss": 1.6099,
+      "step": 10700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.50959595959596e-05,
+      "loss": 1.5948,
+      "step": 10710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5090909090909095e-05,
+      "loss": 1.595,
+      "step": 10720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5085858585858585e-05,
+      "loss": 1.5982,
+      "step": 10730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.508080808080809e-05,
+      "loss": 1.6015,
+      "step": 10740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5075757575757577e-05,
+      "loss": 1.6013,
+      "step": 10750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.507070707070707e-05,
+      "loss": 1.6056,
+      "step": 10760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.506565656565657e-05,
+      "loss": 1.6101,
+      "step": 10770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5060606060606065e-05,
+      "loss": 1.6208,
+      "step": 10780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5055555555555554e-05,
+      "loss": 1.5985,
+      "step": 10790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5050505050505056e-05,
+      "loss": 1.6174,
+      "step": 10800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5045454545454546e-05,
+      "loss": 1.6202,
+      "step": 10810
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.504040404040404e-05,
+      "loss": 1.609,
+      "step": 10820
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.503535353535354e-05,
+      "loss": 1.6122,
+      "step": 10830
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5030303030303034e-05,
+      "loss": 1.6057,
+      "step": 10840
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.502525252525252e-05,
+      "loss": 1.6165,
+      "step": 10850
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5020202020202026e-05,
+      "loss": 1.6257,
+      "step": 10860
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5015151515151515e-05,
+      "loss": 1.6208,
+      "step": 10870
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.501010101010101e-05,
+      "loss": 1.616,
+      "step": 10880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.500505050505051e-05,
+      "loss": 1.59,
+      "step": 10890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.5e-05,
+      "loss": 1.5925,
+      "step": 10900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.499494949494949e-05,
+      "loss": 1.619,
+      "step": 10910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.4989898989898995e-05,
+      "loss": 1.5857,
+      "step": 10920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.4984848484848484e-05,
+      "loss": 1.5959,
+      "step": 10930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.497979797979798e-05,
+      "loss": 1.5947,
+      "step": 10940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.4974747474747476e-05,
+      "loss": 1.6276,
+      "step": 10950
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.496969696969697e-05,
+      "loss": 1.6081,
+      "step": 10960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.496464646464647e-05,
+      "loss": 1.6095,
+      "step": 10970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4959595959595964e-05,
+      "loss": 1.5979,
+      "step": 10980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.495454545454545e-05,
+      "loss": 1.6041,
+      "step": 10990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.494949494949495e-05,
+      "loss": 1.5675,
+      "step": 11000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4944444444444445e-05,
+      "loss": 1.5869,
+      "step": 11010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.493939393939394e-05,
+      "loss": 1.617,
+      "step": 11020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.493434343434344e-05,
+      "loss": 1.5902,
+      "step": 11030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.492929292929293e-05,
+      "loss": 1.5765,
+      "step": 11040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.492424242424242e-05,
+      "loss": 1.5974,
+      "step": 11050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4919191919191925e-05,
+      "loss": 1.5939,
+      "step": 11060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4914141414141414e-05,
+      "loss": 1.6242,
+      "step": 11070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.490909090909091e-05,
+      "loss": 1.5983,
+      "step": 11080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4904040404040406e-05,
+      "loss": 1.59,
+      "step": 11090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.48989898989899e-05,
+      "loss": 1.6444,
+      "step": 11100
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.489393939393939e-05,
+      "loss": 1.6091,
+      "step": 11110
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4888888888888894e-05,
+      "loss": 1.6106,
+      "step": 11120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4883838383838384e-05,
+      "loss": 1.5979,
+      "step": 11130
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.487878787878788e-05,
+      "loss": 1.6216,
+      "step": 11140
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4873737373737375e-05,
+      "loss": 1.5838,
+      "step": 11150
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.486868686868687e-05,
+      "loss": 1.6016,
+      "step": 11160
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.486363636363636e-05,
+      "loss": 1.6073,
+      "step": 11170
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4858585858585863e-05,
+      "loss": 1.5875,
+      "step": 11180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.485353535353535e-05,
+      "loss": 1.5946,
+      "step": 11190
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.484848484848485e-05,
+      "loss": 1.6114,
+      "step": 11200
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4843434343434345e-05,
+      "loss": 1.6078,
+      "step": 11210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.483838383838384e-05,
+      "loss": 1.6211,
+      "step": 11220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.483333333333333e-05,
+      "loss": 1.6072,
+      "step": 11230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.482828282828283e-05,
+      "loss": 1.612,
+      "step": 11240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.482323232323233e-05,
+      "loss": 1.5993,
+      "step": 11250
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.481818181818182e-05,
+      "loss": 1.6007,
+      "step": 11260
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.481313131313132e-05,
+      "loss": 1.5922,
+      "step": 11270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.480808080808081e-05,
+      "loss": 1.5958,
+      "step": 11280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4803030303030306e-05,
+      "loss": 1.6094,
+      "step": 11290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.47979797979798e-05,
+      "loss": 1.5903,
+      "step": 11300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.47929292929293e-05,
+      "loss": 1.6127,
+      "step": 11310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.478787878787879e-05,
+      "loss": 1.6201,
+      "step": 11320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.478282828282829e-05,
+      "loss": 1.6107,
+      "step": 11330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.477777777777778e-05,
+      "loss": 1.5696,
+      "step": 11340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4772727272727275e-05,
+      "loss": 1.6087,
+      "step": 11350
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.476767676767677e-05,
+      "loss": 1.6328,
+      "step": 11360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.476262626262627e-05,
+      "loss": 1.6092,
+      "step": 11370
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4757575757575756e-05,
+      "loss": 1.6101,
+      "step": 11380
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.475252525252526e-05,
+      "loss": 1.5844,
+      "step": 11390
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.474747474747475e-05,
+      "loss": 1.6038,
+      "step": 11400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4742424242424244e-05,
+      "loss": 1.6098,
+      "step": 11410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.473737373737374e-05,
+      "loss": 1.5992,
+      "step": 11420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4732323232323236e-05,
+      "loss": 1.582,
+      "step": 11430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.472727272727273e-05,
+      "loss": 1.6043,
+      "step": 11440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.472222222222223e-05,
+      "loss": 1.6199,
+      "step": 11450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.471717171717172e-05,
+      "loss": 1.6009,
+      "step": 11460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.471212121212121e-05,
+      "loss": 1.6108,
+      "step": 11470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.470707070707071e-05,
+      "loss": 1.5907,
+      "step": 11480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4702020202020205e-05,
+      "loss": 1.5982,
+      "step": 11490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.46969696969697e-05,
+      "loss": 1.5995,
+      "step": 11500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.46919191919192e-05,
+      "loss": 1.5867,
+      "step": 11510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4686868686868686e-05,
+      "loss": 1.5916,
+      "step": 11520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.468181818181818e-05,
+      "loss": 1.6173,
+      "step": 11530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.467676767676768e-05,
+      "loss": 1.6022,
+      "step": 11540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4671717171717174e-05,
+      "loss": 1.6037,
+      "step": 11550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.466666666666667e-05,
+      "loss": 1.5909,
+      "step": 11560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4661616161616166e-05,
+      "loss": 1.5724,
+      "step": 11570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4656565656565656e-05,
+      "loss": 1.6087,
+      "step": 11580
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.465151515151516e-05,
+      "loss": 1.6019,
+      "step": 11590
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.464646464646465e-05,
+      "loss": 1.5946,
+      "step": 11600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4641414141414144e-05,
+      "loss": 1.579,
+      "step": 11610
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.463636363636364e-05,
+      "loss": 1.6125,
+      "step": 11620
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4631313131313136e-05,
+      "loss": 1.6305,
+      "step": 11630
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.4626262626262625e-05,
+      "loss": 1.6094,
+      "step": 11640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.462121212121213e-05,
+      "loss": 1.591,
+      "step": 11650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.461616161616162e-05,
+      "loss": 1.5873,
+      "step": 11660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.461111111111111e-05,
+      "loss": 1.5846,
+      "step": 11670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.460606060606061e-05,
+      "loss": 1.6067,
+      "step": 11680
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4601010101010105e-05,
+      "loss": 1.6027,
+      "step": 11690
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4595959595959594e-05,
+      "loss": 1.6035,
+      "step": 11700
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.45909090909091e-05,
+      "loss": 1.6136,
+      "step": 11710
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4585858585858586e-05,
+      "loss": 1.5999,
+      "step": 11720
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.458080808080808e-05,
+      "loss": 1.6261,
+      "step": 11730
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.457575757575758e-05,
+      "loss": 1.6152,
+      "step": 11740
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4570707070707074e-05,
+      "loss": 1.6117,
+      "step": 11750
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.456565656565656e-05,
+      "loss": 1.6182,
+      "step": 11760
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4560606060606066e-05,
+      "loss": 1.6185,
+      "step": 11770
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4555555555555555e-05,
+      "loss": 1.5709,
+      "step": 11780
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.455050505050505e-05,
+      "loss": 1.6259,
+      "step": 11790
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.454545454545455e-05,
+      "loss": 1.6229,
+      "step": 11800
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.454040404040404e-05,
+      "loss": 1.5969,
+      "step": 11810
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.453535353535353e-05,
+      "loss": 1.5923,
+      "step": 11820
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4530303030303035e-05,
+      "loss": 1.5935,
+      "step": 11830
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4525252525252524e-05,
+      "loss": 1.5921,
+      "step": 11840
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.452020202020202e-05,
+      "loss": 1.5889,
+      "step": 11850
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4515151515151516e-05,
+      "loss": 1.6032,
+      "step": 11860
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.451010101010101e-05,
+      "loss": 1.5997,
+      "step": 11870
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.450505050505051e-05,
+      "loss": 1.6034,
+      "step": 11880
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4500000000000004e-05,
+      "loss": 1.6085,
+      "step": 11890
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4494949494949493e-05,
+      "loss": 1.6028,
+      "step": 11900
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.448989898989899e-05,
+      "loss": 1.5877,
+      "step": 11910
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4484848484848485e-05,
+      "loss": 1.6052,
+      "step": 11920
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.447979797979798e-05,
+      "loss": 1.5812,
+      "step": 11930
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.447474747474748e-05,
+      "loss": 1.5969,
+      "step": 11940
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4469696969696973e-05,
+      "loss": 1.5928,
+      "step": 11950
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.446464646464646e-05,
+      "loss": 1.6188,
+      "step": 11960
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4459595959595965e-05,
+      "loss": 1.6292,
+      "step": 11970
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4454545454545455e-05,
+      "loss": 1.6098,
+      "step": 11980
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.444949494949495e-05,
+      "loss": 1.6188,
+      "step": 11990
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4444444444444447e-05,
+      "loss": 1.6081,
+      "step": 12000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.443939393939394e-05,
+      "loss": 1.6045,
+      "step": 12010
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.443434343434343e-05,
+      "loss": 1.6026,
+      "step": 12020
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4429292929292935e-05,
+      "loss": 1.6145,
+      "step": 12030
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4424242424242424e-05,
+      "loss": 1.5984,
+      "step": 12040
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.441919191919192e-05,
+      "loss": 1.5975,
+      "step": 12050
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4414141414141416e-05,
+      "loss": 1.5858,
+      "step": 12060
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.440909090909091e-05,
+      "loss": 1.5871,
+      "step": 12070
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.44040404040404e-05,
+      "loss": 1.592,
+      "step": 12080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4398989898989904e-05,
+      "loss": 1.5851,
+      "step": 12090
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.43939393939394e-05,
+      "loss": 1.6217,
+      "step": 12100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.438888888888889e-05,
+      "loss": 1.6063,
+      "step": 12110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.438383838383839e-05,
+      "loss": 1.584,
+      "step": 12120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.437878787878788e-05,
+      "loss": 1.5851,
+      "step": 12130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.437373737373738e-05,
+      "loss": 1.5831,
+      "step": 12140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.436868686868687e-05,
+      "loss": 1.6057,
+      "step": 12150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.436363636363637e-05,
+      "loss": 1.5957,
+      "step": 12160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.435858585858586e-05,
+      "loss": 1.6068,
+      "step": 12170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.435353535353536e-05,
+      "loss": 1.6038,
+      "step": 12180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.434848484848485e-05,
+      "loss": 1.5872,
+      "step": 12190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4343434343434346e-05,
+      "loss": 1.626,
+      "step": 12200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.433838383838384e-05,
+      "loss": 1.6065,
+      "step": 12210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.433333333333334e-05,
+      "loss": 1.5891,
+      "step": 12220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.432828282828283e-05,
+      "loss": 1.6183,
+      "step": 12230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.432323232323233e-05,
+      "loss": 1.6292,
+      "step": 12240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.431818181818182e-05,
+      "loss": 1.5843,
+      "step": 12250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4313131313131315e-05,
+      "loss": 1.591,
+      "step": 12260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.430808080808081e-05,
+      "loss": 1.6177,
+      "step": 12270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.430303030303031e-05,
+      "loss": 1.6192,
+      "step": 12280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.4297979797979796e-05,
+      "loss": 1.6116,
+      "step": 12290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.42929292929293e-05,
+      "loss": 1.5776,
+      "step": 12300
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.428787878787879e-05,
+      "loss": 1.5887,
+      "step": 12310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4282828282828284e-05,
+      "loss": 1.597,
+      "step": 12320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.427777777777778e-05,
+      "loss": 1.5886,
+      "step": 12330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4272727272727276e-05,
+      "loss": 1.607,
+      "step": 12340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.426767676767677e-05,
+      "loss": 1.5803,
+      "step": 12350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.426262626262627e-05,
+      "loss": 1.6149,
+      "step": 12360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.425757575757576e-05,
+      "loss": 1.5964,
+      "step": 12370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4252525252525254e-05,
+      "loss": 1.6038,
+      "step": 12380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.424747474747475e-05,
+      "loss": 1.6164,
+      "step": 12390
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4242424242424246e-05,
+      "loss": 1.6048,
+      "step": 12400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.423737373737374e-05,
+      "loss": 1.5832,
+      "step": 12410
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.423232323232324e-05,
+      "loss": 1.5846,
+      "step": 12420
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.422727272727273e-05,
+      "loss": 1.6015,
+      "step": 12430
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.422222222222222e-05,
+      "loss": 1.6057,
+      "step": 12440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.421717171717172e-05,
+      "loss": 1.5873,
+      "step": 12450
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4212121212121215e-05,
+      "loss": 1.5953,
+      "step": 12460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.420707070707071e-05,
+      "loss": 1.5998,
+      "step": 12470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.420202020202021e-05,
+      "loss": 1.5798,
+      "step": 12480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4196969696969696e-05,
+      "loss": 1.6223,
+      "step": 12490
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.41919191919192e-05,
+      "loss": 1.5745,
+      "step": 12500
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.418686868686869e-05,
+      "loss": 1.6027,
+      "step": 12510
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4181818181818184e-05,
+      "loss": 1.6038,
+      "step": 12520
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.417676767676768e-05,
+      "loss": 1.6025,
+      "step": 12530
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4171717171717176e-05,
+      "loss": 1.6295,
+      "step": 12540
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4166666666666665e-05,
+      "loss": 1.5912,
+      "step": 12550
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.416161616161617e-05,
+      "loss": 1.6098,
+      "step": 12560
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.415656565656566e-05,
+      "loss": 1.5871,
+      "step": 12570
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.415151515151515e-05,
+      "loss": 1.5887,
+      "step": 12580
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.414646464646465e-05,
+      "loss": 1.6019,
+      "step": 12590
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4141414141414145e-05,
+      "loss": 1.5701,
+      "step": 12600
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4136363636363634e-05,
+      "loss": 1.6103,
+      "step": 12610
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.413131313131314e-05,
+      "loss": 1.626,
+      "step": 12620
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4126262626262626e-05,
+      "loss": 1.6116,
+      "step": 12630
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.412121212121212e-05,
+      "loss": 1.6097,
+      "step": 12640
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.411616161616162e-05,
+      "loss": 1.5905,
+      "step": 12650
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4111111111111114e-05,
+      "loss": 1.5998,
+      "step": 12660
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.41060606060606e-05,
+      "loss": 1.5823,
+      "step": 12670
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4101010101010106e-05,
+      "loss": 1.5918,
+      "step": 12680
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4095959595959595e-05,
+      "loss": 1.5833,
+      "step": 12690
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.409090909090909e-05,
+      "loss": 1.6159,
+      "step": 12700
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.408585858585859e-05,
+      "loss": 1.5945,
+      "step": 12710
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.408080808080808e-05,
+      "loss": 1.6051,
+      "step": 12720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.407575757575757e-05,
+      "loss": 1.5747,
+      "step": 12730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4070707070707075e-05,
+      "loss": 1.6063,
+      "step": 12740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4065656565656565e-05,
+      "loss": 1.6023,
+      "step": 12750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.406060606060606e-05,
+      "loss": 1.5877,
+      "step": 12760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4055555555555557e-05,
+      "loss": 1.5947,
+      "step": 12770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.405050505050505e-05,
+      "loss": 1.5936,
+      "step": 12780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.404545454545455e-05,
+      "loss": 1.5989,
+      "step": 12790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4040404040404044e-05,
+      "loss": 1.5887,
+      "step": 12800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4035353535353534e-05,
+      "loss": 1.6241,
+      "step": 12810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.403030303030303e-05,
+      "loss": 1.6066,
+      "step": 12820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4025252525252526e-05,
+      "loss": 1.5943,
+      "step": 12830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.402020202020202e-05,
+      "loss": 1.5886,
+      "step": 12840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.401515151515152e-05,
+      "loss": 1.6021,
+      "step": 12850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4010101010101014e-05,
+      "loss": 1.595,
+      "step": 12860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.40050505050505e-05,
+      "loss": 1.5861,
+      "step": 12870
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 1.5904,
+      "step": 12880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.3994949494949495e-05,
+      "loss": 1.5949,
+      "step": 12890
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.398989898989899e-05,
+      "loss": 1.5959,
+      "step": 12900
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.398484848484849e-05,
+      "loss": 1.6056,
+      "step": 12910
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.397979797979798e-05,
+      "loss": 1.6027,
+      "step": 12920
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.397474747474747e-05,
+      "loss": 1.6065,
+      "step": 12930
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.3969696969696975e-05,
+      "loss": 1.6065,
+      "step": 12940
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.396464646464647e-05,
+      "loss": 1.6077,
+      "step": 12950
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.395959595959596e-05,
+      "loss": 1.5739,
+      "step": 12960
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.3954545454545456e-05,
+      "loss": 1.6039,
+      "step": 12970
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.394949494949495e-05,
+      "loss": 1.5806,
+      "step": 12980
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.394444444444445e-05,
+      "loss": 1.5983,
+      "step": 12990
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3939393939393944e-05,
+      "loss": 1.5801,
+      "step": 13000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.393434343434344e-05,
+      "loss": 1.5793,
+      "step": 13010
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.392929292929293e-05,
+      "loss": 1.5962,
+      "step": 13020
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.392424242424243e-05,
+      "loss": 1.6057,
+      "step": 13030
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.391919191919192e-05,
+      "loss": 1.5895,
+      "step": 13040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.391414141414142e-05,
+      "loss": 1.6221,
+      "step": 13050
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.390909090909091e-05,
+      "loss": 1.5798,
+      "step": 13060
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.390404040404041e-05,
+      "loss": 1.6207,
+      "step": 13070
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.38989898989899e-05,
+      "loss": 1.6115,
+      "step": 13080
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.38939393939394e-05,
+      "loss": 1.5968,
+      "step": 13090
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.388888888888889e-05,
+      "loss": 1.576,
+      "step": 13100
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3883838383838386e-05,
+      "loss": 1.5835,
+      "step": 13110
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.387878787878788e-05,
+      "loss": 1.5886,
+      "step": 13120
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.387373737373738e-05,
+      "loss": 1.6198,
+      "step": 13130
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.386868686868687e-05,
+      "loss": 1.6023,
+      "step": 13140
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.386363636363637e-05,
+      "loss": 1.5731,
+      "step": 13150
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.385858585858586e-05,
+      "loss": 1.5978,
+      "step": 13160
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3853535353535355e-05,
+      "loss": 1.5776,
+      "step": 13170
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.384848484848485e-05,
+      "loss": 1.5741,
+      "step": 13180
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.384343434343435e-05,
+      "loss": 1.5931,
+      "step": 13190
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.383838383838384e-05,
+      "loss": 1.6104,
+      "step": 13200
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.383333333333334e-05,
+      "loss": 1.6144,
+      "step": 13210
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.382828282828283e-05,
+      "loss": 1.6066,
+      "step": 13220
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3823232323232325e-05,
+      "loss": 1.5737,
+      "step": 13230
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.381818181818182e-05,
+      "loss": 1.598,
+      "step": 13240
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.381313131313132e-05,
+      "loss": 1.5802,
+      "step": 13250
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3808080808080806e-05,
+      "loss": 1.5817,
+      "step": 13260
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.380303030303031e-05,
+      "loss": 1.591,
+      "step": 13270
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.37979797979798e-05,
+      "loss": 1.58,
+      "step": 13280
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3792929292929294e-05,
+      "loss": 1.5901,
+      "step": 13290
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.378787878787879e-05,
+      "loss": 1.5961,
+      "step": 13300
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3782828282828286e-05,
+      "loss": 1.6027,
+      "step": 13310
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.377777777777778e-05,
+      "loss": 1.5812,
+      "step": 13320
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.377272727272728e-05,
+      "loss": 1.5966,
+      "step": 13330
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.376767676767677e-05,
+      "loss": 1.5983,
+      "step": 13340
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.376262626262626e-05,
+      "loss": 1.5922,
+      "step": 13350
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.375757575757576e-05,
+      "loss": 1.6068,
+      "step": 13360
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3752525252525255e-05,
+      "loss": 1.5846,
+      "step": 13370
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.374747474747475e-05,
+      "loss": 1.5822,
+      "step": 13380
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.374242424242425e-05,
+      "loss": 1.5986,
+      "step": 13390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3737373737373736e-05,
+      "loss": 1.5856,
+      "step": 13400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.373232323232324e-05,
+      "loss": 1.6076,
+      "step": 13410
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.372727272727273e-05,
+      "loss": 1.5971,
+      "step": 13420
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3722222222222224e-05,
+      "loss": 1.61,
+      "step": 13430
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.371717171717172e-05,
+      "loss": 1.6017,
+      "step": 13440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3712121212121216e-05,
+      "loss": 1.5819,
+      "step": 13450
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3707070707070705e-05,
+      "loss": 1.5958,
+      "step": 13460
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.370202020202021e-05,
+      "loss": 1.5962,
+      "step": 13470
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.36969696969697e-05,
+      "loss": 1.6036,
+      "step": 13480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.369191919191919e-05,
+      "loss": 1.5781,
+      "step": 13490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.368686868686869e-05,
+      "loss": 1.5929,
+      "step": 13500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3681818181818185e-05,
+      "loss": 1.5899,
+      "step": 13510
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3676767676767674e-05,
+      "loss": 1.6081,
+      "step": 13520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.367171717171718e-05,
+      "loss": 1.5808,
+      "step": 13530
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3666666666666666e-05,
+      "loss": 1.5843,
+      "step": 13540
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.366161616161616e-05,
+      "loss": 1.5896,
+      "step": 13550
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.365656565656566e-05,
+      "loss": 1.5919,
+      "step": 13560
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3651515151515154e-05,
+      "loss": 1.5901,
+      "step": 13570
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3646464646464644e-05,
+      "loss": 1.5699,
+      "step": 13580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3641414141414146e-05,
+      "loss": 1.5977,
+      "step": 13590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3636363636363636e-05,
+      "loss": 1.5883,
+      "step": 13600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.363131313131313e-05,
+      "loss": 1.5953,
+      "step": 13610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.362626262626263e-05,
+      "loss": 1.5969,
+      "step": 13620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.3621212121212124e-05,
+      "loss": 1.5952,
+      "step": 13630
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.361616161616161e-05,
+      "loss": 1.6026,
+      "step": 13640
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3611111111111116e-05,
+      "loss": 1.6031,
+      "step": 13650
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3606060606060605e-05,
+      "loss": 1.5811,
+      "step": 13660
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.36010101010101e-05,
+      "loss": 1.5922,
+      "step": 13670
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.35959595959596e-05,
+      "loss": 1.5862,
+      "step": 13680
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.359090909090909e-05,
+      "loss": 1.5904,
+      "step": 13690
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.358585858585859e-05,
+      "loss": 1.6128,
+      "step": 13700
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3580808080808085e-05,
+      "loss": 1.5807,
+      "step": 13710
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3575757575757574e-05,
+      "loss": 1.6089,
+      "step": 13720
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.357070707070707e-05,
+      "loss": 1.5805,
+      "step": 13730
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3565656565656566e-05,
+      "loss": 1.5854,
+      "step": 13740
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.356060606060606e-05,
+      "loss": 1.5839,
+      "step": 13750
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.355555555555556e-05,
+      "loss": 1.6069,
+      "step": 13760
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3550505050505054e-05,
+      "loss": 1.6027,
+      "step": 13770
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.354545454545454e-05,
+      "loss": 1.5777,
+      "step": 13780
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3540404040404046e-05,
+      "loss": 1.5629,
+      "step": 13790
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3535353535353535e-05,
+      "loss": 1.5672,
+      "step": 13800
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.353030303030303e-05,
+      "loss": 1.6119,
+      "step": 13810
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.352525252525253e-05,
+      "loss": 1.5923,
+      "step": 13820
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.352020202020202e-05,
+      "loss": 1.5895,
+      "step": 13830
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.351515151515152e-05,
+      "loss": 1.5971,
+      "step": 13840
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3510101010101015e-05,
+      "loss": 1.5917,
+      "step": 13850
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.350505050505051e-05,
+      "loss": 1.6058,
+      "step": 13860
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.35e-05,
+      "loss": 1.6012,
+      "step": 13870
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3494949494949496e-05,
+      "loss": 1.6131,
+      "step": 13880
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.348989898989899e-05,
+      "loss": 1.5875,
+      "step": 13890
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.348484848484849e-05,
+      "loss": 1.5874,
+      "step": 13900
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3479797979797984e-05,
+      "loss": 1.6223,
+      "step": 13910
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.347474747474748e-05,
+      "loss": 1.5904,
+      "step": 13920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.346969696969697e-05,
+      "loss": 1.6093,
+      "step": 13930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.346464646464647e-05,
+      "loss": 1.5885,
+      "step": 13940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.345959595959596e-05,
+      "loss": 1.6068,
+      "step": 13950
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.345454545454546e-05,
+      "loss": 1.5832,
+      "step": 13960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.344949494949495e-05,
+      "loss": 1.6007,
+      "step": 13970
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.344444444444445e-05,
+      "loss": 1.6064,
+      "step": 13980
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.343939393939394e-05,
+      "loss": 1.6002,
+      "step": 13990
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.343434343434344e-05,
+      "loss": 1.5701,
+      "step": 14000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.342929292929293e-05,
+      "loss": 1.582,
+      "step": 14010
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3424242424242427e-05,
+      "loss": 1.6001,
+      "step": 14020
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.341919191919192e-05,
+      "loss": 1.6115,
+      "step": 14030
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.341414141414142e-05,
+      "loss": 1.5872,
+      "step": 14040
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.340909090909091e-05,
+      "loss": 1.6177,
+      "step": 14050
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.340404040404041e-05,
+      "loss": 1.5755,
+      "step": 14060
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.33989898989899e-05,
+      "loss": 1.587,
+      "step": 14070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3393939393939396e-05,
+      "loss": 1.6144,
+      "step": 14080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.338888888888889e-05,
+      "loss": 1.593,
+      "step": 14090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.338383838383839e-05,
+      "loss": 1.5983,
+      "step": 14100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.337878787878788e-05,
+      "loss": 1.5911,
+      "step": 14110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.337373737373738e-05,
+      "loss": 1.5795,
+      "step": 14120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.336868686868687e-05,
+      "loss": 1.5901,
+      "step": 14130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3363636363636365e-05,
+      "loss": 1.5774,
+      "step": 14140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.335858585858586e-05,
+      "loss": 1.5722,
+      "step": 14150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.335353535353536e-05,
+      "loss": 1.5836,
+      "step": 14160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3348484848484846e-05,
+      "loss": 1.5776,
+      "step": 14170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.334343434343435e-05,
+      "loss": 1.5839,
+      "step": 14180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.333838383838384e-05,
+      "loss": 1.6023,
+      "step": 14190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3333333333333334e-05,
+      "loss": 1.5901,
+      "step": 14200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.332828282828283e-05,
+      "loss": 1.5934,
+      "step": 14210
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3323232323232326e-05,
+      "loss": 1.6002,
+      "step": 14220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.331818181818182e-05,
+      "loss": 1.5975,
+      "step": 14230
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.331313131313132e-05,
+      "loss": 1.5927,
+      "step": 14240
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.330808080808081e-05,
+      "loss": 1.5829,
+      "step": 14250
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.33030303030303e-05,
+      "loss": 1.5856,
+      "step": 14260
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.32979797979798e-05,
+      "loss": 1.6088,
+      "step": 14270
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.3292929292929295e-05,
+      "loss": 1.5747,
+      "step": 14280
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.328787878787879e-05,
+      "loss": 1.5932,
+      "step": 14290
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.328282828282829e-05,
+      "loss": 1.5918,
+      "step": 14300
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3277777777777776e-05,
+      "loss": 1.5819,
+      "step": 14310
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.327272727272728e-05,
+      "loss": 1.5789,
+      "step": 14320
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.326767676767677e-05,
+      "loss": 1.5962,
+      "step": 14330
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3262626262626264e-05,
+      "loss": 1.5701,
+      "step": 14340
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.325757575757576e-05,
+      "loss": 1.5922,
+      "step": 14350
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3252525252525256e-05,
+      "loss": 1.5969,
+      "step": 14360
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3247474747474746e-05,
+      "loss": 1.6106,
+      "step": 14370
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.324242424242425e-05,
+      "loss": 1.6246,
+      "step": 14380
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.323737373737374e-05,
+      "loss": 1.5884,
+      "step": 14390
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3232323232323234e-05,
+      "loss": 1.6057,
+      "step": 14400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.322727272727273e-05,
+      "loss": 1.5858,
+      "step": 14410
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3222222222222226e-05,
+      "loss": 1.575,
+      "step": 14420
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3217171717171715e-05,
+      "loss": 1.5763,
+      "step": 14430
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.321212121212122e-05,
+      "loss": 1.6019,
+      "step": 14440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.320707070707071e-05,
+      "loss": 1.5897,
+      "step": 14450
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.32020202020202e-05,
+      "loss": 1.5697,
+      "step": 14460
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.31969696969697e-05,
+      "loss": 1.5622,
+      "step": 14470
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3191919191919195e-05,
+      "loss": 1.5754,
+      "step": 14480
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3186868686868684e-05,
+      "loss": 1.5926,
+      "step": 14490
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.318181818181819e-05,
+      "loss": 1.5914,
+      "step": 14500
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3176767676767676e-05,
+      "loss": 1.6096,
+      "step": 14510
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.317171717171717e-05,
+      "loss": 1.6161,
+      "step": 14520
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.316666666666667e-05,
+      "loss": 1.6037,
+      "step": 14530
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3161616161616164e-05,
+      "loss": 1.581,
+      "step": 14540
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.315656565656565e-05,
+      "loss": 1.5892,
+      "step": 14550
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3151515151515156e-05,
+      "loss": 1.601,
+      "step": 14560
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3146464646464645e-05,
+      "loss": 1.6088,
+      "step": 14570
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.314141414141414e-05,
+      "loss": 1.5776,
+      "step": 14580
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.313636363636364e-05,
+      "loss": 1.5786,
+      "step": 14590
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.313131313131313e-05,
+      "loss": 1.5822,
+      "step": 14600
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.312626262626263e-05,
+      "loss": 1.609,
+      "step": 14610
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3121212121212125e-05,
+      "loss": 1.5971,
+      "step": 14620
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3116161616161614e-05,
+      "loss": 1.5756,
+      "step": 14630
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.311111111111111e-05,
+      "loss": 1.5889,
+      "step": 14640
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3106060606060606e-05,
+      "loss": 1.6019,
+      "step": 14650
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.31010101010101e-05,
+      "loss": 1.5931,
+      "step": 14660
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.30959595959596e-05,
+      "loss": 1.5699,
+      "step": 14670
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3090909090909094e-05,
+      "loss": 1.5887,
+      "step": 14680
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.308585858585859e-05,
+      "loss": 1.5871,
+      "step": 14690
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.308080808080808e-05,
+      "loss": 1.5917,
+      "step": 14700
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.307575757575758e-05,
+      "loss": 1.5776,
+      "step": 14710
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.307070707070707e-05,
+      "loss": 1.5728,
+      "step": 14720
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.306565656565657e-05,
+      "loss": 1.5731,
+      "step": 14730
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.306060606060606e-05,
+      "loss": 1.6105,
+      "step": 14740
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.305555555555556e-05,
+      "loss": 1.6002,
+      "step": 14750
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3050505050505055e-05,
+      "loss": 1.5855,
+      "step": 14760
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.304545454545455e-05,
+      "loss": 1.6013,
+      "step": 14770
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.304040404040404e-05,
+      "loss": 1.6025,
+      "step": 14780
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3035353535353536e-05,
+      "loss": 1.5742,
+      "step": 14790
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.303030303030303e-05,
+      "loss": 1.6065,
+      "step": 14800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.302525252525253e-05,
+      "loss": 1.5783,
+      "step": 14810
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3020202020202024e-05,
+      "loss": 1.591,
+      "step": 14820
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.301515151515152e-05,
+      "loss": 1.5802,
+      "step": 14830
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.301010101010101e-05,
+      "loss": 1.5853,
+      "step": 14840
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.300505050505051e-05,
+      "loss": 1.61,
+      "step": 14850
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.3e-05,
+      "loss": 1.5801,
+      "step": 14860
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.29949494949495e-05,
+      "loss": 1.5866,
+      "step": 14870
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.2989898989898994e-05,
+      "loss": 1.5868,
+      "step": 14880
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.298484848484849e-05,
+      "loss": 1.5877,
+      "step": 14890
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.297979797979798e-05,
+      "loss": 1.5946,
+      "step": 14900
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.297474747474748e-05,
+      "loss": 1.5855,
+      "step": 14910
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.296969696969697e-05,
+      "loss": 1.6013,
+      "step": 14920
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.296464646464647e-05,
+      "loss": 1.5873,
+      "step": 14930
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.295959595959596e-05,
+      "loss": 1.5937,
+      "step": 14940
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.295454545454546e-05,
+      "loss": 1.5859,
+      "step": 14950
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.294949494949495e-05,
+      "loss": 1.5831,
+      "step": 14960
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.294444444444445e-05,
+      "loss": 1.5921,
+      "step": 14970
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.293939393939394e-05,
+      "loss": 1.59,
+      "step": 14980
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2934343434343436e-05,
+      "loss": 1.5905,
+      "step": 14990
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.292929292929293e-05,
+      "loss": 1.5944,
+      "step": 15000
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.292424242424243e-05,
+      "loss": 1.6117,
+      "step": 15010
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.291919191919192e-05,
+      "loss": 1.5864,
+      "step": 15020
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.291414141414142e-05,
+      "loss": 1.6016,
+      "step": 15030
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.290909090909091e-05,
+      "loss": 1.6028,
+      "step": 15040
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2904040404040405e-05,
+      "loss": 1.6017,
+      "step": 15050
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.28989898989899e-05,
+      "loss": 1.5799,
+      "step": 15060
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.28939393939394e-05,
+      "loss": 1.5735,
+      "step": 15070
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2888888888888886e-05,
+      "loss": 1.5759,
+      "step": 15080
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.288383838383839e-05,
+      "loss": 1.5991,
+      "step": 15090
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.287878787878788e-05,
+      "loss": 1.5906,
+      "step": 15100
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2873737373737374e-05,
+      "loss": 1.5986,
+      "step": 15110
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.286868686868687e-05,
+      "loss": 1.5693,
+      "step": 15120
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2863636363636366e-05,
+      "loss": 1.597,
+      "step": 15130
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.285858585858586e-05,
+      "loss": 1.5967,
+      "step": 15140
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.285353535353536e-05,
+      "loss": 1.5678,
+      "step": 15150
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.284848484848485e-05,
+      "loss": 1.5811,
+      "step": 15160
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2843434343434343e-05,
+      "loss": 1.5953,
+      "step": 15170
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.283838383838384e-05,
+      "loss": 1.5936,
+      "step": 15180
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2833333333333335e-05,
+      "loss": 1.6117,
+      "step": 15190
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.282828282828283e-05,
+      "loss": 1.5978,
+      "step": 15200
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.282323232323233e-05,
+      "loss": 1.5781,
+      "step": 15210
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.281818181818182e-05,
+      "loss": 1.5867,
+      "step": 15220
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.281313131313132e-05,
+      "loss": 1.5803,
+      "step": 15230
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.280808080808081e-05,
+      "loss": 1.5785,
+      "step": 15240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2803030303030305e-05,
+      "loss": 1.579,
+      "step": 15250
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.27979797979798e-05,
+      "loss": 1.6106,
+      "step": 15260
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2792929292929297e-05,
+      "loss": 1.5714,
+      "step": 15270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2787878787878786e-05,
+      "loss": 1.5706,
+      "step": 15280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.278282828282829e-05,
+      "loss": 1.5848,
+      "step": 15290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.277777777777778e-05,
+      "loss": 1.5933,
+      "step": 15300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2772727272727274e-05,
+      "loss": 1.583,
+      "step": 15310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.276767676767677e-05,
+      "loss": 1.5835,
+      "step": 15320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2762626262626266e-05,
+      "loss": 1.5852,
+      "step": 15330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2757575757575755e-05,
+      "loss": 1.5978,
+      "step": 15340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.275252525252526e-05,
+      "loss": 1.5953,
+      "step": 15350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.274747474747475e-05,
+      "loss": 1.5675,
+      "step": 15360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.274242424242424e-05,
+      "loss": 1.6158,
+      "step": 15370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.273737373737374e-05,
+      "loss": 1.5806,
+      "step": 15380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2732323232323235e-05,
+      "loss": 1.5946,
+      "step": 15390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2727272727272724e-05,
+      "loss": 1.5846,
+      "step": 15400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.272222222222223e-05,
+      "loss": 1.6081,
+      "step": 15410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2717171717171716e-05,
+      "loss": 1.5959,
+      "step": 15420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.271212121212121e-05,
+      "loss": 1.5759,
+      "step": 15430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.270707070707071e-05,
+      "loss": 1.5683,
+      "step": 15440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2702020202020204e-05,
+      "loss": 1.5752,
+      "step": 15450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.269696969696969e-05,
+      "loss": 1.5833,
+      "step": 15460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2691919191919196e-05,
+      "loss": 1.5712,
+      "step": 15470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2686868686868685e-05,
+      "loss": 1.6014,
+      "step": 15480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.268181818181818e-05,
+      "loss": 1.5985,
+      "step": 15490
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.267676767676768e-05,
+      "loss": 1.5789,
+      "step": 15500
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.267171717171717e-05,
+      "loss": 1.5998,
+      "step": 15510
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.266666666666667e-05,
+      "loss": 1.6018,
+      "step": 15520
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2661616161616165e-05,
+      "loss": 1.5919,
+      "step": 15530
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.265656565656566e-05,
+      "loss": 1.5937,
+      "step": 15540
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.265151515151515e-05,
+      "loss": 1.5752,
+      "step": 15550
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.264646464646465e-05,
+      "loss": 1.5905,
+      "step": 15560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.264141414141414e-05,
+      "loss": 1.5514,
+      "step": 15570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.263636363636364e-05,
+      "loss": 1.5797,
+      "step": 15580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.2631313131313134e-05,
+      "loss": 1.5775,
+      "step": 15590
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.262626262626263e-05,
+      "loss": 1.5646,
+      "step": 15600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.262121212121212e-05,
+      "loss": 1.5951,
+      "step": 15610
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.261616161616162e-05,
+      "loss": 1.596,
+      "step": 15620
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.261111111111111e-05,
+      "loss": 1.5807,
+      "step": 15630
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.260606060606061e-05,
+      "loss": 1.586,
+      "step": 15640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2601010101010104e-05,
+      "loss": 1.6004,
+      "step": 15650
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.25959595959596e-05,
+      "loss": 1.6135,
+      "step": 15660
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2590909090909096e-05,
+      "loss": 1.5842,
+      "step": 15670
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.258585858585859e-05,
+      "loss": 1.5852,
+      "step": 15680
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.258080808080808e-05,
+      "loss": 1.5757,
+      "step": 15690
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.257575757575758e-05,
+      "loss": 1.6023,
+      "step": 15700
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.257070707070707e-05,
+      "loss": 1.593,
+      "step": 15710
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.256565656565657e-05,
+      "loss": 1.5796,
+      "step": 15720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2560606060606065e-05,
+      "loss": 1.5849,
+      "step": 15730
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.255555555555556e-05,
+      "loss": 1.5663,
+      "step": 15740
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.255050505050505e-05,
+      "loss": 1.5935,
+      "step": 15750
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.254545454545455e-05,
+      "loss": 1.5745,
+      "step": 15760
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.254040404040404e-05,
+      "loss": 1.5846,
+      "step": 15770
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.253535353535354e-05,
+      "loss": 1.5855,
+      "step": 15780
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2530303030303034e-05,
+      "loss": 1.5821,
+      "step": 15790
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.252525252525253e-05,
+      "loss": 1.6051,
+      "step": 15800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.252020202020202e-05,
+      "loss": 1.5824,
+      "step": 15810
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.251515151515152e-05,
+      "loss": 1.5928,
+      "step": 15820
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.251010101010101e-05,
+      "loss": 1.5895,
+      "step": 15830
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.250505050505051e-05,
+      "loss": 1.5594,
+      "step": 15840
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.25e-05,
+      "loss": 1.5603,
+      "step": 15850
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.24949494949495e-05,
+      "loss": 1.5762,
+      "step": 15860
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.248989898989899e-05,
+      "loss": 1.6143,
+      "step": 15870
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.248484848484849e-05,
+      "loss": 1.5938,
+      "step": 15880
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.247979797979798e-05,
+      "loss": 1.5913,
+      "step": 15890
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2474747474747476e-05,
+      "loss": 1.5633,
+      "step": 15900
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.246969696969697e-05,
+      "loss": 1.5731,
+      "step": 15910
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.246464646464647e-05,
+      "loss": 1.5616,
+      "step": 15920
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.245959595959596e-05,
+      "loss": 1.5696,
+      "step": 15930
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.245454545454546e-05,
+      "loss": 1.5856,
+      "step": 15940
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.244949494949495e-05,
+      "loss": 1.5965,
+      "step": 15950
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2444444444444445e-05,
+      "loss": 1.5887,
+      "step": 15960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.243939393939394e-05,
+      "loss": 1.5805,
+      "step": 15970
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.243434343434344e-05,
+      "loss": 1.5909,
+      "step": 15980
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2429292929292927e-05,
+      "loss": 1.6004,
+      "step": 15990
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.242424242424243e-05,
+      "loss": 1.5832,
+      "step": 16000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.241919191919192e-05,
+      "loss": 1.5794,
+      "step": 16010
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2414141414141415e-05,
+      "loss": 1.5784,
+      "step": 16020
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.240909090909091e-05,
+      "loss": 1.5668,
+      "step": 16030
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2404040404040407e-05,
+      "loss": 1.5921,
+      "step": 16040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.23989898989899e-05,
+      "loss": 1.5799,
+      "step": 16050
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.23939393939394e-05,
+      "loss": 1.598,
+      "step": 16060
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.238888888888889e-05,
+      "loss": 1.5641,
+      "step": 16070
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2383838383838384e-05,
+      "loss": 1.5965,
+      "step": 16080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.237878787878788e-05,
+      "loss": 1.584,
+      "step": 16090
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2373737373737376e-05,
+      "loss": 1.5885,
+      "step": 16100
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.236868686868687e-05,
+      "loss": 1.6178,
+      "step": 16110
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.236363636363637e-05,
+      "loss": 1.5726,
+      "step": 16120
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.235858585858586e-05,
+      "loss": 1.6265,
+      "step": 16130
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.235353535353536e-05,
+      "loss": 1.5856,
+      "step": 16140
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.234848484848485e-05,
+      "loss": 1.5513,
+      "step": 16150
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2343434343434345e-05,
+      "loss": 1.5967,
+      "step": 16160
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.233838383838384e-05,
+      "loss": 1.5864,
+      "step": 16170
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.233333333333334e-05,
+      "loss": 1.5953,
+      "step": 16180
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2328282828282826e-05,
+      "loss": 1.5811,
+      "step": 16190
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.232323232323233e-05,
+      "loss": 1.5929,
+      "step": 16200
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.231818181818182e-05,
+      "loss": 1.5809,
+      "step": 16210
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2313131313131314e-05,
+      "loss": 1.5837,
+      "step": 16220
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.230808080808081e-05,
+      "loss": 1.576,
+      "step": 16230
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2303030303030306e-05,
+      "loss": 1.5838,
+      "step": 16240
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.2297979797979795e-05,
+      "loss": 1.5894,
+      "step": 16250
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.22929292929293e-05,
+      "loss": 1.592,
+      "step": 16260
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.228787878787879e-05,
+      "loss": 1.5847,
+      "step": 16270
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.228282828282828e-05,
+      "loss": 1.5874,
+      "step": 16280
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.227777777777778e-05,
+      "loss": 1.5777,
+      "step": 16290
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2272727272727275e-05,
+      "loss": 1.6059,
+      "step": 16300
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2267676767676764e-05,
+      "loss": 1.5675,
+      "step": 16310
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.226262626262627e-05,
+      "loss": 1.5926,
+      "step": 16320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2257575757575756e-05,
+      "loss": 1.5791,
+      "step": 16330
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.225252525252525e-05,
+      "loss": 1.5695,
+      "step": 16340
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.224747474747475e-05,
+      "loss": 1.5694,
+      "step": 16350
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2242424242424244e-05,
+      "loss": 1.587,
+      "step": 16360
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.223737373737374e-05,
+      "loss": 1.5558,
+      "step": 16370
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2232323232323236e-05,
+      "loss": 1.5985,
+      "step": 16380
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.222727272727273e-05,
+      "loss": 1.5878,
+      "step": 16390
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.222222222222222e-05,
+      "loss": 1.6087,
+      "step": 16400
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2217171717171724e-05,
+      "loss": 1.6023,
+      "step": 16410
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2212121212121214e-05,
+      "loss": 1.5923,
+      "step": 16420
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.220707070707071e-05,
+      "loss": 1.5773,
+      "step": 16430
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2202020202020205e-05,
+      "loss": 1.5689,
+      "step": 16440
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.21969696969697e-05,
+      "loss": 1.5948,
+      "step": 16450
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.219191919191919e-05,
+      "loss": 1.5614,
+      "step": 16460
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2186868686868693e-05,
+      "loss": 1.5879,
+      "step": 16470
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.218181818181818e-05,
+      "loss": 1.5936,
+      "step": 16480
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.217676767676768e-05,
+      "loss": 1.5822,
+      "step": 16490
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2171717171717175e-05,
+      "loss": 1.5785,
+      "step": 16500
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.216666666666667e-05,
+      "loss": 1.5824,
+      "step": 16510
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.216161616161616e-05,
+      "loss": 1.5984,
+      "step": 16520
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.215656565656566e-05,
+      "loss": 1.5657,
+      "step": 16530
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.215151515151515e-05,
+      "loss": 1.5937,
+      "step": 16540
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.214646464646465e-05,
+      "loss": 1.5809,
+      "step": 16550
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2141414141414144e-05,
+      "loss": 1.5859,
+      "step": 16560
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.213636363636364e-05,
+      "loss": 1.5806,
+      "step": 16570
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2131313131313136e-05,
+      "loss": 1.5906,
+      "step": 16580
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.212626262626263e-05,
+      "loss": 1.5718,
+      "step": 16590
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.212121212121212e-05,
+      "loss": 1.5793,
+      "step": 16600
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.211616161616162e-05,
+      "loss": 1.5736,
+      "step": 16610
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.211111111111111e-05,
+      "loss": 1.5772,
+      "step": 16620
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.210606060606061e-05,
+      "loss": 1.5843,
+      "step": 16630
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2101010101010105e-05,
+      "loss": 1.5854,
+      "step": 16640
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.20959595959596e-05,
+      "loss": 1.5772,
+      "step": 16650
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.209090909090909e-05,
+      "loss": 1.6011,
+      "step": 16660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.208585858585859e-05,
+      "loss": 1.5876,
+      "step": 16670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.208080808080808e-05,
+      "loss": 1.5799,
+      "step": 16680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.207575757575758e-05,
+      "loss": 1.5875,
+      "step": 16690
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2070707070707074e-05,
+      "loss": 1.5872,
+      "step": 16700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.206565656565657e-05,
+      "loss": 1.5914,
+      "step": 16710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.206060606060606e-05,
+      "loss": 1.5941,
+      "step": 16720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.205555555555556e-05,
+      "loss": 1.5988,
+      "step": 16730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.205050505050505e-05,
+      "loss": 1.5928,
+      "step": 16740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.204545454545455e-05,
+      "loss": 1.5818,
+      "step": 16750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.204040404040404e-05,
+      "loss": 1.6047,
+      "step": 16760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.203535353535354e-05,
+      "loss": 1.5766,
+      "step": 16770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.203030303030303e-05,
+      "loss": 1.5584,
+      "step": 16780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.202525252525253e-05,
+      "loss": 1.5805,
+      "step": 16790
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.202020202020202e-05,
+      "loss": 1.5704,
+      "step": 16800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2015151515151516e-05,
+      "loss": 1.5664,
+      "step": 16810
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.201010101010101e-05,
+      "loss": 1.5921,
+      "step": 16820
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.200505050505051e-05,
+      "loss": 1.5877,
+      "step": 16830
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.2e-05,
+      "loss": 1.5899,
+      "step": 16840
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.19949494949495e-05,
+      "loss": 1.5859,
+      "step": 16850
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.198989898989899e-05,
+      "loss": 1.5893,
+      "step": 16860
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.1984848484848486e-05,
+      "loss": 1.5727,
+      "step": 16870
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.197979797979798e-05,
+      "loss": 1.5705,
+      "step": 16880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.197474747474748e-05,
+      "loss": 1.6082,
+      "step": 16890
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.196969696969697e-05,
+      "loss": 1.5972,
+      "step": 16900
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.196464646464647e-05,
+      "loss": 1.5651,
+      "step": 16910
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.195959595959596e-05,
+      "loss": 1.5877,
+      "step": 16920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.1954545454545455e-05,
+      "loss": 1.5779,
+      "step": 16930
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.194949494949495e-05,
+      "loss": 1.5814,
+      "step": 16940
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.194444444444445e-05,
+      "loss": 1.5754,
+      "step": 16950
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.193939393939394e-05,
+      "loss": 1.5919,
+      "step": 16960
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.193434343434344e-05,
+      "loss": 1.5993,
+      "step": 16970
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.192929292929293e-05,
+      "loss": 1.585,
+      "step": 16980
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1924242424242424e-05,
+      "loss": 1.5825,
+      "step": 16990
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.191919191919192e-05,
+      "loss": 1.5636,
+      "step": 17000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1914141414141416e-05,
+      "loss": 1.5864,
+      "step": 17010
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.190909090909091e-05,
+      "loss": 1.5646,
+      "step": 17020
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.190404040404041e-05,
+      "loss": 1.5783,
+      "step": 17030
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.18989898989899e-05,
+      "loss": 1.5796,
+      "step": 17040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.189393939393939e-05,
+      "loss": 1.5804,
+      "step": 17050
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.188888888888889e-05,
+      "loss": 1.5728,
+      "step": 17060
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1883838383838385e-05,
+      "loss": 1.5875,
+      "step": 17070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.187878787878788e-05,
+      "loss": 1.574,
+      "step": 17080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.187373737373738e-05,
+      "loss": 1.5693,
+      "step": 17090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1868686868686866e-05,
+      "loss": 1.5889,
+      "step": 17100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.186363636363637e-05,
+      "loss": 1.5864,
+      "step": 17110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.185858585858586e-05,
+      "loss": 1.5989,
+      "step": 17120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1853535353535354e-05,
+      "loss": 1.5926,
+      "step": 17130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.184848484848485e-05,
+      "loss": 1.5744,
+      "step": 17140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1843434343434346e-05,
+      "loss": 1.5905,
+      "step": 17150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1838383838383835e-05,
+      "loss": 1.5573,
+      "step": 17160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.183333333333334e-05,
+      "loss": 1.5854,
+      "step": 17170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.182828282828283e-05,
+      "loss": 1.6059,
+      "step": 17180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1823232323232323e-05,
+      "loss": 1.5733,
+      "step": 17190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.181818181818182e-05,
+      "loss": 1.5908,
+      "step": 17200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1813131313131315e-05,
+      "loss": 1.5577,
+      "step": 17210
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.180808080808081e-05,
+      "loss": 1.5702,
+      "step": 17220
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.180303030303031e-05,
+      "loss": 1.5888,
+      "step": 17230
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1797979797979803e-05,
+      "loss": 1.6065,
+      "step": 17240
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.179292929292929e-05,
+      "loss": 1.5913,
+      "step": 17250
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1787878787878795e-05,
+      "loss": 1.5653,
+      "step": 17260
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1782828282828285e-05,
+      "loss": 1.5703,
+      "step": 17270
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.177777777777778e-05,
+      "loss": 1.5767,
+      "step": 17280
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1772727272727277e-05,
+      "loss": 1.5769,
+      "step": 17290
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.176767676767677e-05,
+      "loss": 1.6098,
+      "step": 17300
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.176262626262626e-05,
+      "loss": 1.5816,
+      "step": 17310
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1757575757575765e-05,
+      "loss": 1.5739,
+      "step": 17320
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1752525252525254e-05,
+      "loss": 1.5927,
+      "step": 17330
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.174747474747475e-05,
+      "loss": 1.5707,
+      "step": 17340
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1742424242424246e-05,
+      "loss": 1.5738,
+      "step": 17350
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.173737373737374e-05,
+      "loss": 1.5948,
+      "step": 17360
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.173232323232323e-05,
+      "loss": 1.602,
+      "step": 17370
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1727272727272734e-05,
+      "loss": 1.5725,
+      "step": 17380
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.172222222222222e-05,
+      "loss": 1.577,
+      "step": 17390
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.171717171717172e-05,
+      "loss": 1.5695,
+      "step": 17400
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1712121212121215e-05,
+      "loss": 1.5674,
+      "step": 17410
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.170707070707071e-05,
+      "loss": 1.6036,
+      "step": 17420
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.17020202020202e-05,
+      "loss": 1.5858,
+      "step": 17430
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.16969696969697e-05,
+      "loss": 1.5815,
+      "step": 17440
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.169191919191919e-05,
+      "loss": 1.5873,
+      "step": 17450
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.168686868686869e-05,
+      "loss": 1.5688,
+      "step": 17460
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1681818181818184e-05,
+      "loss": 1.5602,
+      "step": 17470
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.167676767676768e-05,
+      "loss": 1.5804,
+      "step": 17480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1671717171717176e-05,
+      "loss": 1.5857,
+      "step": 17490
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.166666666666667e-05,
+      "loss": 1.5692,
+      "step": 17500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.166161616161616e-05,
+      "loss": 1.5893,
+      "step": 17510
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.165656565656566e-05,
+      "loss": 1.5767,
+      "step": 17520
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.165151515151515e-05,
+      "loss": 1.5673,
+      "step": 17530
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.164646464646465e-05,
+      "loss": 1.5755,
+      "step": 17540
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1641414141414145e-05,
+      "loss": 1.5871,
+      "step": 17550
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.163636363636364e-05,
+      "loss": 1.603,
+      "step": 17560
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.163131313131313e-05,
+      "loss": 1.5805,
+      "step": 17570
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.162626262626263e-05,
+      "loss": 1.573,
+      "step": 17580
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.162121212121212e-05,
+      "loss": 1.579,
+      "step": 17590
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.161616161616162e-05,
+      "loss": 1.5841,
+      "step": 17600
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.1611111111111114e-05,
+      "loss": 1.5805,
+      "step": 17610
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.160606060606061e-05,
+      "loss": 1.5494,
+      "step": 17620
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.16010101010101e-05,
+      "loss": 1.5965,
+      "step": 17630
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.15959595959596e-05,
+      "loss": 1.587,
+      "step": 17640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.159090909090909e-05,
+      "loss": 1.6161,
+      "step": 17650
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.158585858585859e-05,
+      "loss": 1.5954,
+      "step": 17660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1580808080808084e-05,
+      "loss": 1.5862,
+      "step": 17670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.157575757575758e-05,
+      "loss": 1.5756,
+      "step": 17680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.157070707070707e-05,
+      "loss": 1.5904,
+      "step": 17690
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.156565656565657e-05,
+      "loss": 1.577,
+      "step": 17700
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.156060606060606e-05,
+      "loss": 1.5644,
+      "step": 17710
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.155555555555556e-05,
+      "loss": 1.5651,
+      "step": 17720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.155050505050505e-05,
+      "loss": 1.5732,
+      "step": 17730
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.154545454545455e-05,
+      "loss": 1.5988,
+      "step": 17740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.154040404040404e-05,
+      "loss": 1.5837,
+      "step": 17750
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.153535353535354e-05,
+      "loss": 1.5714,
+      "step": 17760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.153030303030303e-05,
+      "loss": 1.5883,
+      "step": 17770
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1525252525252526e-05,
+      "loss": 1.5682,
+      "step": 17780
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.152020202020202e-05,
+      "loss": 1.5783,
+      "step": 17790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.151515151515152e-05,
+      "loss": 1.5785,
+      "step": 17800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.151010101010101e-05,
+      "loss": 1.5688,
+      "step": 17810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.150505050505051e-05,
+      "loss": 1.5456,
+      "step": 17820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.15e-05,
+      "loss": 1.557,
+      "step": 17830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1494949494949495e-05,
+      "loss": 1.5876,
+      "step": 17840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.148989898989899e-05,
+      "loss": 1.5692,
+      "step": 17850
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.148484848484849e-05,
+      "loss": 1.5866,
+      "step": 17860
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.147979797979798e-05,
+      "loss": 1.5984,
+      "step": 17870
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.147474747474748e-05,
+      "loss": 1.5688,
+      "step": 17880
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.146969696969697e-05,
+      "loss": 1.5747,
+      "step": 17890
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1464646464646464e-05,
+      "loss": 1.5899,
+      "step": 17900
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.145959595959596e-05,
+      "loss": 1.5666,
+      "step": 17910
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1454545454545456e-05,
+      "loss": 1.5839,
+      "step": 17920
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.144949494949495e-05,
+      "loss": 1.5903,
+      "step": 17930
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.144444444444445e-05,
+      "loss": 1.5755,
+      "step": 17940
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.143939393939394e-05,
+      "loss": 1.6011,
+      "step": 17950
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1434343434343433e-05,
+      "loss": 1.5686,
+      "step": 17960
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.142929292929293e-05,
+      "loss": 1.5607,
+      "step": 17970
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1424242424242425e-05,
+      "loss": 1.5827,
+      "step": 17980
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.141919191919192e-05,
+      "loss": 1.5841,
+      "step": 17990
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.141414141414142e-05,
+      "loss": 1.5709,
+      "step": 18000
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1409090909090907e-05,
+      "loss": 1.5851,
+      "step": 18010
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.140404040404041e-05,
+      "loss": 1.5713,
+      "step": 18020
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.13989898989899e-05,
+      "loss": 1.5861,
+      "step": 18030
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1393939393939395e-05,
+      "loss": 1.5929,
+      "step": 18040
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.138888888888889e-05,
+      "loss": 1.5789,
+      "step": 18050
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1383838383838387e-05,
+      "loss": 1.5684,
+      "step": 18060
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.137878787878788e-05,
+      "loss": 1.5649,
+      "step": 18070
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.137373737373738e-05,
+      "loss": 1.562,
+      "step": 18080
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1368686868686874e-05,
+      "loss": 1.5855,
+      "step": 18090
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1363636363636364e-05,
+      "loss": 1.5898,
+      "step": 18100
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1358585858585866e-05,
+      "loss": 1.5684,
+      "step": 18110
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1353535353535356e-05,
+      "loss": 1.5854,
+      "step": 18120
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.134848484848485e-05,
+      "loss": 1.5923,
+      "step": 18130
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.134343434343435e-05,
+      "loss": 1.5819,
+      "step": 18140
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1338383838383844e-05,
+      "loss": 1.5585,
+      "step": 18150
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.133333333333333e-05,
+      "loss": 1.5878,
+      "step": 18160
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1328282828282836e-05,
+      "loss": 1.5921,
+      "step": 18170
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1323232323232325e-05,
+      "loss": 1.5849,
+      "step": 18180
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.131818181818182e-05,
+      "loss": 1.5945,
+      "step": 18190
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.131313131313132e-05,
+      "loss": 1.5889,
+      "step": 18200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.130808080808081e-05,
+      "loss": 1.5618,
+      "step": 18210
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.13030303030303e-05,
+      "loss": 1.5746,
+      "step": 18220
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1297979797979805e-05,
+      "loss": 1.567,
+      "step": 18230
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1292929292929294e-05,
+      "loss": 1.5743,
+      "step": 18240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.128787878787879e-05,
+      "loss": 1.5742,
+      "step": 18250
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1282828282828286e-05,
+      "loss": 1.5904,
+      "step": 18260
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.127777777777778e-05,
+      "loss": 1.5852,
+      "step": 18270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.127272727272727e-05,
+      "loss": 1.5974,
+      "step": 18280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.1267676767676774e-05,
+      "loss": 1.5764,
+      "step": 18290
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.126262626262626e-05,
+      "loss": 1.5878,
+      "step": 18300
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.125757575757576e-05,
+      "loss": 1.5702,
+      "step": 18310
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.1252525252525255e-05,
+      "loss": 1.5725,
+      "step": 18320
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.124747474747475e-05,
+      "loss": 1.5637,
+      "step": 18330
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.124242424242424e-05,
+      "loss": 1.568,
+      "step": 18340
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.123737373737374e-05,
+      "loss": 1.5828,
+      "step": 18350
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.123232323232323e-05,
+      "loss": 1.6045,
+      "step": 18360
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.122727272727273e-05,
+      "loss": 1.5875,
+      "step": 18370
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.1222222222222224e-05,
+      "loss": 1.5822,
+      "step": 18380
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.121717171717172e-05,
+      "loss": 1.5807,
+      "step": 18390
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.1212121212121216e-05,
+      "loss": 1.5789,
+      "step": 18400
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.120707070707071e-05,
+      "loss": 1.5731,
+      "step": 18410
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.12020202020202e-05,
+      "loss": 1.5435,
+      "step": 18420
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.11969696969697e-05,
+      "loss": 1.5574,
+      "step": 18430
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.1191919191919193e-05,
+      "loss": 1.5838,
+      "step": 18440
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.118686868686869e-05,
+      "loss": 1.5413,
+      "step": 18450
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.1181818181818185e-05,
+      "loss": 1.5638,
+      "step": 18460
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.117676767676768e-05,
+      "loss": 1.5541,
+      "step": 18470
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.117171717171717e-05,
+      "loss": 1.5668,
+      "step": 18480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.116666666666667e-05,
+      "loss": 1.5616,
+      "step": 18490
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.116161616161616e-05,
+      "loss": 1.5792,
+      "step": 18500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.115656565656566e-05,
+      "loss": 1.5686,
+      "step": 18510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.1151515151515155e-05,
+      "loss": 1.579,
+      "step": 18520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.114646464646465e-05,
+      "loss": 1.5715,
+      "step": 18530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.114141414141414e-05,
+      "loss": 1.5806,
+      "step": 18540
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.113636363636364e-05,
+      "loss": 1.5709,
+      "step": 18550
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.113131313131313e-05,
+      "loss": 1.5976,
+      "step": 18560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.112626262626263e-05,
+      "loss": 1.5786,
+      "step": 18570
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.1121212121212124e-05,
+      "loss": 1.5721,
+      "step": 18580
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.111616161616162e-05,
+      "loss": 1.5689,
+      "step": 18590
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.111111111111111e-05,
+      "loss": 1.5646,
+      "step": 18600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.110606060606061e-05,
+      "loss": 1.5859,
+      "step": 18610
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.11010101010101e-05,
+      "loss": 1.5673,
+      "step": 18620
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.10959595959596e-05,
+      "loss": 1.5494,
+      "step": 18630
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.109090909090909e-05,
+      "loss": 1.5886,
+      "step": 18640
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.108585858585859e-05,
+      "loss": 1.5742,
+      "step": 18650
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.108080808080808e-05,
+      "loss": 1.5921,
+      "step": 18660
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.107575757575758e-05,
+      "loss": 1.5923,
+      "step": 18670
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.107070707070707e-05,
+      "loss": 1.5854,
+      "step": 18680
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.1065656565656566e-05,
+      "loss": 1.5775,
+      "step": 18690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.106060606060606e-05,
+      "loss": 1.5914,
+      "step": 18700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.105555555555556e-05,
+      "loss": 1.5553,
+      "step": 18710
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.105050505050505e-05,
+      "loss": 1.5633,
+      "step": 18720
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.104545454545455e-05,
+      "loss": 1.5848,
+      "step": 18730
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.104040404040404e-05,
+      "loss": 1.5699,
+      "step": 18740
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.1035353535353535e-05,
+      "loss": 1.5651,
+      "step": 18750
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.103030303030303e-05,
+      "loss": 1.5814,
+      "step": 18760
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.102525252525253e-05,
+      "loss": 1.5879,
+      "step": 18770
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.1020202020202017e-05,
+      "loss": 1.5763,
+      "step": 18780
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.101515151515152e-05,
+      "loss": 1.5901,
+      "step": 18790
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.101010101010101e-05,
+      "loss": 1.5802,
+      "step": 18800
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.1005050505050504e-05,
+      "loss": 1.5752,
+      "step": 18810
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.1e-05,
+      "loss": 1.5782,
+      "step": 18820
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.0994949494949496e-05,
+      "loss": 1.5698,
+      "step": 18830
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.098989898989899e-05,
+      "loss": 1.5853,
+      "step": 18840
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.098484848484849e-05,
+      "loss": 1.562,
+      "step": 18850
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.097979797979798e-05,
+      "loss": 1.582,
+      "step": 18860
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.0974747474747474e-05,
+      "loss": 1.5671,
+      "step": 18870
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.096969696969697e-05,
+      "loss": 1.5762,
+      "step": 18880
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.0964646464646466e-05,
+      "loss": 1.5556,
+      "step": 18890
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.095959595959596e-05,
+      "loss": 1.5663,
+      "step": 18900
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.095454545454546e-05,
+      "loss": 1.5911,
+      "step": 18910
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.094949494949495e-05,
+      "loss": 1.556,
+      "step": 18920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.094444444444445e-05,
+      "loss": 1.5464,
+      "step": 18930
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.0939393939393946e-05,
+      "loss": 1.563,
+      "step": 18940
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.0934343434343435e-05,
+      "loss": 1.569,
+      "step": 18950
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.092929292929293e-05,
+      "loss": 1.5764,
+      "step": 18960
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.092424242424243e-05,
+      "loss": 1.5878,
+      "step": 18970
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.091919191919192e-05,
+      "loss": 1.5795,
+      "step": 18980
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.091414141414142e-05,
+      "loss": 1.5451,
+      "step": 18990
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0909090909090915e-05,
+      "loss": 1.5704,
+      "step": 19000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0904040404040404e-05,
+      "loss": 1.5838,
+      "step": 19010
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.089898989898991e-05,
+      "loss": 1.5644,
+      "step": 19020
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0893939393939396e-05,
+      "loss": 1.5908,
+      "step": 19030
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.088888888888889e-05,
+      "loss": 1.5869,
+      "step": 19040
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.088383838383839e-05,
+      "loss": 1.5758,
+      "step": 19050
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0878787878787884e-05,
+      "loss": 1.5759,
+      "step": 19060
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.087373737373737e-05,
+      "loss": 1.5843,
+      "step": 19070
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0868686868686876e-05,
+      "loss": 1.5678,
+      "step": 19080
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0863636363636365e-05,
+      "loss": 1.5628,
+      "step": 19090
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.085858585858586e-05,
+      "loss": 1.5526,
+      "step": 19100
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.085353535353536e-05,
+      "loss": 1.5729,
+      "step": 19110
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.084848484848485e-05,
+      "loss": 1.5641,
+      "step": 19120
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.084343434343434e-05,
+      "loss": 1.5614,
+      "step": 19130
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0838383838383845e-05,
+      "loss": 1.603,
+      "step": 19140
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0833333333333334e-05,
+      "loss": 1.5693,
+      "step": 19150
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.082828282828283e-05,
+      "loss": 1.5894,
+      "step": 19160
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0823232323232326e-05,
+      "loss": 1.5601,
+      "step": 19170
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.081818181818182e-05,
+      "loss": 1.5463,
+      "step": 19180
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.081313131313131e-05,
+      "loss": 1.5719,
+      "step": 19190
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0808080808080814e-05,
+      "loss": 1.5636,
+      "step": 19200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0803030303030303e-05,
+      "loss": 1.574,
+      "step": 19210
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.07979797979798e-05,
+      "loss": 1.5568,
+      "step": 19220
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0792929292929295e-05,
+      "loss": 1.5408,
+      "step": 19230
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.078787878787879e-05,
+      "loss": 1.5664,
+      "step": 19240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.078282828282828e-05,
+      "loss": 1.5699,
+      "step": 19250
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0777777777777783e-05,
+      "loss": 1.5688,
+      "step": 19260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.077272727272727e-05,
+      "loss": 1.5973,
+      "step": 19270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.076767676767677e-05,
+      "loss": 1.5766,
+      "step": 19280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0762626262626265e-05,
+      "loss": 1.5442,
+      "step": 19290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.075757575757576e-05,
+      "loss": 1.5761,
+      "step": 19300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0752525252525257e-05,
+      "loss": 1.5715,
+      "step": 19310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.074747474747475e-05,
+      "loss": 1.5666,
+      "step": 19320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.074242424242424e-05,
+      "loss": 1.5867,
+      "step": 19330
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.073737373737374e-05,
+      "loss": 1.5742,
+      "step": 19340
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0732323232323234e-05,
+      "loss": 1.5749,
+      "step": 19350
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.072727272727273e-05,
+      "loss": 1.5953,
+      "step": 19360
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0722222222222226e-05,
+      "loss": 1.5942,
+      "step": 19370
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.071717171717172e-05,
+      "loss": 1.5669,
+      "step": 19380
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.071212121212121e-05,
+      "loss": 1.5838,
+      "step": 19390
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.070707070707071e-05,
+      "loss": 1.5945,
+      "step": 19400
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.07020202020202e-05,
+      "loss": 1.5692,
+      "step": 19410
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.06969696969697e-05,
+      "loss": 1.5682,
+      "step": 19420
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0691919191919195e-05,
+      "loss": 1.5685,
+      "step": 19430
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.068686868686869e-05,
+      "loss": 1.5567,
+      "step": 19440
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.068181818181818e-05,
+      "loss": 1.578,
+      "step": 19450
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.067676767676768e-05,
+      "loss": 1.5572,
+      "step": 19460
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.067171717171717e-05,
+      "loss": 1.5696,
+      "step": 19470
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.066666666666667e-05,
+      "loss": 1.5611,
+      "step": 19480
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0661616161616164e-05,
+      "loss": 1.577,
+      "step": 19490
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.065656565656566e-05,
+      "loss": 1.5395,
+      "step": 19500
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.065151515151515e-05,
+      "loss": 1.5982,
+      "step": 19510
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.064646464646465e-05,
+      "loss": 1.563,
+      "step": 19520
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.064141414141414e-05,
+      "loss": 1.5498,
+      "step": 19530
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.063636363636364e-05,
+      "loss": 1.5988,
+      "step": 19540
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.063131313131313e-05,
+      "loss": 1.5804,
+      "step": 19550
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.062626262626263e-05,
+      "loss": 1.5701,
+      "step": 19560
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.062121212121212e-05,
+      "loss": 1.5778,
+      "step": 19570
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.061616161616162e-05,
+      "loss": 1.5655,
+      "step": 19580
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.061111111111111e-05,
+      "loss": 1.5896,
+      "step": 19590
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.0606060606060606e-05,
+      "loss": 1.5757,
+      "step": 19600
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.06010101010101e-05,
+      "loss": 1.5736,
+      "step": 19610
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.05959595959596e-05,
+      "loss": 1.5719,
+      "step": 19620
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.059090909090909e-05,
+      "loss": 1.576,
+      "step": 19630
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.058585858585859e-05,
+      "loss": 1.563,
+      "step": 19640
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.058080808080808e-05,
+      "loss": 1.5844,
+      "step": 19650
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0575757575757576e-05,
+      "loss": 1.5593,
+      "step": 19660
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.057070707070707e-05,
+      "loss": 1.5646,
+      "step": 19670
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.056565656565657e-05,
+      "loss": 1.5808,
+      "step": 19680
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.056060606060606e-05,
+      "loss": 1.5284,
+      "step": 19690
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.055555555555556e-05,
+      "loss": 1.581,
+      "step": 19700
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.055050505050505e-05,
+      "loss": 1.5745,
+      "step": 19710
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0545454545454545e-05,
+      "loss": 1.5613,
+      "step": 19720
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.054040404040404e-05,
+      "loss": 1.5644,
+      "step": 19730
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.053535353535354e-05,
+      "loss": 1.5785,
+      "step": 19740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.053030303030303e-05,
+      "loss": 1.5562,
+      "step": 19750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.052525252525253e-05,
+      "loss": 1.5643,
+      "step": 19760
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.052020202020202e-05,
+      "loss": 1.5682,
+      "step": 19770
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0515151515151514e-05,
+      "loss": 1.578,
+      "step": 19780
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.051010101010102e-05,
+      "loss": 1.5564,
+      "step": 19790
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0505050505050506e-05,
+      "loss": 1.5606,
+      "step": 19800
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.05e-05,
+      "loss": 1.556,
+      "step": 19810
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.04949494949495e-05,
+      "loss": 1.5654,
+      "step": 19820
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0489898989898994e-05,
+      "loss": 1.5617,
+      "step": 19830
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.048484848484849e-05,
+      "loss": 1.5533,
+      "step": 19840
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0479797979797986e-05,
+      "loss": 1.5839,
+      "step": 19850
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0474747474747475e-05,
+      "loss": 1.5885,
+      "step": 19860
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.046969696969697e-05,
+      "loss": 1.59,
+      "step": 19870
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.046464646464647e-05,
+      "loss": 1.5596,
+      "step": 19880
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.045959595959596e-05,
+      "loss": 1.5688,
+      "step": 19890
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.045454545454546e-05,
+      "loss": 1.5704,
+      "step": 19900
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0449494949494955e-05,
+      "loss": 1.5608,
+      "step": 19910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0444444444444444e-05,
+      "loss": 1.5747,
+      "step": 19920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.043939393939395e-05,
+      "loss": 1.5805,
+      "step": 19930
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0434343434343436e-05,
+      "loss": 1.5965,
+      "step": 19940
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.042929292929293e-05,
+      "loss": 1.5595,
+      "step": 19950
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.042424242424243e-05,
+      "loss": 1.5846,
+      "step": 19960
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0419191919191924e-05,
+      "loss": 1.5636,
+      "step": 19970
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.041414141414141e-05,
+      "loss": 1.5752,
+      "step": 19980
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0409090909090916e-05,
+      "loss": 1.556,
+      "step": 19990
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0404040404040405e-05,
+      "loss": 1.5802,
+      "step": 20000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.03989898989899e-05,
+      "loss": 1.5632,
+      "step": 20010
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.03939393939394e-05,
+      "loss": 1.5809,
+      "step": 20020
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.038888888888889e-05,
+      "loss": 1.5689,
+      "step": 20030
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.038383838383838e-05,
+      "loss": 1.5687,
+      "step": 20040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0378787878787885e-05,
+      "loss": 1.5733,
+      "step": 20050
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0373737373737375e-05,
+      "loss": 1.5644,
+      "step": 20060
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.036868686868687e-05,
+      "loss": 1.5745,
+      "step": 20070
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0363636363636367e-05,
+      "loss": 1.5633,
+      "step": 20080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.035858585858586e-05,
+      "loss": 1.5673,
+      "step": 20090
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.035353535353535e-05,
+      "loss": 1.5857,
+      "step": 20100
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0348484848484854e-05,
+      "loss": 1.5809,
+      "step": 20110
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0343434343434344e-05,
+      "loss": 1.5851,
+      "step": 20120
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.033838383838384e-05,
+      "loss": 1.578,
+      "step": 20130
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0333333333333336e-05,
+      "loss": 1.5979,
+      "step": 20140
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.032828282828283e-05,
+      "loss": 1.5575,
+      "step": 20150
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.032323232323232e-05,
+      "loss": 1.5884,
+      "step": 20160
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0318181818181824e-05,
+      "loss": 1.5473,
+      "step": 20170
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.031313131313131e-05,
+      "loss": 1.578,
+      "step": 20180
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.030808080808081e-05,
+      "loss": 1.571,
+      "step": 20190
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0303030303030305e-05,
+      "loss": 1.5608,
+      "step": 20200
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.02979797979798e-05,
+      "loss": 1.542,
+      "step": 20210
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.029292929292929e-05,
+      "loss": 1.5899,
+      "step": 20220
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.028787878787879e-05,
+      "loss": 1.5661,
+      "step": 20230
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.028282828282828e-05,
+      "loss": 1.5571,
+      "step": 20240
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.027777777777778e-05,
+      "loss": 1.5752,
+      "step": 20250
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0272727272727274e-05,
+      "loss": 1.5669,
+      "step": 20260
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.026767676767677e-05,
+      "loss": 1.5739,
+      "step": 20270
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.0262626262626266e-05,
+      "loss": 1.5724,
+      "step": 20280
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.025757575757576e-05,
+      "loss": 1.5604,
+      "step": 20290
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.025252525252525e-05,
+      "loss": 1.5789,
+      "step": 20300
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.024747474747475e-05,
+      "loss": 1.5817,
+      "step": 20310
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.024242424242424e-05,
+      "loss": 1.5802,
+      "step": 20320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.023737373737374e-05,
+      "loss": 1.569,
+      "step": 20330
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0232323232323235e-05,
+      "loss": 1.5946,
+      "step": 20340
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.022727272727273e-05,
+      "loss": 1.5872,
+      "step": 20350
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.022222222222222e-05,
+      "loss": 1.5572,
+      "step": 20360
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.021717171717172e-05,
+      "loss": 1.5573,
+      "step": 20370
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.021212121212121e-05,
+      "loss": 1.5752,
+      "step": 20380
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.020707070707071e-05,
+      "loss": 1.5549,
+      "step": 20390
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0202020202020204e-05,
+      "loss": 1.5685,
+      "step": 20400
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.01969696969697e-05,
+      "loss": 1.5539,
+      "step": 20410
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.019191919191919e-05,
+      "loss": 1.5717,
+      "step": 20420
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.018686868686869e-05,
+      "loss": 1.5706,
+      "step": 20430
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.018181818181818e-05,
+      "loss": 1.5737,
+      "step": 20440
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.017676767676768e-05,
+      "loss": 1.5686,
+      "step": 20450
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0171717171717173e-05,
+      "loss": 1.5656,
+      "step": 20460
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.016666666666667e-05,
+      "loss": 1.5867,
+      "step": 20470
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.016161616161616e-05,
+      "loss": 1.559,
+      "step": 20480
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.015656565656566e-05,
+      "loss": 1.5676,
+      "step": 20490
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.015151515151515e-05,
+      "loss": 1.5646,
+      "step": 20500
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.014646464646465e-05,
+      "loss": 1.5632,
+      "step": 20510
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.014141414141414e-05,
+      "loss": 1.5739,
+      "step": 20520
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.013636363636364e-05,
+      "loss": 1.5725,
+      "step": 20530
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.013131313131313e-05,
+      "loss": 1.5678,
+      "step": 20540
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.012626262626263e-05,
+      "loss": 1.5766,
+      "step": 20550
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.012121212121212e-05,
+      "loss": 1.5344,
+      "step": 20560
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0116161616161616e-05,
+      "loss": 1.5797,
+      "step": 20570
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.011111111111111e-05,
+      "loss": 1.5706,
+      "step": 20580
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.010606060606061e-05,
+      "loss": 1.5396,
+      "step": 20590
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.01010101010101e-05,
+      "loss": 1.6032,
+      "step": 20600
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.00959595959596e-05,
+      "loss": 1.5736,
+      "step": 20610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.009090909090909e-05,
+      "loss": 1.551,
+      "step": 20620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0085858585858585e-05,
+      "loss": 1.578,
+      "step": 20630
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.008080808080809e-05,
+      "loss": 1.5586,
+      "step": 20640
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.007575757575758e-05,
+      "loss": 1.5625,
+      "step": 20650
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.007070707070707e-05,
+      "loss": 1.5737,
+      "step": 20660
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.006565656565657e-05,
+      "loss": 1.5494,
+      "step": 20670
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0060606060606065e-05,
+      "loss": 1.5738,
+      "step": 20680
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0055555555555554e-05,
+      "loss": 1.5921,
+      "step": 20690
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.005050505050506e-05,
+      "loss": 1.5554,
+      "step": 20700
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0045454545454546e-05,
+      "loss": 1.5753,
+      "step": 20710
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.004040404040404e-05,
+      "loss": 1.5574,
+      "step": 20720
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.003535353535354e-05,
+      "loss": 1.5702,
+      "step": 20730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0030303030303034e-05,
+      "loss": 1.6013,
+      "step": 20740
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.002525252525253e-05,
+      "loss": 1.5569,
+      "step": 20750
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0020202020202026e-05,
+      "loss": 1.5729,
+      "step": 20760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.0015151515151515e-05,
+      "loss": 1.5691,
+      "step": 20770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.001010101010101e-05,
+      "loss": 1.5579,
+      "step": 20780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.000505050505051e-05,
+      "loss": 1.5685,
+      "step": 20790
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4e-05,
+      "loss": 1.604,
+      "step": 20800
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.99949494949495e-05,
+      "loss": 1.5822,
+      "step": 20810
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.9989898989898995e-05,
+      "loss": 1.5563,
+      "step": 20820
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.9984848484848484e-05,
+      "loss": 1.5773,
+      "step": 20830
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.997979797979798e-05,
+      "loss": 1.5705,
+      "step": 20840
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.9974747474747476e-05,
+      "loss": 1.566,
+      "step": 20850
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.996969696969697e-05,
+      "loss": 1.5918,
+      "step": 20860
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.996464646464647e-05,
+      "loss": 1.5892,
+      "step": 20870
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.9959595959595964e-05,
+      "loss": 1.568,
+      "step": 20880
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.9954545454545454e-05,
+      "loss": 1.5602,
+      "step": 20890
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.9949494949494956e-05,
+      "loss": 1.5721,
+      "step": 20900
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.9944444444444446e-05,
+      "loss": 1.5608,
+      "step": 20910
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.993939393939394e-05,
+      "loss": 1.5651,
+      "step": 20920
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.993434343434344e-05,
+      "loss": 1.5688,
+      "step": 20930
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.9929292929292934e-05,
+      "loss": 1.5644,
+      "step": 20940
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 3.992424242424242e-05,
+      "loss": 1.5887,
+      "step": 20950
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9919191919191926e-05,
+      "loss": 1.5565,
+      "step": 20960
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9914141414141415e-05,
+      "loss": 1.5645,
+      "step": 20970
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.990909090909091e-05,
+      "loss": 1.5833,
+      "step": 20980
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.990404040404041e-05,
+      "loss": 1.5418,
+      "step": 20990
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.98989898989899e-05,
+      "loss": 1.5632,
+      "step": 21000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.989393939393939e-05,
+      "loss": 1.5443,
+      "step": 21010
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9888888888888895e-05,
+      "loss": 1.572,
+      "step": 21020
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9883838383838384e-05,
+      "loss": 1.5889,
+      "step": 21030
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.987878787878788e-05,
+      "loss": 1.5675,
+      "step": 21040
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9873737373737376e-05,
+      "loss": 1.5623,
+      "step": 21050
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.986868686868687e-05,
+      "loss": 1.5639,
+      "step": 21060
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.986363636363636e-05,
+      "loss": 1.5464,
+      "step": 21070
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9858585858585864e-05,
+      "loss": 1.5754,
+      "step": 21080
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.985353535353535e-05,
+      "loss": 1.5692,
+      "step": 21090
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.984848484848485e-05,
+      "loss": 1.5584,
+      "step": 21100
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9843434343434345e-05,
+      "loss": 1.5943,
+      "step": 21110
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.983838383838384e-05,
+      "loss": 1.5776,
+      "step": 21120
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.983333333333333e-05,
+      "loss": 1.5778,
+      "step": 21130
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.982828282828283e-05,
+      "loss": 1.5486,
+      "step": 21140
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.982323232323232e-05,
+      "loss": 1.5607,
+      "step": 21150
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.981818181818182e-05,
+      "loss": 1.5787,
+      "step": 21160
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9813131313131314e-05,
+      "loss": 1.5812,
+      "step": 21170
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.980808080808081e-05,
+      "loss": 1.5643,
+      "step": 21180
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9803030303030306e-05,
+      "loss": 1.5707,
+      "step": 21190
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.97979797979798e-05,
+      "loss": 1.564,
+      "step": 21200
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.979292929292929e-05,
+      "loss": 1.5831,
+      "step": 21210
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.978787878787879e-05,
+      "loss": 1.5743,
+      "step": 21220
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9782828282828283e-05,
+      "loss": 1.5576,
+      "step": 21230
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.977777777777778e-05,
+      "loss": 1.5774,
+      "step": 21240
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9772727272727275e-05,
+      "loss": 1.5579,
+      "step": 21250
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.976767676767677e-05,
+      "loss": 1.5809,
+      "step": 21260
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.976262626262626e-05,
+      "loss": 1.5585,
+      "step": 21270
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.975757575757576e-05,
+      "loss": 1.5839,
+      "step": 21280
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.975252525252525e-05,
+      "loss": 1.5761,
+      "step": 21290
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.974747474747475e-05,
+      "loss": 1.5797,
+      "step": 21300
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9742424242424245e-05,
+      "loss": 1.547,
+      "step": 21310
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.973737373737374e-05,
+      "loss": 1.5491,
+      "step": 21320
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.973232323232323e-05,
+      "loss": 1.5646,
+      "step": 21330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.972727272727273e-05,
+      "loss": 1.5739,
+      "step": 21340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.972222222222222e-05,
+      "loss": 1.5841,
+      "step": 21350
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.971717171717172e-05,
+      "loss": 1.5931,
+      "step": 21360
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9712121212121214e-05,
+      "loss": 1.5626,
+      "step": 21370
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.970707070707071e-05,
+      "loss": 1.5565,
+      "step": 21380
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.97020202020202e-05,
+      "loss": 1.5777,
+      "step": 21390
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.96969696969697e-05,
+      "loss": 1.5623,
+      "step": 21400
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.969191919191919e-05,
+      "loss": 1.573,
+      "step": 21410
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.968686868686869e-05,
+      "loss": 1.5851,
+      "step": 21420
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.968181818181818e-05,
+      "loss": 1.5758,
+      "step": 21430
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.967676767676768e-05,
+      "loss": 1.5657,
+      "step": 21440
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.967171717171717e-05,
+      "loss": 1.5772,
+      "step": 21450
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.966666666666667e-05,
+      "loss": 1.5609,
+      "step": 21460
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.966161616161616e-05,
+      "loss": 1.5455,
+      "step": 21470
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9656565656565656e-05,
+      "loss": 1.5947,
+      "step": 21480
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.965151515151516e-05,
+      "loss": 1.5642,
+      "step": 21490
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.964646464646465e-05,
+      "loss": 1.5387,
+      "step": 21500
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9641414141414144e-05,
+      "loss": 1.5755,
+      "step": 21510
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.963636363636364e-05,
+      "loss": 1.5721,
+      "step": 21520
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9631313131313136e-05,
+      "loss": 1.5574,
+      "step": 21530
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9626262626262625e-05,
+      "loss": 1.5584,
+      "step": 21540
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.962121212121213e-05,
+      "loss": 1.5567,
+      "step": 21550
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.961616161616162e-05,
+      "loss": 1.5598,
+      "step": 21560
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.961111111111111e-05,
+      "loss": 1.5721,
+      "step": 21570
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.960606060606061e-05,
+      "loss": 1.5922,
+      "step": 21580
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9601010101010105e-05,
+      "loss": 1.5822,
+      "step": 21590
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.9595959595959594e-05,
+      "loss": 1.5525,
+      "step": 21600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 3.95909090909091e-05,
+      "loss": 1.5749,
+      "step": 21610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9585858585858586e-05,
+      "loss": 1.5991,
+      "step": 21620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.958080808080808e-05,
+      "loss": 1.5574,
+      "step": 21630
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.957575757575758e-05,
+      "loss": 1.5763,
+      "step": 21640
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9570707070707074e-05,
+      "loss": 1.5615,
+      "step": 21650
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.956565656565657e-05,
+      "loss": 1.5683,
+      "step": 21660
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9560606060606066e-05,
+      "loss": 1.5672,
+      "step": 21670
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9555555555555556e-05,
+      "loss": 1.5937,
+      "step": 21680
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.955050505050505e-05,
+      "loss": 1.543,
+      "step": 21690
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.954545454545455e-05,
+      "loss": 1.5692,
+      "step": 21700
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9540404040404044e-05,
+      "loss": 1.5453,
+      "step": 21710
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.953535353535354e-05,
+      "loss": 1.5769,
+      "step": 21720
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9530303030303036e-05,
+      "loss": 1.5609,
+      "step": 21730
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9525252525252525e-05,
+      "loss": 1.5564,
+      "step": 21740
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.952020202020202e-05,
+      "loss": 1.5667,
+      "step": 21750
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.951515151515152e-05,
+      "loss": 1.5721,
+      "step": 21760
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.951010101010101e-05,
+      "loss": 1.5696,
+      "step": 21770
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.950505050505051e-05,
+      "loss": 1.5659,
+      "step": 21780
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9500000000000005e-05,
+      "loss": 1.5605,
+      "step": 21790
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9494949494949494e-05,
+      "loss": 1.578,
+      "step": 21800
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9489898989899e-05,
+      "loss": 1.5578,
+      "step": 21810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9484848484848486e-05,
+      "loss": 1.5597,
+      "step": 21820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.947979797979798e-05,
+      "loss": 1.5567,
+      "step": 21830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.947474747474748e-05,
+      "loss": 1.5834,
+      "step": 21840
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9469696969696974e-05,
+      "loss": 1.5568,
+      "step": 21850
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.946464646464646e-05,
+      "loss": 1.5581,
+      "step": 21860
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9459595959595966e-05,
+      "loss": 1.5445,
+      "step": 21870
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9454545454545455e-05,
+      "loss": 1.5575,
+      "step": 21880
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.944949494949495e-05,
+      "loss": 1.5739,
+      "step": 21890
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.944444444444445e-05,
+      "loss": 1.5597,
+      "step": 21900
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.943939393939394e-05,
+      "loss": 1.559,
+      "step": 21910
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.943434343434343e-05,
+      "loss": 1.5623,
+      "step": 21920
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9429292929292935e-05,
+      "loss": 1.5793,
+      "step": 21930
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9424242424242424e-05,
+      "loss": 1.5626,
+      "step": 21940
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.941919191919192e-05,
+      "loss": 1.5668,
+      "step": 21950
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9414141414141416e-05,
+      "loss": 1.5643,
+      "step": 21960
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.940909090909091e-05,
+      "loss": 1.5706,
+      "step": 21970
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.94040404040404e-05,
+      "loss": 1.5737,
+      "step": 21980
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9398989898989904e-05,
+      "loss": 1.5492,
+      "step": 21990
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.939393939393939e-05,
+      "loss": 1.5756,
+      "step": 22000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.938888888888889e-05,
+      "loss": 1.5553,
+      "step": 22010
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9383838383838385e-05,
+      "loss": 1.5864,
+      "step": 22020
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.937878787878788e-05,
+      "loss": 1.5537,
+      "step": 22030
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.937373737373737e-05,
+      "loss": 1.5615,
+      "step": 22040
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.936868686868687e-05,
+      "loss": 1.5545,
+      "step": 22050
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.936363636363636e-05,
+      "loss": 1.575,
+      "step": 22060
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.935858585858586e-05,
+      "loss": 1.5814,
+      "step": 22070
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9353535353535355e-05,
+      "loss": 1.5626,
+      "step": 22080
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.934848484848485e-05,
+      "loss": 1.5617,
+      "step": 22090
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9343434343434346e-05,
+      "loss": 1.5581,
+      "step": 22100
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.933838383838384e-05,
+      "loss": 1.5585,
+      "step": 22110
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.933333333333333e-05,
+      "loss": 1.5716,
+      "step": 22120
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.932828282828283e-05,
+      "loss": 1.5747,
+      "step": 22130
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9323232323232324e-05,
+      "loss": 1.5664,
+      "step": 22140
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.931818181818182e-05,
+      "loss": 1.5663,
+      "step": 22150
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9313131313131316e-05,
+      "loss": 1.5522,
+      "step": 22160
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.930808080808081e-05,
+      "loss": 1.5661,
+      "step": 22170
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.93030303030303e-05,
+      "loss": 1.5886,
+      "step": 22180
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9297979797979804e-05,
+      "loss": 1.5871,
+      "step": 22190
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.929292929292929e-05,
+      "loss": 1.56,
+      "step": 22200
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.928787878787879e-05,
+      "loss": 1.5703,
+      "step": 22210
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9282828282828285e-05,
+      "loss": 1.5735,
+      "step": 22220
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.927777777777778e-05,
+      "loss": 1.5496,
+      "step": 22230
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.927272727272727e-05,
+      "loss": 1.5894,
+      "step": 22240
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.926767676767677e-05,
+      "loss": 1.5619,
+      "step": 22250
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.926262626262626e-05,
+      "loss": 1.5581,
+      "step": 22260
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.925757575757576e-05,
+      "loss": 1.5562,
+      "step": 22270
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 3.9252525252525254e-05,
+      "loss": 1.5565,
+      "step": 22280
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.924747474747475e-05,
+      "loss": 1.5706,
+      "step": 22290
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.924242424242424e-05,
+      "loss": 1.5666,
+      "step": 22300
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.923737373737374e-05,
+      "loss": 1.5725,
+      "step": 22310
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.923232323232323e-05,
+      "loss": 1.5682,
+      "step": 22320
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.922727272727273e-05,
+      "loss": 1.5689,
+      "step": 22330
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.922222222222223e-05,
+      "loss": 1.5576,
+      "step": 22340
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.921717171717172e-05,
+      "loss": 1.5715,
+      "step": 22350
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9212121212121215e-05,
+      "loss": 1.5547,
+      "step": 22360
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.920707070707071e-05,
+      "loss": 1.5628,
+      "step": 22370
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.920202020202021e-05,
+      "loss": 1.582,
+      "step": 22380
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9196969696969696e-05,
+      "loss": 1.5716,
+      "step": 22390
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.91919191919192e-05,
+      "loss": 1.568,
+      "step": 22400
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.918686868686869e-05,
+      "loss": 1.5556,
+      "step": 22410
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9181818181818184e-05,
+      "loss": 1.5493,
+      "step": 22420
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.917676767676768e-05,
+      "loss": 1.5686,
+      "step": 22430
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9171717171717176e-05,
+      "loss": 1.5597,
+      "step": 22440
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9166666666666665e-05,
+      "loss": 1.5687,
+      "step": 22450
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.916161616161617e-05,
+      "loss": 1.561,
+      "step": 22460
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.915656565656566e-05,
+      "loss": 1.5597,
+      "step": 22470
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9151515151515153e-05,
+      "loss": 1.5492,
+      "step": 22480
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.914646464646465e-05,
+      "loss": 1.5771,
+      "step": 22490
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9141414141414145e-05,
+      "loss": 1.5843,
+      "step": 22500
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9136363636363635e-05,
+      "loss": 1.5621,
+      "step": 22510
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.913131313131314e-05,
+      "loss": 1.5562,
+      "step": 22520
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.912626262626263e-05,
+      "loss": 1.5593,
+      "step": 22530
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.912121212121212e-05,
+      "loss": 1.5595,
+      "step": 22540
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.911616161616162e-05,
+      "loss": 1.5745,
+      "step": 22550
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9111111111111115e-05,
+      "loss": 1.5607,
+      "step": 22560
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9106060606060604e-05,
+      "loss": 1.571,
+      "step": 22570
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9101010101010107e-05,
+      "loss": 1.5717,
+      "step": 22580
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9095959595959596e-05,
+      "loss": 1.5456,
+      "step": 22590
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.909090909090909e-05,
+      "loss": 1.5675,
+      "step": 22600
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.908585858585859e-05,
+      "loss": 1.5575,
+      "step": 22610
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9080808080808084e-05,
+      "loss": 1.5568,
+      "step": 22620
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.907575757575758e-05,
+      "loss": 1.5695,
+      "step": 22630
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9070707070707076e-05,
+      "loss": 1.5641,
+      "step": 22640
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9065656565656565e-05,
+      "loss": 1.5656,
+      "step": 22650
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.906060606060606e-05,
+      "loss": 1.5679,
+      "step": 22660
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.905555555555556e-05,
+      "loss": 1.5534,
+      "step": 22670
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.905050505050505e-05,
+      "loss": 1.5482,
+      "step": 22680
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.904545454545455e-05,
+      "loss": 1.5643,
+      "step": 22690
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9040404040404045e-05,
+      "loss": 1.5698,
+      "step": 22700
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9035353535353534e-05,
+      "loss": 1.5607,
+      "step": 22710
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.903030303030304e-05,
+      "loss": 1.5579,
+      "step": 22720
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9025252525252526e-05,
+      "loss": 1.5761,
+      "step": 22730
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.902020202020202e-05,
+      "loss": 1.5736,
+      "step": 22740
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.901515151515152e-05,
+      "loss": 1.5583,
+      "step": 22750
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9010101010101014e-05,
+      "loss": 1.5774,
+      "step": 22760
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.90050505050505e-05,
+      "loss": 1.5704,
+      "step": 22770
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.9000000000000006e-05,
+      "loss": 1.5515,
+      "step": 22780
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8994949494949495e-05,
+      "loss": 1.575,
+      "step": 22790
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.898989898989899e-05,
+      "loss": 1.5598,
+      "step": 22800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.898484848484849e-05,
+      "loss": 1.5536,
+      "step": 22810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.897979797979798e-05,
+      "loss": 1.5571,
+      "step": 22820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.897474747474747e-05,
+      "loss": 1.5378,
+      "step": 22830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8969696969696975e-05,
+      "loss": 1.5679,
+      "step": 22840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8964646464646464e-05,
+      "loss": 1.5781,
+      "step": 22850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.895959595959596e-05,
+      "loss": 1.5635,
+      "step": 22860
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8954545454545456e-05,
+      "loss": 1.5896,
+      "step": 22870
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.894949494949495e-05,
+      "loss": 1.5551,
+      "step": 22880
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.894444444444444e-05,
+      "loss": 1.5435,
+      "step": 22890
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8939393939393944e-05,
+      "loss": 1.5725,
+      "step": 22900
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8934343434343434e-05,
+      "loss": 1.5847,
+      "step": 22910
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.892929292929293e-05,
+      "loss": 1.5489,
+      "step": 22920
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.8924242424242426e-05,
+      "loss": 1.5712,
+      "step": 22930
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.891919191919192e-05,
+      "loss": 1.5766,
+      "step": 22940
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 3.891414141414141e-05,
+      "loss": 1.5722,
+      "step": 22950
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8909090909090914e-05,
+      "loss": 1.5584,
+      "step": 22960
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.89040404040404e-05,
+      "loss": 1.5377,
+      "step": 22970
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.88989898989899e-05,
+      "loss": 1.5761,
+      "step": 22980
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8893939393939395e-05,
+      "loss": 1.5474,
+      "step": 22990
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.888888888888889e-05,
+      "loss": 1.5915,
+      "step": 23000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.888383838383839e-05,
+      "loss": 1.5655,
+      "step": 23010
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.887878787878788e-05,
+      "loss": 1.5748,
+      "step": 23020
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.887373737373737e-05,
+      "loss": 1.5481,
+      "step": 23030
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.886868686868687e-05,
+      "loss": 1.5641,
+      "step": 23040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8863636363636364e-05,
+      "loss": 1.5562,
+      "step": 23050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.885858585858586e-05,
+      "loss": 1.5596,
+      "step": 23060
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8853535353535356e-05,
+      "loss": 1.5456,
+      "step": 23070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.884848484848485e-05,
+      "loss": 1.5525,
+      "step": 23080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.884343434343434e-05,
+      "loss": 1.5772,
+      "step": 23090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8838383838383844e-05,
+      "loss": 1.5729,
+      "step": 23100
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.883333333333333e-05,
+      "loss": 1.5661,
+      "step": 23110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.882828282828283e-05,
+      "loss": 1.5544,
+      "step": 23120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8823232323232325e-05,
+      "loss": 1.578,
+      "step": 23130
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.881818181818182e-05,
+      "loss": 1.5568,
+      "step": 23140
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.881313131313131e-05,
+      "loss": 1.5725,
+      "step": 23150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.880808080808081e-05,
+      "loss": 1.5639,
+      "step": 23160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.88030303030303e-05,
+      "loss": 1.5415,
+      "step": 23170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.87979797979798e-05,
+      "loss": 1.5595,
+      "step": 23180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8792929292929294e-05,
+      "loss": 1.5681,
+      "step": 23190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.878787878787879e-05,
+      "loss": 1.5683,
+      "step": 23200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8782828282828286e-05,
+      "loss": 1.5717,
+      "step": 23210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.877777777777778e-05,
+      "loss": 1.5658,
+      "step": 23220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.877272727272728e-05,
+      "loss": 1.5734,
+      "step": 23230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.876767676767677e-05,
+      "loss": 1.5656,
+      "step": 23240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.876262626262627e-05,
+      "loss": 1.5429,
+      "step": 23250
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.875757575757576e-05,
+      "loss": 1.5759,
+      "step": 23260
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8752525252525255e-05,
+      "loss": 1.5838,
+      "step": 23270
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.874747474747475e-05,
+      "loss": 1.5517,
+      "step": 23280
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.874242424242425e-05,
+      "loss": 1.5618,
+      "step": 23290
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8737373737373737e-05,
+      "loss": 1.5478,
+      "step": 23300
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.873232323232324e-05,
+      "loss": 1.5276,
+      "step": 23310
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.872727272727273e-05,
+      "loss": 1.5805,
+      "step": 23320
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8722222222222225e-05,
+      "loss": 1.5747,
+      "step": 23330
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.871717171717172e-05,
+      "loss": 1.5666,
+      "step": 23340
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8712121212121217e-05,
+      "loss": 1.5657,
+      "step": 23350
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8707070707070706e-05,
+      "loss": 1.5711,
+      "step": 23360
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.870202020202021e-05,
+      "loss": 1.5873,
+      "step": 23370
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.86969696969697e-05,
+      "loss": 1.5489,
+      "step": 23380
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8691919191919194e-05,
+      "loss": 1.5811,
+      "step": 23390
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.868686868686869e-05,
+      "loss": 1.5604,
+      "step": 23400
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8681818181818186e-05,
+      "loss": 1.5535,
+      "step": 23410
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8676767676767675e-05,
+      "loss": 1.5535,
+      "step": 23420
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.867171717171718e-05,
+      "loss": 1.5819,
+      "step": 23430
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.866666666666667e-05,
+      "loss": 1.558,
+      "step": 23440
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.866161616161616e-05,
+      "loss": 1.5625,
+      "step": 23450
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.865656565656566e-05,
+      "loss": 1.5543,
+      "step": 23460
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8651515151515155e-05,
+      "loss": 1.5615,
+      "step": 23470
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8646464646464644e-05,
+      "loss": 1.5473,
+      "step": 23480
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.864141414141415e-05,
+      "loss": 1.5656,
+      "step": 23490
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8636363636363636e-05,
+      "loss": 1.5608,
+      "step": 23500
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.863131313131313e-05,
+      "loss": 1.5669,
+      "step": 23510
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.862626262626263e-05,
+      "loss": 1.5611,
+      "step": 23520
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8621212121212124e-05,
+      "loss": 1.5696,
+      "step": 23530
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.861616161616162e-05,
+      "loss": 1.5585,
+      "step": 23540
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8611111111111116e-05,
+      "loss": 1.5534,
+      "step": 23550
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8606060606060605e-05,
+      "loss": 1.563,
+      "step": 23560
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.86010101010101e-05,
+      "loss": 1.5444,
+      "step": 23570
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.85959595959596e-05,
+      "loss": 1.5747,
+      "step": 23580
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.859090909090909e-05,
+      "loss": 1.5734,
+      "step": 23590
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.858585858585859e-05,
+      "loss": 1.5581,
+      "step": 23600
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 3.8580808080808085e-05,
+      "loss": 1.5636,
+      "step": 23610
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8575757575757574e-05,
+      "loss": 1.567,
+      "step": 23620
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.857070707070708e-05,
+      "loss": 1.5547,
+      "step": 23630
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8565656565656566e-05,
+      "loss": 1.5877,
+      "step": 23640
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.856060606060606e-05,
+      "loss": 1.5686,
+      "step": 23650
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.855555555555556e-05,
+      "loss": 1.587,
+      "step": 23660
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8550505050505054e-05,
+      "loss": 1.5868,
+      "step": 23670
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8545454545454544e-05,
+      "loss": 1.5587,
+      "step": 23680
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8540404040404046e-05,
+      "loss": 1.5771,
+      "step": 23690
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8535353535353536e-05,
+      "loss": 1.5643,
+      "step": 23700
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.853030303030303e-05,
+      "loss": 1.5728,
+      "step": 23710
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.852525252525253e-05,
+      "loss": 1.5656,
+      "step": 23720
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8520202020202024e-05,
+      "loss": 1.5815,
+      "step": 23730
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.851515151515151e-05,
+      "loss": 1.5721,
+      "step": 23740
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8510101010101015e-05,
+      "loss": 1.5499,
+      "step": 23750
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8505050505050505e-05,
+      "loss": 1.5803,
+      "step": 23760
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.85e-05,
+      "loss": 1.5674,
+      "step": 23770
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.84949494949495e-05,
+      "loss": 1.5537,
+      "step": 23780
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.848989898989899e-05,
+      "loss": 1.5605,
+      "step": 23790
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.848484848484848e-05,
+      "loss": 1.5605,
+      "step": 23800
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8479797979797985e-05,
+      "loss": 1.5756,
+      "step": 23810
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8474747474747474e-05,
+      "loss": 1.5729,
+      "step": 23820
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.846969696969697e-05,
+      "loss": 1.5629,
+      "step": 23830
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8464646464646466e-05,
+      "loss": 1.5781,
+      "step": 23840
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.845959595959596e-05,
+      "loss": 1.5808,
+      "step": 23850
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.845454545454545e-05,
+      "loss": 1.5777,
+      "step": 23860
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8449494949494954e-05,
+      "loss": 1.543,
+      "step": 23870
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.844444444444444e-05,
+      "loss": 1.5831,
+      "step": 23880
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.843939393939394e-05,
+      "loss": 1.5426,
+      "step": 23890
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8434343434343435e-05,
+      "loss": 1.5596,
+      "step": 23900
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.842929292929293e-05,
+      "loss": 1.6128,
+      "step": 23910
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.842424242424243e-05,
+      "loss": 1.5889,
+      "step": 23920
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.841919191919192e-05,
+      "loss": 1.5552,
+      "step": 23930
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.841414141414141e-05,
+      "loss": 1.5808,
+      "step": 23940
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.840909090909091e-05,
+      "loss": 1.5528,
+      "step": 23950
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8404040404040404e-05,
+      "loss": 1.5549,
+      "step": 23960
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.83989898989899e-05,
+      "loss": 1.5455,
+      "step": 23970
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8393939393939396e-05,
+      "loss": 1.5416,
+      "step": 23980
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.838888888888889e-05,
+      "loss": 1.5791,
+      "step": 23990
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.838383838383838e-05,
+      "loss": 1.5594,
+      "step": 24000
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.837878787878788e-05,
+      "loss": 1.5409,
+      "step": 24010
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.837373737373737e-05,
+      "loss": 1.5465,
+      "step": 24020
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.836868686868687e-05,
+      "loss": 1.5485,
+      "step": 24030
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8363636363636365e-05,
+      "loss": 1.5387,
+      "step": 24040
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.835858585858586e-05,
+      "loss": 1.5482,
+      "step": 24050
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.835353535353536e-05,
+      "loss": 1.5564,
+      "step": 24060
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.834848484848485e-05,
+      "loss": 1.5709,
+      "step": 24070
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.834343434343435e-05,
+      "loss": 1.5595,
+      "step": 24080
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.833838383838384e-05,
+      "loss": 1.5469,
+      "step": 24090
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8333333333333334e-05,
+      "loss": 1.5809,
+      "step": 24100
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.832828282828283e-05,
+      "loss": 1.5617,
+      "step": 24110
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8323232323232326e-05,
+      "loss": 1.5672,
+      "step": 24120
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.831818181818182e-05,
+      "loss": 1.5599,
+      "step": 24130
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.831313131313132e-05,
+      "loss": 1.5485,
+      "step": 24140
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.830808080808081e-05,
+      "loss": 1.5781,
+      "step": 24150
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.830303030303031e-05,
+      "loss": 1.5588,
+      "step": 24160
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.82979797979798e-05,
+      "loss": 1.559,
+      "step": 24170
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8292929292929296e-05,
+      "loss": 1.5508,
+      "step": 24180
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.828787878787879e-05,
+      "loss": 1.5533,
+      "step": 24190
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.828282828282829e-05,
+      "loss": 1.5769,
+      "step": 24200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.827777777777778e-05,
+      "loss": 1.5684,
+      "step": 24210
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.827272727272728e-05,
+      "loss": 1.5711,
+      "step": 24220
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.826767676767677e-05,
+      "loss": 1.5594,
+      "step": 24230
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8262626262626265e-05,
+      "loss": 1.5558,
+      "step": 24240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.825757575757576e-05,
+      "loss": 1.5494,
+      "step": 24250
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.825252525252526e-05,
+      "loss": 1.5687,
+      "step": 24260
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.8247474747474746e-05,
+      "loss": 1.5436,
+      "step": 24270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 3.824242424242425e-05,
+      "loss": 1.5571,
+      "step": 24280
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.823737373737374e-05,
+      "loss": 1.5523,
+      "step": 24290
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8232323232323234e-05,
+      "loss": 1.5509,
+      "step": 24300
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.822727272727273e-05,
+      "loss": 1.5767,
+      "step": 24310
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8222222222222226e-05,
+      "loss": 1.567,
+      "step": 24320
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8217171717171715e-05,
+      "loss": 1.5521,
+      "step": 24330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.821212121212122e-05,
+      "loss": 1.5423,
+      "step": 24340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.820707070707071e-05,
+      "loss": 1.5733,
+      "step": 24350
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.82020202020202e-05,
+      "loss": 1.5483,
+      "step": 24360
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.81969696969697e-05,
+      "loss": 1.5493,
+      "step": 24370
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8191919191919195e-05,
+      "loss": 1.5718,
+      "step": 24380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8186868686868684e-05,
+      "loss": 1.545,
+      "step": 24390
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.818181818181819e-05,
+      "loss": 1.5478,
+      "step": 24400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8176767676767676e-05,
+      "loss": 1.5683,
+      "step": 24410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.817171717171717e-05,
+      "loss": 1.5686,
+      "step": 24420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.816666666666667e-05,
+      "loss": 1.5536,
+      "step": 24430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8161616161616164e-05,
+      "loss": 1.5433,
+      "step": 24440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.815656565656566e-05,
+      "loss": 1.5605,
+      "step": 24450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8151515151515156e-05,
+      "loss": 1.5711,
+      "step": 24460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8146464646464645e-05,
+      "loss": 1.55,
+      "step": 24470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.814141414141414e-05,
+      "loss": 1.5606,
+      "step": 24480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.813636363636364e-05,
+      "loss": 1.5428,
+      "step": 24490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8131313131313133e-05,
+      "loss": 1.5493,
+      "step": 24500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.812626262626263e-05,
+      "loss": 1.5633,
+      "step": 24510
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8121212121212125e-05,
+      "loss": 1.5582,
+      "step": 24520
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8116161616161615e-05,
+      "loss": 1.5573,
+      "step": 24530
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.811111111111112e-05,
+      "loss": 1.5733,
+      "step": 24540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.810606060606061e-05,
+      "loss": 1.5311,
+      "step": 24550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.81010101010101e-05,
+      "loss": 1.5576,
+      "step": 24560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.80959595959596e-05,
+      "loss": 1.5684,
+      "step": 24570
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8090909090909095e-05,
+      "loss": 1.5564,
+      "step": 24580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8085858585858584e-05,
+      "loss": 1.5503,
+      "step": 24590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8080808080808087e-05,
+      "loss": 1.5601,
+      "step": 24600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8075757575757576e-05,
+      "loss": 1.5647,
+      "step": 24610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.807070707070707e-05,
+      "loss": 1.5502,
+      "step": 24620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.806565656565657e-05,
+      "loss": 1.5374,
+      "step": 24630
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8060606060606064e-05,
+      "loss": 1.5726,
+      "step": 24640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.805555555555555e-05,
+      "loss": 1.5616,
+      "step": 24650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8050505050505056e-05,
+      "loss": 1.5659,
+      "step": 24660
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8045454545454545e-05,
+      "loss": 1.5509,
+      "step": 24670
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.804040404040404e-05,
+      "loss": 1.5835,
+      "step": 24680
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.803535353535354e-05,
+      "loss": 1.5797,
+      "step": 24690
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.803030303030303e-05,
+      "loss": 1.5678,
+      "step": 24700
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.802525252525252e-05,
+      "loss": 1.5638,
+      "step": 24710
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8020202020202025e-05,
+      "loss": 1.5725,
+      "step": 24720
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8015151515151514e-05,
+      "loss": 1.5517,
+      "step": 24730
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.801010101010101e-05,
+      "loss": 1.5685,
+      "step": 24740
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8005050505050506e-05,
+      "loss": 1.582,
+      "step": 24750
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.8e-05,
+      "loss": 1.5702,
+      "step": 24760
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.799494949494949e-05,
+      "loss": 1.548,
+      "step": 24770
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.7989898989898994e-05,
+      "loss": 1.5823,
+      "step": 24780
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.798484848484848e-05,
+      "loss": 1.5508,
+      "step": 24790
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.797979797979798e-05,
+      "loss": 1.5693,
+      "step": 24800
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.7974747474747475e-05,
+      "loss": 1.5409,
+      "step": 24810
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.796969696969697e-05,
+      "loss": 1.5692,
+      "step": 24820
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.796464646464647e-05,
+      "loss": 1.5811,
+      "step": 24830
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.795959595959596e-05,
+      "loss": 1.5716,
+      "step": 24840
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.795454545454545e-05,
+      "loss": 1.5421,
+      "step": 24850
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.794949494949495e-05,
+      "loss": 1.584,
+      "step": 24860
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.7944444444444444e-05,
+      "loss": 1.5594,
+      "step": 24870
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.793939393939394e-05,
+      "loss": 1.5773,
+      "step": 24880
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.7934343434343436e-05,
+      "loss": 1.553,
+      "step": 24890
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.792929292929293e-05,
+      "loss": 1.5424,
+      "step": 24900
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.792424242424243e-05,
+      "loss": 1.5475,
+      "step": 24910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.791919191919192e-05,
+      "loss": 1.5342,
+      "step": 24920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.791414141414142e-05,
+      "loss": 1.5512,
+      "step": 24930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 3.790909090909091e-05,
+      "loss": 1.5429,
+      "step": 24940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7904040404040406e-05,
+      "loss": 1.5798,
+      "step": 24950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.78989898989899e-05,
+      "loss": 1.5491,
+      "step": 24960
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.78939393939394e-05,
+      "loss": 1.5732,
+      "step": 24970
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7888888888888894e-05,
+      "loss": 1.5379,
+      "step": 24980
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.788383838383839e-05,
+      "loss": 1.5306,
+      "step": 24990
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.787878787878788e-05,
+      "loss": 1.5582,
+      "step": 25000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7873737373737375e-05,
+      "loss": 1.5496,
+      "step": 25010
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.786868686868687e-05,
+      "loss": 1.548,
+      "step": 25020
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.786363636363637e-05,
+      "loss": 1.5765,
+      "step": 25030
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.785858585858586e-05,
+      "loss": 1.5684,
+      "step": 25040
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.785353535353536e-05,
+      "loss": 1.5359,
+      "step": 25050
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.784848484848485e-05,
+      "loss": 1.5509,
+      "step": 25060
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.784343434343435e-05,
+      "loss": 1.5554,
+      "step": 25070
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.783838383838384e-05,
+      "loss": 1.5779,
+      "step": 25080
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7833333333333336e-05,
+      "loss": 1.5623,
+      "step": 25090
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.782828282828283e-05,
+      "loss": 1.5692,
+      "step": 25100
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.782323232323233e-05,
+      "loss": 1.5594,
+      "step": 25110
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.781818181818182e-05,
+      "loss": 1.5445,
+      "step": 25120
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.781313131313132e-05,
+      "loss": 1.5675,
+      "step": 25130
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.780808080808081e-05,
+      "loss": 1.5609,
+      "step": 25140
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7803030303030305e-05,
+      "loss": 1.5766,
+      "step": 25150
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.77979797979798e-05,
+      "loss": 1.5776,
+      "step": 25160
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.77929292929293e-05,
+      "loss": 1.5415,
+      "step": 25170
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7787878787878786e-05,
+      "loss": 1.559,
+      "step": 25180
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.778282828282829e-05,
+      "loss": 1.5727,
+      "step": 25190
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.777777777777778e-05,
+      "loss": 1.5644,
+      "step": 25200
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7772727272727274e-05,
+      "loss": 1.5757,
+      "step": 25210
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.776767676767677e-05,
+      "loss": 1.579,
+      "step": 25220
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7762626262626266e-05,
+      "loss": 1.5374,
+      "step": 25230
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7757575757575755e-05,
+      "loss": 1.5662,
+      "step": 25240
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.775252525252526e-05,
+      "loss": 1.5346,
+      "step": 25250
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.774747474747475e-05,
+      "loss": 1.5842,
+      "step": 25260
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7742424242424243e-05,
+      "loss": 1.5564,
+      "step": 25270
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.773737373737374e-05,
+      "loss": 1.5511,
+      "step": 25280
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7732323232323235e-05,
+      "loss": 1.574,
+      "step": 25290
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7727272727272725e-05,
+      "loss": 1.5532,
+      "step": 25300
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.772222222222223e-05,
+      "loss": 1.5551,
+      "step": 25310
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7717171717171717e-05,
+      "loss": 1.5453,
+      "step": 25320
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.771212121212121e-05,
+      "loss": 1.5544,
+      "step": 25330
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.770707070707071e-05,
+      "loss": 1.559,
+      "step": 25340
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7702020202020205e-05,
+      "loss": 1.558,
+      "step": 25350
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.76969696969697e-05,
+      "loss": 1.5557,
+      "step": 25360
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7691919191919197e-05,
+      "loss": 1.5205,
+      "step": 25370
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7686868686868686e-05,
+      "loss": 1.5426,
+      "step": 25380
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.768181818181818e-05,
+      "loss": 1.558,
+      "step": 25390
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.767676767676768e-05,
+      "loss": 1.5669,
+      "step": 25400
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7671717171717174e-05,
+      "loss": 1.5501,
+      "step": 25410
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.766666666666667e-05,
+      "loss": 1.583,
+      "step": 25420
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7661616161616166e-05,
+      "loss": 1.5643,
+      "step": 25430
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7656565656565655e-05,
+      "loss": 1.5472,
+      "step": 25440
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.765151515151516e-05,
+      "loss": 1.5446,
+      "step": 25450
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.764646464646465e-05,
+      "loss": 1.5536,
+      "step": 25460
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.764141414141414e-05,
+      "loss": 1.5687,
+      "step": 25470
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.763636363636364e-05,
+      "loss": 1.563,
+      "step": 25480
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7631313131313135e-05,
+      "loss": 1.5546,
+      "step": 25490
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7626262626262624e-05,
+      "loss": 1.5547,
+      "step": 25500
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.762121212121213e-05,
+      "loss": 1.5351,
+      "step": 25510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7616161616161616e-05,
+      "loss": 1.5539,
+      "step": 25520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.761111111111111e-05,
+      "loss": 1.5358,
+      "step": 25530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.760606060606061e-05,
+      "loss": 1.5527,
+      "step": 25540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7601010101010104e-05,
+      "loss": 1.5621,
+      "step": 25550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.759595959595959e-05,
+      "loss": 1.5393,
+      "step": 25560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7590909090909096e-05,
+      "loss": 1.54,
+      "step": 25570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.7585858585858585e-05,
+      "loss": 1.5669,
+      "step": 25580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.758080808080808e-05,
+      "loss": 1.5472,
+      "step": 25590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.757575757575758e-05,
+      "loss": 1.5483,
+      "step": 25600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 3.757070707070707e-05,
+      "loss": 1.5412,
+      "step": 25610
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.756565656565656e-05,
+      "loss": 1.5408,
+      "step": 25620
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7560606060606065e-05,
+      "loss": 1.5483,
+      "step": 25630
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7555555555555554e-05,
+      "loss": 1.5427,
+      "step": 25640
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.755050505050505e-05,
+      "loss": 1.5567,
+      "step": 25650
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7545454545454546e-05,
+      "loss": 1.5504,
+      "step": 25660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.754040404040404e-05,
+      "loss": 1.5624,
+      "step": 25670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.753535353535353e-05,
+      "loss": 1.5732,
+      "step": 25680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7530303030303034e-05,
+      "loss": 1.5698,
+      "step": 25690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7525252525252524e-05,
+      "loss": 1.5518,
+      "step": 25700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.752020202020202e-05,
+      "loss": 1.5743,
+      "step": 25710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7515151515151516e-05,
+      "loss": 1.5728,
+      "step": 25720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.751010101010101e-05,
+      "loss": 1.5559,
+      "step": 25730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.75050505050505e-05,
+      "loss": 1.55,
+      "step": 25740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7500000000000003e-05,
+      "loss": 1.5655,
+      "step": 25750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.74949494949495e-05,
+      "loss": 1.5843,
+      "step": 25760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.748989898989899e-05,
+      "loss": 1.554,
+      "step": 25770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.748484848484849e-05,
+      "loss": 1.5678,
+      "step": 25780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.747979797979798e-05,
+      "loss": 1.5522,
+      "step": 25790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.747474747474748e-05,
+      "loss": 1.5418,
+      "step": 25800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.746969696969697e-05,
+      "loss": 1.543,
+      "step": 25810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.746464646464647e-05,
+      "loss": 1.5535,
+      "step": 25820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.745959595959596e-05,
+      "loss": 1.5541,
+      "step": 25830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.745454545454546e-05,
+      "loss": 1.5447,
+      "step": 25840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.744949494949495e-05,
+      "loss": 1.5472,
+      "step": 25850
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7444444444444446e-05,
+      "loss": 1.5662,
+      "step": 25860
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.743939393939394e-05,
+      "loss": 1.5784,
+      "step": 25870
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.743434343434344e-05,
+      "loss": 1.5508,
+      "step": 25880
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7429292929292934e-05,
+      "loss": 1.5674,
+      "step": 25890
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.742424242424243e-05,
+      "loss": 1.5568,
+      "step": 25900
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.741919191919192e-05,
+      "loss": 1.5485,
+      "step": 25910
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7414141414141415e-05,
+      "loss": 1.5548,
+      "step": 25920
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.740909090909091e-05,
+      "loss": 1.5644,
+      "step": 25930
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.740404040404041e-05,
+      "loss": 1.5346,
+      "step": 25940
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.73989898989899e-05,
+      "loss": 1.5469,
+      "step": 25950
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.73939393939394e-05,
+      "loss": 1.5519,
+      "step": 25960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.738888888888889e-05,
+      "loss": 1.5688,
+      "step": 25970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.738383838383839e-05,
+      "loss": 1.5719,
+      "step": 25980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.737878787878788e-05,
+      "loss": 1.5697,
+      "step": 25990
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7373737373737376e-05,
+      "loss": 1.5571,
+      "step": 26000
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.736868686868687e-05,
+      "loss": 1.5295,
+      "step": 26010
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.736363636363637e-05,
+      "loss": 1.5609,
+      "step": 26020
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.735858585858586e-05,
+      "loss": 1.5662,
+      "step": 26030
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.735353535353536e-05,
+      "loss": 1.5501,
+      "step": 26040
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.734848484848485e-05,
+      "loss": 1.5735,
+      "step": 26050
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7343434343434345e-05,
+      "loss": 1.5495,
+      "step": 26060
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.733838383838384e-05,
+      "loss": 1.5666,
+      "step": 26070
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.733333333333334e-05,
+      "loss": 1.5492,
+      "step": 26080
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7328282828282827e-05,
+      "loss": 1.5573,
+      "step": 26090
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.732323232323233e-05,
+      "loss": 1.5617,
+      "step": 26100
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.731818181818182e-05,
+      "loss": 1.5532,
+      "step": 26110
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7313131313131314e-05,
+      "loss": 1.5427,
+      "step": 26120
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.730808080808081e-05,
+      "loss": 1.5584,
+      "step": 26130
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7303030303030306e-05,
+      "loss": 1.5748,
+      "step": 26140
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7297979797979796e-05,
+      "loss": 1.5556,
+      "step": 26150
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.72929292929293e-05,
+      "loss": 1.5459,
+      "step": 26160
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.728787878787879e-05,
+      "loss": 1.524,
+      "step": 26170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7282828282828284e-05,
+      "loss": 1.5464,
+      "step": 26180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.727777777777778e-05,
+      "loss": 1.568,
+      "step": 26190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7272727272727276e-05,
+      "loss": 1.5427,
+      "step": 26200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7267676767676765e-05,
+      "loss": 1.564,
+      "step": 26210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.726262626262627e-05,
+      "loss": 1.5742,
+      "step": 26220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.725757575757576e-05,
+      "loss": 1.5476,
+      "step": 26230
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.725252525252525e-05,
+      "loss": 1.5612,
+      "step": 26240
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.724747474747475e-05,
+      "loss": 1.5495,
+      "step": 26250
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.7242424242424245e-05,
+      "loss": 1.5559,
+      "step": 26260
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 3.723737373737374e-05,
+      "loss": 1.5639,
+      "step": 26270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.723232323232324e-05,
+      "loss": 1.5282,
+      "step": 26280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7227272727272726e-05,
+      "loss": 1.5742,
+      "step": 26290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.722222222222222e-05,
+      "loss": 1.5337,
+      "step": 26300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.721717171717172e-05,
+      "loss": 1.5368,
+      "step": 26310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7212121212121214e-05,
+      "loss": 1.5536,
+      "step": 26320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.720707070707071e-05,
+      "loss": 1.5585,
+      "step": 26330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7202020202020206e-05,
+      "loss": 1.5574,
+      "step": 26340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7196969696969695e-05,
+      "loss": 1.5434,
+      "step": 26350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.719191919191919e-05,
+      "loss": 1.5543,
+      "step": 26360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.718686868686869e-05,
+      "loss": 1.5528,
+      "step": 26370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.718181818181818e-05,
+      "loss": 1.5694,
+      "step": 26380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.717676767676768e-05,
+      "loss": 1.5451,
+      "step": 26390
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7171717171717175e-05,
+      "loss": 1.5534,
+      "step": 26400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7166666666666664e-05,
+      "loss": 1.5477,
+      "step": 26410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.716161616161617e-05,
+      "loss": 1.5412,
+      "step": 26420
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7156565656565656e-05,
+      "loss": 1.555,
+      "step": 26430
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.715151515151515e-05,
+      "loss": 1.5433,
+      "step": 26440
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.714646464646465e-05,
+      "loss": 1.5646,
+      "step": 26450
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7141414141414144e-05,
+      "loss": 1.5635,
+      "step": 26460
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7136363636363633e-05,
+      "loss": 1.5552,
+      "step": 26470
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7131313131313136e-05,
+      "loss": 1.543,
+      "step": 26480
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7126262626262625e-05,
+      "loss": 1.5617,
+      "step": 26490
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.712121212121212e-05,
+      "loss": 1.5641,
+      "step": 26500
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.711616161616162e-05,
+      "loss": 1.5495,
+      "step": 26510
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7111111111111113e-05,
+      "loss": 1.5586,
+      "step": 26520
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.71060606060606e-05,
+      "loss": 1.5537,
+      "step": 26530
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7101010101010105e-05,
+      "loss": 1.5555,
+      "step": 26540
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7095959595959595e-05,
+      "loss": 1.5407,
+      "step": 26550
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.709090909090909e-05,
+      "loss": 1.5448,
+      "step": 26560
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7085858585858587e-05,
+      "loss": 1.5597,
+      "step": 26570
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.708080808080808e-05,
+      "loss": 1.5525,
+      "step": 26580
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.707575757575757e-05,
+      "loss": 1.5594,
+      "step": 26590
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7070707070707075e-05,
+      "loss": 1.5522,
+      "step": 26600
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.706565656565657e-05,
+      "loss": 1.5369,
+      "step": 26610
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.706060606060606e-05,
+      "loss": 1.5292,
+      "step": 26620
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.705555555555556e-05,
+      "loss": 1.5717,
+      "step": 26630
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.705050505050505e-05,
+      "loss": 1.5742,
+      "step": 26640
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.704545454545455e-05,
+      "loss": 1.5574,
+      "step": 26650
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7040404040404044e-05,
+      "loss": 1.5557,
+      "step": 26660
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.703535353535354e-05,
+      "loss": 1.5617,
+      "step": 26670
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.703030303030303e-05,
+      "loss": 1.5573,
+      "step": 26680
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.702525252525253e-05,
+      "loss": 1.5687,
+      "step": 26690
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.702020202020202e-05,
+      "loss": 1.5691,
+      "step": 26700
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.701515151515152e-05,
+      "loss": 1.5685,
+      "step": 26710
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.701010101010101e-05,
+      "loss": 1.5565,
+      "step": 26720
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.700505050505051e-05,
+      "loss": 1.544,
+      "step": 26730
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.7e-05,
+      "loss": 1.5732,
+      "step": 26740
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.69949494949495e-05,
+      "loss": 1.5827,
+      "step": 26750
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.698989898989899e-05,
+      "loss": 1.5426,
+      "step": 26760
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.6984848484848486e-05,
+      "loss": 1.5645,
+      "step": 26770
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.697979797979798e-05,
+      "loss": 1.5551,
+      "step": 26780
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.697474747474748e-05,
+      "loss": 1.5487,
+      "step": 26790
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.6969696969696974e-05,
+      "loss": 1.558,
+      "step": 26800
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.696464646464647e-05,
+      "loss": 1.5555,
+      "step": 26810
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.695959595959596e-05,
+      "loss": 1.5526,
+      "step": 26820
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.6954545454545455e-05,
+      "loss": 1.571,
+      "step": 26830
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.694949494949495e-05,
+      "loss": 1.5675,
+      "step": 26840
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.694444444444445e-05,
+      "loss": 1.5595,
+      "step": 26850
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.693939393939394e-05,
+      "loss": 1.5604,
+      "step": 26860
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.693434343434344e-05,
+      "loss": 1.5364,
+      "step": 26870
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.692929292929293e-05,
+      "loss": 1.555,
+      "step": 26880
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.692424242424243e-05,
+      "loss": 1.5319,
+      "step": 26890
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.691919191919192e-05,
+      "loss": 1.5587,
+      "step": 26900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.6914141414141416e-05,
+      "loss": 1.5507,
+      "step": 26910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.690909090909091e-05,
+      "loss": 1.5465,
+      "step": 26920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.690404040404041e-05,
+      "loss": 1.5402,
+      "step": 26930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 3.68989898989899e-05,
+      "loss": 1.5467,
+      "step": 26940
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.68939393939394e-05,
+      "loss": 1.5703,
+      "step": 26950
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.688888888888889e-05,
+      "loss": 1.5681,
+      "step": 26960
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6883838383838386e-05,
+      "loss": 1.5516,
+      "step": 26970
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.687878787878788e-05,
+      "loss": 1.5401,
+      "step": 26980
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.687373737373738e-05,
+      "loss": 1.5369,
+      "step": 26990
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.686868686868687e-05,
+      "loss": 1.561,
+      "step": 27000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.686363636363637e-05,
+      "loss": 1.5653,
+      "step": 27010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.685858585858586e-05,
+      "loss": 1.5676,
+      "step": 27020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6853535353535355e-05,
+      "loss": 1.5637,
+      "step": 27030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.684848484848485e-05,
+      "loss": 1.5719,
+      "step": 27040
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.684343434343435e-05,
+      "loss": 1.5567,
+      "step": 27050
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6838383838383836e-05,
+      "loss": 1.5673,
+      "step": 27060
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.683333333333334e-05,
+      "loss": 1.536,
+      "step": 27070
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.682828282828283e-05,
+      "loss": 1.5369,
+      "step": 27080
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6823232323232324e-05,
+      "loss": 1.5692,
+      "step": 27090
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.681818181818182e-05,
+      "loss": 1.5735,
+      "step": 27100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6813131313131316e-05,
+      "loss": 1.553,
+      "step": 27110
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6808080808080805e-05,
+      "loss": 1.5496,
+      "step": 27120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.680303030303031e-05,
+      "loss": 1.5522,
+      "step": 27130
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.67979797979798e-05,
+      "loss": 1.5502,
+      "step": 27140
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.679292929292929e-05,
+      "loss": 1.5516,
+      "step": 27150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.678787878787879e-05,
+      "loss": 1.5504,
+      "step": 27160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6782828282828285e-05,
+      "loss": 1.5569,
+      "step": 27170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.677777777777778e-05,
+      "loss": 1.5635,
+      "step": 27180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.677272727272728e-05,
+      "loss": 1.5907,
+      "step": 27190
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6767676767676766e-05,
+      "loss": 1.5453,
+      "step": 27200
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.676262626262626e-05,
+      "loss": 1.5607,
+      "step": 27210
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.675757575757576e-05,
+      "loss": 1.5703,
+      "step": 27220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6752525252525254e-05,
+      "loss": 1.5397,
+      "step": 27230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.674747474747475e-05,
+      "loss": 1.5522,
+      "step": 27240
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6742424242424246e-05,
+      "loss": 1.5607,
+      "step": 27250
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6737373737373735e-05,
+      "loss": 1.562,
+      "step": 27260
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.673232323232323e-05,
+      "loss": 1.577,
+      "step": 27270
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.672727272727273e-05,
+      "loss": 1.5589,
+      "step": 27280
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.672222222222222e-05,
+      "loss": 1.5393,
+      "step": 27290
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.671717171717172e-05,
+      "loss": 1.5727,
+      "step": 27300
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6712121212121215e-05,
+      "loss": 1.5412,
+      "step": 27310
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6707070707070705e-05,
+      "loss": 1.5646,
+      "step": 27320
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.670202020202021e-05,
+      "loss": 1.5804,
+      "step": 27330
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6696969696969697e-05,
+      "loss": 1.5505,
+      "step": 27340
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.669191919191919e-05,
+      "loss": 1.5619,
+      "step": 27350
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.668686868686869e-05,
+      "loss": 1.5569,
+      "step": 27360
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6681818181818185e-05,
+      "loss": 1.5577,
+      "step": 27370
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6676767676767674e-05,
+      "loss": 1.5725,
+      "step": 27380
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6671717171717176e-05,
+      "loss": 1.5655,
+      "step": 27390
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6666666666666666e-05,
+      "loss": 1.5452,
+      "step": 27400
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.666161616161616e-05,
+      "loss": 1.5627,
+      "step": 27410
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.665656565656566e-05,
+      "loss": 1.5622,
+      "step": 27420
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6651515151515154e-05,
+      "loss": 1.5481,
+      "step": 27430
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.664646464646464e-05,
+      "loss": 1.5452,
+      "step": 27440
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6641414141414146e-05,
+      "loss": 1.5667,
+      "step": 27450
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.663636363636364e-05,
+      "loss": 1.5387,
+      "step": 27460
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.663131313131313e-05,
+      "loss": 1.5455,
+      "step": 27470
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6626262626262634e-05,
+      "loss": 1.5386,
+      "step": 27480
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.662121212121212e-05,
+      "loss": 1.5537,
+      "step": 27490
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.661616161616162e-05,
+      "loss": 1.557,
+      "step": 27500
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6611111111111115e-05,
+      "loss": 1.551,
+      "step": 27510
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.660606060606061e-05,
+      "loss": 1.5574,
+      "step": 27520
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.66010101010101e-05,
+      "loss": 1.5684,
+      "step": 27530
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.65959595959596e-05,
+      "loss": 1.5377,
+      "step": 27540
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.659090909090909e-05,
+      "loss": 1.5741,
+      "step": 27550
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.658585858585859e-05,
+      "loss": 1.5604,
+      "step": 27560
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.6580808080808084e-05,
+      "loss": 1.5646,
+      "step": 27570
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.657575757575758e-05,
+      "loss": 1.5636,
+      "step": 27580
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.657070707070707e-05,
+      "loss": 1.5364,
+      "step": 27590
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 3.656565656565657e-05,
+      "loss": 1.5527,
+      "step": 27600
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.656060606060606e-05,
+      "loss": 1.5505,
+      "step": 27610
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.655555555555556e-05,
+      "loss": 1.5774,
+      "step": 27620
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.655050505050505e-05,
+      "loss": 1.5516,
+      "step": 27630
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.654545454545455e-05,
+      "loss": 1.5711,
+      "step": 27640
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.654040404040404e-05,
+      "loss": 1.5784,
+      "step": 27650
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.653535353535354e-05,
+      "loss": 1.5461,
+      "step": 27660
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.653030303030303e-05,
+      "loss": 1.5668,
+      "step": 27670
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6525252525252526e-05,
+      "loss": 1.5506,
+      "step": 27680
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.652020202020202e-05,
+      "loss": 1.5524,
+      "step": 27690
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.651515151515152e-05,
+      "loss": 1.5527,
+      "step": 27700
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6510101010101014e-05,
+      "loss": 1.5344,
+      "step": 27710
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.650505050505051e-05,
+      "loss": 1.5446,
+      "step": 27720
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.65e-05,
+      "loss": 1.5383,
+      "step": 27730
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6494949494949496e-05,
+      "loss": 1.5507,
+      "step": 27740
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.648989898989899e-05,
+      "loss": 1.5442,
+      "step": 27750
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.648484848484849e-05,
+      "loss": 1.5613,
+      "step": 27760
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6479797979797983e-05,
+      "loss": 1.5456,
+      "step": 27770
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.647474747474748e-05,
+      "loss": 1.5505,
+      "step": 27780
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.646969696969697e-05,
+      "loss": 1.543,
+      "step": 27790
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6464646464646465e-05,
+      "loss": 1.5645,
+      "step": 27800
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.645959595959596e-05,
+      "loss": 1.5663,
+      "step": 27810
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.645454545454546e-05,
+      "loss": 1.5626,
+      "step": 27820
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.644949494949495e-05,
+      "loss": 1.5691,
+      "step": 27830
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.644444444444445e-05,
+      "loss": 1.5325,
+      "step": 27840
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.643939393939394e-05,
+      "loss": 1.5361,
+      "step": 27850
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.643434343434344e-05,
+      "loss": 1.5439,
+      "step": 27860
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.642929292929293e-05,
+      "loss": 1.5665,
+      "step": 27870
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6424242424242426e-05,
+      "loss": 1.5448,
+      "step": 27880
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.641919191919192e-05,
+      "loss": 1.5412,
+      "step": 27890
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.641414141414142e-05,
+      "loss": 1.5364,
+      "step": 27900
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.640909090909091e-05,
+      "loss": 1.5185,
+      "step": 27910
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.640404040404041e-05,
+      "loss": 1.5698,
+      "step": 27920
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.63989898989899e-05,
+      "loss": 1.5475,
+      "step": 27930
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6393939393939395e-05,
+      "loss": 1.5689,
+      "step": 27940
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.638888888888889e-05,
+      "loss": 1.5455,
+      "step": 27950
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.638383838383839e-05,
+      "loss": 1.5481,
+      "step": 27960
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6378787878787876e-05,
+      "loss": 1.564,
+      "step": 27970
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.637373737373738e-05,
+      "loss": 1.567,
+      "step": 27980
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.636868686868687e-05,
+      "loss": 1.5693,
+      "step": 27990
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6363636363636364e-05,
+      "loss": 1.5542,
+      "step": 28000
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.635858585858586e-05,
+      "loss": 1.5804,
+      "step": 28010
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6353535353535356e-05,
+      "loss": 1.5486,
+      "step": 28020
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6348484848484845e-05,
+      "loss": 1.5247,
+      "step": 28030
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.634343434343435e-05,
+      "loss": 1.5523,
+      "step": 28040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.633838383838384e-05,
+      "loss": 1.561,
+      "step": 28050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.633333333333333e-05,
+      "loss": 1.5581,
+      "step": 28060
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.632828282828283e-05,
+      "loss": 1.5398,
+      "step": 28070
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6323232323232325e-05,
+      "loss": 1.5322,
+      "step": 28080
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6318181818181815e-05,
+      "loss": 1.5408,
+      "step": 28090
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.631313131313132e-05,
+      "loss": 1.5612,
+      "step": 28100
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6308080808080806e-05,
+      "loss": 1.5761,
+      "step": 28110
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.63030303030303e-05,
+      "loss": 1.5541,
+      "step": 28120
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.62979797979798e-05,
+      "loss": 1.5366,
+      "step": 28130
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6292929292929294e-05,
+      "loss": 1.558,
+      "step": 28140
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.628787878787879e-05,
+      "loss": 1.5689,
+      "step": 28150
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6282828282828286e-05,
+      "loss": 1.5549,
+      "step": 28160
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6277777777777776e-05,
+      "loss": 1.5401,
+      "step": 28170
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.627272727272727e-05,
+      "loss": 1.5436,
+      "step": 28180
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.626767676767677e-05,
+      "loss": 1.5549,
+      "step": 28190
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6262626262626264e-05,
+      "loss": 1.5651,
+      "step": 28200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.625757575757576e-05,
+      "loss": 1.5447,
+      "step": 28210
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6252525252525256e-05,
+      "loss": 1.5561,
+      "step": 28220
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.6247474747474745e-05,
+      "loss": 1.5393,
+      "step": 28230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.624242424242425e-05,
+      "loss": 1.562,
+      "step": 28240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.623737373737374e-05,
+      "loss": 1.5242,
+      "step": 28250
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.623232323232323e-05,
+      "loss": 1.5522,
+      "step": 28260
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 3.622727272727273e-05,
+      "loss": 1.5549,
+      "step": 28270
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6222222222222225e-05,
+      "loss": 1.547,
+      "step": 28280
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6217171717171714e-05,
+      "loss": 1.5657,
+      "step": 28290
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.621212121212122e-05,
+      "loss": 1.5427,
+      "step": 28300
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6207070707070706e-05,
+      "loss": 1.5572,
+      "step": 28310
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.62020202020202e-05,
+      "loss": 1.542,
+      "step": 28320
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6196969696969705e-05,
+      "loss": 1.5289,
+      "step": 28330
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6191919191919194e-05,
+      "loss": 1.5576,
+      "step": 28340
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.618686868686869e-05,
+      "loss": 1.5661,
+      "step": 28350
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6181818181818186e-05,
+      "loss": 1.531,
+      "step": 28360
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.617676767676768e-05,
+      "loss": 1.5339,
+      "step": 28370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.617171717171717e-05,
+      "loss": 1.5567,
+      "step": 28380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6166666666666674e-05,
+      "loss": 1.5425,
+      "step": 28390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.616161616161616e-05,
+      "loss": 1.5443,
+      "step": 28400
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.615656565656566e-05,
+      "loss": 1.54,
+      "step": 28410
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6151515151515155e-05,
+      "loss": 1.5679,
+      "step": 28420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.614646464646465e-05,
+      "loss": 1.5387,
+      "step": 28430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.614141414141414e-05,
+      "loss": 1.5497,
+      "step": 28440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.613636363636364e-05,
+      "loss": 1.5477,
+      "step": 28450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.613131313131313e-05,
+      "loss": 1.564,
+      "step": 28460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.612626262626263e-05,
+      "loss": 1.5484,
+      "step": 28470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6121212121212124e-05,
+      "loss": 1.5523,
+      "step": 28480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.611616161616162e-05,
+      "loss": 1.54,
+      "step": 28490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.611111111111111e-05,
+      "loss": 1.5608,
+      "step": 28500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.610606060606061e-05,
+      "loss": 1.5424,
+      "step": 28510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.61010101010101e-05,
+      "loss": 1.5246,
+      "step": 28520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.60959595959596e-05,
+      "loss": 1.5488,
+      "step": 28530
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6090909090909093e-05,
+      "loss": 1.5674,
+      "step": 28540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.608585858585859e-05,
+      "loss": 1.549,
+      "step": 28550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.608080808080808e-05,
+      "loss": 1.5862,
+      "step": 28560
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.607575757575758e-05,
+      "loss": 1.554,
+      "step": 28570
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.607070707070707e-05,
+      "loss": 1.5785,
+      "step": 28580
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6065656565656567e-05,
+      "loss": 1.5595,
+      "step": 28590
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.606060606060606e-05,
+      "loss": 1.5482,
+      "step": 28600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.605555555555556e-05,
+      "loss": 1.537,
+      "step": 28610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6050505050505055e-05,
+      "loss": 1.5632,
+      "step": 28620
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.604545454545455e-05,
+      "loss": 1.5627,
+      "step": 28630
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.604040404040404e-05,
+      "loss": 1.5767,
+      "step": 28640
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6035353535353536e-05,
+      "loss": 1.5407,
+      "step": 28650
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.603030303030303e-05,
+      "loss": 1.5538,
+      "step": 28660
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.602525252525253e-05,
+      "loss": 1.5655,
+      "step": 28670
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6020202020202024e-05,
+      "loss": 1.5456,
+      "step": 28680
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.601515151515152e-05,
+      "loss": 1.5452,
+      "step": 28690
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.601010101010101e-05,
+      "loss": 1.5574,
+      "step": 28700
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6005050505050505e-05,
+      "loss": 1.542,
+      "step": 28710
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.6e-05,
+      "loss": 1.5618,
+      "step": 28720
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.59949494949495e-05,
+      "loss": 1.552,
+      "step": 28730
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.598989898989899e-05,
+      "loss": 1.5446,
+      "step": 28740
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.598484848484849e-05,
+      "loss": 1.5611,
+      "step": 28750
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.597979797979798e-05,
+      "loss": 1.5263,
+      "step": 28760
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.597474747474748e-05,
+      "loss": 1.5602,
+      "step": 28770
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.596969696969697e-05,
+      "loss": 1.5454,
+      "step": 28780
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.5964646464646466e-05,
+      "loss": 1.56,
+      "step": 28790
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.595959595959596e-05,
+      "loss": 1.563,
+      "step": 28800
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.595454545454546e-05,
+      "loss": 1.5626,
+      "step": 28810
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.594949494949495e-05,
+      "loss": 1.5257,
+      "step": 28820
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.594444444444445e-05,
+      "loss": 1.5583,
+      "step": 28830
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.593939393939394e-05,
+      "loss": 1.5595,
+      "step": 28840
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.5934343434343435e-05,
+      "loss": 1.5299,
+      "step": 28850
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.592929292929293e-05,
+      "loss": 1.5393,
+      "step": 28860
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.592424242424243e-05,
+      "loss": 1.5548,
+      "step": 28870
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.5919191919191916e-05,
+      "loss": 1.5642,
+      "step": 28880
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.591414141414142e-05,
+      "loss": 1.556,
+      "step": 28890
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.590909090909091e-05,
+      "loss": 1.564,
+      "step": 28900
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.5904040404040404e-05,
+      "loss": 1.5595,
+      "step": 28910
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.58989898989899e-05,
+      "loss": 1.5302,
+      "step": 28920
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 3.5893939393939396e-05,
+      "loss": 1.5561,
+      "step": 28930
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5888888888888886e-05,
+      "loss": 1.5506,
+      "step": 28940
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.588383838383839e-05,
+      "loss": 1.5529,
+      "step": 28950
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.587878787878788e-05,
+      "loss": 1.5588,
+      "step": 28960
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5873737373737374e-05,
+      "loss": 1.5146,
+      "step": 28970
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.586868686868687e-05,
+      "loss": 1.5604,
+      "step": 28980
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5863636363636366e-05,
+      "loss": 1.5427,
+      "step": 28990
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5858585858585855e-05,
+      "loss": 1.5433,
+      "step": 29000
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.585353535353536e-05,
+      "loss": 1.5362,
+      "step": 29010
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.584848484848485e-05,
+      "loss": 1.5543,
+      "step": 29020
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.584343434343434e-05,
+      "loss": 1.5191,
+      "step": 29030
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.583838383838384e-05,
+      "loss": 1.534,
+      "step": 29040
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5833333333333335e-05,
+      "loss": 1.5647,
+      "step": 29050
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.582828282828283e-05,
+      "loss": 1.5555,
+      "step": 29060
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.582323232323233e-05,
+      "loss": 1.5577,
+      "step": 29070
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5818181818181816e-05,
+      "loss": 1.5609,
+      "step": 29080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.581313131313131e-05,
+      "loss": 1.5585,
+      "step": 29090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.580808080808081e-05,
+      "loss": 1.5185,
+      "step": 29100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5803030303030304e-05,
+      "loss": 1.5493,
+      "step": 29110
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.57979797979798e-05,
+      "loss": 1.5619,
+      "step": 29120
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5792929292929296e-05,
+      "loss": 1.5432,
+      "step": 29130
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5787878787878785e-05,
+      "loss": 1.5418,
+      "step": 29140
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.578282828282829e-05,
+      "loss": 1.5576,
+      "step": 29150
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.577777777777778e-05,
+      "loss": 1.5551,
+      "step": 29160
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.577272727272727e-05,
+      "loss": 1.5373,
+      "step": 29170
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.576767676767677e-05,
+      "loss": 1.5497,
+      "step": 29180
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5762626262626265e-05,
+      "loss": 1.5439,
+      "step": 29190
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.575757575757576e-05,
+      "loss": 1.5587,
+      "step": 29200
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.575252525252526e-05,
+      "loss": 1.5585,
+      "step": 29210
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.574747474747475e-05,
+      "loss": 1.5647,
+      "step": 29220
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.574242424242424e-05,
+      "loss": 1.5342,
+      "step": 29230
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5737373737373745e-05,
+      "loss": 1.5351,
+      "step": 29240
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5732323232323234e-05,
+      "loss": 1.5445,
+      "step": 29250
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.572727272727273e-05,
+      "loss": 1.5286,
+      "step": 29260
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5722222222222226e-05,
+      "loss": 1.548,
+      "step": 29270
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.571717171717172e-05,
+      "loss": 1.5528,
+      "step": 29280
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.571212121212121e-05,
+      "loss": 1.528,
+      "step": 29290
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5707070707070714e-05,
+      "loss": 1.5735,
+      "step": 29300
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.57020202020202e-05,
+      "loss": 1.5746,
+      "step": 29310
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.56969696969697e-05,
+      "loss": 1.5527,
+      "step": 29320
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5691919191919195e-05,
+      "loss": 1.548,
+      "step": 29330
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.568686868686869e-05,
+      "loss": 1.5427,
+      "step": 29340
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.568181818181818e-05,
+      "loss": 1.5647,
+      "step": 29350
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.567676767676768e-05,
+      "loss": 1.5714,
+      "step": 29360
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.567171717171717e-05,
+      "loss": 1.5245,
+      "step": 29370
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.566666666666667e-05,
+      "loss": 1.5477,
+      "step": 29380
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5661616161616165e-05,
+      "loss": 1.5564,
+      "step": 29390
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.565656565656566e-05,
+      "loss": 1.5596,
+      "step": 29400
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.565151515151515e-05,
+      "loss": 1.5652,
+      "step": 29410
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.564646464646465e-05,
+      "loss": 1.5546,
+      "step": 29420
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.564141414141414e-05,
+      "loss": 1.5539,
+      "step": 29430
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.563636363636364e-05,
+      "loss": 1.5339,
+      "step": 29440
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5631313131313134e-05,
+      "loss": 1.5351,
+      "step": 29450
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.562626262626263e-05,
+      "loss": 1.5393,
+      "step": 29460
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.562121212121212e-05,
+      "loss": 1.5432,
+      "step": 29470
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.561616161616162e-05,
+      "loss": 1.5767,
+      "step": 29480
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.561111111111111e-05,
+      "loss": 1.5419,
+      "step": 29490
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.560606060606061e-05,
+      "loss": 1.57,
+      "step": 29500
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.56010101010101e-05,
+      "loss": 1.5622,
+      "step": 29510
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.55959595959596e-05,
+      "loss": 1.5446,
+      "step": 29520
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.559090909090909e-05,
+      "loss": 1.5511,
+      "step": 29530
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.558585858585859e-05,
+      "loss": 1.5444,
+      "step": 29540
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.558080808080808e-05,
+      "loss": 1.5478,
+      "step": 29550
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5575757575757576e-05,
+      "loss": 1.5617,
+      "step": 29560
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.557070707070707e-05,
+      "loss": 1.5618,
+      "step": 29570
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.556565656565657e-05,
+      "loss": 1.5317,
+      "step": 29580
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.5560606060606064e-05,
+      "loss": 1.5436,
+      "step": 29590
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 3.555555555555556e-05,
+      "loss": 1.5664,
+      "step": 29600
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.555050505050505e-05,
+      "loss": 1.5471,
+      "step": 29610
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5545454545454545e-05,
+      "loss": 1.5439,
+      "step": 29620
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.554040404040404e-05,
+      "loss": 1.5498,
+      "step": 29630
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.553535353535354e-05,
+      "loss": 1.5551,
+      "step": 29640
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.553030303030303e-05,
+      "loss": 1.529,
+      "step": 29650
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.552525252525253e-05,
+      "loss": 1.5348,
+      "step": 29660
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.552020202020202e-05,
+      "loss": 1.5559,
+      "step": 29670
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.551515151515152e-05,
+      "loss": 1.552,
+      "step": 29680
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.551010101010101e-05,
+      "loss": 1.5493,
+      "step": 29690
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5505050505050506e-05,
+      "loss": 1.5725,
+      "step": 29700
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.55e-05,
+      "loss": 1.5501,
+      "step": 29710
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.54949494949495e-05,
+      "loss": 1.5315,
+      "step": 29720
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.548989898989899e-05,
+      "loss": 1.5459,
+      "step": 29730
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.548484848484849e-05,
+      "loss": 1.5557,
+      "step": 29740
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.547979797979798e-05,
+      "loss": 1.539,
+      "step": 29750
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5474747474747475e-05,
+      "loss": 1.5481,
+      "step": 29760
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.546969696969697e-05,
+      "loss": 1.559,
+      "step": 29770
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.546464646464647e-05,
+      "loss": 1.5582,
+      "step": 29780
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.545959595959596e-05,
+      "loss": 1.5484,
+      "step": 29790
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.545454545454546e-05,
+      "loss": 1.5498,
+      "step": 29800
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.544949494949495e-05,
+      "loss": 1.5483,
+      "step": 29810
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5444444444444445e-05,
+      "loss": 1.5391,
+      "step": 29820
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.543939393939394e-05,
+      "loss": 1.5564,
+      "step": 29830
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.543434343434344e-05,
+      "loss": 1.5469,
+      "step": 29840
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5429292929292926e-05,
+      "loss": 1.5425,
+      "step": 29850
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.542424242424243e-05,
+      "loss": 1.5447,
+      "step": 29860
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.541919191919192e-05,
+      "loss": 1.5343,
+      "step": 29870
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5414141414141414e-05,
+      "loss": 1.5613,
+      "step": 29880
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.540909090909091e-05,
+      "loss": 1.5247,
+      "step": 29890
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5404040404040406e-05,
+      "loss": 1.5409,
+      "step": 29900
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5398989898989895e-05,
+      "loss": 1.5392,
+      "step": 29910
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.53939393939394e-05,
+      "loss": 1.5536,
+      "step": 29920
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.538888888888889e-05,
+      "loss": 1.5453,
+      "step": 29930
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.538383838383838e-05,
+      "loss": 1.5535,
+      "step": 29940
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.537878787878788e-05,
+      "loss": 1.5537,
+      "step": 29950
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5373737373737375e-05,
+      "loss": 1.5443,
+      "step": 29960
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.536868686868687e-05,
+      "loss": 1.553,
+      "step": 29970
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.536363636363637e-05,
+      "loss": 1.5392,
+      "step": 29980
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5358585858585856e-05,
+      "loss": 1.563,
+      "step": 29990
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.535353535353535e-05,
+      "loss": 1.5404,
+      "step": 30000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.534848484848485e-05,
+      "loss": 1.5418,
+      "step": 30010
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5343434343434344e-05,
+      "loss": 1.5577,
+      "step": 30020
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.533838383838384e-05,
+      "loss": 1.5488,
+      "step": 30030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5333333333333336e-05,
+      "loss": 1.5445,
+      "step": 30040
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.532828282828283e-05,
+      "loss": 1.5599,
+      "step": 30050
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.532323232323233e-05,
+      "loss": 1.5266,
+      "step": 30060
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5318181818181824e-05,
+      "loss": 1.541,
+      "step": 30070
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.531313131313131e-05,
+      "loss": 1.5364,
+      "step": 30080
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.530808080808081e-05,
+      "loss": 1.538,
+      "step": 30090
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5303030303030305e-05,
+      "loss": 1.53,
+      "step": 30100
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.52979797979798e-05,
+      "loss": 1.5384,
+      "step": 30110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.52929292929293e-05,
+      "loss": 1.5786,
+      "step": 30120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.528787878787879e-05,
+      "loss": 1.5554,
+      "step": 30130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.528282828282828e-05,
+      "loss": 1.5582,
+      "step": 30140
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.527777777777778e-05,
+      "loss": 1.5514,
+      "step": 30150
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5272727272727274e-05,
+      "loss": 1.56,
+      "step": 30160
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.526767676767677e-05,
+      "loss": 1.571,
+      "step": 30170
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5262626262626266e-05,
+      "loss": 1.5457,
+      "step": 30180
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.525757575757576e-05,
+      "loss": 1.5519,
+      "step": 30190
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.525252525252525e-05,
+      "loss": 1.5474,
+      "step": 30200
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5247474747474754e-05,
+      "loss": 1.5586,
+      "step": 30210
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5242424242424244e-05,
+      "loss": 1.5533,
+      "step": 30220
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.523737373737374e-05,
+      "loss": 1.563,
+      "step": 30230
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.5232323232323236e-05,
+      "loss": 1.5197,
+      "step": 30240
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.522727272727273e-05,
+      "loss": 1.5581,
+      "step": 30250
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 3.522222222222222e-05,
+      "loss": 1.5679,
+      "step": 30260
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.5217171717171724e-05,
+      "loss": 1.5262,
+      "step": 30270
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.521212121212121e-05,
+      "loss": 1.5536,
+      "step": 30280
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.520707070707071e-05,
+      "loss": 1.5399,
+      "step": 30290
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.5202020202020205e-05,
+      "loss": 1.5521,
+      "step": 30300
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.51969696969697e-05,
+      "loss": 1.5624,
+      "step": 30310
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.519191919191919e-05,
+      "loss": 1.5393,
+      "step": 30320
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.518686868686869e-05,
+      "loss": 1.5434,
+      "step": 30330
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.518181818181818e-05,
+      "loss": 1.5626,
+      "step": 30340
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.517676767676768e-05,
+      "loss": 1.5347,
+      "step": 30350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.5171717171717174e-05,
+      "loss": 1.5462,
+      "step": 30360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.516666666666667e-05,
+      "loss": 1.5577,
+      "step": 30370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.516161616161616e-05,
+      "loss": 1.5639,
+      "step": 30380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.515656565656566e-05,
+      "loss": 1.5614,
+      "step": 30390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.515151515151515e-05,
+      "loss": 1.5287,
+      "step": 30400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.514646464646465e-05,
+      "loss": 1.5426,
+      "step": 30410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.514141414141414e-05,
+      "loss": 1.5515,
+      "step": 30420
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.513636363636364e-05,
+      "loss": 1.5626,
+      "step": 30430
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.513131313131313e-05,
+      "loss": 1.5655,
+      "step": 30440
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.512626262626263e-05,
+      "loss": 1.5232,
+      "step": 30450
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.512121212121212e-05,
+      "loss": 1.5448,
+      "step": 30460
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.5116161616161616e-05,
+      "loss": 1.5383,
+      "step": 30470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.511111111111111e-05,
+      "loss": 1.5427,
+      "step": 30480
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.510606060606061e-05,
+      "loss": 1.5537,
+      "step": 30490
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.5101010101010104e-05,
+      "loss": 1.5304,
+      "step": 30500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.50959595959596e-05,
+      "loss": 1.553,
+      "step": 30510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.509090909090909e-05,
+      "loss": 1.5413,
+      "step": 30520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.5085858585858585e-05,
+      "loss": 1.5582,
+      "step": 30530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.508080808080808e-05,
+      "loss": 1.5782,
+      "step": 30540
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.507575757575758e-05,
+      "loss": 1.5326,
+      "step": 30550
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.5070707070707073e-05,
+      "loss": 1.5477,
+      "step": 30560
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.506565656565657e-05,
+      "loss": 1.5476,
+      "step": 30570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.506060606060606e-05,
+      "loss": 1.5377,
+      "step": 30580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.505555555555556e-05,
+      "loss": 1.5496,
+      "step": 30590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.505050505050505e-05,
+      "loss": 1.524,
+      "step": 30600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.5045454545454547e-05,
+      "loss": 1.5423,
+      "step": 30610
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.504040404040404e-05,
+      "loss": 1.5517,
+      "step": 30620
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.503535353535354e-05,
+      "loss": 1.5574,
+      "step": 30630
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.503030303030303e-05,
+      "loss": 1.5359,
+      "step": 30640
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.502525252525253e-05,
+      "loss": 1.5387,
+      "step": 30650
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.502020202020202e-05,
+      "loss": 1.5526,
+      "step": 30660
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.5015151515151516e-05,
+      "loss": 1.5438,
+      "step": 30670
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.501010101010101e-05,
+      "loss": 1.556,
+      "step": 30680
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.500505050505051e-05,
+      "loss": 1.545,
+      "step": 30690
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.5e-05,
+      "loss": 1.5468,
+      "step": 30700
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.49949494949495e-05,
+      "loss": 1.537,
+      "step": 30710
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.498989898989899e-05,
+      "loss": 1.536,
+      "step": 30720
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.4984848484848485e-05,
+      "loss": 1.5518,
+      "step": 30730
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.497979797979798e-05,
+      "loss": 1.5594,
+      "step": 30740
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.497474747474748e-05,
+      "loss": 1.5369,
+      "step": 30750
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.4969696969696966e-05,
+      "loss": 1.5433,
+      "step": 30760
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.496464646464647e-05,
+      "loss": 1.5527,
+      "step": 30770
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.495959595959596e-05,
+      "loss": 1.5429,
+      "step": 30780
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.4954545454545454e-05,
+      "loss": 1.5404,
+      "step": 30790
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.494949494949495e-05,
+      "loss": 1.5444,
+      "step": 30800
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.4944444444444446e-05,
+      "loss": 1.5418,
+      "step": 30810
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.4939393939393935e-05,
+      "loss": 1.5303,
+      "step": 30820
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.493434343434344e-05,
+      "loss": 1.552,
+      "step": 30830
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.492929292929293e-05,
+      "loss": 1.5416,
+      "step": 30840
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.492424242424242e-05,
+      "loss": 1.5576,
+      "step": 30850
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.491919191919192e-05,
+      "loss": 1.5327,
+      "step": 30860
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.4914141414141415e-05,
+      "loss": 1.5767,
+      "step": 30870
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.490909090909091e-05,
+      "loss": 1.5444,
+      "step": 30880
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.490404040404041e-05,
+      "loss": 1.5572,
+      "step": 30890
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.48989898989899e-05,
+      "loss": 1.5388,
+      "step": 30900
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.489393939393939e-05,
+      "loss": 1.5445,
+      "step": 30910
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.4888888888888895e-05,
+      "loss": 1.5555,
+      "step": 30920
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 3.4883838383838384e-05,
+      "loss": 1.5573,
+      "step": 30930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.487878787878788e-05,
+      "loss": 1.5265,
+      "step": 30940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4873737373737376e-05,
+      "loss": 1.5521,
+      "step": 30950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.486868686868687e-05,
+      "loss": 1.5423,
+      "step": 30960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.486363636363637e-05,
+      "loss": 1.5469,
+      "step": 30970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4858585858585864e-05,
+      "loss": 1.5689,
+      "step": 30980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4853535353535354e-05,
+      "loss": 1.5507,
+      "step": 30990
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.484848484848485e-05,
+      "loss": 1.5351,
+      "step": 31000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4843434343434346e-05,
+      "loss": 1.5481,
+      "step": 31010
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.483838383838384e-05,
+      "loss": 1.5382,
+      "step": 31020
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.483333333333334e-05,
+      "loss": 1.5361,
+      "step": 31030
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4828282828282834e-05,
+      "loss": 1.5454,
+      "step": 31040
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.482323232323232e-05,
+      "loss": 1.5397,
+      "step": 31050
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.481818181818182e-05,
+      "loss": 1.565,
+      "step": 31060
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4813131313131315e-05,
+      "loss": 1.5265,
+      "step": 31070
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.480808080808081e-05,
+      "loss": 1.5293,
+      "step": 31080
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.480303030303031e-05,
+      "loss": 1.5346,
+      "step": 31090
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.47979797979798e-05,
+      "loss": 1.5516,
+      "step": 31100
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.479292929292929e-05,
+      "loss": 1.5437,
+      "step": 31110
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4787878787878795e-05,
+      "loss": 1.5546,
+      "step": 31120
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4782828282828284e-05,
+      "loss": 1.5331,
+      "step": 31130
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.477777777777778e-05,
+      "loss": 1.5459,
+      "step": 31140
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4772727272727276e-05,
+      "loss": 1.5462,
+      "step": 31150
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.476767676767677e-05,
+      "loss": 1.5321,
+      "step": 31160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.476262626262626e-05,
+      "loss": 1.5798,
+      "step": 31170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4757575757575764e-05,
+      "loss": 1.5438,
+      "step": 31180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.475252525252525e-05,
+      "loss": 1.5149,
+      "step": 31190
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.474747474747475e-05,
+      "loss": 1.5729,
+      "step": 31200
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4742424242424245e-05,
+      "loss": 1.5393,
+      "step": 31210
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.473737373737374e-05,
+      "loss": 1.5337,
+      "step": 31220
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.473232323232323e-05,
+      "loss": 1.5575,
+      "step": 31230
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.472727272727273e-05,
+      "loss": 1.5459,
+      "step": 31240
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.472222222222222e-05,
+      "loss": 1.5471,
+      "step": 31250
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.471717171717172e-05,
+      "loss": 1.5388,
+      "step": 31260
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4712121212121214e-05,
+      "loss": 1.5276,
+      "step": 31270
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.470707070707071e-05,
+      "loss": 1.5408,
+      "step": 31280
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.47020202020202e-05,
+      "loss": 1.549,
+      "step": 31290
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.46969696969697e-05,
+      "loss": 1.5367,
+      "step": 31300
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.469191919191919e-05,
+      "loss": 1.5313,
+      "step": 31310
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.468686868686869e-05,
+      "loss": 1.5498,
+      "step": 31320
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.468181818181818e-05,
+      "loss": 1.5456,
+      "step": 31330
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.467676767676768e-05,
+      "loss": 1.5468,
+      "step": 31340
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.467171717171717e-05,
+      "loss": 1.5283,
+      "step": 31350
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.466666666666667e-05,
+      "loss": 1.5452,
+      "step": 31360
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.466161616161616e-05,
+      "loss": 1.5298,
+      "step": 31370
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4656565656565657e-05,
+      "loss": 1.5306,
+      "step": 31380
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.465151515151515e-05,
+      "loss": 1.5328,
+      "step": 31390
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.464646464646465e-05,
+      "loss": 1.5677,
+      "step": 31400
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4641414141414144e-05,
+      "loss": 1.5359,
+      "step": 31410
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.463636363636364e-05,
+      "loss": 1.552,
+      "step": 31420
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.463131313131313e-05,
+      "loss": 1.5288,
+      "step": 31430
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4626262626262626e-05,
+      "loss": 1.5634,
+      "step": 31440
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.462121212121212e-05,
+      "loss": 1.5478,
+      "step": 31450
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.461616161616162e-05,
+      "loss": 1.5644,
+      "step": 31460
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4611111111111114e-05,
+      "loss": 1.525,
+      "step": 31470
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.460606060606061e-05,
+      "loss": 1.5369,
+      "step": 31480
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.46010101010101e-05,
+      "loss": 1.5382,
+      "step": 31490
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.45959595959596e-05,
+      "loss": 1.5721,
+      "step": 31500
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.459090909090909e-05,
+      "loss": 1.5418,
+      "step": 31510
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.458585858585859e-05,
+      "loss": 1.5388,
+      "step": 31520
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.458080808080808e-05,
+      "loss": 1.5332,
+      "step": 31530
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.457575757575758e-05,
+      "loss": 1.5477,
+      "step": 31540
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.457070707070707e-05,
+      "loss": 1.5319,
+      "step": 31550
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.456565656565657e-05,
+      "loss": 1.5607,
+      "step": 31560
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.456060606060606e-05,
+      "loss": 1.5572,
+      "step": 31570
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.4555555555555556e-05,
+      "loss": 1.5464,
+      "step": 31580
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 3.455050505050505e-05,
+      "loss": 1.5497,
+      "step": 31590
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.454545454545455e-05,
+      "loss": 1.5379,
+      "step": 31600
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.454040404040404e-05,
+      "loss": 1.5384,
+      "step": 31610
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.453535353535354e-05,
+      "loss": 1.5322,
+      "step": 31620
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.453030303030303e-05,
+      "loss": 1.5323,
+      "step": 31630
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4525252525252525e-05,
+      "loss": 1.5566,
+      "step": 31640
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.452020202020202e-05,
+      "loss": 1.5325,
+      "step": 31650
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.451515151515152e-05,
+      "loss": 1.5529,
+      "step": 31660
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4510101010101006e-05,
+      "loss": 1.546,
+      "step": 31670
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.450505050505051e-05,
+      "loss": 1.559,
+      "step": 31680
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.45e-05,
+      "loss": 1.5355,
+      "step": 31690
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4494949494949494e-05,
+      "loss": 1.5399,
+      "step": 31700
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.448989898989899e-05,
+      "loss": 1.5474,
+      "step": 31710
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4484848484848486e-05,
+      "loss": 1.5343,
+      "step": 31720
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.447979797979798e-05,
+      "loss": 1.5714,
+      "step": 31730
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.447474747474748e-05,
+      "loss": 1.5602,
+      "step": 31740
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4469696969696974e-05,
+      "loss": 1.5356,
+      "step": 31750
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4464646464646463e-05,
+      "loss": 1.5565,
+      "step": 31760
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4459595959595966e-05,
+      "loss": 1.5488,
+      "step": 31770
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4454545454545455e-05,
+      "loss": 1.5379,
+      "step": 31780
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.444949494949495e-05,
+      "loss": 1.5423,
+      "step": 31790
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.444444444444445e-05,
+      "loss": 1.539,
+      "step": 31800
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4439393939393943e-05,
+      "loss": 1.5573,
+      "step": 31810
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.443434343434343e-05,
+      "loss": 1.5449,
+      "step": 31820
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4429292929292935e-05,
+      "loss": 1.5381,
+      "step": 31830
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4424242424242425e-05,
+      "loss": 1.5344,
+      "step": 31840
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.441919191919192e-05,
+      "loss": 1.5243,
+      "step": 31850
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.441414141414142e-05,
+      "loss": 1.5363,
+      "step": 31860
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.440909090909091e-05,
+      "loss": 1.546,
+      "step": 31870
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.44040404040404e-05,
+      "loss": 1.5203,
+      "step": 31880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4398989898989905e-05,
+      "loss": 1.5156,
+      "step": 31890
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4393939393939394e-05,
+      "loss": 1.5272,
+      "step": 31900
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.438888888888889e-05,
+      "loss": 1.5671,
+      "step": 31910
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4383838383838386e-05,
+      "loss": 1.5531,
+      "step": 31920
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.437878787878788e-05,
+      "loss": 1.5392,
+      "step": 31930
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.437373737373738e-05,
+      "loss": 1.553,
+      "step": 31940
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4368686868686874e-05,
+      "loss": 1.5448,
+      "step": 31950
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.436363636363636e-05,
+      "loss": 1.5517,
+      "step": 31960
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.435858585858586e-05,
+      "loss": 1.5355,
+      "step": 31970
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4353535353535355e-05,
+      "loss": 1.5437,
+      "step": 31980
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.434848484848485e-05,
+      "loss": 1.5564,
+      "step": 31990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.434343434343435e-05,
+      "loss": 1.5568,
+      "step": 32000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.433838383838384e-05,
+      "loss": 1.552,
+      "step": 32010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.433333333333333e-05,
+      "loss": 1.5454,
+      "step": 32020
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4328282828282835e-05,
+      "loss": 1.536,
+      "step": 32030
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4323232323232324e-05,
+      "loss": 1.5461,
+      "step": 32040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.431818181818182e-05,
+      "loss": 1.5411,
+      "step": 32050
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4313131313131316e-05,
+      "loss": 1.5351,
+      "step": 32060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.430808080808081e-05,
+      "loss": 1.5351,
+      "step": 32070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.43030303030303e-05,
+      "loss": 1.5504,
+      "step": 32080
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4297979797979804e-05,
+      "loss": 1.5546,
+      "step": 32090
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.429292929292929e-05,
+      "loss": 1.5625,
+      "step": 32100
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.428787878787879e-05,
+      "loss": 1.5527,
+      "step": 32110
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4282828282828285e-05,
+      "loss": 1.5381,
+      "step": 32120
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.427777777777778e-05,
+      "loss": 1.5623,
+      "step": 32130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.427272727272727e-05,
+      "loss": 1.5392,
+      "step": 32140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.426767676767677e-05,
+      "loss": 1.5497,
+      "step": 32150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.426262626262626e-05,
+      "loss": 1.5393,
+      "step": 32160
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.425757575757576e-05,
+      "loss": 1.541,
+      "step": 32170
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4252525252525254e-05,
+      "loss": 1.5401,
+      "step": 32180
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.424747474747475e-05,
+      "loss": 1.5416,
+      "step": 32190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.424242424242424e-05,
+      "loss": 1.5529,
+      "step": 32200
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.423737373737374e-05,
+      "loss": 1.5637,
+      "step": 32210
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.423232323232323e-05,
+      "loss": 1.5474,
+      "step": 32220
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.422727272727273e-05,
+      "loss": 1.5666,
+      "step": 32230
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.4222222222222224e-05,
+      "loss": 1.5355,
+      "step": 32240
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.421717171717172e-05,
+      "loss": 1.5524,
+      "step": 32250
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 3.421212121212121e-05,
+      "loss": 1.5443,
+      "step": 32260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.420707070707071e-05,
+      "loss": 1.5202,
+      "step": 32270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.42020202020202e-05,
+      "loss": 1.5403,
+      "step": 32280
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.41969696969697e-05,
+      "loss": 1.5524,
+      "step": 32290
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.419191919191919e-05,
+      "loss": 1.5289,
+      "step": 32300
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.418686868686869e-05,
+      "loss": 1.5399,
+      "step": 32310
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.4181818181818185e-05,
+      "loss": 1.5589,
+      "step": 32320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.417676767676768e-05,
+      "loss": 1.5358,
+      "step": 32330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.417171717171717e-05,
+      "loss": 1.5373,
+      "step": 32340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.4166666666666666e-05,
+      "loss": 1.5557,
+      "step": 32350
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.416161616161616e-05,
+      "loss": 1.5563,
+      "step": 32360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.415656565656566e-05,
+      "loss": 1.5326,
+      "step": 32370
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.4151515151515154e-05,
+      "loss": 1.5434,
+      "step": 32380
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.414646464646465e-05,
+      "loss": 1.5363,
+      "step": 32390
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.414141414141414e-05,
+      "loss": 1.5222,
+      "step": 32400
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.413636363636364e-05,
+      "loss": 1.5421,
+      "step": 32410
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.413131313131313e-05,
+      "loss": 1.5454,
+      "step": 32420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.412626262626263e-05,
+      "loss": 1.5203,
+      "step": 32430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.412121212121212e-05,
+      "loss": 1.516,
+      "step": 32440
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.411616161616162e-05,
+      "loss": 1.5528,
+      "step": 32450
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.411111111111111e-05,
+      "loss": 1.5562,
+      "step": 32460
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.410606060606061e-05,
+      "loss": 1.5417,
+      "step": 32470
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.41010101010101e-05,
+      "loss": 1.5493,
+      "step": 32480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.4095959595959596e-05,
+      "loss": 1.5239,
+      "step": 32490
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.409090909090909e-05,
+      "loss": 1.5306,
+      "step": 32500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.408585858585859e-05,
+      "loss": 1.5396,
+      "step": 32510
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.408080808080808e-05,
+      "loss": 1.552,
+      "step": 32520
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.407575757575758e-05,
+      "loss": 1.5561,
+      "step": 32530
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.407070707070707e-05,
+      "loss": 1.5293,
+      "step": 32540
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.4065656565656565e-05,
+      "loss": 1.5456,
+      "step": 32550
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.406060606060606e-05,
+      "loss": 1.54,
+      "step": 32560
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.405555555555556e-05,
+      "loss": 1.5433,
+      "step": 32570
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.405050505050505e-05,
+      "loss": 1.5528,
+      "step": 32580
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.404545454545455e-05,
+      "loss": 1.5443,
+      "step": 32590
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.4040404040404045e-05,
+      "loss": 1.5309,
+      "step": 32600
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.4035353535353535e-05,
+      "loss": 1.5294,
+      "step": 32610
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.403030303030304e-05,
+      "loss": 1.5403,
+      "step": 32620
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.4025252525252527e-05,
+      "loss": 1.5267,
+      "step": 32630
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.402020202020202e-05,
+      "loss": 1.5221,
+      "step": 32640
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.401515151515152e-05,
+      "loss": 1.531,
+      "step": 32650
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.4010101010101015e-05,
+      "loss": 1.5402,
+      "step": 32660
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.4005050505050504e-05,
+      "loss": 1.5598,
+      "step": 32670
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 1.5625,
+      "step": 32680
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.3994949494949496e-05,
+      "loss": 1.5317,
+      "step": 32690
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.398989898989899e-05,
+      "loss": 1.547,
+      "step": 32700
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.398484848484849e-05,
+      "loss": 1.5203,
+      "step": 32710
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.3979797979797984e-05,
+      "loss": 1.5378,
+      "step": 32720
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.397474747474747e-05,
+      "loss": 1.5333,
+      "step": 32730
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.3969696969696976e-05,
+      "loss": 1.54,
+      "step": 32740
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.3964646464646465e-05,
+      "loss": 1.5611,
+      "step": 32750
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.395959595959596e-05,
+      "loss": 1.5429,
+      "step": 32760
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.395454545454546e-05,
+      "loss": 1.5185,
+      "step": 32770
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.394949494949495e-05,
+      "loss": 1.5449,
+      "step": 32780
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.394444444444444e-05,
+      "loss": 1.5338,
+      "step": 32790
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.3939393939393945e-05,
+      "loss": 1.5273,
+      "step": 32800
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.3934343434343434e-05,
+      "loss": 1.5481,
+      "step": 32810
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.392929292929293e-05,
+      "loss": 1.5527,
+      "step": 32820
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.3924242424242426e-05,
+      "loss": 1.5605,
+      "step": 32830
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.391919191919192e-05,
+      "loss": 1.537,
+      "step": 32840
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.391414141414142e-05,
+      "loss": 1.5667,
+      "step": 32850
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.3909090909090914e-05,
+      "loss": 1.5401,
+      "step": 32860
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.39040404040404e-05,
+      "loss": 1.5178,
+      "step": 32870
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.38989898989899e-05,
+      "loss": 1.5436,
+      "step": 32880
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.3893939393939395e-05,
+      "loss": 1.5357,
+      "step": 32890
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.388888888888889e-05,
+      "loss": 1.5386,
+      "step": 32900
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.388383838383839e-05,
+      "loss": 1.5634,
+      "step": 32910
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 3.387878787878788e-05,
+      "loss": 1.5284,
+      "step": 32920
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.387373737373737e-05,
+      "loss": 1.5402,
+      "step": 32930
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3868686868686875e-05,
+      "loss": 1.535,
+      "step": 32940
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3863636363636364e-05,
+      "loss": 1.5448,
+      "step": 32950
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.385858585858586e-05,
+      "loss": 1.5067,
+      "step": 32960
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3853535353535356e-05,
+      "loss": 1.5572,
+      "step": 32970
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.384848484848485e-05,
+      "loss": 1.5515,
+      "step": 32980
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.384343434343434e-05,
+      "loss": 1.529,
+      "step": 32990
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3838383838383844e-05,
+      "loss": 1.5691,
+      "step": 33000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3833333333333334e-05,
+      "loss": 1.5455,
+      "step": 33010
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.382828282828283e-05,
+      "loss": 1.5385,
+      "step": 33020
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3823232323232326e-05,
+      "loss": 1.5534,
+      "step": 33030
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.381818181818182e-05,
+      "loss": 1.5214,
+      "step": 33040
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.381313131313131e-05,
+      "loss": 1.5589,
+      "step": 33050
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3808080808080813e-05,
+      "loss": 1.561,
+      "step": 33060
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.38030303030303e-05,
+      "loss": 1.5517,
+      "step": 33070
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.37979797979798e-05,
+      "loss": 1.5459,
+      "step": 33080
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3792929292929295e-05,
+      "loss": 1.5146,
+      "step": 33090
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.378787878787879e-05,
+      "loss": 1.5474,
+      "step": 33100
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.378282828282828e-05,
+      "loss": 1.5107,
+      "step": 33110
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.377777777777778e-05,
+      "loss": 1.5381,
+      "step": 33120
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.377272727272727e-05,
+      "loss": 1.532,
+      "step": 33130
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.376767676767677e-05,
+      "loss": 1.5258,
+      "step": 33140
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3762626262626264e-05,
+      "loss": 1.5421,
+      "step": 33150
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.375757575757576e-05,
+      "loss": 1.5378,
+      "step": 33160
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.375252525252525e-05,
+      "loss": 1.5295,
+      "step": 33170
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.374747474747475e-05,
+      "loss": 1.515,
+      "step": 33180
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.374242424242424e-05,
+      "loss": 1.5394,
+      "step": 33190
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.373737373737374e-05,
+      "loss": 1.5304,
+      "step": 33200
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.373232323232323e-05,
+      "loss": 1.5295,
+      "step": 33210
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.372727272727273e-05,
+      "loss": 1.5555,
+      "step": 33220
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3722222222222225e-05,
+      "loss": 1.5529,
+      "step": 33230
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.371717171717172e-05,
+      "loss": 1.5503,
+      "step": 33240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.371212121212121e-05,
+      "loss": 1.5287,
+      "step": 33250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3707070707070706e-05,
+      "loss": 1.5348,
+      "step": 33260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.37020202020202e-05,
+      "loss": 1.532,
+      "step": 33270
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.36969696969697e-05,
+      "loss": 1.534,
+      "step": 33280
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3691919191919194e-05,
+      "loss": 1.5211,
+      "step": 33290
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.368686868686869e-05,
+      "loss": 1.5439,
+      "step": 33300
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.368181818181818e-05,
+      "loss": 1.5446,
+      "step": 33310
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.367676767676768e-05,
+      "loss": 1.5312,
+      "step": 33320
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.367171717171717e-05,
+      "loss": 1.5652,
+      "step": 33330
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.366666666666667e-05,
+      "loss": 1.5398,
+      "step": 33340
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.366161616161616e-05,
+      "loss": 1.5416,
+      "step": 33350
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.365656565656566e-05,
+      "loss": 1.5563,
+      "step": 33360
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.365151515151515e-05,
+      "loss": 1.5456,
+      "step": 33370
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.364646464646465e-05,
+      "loss": 1.544,
+      "step": 33380
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.364141414141414e-05,
+      "loss": 1.5155,
+      "step": 33390
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3636363636363636e-05,
+      "loss": 1.5502,
+      "step": 33400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.363131313131313e-05,
+      "loss": 1.5538,
+      "step": 33410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.362626262626263e-05,
+      "loss": 1.5576,
+      "step": 33420
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.362121212121212e-05,
+      "loss": 1.5361,
+      "step": 33430
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.361616161616162e-05,
+      "loss": 1.5243,
+      "step": 33440
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3611111111111116e-05,
+      "loss": 1.5198,
+      "step": 33450
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3606060606060606e-05,
+      "loss": 1.5145,
+      "step": 33460
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.360101010101011e-05,
+      "loss": 1.5433,
+      "step": 33470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.35959595959596e-05,
+      "loss": 1.5424,
+      "step": 33480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3590909090909094e-05,
+      "loss": 1.5439,
+      "step": 33490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.358585858585859e-05,
+      "loss": 1.514,
+      "step": 33500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3580808080808086e-05,
+      "loss": 1.563,
+      "step": 33510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3575757575757575e-05,
+      "loss": 1.5464,
+      "step": 33520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.357070707070708e-05,
+      "loss": 1.5425,
+      "step": 33530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.356565656565657e-05,
+      "loss": 1.5434,
+      "step": 33540
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.356060606060606e-05,
+      "loss": 1.5333,
+      "step": 33550
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.355555555555556e-05,
+      "loss": 1.543,
+      "step": 33560
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3550505050505055e-05,
+      "loss": 1.5475,
+      "step": 33570
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.3545454545454544e-05,
+      "loss": 1.5308,
+      "step": 33580
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 3.354040404040405e-05,
+      "loss": 1.5215,
+      "step": 33590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3535353535353536e-05,
+      "loss": 1.5248,
+      "step": 33600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.353030303030303e-05,
+      "loss": 1.5428,
+      "step": 33610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.352525252525253e-05,
+      "loss": 1.5283,
+      "step": 33620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3520202020202024e-05,
+      "loss": 1.5506,
+      "step": 33630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.351515151515151e-05,
+      "loss": 1.5299,
+      "step": 33640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3510101010101016e-05,
+      "loss": 1.5738,
+      "step": 33650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3505050505050505e-05,
+      "loss": 1.5407,
+      "step": 33660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.35e-05,
+      "loss": 1.53,
+      "step": 33670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.34949494949495e-05,
+      "loss": 1.5486,
+      "step": 33680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.348989898989899e-05,
+      "loss": 1.5417,
+      "step": 33690
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.348484848484848e-05,
+      "loss": 1.545,
+      "step": 33700
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3479797979797985e-05,
+      "loss": 1.5497,
+      "step": 33710
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3474747474747474e-05,
+      "loss": 1.5453,
+      "step": 33720
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.346969696969697e-05,
+      "loss": 1.5406,
+      "step": 33730
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3464646464646466e-05,
+      "loss": 1.5429,
+      "step": 33740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.345959595959596e-05,
+      "loss": 1.5578,
+      "step": 33750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.345454545454546e-05,
+      "loss": 1.5422,
+      "step": 33760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3449494949494954e-05,
+      "loss": 1.5396,
+      "step": 33770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3444444444444443e-05,
+      "loss": 1.5385,
+      "step": 33780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.343939393939394e-05,
+      "loss": 1.5383,
+      "step": 33790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3434343434343435e-05,
+      "loss": 1.5485,
+      "step": 33800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.342929292929293e-05,
+      "loss": 1.5565,
+      "step": 33810
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.342424242424243e-05,
+      "loss": 1.547,
+      "step": 33820
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3419191919191923e-05,
+      "loss": 1.5593,
+      "step": 33830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.341414141414141e-05,
+      "loss": 1.5463,
+      "step": 33840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3409090909090915e-05,
+      "loss": 1.5475,
+      "step": 33850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3404040404040405e-05,
+      "loss": 1.5091,
+      "step": 33860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.33989898989899e-05,
+      "loss": 1.5273,
+      "step": 33870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3393939393939397e-05,
+      "loss": 1.5473,
+      "step": 33880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.338888888888889e-05,
+      "loss": 1.5428,
+      "step": 33890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.338383838383838e-05,
+      "loss": 1.5598,
+      "step": 33900
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3378787878787885e-05,
+      "loss": 1.5373,
+      "step": 33910
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3373737373737374e-05,
+      "loss": 1.5397,
+      "step": 33920
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.336868686868687e-05,
+      "loss": 1.5447,
+      "step": 33930
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3363636363636366e-05,
+      "loss": 1.5167,
+      "step": 33940
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.335858585858586e-05,
+      "loss": 1.5625,
+      "step": 33950
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.335353535353535e-05,
+      "loss": 1.5305,
+      "step": 33960
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3348484848484854e-05,
+      "loss": 1.5344,
+      "step": 33970
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.334343434343434e-05,
+      "loss": 1.5404,
+      "step": 33980
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.333838383838384e-05,
+      "loss": 1.5174,
+      "step": 33990
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 1.5385,
+      "step": 34000
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.332828282828283e-05,
+      "loss": 1.519,
+      "step": 34010
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.332323232323232e-05,
+      "loss": 1.5414,
+      "step": 34020
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.331818181818182e-05,
+      "loss": 1.5497,
+      "step": 34030
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.331313131313131e-05,
+      "loss": 1.5396,
+      "step": 34040
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.330808080808081e-05,
+      "loss": 1.5291,
+      "step": 34050
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3303030303030304e-05,
+      "loss": 1.5486,
+      "step": 34060
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.32979797979798e-05,
+      "loss": 1.5469,
+      "step": 34070
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.329292929292929e-05,
+      "loss": 1.5492,
+      "step": 34080
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.328787878787879e-05,
+      "loss": 1.5494,
+      "step": 34090
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.328282828282828e-05,
+      "loss": 1.5237,
+      "step": 34100
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.327777777777778e-05,
+      "loss": 1.5533,
+      "step": 34110
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.327272727272727e-05,
+      "loss": 1.5272,
+      "step": 34120
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.326767676767677e-05,
+      "loss": 1.5258,
+      "step": 34130
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3262626262626265e-05,
+      "loss": 1.5123,
+      "step": 34140
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.325757575757576e-05,
+      "loss": 1.5447,
+      "step": 34150
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.325252525252525e-05,
+      "loss": 1.5465,
+      "step": 34160
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3247474747474746e-05,
+      "loss": 1.5254,
+      "step": 34170
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.324242424242424e-05,
+      "loss": 1.554,
+      "step": 34180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.323737373737374e-05,
+      "loss": 1.5478,
+      "step": 34190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3232323232323234e-05,
+      "loss": 1.5564,
+      "step": 34200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.322727272727273e-05,
+      "loss": 1.5308,
+      "step": 34210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.322222222222222e-05,
+      "loss": 1.5205,
+      "step": 34220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.3217171717171716e-05,
+      "loss": 1.519,
+      "step": 34230
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.321212121212121e-05,
+      "loss": 1.5523,
+      "step": 34240
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 3.320707070707071e-05,
+      "loss": 1.5369,
+      "step": 34250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3202020202020204e-05,
+      "loss": 1.5251,
+      "step": 34260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.31969696969697e-05,
+      "loss": 1.5303,
+      "step": 34270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.319191919191919e-05,
+      "loss": 1.5206,
+      "step": 34280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.318686868686869e-05,
+      "loss": 1.5537,
+      "step": 34290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.318181818181819e-05,
+      "loss": 1.529,
+      "step": 34300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.317676767676768e-05,
+      "loss": 1.5342,
+      "step": 34310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.317171717171717e-05,
+      "loss": 1.5469,
+      "step": 34320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.316666666666667e-05,
+      "loss": 1.5181,
+      "step": 34330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3161616161616165e-05,
+      "loss": 1.5254,
+      "step": 34340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.315656565656566e-05,
+      "loss": 1.5216,
+      "step": 34350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.315151515151516e-05,
+      "loss": 1.5346,
+      "step": 34360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3146464646464646e-05,
+      "loss": 1.554,
+      "step": 34370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.314141414141415e-05,
+      "loss": 1.5474,
+      "step": 34380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.313636363636364e-05,
+      "loss": 1.5349,
+      "step": 34390
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3131313131313134e-05,
+      "loss": 1.547,
+      "step": 34400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.312626262626263e-05,
+      "loss": 1.5472,
+      "step": 34410
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3121212121212126e-05,
+      "loss": 1.5413,
+      "step": 34420
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3116161616161615e-05,
+      "loss": 1.5532,
+      "step": 34430
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.311111111111112e-05,
+      "loss": 1.5263,
+      "step": 34440
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.310606060606061e-05,
+      "loss": 1.5315,
+      "step": 34450
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.31010101010101e-05,
+      "loss": 1.5359,
+      "step": 34460
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.30959595959596e-05,
+      "loss": 1.5166,
+      "step": 34470
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3090909090909095e-05,
+      "loss": 1.5447,
+      "step": 34480
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3085858585858584e-05,
+      "loss": 1.505,
+      "step": 34490
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.308080808080809e-05,
+      "loss": 1.5516,
+      "step": 34500
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3075757575757576e-05,
+      "loss": 1.5126,
+      "step": 34510
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.307070707070707e-05,
+      "loss": 1.5407,
+      "step": 34520
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.306565656565657e-05,
+      "loss": 1.5304,
+      "step": 34530
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3060606060606064e-05,
+      "loss": 1.5456,
+      "step": 34540
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3055555555555553e-05,
+      "loss": 1.538,
+      "step": 34550
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3050505050505056e-05,
+      "loss": 1.5162,
+      "step": 34560
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3045454545454545e-05,
+      "loss": 1.5366,
+      "step": 34570
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.304040404040404e-05,
+      "loss": 1.5326,
+      "step": 34580
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.303535353535354e-05,
+      "loss": 1.5526,
+      "step": 34590
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.303030303030303e-05,
+      "loss": 1.5206,
+      "step": 34600
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.302525252525252e-05,
+      "loss": 1.5379,
+      "step": 34610
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3020202020202025e-05,
+      "loss": 1.5623,
+      "step": 34620
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3015151515151515e-05,
+      "loss": 1.5329,
+      "step": 34630
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.301010101010101e-05,
+      "loss": 1.5504,
+      "step": 34640
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3005050505050507e-05,
+      "loss": 1.5612,
+      "step": 34650
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.3e-05,
+      "loss": 1.5281,
+      "step": 34660
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.29949494949495e-05,
+      "loss": 1.5255,
+      "step": 34670
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.2989898989898995e-05,
+      "loss": 1.5093,
+      "step": 34680
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.2984848484848484e-05,
+      "loss": 1.533,
+      "step": 34690
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.297979797979798e-05,
+      "loss": 1.5222,
+      "step": 34700
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.2974747474747476e-05,
+      "loss": 1.5464,
+      "step": 34710
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.296969696969697e-05,
+      "loss": 1.5425,
+      "step": 34720
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.296464646464647e-05,
+      "loss": 1.536,
+      "step": 34730
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.2959595959595964e-05,
+      "loss": 1.5442,
+      "step": 34740
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.295454545454545e-05,
+      "loss": 1.5343,
+      "step": 34750
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.2949494949494956e-05,
+      "loss": 1.5292,
+      "step": 34760
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.2944444444444445e-05,
+      "loss": 1.5524,
+      "step": 34770
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.293939393939394e-05,
+      "loss": 1.5444,
+      "step": 34780
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.293434343434344e-05,
+      "loss": 1.541,
+      "step": 34790
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.292929292929293e-05,
+      "loss": 1.5431,
+      "step": 34800
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.292424242424242e-05,
+      "loss": 1.5539,
+      "step": 34810
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.2919191919191925e-05,
+      "loss": 1.5317,
+      "step": 34820
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.2914141414141414e-05,
+      "loss": 1.5322,
+      "step": 34830
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.290909090909091e-05,
+      "loss": 1.5365,
+      "step": 34840
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.2904040404040406e-05,
+      "loss": 1.5324,
+      "step": 34850
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.28989898989899e-05,
+      "loss": 1.5486,
+      "step": 34860
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.289393939393939e-05,
+      "loss": 1.5252,
+      "step": 34870
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.2888888888888894e-05,
+      "loss": 1.5612,
+      "step": 34880
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.288383838383838e-05,
+      "loss": 1.5293,
+      "step": 34890
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.287878787878788e-05,
+      "loss": 1.5719,
+      "step": 34900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.2873737373737375e-05,
+      "loss": 1.5477,
+      "step": 34910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 3.286868686868687e-05,
+      "loss": 1.5283,
+      "step": 34920
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.286363636363636e-05,
+      "loss": 1.5403,
+      "step": 34930
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.285858585858586e-05,
+      "loss": 1.5172,
+      "step": 34940
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.285353535353535e-05,
+      "loss": 1.5329,
+      "step": 34950
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.284848484848485e-05,
+      "loss": 1.5453,
+      "step": 34960
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2843434343434344e-05,
+      "loss": 1.5281,
+      "step": 34970
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.283838383838384e-05,
+      "loss": 1.5314,
+      "step": 34980
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.283333333333333e-05,
+      "loss": 1.564,
+      "step": 34990
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.282828282828283e-05,
+      "loss": 1.5405,
+      "step": 35000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.282323232323232e-05,
+      "loss": 1.5369,
+      "step": 35010
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.281818181818182e-05,
+      "loss": 1.5394,
+      "step": 35020
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2813131313131314e-05,
+      "loss": 1.5361,
+      "step": 35030
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.280808080808081e-05,
+      "loss": 1.5292,
+      "step": 35040
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2803030303030305e-05,
+      "loss": 1.5514,
+      "step": 35050
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.27979797979798e-05,
+      "loss": 1.536,
+      "step": 35060
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.279292929292929e-05,
+      "loss": 1.5288,
+      "step": 35070
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.278787878787879e-05,
+      "loss": 1.5498,
+      "step": 35080
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.278282828282828e-05,
+      "loss": 1.5462,
+      "step": 35090
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.277777777777778e-05,
+      "loss": 1.5203,
+      "step": 35100
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2772727272727275e-05,
+      "loss": 1.5595,
+      "step": 35110
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.276767676767677e-05,
+      "loss": 1.5188,
+      "step": 35120
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.276262626262626e-05,
+      "loss": 1.5168,
+      "step": 35130
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2757575757575756e-05,
+      "loss": 1.5319,
+      "step": 35140
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.275252525252526e-05,
+      "loss": 1.5622,
+      "step": 35150
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.274747474747475e-05,
+      "loss": 1.5385,
+      "step": 35160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2742424242424244e-05,
+      "loss": 1.5388,
+      "step": 35170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.273737373737374e-05,
+      "loss": 1.5576,
+      "step": 35180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2732323232323236e-05,
+      "loss": 1.5401,
+      "step": 35190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.272727272727273e-05,
+      "loss": 1.5463,
+      "step": 35200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.272222222222223e-05,
+      "loss": 1.5343,
+      "step": 35210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.271717171717172e-05,
+      "loss": 1.5489,
+      "step": 35220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.271212121212121e-05,
+      "loss": 1.5345,
+      "step": 35230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.270707070707071e-05,
+      "loss": 1.5425,
+      "step": 35240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2702020202020205e-05,
+      "loss": 1.5301,
+      "step": 35250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.26969696969697e-05,
+      "loss": 1.5727,
+      "step": 35260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.26919191919192e-05,
+      "loss": 1.5273,
+      "step": 35270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2686868686868686e-05,
+      "loss": 1.5479,
+      "step": 35280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.268181818181819e-05,
+      "loss": 1.5564,
+      "step": 35290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.267676767676768e-05,
+      "loss": 1.5346,
+      "step": 35300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2671717171717174e-05,
+      "loss": 1.5507,
+      "step": 35310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.266666666666667e-05,
+      "loss": 1.5365,
+      "step": 35320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2661616161616166e-05,
+      "loss": 1.5534,
+      "step": 35330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2656565656565655e-05,
+      "loss": 1.5265,
+      "step": 35340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.265151515151516e-05,
+      "loss": 1.5343,
+      "step": 35350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.264646464646465e-05,
+      "loss": 1.5325,
+      "step": 35360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.264141414141414e-05,
+      "loss": 1.5489,
+      "step": 35370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.263636363636364e-05,
+      "loss": 1.5214,
+      "step": 35380
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2631313131313135e-05,
+      "loss": 1.542,
+      "step": 35390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2626262626262624e-05,
+      "loss": 1.5328,
+      "step": 35400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.262121212121213e-05,
+      "loss": 1.5293,
+      "step": 35410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2616161616161616e-05,
+      "loss": 1.5344,
+      "step": 35420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.261111111111111e-05,
+      "loss": 1.5301,
+      "step": 35430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.260606060606061e-05,
+      "loss": 1.5463,
+      "step": 35440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2601010101010104e-05,
+      "loss": 1.535,
+      "step": 35450
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2595959595959594e-05,
+      "loss": 1.5448,
+      "step": 35460
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2590909090909096e-05,
+      "loss": 1.5398,
+      "step": 35470
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2585858585858586e-05,
+      "loss": 1.5405,
+      "step": 35480
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.258080808080808e-05,
+      "loss": 1.5441,
+      "step": 35490
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.257575757575758e-05,
+      "loss": 1.5436,
+      "step": 35500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2570707070707074e-05,
+      "loss": 1.5331,
+      "step": 35510
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.256565656565656e-05,
+      "loss": 1.5393,
+      "step": 35520
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2560606060606066e-05,
+      "loss": 1.5342,
+      "step": 35530
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.2555555555555555e-05,
+      "loss": 1.5143,
+      "step": 35540
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.255050505050505e-05,
+      "loss": 1.5335,
+      "step": 35550
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.254545454545455e-05,
+      "loss": 1.5385,
+      "step": 35560
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.254040404040404e-05,
+      "loss": 1.5327,
+      "step": 35570
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 3.253535353535354e-05,
+      "loss": 1.5268,
+      "step": 35580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2530303030303035e-05,
+      "loss": 1.5512,
+      "step": 35590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2525252525252524e-05,
+      "loss": 1.5323,
+      "step": 35600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.252020202020202e-05,
+      "loss": 1.5356,
+      "step": 35610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2515151515151516e-05,
+      "loss": 1.5404,
+      "step": 35620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.251010101010101e-05,
+      "loss": 1.5651,
+      "step": 35630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.250505050505051e-05,
+      "loss": 1.5415,
+      "step": 35640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2500000000000004e-05,
+      "loss": 1.5451,
+      "step": 35650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.249494949494949e-05,
+      "loss": 1.534,
+      "step": 35660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.248989898989899e-05,
+      "loss": 1.561,
+      "step": 35670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2484848484848485e-05,
+      "loss": 1.5476,
+      "step": 35680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.247979797979798e-05,
+      "loss": 1.5237,
+      "step": 35690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.247474747474748e-05,
+      "loss": 1.5299,
+      "step": 35700
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.246969696969697e-05,
+      "loss": 1.574,
+      "step": 35710
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.246464646464646e-05,
+      "loss": 1.536,
+      "step": 35720
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2459595959595965e-05,
+      "loss": 1.5471,
+      "step": 35730
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2454545454545454e-05,
+      "loss": 1.5346,
+      "step": 35740
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.244949494949495e-05,
+      "loss": 1.5379,
+      "step": 35750
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2444444444444446e-05,
+      "loss": 1.5433,
+      "step": 35760
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.243939393939394e-05,
+      "loss": 1.5528,
+      "step": 35770
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.243434343434343e-05,
+      "loss": 1.5285,
+      "step": 35780
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2429292929292934e-05,
+      "loss": 1.5513,
+      "step": 35790
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2424242424242423e-05,
+      "loss": 1.5435,
+      "step": 35800
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.241919191919192e-05,
+      "loss": 1.53,
+      "step": 35810
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2414141414141415e-05,
+      "loss": 1.5513,
+      "step": 35820
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.240909090909091e-05,
+      "loss": 1.55,
+      "step": 35830
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.24040404040404e-05,
+      "loss": 1.5425,
+      "step": 35840
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2398989898989903e-05,
+      "loss": 1.5319,
+      "step": 35850
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.239393939393939e-05,
+      "loss": 1.5259,
+      "step": 35860
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.238888888888889e-05,
+      "loss": 1.5352,
+      "step": 35870
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2383838383838385e-05,
+      "loss": 1.5262,
+      "step": 35880
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.237878787878788e-05,
+      "loss": 1.5248,
+      "step": 35890
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.237373737373737e-05,
+      "loss": 1.5289,
+      "step": 35900
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.236868686868687e-05,
+      "loss": 1.5539,
+      "step": 35910
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.236363636363636e-05,
+      "loss": 1.5429,
+      "step": 35920
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.235858585858586e-05,
+      "loss": 1.5332,
+      "step": 35930
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2353535353535354e-05,
+      "loss": 1.5472,
+      "step": 35940
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.234848484848485e-05,
+      "loss": 1.5174,
+      "step": 35950
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.234343434343434e-05,
+      "loss": 1.5532,
+      "step": 35960
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.233838383838384e-05,
+      "loss": 1.5387,
+      "step": 35970
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.233333333333333e-05,
+      "loss": 1.5292,
+      "step": 35980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.232828282828283e-05,
+      "loss": 1.5362,
+      "step": 35990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.232323232323233e-05,
+      "loss": 1.5235,
+      "step": 36000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.231818181818182e-05,
+      "loss": 1.5344,
+      "step": 36010
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2313131313131315e-05,
+      "loss": 1.5399,
+      "step": 36020
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.230808080808081e-05,
+      "loss": 1.559,
+      "step": 36030
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.230303030303031e-05,
+      "loss": 1.548,
+      "step": 36040
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2297979797979796e-05,
+      "loss": 1.5253,
+      "step": 36050
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.22929292929293e-05,
+      "loss": 1.5388,
+      "step": 36060
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.228787878787879e-05,
+      "loss": 1.5546,
+      "step": 36070
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2282828282828284e-05,
+      "loss": 1.5503,
+      "step": 36080
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.227777777777778e-05,
+      "loss": 1.5209,
+      "step": 36090
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2272727272727276e-05,
+      "loss": 1.5467,
+      "step": 36100
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.226767676767677e-05,
+      "loss": 1.5298,
+      "step": 36110
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.226262626262627e-05,
+      "loss": 1.533,
+      "step": 36120
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.225757575757576e-05,
+      "loss": 1.5381,
+      "step": 36130
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.225252525252525e-05,
+      "loss": 1.5337,
+      "step": 36140
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.224747474747475e-05,
+      "loss": 1.547,
+      "step": 36150
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2242424242424245e-05,
+      "loss": 1.5471,
+      "step": 36160
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.223737373737374e-05,
+      "loss": 1.5586,
+      "step": 36170
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.223232323232324e-05,
+      "loss": 1.5472,
+      "step": 36180
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2227272727272726e-05,
+      "loss": 1.5239,
+      "step": 36190
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.222222222222223e-05,
+      "loss": 1.5284,
+      "step": 36200
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.221717171717172e-05,
+      "loss": 1.5553,
+      "step": 36210
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2212121212121214e-05,
+      "loss": 1.5439,
+      "step": 36220
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.220707070707071e-05,
+      "loss": 1.5295,
+      "step": 36230
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2202020202020206e-05,
+      "loss": 1.5144,
+      "step": 36240
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 3.2196969696969696e-05,
+      "loss": 1.535,
+      "step": 36250
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.21919191919192e-05,
+      "loss": 1.5605,
+      "step": 36260
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.218686868686869e-05,
+      "loss": 1.5361,
+      "step": 36270
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2181818181818184e-05,
+      "loss": 1.5267,
+      "step": 36280
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.217676767676768e-05,
+      "loss": 1.5491,
+      "step": 36290
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2171717171717176e-05,
+      "loss": 1.5389,
+      "step": 36300
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2166666666666665e-05,
+      "loss": 1.5366,
+      "step": 36310
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.216161616161617e-05,
+      "loss": 1.5365,
+      "step": 36320
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.215656565656566e-05,
+      "loss": 1.5389,
+      "step": 36330
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.215151515151515e-05,
+      "loss": 1.5526,
+      "step": 36340
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.214646464646465e-05,
+      "loss": 1.5639,
+      "step": 36350
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2141414141414145e-05,
+      "loss": 1.5505,
+      "step": 36360
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2136363636363634e-05,
+      "loss": 1.5395,
+      "step": 36370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.213131313131314e-05,
+      "loss": 1.5221,
+      "step": 36380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2126262626262626e-05,
+      "loss": 1.5403,
+      "step": 36390
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.212121212121212e-05,
+      "loss": 1.534,
+      "step": 36400
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.211616161616162e-05,
+      "loss": 1.5455,
+      "step": 36410
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2111111111111114e-05,
+      "loss": 1.5337,
+      "step": 36420
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.21060606060606e-05,
+      "loss": 1.5222,
+      "step": 36430
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2101010101010106e-05,
+      "loss": 1.5406,
+      "step": 36440
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2095959595959595e-05,
+      "loss": 1.5501,
+      "step": 36450
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.209090909090909e-05,
+      "loss": 1.5334,
+      "step": 36460
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.208585858585859e-05,
+      "loss": 1.5239,
+      "step": 36470
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.208080808080808e-05,
+      "loss": 1.535,
+      "step": 36480
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.207575757575758e-05,
+      "loss": 1.5461,
+      "step": 36490
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2070707070707075e-05,
+      "loss": 1.5319,
+      "step": 36500
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2065656565656564e-05,
+      "loss": 1.5461,
+      "step": 36510
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.206060606060606e-05,
+      "loss": 1.5315,
+      "step": 36520
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2055555555555556e-05,
+      "loss": 1.5318,
+      "step": 36530
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.205050505050505e-05,
+      "loss": 1.5325,
+      "step": 36540
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.204545454545455e-05,
+      "loss": 1.5091,
+      "step": 36550
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2040404040404044e-05,
+      "loss": 1.5552,
+      "step": 36560
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2035353535353533e-05,
+      "loss": 1.5262,
+      "step": 36570
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.203030303030303e-05,
+      "loss": 1.5558,
+      "step": 36580
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2025252525252525e-05,
+      "loss": 1.5255,
+      "step": 36590
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.202020202020202e-05,
+      "loss": 1.5099,
+      "step": 36600
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.201515151515152e-05,
+      "loss": 1.535,
+      "step": 36610
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.201010101010101e-05,
+      "loss": 1.5051,
+      "step": 36620
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.20050505050505e-05,
+      "loss": 1.5622,
+      "step": 36630
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 1.543,
+      "step": 36640
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.1994949494949495e-05,
+      "loss": 1.5401,
+      "step": 36650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.198989898989899e-05,
+      "loss": 1.5302,
+      "step": 36660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.1984848484848487e-05,
+      "loss": 1.5473,
+      "step": 36670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.197979797979798e-05,
+      "loss": 1.5388,
+      "step": 36680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.197474747474747e-05,
+      "loss": 1.5489,
+      "step": 36690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.1969696969696974e-05,
+      "loss": 1.5375,
+      "step": 36700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.1964646464646464e-05,
+      "loss": 1.5143,
+      "step": 36710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.195959595959596e-05,
+      "loss": 1.5329,
+      "step": 36720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.1954545454545456e-05,
+      "loss": 1.5218,
+      "step": 36730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.194949494949495e-05,
+      "loss": 1.5429,
+      "step": 36740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.194444444444444e-05,
+      "loss": 1.5304,
+      "step": 36750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.1939393939393944e-05,
+      "loss": 1.5125,
+      "step": 36760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.193434343434343e-05,
+      "loss": 1.54,
+      "step": 36770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.192929292929293e-05,
+      "loss": 1.5499,
+      "step": 36780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.1924242424242425e-05,
+      "loss": 1.5099,
+      "step": 36790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.191919191919192e-05,
+      "loss": 1.5387,
+      "step": 36800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.191414141414141e-05,
+      "loss": 1.529,
+      "step": 36810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.190909090909091e-05,
+      "loss": 1.5386,
+      "step": 36820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.19040404040404e-05,
+      "loss": 1.5294,
+      "step": 36830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.18989898989899e-05,
+      "loss": 1.5323,
+      "step": 36840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.18939393939394e-05,
+      "loss": 1.5248,
+      "step": 36850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.188888888888889e-05,
+      "loss": 1.5354,
+      "step": 36860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.1883838383838386e-05,
+      "loss": 1.5216,
+      "step": 36870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.187878787878788e-05,
+      "loss": 1.5334,
+      "step": 36880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.187373737373738e-05,
+      "loss": 1.5213,
+      "step": 36890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.186868686868687e-05,
+      "loss": 1.5348,
+      "step": 36900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 3.186363636363637e-05,
+      "loss": 1.5403,
+      "step": 36910
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.185858585858586e-05,
+      "loss": 1.5541,
+      "step": 36920
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1853535353535355e-05,
+      "loss": 1.5172,
+      "step": 36930
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.184848484848485e-05,
+      "loss": 1.5337,
+      "step": 36940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.184343434343435e-05,
+      "loss": 1.5228,
+      "step": 36950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1838383838383836e-05,
+      "loss": 1.5604,
+      "step": 36960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.183333333333334e-05,
+      "loss": 1.5426,
+      "step": 36970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.182828282828283e-05,
+      "loss": 1.5538,
+      "step": 36980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1823232323232324e-05,
+      "loss": 1.5607,
+      "step": 36990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.181818181818182e-05,
+      "loss": 1.5296,
+      "step": 37000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1813131313131316e-05,
+      "loss": 1.5261,
+      "step": 37010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.180808080808081e-05,
+      "loss": 1.5335,
+      "step": 37020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.180303030303031e-05,
+      "loss": 1.5371,
+      "step": 37030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.17979797979798e-05,
+      "loss": 1.5306,
+      "step": 37040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1792929292929293e-05,
+      "loss": 1.5329,
+      "step": 37050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.178787878787879e-05,
+      "loss": 1.5143,
+      "step": 37060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1782828282828285e-05,
+      "loss": 1.5298,
+      "step": 37070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.177777777777778e-05,
+      "loss": 1.5133,
+      "step": 37080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.177272727272728e-05,
+      "loss": 1.5337,
+      "step": 37090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.176767676767677e-05,
+      "loss": 1.5248,
+      "step": 37100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.176262626262627e-05,
+      "loss": 1.5461,
+      "step": 37110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.175757575757576e-05,
+      "loss": 1.5451,
+      "step": 37120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1752525252525255e-05,
+      "loss": 1.5399,
+      "step": 37130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.174747474747475e-05,
+      "loss": 1.5236,
+      "step": 37140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.174242424242425e-05,
+      "loss": 1.5199,
+      "step": 37150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1737373737373736e-05,
+      "loss": 1.5431,
+      "step": 37160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.173232323232324e-05,
+      "loss": 1.5293,
+      "step": 37170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.172727272727273e-05,
+      "loss": 1.5101,
+      "step": 37180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1722222222222224e-05,
+      "loss": 1.5239,
+      "step": 37190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.171717171717172e-05,
+      "loss": 1.5245,
+      "step": 37200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1712121212121216e-05,
+      "loss": 1.5433,
+      "step": 37210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1707070707070705e-05,
+      "loss": 1.5371,
+      "step": 37220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.170202020202021e-05,
+      "loss": 1.5409,
+      "step": 37230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.16969696969697e-05,
+      "loss": 1.5318,
+      "step": 37240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.169191919191919e-05,
+      "loss": 1.5424,
+      "step": 37250
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.168686868686869e-05,
+      "loss": 1.5136,
+      "step": 37260
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1681818181818185e-05,
+      "loss": 1.5141,
+      "step": 37270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1676767676767674e-05,
+      "loss": 1.5322,
+      "step": 37280
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.167171717171718e-05,
+      "loss": 1.5311,
+      "step": 37290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1666666666666666e-05,
+      "loss": 1.5228,
+      "step": 37300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.166161616161616e-05,
+      "loss": 1.537,
+      "step": 37310
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.165656565656566e-05,
+      "loss": 1.559,
+      "step": 37320
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1651515151515154e-05,
+      "loss": 1.5514,
+      "step": 37330
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.164646464646464e-05,
+      "loss": 1.5421,
+      "step": 37340
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1641414141414146e-05,
+      "loss": 1.5482,
+      "step": 37350
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1636363636363635e-05,
+      "loss": 1.5147,
+      "step": 37360
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.163131313131313e-05,
+      "loss": 1.5236,
+      "step": 37370
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.162626262626263e-05,
+      "loss": 1.5273,
+      "step": 37380
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.162121212121212e-05,
+      "loss": 1.547,
+      "step": 37390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.161616161616161e-05,
+      "loss": 1.5364,
+      "step": 37400
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1611111111111115e-05,
+      "loss": 1.5301,
+      "step": 37410
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1606060606060604e-05,
+      "loss": 1.5263,
+      "step": 37420
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.16010101010101e-05,
+      "loss": 1.5447,
+      "step": 37430
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1595959595959596e-05,
+      "loss": 1.5328,
+      "step": 37440
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.159090909090909e-05,
+      "loss": 1.5311,
+      "step": 37450
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.158585858585859e-05,
+      "loss": 1.5223,
+      "step": 37460
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1580808080808084e-05,
+      "loss": 1.5473,
+      "step": 37470
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1575757575757574e-05,
+      "loss": 1.5359,
+      "step": 37480
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.157070707070707e-05,
+      "loss": 1.5304,
+      "step": 37490
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1565656565656566e-05,
+      "loss": 1.5292,
+      "step": 37500
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.156060606060606e-05,
+      "loss": 1.5283,
+      "step": 37510
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.155555555555556e-05,
+      "loss": 1.53,
+      "step": 37520
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1550505050505054e-05,
+      "loss": 1.5144,
+      "step": 37530
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.154545454545454e-05,
+      "loss": 1.5342,
+      "step": 37540
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1540404040404046e-05,
+      "loss": 1.5237,
+      "step": 37550
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.1535353535353535e-05,
+      "loss": 1.5559,
+      "step": 37560
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.153030303030303e-05,
+      "loss": 1.532,
+      "step": 37570
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.152525252525253e-05,
+      "loss": 1.5193,
+      "step": 37580
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.152020202020202e-05,
+      "loss": 1.5262,
+      "step": 37590
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.151515151515151e-05,
+      "loss": 1.535,
+      "step": 37600
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1510101010101015e-05,
+      "loss": 1.5362,
+      "step": 37610
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1505050505050504e-05,
+      "loss": 1.5373,
+      "step": 37620
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.15e-05,
+      "loss": 1.5263,
+      "step": 37630
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1494949494949496e-05,
+      "loss": 1.5316,
+      "step": 37640
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.148989898989899e-05,
+      "loss": 1.5375,
+      "step": 37650
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.148484848484848e-05,
+      "loss": 1.53,
+      "step": 37660
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1479797979797984e-05,
+      "loss": 1.5215,
+      "step": 37670
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.147474747474747e-05,
+      "loss": 1.5563,
+      "step": 37680
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.146969696969697e-05,
+      "loss": 1.5519,
+      "step": 37690
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.146464646464647e-05,
+      "loss": 1.5355,
+      "step": 37700
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.145959595959596e-05,
+      "loss": 1.5336,
+      "step": 37710
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.145454545454546e-05,
+      "loss": 1.5443,
+      "step": 37720
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.144949494949495e-05,
+      "loss": 1.5441,
+      "step": 37730
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.144444444444445e-05,
+      "loss": 1.519,
+      "step": 37740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.143939393939394e-05,
+      "loss": 1.5292,
+      "step": 37750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.143434343434344e-05,
+      "loss": 1.5157,
+      "step": 37760
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.142929292929293e-05,
+      "loss": 1.5244,
+      "step": 37770
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1424242424242426e-05,
+      "loss": 1.539,
+      "step": 37780
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.141919191919192e-05,
+      "loss": 1.5199,
+      "step": 37790
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.141414141414142e-05,
+      "loss": 1.5343,
+      "step": 37800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.140909090909091e-05,
+      "loss": 1.5524,
+      "step": 37810
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.140404040404041e-05,
+      "loss": 1.554,
+      "step": 37820
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.13989898989899e-05,
+      "loss": 1.5369,
+      "step": 37830
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1393939393939395e-05,
+      "loss": 1.5331,
+      "step": 37840
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.138888888888889e-05,
+      "loss": 1.5312,
+      "step": 37850
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.138383838383839e-05,
+      "loss": 1.4988,
+      "step": 37860
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.137878787878788e-05,
+      "loss": 1.5365,
+      "step": 37870
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.137373737373738e-05,
+      "loss": 1.525,
+      "step": 37880
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.136868686868687e-05,
+      "loss": 1.5142,
+      "step": 37890
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1363636363636365e-05,
+      "loss": 1.5242,
+      "step": 37900
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.135858585858586e-05,
+      "loss": 1.5212,
+      "step": 37910
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1353535353535357e-05,
+      "loss": 1.5054,
+      "step": 37920
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.134848484848485e-05,
+      "loss": 1.5323,
+      "step": 37930
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.134343434343435e-05,
+      "loss": 1.522,
+      "step": 37940
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.133838383838384e-05,
+      "loss": 1.5221,
+      "step": 37950
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1333333333333334e-05,
+      "loss": 1.5254,
+      "step": 37960
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.132828282828283e-05,
+      "loss": 1.5232,
+      "step": 37970
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1323232323232326e-05,
+      "loss": 1.5436,
+      "step": 37980
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.131818181818182e-05,
+      "loss": 1.5133,
+      "step": 37990
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.131313131313132e-05,
+      "loss": 1.5312,
+      "step": 38000
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.130808080808081e-05,
+      "loss": 1.5306,
+      "step": 38010
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.13030303030303e-05,
+      "loss": 1.5182,
+      "step": 38020
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.12979797979798e-05,
+      "loss": 1.5295,
+      "step": 38030
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1292929292929295e-05,
+      "loss": 1.535,
+      "step": 38040
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.128787878787879e-05,
+      "loss": 1.5435,
+      "step": 38050
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.128282828282829e-05,
+      "loss": 1.5298,
+      "step": 38060
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1277777777777776e-05,
+      "loss": 1.5166,
+      "step": 38070
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.127272727272728e-05,
+      "loss": 1.5316,
+      "step": 38080
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.126767676767677e-05,
+      "loss": 1.5328,
+      "step": 38090
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1262626262626264e-05,
+      "loss": 1.5269,
+      "step": 38100
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.125757575757576e-05,
+      "loss": 1.5196,
+      "step": 38110
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1252525252525256e-05,
+      "loss": 1.542,
+      "step": 38120
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1247474747474745e-05,
+      "loss": 1.5264,
+      "step": 38130
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.124242424242425e-05,
+      "loss": 1.5306,
+      "step": 38140
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.123737373737374e-05,
+      "loss": 1.5269,
+      "step": 38150
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.123232323232323e-05,
+      "loss": 1.5325,
+      "step": 38160
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.122727272727273e-05,
+      "loss": 1.5358,
+      "step": 38170
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1222222222222225e-05,
+      "loss": 1.5278,
+      "step": 38180
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1217171717171714e-05,
+      "loss": 1.5353,
+      "step": 38190
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.121212121212122e-05,
+      "loss": 1.5485,
+      "step": 38200
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1207070707070706e-05,
+      "loss": 1.5604,
+      "step": 38210
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.12020202020202e-05,
+      "loss": 1.5429,
+      "step": 38220
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.11969696969697e-05,
+      "loss": 1.5444,
+      "step": 38230
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1191919191919194e-05,
+      "loss": 1.537,
+      "step": 38240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.1186868686868684e-05,
+      "loss": 1.5112,
+      "step": 38250
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1181818181818186e-05,
+      "loss": 1.5537,
+      "step": 38260
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1176767676767676e-05,
+      "loss": 1.5269,
+      "step": 38270
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.117171717171717e-05,
+      "loss": 1.5084,
+      "step": 38280
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.116666666666667e-05,
+      "loss": 1.5132,
+      "step": 38290
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1161616161616164e-05,
+      "loss": 1.5149,
+      "step": 38300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.115656565656565e-05,
+      "loss": 1.5071,
+      "step": 38310
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1151515151515156e-05,
+      "loss": 1.5404,
+      "step": 38320
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1146464646464645e-05,
+      "loss": 1.5219,
+      "step": 38330
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.114141414141414e-05,
+      "loss": 1.5481,
+      "step": 38340
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.113636363636364e-05,
+      "loss": 1.5295,
+      "step": 38350
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.113131313131313e-05,
+      "loss": 1.5588,
+      "step": 38360
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.112626262626263e-05,
+      "loss": 1.5264,
+      "step": 38370
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1121212121212125e-05,
+      "loss": 1.5211,
+      "step": 38380
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1116161616161614e-05,
+      "loss": 1.5396,
+      "step": 38390
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.111111111111111e-05,
+      "loss": 1.5233,
+      "step": 38400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1106060606060606e-05,
+      "loss": 1.5083,
+      "step": 38410
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.11010101010101e-05,
+      "loss": 1.5377,
+      "step": 38420
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.10959595959596e-05,
+      "loss": 1.5345,
+      "step": 38430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1090909090909094e-05,
+      "loss": 1.5124,
+      "step": 38440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.108585858585858e-05,
+      "loss": 1.5258,
+      "step": 38450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1080808080808086e-05,
+      "loss": 1.5185,
+      "step": 38460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1075757575757575e-05,
+      "loss": 1.5312,
+      "step": 38470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.107070707070707e-05,
+      "loss": 1.5429,
+      "step": 38480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.106565656565657e-05,
+      "loss": 1.5429,
+      "step": 38490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.106060606060606e-05,
+      "loss": 1.5299,
+      "step": 38500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.105555555555555e-05,
+      "loss": 1.5319,
+      "step": 38510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1050505050505055e-05,
+      "loss": 1.551,
+      "step": 38520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1045454545454544e-05,
+      "loss": 1.5423,
+      "step": 38530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.104040404040404e-05,
+      "loss": 1.541,
+      "step": 38540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1035353535353536e-05,
+      "loss": 1.5385,
+      "step": 38550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.103030303030303e-05,
+      "loss": 1.5347,
+      "step": 38560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.102525252525253e-05,
+      "loss": 1.5466,
+      "step": 38570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1020202020202024e-05,
+      "loss": 1.5252,
+      "step": 38580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.101515151515152e-05,
+      "loss": 1.5213,
+      "step": 38590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.101010101010101e-05,
+      "loss": 1.5057,
+      "step": 38600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.100505050505051e-05,
+      "loss": 1.5493,
+      "step": 38610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.1e-05,
+      "loss": 1.546,
+      "step": 38620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.09949494949495e-05,
+      "loss": 1.5438,
+      "step": 38630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.098989898989899e-05,
+      "loss": 1.5346,
+      "step": 38640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.098484848484849e-05,
+      "loss": 1.5426,
+      "step": 38650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.097979797979798e-05,
+      "loss": 1.5378,
+      "step": 38660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.097474747474748e-05,
+      "loss": 1.5359,
+      "step": 38670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.096969696969697e-05,
+      "loss": 1.534,
+      "step": 38680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.0964646464646467e-05,
+      "loss": 1.5273,
+      "step": 38690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.095959595959596e-05,
+      "loss": 1.5393,
+      "step": 38700
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.095454545454546e-05,
+      "loss": 1.5302,
+      "step": 38710
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.094949494949495e-05,
+      "loss": 1.5246,
+      "step": 38720
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.094444444444445e-05,
+      "loss": 1.5371,
+      "step": 38730
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.093939393939394e-05,
+      "loss": 1.5372,
+      "step": 38740
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.0934343434343436e-05,
+      "loss": 1.5261,
+      "step": 38750
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.092929292929293e-05,
+      "loss": 1.5149,
+      "step": 38760
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.092424242424243e-05,
+      "loss": 1.5336,
+      "step": 38770
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.091919191919192e-05,
+      "loss": 1.5311,
+      "step": 38780
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.091414141414142e-05,
+      "loss": 1.5181,
+      "step": 38790
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.090909090909091e-05,
+      "loss": 1.5368,
+      "step": 38800
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.0904040404040405e-05,
+      "loss": 1.5367,
+      "step": 38810
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.08989898989899e-05,
+      "loss": 1.5534,
+      "step": 38820
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.08939393939394e-05,
+      "loss": 1.5349,
+      "step": 38830
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.088888888888889e-05,
+      "loss": 1.5314,
+      "step": 38840
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.088383838383839e-05,
+      "loss": 1.5324,
+      "step": 38850
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.087878787878788e-05,
+      "loss": 1.551,
+      "step": 38860
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.0873737373737374e-05,
+      "loss": 1.526,
+      "step": 38870
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.086868686868687e-05,
+      "loss": 1.5306,
+      "step": 38880
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.0863636363636366e-05,
+      "loss": 1.5385,
+      "step": 38890
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.085858585858586e-05,
+      "loss": 1.5465,
+      "step": 38900
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.085353535353536e-05,
+      "loss": 1.5193,
+      "step": 38910
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.084848484848485e-05,
+      "loss": 1.5271,
+      "step": 38920
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.084343434343434e-05,
+      "loss": 1.5484,
+      "step": 38930
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.083838383838384e-05,
+      "loss": 1.5284,
+      "step": 38940
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0833333333333335e-05,
+      "loss": 1.5419,
+      "step": 38950
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.082828282828283e-05,
+      "loss": 1.5295,
+      "step": 38960
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.082323232323233e-05,
+      "loss": 1.5268,
+      "step": 38970
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0818181818181816e-05,
+      "loss": 1.5521,
+      "step": 38980
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.081313131313132e-05,
+      "loss": 1.5165,
+      "step": 38990
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.080808080808081e-05,
+      "loss": 1.527,
+      "step": 39000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0803030303030304e-05,
+      "loss": 1.538,
+      "step": 39010
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.07979797979798e-05,
+      "loss": 1.5473,
+      "step": 39020
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0792929292929296e-05,
+      "loss": 1.5211,
+      "step": 39030
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0787878787878786e-05,
+      "loss": 1.5114,
+      "step": 39040
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.078282828282829e-05,
+      "loss": 1.5236,
+      "step": 39050
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.077777777777778e-05,
+      "loss": 1.5458,
+      "step": 39060
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0772727272727273e-05,
+      "loss": 1.5253,
+      "step": 39070
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.076767676767677e-05,
+      "loss": 1.5436,
+      "step": 39080
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0762626262626265e-05,
+      "loss": 1.5185,
+      "step": 39090
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0757575757575755e-05,
+      "loss": 1.5487,
+      "step": 39100
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.075252525252526e-05,
+      "loss": 1.5062,
+      "step": 39110
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.074747474747475e-05,
+      "loss": 1.5121,
+      "step": 39120
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.074242424242424e-05,
+      "loss": 1.5202,
+      "step": 39130
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.073737373737374e-05,
+      "loss": 1.531,
+      "step": 39140
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0732323232323235e-05,
+      "loss": 1.5289,
+      "step": 39150
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0727272727272724e-05,
+      "loss": 1.5436,
+      "step": 39160
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0722222222222227e-05,
+      "loss": 1.5204,
+      "step": 39170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0717171717171716e-05,
+      "loss": 1.5459,
+      "step": 39180
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.071212121212121e-05,
+      "loss": 1.5529,
+      "step": 39190
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.070707070707071e-05,
+      "loss": 1.5346,
+      "step": 39200
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0702020202020204e-05,
+      "loss": 1.5488,
+      "step": 39210
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.069696969696969e-05,
+      "loss": 1.5358,
+      "step": 39220
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0691919191919196e-05,
+      "loss": 1.555,
+      "step": 39230
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0686868686868685e-05,
+      "loss": 1.5387,
+      "step": 39240
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.068181818181818e-05,
+      "loss": 1.5309,
+      "step": 39250
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.067676767676768e-05,
+      "loss": 1.5286,
+      "step": 39260
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.067171717171717e-05,
+      "loss": 1.5269,
+      "step": 39270
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.066666666666667e-05,
+      "loss": 1.543,
+      "step": 39280
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0661616161616165e-05,
+      "loss": 1.536,
+      "step": 39290
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0656565656565654e-05,
+      "loss": 1.5256,
+      "step": 39300
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.065151515151515e-05,
+      "loss": 1.5211,
+      "step": 39310
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0646464646464646e-05,
+      "loss": 1.5202,
+      "step": 39320
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.064141414141414e-05,
+      "loss": 1.5482,
+      "step": 39330
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.063636363636364e-05,
+      "loss": 1.5489,
+      "step": 39340
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0631313131313134e-05,
+      "loss": 1.5422,
+      "step": 39350
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.062626262626262e-05,
+      "loss": 1.5426,
+      "step": 39360
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0621212121212126e-05,
+      "loss": 1.5372,
+      "step": 39370
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0616161616161615e-05,
+      "loss": 1.5407,
+      "step": 39380
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.061111111111111e-05,
+      "loss": 1.5239,
+      "step": 39390
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.060606060606061e-05,
+      "loss": 1.5469,
+      "step": 39400
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.06010101010101e-05,
+      "loss": 1.54,
+      "step": 39410
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.05959595959596e-05,
+      "loss": 1.5125,
+      "step": 39420
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0590909090909095e-05,
+      "loss": 1.5303,
+      "step": 39430
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.058585858585859e-05,
+      "loss": 1.522,
+      "step": 39440
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.058080808080808e-05,
+      "loss": 1.5339,
+      "step": 39450
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0575757575757576e-05,
+      "loss": 1.5382,
+      "step": 39460
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.057070707070707e-05,
+      "loss": 1.5192,
+      "step": 39470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.056565656565657e-05,
+      "loss": 1.5423,
+      "step": 39480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0560606060606064e-05,
+      "loss": 1.5346,
+      "step": 39490
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.055555555555556e-05,
+      "loss": 1.5285,
+      "step": 39500
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.055050505050505e-05,
+      "loss": 1.5124,
+      "step": 39510
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.054545454545455e-05,
+      "loss": 1.5205,
+      "step": 39520
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.054040404040404e-05,
+      "loss": 1.5294,
+      "step": 39530
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.053535353535354e-05,
+      "loss": 1.5309,
+      "step": 39540
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0530303030303034e-05,
+      "loss": 1.5144,
+      "step": 39550
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.052525252525253e-05,
+      "loss": 1.5191,
+      "step": 39560
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.052020202020202e-05,
+      "loss": 1.5244,
+      "step": 39570
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.0515151515151518e-05,
+      "loss": 1.5368,
+      "step": 39580
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.051010101010101e-05,
+      "loss": 1.534,
+      "step": 39590
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.050505050505051e-05,
+      "loss": 1.5355,
+      "step": 39600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.05e-05,
+      "loss": 1.5276,
+      "step": 39610
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.04949494949495e-05,
+      "loss": 1.5111,
+      "step": 39620
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.048989898989899e-05,
+      "loss": 1.5364,
+      "step": 39630
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0484848484848487e-05,
+      "loss": 1.5255,
+      "step": 39640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.047979797979798e-05,
+      "loss": 1.5137,
+      "step": 39650
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.047474747474748e-05,
+      "loss": 1.525,
+      "step": 39660
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0469696969696972e-05,
+      "loss": 1.5412,
+      "step": 39670
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0464646464646468e-05,
+      "loss": 1.5497,
+      "step": 39680
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.045959595959596e-05,
+      "loss": 1.5366,
+      "step": 39690
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0454545454545456e-05,
+      "loss": 1.5274,
+      "step": 39700
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.044949494949495e-05,
+      "loss": 1.5126,
+      "step": 39710
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.044444444444445e-05,
+      "loss": 1.5322,
+      "step": 39720
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.043939393939394e-05,
+      "loss": 1.5313,
+      "step": 39730
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0434343434343437e-05,
+      "loss": 1.539,
+      "step": 39740
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.042929292929293e-05,
+      "loss": 1.5225,
+      "step": 39750
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0424242424242426e-05,
+      "loss": 1.5436,
+      "step": 39760
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0419191919191918e-05,
+      "loss": 1.5332,
+      "step": 39770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0414141414141418e-05,
+      "loss": 1.5329,
+      "step": 39780
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.040909090909091e-05,
+      "loss": 1.5429,
+      "step": 39790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0404040404040406e-05,
+      "loss": 1.5431,
+      "step": 39800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.03989898989899e-05,
+      "loss": 1.54,
+      "step": 39810
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0393939393939398e-05,
+      "loss": 1.5128,
+      "step": 39820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0388888888888887e-05,
+      "loss": 1.5208,
+      "step": 39830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0383838383838387e-05,
+      "loss": 1.5138,
+      "step": 39840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.037878787878788e-05,
+      "loss": 1.54,
+      "step": 39850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0373737373737375e-05,
+      "loss": 1.5304,
+      "step": 39860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0368686868686868e-05,
+      "loss": 1.5284,
+      "step": 39870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0363636363636367e-05,
+      "loss": 1.5264,
+      "step": 39880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.035858585858586e-05,
+      "loss": 1.5439,
+      "step": 39890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0353535353535356e-05,
+      "loss": 1.5516,
+      "step": 39900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.034848484848485e-05,
+      "loss": 1.5395,
+      "step": 39910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0343434343434345e-05,
+      "loss": 1.5517,
+      "step": 39920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0338383838383837e-05,
+      "loss": 1.5459,
+      "step": 39930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0333333333333337e-05,
+      "loss": 1.5296,
+      "step": 39940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.032828282828283e-05,
+      "loss": 1.5101,
+      "step": 39950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0323232323232325e-05,
+      "loss": 1.5154,
+      "step": 39960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0318181818181818e-05,
+      "loss": 1.5149,
+      "step": 39970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0313131313131317e-05,
+      "loss": 1.5423,
+      "step": 39980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0308080808080806e-05,
+      "loss": 1.5472,
+      "step": 39990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0303030303030306e-05,
+      "loss": 1.5659,
+      "step": 40000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.02979797979798e-05,
+      "loss": 1.5554,
+      "step": 40010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0292929292929294e-05,
+      "loss": 1.5302,
+      "step": 40020
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0287878787878787e-05,
+      "loss": 1.5242,
+      "step": 40030
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0282828282828286e-05,
+      "loss": 1.5131,
+      "step": 40040
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0277777777777776e-05,
+      "loss": 1.5426,
+      "step": 40050
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0272727272727275e-05,
+      "loss": 1.5252,
+      "step": 40060
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0267676767676767e-05,
+      "loss": 1.5092,
+      "step": 40070
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0262626262626263e-05,
+      "loss": 1.5176,
+      "step": 40080
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0257575757575756e-05,
+      "loss": 1.5113,
+      "step": 40090
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0252525252525255e-05,
+      "loss": 1.523,
+      "step": 40100
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0247474747474748e-05,
+      "loss": 1.529,
+      "step": 40110
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0242424242424244e-05,
+      "loss": 1.5157,
+      "step": 40120
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0237373737373737e-05,
+      "loss": 1.5291,
+      "step": 40130
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0232323232323233e-05,
+      "loss": 1.5145,
+      "step": 40140
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0227272727272725e-05,
+      "loss": 1.5243,
+      "step": 40150
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0222222222222225e-05,
+      "loss": 1.5507,
+      "step": 40160
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0217171717171717e-05,
+      "loss": 1.533,
+      "step": 40170
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0212121212121213e-05,
+      "loss": 1.5112,
+      "step": 40180
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0207070707070706e-05,
+      "loss": 1.5258,
+      "step": 40190
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0202020202020205e-05,
+      "loss": 1.5211,
+      "step": 40200
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0196969696969694e-05,
+      "loss": 1.5501,
+      "step": 40210
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0191919191919194e-05,
+      "loss": 1.5352,
+      "step": 40220
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0186868686868686e-05,
+      "loss": 1.5043,
+      "step": 40230
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.0181818181818182e-05,
+      "loss": 1.5284,
+      "step": 40240
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0176767676767675e-05,
+      "loss": 1.5191,
+      "step": 40250
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0171717171717174e-05,
+      "loss": 1.5072,
+      "step": 40260
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.016666666666667e-05,
+      "loss": 1.5353,
+      "step": 40270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0161616161616163e-05,
+      "loss": 1.5318,
+      "step": 40280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0156565656565662e-05,
+      "loss": 1.5508,
+      "step": 40290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.015151515151515e-05,
+      "loss": 1.5205,
+      "step": 40300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.014646464646465e-05,
+      "loss": 1.5119,
+      "step": 40310
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0141414141414144e-05,
+      "loss": 1.5357,
+      "step": 40320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.013636363636364e-05,
+      "loss": 1.5286,
+      "step": 40330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0131313131313132e-05,
+      "loss": 1.5181,
+      "step": 40340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.012626262626263e-05,
+      "loss": 1.529,
+      "step": 40350
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.012121212121212e-05,
+      "loss": 1.5078,
+      "step": 40360
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.011616161616162e-05,
+      "loss": 1.5374,
+      "step": 40370
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0111111111111113e-05,
+      "loss": 1.5508,
+      "step": 40380
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.010606060606061e-05,
+      "loss": 1.5438,
+      "step": 40390
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.01010101010101e-05,
+      "loss": 1.5121,
+      "step": 40400
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.00959595959596e-05,
+      "loss": 1.5494,
+      "step": 40410
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0090909090909093e-05,
+      "loss": 1.5056,
+      "step": 40420
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.008585858585859e-05,
+      "loss": 1.5312,
+      "step": 40430
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0080808080808082e-05,
+      "loss": 1.512,
+      "step": 40440
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0075757575757578e-05,
+      "loss": 1.5109,
+      "step": 40450
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.007070707070707e-05,
+      "loss": 1.5229,
+      "step": 40460
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.006565656565657e-05,
+      "loss": 1.5153,
+      "step": 40470
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0060606060606062e-05,
+      "loss": 1.5469,
+      "step": 40480
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.005555555555556e-05,
+      "loss": 1.523,
+      "step": 40490
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.005050505050505e-05,
+      "loss": 1.5273,
+      "step": 40500
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.004545454545455e-05,
+      "loss": 1.5448,
+      "step": 40510
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.004040404040404e-05,
+      "loss": 1.5185,
+      "step": 40520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.003535353535354e-05,
+      "loss": 1.5276,
+      "step": 40530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.003030303030303e-05,
+      "loss": 1.5042,
+      "step": 40540
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0025252525252528e-05,
+      "loss": 1.5391,
+      "step": 40550
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.002020202020202e-05,
+      "loss": 1.542,
+      "step": 40560
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.001515151515152e-05,
+      "loss": 1.5533,
+      "step": 40570
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0010101010101012e-05,
+      "loss": 1.5249,
+      "step": 40580
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.0005050505050508e-05,
+      "loss": 1.5364,
+      "step": 40590
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3e-05,
+      "loss": 1.5204,
+      "step": 40600
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9994949494949497e-05,
+      "loss": 1.5453,
+      "step": 40610
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.998989898989899e-05,
+      "loss": 1.5359,
+      "step": 40620
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.998484848484849e-05,
+      "loss": 1.5266,
+      "step": 40630
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.997979797979798e-05,
+      "loss": 1.5159,
+      "step": 40640
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9974747474747477e-05,
+      "loss": 1.5128,
+      "step": 40650
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.996969696969697e-05,
+      "loss": 1.5351,
+      "step": 40660
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9964646464646466e-05,
+      "loss": 1.5269,
+      "step": 40670
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.995959595959596e-05,
+      "loss": 1.5398,
+      "step": 40680
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9954545454545458e-05,
+      "loss": 1.5017,
+      "step": 40690
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.994949494949495e-05,
+      "loss": 1.5257,
+      "step": 40700
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9944444444444446e-05,
+      "loss": 1.5025,
+      "step": 40710
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.993939393939394e-05,
+      "loss": 1.5177,
+      "step": 40720
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.993434343434344e-05,
+      "loss": 1.5153,
+      "step": 40730
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9929292929292928e-05,
+      "loss": 1.5536,
+      "step": 40740
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9924242424242427e-05,
+      "loss": 1.527,
+      "step": 40750
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.991919191919192e-05,
+      "loss": 1.5226,
+      "step": 40760
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9914141414141416e-05,
+      "loss": 1.5299,
+      "step": 40770
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9909090909090908e-05,
+      "loss": 1.5385,
+      "step": 40780
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9904040404040408e-05,
+      "loss": 1.5468,
+      "step": 40790
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.98989898989899e-05,
+      "loss": 1.524,
+      "step": 40800
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9893939393939396e-05,
+      "loss": 1.5481,
+      "step": 40810
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.988888888888889e-05,
+      "loss": 1.5121,
+      "step": 40820
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9883838383838385e-05,
+      "loss": 1.5205,
+      "step": 40830
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9878787878787877e-05,
+      "loss": 1.5234,
+      "step": 40840
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9873737373737377e-05,
+      "loss": 1.5385,
+      "step": 40850
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.986868686868687e-05,
+      "loss": 1.5363,
+      "step": 40860
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9863636363636365e-05,
+      "loss": 1.5174,
+      "step": 40870
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9858585858585858e-05,
+      "loss": 1.5533,
+      "step": 40880
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9853535353535357e-05,
+      "loss": 1.5324,
+      "step": 40890
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9848484848484847e-05,
+      "loss": 1.5401,
+      "step": 40900
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2.9843434343434346e-05,
+      "loss": 1.5051,
+      "step": 40910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.983838383838384e-05,
+      "loss": 1.5534,
+      "step": 40920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9833333333333335e-05,
+      "loss": 1.5154,
+      "step": 40930
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9828282828282827e-05,
+      "loss": 1.5215,
+      "step": 40940
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9823232323232327e-05,
+      "loss": 1.5336,
+      "step": 40950
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9818181818181816e-05,
+      "loss": 1.5207,
+      "step": 40960
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9813131313131315e-05,
+      "loss": 1.5273,
+      "step": 40970
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9808080808080808e-05,
+      "loss": 1.5423,
+      "step": 40980
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9803030303030304e-05,
+      "loss": 1.5308,
+      "step": 40990
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9797979797979796e-05,
+      "loss": 1.5182,
+      "step": 41000
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9792929292929296e-05,
+      "loss": 1.5306,
+      "step": 41010
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.978787878787879e-05,
+      "loss": 1.5392,
+      "step": 41020
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9782828282828284e-05,
+      "loss": 1.5313,
+      "step": 41030
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9777777777777777e-05,
+      "loss": 1.532,
+      "step": 41040
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9772727272727273e-05,
+      "loss": 1.5227,
+      "step": 41050
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9767676767676765e-05,
+      "loss": 1.5377,
+      "step": 41060
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9762626262626265e-05,
+      "loss": 1.5179,
+      "step": 41070
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9757575757575757e-05,
+      "loss": 1.5388,
+      "step": 41080
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9752525252525253e-05,
+      "loss": 1.5547,
+      "step": 41090
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9747474747474746e-05,
+      "loss": 1.5308,
+      "step": 41100
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9742424242424245e-05,
+      "loss": 1.519,
+      "step": 41110
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.973737373737374e-05,
+      "loss": 1.5288,
+      "step": 41120
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9732323232323234e-05,
+      "loss": 1.5221,
+      "step": 41130
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.972727272727273e-05,
+      "loss": 1.5551,
+      "step": 41140
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9722222222222223e-05,
+      "loss": 1.5321,
+      "step": 41150
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9717171717171722e-05,
+      "loss": 1.5191,
+      "step": 41160
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9712121212121215e-05,
+      "loss": 1.5274,
+      "step": 41170
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.970707070707071e-05,
+      "loss": 1.5313,
+      "step": 41180
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9702020202020203e-05,
+      "loss": 1.5084,
+      "step": 41190
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.96969696969697e-05,
+      "loss": 1.5211,
+      "step": 41200
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9691919191919192e-05,
+      "loss": 1.5416,
+      "step": 41210
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.968686868686869e-05,
+      "loss": 1.5506,
+      "step": 41220
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9681818181818184e-05,
+      "loss": 1.5121,
+      "step": 41230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.967676767676768e-05,
+      "loss": 1.5288,
+      "step": 41240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9671717171717172e-05,
+      "loss": 1.534,
+      "step": 41250
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9666666666666672e-05,
+      "loss": 1.5374,
+      "step": 41260
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.966161616161616e-05,
+      "loss": 1.5412,
+      "step": 41270
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.965656565656566e-05,
+      "loss": 1.5429,
+      "step": 41280
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9651515151515153e-05,
+      "loss": 1.5289,
+      "step": 41290
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.964646464646465e-05,
+      "loss": 1.5341,
+      "step": 41300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.964141414141414e-05,
+      "loss": 1.5365,
+      "step": 41310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.963636363636364e-05,
+      "loss": 1.5483,
+      "step": 41320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9631313131313134e-05,
+      "loss": 1.5043,
+      "step": 41330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.962626262626263e-05,
+      "loss": 1.5319,
+      "step": 41340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9621212121212122e-05,
+      "loss": 1.5154,
+      "step": 41350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9616161616161618e-05,
+      "loss": 1.5149,
+      "step": 41360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.961111111111111e-05,
+      "loss": 1.5315,
+      "step": 41370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.960606060606061e-05,
+      "loss": 1.5263,
+      "step": 41380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9601010101010103e-05,
+      "loss": 1.5275,
+      "step": 41390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.95959595959596e-05,
+      "loss": 1.5278,
+      "step": 41400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.959090909090909e-05,
+      "loss": 1.5361,
+      "step": 41410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.958585858585859e-05,
+      "loss": 1.5505,
+      "step": 41420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.958080808080808e-05,
+      "loss": 1.5357,
+      "step": 41430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.957575757575758e-05,
+      "loss": 1.5337,
+      "step": 41440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9570707070707072e-05,
+      "loss": 1.521,
+      "step": 41450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9565656565656568e-05,
+      "loss": 1.5413,
+      "step": 41460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.956060606060606e-05,
+      "loss": 1.5309,
+      "step": 41470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.955555555555556e-05,
+      "loss": 1.515,
+      "step": 41480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.955050505050505e-05,
+      "loss": 1.5302,
+      "step": 41490
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.954545454545455e-05,
+      "loss": 1.5181,
+      "step": 41500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.954040404040404e-05,
+      "loss": 1.5142,
+      "step": 41510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9535353535353537e-05,
+      "loss": 1.5158,
+      "step": 41520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.953030303030303e-05,
+      "loss": 1.5215,
+      "step": 41530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.952525252525253e-05,
+      "loss": 1.5206,
+      "step": 41540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.952020202020202e-05,
+      "loss": 1.5258,
+      "step": 41550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.9515151515151518e-05,
+      "loss": 1.5283,
+      "step": 41560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2.951010101010101e-05,
+      "loss": 1.5265,
+      "step": 41570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9505050505050506e-05,
+      "loss": 1.5182,
+      "step": 41580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.95e-05,
+      "loss": 1.5326,
+      "step": 41590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9494949494949498e-05,
+      "loss": 1.5312,
+      "step": 41600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.948989898989899e-05,
+      "loss": 1.5364,
+      "step": 41610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9484848484848487e-05,
+      "loss": 1.5276,
+      "step": 41620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.947979797979798e-05,
+      "loss": 1.512,
+      "step": 41630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.947474747474748e-05,
+      "loss": 1.504,
+      "step": 41640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9469696969696968e-05,
+      "loss": 1.5545,
+      "step": 41650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9464646464646467e-05,
+      "loss": 1.5099,
+      "step": 41660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.945959595959596e-05,
+      "loss": 1.5454,
+      "step": 41670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9454545454545456e-05,
+      "loss": 1.5351,
+      "step": 41680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.944949494949495e-05,
+      "loss": 1.5144,
+      "step": 41690
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9444444444444448e-05,
+      "loss": 1.5181,
+      "step": 41700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.943939393939394e-05,
+      "loss": 1.5066,
+      "step": 41710
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9434343434343436e-05,
+      "loss": 1.5286,
+      "step": 41720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.942929292929293e-05,
+      "loss": 1.5315,
+      "step": 41730
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9424242424242425e-05,
+      "loss": 1.5431,
+      "step": 41740
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9419191919191918e-05,
+      "loss": 1.5399,
+      "step": 41750
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9414141414141417e-05,
+      "loss": 1.4949,
+      "step": 41760
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.940909090909091e-05,
+      "loss": 1.5401,
+      "step": 41770
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9404040404040406e-05,
+      "loss": 1.5458,
+      "step": 41780
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9398989898989898e-05,
+      "loss": 1.5317,
+      "step": 41790
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9393939393939394e-05,
+      "loss": 1.532,
+      "step": 41800
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9388888888888887e-05,
+      "loss": 1.5096,
+      "step": 41810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9383838383838386e-05,
+      "loss": 1.5148,
+      "step": 41820
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.937878787878788e-05,
+      "loss": 1.5286,
+      "step": 41830
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9373737373737375e-05,
+      "loss": 1.517,
+      "step": 41840
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9368686868686867e-05,
+      "loss": 1.527,
+      "step": 41850
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9363636363636367e-05,
+      "loss": 1.5158,
+      "step": 41860
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9358585858585856e-05,
+      "loss": 1.5385,
+      "step": 41870
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9353535353535355e-05,
+      "loss": 1.5473,
+      "step": 41880
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9348484848484848e-05,
+      "loss": 1.5417,
+      "step": 41890
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9343434343434344e-05,
+      "loss": 1.5254,
+      "step": 41900
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9338383838383837e-05,
+      "loss": 1.5176,
+      "step": 41910
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9333333333333336e-05,
+      "loss": 1.5293,
+      "step": 41920
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.932828282828283e-05,
+      "loss": 1.5402,
+      "step": 41930
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9323232323232325e-05,
+      "loss": 1.5352,
+      "step": 41940
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9318181818181817e-05,
+      "loss": 1.5226,
+      "step": 41950
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9313131313131313e-05,
+      "loss": 1.5062,
+      "step": 41960
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9308080808080813e-05,
+      "loss": 1.5311,
+      "step": 41970
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9303030303030305e-05,
+      "loss": 1.5459,
+      "step": 41980
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.92979797979798e-05,
+      "loss": 1.522,
+      "step": 41990
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9292929292929294e-05,
+      "loss": 1.5478,
+      "step": 42000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9287878787878793e-05,
+      "loss": 1.5366,
+      "step": 42010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9282828282828286e-05,
+      "loss": 1.5232,
+      "step": 42020
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.927777777777778e-05,
+      "loss": 1.5404,
+      "step": 42030
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9272727272727274e-05,
+      "loss": 1.5472,
+      "step": 42040
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.926767676767677e-05,
+      "loss": 1.5168,
+      "step": 42050
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9262626262626263e-05,
+      "loss": 1.5451,
+      "step": 42060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9257575757575762e-05,
+      "loss": 1.5329,
+      "step": 42070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9252525252525255e-05,
+      "loss": 1.5333,
+      "step": 42080
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.924747474747475e-05,
+      "loss": 1.5182,
+      "step": 42090
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9242424242424243e-05,
+      "loss": 1.5214,
+      "step": 42100
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.923737373737374e-05,
+      "loss": 1.5035,
+      "step": 42110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9232323232323232e-05,
+      "loss": 1.5237,
+      "step": 42120
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.922727272727273e-05,
+      "loss": 1.5137,
+      "step": 42130
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9222222222222224e-05,
+      "loss": 1.5094,
+      "step": 42140
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.921717171717172e-05,
+      "loss": 1.562,
+      "step": 42150
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9212121212121213e-05,
+      "loss": 1.5262,
+      "step": 42160
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9207070707070712e-05,
+      "loss": 1.5036,
+      "step": 42170
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.92020202020202e-05,
+      "loss": 1.521,
+      "step": 42180
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.91969696969697e-05,
+      "loss": 1.5342,
+      "step": 42190
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9191919191919193e-05,
+      "loss": 1.5108,
+      "step": 42200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.918686868686869e-05,
+      "loss": 1.4947,
+      "step": 42210
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9181818181818182e-05,
+      "loss": 1.5335,
+      "step": 42220
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.917676767676768e-05,
+      "loss": 1.5429,
+      "step": 42230
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9171717171717174e-05,
+      "loss": 1.5323,
+      "step": 42240
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.916666666666667e-05,
+      "loss": 1.5339,
+      "step": 42250
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9161616161616162e-05,
+      "loss": 1.5285,
+      "step": 42260
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.915656565656566e-05,
+      "loss": 1.5514,
+      "step": 42270
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.915151515151515e-05,
+      "loss": 1.5329,
+      "step": 42280
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.914646464646465e-05,
+      "loss": 1.5256,
+      "step": 42290
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9141414141414143e-05,
+      "loss": 1.524,
+      "step": 42300
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.913636363636364e-05,
+      "loss": 1.5311,
+      "step": 42310
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.913131313131313e-05,
+      "loss": 1.5335,
+      "step": 42320
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.912626262626263e-05,
+      "loss": 1.5206,
+      "step": 42330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.912121212121212e-05,
+      "loss": 1.5403,
+      "step": 42340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.911616161616162e-05,
+      "loss": 1.5147,
+      "step": 42350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9111111111111112e-05,
+      "loss": 1.5327,
+      "step": 42360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9106060606060608e-05,
+      "loss": 1.536,
+      "step": 42370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.91010101010101e-05,
+      "loss": 1.5352,
+      "step": 42380
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.90959595959596e-05,
+      "loss": 1.5242,
+      "step": 42390
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.909090909090909e-05,
+      "loss": 1.526,
+      "step": 42400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.908585858585859e-05,
+      "loss": 1.5328,
+      "step": 42410
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.908080808080808e-05,
+      "loss": 1.542,
+      "step": 42420
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9075757575757577e-05,
+      "loss": 1.5217,
+      "step": 42430
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.907070707070707e-05,
+      "loss": 1.5169,
+      "step": 42440
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.906565656565657e-05,
+      "loss": 1.5176,
+      "step": 42450
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9060606060606062e-05,
+      "loss": 1.5301,
+      "step": 42460
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9055555555555558e-05,
+      "loss": 1.5334,
+      "step": 42470
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.905050505050505e-05,
+      "loss": 1.4987,
+      "step": 42480
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9045454545454546e-05,
+      "loss": 1.5252,
+      "step": 42490
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.904040404040404e-05,
+      "loss": 1.5442,
+      "step": 42500
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.903535353535354e-05,
+      "loss": 1.5276,
+      "step": 42510
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.903030303030303e-05,
+      "loss": 1.5146,
+      "step": 42520
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9025252525252527e-05,
+      "loss": 1.523,
+      "step": 42530
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.902020202020202e-05,
+      "loss": 1.5213,
+      "step": 42540
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.901515151515152e-05,
+      "loss": 1.5164,
+      "step": 42550
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9010101010101008e-05,
+      "loss": 1.5224,
+      "step": 42560
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9005050505050508e-05,
+      "loss": 1.5371,
+      "step": 42570
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9e-05,
+      "loss": 1.5053,
+      "step": 42580
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8994949494949496e-05,
+      "loss": 1.5177,
+      "step": 42590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.898989898989899e-05,
+      "loss": 1.5263,
+      "step": 42600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8984848484848488e-05,
+      "loss": 1.5476,
+      "step": 42610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.897979797979798e-05,
+      "loss": 1.5306,
+      "step": 42620
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8974747474747477e-05,
+      "loss": 1.5267,
+      "step": 42630
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.896969696969697e-05,
+      "loss": 1.5166,
+      "step": 42640
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8964646464646465e-05,
+      "loss": 1.5131,
+      "step": 42650
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8959595959595958e-05,
+      "loss": 1.513,
+      "step": 42660
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8954545454545457e-05,
+      "loss": 1.5158,
+      "step": 42670
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.894949494949495e-05,
+      "loss": 1.5061,
+      "step": 42680
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8944444444444446e-05,
+      "loss": 1.5107,
+      "step": 42690
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.893939393939394e-05,
+      "loss": 1.5065,
+      "step": 42700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8934343434343434e-05,
+      "loss": 1.5201,
+      "step": 42710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8929292929292927e-05,
+      "loss": 1.5126,
+      "step": 42720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8924242424242426e-05,
+      "loss": 1.5121,
+      "step": 42730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.891919191919192e-05,
+      "loss": 1.5452,
+      "step": 42740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8914141414141415e-05,
+      "loss": 1.5185,
+      "step": 42750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8909090909090908e-05,
+      "loss": 1.5207,
+      "step": 42760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8904040404040407e-05,
+      "loss": 1.5097,
+      "step": 42770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8898989898989896e-05,
+      "loss": 1.5466,
+      "step": 42780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8893939393939396e-05,
+      "loss": 1.5338,
+      "step": 42790
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8888888888888888e-05,
+      "loss": 1.5026,
+      "step": 42800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8883838383838384e-05,
+      "loss": 1.5136,
+      "step": 42810
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8878787878787884e-05,
+      "loss": 1.5321,
+      "step": 42820
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8873737373737376e-05,
+      "loss": 1.5117,
+      "step": 42830
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8868686868686872e-05,
+      "loss": 1.5359,
+      "step": 42840
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8863636363636365e-05,
+      "loss": 1.5446,
+      "step": 42850
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8858585858585864e-05,
+      "loss": 1.5093,
+      "step": 42860
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8853535353535353e-05,
+      "loss": 1.5141,
+      "step": 42870
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8848484848484853e-05,
+      "loss": 1.5269,
+      "step": 42880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8843434343434345e-05,
+      "loss": 1.5149,
+      "step": 42890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.883838383838384e-05,
+      "loss": 1.5249,
+      "step": 42900
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8833333333333334e-05,
+      "loss": 1.5264,
+      "step": 42910
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8828282828282833e-05,
+      "loss": 1.5242,
+      "step": 42920
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8823232323232323e-05,
+      "loss": 1.5155,
+      "step": 42930
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8818181818181822e-05,
+      "loss": 1.5305,
+      "step": 42940
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8813131313131315e-05,
+      "loss": 1.5258,
+      "step": 42950
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.880808080808081e-05,
+      "loss": 1.5175,
+      "step": 42960
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8803030303030303e-05,
+      "loss": 1.5253,
+      "step": 42970
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8797979797979803e-05,
+      "loss": 1.5257,
+      "step": 42980
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8792929292929295e-05,
+      "loss": 1.5161,
+      "step": 42990
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.878787878787879e-05,
+      "loss": 1.5172,
+      "step": 43000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8782828282828284e-05,
+      "loss": 1.5194,
+      "step": 43010
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.877777777777778e-05,
+      "loss": 1.5416,
+      "step": 43020
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8772727272727272e-05,
+      "loss": 1.5343,
+      "step": 43030
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.876767676767677e-05,
+      "loss": 1.5165,
+      "step": 43040
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8762626262626264e-05,
+      "loss": 1.5283,
+      "step": 43050
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.875757575757576e-05,
+      "loss": 1.5128,
+      "step": 43060
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8752525252525253e-05,
+      "loss": 1.5028,
+      "step": 43070
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8747474747474752e-05,
+      "loss": 1.5376,
+      "step": 43080
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.874242424242424e-05,
+      "loss": 1.5112,
+      "step": 43090
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.873737373737374e-05,
+      "loss": 1.5274,
+      "step": 43100
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8732323232323233e-05,
+      "loss": 1.5278,
+      "step": 43110
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.872727272727273e-05,
+      "loss": 1.5011,
+      "step": 43120
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8722222222222222e-05,
+      "loss": 1.5286,
+      "step": 43130
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.871717171717172e-05,
+      "loss": 1.5275,
+      "step": 43140
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8712121212121214e-05,
+      "loss": 1.5061,
+      "step": 43150
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.870707070707071e-05,
+      "loss": 1.5284,
+      "step": 43160
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8702020202020203e-05,
+      "loss": 1.5372,
+      "step": 43170
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.86969696969697e-05,
+      "loss": 1.5054,
+      "step": 43180
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.869191919191919e-05,
+      "loss": 1.5221,
+      "step": 43190
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.868686868686869e-05,
+      "loss": 1.5314,
+      "step": 43200
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8681818181818183e-05,
+      "loss": 1.5143,
+      "step": 43210
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.867676767676768e-05,
+      "loss": 1.5089,
+      "step": 43220
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8671717171717172e-05,
+      "loss": 1.5039,
+      "step": 43230
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8666666666666668e-05,
+      "loss": 1.5232,
+      "step": 43240
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.866161616161616e-05,
+      "loss": 1.532,
+      "step": 43250
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.865656565656566e-05,
+      "loss": 1.5463,
+      "step": 43260
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8651515151515152e-05,
+      "loss": 1.5178,
+      "step": 43270
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.864646464646465e-05,
+      "loss": 1.5331,
+      "step": 43280
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.864141414141414e-05,
+      "loss": 1.525,
+      "step": 43290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.863636363636364e-05,
+      "loss": 1.5405,
+      "step": 43300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.863131313131313e-05,
+      "loss": 1.5127,
+      "step": 43310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.862626262626263e-05,
+      "loss": 1.5266,
+      "step": 43320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.862121212121212e-05,
+      "loss": 1.5442,
+      "step": 43330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8616161616161618e-05,
+      "loss": 1.5325,
+      "step": 43340
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.861111111111111e-05,
+      "loss": 1.5498,
+      "step": 43350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.860606060606061e-05,
+      "loss": 1.5392,
+      "step": 43360
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8601010101010102e-05,
+      "loss": 1.5424,
+      "step": 43370
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8595959595959598e-05,
+      "loss": 1.5079,
+      "step": 43380
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.859090909090909e-05,
+      "loss": 1.5478,
+      "step": 43390
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8585858585858587e-05,
+      "loss": 1.5359,
+      "step": 43400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.858080808080808e-05,
+      "loss": 1.524,
+      "step": 43410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.857575757575758e-05,
+      "loss": 1.5341,
+      "step": 43420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.857070707070707e-05,
+      "loss": 1.535,
+      "step": 43430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8565656565656567e-05,
+      "loss": 1.5255,
+      "step": 43440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.856060606060606e-05,
+      "loss": 1.5363,
+      "step": 43450
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.855555555555556e-05,
+      "loss": 1.5375,
+      "step": 43460
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.855050505050505e-05,
+      "loss": 1.527,
+      "step": 43470
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8545454545454548e-05,
+      "loss": 1.5405,
+      "step": 43480
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.854040404040404e-05,
+      "loss": 1.5411,
+      "step": 43490
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8535353535353536e-05,
+      "loss": 1.5203,
+      "step": 43500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.853030303030303e-05,
+      "loss": 1.5302,
+      "step": 43510
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.852525252525253e-05,
+      "loss": 1.5192,
+      "step": 43520
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8520202020202018e-05,
+      "loss": 1.5172,
+      "step": 43530
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8515151515151517e-05,
+      "loss": 1.5317,
+      "step": 43540
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.851010101010101e-05,
+      "loss": 1.5141,
+      "step": 43550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8505050505050506e-05,
+      "loss": 1.5017,
+      "step": 43560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8499999999999998e-05,
+      "loss": 1.5247,
+      "step": 43570
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8494949494949498e-05,
+      "loss": 1.5165,
+      "step": 43580
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.848989898989899e-05,
+      "loss": 1.5323,
+      "step": 43590
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8484848484848486e-05,
+      "loss": 1.5126,
+      "step": 43600
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.847979797979798e-05,
+      "loss": 1.51,
+      "step": 43610
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8474747474747475e-05,
+      "loss": 1.508,
+      "step": 43620
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8469696969696967e-05,
+      "loss": 1.5179,
+      "step": 43630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8464646464646467e-05,
+      "loss": 1.5182,
+      "step": 43640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.845959595959596e-05,
+      "loss": 1.5446,
+      "step": 43650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8454545454545455e-05,
+      "loss": 1.5152,
+      "step": 43660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8449494949494948e-05,
+      "loss": 1.5164,
+      "step": 43670
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8444444444444447e-05,
+      "loss": 1.5345,
+      "step": 43680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8439393939393943e-05,
+      "loss": 1.5192,
+      "step": 43690
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8434343434343436e-05,
+      "loss": 1.5476,
+      "step": 43700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8429292929292932e-05,
+      "loss": 1.5183,
+      "step": 43710
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8424242424242424e-05,
+      "loss": 1.536,
+      "step": 43720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8419191919191924e-05,
+      "loss": 1.5281,
+      "step": 43730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8414141414141416e-05,
+      "loss": 1.4904,
+      "step": 43740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8409090909090912e-05,
+      "loss": 1.5077,
+      "step": 43750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8404040404040405e-05,
+      "loss": 1.5283,
+      "step": 43760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8398989898989904e-05,
+      "loss": 1.5133,
+      "step": 43770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8393939393939394e-05,
+      "loss": 1.523,
+      "step": 43780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8388888888888893e-05,
+      "loss": 1.5285,
+      "step": 43790
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8383838383838386e-05,
+      "loss": 1.5296,
+      "step": 43800
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.837878787878788e-05,
+      "loss": 1.5205,
+      "step": 43810
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8373737373737374e-05,
+      "loss": 1.5213,
+      "step": 43820
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8368686868686874e-05,
+      "loss": 1.5108,
+      "step": 43830
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8363636363636363e-05,
+      "loss": 1.4978,
+      "step": 43840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8358585858585862e-05,
+      "loss": 1.5343,
+      "step": 43850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8353535353535355e-05,
+      "loss": 1.5053,
+      "step": 43860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.834848484848485e-05,
+      "loss": 1.5113,
+      "step": 43870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8343434343434343e-05,
+      "loss": 1.5174,
+      "step": 43880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8338383838383843e-05,
+      "loss": 1.5041,
+      "step": 43890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8333333333333335e-05,
+      "loss": 1.5018,
+      "step": 43900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.832828282828283e-05,
+      "loss": 1.5228,
+      "step": 43910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8323232323232324e-05,
+      "loss": 1.534,
+      "step": 43920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.831818181818182e-05,
+      "loss": 1.5069,
+      "step": 43930
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8313131313131313e-05,
+      "loss": 1.5138,
+      "step": 43940
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8308080808080812e-05,
+      "loss": 1.5164,
+      "step": 43950
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8303030303030305e-05,
+      "loss": 1.5501,
+      "step": 43960
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.82979797979798e-05,
+      "loss": 1.527,
+      "step": 43970
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8292929292929293e-05,
+      "loss": 1.5339,
+      "step": 43980
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8287878787878793e-05,
+      "loss": 1.5345,
+      "step": 43990
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8282828282828282e-05,
+      "loss": 1.5302,
+      "step": 44000
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.827777777777778e-05,
+      "loss": 1.5234,
+      "step": 44010
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8272727272727274e-05,
+      "loss": 1.5273,
+      "step": 44020
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.826767676767677e-05,
+      "loss": 1.5163,
+      "step": 44030
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8262626262626262e-05,
+      "loss": 1.5095,
+      "step": 44040
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.825757575757576e-05,
+      "loss": 1.5491,
+      "step": 44050
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8252525252525254e-05,
+      "loss": 1.5266,
+      "step": 44060
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.824747474747475e-05,
+      "loss": 1.524,
+      "step": 44070
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8242424242424243e-05,
+      "loss": 1.5248,
+      "step": 44080
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.823737373737374e-05,
+      "loss": 1.5336,
+      "step": 44090
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.823232323232323e-05,
+      "loss": 1.5013,
+      "step": 44100
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.822727272727273e-05,
+      "loss": 1.5182,
+      "step": 44110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8222222222222223e-05,
+      "loss": 1.5096,
+      "step": 44120
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.821717171717172e-05,
+      "loss": 1.5322,
+      "step": 44130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8212121212121212e-05,
+      "loss": 1.531,
+      "step": 44140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8207070707070708e-05,
+      "loss": 1.5143,
+      "step": 44150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.82020202020202e-05,
+      "loss": 1.526,
+      "step": 44160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.81969696969697e-05,
+      "loss": 1.5148,
+      "step": 44170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.8191919191919193e-05,
+      "loss": 1.5307,
+      "step": 44180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.818686868686869e-05,
+      "loss": 1.5138,
+      "step": 44190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.818181818181818e-05,
+      "loss": 1.5232,
+      "step": 44200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.817676767676768e-05,
+      "loss": 1.5407,
+      "step": 44210
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.817171717171717e-05,
+      "loss": 1.5237,
+      "step": 44220
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.816666666666667e-05,
+      "loss": 1.5391,
+      "step": 44230
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8161616161616162e-05,
+      "loss": 1.535,
+      "step": 44240
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8156565656565658e-05,
+      "loss": 1.5134,
+      "step": 44250
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.815151515151515e-05,
+      "loss": 1.5136,
+      "step": 44260
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.814646464646465e-05,
+      "loss": 1.525,
+      "step": 44270
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8141414141414142e-05,
+      "loss": 1.513,
+      "step": 44280
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.813636363636364e-05,
+      "loss": 1.5561,
+      "step": 44290
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.813131313131313e-05,
+      "loss": 1.5079,
+      "step": 44300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8126262626262627e-05,
+      "loss": 1.5046,
+      "step": 44310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.812121212121212e-05,
+      "loss": 1.5382,
+      "step": 44320
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.811616161616162e-05,
+      "loss": 1.5201,
+      "step": 44330
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.811111111111111e-05,
+      "loss": 1.5414,
+      "step": 44340
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8106060606060607e-05,
+      "loss": 1.516,
+      "step": 44350
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.81010101010101e-05,
+      "loss": 1.5166,
+      "step": 44360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.80959595959596e-05,
+      "loss": 1.5066,
+      "step": 44370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.809090909090909e-05,
+      "loss": 1.5163,
+      "step": 44380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8085858585858588e-05,
+      "loss": 1.5404,
+      "step": 44390
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.808080808080808e-05,
+      "loss": 1.5127,
+      "step": 44400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8075757575757577e-05,
+      "loss": 1.5233,
+      "step": 44410
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.807070707070707e-05,
+      "loss": 1.5003,
+      "step": 44420
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.806565656565657e-05,
+      "loss": 1.5357,
+      "step": 44430
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8060606060606058e-05,
+      "loss": 1.5401,
+      "step": 44440
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8055555555555557e-05,
+      "loss": 1.528,
+      "step": 44450
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.805050505050505e-05,
+      "loss": 1.5329,
+      "step": 44460
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8045454545454546e-05,
+      "loss": 1.4999,
+      "step": 44470
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.804040404040404e-05,
+      "loss": 1.5097,
+      "step": 44480
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8035353535353538e-05,
+      "loss": 1.5177,
+      "step": 44490
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.803030303030303e-05,
+      "loss": 1.5395,
+      "step": 44500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8025252525252526e-05,
+      "loss": 1.515,
+      "step": 44510
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.802020202020202e-05,
+      "loss": 1.5055,
+      "step": 44520
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8015151515151515e-05,
+      "loss": 1.5206,
+      "step": 44530
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8010101010101014e-05,
+      "loss": 1.5131,
+      "step": 44540
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8005050505050507e-05,
+      "loss": 1.5129,
+      "step": 44550
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 1.5207,
+      "step": 44560
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7994949494949496e-05,
+      "loss": 1.5461,
+      "step": 44570
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7989898989898995e-05,
+      "loss": 1.5169,
+      "step": 44580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7984848484848488e-05,
+      "loss": 1.5011,
+      "step": 44590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7979797979797984e-05,
+      "loss": 1.5431,
+      "step": 44600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7974747474747476e-05,
+      "loss": 1.5395,
+      "step": 44610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7969696969696972e-05,
+      "loss": 1.5313,
+      "step": 44620
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7964646464646465e-05,
+      "loss": 1.5358,
+      "step": 44630
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7959595959595964e-05,
+      "loss": 1.5042,
+      "step": 44640
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7954545454545457e-05,
+      "loss": 1.5324,
+      "step": 44650
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7949494949494953e-05,
+      "loss": 1.4989,
+      "step": 44660
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7944444444444445e-05,
+      "loss": 1.5039,
+      "step": 44670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7939393939393945e-05,
+      "loss": 1.5423,
+      "step": 44680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7934343434343434e-05,
+      "loss": 1.5072,
+      "step": 44690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7929292929292933e-05,
+      "loss": 1.5159,
+      "step": 44700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7924242424242426e-05,
+      "loss": 1.5162,
+      "step": 44710
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7919191919191922e-05,
+      "loss": 1.5061,
+      "step": 44720
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7914141414141414e-05,
+      "loss": 1.5021,
+      "step": 44730
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7909090909090914e-05,
+      "loss": 1.5256,
+      "step": 44740
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7904040404040403e-05,
+      "loss": 1.5415,
+      "step": 44750
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7898989898989902e-05,
+      "loss": 1.5142,
+      "step": 44760
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7893939393939395e-05,
+      "loss": 1.5414,
+      "step": 44770
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.788888888888889e-05,
+      "loss": 1.4999,
+      "step": 44780
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7883838383838384e-05,
+      "loss": 1.5247,
+      "step": 44790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7878787878787883e-05,
+      "loss": 1.5269,
+      "step": 44800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7873737373737376e-05,
+      "loss": 1.5317,
+      "step": 44810
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.786868686868687e-05,
+      "loss": 1.5063,
+      "step": 44820
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7863636363636364e-05,
+      "loss": 1.5187,
+      "step": 44830
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.785858585858586e-05,
+      "loss": 1.5227,
+      "step": 44840
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7853535353535353e-05,
+      "loss": 1.5515,
+      "step": 44850
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7848484848484852e-05,
+      "loss": 1.5226,
+      "step": 44860
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7843434343434345e-05,
+      "loss": 1.5441,
+      "step": 44870
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.783838383838384e-05,
+      "loss": 1.5098,
+      "step": 44880
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7833333333333333e-05,
+      "loss": 1.5397,
+      "step": 44890
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.7828282828282833e-05,
+      "loss": 1.5186,
+      "step": 44900
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7823232323232322e-05,
+      "loss": 1.5339,
+      "step": 44910
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.781818181818182e-05,
+      "loss": 1.5216,
+      "step": 44920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7813131313131314e-05,
+      "loss": 1.5142,
+      "step": 44930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.780808080808081e-05,
+      "loss": 1.5221,
+      "step": 44940
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7803030303030303e-05,
+      "loss": 1.5275,
+      "step": 44950
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7797979797979802e-05,
+      "loss": 1.5193,
+      "step": 44960
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.779292929292929e-05,
+      "loss": 1.504,
+      "step": 44970
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.778787878787879e-05,
+      "loss": 1.5056,
+      "step": 44980
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7782828282828283e-05,
+      "loss": 1.5119,
+      "step": 44990
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.777777777777778e-05,
+      "loss": 1.5176,
+      "step": 45000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7772727272727272e-05,
+      "loss": 1.5133,
+      "step": 45010
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.776767676767677e-05,
+      "loss": 1.5128,
+      "step": 45020
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7762626262626264e-05,
+      "loss": 1.5044,
+      "step": 45030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.775757575757576e-05,
+      "loss": 1.5198,
+      "step": 45040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7752525252525252e-05,
+      "loss": 1.5172,
+      "step": 45050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7747474747474748e-05,
+      "loss": 1.5268,
+      "step": 45060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.774242424242424e-05,
+      "loss": 1.5142,
+      "step": 45070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.773737373737374e-05,
+      "loss": 1.5124,
+      "step": 45080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7732323232323233e-05,
+      "loss": 1.5403,
+      "step": 45090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.772727272727273e-05,
+      "loss": 1.5179,
+      "step": 45100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.772222222222222e-05,
+      "loss": 1.5211,
+      "step": 45110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.771717171717172e-05,
+      "loss": 1.5072,
+      "step": 45120
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.771212121212121e-05,
+      "loss": 1.5104,
+      "step": 45130
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.770707070707071e-05,
+      "loss": 1.5146,
+      "step": 45140
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7702020202020202e-05,
+      "loss": 1.5299,
+      "step": 45150
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7696969696969698e-05,
+      "loss": 1.5256,
+      "step": 45160
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.769191919191919e-05,
+      "loss": 1.525,
+      "step": 45170
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.768686868686869e-05,
+      "loss": 1.5106,
+      "step": 45180
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7681818181818183e-05,
+      "loss": 1.5277,
+      "step": 45190
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.767676767676768e-05,
+      "loss": 1.5181,
+      "step": 45200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.767171717171717e-05,
+      "loss": 1.5177,
+      "step": 45210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7666666666666667e-05,
+      "loss": 1.5067,
+      "step": 45220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.766161616161616e-05,
+      "loss": 1.5375,
+      "step": 45230
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.765656565656566e-05,
+      "loss": 1.5195,
+      "step": 45240
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7651515151515152e-05,
+      "loss": 1.5209,
+      "step": 45250
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7646464646464648e-05,
+      "loss": 1.5146,
+      "step": 45260
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.764141414141414e-05,
+      "loss": 1.5337,
+      "step": 45270
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7636363636363636e-05,
+      "loss": 1.4939,
+      "step": 45280
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.763131313131313e-05,
+      "loss": 1.5203,
+      "step": 45290
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.762626262626263e-05,
+      "loss": 1.5251,
+      "step": 45300
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.762121212121212e-05,
+      "loss": 1.5243,
+      "step": 45310
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7616161616161617e-05,
+      "loss": 1.5126,
+      "step": 45320
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.761111111111111e-05,
+      "loss": 1.5339,
+      "step": 45330
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.760606060606061e-05,
+      "loss": 1.5492,
+      "step": 45340
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7601010101010098e-05,
+      "loss": 1.5143,
+      "step": 45350
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7595959595959597e-05,
+      "loss": 1.5234,
+      "step": 45360
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.759090909090909e-05,
+      "loss": 1.5238,
+      "step": 45370
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7585858585858586e-05,
+      "loss": 1.5202,
+      "step": 45380
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7580808080808085e-05,
+      "loss": 1.5075,
+      "step": 45390
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7575757575757578e-05,
+      "loss": 1.5142,
+      "step": 45400
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7570707070707074e-05,
+      "loss": 1.5005,
+      "step": 45410
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7565656565656567e-05,
+      "loss": 1.5175,
+      "step": 45420
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7560606060606066e-05,
+      "loss": 1.4952,
+      "step": 45430
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7555555555555555e-05,
+      "loss": 1.5282,
+      "step": 45440
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7550505050505055e-05,
+      "loss": 1.5213,
+      "step": 45450
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7545454545454547e-05,
+      "loss": 1.5245,
+      "step": 45460
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7540404040404043e-05,
+      "loss": 1.5083,
+      "step": 45470
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7535353535353536e-05,
+      "loss": 1.4903,
+      "step": 45480
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7530303030303035e-05,
+      "loss": 1.5285,
+      "step": 45490
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7525252525252528e-05,
+      "loss": 1.4978,
+      "step": 45500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7520202020202024e-05,
+      "loss": 1.5316,
+      "step": 45510
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7515151515151516e-05,
+      "loss": 1.514,
+      "step": 45520
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7510101010101012e-05,
+      "loss": 1.509,
+      "step": 45530
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7505050505050505e-05,
+      "loss": 1.5149,
+      "step": 45540
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7500000000000004e-05,
+      "loss": 1.5066,
+      "step": 45550
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.7494949494949497e-05,
+      "loss": 1.5409,
+      "step": 45560
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7489898989898993e-05,
+      "loss": 1.5576,
+      "step": 45570
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7484848484848486e-05,
+      "loss": 1.5279,
+      "step": 45580
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.747979797979798e-05,
+      "loss": 1.5242,
+      "step": 45590
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7474747474747474e-05,
+      "loss": 1.5057,
+      "step": 45600
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7469696969696974e-05,
+      "loss": 1.5364,
+      "step": 45610
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7464646464646466e-05,
+      "loss": 1.549,
+      "step": 45620
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7459595959595962e-05,
+      "loss": 1.5011,
+      "step": 45630
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7454545454545455e-05,
+      "loss": 1.5232,
+      "step": 45640
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7449494949494954e-05,
+      "loss": 1.4953,
+      "step": 45650
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7444444444444443e-05,
+      "loss": 1.5039,
+      "step": 45660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7439393939393943e-05,
+      "loss": 1.5265,
+      "step": 45670
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7434343434343435e-05,
+      "loss": 1.5433,
+      "step": 45680
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.742929292929293e-05,
+      "loss": 1.5466,
+      "step": 45690
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7424242424242424e-05,
+      "loss": 1.5488,
+      "step": 45700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7419191919191923e-05,
+      "loss": 1.5083,
+      "step": 45710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7414141414141416e-05,
+      "loss": 1.5256,
+      "step": 45720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7409090909090912e-05,
+      "loss": 1.5096,
+      "step": 45730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7404040404040404e-05,
+      "loss": 1.5235,
+      "step": 45740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.73989898989899e-05,
+      "loss": 1.5013,
+      "step": 45750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7393939393939393e-05,
+      "loss": 1.5261,
+      "step": 45760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7388888888888892e-05,
+      "loss": 1.4908,
+      "step": 45770
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7383838383838385e-05,
+      "loss": 1.5364,
+      "step": 45780
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.737878787878788e-05,
+      "loss": 1.5288,
+      "step": 45790
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7373737373737374e-05,
+      "loss": 1.5287,
+      "step": 45800
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7368686868686873e-05,
+      "loss": 1.531,
+      "step": 45810
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7363636363636362e-05,
+      "loss": 1.5104,
+      "step": 45820
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.735858585858586e-05,
+      "loss": 1.5215,
+      "step": 45830
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7353535353535354e-05,
+      "loss": 1.5201,
+      "step": 45840
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.734848484848485e-05,
+      "loss": 1.5192,
+      "step": 45850
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7343434343434343e-05,
+      "loss": 1.502,
+      "step": 45860
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7338383838383842e-05,
+      "loss": 1.5392,
+      "step": 45870
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.733333333333333e-05,
+      "loss": 1.5337,
+      "step": 45880
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.732828282828283e-05,
+      "loss": 1.5158,
+      "step": 45890
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7323232323232323e-05,
+      "loss": 1.5207,
+      "step": 45900
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.731818181818182e-05,
+      "loss": 1.509,
+      "step": 45910
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7313131313131312e-05,
+      "loss": 1.5382,
+      "step": 45920
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.730808080808081e-05,
+      "loss": 1.5036,
+      "step": 45930
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7303030303030304e-05,
+      "loss": 1.5184,
+      "step": 45940
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.72979797979798e-05,
+      "loss": 1.5155,
+      "step": 45950
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7292929292929293e-05,
+      "loss": 1.5109,
+      "step": 45960
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.728787878787879e-05,
+      "loss": 1.5447,
+      "step": 45970
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.728282828282828e-05,
+      "loss": 1.5187,
+      "step": 45980
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.727777777777778e-05,
+      "loss": 1.5502,
+      "step": 45990
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7272727272727273e-05,
+      "loss": 1.5073,
+      "step": 46000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.726767676767677e-05,
+      "loss": 1.5235,
+      "step": 46010
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7262626262626262e-05,
+      "loss": 1.529,
+      "step": 46020
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.725757575757576e-05,
+      "loss": 1.5182,
+      "step": 46030
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.725252525252525e-05,
+      "loss": 1.5104,
+      "step": 46040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.724747474747475e-05,
+      "loss": 1.5052,
+      "step": 46050
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7242424242424242e-05,
+      "loss": 1.5239,
+      "step": 46060
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7237373737373738e-05,
+      "loss": 1.485,
+      "step": 46070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.723232323232323e-05,
+      "loss": 1.5151,
+      "step": 46080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.722727272727273e-05,
+      "loss": 1.5405,
+      "step": 46090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7222222222222223e-05,
+      "loss": 1.5319,
+      "step": 46100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.721717171717172e-05,
+      "loss": 1.5189,
+      "step": 46110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.721212121212121e-05,
+      "loss": 1.5194,
+      "step": 46120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7207070707070707e-05,
+      "loss": 1.5146,
+      "step": 46130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.72020202020202e-05,
+      "loss": 1.5162,
+      "step": 46140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.71969696969697e-05,
+      "loss": 1.5168,
+      "step": 46150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7191919191919192e-05,
+      "loss": 1.5076,
+      "step": 46160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7186868686868688e-05,
+      "loss": 1.5343,
+      "step": 46170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.718181818181818e-05,
+      "loss": 1.5119,
+      "step": 46180
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7176767676767677e-05,
+      "loss": 1.5211,
+      "step": 46190
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.717171717171717e-05,
+      "loss": 1.544,
+      "step": 46200
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.716666666666667e-05,
+      "loss": 1.5305,
+      "step": 46210
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.716161616161616e-05,
+      "loss": 1.5134,
+      "step": 46220
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.7156565656565657e-05,
+      "loss": 1.5363,
+      "step": 46230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7151515151515157e-05,
+      "loss": 1.5068,
+      "step": 46240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.714646464646465e-05,
+      "loss": 1.512,
+      "step": 46250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7141414141414145e-05,
+      "loss": 1.5076,
+      "step": 46260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7136363636363638e-05,
+      "loss": 1.5199,
+      "step": 46270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7131313131313134e-05,
+      "loss": 1.5339,
+      "step": 46280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7126262626262626e-05,
+      "loss": 1.5248,
+      "step": 46290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7121212121212126e-05,
+      "loss": 1.5319,
+      "step": 46300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.711616161616162e-05,
+      "loss": 1.5237,
+      "step": 46310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7111111111111114e-05,
+      "loss": 1.5353,
+      "step": 46320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7106060606060607e-05,
+      "loss": 1.5107,
+      "step": 46330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7101010101010106e-05,
+      "loss": 1.5205,
+      "step": 46340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7095959595959596e-05,
+      "loss": 1.4975,
+      "step": 46350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7090909090909095e-05,
+      "loss": 1.5321,
+      "step": 46360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7085858585858587e-05,
+      "loss": 1.5086,
+      "step": 46370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7080808080808083e-05,
+      "loss": 1.5438,
+      "step": 46380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7075757575757576e-05,
+      "loss": 1.5182,
+      "step": 46390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7070707070707075e-05,
+      "loss": 1.5398,
+      "step": 46400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7065656565656568e-05,
+      "loss": 1.5144,
+      "step": 46410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7060606060606064e-05,
+      "loss": 1.5176,
+      "step": 46420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7055555555555557e-05,
+      "loss": 1.5116,
+      "step": 46430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7050505050505053e-05,
+      "loss": 1.5143,
+      "step": 46440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7045454545454545e-05,
+      "loss": 1.5243,
+      "step": 46450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7040404040404045e-05,
+      "loss": 1.5275,
+      "step": 46460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7035353535353537e-05,
+      "loss": 1.5135,
+      "step": 46470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7030303030303033e-05,
+      "loss": 1.5259,
+      "step": 46480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7025252525252526e-05,
+      "loss": 1.5296,
+      "step": 46490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7020202020202022e-05,
+      "loss": 1.5306,
+      "step": 46500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7015151515151514e-05,
+      "loss": 1.5304,
+      "step": 46510
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7010101010101014e-05,
+      "loss": 1.5294,
+      "step": 46520
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7005050505050506e-05,
+      "loss": 1.4998,
+      "step": 46530
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.7000000000000002e-05,
+      "loss": 1.5251,
+      "step": 46540
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6994949494949495e-05,
+      "loss": 1.5406,
+      "step": 46550
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6989898989898994e-05,
+      "loss": 1.511,
+      "step": 46560
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6984848484848484e-05,
+      "loss": 1.512,
+      "step": 46570
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6979797979797983e-05,
+      "loss": 1.5156,
+      "step": 46580
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6974747474747476e-05,
+      "loss": 1.5231,
+      "step": 46590
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.696969696969697e-05,
+      "loss": 1.5102,
+      "step": 46600
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6964646464646464e-05,
+      "loss": 1.5218,
+      "step": 46610
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6959595959595964e-05,
+      "loss": 1.5242,
+      "step": 46620
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6954545454545456e-05,
+      "loss": 1.5216,
+      "step": 46630
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6949494949494952e-05,
+      "loss": 1.533,
+      "step": 46640
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6944444444444445e-05,
+      "loss": 1.5233,
+      "step": 46650
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.693939393939394e-05,
+      "loss": 1.5082,
+      "step": 46660
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6934343434343433e-05,
+      "loss": 1.5082,
+      "step": 46670
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6929292929292933e-05,
+      "loss": 1.519,
+      "step": 46680
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6924242424242425e-05,
+      "loss": 1.5242,
+      "step": 46690
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.691919191919192e-05,
+      "loss": 1.5293,
+      "step": 46700
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6914141414141414e-05,
+      "loss": 1.5297,
+      "step": 46710
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6909090909090913e-05,
+      "loss": 1.5374,
+      "step": 46720
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6904040404040402e-05,
+      "loss": 1.4889,
+      "step": 46730
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6898989898989902e-05,
+      "loss": 1.5239,
+      "step": 46740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6893939393939394e-05,
+      "loss": 1.5105,
+      "step": 46750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.688888888888889e-05,
+      "loss": 1.5133,
+      "step": 46760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6883838383838383e-05,
+      "loss": 1.499,
+      "step": 46770
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6878787878787882e-05,
+      "loss": 1.5292,
+      "step": 46780
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.687373737373737e-05,
+      "loss": 1.5194,
+      "step": 46790
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.686868686868687e-05,
+      "loss": 1.5091,
+      "step": 46800
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6863636363636364e-05,
+      "loss": 1.5194,
+      "step": 46810
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.685858585858586e-05,
+      "loss": 1.5374,
+      "step": 46820
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6853535353535352e-05,
+      "loss": 1.5024,
+      "step": 46830
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.684848484848485e-05,
+      "loss": 1.4983,
+      "step": 46840
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6843434343434344e-05,
+      "loss": 1.5281,
+      "step": 46850
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.683838383838384e-05,
+      "loss": 1.5021,
+      "step": 46860
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.6833333333333333e-05,
+      "loss": 1.5356,
+      "step": 46870
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.682828282828283e-05,
+      "loss": 1.5071,
+      "step": 46880
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.682323232323232e-05,
+      "loss": 1.5092,
+      "step": 46890
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.681818181818182e-05,
+      "loss": 1.5115,
+      "step": 46900
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6813131313131313e-05,
+      "loss": 1.5299,
+      "step": 46910
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.680808080808081e-05,
+      "loss": 1.5356,
+      "step": 46920
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6803030303030302e-05,
+      "loss": 1.5244,
+      "step": 46930
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.67979797979798e-05,
+      "loss": 1.5139,
+      "step": 46940
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.679292929292929e-05,
+      "loss": 1.5207,
+      "step": 46950
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.678787878787879e-05,
+      "loss": 1.5411,
+      "step": 46960
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6782828282828283e-05,
+      "loss": 1.5416,
+      "step": 46970
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.677777777777778e-05,
+      "loss": 1.5291,
+      "step": 46980
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.677272727272727e-05,
+      "loss": 1.5258,
+      "step": 46990
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.676767676767677e-05,
+      "loss": 1.5058,
+      "step": 47000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.676262626262626e-05,
+      "loss": 1.5258,
+      "step": 47010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.675757575757576e-05,
+      "loss": 1.5136,
+      "step": 47020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6752525252525252e-05,
+      "loss": 1.523,
+      "step": 47030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6747474747474748e-05,
+      "loss": 1.5264,
+      "step": 47040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.674242424242424e-05,
+      "loss": 1.5116,
+      "step": 47050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.673737373737374e-05,
+      "loss": 1.5213,
+      "step": 47060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6732323232323232e-05,
+      "loss": 1.5247,
+      "step": 47070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6727272727272728e-05,
+      "loss": 1.5146,
+      "step": 47080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6722222222222228e-05,
+      "loss": 1.5168,
+      "step": 47090
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6717171717171717e-05,
+      "loss": 1.5138,
+      "step": 47100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6712121212121216e-05,
+      "loss": 1.5168,
+      "step": 47110
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.670707070707071e-05,
+      "loss": 1.5246,
+      "step": 47120
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6702020202020205e-05,
+      "loss": 1.5252,
+      "step": 47130
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6696969696969697e-05,
+      "loss": 1.4908,
+      "step": 47140
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6691919191919197e-05,
+      "loss": 1.5106,
+      "step": 47150
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.668686868686869e-05,
+      "loss": 1.5192,
+      "step": 47160
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6681818181818185e-05,
+      "loss": 1.5262,
+      "step": 47170
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6676767676767678e-05,
+      "loss": 1.5157,
+      "step": 47180
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6671717171717174e-05,
+      "loss": 1.5149,
+      "step": 47190
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6666666666666667e-05,
+      "loss": 1.5064,
+      "step": 47200
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6661616161616166e-05,
+      "loss": 1.5187,
+      "step": 47210
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.665656565656566e-05,
+      "loss": 1.4997,
+      "step": 47220
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6651515151515155e-05,
+      "loss": 1.4991,
+      "step": 47230
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6646464646464647e-05,
+      "loss": 1.5405,
+      "step": 47240
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6641414141414147e-05,
+      "loss": 1.5229,
+      "step": 47250
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6636363636363636e-05,
+      "loss": 1.5112,
+      "step": 47260
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6631313131313135e-05,
+      "loss": 1.5328,
+      "step": 47270
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6626262626262628e-05,
+      "loss": 1.52,
+      "step": 47280
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6621212121212124e-05,
+      "loss": 1.5152,
+      "step": 47290
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6616161616161616e-05,
+      "loss": 1.5112,
+      "step": 47300
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6611111111111116e-05,
+      "loss": 1.5258,
+      "step": 47310
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6606060606060605e-05,
+      "loss": 1.5223,
+      "step": 47320
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6601010101010104e-05,
+      "loss": 1.5346,
+      "step": 47330
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6595959595959597e-05,
+      "loss": 1.5201,
+      "step": 47340
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6590909090909093e-05,
+      "loss": 1.5256,
+      "step": 47350
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6585858585858585e-05,
+      "loss": 1.5128,
+      "step": 47360
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6580808080808085e-05,
+      "loss": 1.5115,
+      "step": 47370
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6575757575757577e-05,
+      "loss": 1.5108,
+      "step": 47380
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6570707070707073e-05,
+      "loss": 1.5425,
+      "step": 47390
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6565656565656566e-05,
+      "loss": 1.4912,
+      "step": 47400
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6560606060606062e-05,
+      "loss": 1.508,
+      "step": 47410
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6555555555555555e-05,
+      "loss": 1.5225,
+      "step": 47420
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6550505050505054e-05,
+      "loss": 1.5183,
+      "step": 47430
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6545454545454547e-05,
+      "loss": 1.541,
+      "step": 47440
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6540404040404043e-05,
+      "loss": 1.5396,
+      "step": 47450
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6535353535353535e-05,
+      "loss": 1.5407,
+      "step": 47460
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6530303030303035e-05,
+      "loss": 1.5363,
+      "step": 47470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6525252525252524e-05,
+      "loss": 1.5189,
+      "step": 47480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6520202020202023e-05,
+      "loss": 1.5197,
+      "step": 47490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6515151515151516e-05,
+      "loss": 1.5022,
+      "step": 47500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6510101010101012e-05,
+      "loss": 1.5,
+      "step": 47510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6505050505050504e-05,
+      "loss": 1.514,
+      "step": 47520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6500000000000004e-05,
+      "loss": 1.496,
+      "step": 47530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6494949494949496e-05,
+      "loss": 1.5312,
+      "step": 47540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6489898989898992e-05,
+      "loss": 1.5495,
+      "step": 47550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.6484848484848485e-05,
+      "loss": 1.5061,
+      "step": 47560
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.647979797979798e-05,
+      "loss": 1.5151,
+      "step": 47570
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6474747474747474e-05,
+      "loss": 1.5193,
+      "step": 47580
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6469696969696973e-05,
+      "loss": 1.5051,
+      "step": 47590
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6464646464646466e-05,
+      "loss": 1.5135,
+      "step": 47600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.645959595959596e-05,
+      "loss": 1.5297,
+      "step": 47610
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6454545454545454e-05,
+      "loss": 1.5218,
+      "step": 47620
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.644949494949495e-05,
+      "loss": 1.5052,
+      "step": 47630
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6444444444444443e-05,
+      "loss": 1.5218,
+      "step": 47640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6439393939393942e-05,
+      "loss": 1.4992,
+      "step": 47650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6434343434343435e-05,
+      "loss": 1.5202,
+      "step": 47660
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.642929292929293e-05,
+      "loss": 1.5189,
+      "step": 47670
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6424242424242423e-05,
+      "loss": 1.5095,
+      "step": 47680
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6419191919191923e-05,
+      "loss": 1.5098,
+      "step": 47690
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6414141414141412e-05,
+      "loss": 1.5093,
+      "step": 47700
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.640909090909091e-05,
+      "loss": 1.5471,
+      "step": 47710
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6404040404040404e-05,
+      "loss": 1.5149,
+      "step": 47720
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.63989898989899e-05,
+      "loss": 1.5137,
+      "step": 47730
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6393939393939392e-05,
+      "loss": 1.4998,
+      "step": 47740
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6388888888888892e-05,
+      "loss": 1.5153,
+      "step": 47750
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6383838383838384e-05,
+      "loss": 1.5012,
+      "step": 47760
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.637878787878788e-05,
+      "loss": 1.5373,
+      "step": 47770
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6373737373737373e-05,
+      "loss": 1.5027,
+      "step": 47780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.636868686868687e-05,
+      "loss": 1.5158,
+      "step": 47790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.636363636363636e-05,
+      "loss": 1.5184,
+      "step": 47800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.635858585858586e-05,
+      "loss": 1.5235,
+      "step": 47810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6353535353535354e-05,
+      "loss": 1.5273,
+      "step": 47820
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.634848484848485e-05,
+      "loss": 1.5238,
+      "step": 47830
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6343434343434342e-05,
+      "loss": 1.5026,
+      "step": 47840
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.633838383838384e-05,
+      "loss": 1.5051,
+      "step": 47850
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.633333333333333e-05,
+      "loss": 1.4905,
+      "step": 47860
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.632828282828283e-05,
+      "loss": 1.5422,
+      "step": 47870
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6323232323232323e-05,
+      "loss": 1.507,
+      "step": 47880
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.631818181818182e-05,
+      "loss": 1.5161,
+      "step": 47890
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.631313131313131e-05,
+      "loss": 1.5062,
+      "step": 47900
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.630808080808081e-05,
+      "loss": 1.5086,
+      "step": 47910
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.63030303030303e-05,
+      "loss": 1.5213,
+      "step": 47920
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.62979797979798e-05,
+      "loss": 1.4916,
+      "step": 47930
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6292929292929295e-05,
+      "loss": 1.5291,
+      "step": 47940
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6287878787878788e-05,
+      "loss": 1.5024,
+      "step": 47950
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6282828282828287e-05,
+      "loss": 1.5027,
+      "step": 47960
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.627777777777778e-05,
+      "loss": 1.5185,
+      "step": 47970
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6272727272727276e-05,
+      "loss": 1.5,
+      "step": 47980
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.626767676767677e-05,
+      "loss": 1.5183,
+      "step": 47990
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6262626262626268e-05,
+      "loss": 1.4913,
+      "step": 48000
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6257575757575757e-05,
+      "loss": 1.5395,
+      "step": 48010
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6252525252525256e-05,
+      "loss": 1.5212,
+      "step": 48020
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.624747474747475e-05,
+      "loss": 1.5182,
+      "step": 48030
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6242424242424245e-05,
+      "loss": 1.4921,
+      "step": 48040
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6237373737373738e-05,
+      "loss": 1.4995,
+      "step": 48050
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6232323232323237e-05,
+      "loss": 1.5137,
+      "step": 48060
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.622727272727273e-05,
+      "loss": 1.5188,
+      "step": 48070
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6222222222222226e-05,
+      "loss": 1.52,
+      "step": 48080
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6217171717171718e-05,
+      "loss": 1.5134,
+      "step": 48090
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6212121212121214e-05,
+      "loss": 1.5125,
+      "step": 48100
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6207070707070707e-05,
+      "loss": 1.4979,
+      "step": 48110
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6202020202020206e-05,
+      "loss": 1.5172,
+      "step": 48120
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.61969696969697e-05,
+      "loss": 1.5183,
+      "step": 48130
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6191919191919195e-05,
+      "loss": 1.5108,
+      "step": 48140
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6186868686868687e-05,
+      "loss": 1.5271,
+      "step": 48150
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6181818181818187e-05,
+      "loss": 1.5181,
+      "step": 48160
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6176767676767676e-05,
+      "loss": 1.5375,
+      "step": 48170
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6171717171717175e-05,
+      "loss": 1.5148,
+      "step": 48180
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6166666666666668e-05,
+      "loss": 1.511,
+      "step": 48190
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6161616161616164e-05,
+      "loss": 1.5164,
+      "step": 48200
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6156565656565657e-05,
+      "loss": 1.5072,
+      "step": 48210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2.6151515151515156e-05,
+      "loss": 1.5369,
+      "step": 48220
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6146464646464645e-05,
+      "loss": 1.5187,
+      "step": 48230
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6141414141414145e-05,
+      "loss": 1.5363,
+      "step": 48240
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6136363636363637e-05,
+      "loss": 1.5214,
+      "step": 48250
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6131313131313133e-05,
+      "loss": 1.4988,
+      "step": 48260
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6126262626262626e-05,
+      "loss": 1.5559,
+      "step": 48270
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6121212121212125e-05,
+      "loss": 1.5166,
+      "step": 48280
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6116161616161618e-05,
+      "loss": 1.5211,
+      "step": 48290
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6111111111111114e-05,
+      "loss": 1.5009,
+      "step": 48300
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6106060606060606e-05,
+      "loss": 1.5068,
+      "step": 48310
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6101010101010102e-05,
+      "loss": 1.5425,
+      "step": 48320
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6095959595959595e-05,
+      "loss": 1.5227,
+      "step": 48330
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6090909090909094e-05,
+      "loss": 1.5084,
+      "step": 48340
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6085858585858587e-05,
+      "loss": 1.4982,
+      "step": 48350
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6080808080808083e-05,
+      "loss": 1.5382,
+      "step": 48360
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6075757575757575e-05,
+      "loss": 1.5168,
+      "step": 48370
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6070707070707075e-05,
+      "loss": 1.5091,
+      "step": 48380
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6065656565656564e-05,
+      "loss": 1.5172,
+      "step": 48390
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6060606060606063e-05,
+      "loss": 1.5003,
+      "step": 48400
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6055555555555556e-05,
+      "loss": 1.5071,
+      "step": 48410
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6050505050505052e-05,
+      "loss": 1.5131,
+      "step": 48420
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6045454545454545e-05,
+      "loss": 1.5243,
+      "step": 48430
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6040404040404044e-05,
+      "loss": 1.5291,
+      "step": 48440
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6035353535353537e-05,
+      "loss": 1.531,
+      "step": 48450
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6030303030303033e-05,
+      "loss": 1.5127,
+      "step": 48460
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6025252525252525e-05,
+      "loss": 1.5235,
+      "step": 48470
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.602020202020202e-05,
+      "loss": 1.5108,
+      "step": 48480
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6015151515151514e-05,
+      "loss": 1.5309,
+      "step": 48490
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6010101010101013e-05,
+      "loss": 1.5265,
+      "step": 48500
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6005050505050506e-05,
+      "loss": 1.5031,
+      "step": 48510
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.6000000000000002e-05,
+      "loss": 1.5285,
+      "step": 48520
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5994949494949494e-05,
+      "loss": 1.5228,
+      "step": 48530
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.598989898989899e-05,
+      "loss": 1.5423,
+      "step": 48540
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5984848484848483e-05,
+      "loss": 1.5045,
+      "step": 48550
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5979797979797982e-05,
+      "loss": 1.5328,
+      "step": 48560
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5974747474747475e-05,
+      "loss": 1.509,
+      "step": 48570
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.596969696969697e-05,
+      "loss": 1.5202,
+      "step": 48580
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5964646464646464e-05,
+      "loss": 1.5079,
+      "step": 48590
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5959595959595963e-05,
+      "loss": 1.5366,
+      "step": 48600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5954545454545452e-05,
+      "loss": 1.4988,
+      "step": 48610
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.594949494949495e-05,
+      "loss": 1.5322,
+      "step": 48620
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5944444444444444e-05,
+      "loss": 1.507,
+      "step": 48630
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.593939393939394e-05,
+      "loss": 1.4947,
+      "step": 48640
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5934343434343433e-05,
+      "loss": 1.518,
+      "step": 48650
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5929292929292932e-05,
+      "loss": 1.4996,
+      "step": 48660
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5924242424242425e-05,
+      "loss": 1.5393,
+      "step": 48670
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.591919191919192e-05,
+      "loss": 1.5339,
+      "step": 48680
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5914141414141413e-05,
+      "loss": 1.5269,
+      "step": 48690
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.590909090909091e-05,
+      "loss": 1.5159,
+      "step": 48700
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5904040404040402e-05,
+      "loss": 1.5149,
+      "step": 48710
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.58989898989899e-05,
+      "loss": 1.5156,
+      "step": 48720
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5893939393939394e-05,
+      "loss": 1.5174,
+      "step": 48730
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.588888888888889e-05,
+      "loss": 1.5146,
+      "step": 48740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5883838383838382e-05,
+      "loss": 1.5155,
+      "step": 48750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.587878787878788e-05,
+      "loss": 1.4977,
+      "step": 48760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.587373737373737e-05,
+      "loss": 1.5074,
+      "step": 48770
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.586868686868687e-05,
+      "loss": 1.5352,
+      "step": 48780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5863636363636363e-05,
+      "loss": 1.5053,
+      "step": 48790
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.585858585858586e-05,
+      "loss": 1.526,
+      "step": 48800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.585353535353536e-05,
+      "loss": 1.513,
+      "step": 48810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.584848484848485e-05,
+      "loss": 1.5158,
+      "step": 48820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5843434343434347e-05,
+      "loss": 1.5241,
+      "step": 48830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.583838383838384e-05,
+      "loss": 1.5123,
+      "step": 48840
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5833333333333336e-05,
+      "loss": 1.5208,
+      "step": 48850
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5828282828282828e-05,
+      "loss": 1.5034,
+      "step": 48860
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5823232323232328e-05,
+      "loss": 1.505,
+      "step": 48870
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.581818181818182e-05,
+      "loss": 1.5082,
+      "step": 48880
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2.5813131313131316e-05,
+      "loss": 1.5269,
+      "step": 48890
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.580808080808081e-05,
+      "loss": 1.4989,
+      "step": 48900
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5803030303030308e-05,
+      "loss": 1.5138,
+      "step": 48910
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5797979797979797e-05,
+      "loss": 1.5231,
+      "step": 48920
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5792929292929297e-05,
+      "loss": 1.5259,
+      "step": 48930
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.578787878787879e-05,
+      "loss": 1.5373,
+      "step": 48940
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5782828282828285e-05,
+      "loss": 1.5206,
+      "step": 48950
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5777777777777778e-05,
+      "loss": 1.53,
+      "step": 48960
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5772727272727277e-05,
+      "loss": 1.5244,
+      "step": 48970
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.576767676767677e-05,
+      "loss": 1.5219,
+      "step": 48980
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5762626262626266e-05,
+      "loss": 1.5407,
+      "step": 48990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.575757575757576e-05,
+      "loss": 1.5019,
+      "step": 49000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5752525252525254e-05,
+      "loss": 1.5254,
+      "step": 49010
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5747474747474747e-05,
+      "loss": 1.5148,
+      "step": 49020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5742424242424246e-05,
+      "loss": 1.5097,
+      "step": 49030
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.573737373737374e-05,
+      "loss": 1.5116,
+      "step": 49040
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5732323232323235e-05,
+      "loss": 1.5032,
+      "step": 49050
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5727272727272728e-05,
+      "loss": 1.5093,
+      "step": 49060
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5722222222222224e-05,
+      "loss": 1.5155,
+      "step": 49070
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5717171717171716e-05,
+      "loss": 1.5093,
+      "step": 49080
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5712121212121216e-05,
+      "loss": 1.5051,
+      "step": 49090
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5707070707070708e-05,
+      "loss": 1.5162,
+      "step": 49100
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5702020202020204e-05,
+      "loss": 1.5269,
+      "step": 49110
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5696969696969697e-05,
+      "loss": 1.5089,
+      "step": 49120
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5691919191919196e-05,
+      "loss": 1.5238,
+      "step": 49130
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5686868686868685e-05,
+      "loss": 1.4818,
+      "step": 49140
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5681818181818185e-05,
+      "loss": 1.4925,
+      "step": 49150
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5676767676767677e-05,
+      "loss": 1.489,
+      "step": 49160
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5671717171717173e-05,
+      "loss": 1.5343,
+      "step": 49170
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5666666666666666e-05,
+      "loss": 1.5088,
+      "step": 49180
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5661616161616165e-05,
+      "loss": 1.5152,
+      "step": 49190
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5656565656565658e-05,
+      "loss": 1.5205,
+      "step": 49200
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5651515151515154e-05,
+      "loss": 1.4989,
+      "step": 49210
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5646464646464647e-05,
+      "loss": 1.4965,
+      "step": 49220
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5641414141414143e-05,
+      "loss": 1.5337,
+      "step": 49230
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5636363636363635e-05,
+      "loss": 1.5156,
+      "step": 49240
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5631313131313135e-05,
+      "loss": 1.5321,
+      "step": 49250
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5626262626262627e-05,
+      "loss": 1.5082,
+      "step": 49260
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5621212121212123e-05,
+      "loss": 1.5224,
+      "step": 49270
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5616161616161616e-05,
+      "loss": 1.5313,
+      "step": 49280
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5611111111111115e-05,
+      "loss": 1.5171,
+      "step": 49290
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5606060606060604e-05,
+      "loss": 1.5195,
+      "step": 49300
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5601010101010104e-05,
+      "loss": 1.4974,
+      "step": 49310
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5595959595959596e-05,
+      "loss": 1.5028,
+      "step": 49320
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5590909090909092e-05,
+      "loss": 1.505,
+      "step": 49330
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5585858585858585e-05,
+      "loss": 1.527,
+      "step": 49340
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5580808080808084e-05,
+      "loss": 1.5261,
+      "step": 49350
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5575757575757573e-05,
+      "loss": 1.5133,
+      "step": 49360
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5570707070707073e-05,
+      "loss": 1.53,
+      "step": 49370
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5565656565656565e-05,
+      "loss": 1.5171,
+      "step": 49380
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.556060606060606e-05,
+      "loss": 1.5073,
+      "step": 49390
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5555555555555554e-05,
+      "loss": 1.4798,
+      "step": 49400
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5550505050505053e-05,
+      "loss": 1.5075,
+      "step": 49410
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5545454545454546e-05,
+      "loss": 1.5116,
+      "step": 49420
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5540404040404042e-05,
+      "loss": 1.5183,
+      "step": 49430
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5535353535353535e-05,
+      "loss": 1.5052,
+      "step": 49440
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.553030303030303e-05,
+      "loss": 1.4819,
+      "step": 49450
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5525252525252523e-05,
+      "loss": 1.5015,
+      "step": 49460
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5520202020202023e-05,
+      "loss": 1.5316,
+      "step": 49470
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5515151515151515e-05,
+      "loss": 1.5361,
+      "step": 49480
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.551010101010101e-05,
+      "loss": 1.5068,
+      "step": 49490
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5505050505050504e-05,
+      "loss": 1.5129,
+      "step": 49500
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5500000000000003e-05,
+      "loss": 1.5232,
+      "step": 49510
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5494949494949492e-05,
+      "loss": 1.4977,
+      "step": 49520
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5489898989898992e-05,
+      "loss": 1.5046,
+      "step": 49530
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.5484848484848484e-05,
+      "loss": 1.5241,
+      "step": 49540
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2.547979797979798e-05,
+      "loss": 1.5047,
+      "step": 49550
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5474747474747473e-05,
+      "loss": 1.5116,
+      "step": 49560
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5469696969696972e-05,
+      "loss": 1.5118,
+      "step": 49570
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5464646464646465e-05,
+      "loss": 1.5221,
+      "step": 49580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.545959595959596e-05,
+      "loss": 1.5143,
+      "step": 49590
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5454545454545454e-05,
+      "loss": 1.5254,
+      "step": 49600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.544949494949495e-05,
+      "loss": 1.505,
+      "step": 49610
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5444444444444442e-05,
+      "loss": 1.5126,
+      "step": 49620
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.543939393939394e-05,
+      "loss": 1.5038,
+      "step": 49630
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5434343434343434e-05,
+      "loss": 1.5065,
+      "step": 49640
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.542929292929293e-05,
+      "loss": 1.5342,
+      "step": 49650
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.542424242424243e-05,
+      "loss": 1.5039,
+      "step": 49660
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.541919191919192e-05,
+      "loss": 1.5208,
+      "step": 49670
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5414141414141418e-05,
+      "loss": 1.5041,
+      "step": 49680
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.540909090909091e-05,
+      "loss": 1.494,
+      "step": 49690
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5404040404040407e-05,
+      "loss": 1.5138,
+      "step": 49700
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.53989898989899e-05,
+      "loss": 1.512,
+      "step": 49710
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.53939393939394e-05,
+      "loss": 1.5059,
+      "step": 49720
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.538888888888889e-05,
+      "loss": 1.5272,
+      "step": 49730
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5383838383838387e-05,
+      "loss": 1.5064,
+      "step": 49740
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.537878787878788e-05,
+      "loss": 1.513,
+      "step": 49750
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5373737373737376e-05,
+      "loss": 1.5136,
+      "step": 49760
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.536868686868687e-05,
+      "loss": 1.5183,
+      "step": 49770
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5363636363636368e-05,
+      "loss": 1.4958,
+      "step": 49780
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.535858585858586e-05,
+      "loss": 1.519,
+      "step": 49790
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5353535353535356e-05,
+      "loss": 1.5111,
+      "step": 49800
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.534848484848485e-05,
+      "loss": 1.489,
+      "step": 49810
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.534343434343435e-05,
+      "loss": 1.5231,
+      "step": 49820
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5338383838383838e-05,
+      "loss": 1.5224,
+      "step": 49830
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5333333333333337e-05,
+      "loss": 1.5179,
+      "step": 49840
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.532828282828283e-05,
+      "loss": 1.5138,
+      "step": 49850
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5323232323232326e-05,
+      "loss": 1.5192,
+      "step": 49860
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5318181818181818e-05,
+      "loss": 1.5041,
+      "step": 49870
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5313131313131318e-05,
+      "loss": 1.5017,
+      "step": 49880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.530808080808081e-05,
+      "loss": 1.5111,
+      "step": 49890
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5303030303030306e-05,
+      "loss": 1.5237,
+      "step": 49900
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.52979797979798e-05,
+      "loss": 1.5332,
+      "step": 49910
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5292929292929295e-05,
+      "loss": 1.506,
+      "step": 49920
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5287878787878787e-05,
+      "loss": 1.5216,
+      "step": 49930
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5282828282828287e-05,
+      "loss": 1.5265,
+      "step": 49940
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.527777777777778e-05,
+      "loss": 1.5264,
+      "step": 49950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5272727272727275e-05,
+      "loss": 1.5202,
+      "step": 49960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5267676767676768e-05,
+      "loss": 1.4894,
+      "step": 49970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5262626262626264e-05,
+      "loss": 1.5218,
+      "step": 49980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5257575757575757e-05,
+      "loss": 1.5098,
+      "step": 49990
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5252525252525256e-05,
+      "loss": 1.5097,
+      "step": 50000
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.524747474747475e-05,
+      "loss": 1.5191,
+      "step": 50010
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5242424242424244e-05,
+      "loss": 1.4913,
+      "step": 50020
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5237373737373737e-05,
+      "loss": 1.5127,
+      "step": 50030
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5232323232323236e-05,
+      "loss": 1.4992,
+      "step": 50040
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5227272727272726e-05,
+      "loss": 1.4961,
+      "step": 50050
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5222222222222225e-05,
+      "loss": 1.519,
+      "step": 50060
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5217171717171718e-05,
+      "loss": 1.5456,
+      "step": 50070
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5212121212121214e-05,
+      "loss": 1.4901,
+      "step": 50080
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5207070707070706e-05,
+      "loss": 1.5127,
+      "step": 50090
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5202020202020206e-05,
+      "loss": 1.5162,
+      "step": 50100
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5196969696969698e-05,
+      "loss": 1.5291,
+      "step": 50110
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5191919191919194e-05,
+      "loss": 1.518,
+      "step": 50120
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5186868686868687e-05,
+      "loss": 1.4893,
+      "step": 50130
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5181818181818183e-05,
+      "loss": 1.527,
+      "step": 50140
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5176767676767675e-05,
+      "loss": 1.509,
+      "step": 50150
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5171717171717175e-05,
+      "loss": 1.5096,
+      "step": 50160
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5166666666666667e-05,
+      "loss": 1.5291,
+      "step": 50170
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5161616161616163e-05,
+      "loss": 1.5193,
+      "step": 50180
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5156565656565656e-05,
+      "loss": 1.5147,
+      "step": 50190
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5151515151515155e-05,
+      "loss": 1.5054,
+      "step": 50200
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5146464646464645e-05,
+      "loss": 1.5021,
+      "step": 50210
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.5141414141414144e-05,
+      "loss": 1.5257,
+      "step": 50220
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5136363636363637e-05,
+      "loss": 1.4839,
+      "step": 50230
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5131313131313133e-05,
+      "loss": 1.5118,
+      "step": 50240
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5126262626262625e-05,
+      "loss": 1.505,
+      "step": 50250
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5121212121212125e-05,
+      "loss": 1.5316,
+      "step": 50260
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5116161616161614e-05,
+      "loss": 1.5212,
+      "step": 50270
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5111111111111113e-05,
+      "loss": 1.506,
+      "step": 50280
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5106060606060606e-05,
+      "loss": 1.508,
+      "step": 50290
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5101010101010102e-05,
+      "loss": 1.5135,
+      "step": 50300
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5095959595959594e-05,
+      "loss": 1.4995,
+      "step": 50310
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5090909090909094e-05,
+      "loss": 1.5075,
+      "step": 50320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5085858585858586e-05,
+      "loss": 1.4974,
+      "step": 50330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5080808080808082e-05,
+      "loss": 1.5185,
+      "step": 50340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5075757575757575e-05,
+      "loss": 1.4996,
+      "step": 50350
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.507070707070707e-05,
+      "loss": 1.5044,
+      "step": 50360
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5065656565656563e-05,
+      "loss": 1.5076,
+      "step": 50370
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5060606060606063e-05,
+      "loss": 1.5238,
+      "step": 50380
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5055555555555555e-05,
+      "loss": 1.516,
+      "step": 50390
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.505050505050505e-05,
+      "loss": 1.5304,
+      "step": 50400
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5045454545454544e-05,
+      "loss": 1.4839,
+      "step": 50410
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5040404040404043e-05,
+      "loss": 1.5353,
+      "step": 50420
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5035353535353533e-05,
+      "loss": 1.5187,
+      "step": 50430
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5030303030303032e-05,
+      "loss": 1.5099,
+      "step": 50440
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5025252525252525e-05,
+      "loss": 1.51,
+      "step": 50450
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.502020202020202e-05,
+      "loss": 1.5282,
+      "step": 50460
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5015151515151513e-05,
+      "loss": 1.5307,
+      "step": 50470
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5010101010101013e-05,
+      "loss": 1.4854,
+      "step": 50480
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5005050505050502e-05,
+      "loss": 1.52,
+      "step": 50490
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.5e-05,
+      "loss": 1.493,
+      "step": 50500
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4994949494949497e-05,
+      "loss": 1.518,
+      "step": 50510
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.498989898989899e-05,
+      "loss": 1.5292,
+      "step": 50520
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4984848484848486e-05,
+      "loss": 1.4896,
+      "step": 50530
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4979797979797982e-05,
+      "loss": 1.4805,
+      "step": 50540
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4974747474747474e-05,
+      "loss": 1.5153,
+      "step": 50550
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.496969696969697e-05,
+      "loss": 1.5058,
+      "step": 50560
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4964646464646466e-05,
+      "loss": 1.5287,
+      "step": 50570
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.495959595959596e-05,
+      "loss": 1.5007,
+      "step": 50580
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4954545454545455e-05,
+      "loss": 1.4961,
+      "step": 50590
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.494949494949495e-05,
+      "loss": 1.5127,
+      "step": 50600
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4944444444444447e-05,
+      "loss": 1.5096,
+      "step": 50610
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.493939393939394e-05,
+      "loss": 1.5163,
+      "step": 50620
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4934343434343436e-05,
+      "loss": 1.4753,
+      "step": 50630
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.492929292929293e-05,
+      "loss": 1.5077,
+      "step": 50640
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4924242424242424e-05,
+      "loss": 1.5245,
+      "step": 50650
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.491919191919192e-05,
+      "loss": 1.5082,
+      "step": 50660
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4914141414141416e-05,
+      "loss": 1.5287,
+      "step": 50670
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.490909090909091e-05,
+      "loss": 1.5184,
+      "step": 50680
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4904040404040405e-05,
+      "loss": 1.524,
+      "step": 50690
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.48989898989899e-05,
+      "loss": 1.5231,
+      "step": 50700
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4893939393939393e-05,
+      "loss": 1.4916,
+      "step": 50710
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.488888888888889e-05,
+      "loss": 1.5053,
+      "step": 50720
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4883838383838385e-05,
+      "loss": 1.5153,
+      "step": 50730
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4878787878787878e-05,
+      "loss": 1.5204,
+      "step": 50740
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4873737373737374e-05,
+      "loss": 1.5188,
+      "step": 50750
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.486868686868687e-05,
+      "loss": 1.5043,
+      "step": 50760
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4863636363636362e-05,
+      "loss": 1.5243,
+      "step": 50770
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.485858585858586e-05,
+      "loss": 1.5164,
+      "step": 50780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4853535353535354e-05,
+      "loss": 1.5105,
+      "step": 50790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4848484848484847e-05,
+      "loss": 1.5209,
+      "step": 50800
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4843434343434343e-05,
+      "loss": 1.509,
+      "step": 50810
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.483838383838384e-05,
+      "loss": 1.5111,
+      "step": 50820
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4833333333333335e-05,
+      "loss": 1.5094,
+      "step": 50830
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4828282828282828e-05,
+      "loss": 1.5116,
+      "step": 50840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4823232323232324e-05,
+      "loss": 1.4997,
+      "step": 50850
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.481818181818182e-05,
+      "loss": 1.4956,
+      "step": 50860
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4813131313131312e-05,
+      "loss": 1.5263,
+      "step": 50870
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.4808080808080808e-05,
+      "loss": 1.5286,
+      "step": 50880
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4803030303030304e-05,
+      "loss": 1.5104,
+      "step": 50890
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4797979797979797e-05,
+      "loss": 1.5122,
+      "step": 50900
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4792929292929293e-05,
+      "loss": 1.503,
+      "step": 50910
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.478787878787879e-05,
+      "loss": 1.5158,
+      "step": 50920
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4782828282828285e-05,
+      "loss": 1.5216,
+      "step": 50930
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.477777777777778e-05,
+      "loss": 1.4961,
+      "step": 50940
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4772727272727277e-05,
+      "loss": 1.5265,
+      "step": 50950
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.476767676767677e-05,
+      "loss": 1.5306,
+      "step": 50960
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4762626262626265e-05,
+      "loss": 1.5116,
+      "step": 50970
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.475757575757576e-05,
+      "loss": 1.514,
+      "step": 50980
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4752525252525254e-05,
+      "loss": 1.5433,
+      "step": 50990
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.474747474747475e-05,
+      "loss": 1.5071,
+      "step": 51000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4742424242424246e-05,
+      "loss": 1.5104,
+      "step": 51010
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.473737373737374e-05,
+      "loss": 1.5069,
+      "step": 51020
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4732323232323234e-05,
+      "loss": 1.5121,
+      "step": 51030
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.472727272727273e-05,
+      "loss": 1.4878,
+      "step": 51040
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4722222222222223e-05,
+      "loss": 1.5137,
+      "step": 51050
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.471717171717172e-05,
+      "loss": 1.5119,
+      "step": 51060
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4712121212121215e-05,
+      "loss": 1.5004,
+      "step": 51070
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4707070707070708e-05,
+      "loss": 1.5236,
+      "step": 51080
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4702020202020204e-05,
+      "loss": 1.5094,
+      "step": 51090
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.46969696969697e-05,
+      "loss": 1.5216,
+      "step": 51100
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4691919191919192e-05,
+      "loss": 1.4987,
+      "step": 51110
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4686868686868688e-05,
+      "loss": 1.5152,
+      "step": 51120
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4681818181818184e-05,
+      "loss": 1.5232,
+      "step": 51130
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.467676767676768e-05,
+      "loss": 1.5163,
+      "step": 51140
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4671717171717173e-05,
+      "loss": 1.519,
+      "step": 51150
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.466666666666667e-05,
+      "loss": 1.5458,
+      "step": 51160
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4661616161616165e-05,
+      "loss": 1.5027,
+      "step": 51170
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4656565656565657e-05,
+      "loss": 1.5001,
+      "step": 51180
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4651515151515153e-05,
+      "loss": 1.5088,
+      "step": 51190
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.464646464646465e-05,
+      "loss": 1.4837,
+      "step": 51200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4641414141414142e-05,
+      "loss": 1.5029,
+      "step": 51210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4636363636363638e-05,
+      "loss": 1.5042,
+      "step": 51220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4631313131313134e-05,
+      "loss": 1.5145,
+      "step": 51230
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4626262626262627e-05,
+      "loss": 1.4932,
+      "step": 51240
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4621212121212123e-05,
+      "loss": 1.5045,
+      "step": 51250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.461616161616162e-05,
+      "loss": 1.5061,
+      "step": 51260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.461111111111111e-05,
+      "loss": 1.498,
+      "step": 51270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4606060606060607e-05,
+      "loss": 1.5088,
+      "step": 51280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4601010101010103e-05,
+      "loss": 1.5105,
+      "step": 51290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4595959595959596e-05,
+      "loss": 1.5028,
+      "step": 51300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4590909090909092e-05,
+      "loss": 1.5132,
+      "step": 51310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4585858585858588e-05,
+      "loss": 1.5029,
+      "step": 51320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4580808080808084e-05,
+      "loss": 1.4993,
+      "step": 51330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4575757575757576e-05,
+      "loss": 1.5195,
+      "step": 51340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4570707070707072e-05,
+      "loss": 1.4953,
+      "step": 51350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4565656565656568e-05,
+      "loss": 1.4951,
+      "step": 51360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.456060606060606e-05,
+      "loss": 1.5191,
+      "step": 51370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4555555555555557e-05,
+      "loss": 1.5126,
+      "step": 51380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4550505050505053e-05,
+      "loss": 1.5091,
+      "step": 51390
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4545454545454545e-05,
+      "loss": 1.5091,
+      "step": 51400
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.454040404040404e-05,
+      "loss": 1.5001,
+      "step": 51410
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4535353535353537e-05,
+      "loss": 1.498,
+      "step": 51420
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.453030303030303e-05,
+      "loss": 1.4769,
+      "step": 51430
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4525252525252526e-05,
+      "loss": 1.5083,
+      "step": 51440
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4520202020202022e-05,
+      "loss": 1.5163,
+      "step": 51450
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4515151515151515e-05,
+      "loss": 1.5104,
+      "step": 51460
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.451010101010101e-05,
+      "loss": 1.5045,
+      "step": 51470
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4505050505050507e-05,
+      "loss": 1.509,
+      "step": 51480
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.45e-05,
+      "loss": 1.5171,
+      "step": 51490
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4494949494949495e-05,
+      "loss": 1.5259,
+      "step": 51500
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.448989898989899e-05,
+      "loss": 1.5228,
+      "step": 51510
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4484848484848484e-05,
+      "loss": 1.5178,
+      "step": 51520
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.447979797979798e-05,
+      "loss": 1.5041,
+      "step": 51530
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4474747474747476e-05,
+      "loss": 1.5093,
+      "step": 51540
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.4469696969696972e-05,
+      "loss": 1.5196,
+      "step": 51550
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4464646464646464e-05,
+      "loss": 1.4865,
+      "step": 51560
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.445959595959596e-05,
+      "loss": 1.5045,
+      "step": 51570
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4454545454545456e-05,
+      "loss": 1.5052,
+      "step": 51580
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.444949494949495e-05,
+      "loss": 1.5087,
+      "step": 51590
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4444444444444445e-05,
+      "loss": 1.5163,
+      "step": 51600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.443939393939394e-05,
+      "loss": 1.521,
+      "step": 51610
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4434343434343434e-05,
+      "loss": 1.5311,
+      "step": 51620
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.442929292929293e-05,
+      "loss": 1.5191,
+      "step": 51630
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4424242424242426e-05,
+      "loss": 1.5087,
+      "step": 51640
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4419191919191918e-05,
+      "loss": 1.5455,
+      "step": 51650
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4414141414141414e-05,
+      "loss": 1.5389,
+      "step": 51660
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.440909090909091e-05,
+      "loss": 1.5115,
+      "step": 51670
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4404040404040403e-05,
+      "loss": 1.5083,
+      "step": 51680
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.43989898989899e-05,
+      "loss": 1.5239,
+      "step": 51690
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4393939393939395e-05,
+      "loss": 1.5132,
+      "step": 51700
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4388888888888887e-05,
+      "loss": 1.4837,
+      "step": 51710
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4383838383838383e-05,
+      "loss": 1.5132,
+      "step": 51720
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.437878787878788e-05,
+      "loss": 1.5394,
+      "step": 51730
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4373737373737375e-05,
+      "loss": 1.5086,
+      "step": 51740
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4368686868686868e-05,
+      "loss": 1.5087,
+      "step": 51750
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4363636363636364e-05,
+      "loss": 1.4951,
+      "step": 51760
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.435858585858586e-05,
+      "loss": 1.499,
+      "step": 51770
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4353535353535352e-05,
+      "loss": 1.5077,
+      "step": 51780
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4348484848484852e-05,
+      "loss": 1.5222,
+      "step": 51790
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4343434343434344e-05,
+      "loss": 1.5118,
+      "step": 51800
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.433838383838384e-05,
+      "loss": 1.5024,
+      "step": 51810
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4333333333333336e-05,
+      "loss": 1.5218,
+      "step": 51820
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.432828282828283e-05,
+      "loss": 1.5174,
+      "step": 51830
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4323232323232325e-05,
+      "loss": 1.4852,
+      "step": 51840
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.431818181818182e-05,
+      "loss": 1.5078,
+      "step": 51850
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4313131313131317e-05,
+      "loss": 1.5252,
+      "step": 51860
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.430808080808081e-05,
+      "loss": 1.5233,
+      "step": 51870
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4303030303030306e-05,
+      "loss": 1.5288,
+      "step": 51880
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.42979797979798e-05,
+      "loss": 1.5089,
+      "step": 51890
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4292929292929294e-05,
+      "loss": 1.5442,
+      "step": 51900
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.428787878787879e-05,
+      "loss": 1.498,
+      "step": 51910
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4282828282828286e-05,
+      "loss": 1.5206,
+      "step": 51920
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.427777777777778e-05,
+      "loss": 1.5242,
+      "step": 51930
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4272727272727275e-05,
+      "loss": 1.5144,
+      "step": 51940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.426767676767677e-05,
+      "loss": 1.518,
+      "step": 51950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4262626262626263e-05,
+      "loss": 1.5126,
+      "step": 51960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.425757575757576e-05,
+      "loss": 1.512,
+      "step": 51970
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4252525252525255e-05,
+      "loss": 1.5008,
+      "step": 51980
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4247474747474748e-05,
+      "loss": 1.5078,
+      "step": 51990
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4242424242424244e-05,
+      "loss": 1.5404,
+      "step": 52000
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.423737373737374e-05,
+      "loss": 1.5104,
+      "step": 52010
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4232323232323232e-05,
+      "loss": 1.5141,
+      "step": 52020
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.422727272727273e-05,
+      "loss": 1.5095,
+      "step": 52030
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4222222222222224e-05,
+      "loss": 1.5063,
+      "step": 52040
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.421717171717172e-05,
+      "loss": 1.5286,
+      "step": 52050
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4212121212121213e-05,
+      "loss": 1.5151,
+      "step": 52060
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.420707070707071e-05,
+      "loss": 1.5064,
+      "step": 52070
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4202020202020205e-05,
+      "loss": 1.5218,
+      "step": 52080
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4196969696969698e-05,
+      "loss": 1.5113,
+      "step": 52090
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4191919191919194e-05,
+      "loss": 1.5153,
+      "step": 52100
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.418686868686869e-05,
+      "loss": 1.5109,
+      "step": 52110
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4181818181818182e-05,
+      "loss": 1.5164,
+      "step": 52120
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4176767676767678e-05,
+      "loss": 1.4769,
+      "step": 52130
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4171717171717174e-05,
+      "loss": 1.491,
+      "step": 52140
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4166666666666667e-05,
+      "loss": 1.5179,
+      "step": 52150
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4161616161616163e-05,
+      "loss": 1.5142,
+      "step": 52160
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.415656565656566e-05,
+      "loss": 1.4863,
+      "step": 52170
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.415151515151515e-05,
+      "loss": 1.5172,
+      "step": 52180
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4146464646464647e-05,
+      "loss": 1.5231,
+      "step": 52190
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4141414141414143e-05,
+      "loss": 1.5151,
+      "step": 52200
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4136363636363636e-05,
+      "loss": 1.51,
+      "step": 52210
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4131313131313132e-05,
+      "loss": 1.5193,
+      "step": 52220
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4126262626262628e-05,
+      "loss": 1.4773,
+      "step": 52230
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4121212121212124e-05,
+      "loss": 1.5193,
+      "step": 52240
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4116161616161617e-05,
+      "loss": 1.4975,
+      "step": 52250
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4111111111111113e-05,
+      "loss": 1.5077,
+      "step": 52260
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.410606060606061e-05,
+      "loss": 1.5229,
+      "step": 52270
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.41010101010101e-05,
+      "loss": 1.5195,
+      "step": 52280
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4095959595959597e-05,
+      "loss": 1.5087,
+      "step": 52290
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4090909090909093e-05,
+      "loss": 1.5026,
+      "step": 52300
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4085858585858586e-05,
+      "loss": 1.4957,
+      "step": 52310
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4080808080808082e-05,
+      "loss": 1.514,
+      "step": 52320
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4075757575757578e-05,
+      "loss": 1.5033,
+      "step": 52330
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.407070707070707e-05,
+      "loss": 1.5191,
+      "step": 52340
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4065656565656566e-05,
+      "loss": 1.5135,
+      "step": 52350
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4060606060606062e-05,
+      "loss": 1.5135,
+      "step": 52360
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4055555555555555e-05,
+      "loss": 1.4975,
+      "step": 52370
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.405050505050505e-05,
+      "loss": 1.4778,
+      "step": 52380
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4045454545454547e-05,
+      "loss": 1.501,
+      "step": 52390
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.404040404040404e-05,
+      "loss": 1.5239,
+      "step": 52400
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4035353535353535e-05,
+      "loss": 1.501,
+      "step": 52410
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.403030303030303e-05,
+      "loss": 1.5083,
+      "step": 52420
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4025252525252524e-05,
+      "loss": 1.481,
+      "step": 52430
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.402020202020202e-05,
+      "loss": 1.5145,
+      "step": 52440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4015151515151516e-05,
+      "loss": 1.5004,
+      "step": 52450
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4010101010101012e-05,
+      "loss": 1.4911,
+      "step": 52460
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4005050505050505e-05,
+      "loss": 1.5215,
+      "step": 52470
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.4e-05,
+      "loss": 1.5158,
+      "step": 52480
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3994949494949497e-05,
+      "loss": 1.5065,
+      "step": 52490
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.398989898989899e-05,
+      "loss": 1.5347,
+      "step": 52500
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3984848484848485e-05,
+      "loss": 1.4687,
+      "step": 52510
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.397979797979798e-05,
+      "loss": 1.5309,
+      "step": 52520
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3974747474747474e-05,
+      "loss": 1.5012,
+      "step": 52530
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.396969696969697e-05,
+      "loss": 1.5074,
+      "step": 52540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3964646464646466e-05,
+      "loss": 1.4806,
+      "step": 52550
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.395959595959596e-05,
+      "loss": 1.5101,
+      "step": 52560
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3954545454545454e-05,
+      "loss": 1.4928,
+      "step": 52570
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.394949494949495e-05,
+      "loss": 1.5228,
+      "step": 52580
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3944444444444443e-05,
+      "loss": 1.4992,
+      "step": 52590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.393939393939394e-05,
+      "loss": 1.5085,
+      "step": 52600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3934343434343435e-05,
+      "loss": 1.5057,
+      "step": 52610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3929292929292928e-05,
+      "loss": 1.5134,
+      "step": 52620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3924242424242424e-05,
+      "loss": 1.4917,
+      "step": 52630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3919191919191923e-05,
+      "loss": 1.4963,
+      "step": 52640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3914141414141416e-05,
+      "loss": 1.4867,
+      "step": 52650
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.390909090909091e-05,
+      "loss": 1.4884,
+      "step": 52660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3904040404040407e-05,
+      "loss": 1.4961,
+      "step": 52670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.38989898989899e-05,
+      "loss": 1.5255,
+      "step": 52680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3893939393939396e-05,
+      "loss": 1.48,
+      "step": 52690
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3888888888888892e-05,
+      "loss": 1.526,
+      "step": 52700
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3883838383838385e-05,
+      "loss": 1.5185,
+      "step": 52710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.387878787878788e-05,
+      "loss": 1.529,
+      "step": 52720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3873737373737377e-05,
+      "loss": 1.5126,
+      "step": 52730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.386868686868687e-05,
+      "loss": 1.4976,
+      "step": 52740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3863636363636365e-05,
+      "loss": 1.5096,
+      "step": 52750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.385858585858586e-05,
+      "loss": 1.4827,
+      "step": 52760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3853535353535357e-05,
+      "loss": 1.4988,
+      "step": 52770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.384848484848485e-05,
+      "loss": 1.5093,
+      "step": 52780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3843434343434346e-05,
+      "loss": 1.521,
+      "step": 52790
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3838383838383842e-05,
+      "loss": 1.4937,
+      "step": 52800
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3833333333333334e-05,
+      "loss": 1.5065,
+      "step": 52810
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.382828282828283e-05,
+      "loss": 1.4921,
+      "step": 52820
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3823232323232326e-05,
+      "loss": 1.4989,
+      "step": 52830
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.381818181818182e-05,
+      "loss": 1.5024,
+      "step": 52840
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3813131313131315e-05,
+      "loss": 1.5186,
+      "step": 52850
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.380808080808081e-05,
+      "loss": 1.5332,
+      "step": 52860
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3803030303030304e-05,
+      "loss": 1.5114,
+      "step": 52870
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.37979797979798e-05,
+      "loss": 1.5032,
+      "step": 52880
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3792929292929296e-05,
+      "loss": 1.4938,
+      "step": 52890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3787878787878788e-05,
+      "loss": 1.4994,
+      "step": 52900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3782828282828284e-05,
+      "loss": 1.4979,
+      "step": 52910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.377777777777778e-05,
+      "loss": 1.5015,
+      "step": 52920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3772727272727273e-05,
+      "loss": 1.5157,
+      "step": 52930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.376767676767677e-05,
+      "loss": 1.4927,
+      "step": 52940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3762626262626265e-05,
+      "loss": 1.4692,
+      "step": 52950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.375757575757576e-05,
+      "loss": 1.5204,
+      "step": 52960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3752525252525253e-05,
+      "loss": 1.5261,
+      "step": 52970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.374747474747475e-05,
+      "loss": 1.4972,
+      "step": 52980
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3742424242424245e-05,
+      "loss": 1.5103,
+      "step": 52990
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3737373737373738e-05,
+      "loss": 1.5208,
+      "step": 53000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3732323232323234e-05,
+      "loss": 1.4967,
+      "step": 53010
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.372727272727273e-05,
+      "loss": 1.5084,
+      "step": 53020
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3722222222222222e-05,
+      "loss": 1.5135,
+      "step": 53030
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.371717171717172e-05,
+      "loss": 1.5118,
+      "step": 53040
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3712121212121214e-05,
+      "loss": 1.5044,
+      "step": 53050
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3707070707070707e-05,
+      "loss": 1.5295,
+      "step": 53060
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3702020202020203e-05,
+      "loss": 1.502,
+      "step": 53070
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.36969696969697e-05,
+      "loss": 1.5087,
+      "step": 53080
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.369191919191919e-05,
+      "loss": 1.5125,
+      "step": 53090
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3686868686868688e-05,
+      "loss": 1.5235,
+      "step": 53100
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3681818181818184e-05,
+      "loss": 1.5179,
+      "step": 53110
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3676767676767676e-05,
+      "loss": 1.5185,
+      "step": 53120
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3671717171717172e-05,
+      "loss": 1.5037,
+      "step": 53130
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3666666666666668e-05,
+      "loss": 1.5287,
+      "step": 53140
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.366161616161616e-05,
+      "loss": 1.5069,
+      "step": 53150
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3656565656565657e-05,
+      "loss": 1.5428,
+      "step": 53160
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3651515151515153e-05,
+      "loss": 1.5245,
+      "step": 53170
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.364646464646465e-05,
+      "loss": 1.5092,
+      "step": 53180
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.364141414141414e-05,
+      "loss": 1.5096,
+      "step": 53190
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3636363636363637e-05,
+      "loss": 1.5156,
+      "step": 53200
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3631313131313133e-05,
+      "loss": 1.5054,
+      "step": 53210
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3626262626262626e-05,
+      "loss": 1.4801,
+      "step": 53220
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3621212121212122e-05,
+      "loss": 1.5065,
+      "step": 53230
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3616161616161618e-05,
+      "loss": 1.5224,
+      "step": 53240
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.361111111111111e-05,
+      "loss": 1.4957,
+      "step": 53250
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3606060606060607e-05,
+      "loss": 1.4957,
+      "step": 53260
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3601010101010103e-05,
+      "loss": 1.4941,
+      "step": 53270
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3595959595959595e-05,
+      "loss": 1.5017,
+      "step": 53280
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.359090909090909e-05,
+      "loss": 1.5654,
+      "step": 53290
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3585858585858587e-05,
+      "loss": 1.5104,
+      "step": 53300
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.358080808080808e-05,
+      "loss": 1.511,
+      "step": 53310
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3575757575757576e-05,
+      "loss": 1.4991,
+      "step": 53320
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3570707070707072e-05,
+      "loss": 1.5119,
+      "step": 53330
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3565656565656564e-05,
+      "loss": 1.5079,
+      "step": 53340
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.356060606060606e-05,
+      "loss": 1.5012,
+      "step": 53350
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3555555555555556e-05,
+      "loss": 1.523,
+      "step": 53360
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3550505050505052e-05,
+      "loss": 1.5263,
+      "step": 53370
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3545454545454545e-05,
+      "loss": 1.5184,
+      "step": 53380
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.354040404040404e-05,
+      "loss": 1.4903,
+      "step": 53390
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3535353535353537e-05,
+      "loss": 1.5066,
+      "step": 53400
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.353030303030303e-05,
+      "loss": 1.5237,
+      "step": 53410
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3525252525252525e-05,
+      "loss": 1.4994,
+      "step": 53420
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.352020202020202e-05,
+      "loss": 1.5094,
+      "step": 53430
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3515151515151514e-05,
+      "loss": 1.5314,
+      "step": 53440
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.351010101010101e-05,
+      "loss": 1.5023,
+      "step": 53450
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3505050505050506e-05,
+      "loss": 1.4981,
+      "step": 53460
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.35e-05,
+      "loss": 1.5153,
+      "step": 53470
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3494949494949495e-05,
+      "loss": 1.5034,
+      "step": 53480
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3489898989898994e-05,
+      "loss": 1.5095,
+      "step": 53490
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3484848484848487e-05,
+      "loss": 1.5033,
+      "step": 53500
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3479797979797983e-05,
+      "loss": 1.4997,
+      "step": 53510
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.347474747474748e-05,
+      "loss": 1.5158,
+      "step": 53520
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.346969696969697e-05,
+      "loss": 1.5173,
+      "step": 53530
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3464646464646467e-05,
+      "loss": 1.5069,
+      "step": 53540
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.3459595959595963e-05,
+      "loss": 1.507,
+      "step": 53550
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3454545454545456e-05,
+      "loss": 1.499,
+      "step": 53560
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3449494949494952e-05,
+      "loss": 1.513,
+      "step": 53570
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3444444444444448e-05,
+      "loss": 1.5253,
+      "step": 53580
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.343939393939394e-05,
+      "loss": 1.5169,
+      "step": 53590
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3434343434343436e-05,
+      "loss": 1.5147,
+      "step": 53600
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3429292929292932e-05,
+      "loss": 1.5066,
+      "step": 53610
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3424242424242425e-05,
+      "loss": 1.5142,
+      "step": 53620
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.341919191919192e-05,
+      "loss": 1.4873,
+      "step": 53630
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3414141414141417e-05,
+      "loss": 1.4971,
+      "step": 53640
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.340909090909091e-05,
+      "loss": 1.5281,
+      "step": 53650
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3404040404040405e-05,
+      "loss": 1.5055,
+      "step": 53660
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.33989898989899e-05,
+      "loss": 1.5141,
+      "step": 53670
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3393939393939397e-05,
+      "loss": 1.491,
+      "step": 53680
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.338888888888889e-05,
+      "loss": 1.5023,
+      "step": 53690
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3383838383838386e-05,
+      "loss": 1.5259,
+      "step": 53700
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3378787878787882e-05,
+      "loss": 1.5214,
+      "step": 53710
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3373737373737375e-05,
+      "loss": 1.5185,
+      "step": 53720
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.336868686868687e-05,
+      "loss": 1.5125,
+      "step": 53730
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3363636363636367e-05,
+      "loss": 1.5198,
+      "step": 53740
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.335858585858586e-05,
+      "loss": 1.521,
+      "step": 53750
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3353535353535355e-05,
+      "loss": 1.5181,
+      "step": 53760
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.334848484848485e-05,
+      "loss": 1.5174,
+      "step": 53770
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3343434343434344e-05,
+      "loss": 1.4879,
+      "step": 53780
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.333838383838384e-05,
+      "loss": 1.5261,
+      "step": 53790
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3333333333333336e-05,
+      "loss": 1.4956,
+      "step": 53800
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.332828282828283e-05,
+      "loss": 1.497,
+      "step": 53810
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3323232323232324e-05,
+      "loss": 1.5069,
+      "step": 53820
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.331818181818182e-05,
+      "loss": 1.5023,
+      "step": 53830
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3313131313131313e-05,
+      "loss": 1.5228,
+      "step": 53840
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.330808080808081e-05,
+      "loss": 1.4956,
+      "step": 53850
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3303030303030305e-05,
+      "loss": 1.5242,
+      "step": 53860
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3297979797979798e-05,
+      "loss": 1.4758,
+      "step": 53870
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3292929292929294e-05,
+      "loss": 1.5133,
+      "step": 53880
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.328787878787879e-05,
+      "loss": 1.5021,
+      "step": 53890
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3282828282828286e-05,
+      "loss": 1.5155,
+      "step": 53900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3277777777777778e-05,
+      "loss": 1.4962,
+      "step": 53910
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3272727272727274e-05,
+      "loss": 1.5358,
+      "step": 53920
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.326767676767677e-05,
+      "loss": 1.4992,
+      "step": 53930
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3262626262626263e-05,
+      "loss": 1.5032,
+      "step": 53940
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.325757575757576e-05,
+      "loss": 1.5221,
+      "step": 53950
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3252525252525255e-05,
+      "loss": 1.5023,
+      "step": 53960
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3247474747474747e-05,
+      "loss": 1.4941,
+      "step": 53970
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3242424242424243e-05,
+      "loss": 1.5175,
+      "step": 53980
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.323737373737374e-05,
+      "loss": 1.5259,
+      "step": 53990
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3232323232323232e-05,
+      "loss": 1.5177,
+      "step": 54000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3227272727272728e-05,
+      "loss": 1.5192,
+      "step": 54010
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3222222222222224e-05,
+      "loss": 1.5225,
+      "step": 54020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3217171717171716e-05,
+      "loss": 1.521,
+      "step": 54030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3212121212121212e-05,
+      "loss": 1.521,
+      "step": 54040
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.320707070707071e-05,
+      "loss": 1.4951,
+      "step": 54050
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.32020202020202e-05,
+      "loss": 1.4958,
+      "step": 54060
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3196969696969697e-05,
+      "loss": 1.4881,
+      "step": 54070
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3191919191919193e-05,
+      "loss": 1.4953,
+      "step": 54080
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.318686868686869e-05,
+      "loss": 1.5282,
+      "step": 54090
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.318181818181818e-05,
+      "loss": 1.5097,
+      "step": 54100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3176767676767678e-05,
+      "loss": 1.4774,
+      "step": 54110
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3171717171717174e-05,
+      "loss": 1.5082,
+      "step": 54120
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3166666666666666e-05,
+      "loss": 1.5305,
+      "step": 54130
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3161616161616162e-05,
+      "loss": 1.5059,
+      "step": 54140
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3156565656565658e-05,
+      "loss": 1.4946,
+      "step": 54150
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.315151515151515e-05,
+      "loss": 1.4948,
+      "step": 54160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3146464646464647e-05,
+      "loss": 1.5314,
+      "step": 54170
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3141414141414143e-05,
+      "loss": 1.5092,
+      "step": 54180
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3136363636363635e-05,
+      "loss": 1.5328,
+      "step": 54190
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.313131313131313e-05,
+      "loss": 1.5086,
+      "step": 54200
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2.3126262626262627e-05,
+      "loss": 1.5295,
+      "step": 54210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.312121212121212e-05,
+      "loss": 1.5071,
+      "step": 54220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3116161616161616e-05,
+      "loss": 1.4863,
+      "step": 54230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3111111111111112e-05,
+      "loss": 1.5136,
+      "step": 54240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3106060606060605e-05,
+      "loss": 1.4792,
+      "step": 54250
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.31010101010101e-05,
+      "loss": 1.4958,
+      "step": 54260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3095959595959597e-05,
+      "loss": 1.4968,
+      "step": 54270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.309090909090909e-05,
+      "loss": 1.5202,
+      "step": 54280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3085858585858585e-05,
+      "loss": 1.5096,
+      "step": 54290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.308080808080808e-05,
+      "loss": 1.5156,
+      "step": 54300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3075757575757577e-05,
+      "loss": 1.5032,
+      "step": 54310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.307070707070707e-05,
+      "loss": 1.5216,
+      "step": 54320
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3065656565656566e-05,
+      "loss": 1.5184,
+      "step": 54330
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.306060606060606e-05,
+      "loss": 1.5102,
+      "step": 54340
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3055555555555558e-05,
+      "loss": 1.4924,
+      "step": 54350
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3050505050505054e-05,
+      "loss": 1.4989,
+      "step": 54360
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3045454545454546e-05,
+      "loss": 1.5204,
+      "step": 54370
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3040404040404042e-05,
+      "loss": 1.4985,
+      "step": 54380
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3035353535353538e-05,
+      "loss": 1.5193,
+      "step": 54390
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3030303030303034e-05,
+      "loss": 1.5086,
+      "step": 54400
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3025252525252527e-05,
+      "loss": 1.5043,
+      "step": 54410
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3020202020202023e-05,
+      "loss": 1.5048,
+      "step": 54420
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.301515151515152e-05,
+      "loss": 1.5118,
+      "step": 54430
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.301010101010101e-05,
+      "loss": 1.5157,
+      "step": 54440
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3005050505050507e-05,
+      "loss": 1.5353,
+      "step": 54450
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.3000000000000003e-05,
+      "loss": 1.5278,
+      "step": 54460
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2994949494949496e-05,
+      "loss": 1.5113,
+      "step": 54470
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2989898989898992e-05,
+      "loss": 1.5174,
+      "step": 54480
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2984848484848488e-05,
+      "loss": 1.5083,
+      "step": 54490
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.297979797979798e-05,
+      "loss": 1.5062,
+      "step": 54500
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2974747474747477e-05,
+      "loss": 1.5061,
+      "step": 54510
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2969696969696973e-05,
+      "loss": 1.5054,
+      "step": 54520
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2964646464646465e-05,
+      "loss": 1.5207,
+      "step": 54530
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.295959595959596e-05,
+      "loss": 1.5208,
+      "step": 54540
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2954545454545457e-05,
+      "loss": 1.4968,
+      "step": 54550
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.294949494949495e-05,
+      "loss": 1.494,
+      "step": 54560
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2944444444444446e-05,
+      "loss": 1.4986,
+      "step": 54570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2939393939393942e-05,
+      "loss": 1.5126,
+      "step": 54580
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2934343434343434e-05,
+      "loss": 1.5108,
+      "step": 54590
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.292929292929293e-05,
+      "loss": 1.5153,
+      "step": 54600
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2924242424242426e-05,
+      "loss": 1.5155,
+      "step": 54610
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2919191919191922e-05,
+      "loss": 1.5159,
+      "step": 54620
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2914141414141415e-05,
+      "loss": 1.5034,
+      "step": 54630
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.290909090909091e-05,
+      "loss": 1.5223,
+      "step": 54640
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2904040404040407e-05,
+      "loss": 1.511,
+      "step": 54650
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.28989898989899e-05,
+      "loss": 1.4884,
+      "step": 54660
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2893939393939395e-05,
+      "loss": 1.5091,
+      "step": 54670
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.288888888888889e-05,
+      "loss": 1.5346,
+      "step": 54680
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2883838383838384e-05,
+      "loss": 1.5082,
+      "step": 54690
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.287878787878788e-05,
+      "loss": 1.5082,
+      "step": 54700
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2873737373737376e-05,
+      "loss": 1.5479,
+      "step": 54710
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.286868686868687e-05,
+      "loss": 1.498,
+      "step": 54720
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2863636363636365e-05,
+      "loss": 1.5231,
+      "step": 54730
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.285858585858586e-05,
+      "loss": 1.5183,
+      "step": 54740
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2853535353535353e-05,
+      "loss": 1.5179,
+      "step": 54750
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.284848484848485e-05,
+      "loss": 1.4822,
+      "step": 54760
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2843434343434345e-05,
+      "loss": 1.4883,
+      "step": 54770
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2838383838383838e-05,
+      "loss": 1.5117,
+      "step": 54780
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2833333333333334e-05,
+      "loss": 1.4914,
+      "step": 54790
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.282828282828283e-05,
+      "loss": 1.5164,
+      "step": 54800
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2823232323232326e-05,
+      "loss": 1.4992,
+      "step": 54810
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.281818181818182e-05,
+      "loss": 1.4934,
+      "step": 54820
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2813131313131314e-05,
+      "loss": 1.5115,
+      "step": 54830
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.280808080808081e-05,
+      "loss": 1.5022,
+      "step": 54840
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2803030303030303e-05,
+      "loss": 1.4963,
+      "step": 54850
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.27979797979798e-05,
+      "loss": 1.5313,
+      "step": 54860
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2792929292929295e-05,
+      "loss": 1.5093,
+      "step": 54870
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2.2787878787878788e-05,
+      "loss": 1.5255,
+      "step": 54880
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2782828282828284e-05,
+      "loss": 1.509,
+      "step": 54890
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.277777777777778e-05,
+      "loss": 1.5211,
+      "step": 54900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2772727272727272e-05,
+      "loss": 1.5025,
+      "step": 54910
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2767676767676768e-05,
+      "loss": 1.4957,
+      "step": 54920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2762626262626264e-05,
+      "loss": 1.5109,
+      "step": 54930
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2757575757575757e-05,
+      "loss": 1.5119,
+      "step": 54940
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2752525252525253e-05,
+      "loss": 1.505,
+      "step": 54950
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.274747474747475e-05,
+      "loss": 1.5159,
+      "step": 54960
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.274242424242424e-05,
+      "loss": 1.5137,
+      "step": 54970
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2737373737373737e-05,
+      "loss": 1.5215,
+      "step": 54980
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2732323232323233e-05,
+      "loss": 1.4968,
+      "step": 54990
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.272727272727273e-05,
+      "loss": 1.5028,
+      "step": 55000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2722222222222222e-05,
+      "loss": 1.5101,
+      "step": 55010
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2717171717171718e-05,
+      "loss": 1.4998,
+      "step": 55020
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2712121212121214e-05,
+      "loss": 1.5146,
+      "step": 55030
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2707070707070706e-05,
+      "loss": 1.5337,
+      "step": 55040
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2702020202020202e-05,
+      "loss": 1.5149,
+      "step": 55050
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.26969696969697e-05,
+      "loss": 1.4695,
+      "step": 55060
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.269191919191919e-05,
+      "loss": 1.4971,
+      "step": 55070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2686868686868687e-05,
+      "loss": 1.5059,
+      "step": 55080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2681818181818183e-05,
+      "loss": 1.5022,
+      "step": 55090
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2676767676767676e-05,
+      "loss": 1.5244,
+      "step": 55100
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.267171717171717e-05,
+      "loss": 1.4891,
+      "step": 55110
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2666666666666668e-05,
+      "loss": 1.5226,
+      "step": 55120
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.266161616161616e-05,
+      "loss": 1.5035,
+      "step": 55130
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2656565656565656e-05,
+      "loss": 1.5174,
+      "step": 55140
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2651515151515152e-05,
+      "loss": 1.5267,
+      "step": 55150
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2646464646464645e-05,
+      "loss": 1.5114,
+      "step": 55160
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.264141414141414e-05,
+      "loss": 1.4792,
+      "step": 55170
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2636363636363637e-05,
+      "loss": 1.5162,
+      "step": 55180
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.263131313131313e-05,
+      "loss": 1.5084,
+      "step": 55190
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.262626262626263e-05,
+      "loss": 1.5029,
+      "step": 55200
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2621212121212125e-05,
+      "loss": 1.5057,
+      "step": 55210
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2616161616161617e-05,
+      "loss": 1.4963,
+      "step": 55220
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2611111111111113e-05,
+      "loss": 1.5134,
+      "step": 55230
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.260606060606061e-05,
+      "loss": 1.4907,
+      "step": 55240
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2601010101010102e-05,
+      "loss": 1.4976,
+      "step": 55250
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2595959595959598e-05,
+      "loss": 1.5096,
+      "step": 55260
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2590909090909094e-05,
+      "loss": 1.4936,
+      "step": 55270
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2585858585858587e-05,
+      "loss": 1.4978,
+      "step": 55280
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2580808080808083e-05,
+      "loss": 1.5353,
+      "step": 55290
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.257575757575758e-05,
+      "loss": 1.5083,
+      "step": 55300
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2570707070707074e-05,
+      "loss": 1.5045,
+      "step": 55310
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2565656565656567e-05,
+      "loss": 1.5096,
+      "step": 55320
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2560606060606063e-05,
+      "loss": 1.508,
+      "step": 55330
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.255555555555556e-05,
+      "loss": 1.5297,
+      "step": 55340
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.255050505050505e-05,
+      "loss": 1.4892,
+      "step": 55350
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2545454545454548e-05,
+      "loss": 1.5017,
+      "step": 55360
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2540404040404044e-05,
+      "loss": 1.5225,
+      "step": 55370
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2535353535353536e-05,
+      "loss": 1.5009,
+      "step": 55380
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2530303030303032e-05,
+      "loss": 1.5124,
+      "step": 55390
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2525252525252528e-05,
+      "loss": 1.5016,
+      "step": 55400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.252020202020202e-05,
+      "loss": 1.5172,
+      "step": 55410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2515151515151517e-05,
+      "loss": 1.498,
+      "step": 55420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2510101010101013e-05,
+      "loss": 1.4879,
+      "step": 55430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2505050505050505e-05,
+      "loss": 1.5069,
+      "step": 55440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.25e-05,
+      "loss": 1.5091,
+      "step": 55450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2494949494949497e-05,
+      "loss": 1.4998,
+      "step": 55460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.248989898989899e-05,
+      "loss": 1.4954,
+      "step": 55470
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2484848484848486e-05,
+      "loss": 1.5054,
+      "step": 55480
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2479797979797982e-05,
+      "loss": 1.4853,
+      "step": 55490
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2474747474747475e-05,
+      "loss": 1.4796,
+      "step": 55500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.246969696969697e-05,
+      "loss": 1.5131,
+      "step": 55510
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2464646464646467e-05,
+      "loss": 1.4996,
+      "step": 55520
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2459595959595963e-05,
+      "loss": 1.4642,
+      "step": 55530
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2.2454545454545455e-05,
+      "loss": 1.5033,
+      "step": 55540
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.244949494949495e-05,
+      "loss": 1.5133,
+      "step": 55550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2444444444444447e-05,
+      "loss": 1.522,
+      "step": 55560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.243939393939394e-05,
+      "loss": 1.5045,
+      "step": 55570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2434343434343436e-05,
+      "loss": 1.5227,
+      "step": 55580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2429292929292932e-05,
+      "loss": 1.5036,
+      "step": 55590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2424242424242424e-05,
+      "loss": 1.5268,
+      "step": 55600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.241919191919192e-05,
+      "loss": 1.5058,
+      "step": 55610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2414141414141416e-05,
+      "loss": 1.5248,
+      "step": 55620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.240909090909091e-05,
+      "loss": 1.5071,
+      "step": 55630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2404040404040405e-05,
+      "loss": 1.505,
+      "step": 55640
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.23989898989899e-05,
+      "loss": 1.5088,
+      "step": 55650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2393939393939393e-05,
+      "loss": 1.5152,
+      "step": 55660
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.238888888888889e-05,
+      "loss": 1.5044,
+      "step": 55670
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2383838383838385e-05,
+      "loss": 1.5165,
+      "step": 55680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2378787878787878e-05,
+      "loss": 1.5192,
+      "step": 55690
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2373737373737374e-05,
+      "loss": 1.5244,
+      "step": 55700
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.236868686868687e-05,
+      "loss": 1.4986,
+      "step": 55710
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2363636363636366e-05,
+      "loss": 1.5006,
+      "step": 55720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.235858585858586e-05,
+      "loss": 1.5136,
+      "step": 55730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2353535353535355e-05,
+      "loss": 1.5248,
+      "step": 55740
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.234848484848485e-05,
+      "loss": 1.5106,
+      "step": 55750
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2343434343434343e-05,
+      "loss": 1.5112,
+      "step": 55760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.233838383838384e-05,
+      "loss": 1.495,
+      "step": 55770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2333333333333335e-05,
+      "loss": 1.5066,
+      "step": 55780
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2328282828282828e-05,
+      "loss": 1.4951,
+      "step": 55790
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2323232323232324e-05,
+      "loss": 1.4985,
+      "step": 55800
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.231818181818182e-05,
+      "loss": 1.4957,
+      "step": 55810
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2313131313131312e-05,
+      "loss": 1.4939,
+      "step": 55820
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.230808080808081e-05,
+      "loss": 1.4908,
+      "step": 55830
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2303030303030304e-05,
+      "loss": 1.4916,
+      "step": 55840
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2297979797979797e-05,
+      "loss": 1.5322,
+      "step": 55850
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2292929292929293e-05,
+      "loss": 1.5188,
+      "step": 55860
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.228787878787879e-05,
+      "loss": 1.4883,
+      "step": 55870
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.228282828282828e-05,
+      "loss": 1.478,
+      "step": 55880
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2277777777777778e-05,
+      "loss": 1.4953,
+      "step": 55890
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2272727272727274e-05,
+      "loss": 1.5114,
+      "step": 55900
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2267676767676766e-05,
+      "loss": 1.5068,
+      "step": 55910
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2262626262626262e-05,
+      "loss": 1.5002,
+      "step": 55920
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2257575757575758e-05,
+      "loss": 1.4993,
+      "step": 55930
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2252525252525254e-05,
+      "loss": 1.5122,
+      "step": 55940
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2247474747474747e-05,
+      "loss": 1.4837,
+      "step": 55950
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2242424242424243e-05,
+      "loss": 1.5181,
+      "step": 55960
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.223737373737374e-05,
+      "loss": 1.5023,
+      "step": 55970
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.223232323232323e-05,
+      "loss": 1.4921,
+      "step": 55980
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2227272727272727e-05,
+      "loss": 1.5047,
+      "step": 55990
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2222222222222223e-05,
+      "loss": 1.4855,
+      "step": 56000
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2217171717171716e-05,
+      "loss": 1.486,
+      "step": 56010
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2212121212121212e-05,
+      "loss": 1.4951,
+      "step": 56020
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2207070707070708e-05,
+      "loss": 1.4964,
+      "step": 56030
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.22020202020202e-05,
+      "loss": 1.5013,
+      "step": 56040
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.21969696969697e-05,
+      "loss": 1.5101,
+      "step": 56050
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2191919191919196e-05,
+      "loss": 1.4853,
+      "step": 56060
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.218686868686869e-05,
+      "loss": 1.5098,
+      "step": 56070
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2181818181818184e-05,
+      "loss": 1.5022,
+      "step": 56080
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.217676767676768e-05,
+      "loss": 1.4861,
+      "step": 56090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2171717171717173e-05,
+      "loss": 1.5052,
+      "step": 56100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.216666666666667e-05,
+      "loss": 1.5134,
+      "step": 56110
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2161616161616165e-05,
+      "loss": 1.5075,
+      "step": 56120
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2156565656565658e-05,
+      "loss": 1.4862,
+      "step": 56130
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2151515151515154e-05,
+      "loss": 1.5165,
+      "step": 56140
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.214646464646465e-05,
+      "loss": 1.5043,
+      "step": 56150
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2141414141414142e-05,
+      "loss": 1.5151,
+      "step": 56160
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2136363636363638e-05,
+      "loss": 1.5148,
+      "step": 56170
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2131313131313134e-05,
+      "loss": 1.5195,
+      "step": 56180
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2126262626262627e-05,
+      "loss": 1.5022,
+      "step": 56190
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.2121212121212123e-05,
+      "loss": 1.49,
+      "step": 56200
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2.211616161616162e-05,
+      "loss": 1.5011,
+      "step": 56210
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.211111111111111e-05,
+      "loss": 1.52,
+      "step": 56220
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2106060606060607e-05,
+      "loss": 1.5067,
+      "step": 56230
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2101010101010103e-05,
+      "loss": 1.5124,
+      "step": 56240
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.20959595959596e-05,
+      "loss": 1.5319,
+      "step": 56250
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2090909090909092e-05,
+      "loss": 1.4888,
+      "step": 56260
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2085858585858588e-05,
+      "loss": 1.5041,
+      "step": 56270
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2080808080808084e-05,
+      "loss": 1.4858,
+      "step": 56280
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2075757575757577e-05,
+      "loss": 1.5128,
+      "step": 56290
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2070707070707073e-05,
+      "loss": 1.4907,
+      "step": 56300
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.206565656565657e-05,
+      "loss": 1.5009,
+      "step": 56310
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.206060606060606e-05,
+      "loss": 1.5102,
+      "step": 56320
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2055555555555557e-05,
+      "loss": 1.5148,
+      "step": 56330
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2050505050505053e-05,
+      "loss": 1.5098,
+      "step": 56340
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2045454545454546e-05,
+      "loss": 1.5058,
+      "step": 56350
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.204040404040404e-05,
+      "loss": 1.4876,
+      "step": 56360
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2035353535353538e-05,
+      "loss": 1.4832,
+      "step": 56370
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.203030303030303e-05,
+      "loss": 1.4908,
+      "step": 56380
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2025252525252526e-05,
+      "loss": 1.4922,
+      "step": 56390
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2020202020202022e-05,
+      "loss": 1.4967,
+      "step": 56400
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2015151515151515e-05,
+      "loss": 1.5013,
+      "step": 56410
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.201010101010101e-05,
+      "loss": 1.5139,
+      "step": 56420
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2005050505050507e-05,
+      "loss": 1.5233,
+      "step": 56430
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 1.4918,
+      "step": 56440
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1994949494949495e-05,
+      "loss": 1.4994,
+      "step": 56450
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.198989898989899e-05,
+      "loss": 1.5256,
+      "step": 56460
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1984848484848487e-05,
+      "loss": 1.5237,
+      "step": 56470
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.197979797979798e-05,
+      "loss": 1.5165,
+      "step": 56480
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1974747474747476e-05,
+      "loss": 1.5028,
+      "step": 56490
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1969696969696972e-05,
+      "loss": 1.4813,
+      "step": 56500
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1964646464646465e-05,
+      "loss": 1.4682,
+      "step": 56510
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.195959595959596e-05,
+      "loss": 1.5221,
+      "step": 56520
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1954545454545457e-05,
+      "loss": 1.498,
+      "step": 56530
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.194949494949495e-05,
+      "loss": 1.5146,
+      "step": 56540
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1944444444444445e-05,
+      "loss": 1.5212,
+      "step": 56550
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.193939393939394e-05,
+      "loss": 1.5077,
+      "step": 56560
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1934343434343434e-05,
+      "loss": 1.5068,
+      "step": 56570
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.192929292929293e-05,
+      "loss": 1.5089,
+      "step": 56580
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1924242424242426e-05,
+      "loss": 1.504,
+      "step": 56590
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.191919191919192e-05,
+      "loss": 1.488,
+      "step": 56600
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1914141414141414e-05,
+      "loss": 1.5097,
+      "step": 56610
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.190909090909091e-05,
+      "loss": 1.5003,
+      "step": 56620
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1904040404040403e-05,
+      "loss": 1.4942,
+      "step": 56630
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.18989898989899e-05,
+      "loss": 1.4981,
+      "step": 56640
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1893939393939395e-05,
+      "loss": 1.5084,
+      "step": 56650
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.188888888888889e-05,
+      "loss": 1.5196,
+      "step": 56660
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1883838383838383e-05,
+      "loss": 1.514,
+      "step": 56670
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.187878787878788e-05,
+      "loss": 1.496,
+      "step": 56680
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1873737373737375e-05,
+      "loss": 1.5183,
+      "step": 56690
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1868686868686868e-05,
+      "loss": 1.498,
+      "step": 56700
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1863636363636364e-05,
+      "loss": 1.522,
+      "step": 56710
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.185858585858586e-05,
+      "loss": 1.4953,
+      "step": 56720
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1853535353535353e-05,
+      "loss": 1.5009,
+      "step": 56730
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.184848484848485e-05,
+      "loss": 1.4874,
+      "step": 56740
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1843434343434345e-05,
+      "loss": 1.4944,
+      "step": 56750
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1838383838383837e-05,
+      "loss": 1.5092,
+      "step": 56760
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1833333333333333e-05,
+      "loss": 1.503,
+      "step": 56770
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.182828282828283e-05,
+      "loss": 1.4838,
+      "step": 56780
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1823232323232322e-05,
+      "loss": 1.4991,
+      "step": 56790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1818181818181818e-05,
+      "loss": 1.507,
+      "step": 56800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1813131313131314e-05,
+      "loss": 1.517,
+      "step": 56810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1808080808080806e-05,
+      "loss": 1.5222,
+      "step": 56820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1803030303030302e-05,
+      "loss": 1.5153,
+      "step": 56830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.17979797979798e-05,
+      "loss": 1.4878,
+      "step": 56840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1792929292929294e-05,
+      "loss": 1.5009,
+      "step": 56850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1787878787878787e-05,
+      "loss": 1.5225,
+      "step": 56860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2.1782828282828283e-05,
+      "loss": 1.4695,
+      "step": 56870
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.177777777777778e-05,
+      "loss": 1.503,
+      "step": 56880
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.177272727272727e-05,
+      "loss": 1.5015,
+      "step": 56890
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1767676767676768e-05,
+      "loss": 1.5211,
+      "step": 56900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1762626262626264e-05,
+      "loss": 1.501,
+      "step": 56910
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.175757575757576e-05,
+      "loss": 1.4789,
+      "step": 56920
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1752525252525256e-05,
+      "loss": 1.4856,
+      "step": 56930
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1747474747474748e-05,
+      "loss": 1.4871,
+      "step": 56940
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1742424242424244e-05,
+      "loss": 1.4787,
+      "step": 56950
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.173737373737374e-05,
+      "loss": 1.5038,
+      "step": 56960
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1732323232323236e-05,
+      "loss": 1.5016,
+      "step": 56970
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.172727272727273e-05,
+      "loss": 1.5057,
+      "step": 56980
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1722222222222225e-05,
+      "loss": 1.4998,
+      "step": 56990
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.171717171717172e-05,
+      "loss": 1.5062,
+      "step": 57000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1712121212121213e-05,
+      "loss": 1.4668,
+      "step": 57010
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.170707070707071e-05,
+      "loss": 1.479,
+      "step": 57020
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1702020202020205e-05,
+      "loss": 1.5157,
+      "step": 57030
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1696969696969698e-05,
+      "loss": 1.5168,
+      "step": 57040
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1691919191919194e-05,
+      "loss": 1.5021,
+      "step": 57050
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.168686868686869e-05,
+      "loss": 1.4922,
+      "step": 57060
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1681818181818182e-05,
+      "loss": 1.4935,
+      "step": 57070
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.167676767676768e-05,
+      "loss": 1.4926,
+      "step": 57080
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1671717171717174e-05,
+      "loss": 1.5203,
+      "step": 57090
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1666666666666667e-05,
+      "loss": 1.5039,
+      "step": 57100
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1661616161616163e-05,
+      "loss": 1.4926,
+      "step": 57110
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.165656565656566e-05,
+      "loss": 1.4805,
+      "step": 57120
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.165151515151515e-05,
+      "loss": 1.5016,
+      "step": 57130
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1646464646464648e-05,
+      "loss": 1.4953,
+      "step": 57140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1641414141414144e-05,
+      "loss": 1.4911,
+      "step": 57150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.163636363636364e-05,
+      "loss": 1.5178,
+      "step": 57160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1631313131313132e-05,
+      "loss": 1.4995,
+      "step": 57170
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1626262626262628e-05,
+      "loss": 1.5208,
+      "step": 57180
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1621212121212124e-05,
+      "loss": 1.5221,
+      "step": 57190
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1616161616161617e-05,
+      "loss": 1.5147,
+      "step": 57200
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1611111111111113e-05,
+      "loss": 1.5131,
+      "step": 57210
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.160606060606061e-05,
+      "loss": 1.4886,
+      "step": 57220
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.16010101010101e-05,
+      "loss": 1.4945,
+      "step": 57230
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1595959595959597e-05,
+      "loss": 1.4965,
+      "step": 57240
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1590909090909093e-05,
+      "loss": 1.5174,
+      "step": 57250
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1585858585858586e-05,
+      "loss": 1.5177,
+      "step": 57260
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1580808080808082e-05,
+      "loss": 1.5232,
+      "step": 57270
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1575757575757578e-05,
+      "loss": 1.4892,
+      "step": 57280
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.157070707070707e-05,
+      "loss": 1.4881,
+      "step": 57290
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1565656565656567e-05,
+      "loss": 1.5016,
+      "step": 57300
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1560606060606062e-05,
+      "loss": 1.5141,
+      "step": 57310
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1555555555555555e-05,
+      "loss": 1.5088,
+      "step": 57320
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.155050505050505e-05,
+      "loss": 1.5184,
+      "step": 57330
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1545454545454547e-05,
+      "loss": 1.4971,
+      "step": 57340
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.154040404040404e-05,
+      "loss": 1.5256,
+      "step": 57350
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1535353535353536e-05,
+      "loss": 1.5262,
+      "step": 57360
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.153030303030303e-05,
+      "loss": 1.5049,
+      "step": 57370
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1525252525252528e-05,
+      "loss": 1.4913,
+      "step": 57380
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.152020202020202e-05,
+      "loss": 1.5121,
+      "step": 57390
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1515151515151516e-05,
+      "loss": 1.5134,
+      "step": 57400
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1510101010101012e-05,
+      "loss": 1.5069,
+      "step": 57410
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1505050505050505e-05,
+      "loss": 1.4814,
+      "step": 57420
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.15e-05,
+      "loss": 1.4994,
+      "step": 57430
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1494949494949497e-05,
+      "loss": 1.5032,
+      "step": 57440
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.148989898989899e-05,
+      "loss": 1.5166,
+      "step": 57450
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1484848484848485e-05,
+      "loss": 1.5034,
+      "step": 57460
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.147979797979798e-05,
+      "loss": 1.5012,
+      "step": 57470
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1474747474747474e-05,
+      "loss": 1.49,
+      "step": 57480
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.146969696969697e-05,
+      "loss": 1.501,
+      "step": 57490
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1464646464646466e-05,
+      "loss": 1.4973,
+      "step": 57500
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.145959595959596e-05,
+      "loss": 1.5141,
+      "step": 57510
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1454545454545455e-05,
+      "loss": 1.5108,
+      "step": 57520
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.144949494949495e-05,
+      "loss": 1.4921,
+      "step": 57530
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2.1444444444444443e-05,
+      "loss": 1.5119,
+      "step": 57540
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.143939393939394e-05,
+      "loss": 1.5052,
+      "step": 57550
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1434343434343435e-05,
+      "loss": 1.507,
+      "step": 57560
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.142929292929293e-05,
+      "loss": 1.4725,
+      "step": 57570
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1424242424242424e-05,
+      "loss": 1.518,
+      "step": 57580
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.141919191919192e-05,
+      "loss": 1.4939,
+      "step": 57590
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1414141414141416e-05,
+      "loss": 1.4916,
+      "step": 57600
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.140909090909091e-05,
+      "loss": 1.5288,
+      "step": 57610
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1404040404040404e-05,
+      "loss": 1.4944,
+      "step": 57620
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.13989898989899e-05,
+      "loss": 1.4969,
+      "step": 57630
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1393939393939393e-05,
+      "loss": 1.4921,
+      "step": 57640
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.138888888888889e-05,
+      "loss": 1.504,
+      "step": 57650
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1383838383838385e-05,
+      "loss": 1.5208,
+      "step": 57660
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1378787878787877e-05,
+      "loss": 1.5098,
+      "step": 57670
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1373737373737373e-05,
+      "loss": 1.5001,
+      "step": 57680
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.136868686868687e-05,
+      "loss": 1.5106,
+      "step": 57690
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1363636363636362e-05,
+      "loss": 1.5153,
+      "step": 57700
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1358585858585858e-05,
+      "loss": 1.4965,
+      "step": 57710
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1353535353535354e-05,
+      "loss": 1.4939,
+      "step": 57720
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1348484848484847e-05,
+      "loss": 1.4849,
+      "step": 57730
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1343434343434343e-05,
+      "loss": 1.4982,
+      "step": 57740
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.133838383838384e-05,
+      "loss": 1.5153,
+      "step": 57750
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1333333333333335e-05,
+      "loss": 1.4869,
+      "step": 57760
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.132828282828283e-05,
+      "loss": 1.4966,
+      "step": 57770
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1323232323232327e-05,
+      "loss": 1.5013,
+      "step": 57780
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.131818181818182e-05,
+      "loss": 1.5212,
+      "step": 57790
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1313131313131315e-05,
+      "loss": 1.508,
+      "step": 57800
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.130808080808081e-05,
+      "loss": 1.5007,
+      "step": 57810
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1303030303030304e-05,
+      "loss": 1.4817,
+      "step": 57820
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.12979797979798e-05,
+      "loss": 1.5094,
+      "step": 57830
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1292929292929296e-05,
+      "loss": 1.4748,
+      "step": 57840
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.128787878787879e-05,
+      "loss": 1.4879,
+      "step": 57850
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1282828282828284e-05,
+      "loss": 1.5109,
+      "step": 57860
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.127777777777778e-05,
+      "loss": 1.4995,
+      "step": 57870
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1272727272727276e-05,
+      "loss": 1.4877,
+      "step": 57880
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.126767676767677e-05,
+      "loss": 1.5354,
+      "step": 57890
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1262626262626265e-05,
+      "loss": 1.5149,
+      "step": 57900
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.125757575757576e-05,
+      "loss": 1.4851,
+      "step": 57910
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1252525252525254e-05,
+      "loss": 1.5092,
+      "step": 57920
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.124747474747475e-05,
+      "loss": 1.5003,
+      "step": 57930
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1242424242424246e-05,
+      "loss": 1.4983,
+      "step": 57940
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1237373737373738e-05,
+      "loss": 1.5018,
+      "step": 57950
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1232323232323234e-05,
+      "loss": 1.5069,
+      "step": 57960
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.122727272727273e-05,
+      "loss": 1.5157,
+      "step": 57970
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1222222222222223e-05,
+      "loss": 1.481,
+      "step": 57980
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.121717171717172e-05,
+      "loss": 1.5119,
+      "step": 57990
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1212121212121215e-05,
+      "loss": 1.5119,
+      "step": 58000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1207070707070707e-05,
+      "loss": 1.478,
+      "step": 58010
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1202020202020203e-05,
+      "loss": 1.4976,
+      "step": 58020
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.11969696969697e-05,
+      "loss": 1.5081,
+      "step": 58030
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1191919191919192e-05,
+      "loss": 1.5118,
+      "step": 58040
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1186868686868688e-05,
+      "loss": 1.5104,
+      "step": 58050
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1181818181818184e-05,
+      "loss": 1.492,
+      "step": 58060
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.117676767676768e-05,
+      "loss": 1.5102,
+      "step": 58070
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1171717171717172e-05,
+      "loss": 1.5161,
+      "step": 58080
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.116666666666667e-05,
+      "loss": 1.5006,
+      "step": 58090
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1161616161616164e-05,
+      "loss": 1.5206,
+      "step": 58100
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1156565656565657e-05,
+      "loss": 1.5081,
+      "step": 58110
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1151515151515153e-05,
+      "loss": 1.4971,
+      "step": 58120
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.114646464646465e-05,
+      "loss": 1.5187,
+      "step": 58130
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.114141414141414e-05,
+      "loss": 1.4897,
+      "step": 58140
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1136363636363638e-05,
+      "loss": 1.5103,
+      "step": 58150
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1131313131313134e-05,
+      "loss": 1.4865,
+      "step": 58160
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1126262626262626e-05,
+      "loss": 1.5048,
+      "step": 58170
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1121212121212122e-05,
+      "loss": 1.4986,
+      "step": 58180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.1116161616161618e-05,
+      "loss": 1.481,
+      "step": 58190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2.111111111111111e-05,
+      "loss": 1.5248,
+      "step": 58200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.1106060606060607e-05,
+      "loss": 1.5064,
+      "step": 58210
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.1101010101010103e-05,
+      "loss": 1.4995,
+      "step": 58220
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.1095959595959595e-05,
+      "loss": 1.5156,
+      "step": 58230
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.109090909090909e-05,
+      "loss": 1.5068,
+      "step": 58240
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.1085858585858587e-05,
+      "loss": 1.4948,
+      "step": 58250
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.108080808080808e-05,
+      "loss": 1.5039,
+      "step": 58260
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.1075757575757576e-05,
+      "loss": 1.513,
+      "step": 58270
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.1070707070707072e-05,
+      "loss": 1.4948,
+      "step": 58280
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.1065656565656568e-05,
+      "loss": 1.4881,
+      "step": 58290
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.106060606060606e-05,
+      "loss": 1.4944,
+      "step": 58300
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.1055555555555556e-05,
+      "loss": 1.5021,
+      "step": 58310
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.1050505050505052e-05,
+      "loss": 1.5183,
+      "step": 58320
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.1045454545454545e-05,
+      "loss": 1.4909,
+      "step": 58330
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.104040404040404e-05,
+      "loss": 1.5003,
+      "step": 58340
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.1035353535353537e-05,
+      "loss": 1.494,
+      "step": 58350
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.103030303030303e-05,
+      "loss": 1.5026,
+      "step": 58360
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.1025252525252526e-05,
+      "loss": 1.5071,
+      "step": 58370
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.102020202020202e-05,
+      "loss": 1.489,
+      "step": 58380
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.1015151515151514e-05,
+      "loss": 1.5134,
+      "step": 58390
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.101010101010101e-05,
+      "loss": 1.5097,
+      "step": 58400
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.1005050505050506e-05,
+      "loss": 1.5252,
+      "step": 58410
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.1e-05,
+      "loss": 1.5019,
+      "step": 58420
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0994949494949495e-05,
+      "loss": 1.4832,
+      "step": 58430
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.098989898989899e-05,
+      "loss": 1.4808,
+      "step": 58440
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0984848484848483e-05,
+      "loss": 1.5027,
+      "step": 58450
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.097979797979798e-05,
+      "loss": 1.4986,
+      "step": 58460
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0974747474747475e-05,
+      "loss": 1.5172,
+      "step": 58470
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.096969696969697e-05,
+      "loss": 1.4971,
+      "step": 58480
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0964646464646464e-05,
+      "loss": 1.491,
+      "step": 58490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.095959595959596e-05,
+      "loss": 1.5038,
+      "step": 58500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0954545454545456e-05,
+      "loss": 1.5104,
+      "step": 58510
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.094949494949495e-05,
+      "loss": 1.506,
+      "step": 58520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0944444444444445e-05,
+      "loss": 1.5049,
+      "step": 58530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.093939393939394e-05,
+      "loss": 1.4924,
+      "step": 58540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0934343434343433e-05,
+      "loss": 1.5146,
+      "step": 58550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.092929292929293e-05,
+      "loss": 1.5211,
+      "step": 58560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0924242424242425e-05,
+      "loss": 1.5001,
+      "step": 58570
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0919191919191918e-05,
+      "loss": 1.5035,
+      "step": 58580
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0914141414141414e-05,
+      "loss": 1.5017,
+      "step": 58590
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.090909090909091e-05,
+      "loss": 1.519,
+      "step": 58600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0904040404040406e-05,
+      "loss": 1.5171,
+      "step": 58610
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0898989898989902e-05,
+      "loss": 1.5209,
+      "step": 58620
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0893939393939398e-05,
+      "loss": 1.5083,
+      "step": 58630
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.088888888888889e-05,
+      "loss": 1.4923,
+      "step": 58640
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0883838383838386e-05,
+      "loss": 1.5013,
+      "step": 58650
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0878787878787882e-05,
+      "loss": 1.515,
+      "step": 58660
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0873737373737375e-05,
+      "loss": 1.5181,
+      "step": 58670
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.086868686868687e-05,
+      "loss": 1.5089,
+      "step": 58680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0863636363636367e-05,
+      "loss": 1.481,
+      "step": 58690
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.085858585858586e-05,
+      "loss": 1.5041,
+      "step": 58700
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0853535353535355e-05,
+      "loss": 1.5261,
+      "step": 58710
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.084848484848485e-05,
+      "loss": 1.5169,
+      "step": 58720
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0843434343434344e-05,
+      "loss": 1.5158,
+      "step": 58730
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.083838383838384e-05,
+      "loss": 1.5017,
+      "step": 58740
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0833333333333336e-05,
+      "loss": 1.4957,
+      "step": 58750
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.082828282828283e-05,
+      "loss": 1.4973,
+      "step": 58760
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0823232323232325e-05,
+      "loss": 1.5032,
+      "step": 58770
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.081818181818182e-05,
+      "loss": 1.5056,
+      "step": 58780
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0813131313131317e-05,
+      "loss": 1.4897,
+      "step": 58790
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.080808080808081e-05,
+      "loss": 1.5235,
+      "step": 58800
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0803030303030305e-05,
+      "loss": 1.5036,
+      "step": 58810
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.07979797979798e-05,
+      "loss": 1.5271,
+      "step": 58820
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0792929292929294e-05,
+      "loss": 1.5032,
+      "step": 58830
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.078787878787879e-05,
+      "loss": 1.4963,
+      "step": 58840
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0782828282828286e-05,
+      "loss": 1.5023,
+      "step": 58850
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.077777777777778e-05,
+      "loss": 1.5043,
+      "step": 58860
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2.0772727272727274e-05,
+      "loss": 1.5116,
+      "step": 58870
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.076767676767677e-05,
+      "loss": 1.4993,
+      "step": 58880
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0762626262626263e-05,
+      "loss": 1.53,
+      "step": 58890
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.075757575757576e-05,
+      "loss": 1.5141,
+      "step": 58900
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0752525252525255e-05,
+      "loss": 1.5088,
+      "step": 58910
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0747474747474748e-05,
+      "loss": 1.5005,
+      "step": 58920
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0742424242424244e-05,
+      "loss": 1.499,
+      "step": 58930
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.073737373737374e-05,
+      "loss": 1.5245,
+      "step": 58940
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0732323232323232e-05,
+      "loss": 1.4942,
+      "step": 58950
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0727272727272728e-05,
+      "loss": 1.5118,
+      "step": 58960
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0722222222222224e-05,
+      "loss": 1.5055,
+      "step": 58970
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0717171717171717e-05,
+      "loss": 1.4785,
+      "step": 58980
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0712121212121213e-05,
+      "loss": 1.5184,
+      "step": 58990
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.070707070707071e-05,
+      "loss": 1.5121,
+      "step": 59000
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0702020202020205e-05,
+      "loss": 1.48,
+      "step": 59010
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0696969696969697e-05,
+      "loss": 1.5099,
+      "step": 59020
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0691919191919193e-05,
+      "loss": 1.5023,
+      "step": 59030
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.068686868686869e-05,
+      "loss": 1.4883,
+      "step": 59040
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0681818181818182e-05,
+      "loss": 1.519,
+      "step": 59050
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0676767676767678e-05,
+      "loss": 1.499,
+      "step": 59060
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0671717171717174e-05,
+      "loss": 1.5015,
+      "step": 59070
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0666666666666666e-05,
+      "loss": 1.4994,
+      "step": 59080
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0661616161616162e-05,
+      "loss": 1.4905,
+      "step": 59090
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.065656565656566e-05,
+      "loss": 1.5242,
+      "step": 59100
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.065151515151515e-05,
+      "loss": 1.5051,
+      "step": 59110
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0646464646464647e-05,
+      "loss": 1.4889,
+      "step": 59120
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0641414141414143e-05,
+      "loss": 1.5047,
+      "step": 59130
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0636363636363636e-05,
+      "loss": 1.5005,
+      "step": 59140
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.063131313131313e-05,
+      "loss": 1.502,
+      "step": 59150
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0626262626262628e-05,
+      "loss": 1.5201,
+      "step": 59160
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.062121212121212e-05,
+      "loss": 1.4861,
+      "step": 59170
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0616161616161616e-05,
+      "loss": 1.4842,
+      "step": 59180
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0611111111111112e-05,
+      "loss": 1.5148,
+      "step": 59190
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0606060606060608e-05,
+      "loss": 1.5091,
+      "step": 59200
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.06010101010101e-05,
+      "loss": 1.4789,
+      "step": 59210
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0595959595959597e-05,
+      "loss": 1.513,
+      "step": 59220
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0590909090909093e-05,
+      "loss": 1.4992,
+      "step": 59230
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0585858585858585e-05,
+      "loss": 1.5066,
+      "step": 59240
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.058080808080808e-05,
+      "loss": 1.5018,
+      "step": 59250
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0575757575757577e-05,
+      "loss": 1.4975,
+      "step": 59260
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.057070707070707e-05,
+      "loss": 1.5092,
+      "step": 59270
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0565656565656566e-05,
+      "loss": 1.4939,
+      "step": 59280
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0560606060606062e-05,
+      "loss": 1.504,
+      "step": 59290
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0555555555555555e-05,
+      "loss": 1.497,
+      "step": 59300
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.055050505050505e-05,
+      "loss": 1.5144,
+      "step": 59310
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0545454545454546e-05,
+      "loss": 1.4859,
+      "step": 59320
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.054040404040404e-05,
+      "loss": 1.5105,
+      "step": 59330
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0535353535353535e-05,
+      "loss": 1.4933,
+      "step": 59340
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.053030303030303e-05,
+      "loss": 1.5114,
+      "step": 59350
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0525252525252524e-05,
+      "loss": 1.4774,
+      "step": 59360
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.052020202020202e-05,
+      "loss": 1.5016,
+      "step": 59370
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0515151515151516e-05,
+      "loss": 1.4981,
+      "step": 59380
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0510101010101008e-05,
+      "loss": 1.4898,
+      "step": 59390
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0505050505050504e-05,
+      "loss": 1.4963,
+      "step": 59400
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.05e-05,
+      "loss": 1.4926,
+      "step": 59410
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0494949494949496e-05,
+      "loss": 1.5035,
+      "step": 59420
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.048989898989899e-05,
+      "loss": 1.5032,
+      "step": 59430
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0484848484848485e-05,
+      "loss": 1.4795,
+      "step": 59440
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.047979797979798e-05,
+      "loss": 1.4985,
+      "step": 59450
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0474747474747473e-05,
+      "loss": 1.502,
+      "step": 59460
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0469696969696973e-05,
+      "loss": 1.4878,
+      "step": 59470
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0464646464646465e-05,
+      "loss": 1.5116,
+      "step": 59480
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.045959595959596e-05,
+      "loss": 1.4988,
+      "step": 59490
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0454545454545457e-05,
+      "loss": 1.4867,
+      "step": 59500
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0449494949494953e-05,
+      "loss": 1.5045,
+      "step": 59510
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0444444444444446e-05,
+      "loss": 1.4962,
+      "step": 59520
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2.0439393939393942e-05,
+      "loss": 1.5011,
+      "step": 59530
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0434343434343438e-05,
+      "loss": 1.5108,
+      "step": 59540
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.042929292929293e-05,
+      "loss": 1.483,
+      "step": 59550
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0424242424242427e-05,
+      "loss": 1.501,
+      "step": 59560
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0419191919191923e-05,
+      "loss": 1.5095,
+      "step": 59570
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0414141414141415e-05,
+      "loss": 1.5047,
+      "step": 59580
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.040909090909091e-05,
+      "loss": 1.5166,
+      "step": 59590
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0404040404040407e-05,
+      "loss": 1.4825,
+      "step": 59600
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.03989898989899e-05,
+      "loss": 1.5034,
+      "step": 59610
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0393939393939396e-05,
+      "loss": 1.5422,
+      "step": 59620
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0388888888888892e-05,
+      "loss": 1.5076,
+      "step": 59630
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0383838383838384e-05,
+      "loss": 1.5039,
+      "step": 59640
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.037878787878788e-05,
+      "loss": 1.5057,
+      "step": 59650
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0373737373737376e-05,
+      "loss": 1.4962,
+      "step": 59660
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.036868686868687e-05,
+      "loss": 1.4867,
+      "step": 59670
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0363636363636365e-05,
+      "loss": 1.5165,
+      "step": 59680
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.035858585858586e-05,
+      "loss": 1.4986,
+      "step": 59690
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0353535353535353e-05,
+      "loss": 1.501,
+      "step": 59700
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.034848484848485e-05,
+      "loss": 1.5182,
+      "step": 59710
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0343434343434345e-05,
+      "loss": 1.5098,
+      "step": 59720
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.033838383838384e-05,
+      "loss": 1.4926,
+      "step": 59730
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0333333333333334e-05,
+      "loss": 1.4826,
+      "step": 59740
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.032828282828283e-05,
+      "loss": 1.5029,
+      "step": 59750
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0323232323232326e-05,
+      "loss": 1.5145,
+      "step": 59760
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.031818181818182e-05,
+      "loss": 1.5059,
+      "step": 59770
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0313131313131315e-05,
+      "loss": 1.5116,
+      "step": 59780
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.030808080808081e-05,
+      "loss": 1.472,
+      "step": 59790
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0303030303030303e-05,
+      "loss": 1.4875,
+      "step": 59800
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.02979797979798e-05,
+      "loss": 1.5101,
+      "step": 59810
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0292929292929295e-05,
+      "loss": 1.5144,
+      "step": 59820
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0287878787878788e-05,
+      "loss": 1.4906,
+      "step": 59830
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0282828282828284e-05,
+      "loss": 1.4808,
+      "step": 59840
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.027777777777778e-05,
+      "loss": 1.5206,
+      "step": 59850
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0272727272727272e-05,
+      "loss": 1.4861,
+      "step": 59860
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.026767676767677e-05,
+      "loss": 1.485,
+      "step": 59870
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0262626262626264e-05,
+      "loss": 1.5113,
+      "step": 59880
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0257575757575757e-05,
+      "loss": 1.5039,
+      "step": 59890
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0252525252525253e-05,
+      "loss": 1.5052,
+      "step": 59900
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.024747474747475e-05,
+      "loss": 1.4794,
+      "step": 59910
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0242424242424245e-05,
+      "loss": 1.4924,
+      "step": 59920
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0237373737373738e-05,
+      "loss": 1.4774,
+      "step": 59930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0232323232323234e-05,
+      "loss": 1.5251,
+      "step": 59940
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.022727272727273e-05,
+      "loss": 1.5033,
+      "step": 59950
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0222222222222222e-05,
+      "loss": 1.5103,
+      "step": 59960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0217171717171718e-05,
+      "loss": 1.4967,
+      "step": 59970
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0212121212121214e-05,
+      "loss": 1.5209,
+      "step": 59980
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0207070707070707e-05,
+      "loss": 1.4886,
+      "step": 59990
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0202020202020203e-05,
+      "loss": 1.5081,
+      "step": 60000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.01969696969697e-05,
+      "loss": 1.5082,
+      "step": 60010
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.019191919191919e-05,
+      "loss": 1.4953,
+      "step": 60020
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0186868686868687e-05,
+      "loss": 1.4899,
+      "step": 60030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0181818181818183e-05,
+      "loss": 1.5031,
+      "step": 60040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0176767676767676e-05,
+      "loss": 1.5099,
+      "step": 60050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0171717171717172e-05,
+      "loss": 1.4879,
+      "step": 60060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0166666666666668e-05,
+      "loss": 1.4864,
+      "step": 60070
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.016161616161616e-05,
+      "loss": 1.4963,
+      "step": 60080
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0156565656565656e-05,
+      "loss": 1.5175,
+      "step": 60090
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0151515151515152e-05,
+      "loss": 1.5206,
+      "step": 60100
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0146464646464645e-05,
+      "loss": 1.4935,
+      "step": 60110
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.014141414141414e-05,
+      "loss": 1.4936,
+      "step": 60120
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0136363636363637e-05,
+      "loss": 1.5122,
+      "step": 60130
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0131313131313133e-05,
+      "loss": 1.501,
+      "step": 60140
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0126262626262626e-05,
+      "loss": 1.5095,
+      "step": 60150
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.012121212121212e-05,
+      "loss": 1.4766,
+      "step": 60160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0116161616161618e-05,
+      "loss": 1.5,
+      "step": 60170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.011111111111111e-05,
+      "loss": 1.4743,
+      "step": 60180
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0106060606060606e-05,
+      "loss": 1.515,
+      "step": 60190
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2.0101010101010102e-05,
+      "loss": 1.4881,
+      "step": 60200
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0095959595959595e-05,
+      "loss": 1.5065,
+      "step": 60210
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.009090909090909e-05,
+      "loss": 1.4966,
+      "step": 60220
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0085858585858587e-05,
+      "loss": 1.5065,
+      "step": 60230
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.008080808080808e-05,
+      "loss": 1.5152,
+      "step": 60240
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0075757575757575e-05,
+      "loss": 1.4856,
+      "step": 60250
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.007070707070707e-05,
+      "loss": 1.5084,
+      "step": 60260
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0065656565656564e-05,
+      "loss": 1.4943,
+      "step": 60270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.006060606060606e-05,
+      "loss": 1.4916,
+      "step": 60280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0055555555555556e-05,
+      "loss": 1.4873,
+      "step": 60290
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.005050505050505e-05,
+      "loss": 1.4953,
+      "step": 60300
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0045454545454544e-05,
+      "loss": 1.5081,
+      "step": 60310
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0040404040404044e-05,
+      "loss": 1.504,
+      "step": 60320
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0035353535353536e-05,
+      "loss": 1.5201,
+      "step": 60330
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0030303030303032e-05,
+      "loss": 1.48,
+      "step": 60340
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.002525252525253e-05,
+      "loss": 1.5109,
+      "step": 60350
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.002020202020202e-05,
+      "loss": 1.4833,
+      "step": 60360
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0015151515151517e-05,
+      "loss": 1.5091,
+      "step": 60370
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0010101010101013e-05,
+      "loss": 1.5113,
+      "step": 60380
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2.0005050505050506e-05,
+      "loss": 1.4997,
+      "step": 60390
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2e-05,
+      "loss": 1.4883,
+      "step": 60400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9994949494949498e-05,
+      "loss": 1.474,
+      "step": 60410
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.998989898989899e-05,
+      "loss": 1.5142,
+      "step": 60420
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9984848484848486e-05,
+      "loss": 1.5047,
+      "step": 60430
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9979797979797982e-05,
+      "loss": 1.5119,
+      "step": 60440
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9974747474747478e-05,
+      "loss": 1.5099,
+      "step": 60450
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.996969696969697e-05,
+      "loss": 1.503,
+      "step": 60460
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9964646464646467e-05,
+      "loss": 1.4851,
+      "step": 60470
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9959595959595963e-05,
+      "loss": 1.4919,
+      "step": 60480
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9954545454545455e-05,
+      "loss": 1.5172,
+      "step": 60490
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.994949494949495e-05,
+      "loss": 1.4989,
+      "step": 60500
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9944444444444447e-05,
+      "loss": 1.4501,
+      "step": 60510
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.993939393939394e-05,
+      "loss": 1.5109,
+      "step": 60520
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9934343434343436e-05,
+      "loss": 1.5003,
+      "step": 60530
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9929292929292932e-05,
+      "loss": 1.5056,
+      "step": 60540
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9924242424242425e-05,
+      "loss": 1.4904,
+      "step": 60550
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.991919191919192e-05,
+      "loss": 1.499,
+      "step": 60560
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9914141414141417e-05,
+      "loss": 1.4683,
+      "step": 60570
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.990909090909091e-05,
+      "loss": 1.5008,
+      "step": 60580
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9904040404040405e-05,
+      "loss": 1.492,
+      "step": 60590
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.98989898989899e-05,
+      "loss": 1.508,
+      "step": 60600
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9893939393939394e-05,
+      "loss": 1.5023,
+      "step": 60610
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.988888888888889e-05,
+      "loss": 1.5196,
+      "step": 60620
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9883838383838386e-05,
+      "loss": 1.5055,
+      "step": 60630
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.987878787878788e-05,
+      "loss": 1.473,
+      "step": 60640
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9873737373737374e-05,
+      "loss": 1.4765,
+      "step": 60650
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.986868686868687e-05,
+      "loss": 1.4961,
+      "step": 60660
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9863636363636366e-05,
+      "loss": 1.4997,
+      "step": 60670
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.985858585858586e-05,
+      "loss": 1.4909,
+      "step": 60680
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9853535353535355e-05,
+      "loss": 1.4964,
+      "step": 60690
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.984848484848485e-05,
+      "loss": 1.4897,
+      "step": 60700
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9843434343434343e-05,
+      "loss": 1.4814,
+      "step": 60710
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.983838383838384e-05,
+      "loss": 1.5012,
+      "step": 60720
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9833333333333335e-05,
+      "loss": 1.4992,
+      "step": 60730
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9828282828282828e-05,
+      "loss": 1.4941,
+      "step": 60740
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9823232323232324e-05,
+      "loss": 1.4871,
+      "step": 60750
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.981818181818182e-05,
+      "loss": 1.4791,
+      "step": 60760
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9813131313131313e-05,
+      "loss": 1.4832,
+      "step": 60770
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.980808080808081e-05,
+      "loss": 1.5115,
+      "step": 60780
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9803030303030305e-05,
+      "loss": 1.488,
+      "step": 60790
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9797979797979797e-05,
+      "loss": 1.5138,
+      "step": 60800
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9792929292929293e-05,
+      "loss": 1.5248,
+      "step": 60810
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.978787878787879e-05,
+      "loss": 1.5013,
+      "step": 60820
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9782828282828285e-05,
+      "loss": 1.5002,
+      "step": 60830
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9777777777777778e-05,
+      "loss": 1.5034,
+      "step": 60840
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.9772727272727274e-05,
+      "loss": 1.5169,
+      "step": 60850
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.976767676767677e-05,
+      "loss": 1.5006,
+      "step": 60860
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9762626262626262e-05,
+      "loss": 1.4999,
+      "step": 60870
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.975757575757576e-05,
+      "loss": 1.4992,
+      "step": 60880
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9752525252525254e-05,
+      "loss": 1.4944,
+      "step": 60890
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9747474747474747e-05,
+      "loss": 1.5051,
+      "step": 60900
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9742424242424243e-05,
+      "loss": 1.5019,
+      "step": 60910
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.973737373737374e-05,
+      "loss": 1.4763,
+      "step": 60920
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.973232323232323e-05,
+      "loss": 1.4745,
+      "step": 60930
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9727272727272728e-05,
+      "loss": 1.4986,
+      "step": 60940
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9722222222222224e-05,
+      "loss": 1.4793,
+      "step": 60950
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9717171717171716e-05,
+      "loss": 1.523,
+      "step": 60960
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9712121212121212e-05,
+      "loss": 1.4893,
+      "step": 60970
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9707070707070708e-05,
+      "loss": 1.4908,
+      "step": 60980
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.97020202020202e-05,
+      "loss": 1.5036,
+      "step": 60990
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9696969696969697e-05,
+      "loss": 1.492,
+      "step": 61000
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9691919191919193e-05,
+      "loss": 1.4892,
+      "step": 61010
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9686868686868685e-05,
+      "loss": 1.4821,
+      "step": 61020
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.968181818181818e-05,
+      "loss": 1.488,
+      "step": 61030
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9676767676767677e-05,
+      "loss": 1.4751,
+      "step": 61040
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9671717171717173e-05,
+      "loss": 1.5047,
+      "step": 61050
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9666666666666666e-05,
+      "loss": 1.5066,
+      "step": 61060
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9661616161616162e-05,
+      "loss": 1.4968,
+      "step": 61070
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9656565656565658e-05,
+      "loss": 1.5273,
+      "step": 61080
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.965151515151515e-05,
+      "loss": 1.5037,
+      "step": 61090
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9646464646464646e-05,
+      "loss": 1.5105,
+      "step": 61100
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9641414141414142e-05,
+      "loss": 1.4875,
+      "step": 61110
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9636363636363635e-05,
+      "loss": 1.5154,
+      "step": 61120
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.963131313131313e-05,
+      "loss": 1.492,
+      "step": 61130
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9626262626262627e-05,
+      "loss": 1.5003,
+      "step": 61140
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.962121212121212e-05,
+      "loss": 1.5015,
+      "step": 61150
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9616161616161616e-05,
+      "loss": 1.5084,
+      "step": 61160
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9611111111111115e-05,
+      "loss": 1.4833,
+      "step": 61170
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9606060606060608e-05,
+      "loss": 1.4737,
+      "step": 61180
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9601010101010104e-05,
+      "loss": 1.5143,
+      "step": 61190
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.95959595959596e-05,
+      "loss": 1.5039,
+      "step": 61200
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9590909090909092e-05,
+      "loss": 1.4858,
+      "step": 61210
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9585858585858588e-05,
+      "loss": 1.497,
+      "step": 61220
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9580808080808084e-05,
+      "loss": 1.5337,
+      "step": 61230
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9575757575757577e-05,
+      "loss": 1.497,
+      "step": 61240
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9570707070707073e-05,
+      "loss": 1.4992,
+      "step": 61250
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.956565656565657e-05,
+      "loss": 1.5003,
+      "step": 61260
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.956060606060606e-05,
+      "loss": 1.4944,
+      "step": 61270
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9555555555555557e-05,
+      "loss": 1.4978,
+      "step": 61280
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9550505050505053e-05,
+      "loss": 1.5,
+      "step": 61290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9545454545454546e-05,
+      "loss": 1.5019,
+      "step": 61300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9540404040404042e-05,
+      "loss": 1.5027,
+      "step": 61310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9535353535353538e-05,
+      "loss": 1.4838,
+      "step": 61320
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.953030303030303e-05,
+      "loss": 1.5027,
+      "step": 61330
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9525252525252526e-05,
+      "loss": 1.488,
+      "step": 61340
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9520202020202022e-05,
+      "loss": 1.5107,
+      "step": 61350
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.951515151515152e-05,
+      "loss": 1.5145,
+      "step": 61360
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.951010101010101e-05,
+      "loss": 1.5053,
+      "step": 61370
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9505050505050507e-05,
+      "loss": 1.4904,
+      "step": 61380
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9500000000000003e-05,
+      "loss": 1.508,
+      "step": 61390
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9494949494949496e-05,
+      "loss": 1.4942,
+      "step": 61400
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.948989898989899e-05,
+      "loss": 1.5211,
+      "step": 61410
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9484848484848488e-05,
+      "loss": 1.4924,
+      "step": 61420
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.947979797979798e-05,
+      "loss": 1.4923,
+      "step": 61430
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9474747474747476e-05,
+      "loss": 1.5026,
+      "step": 61440
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9469696969696972e-05,
+      "loss": 1.5024,
+      "step": 61450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9464646464646465e-05,
+      "loss": 1.5029,
+      "step": 61460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.945959595959596e-05,
+      "loss": 1.5171,
+      "step": 61470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9454545454545457e-05,
+      "loss": 1.5156,
+      "step": 61480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.944949494949495e-05,
+      "loss": 1.5053,
+      "step": 61490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9444444444444445e-05,
+      "loss": 1.4874,
+      "step": 61500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.943939393939394e-05,
+      "loss": 1.5025,
+      "step": 61510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.9434343434343434e-05,
+      "loss": 1.5074,
+      "step": 61520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.942929292929293e-05,
+      "loss": 1.4892,
+      "step": 61530
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9424242424242426e-05,
+      "loss": 1.4891,
+      "step": 61540
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9419191919191922e-05,
+      "loss": 1.5133,
+      "step": 61550
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9414141414141415e-05,
+      "loss": 1.4792,
+      "step": 61560
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.940909090909091e-05,
+      "loss": 1.4813,
+      "step": 61570
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9404040404040407e-05,
+      "loss": 1.4698,
+      "step": 61580
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.93989898989899e-05,
+      "loss": 1.4965,
+      "step": 61590
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9393939393939395e-05,
+      "loss": 1.4959,
+      "step": 61600
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.938888888888889e-05,
+      "loss": 1.5085,
+      "step": 61610
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9383838383838384e-05,
+      "loss": 1.5076,
+      "step": 61620
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.937878787878788e-05,
+      "loss": 1.5062,
+      "step": 61630
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9373737373737376e-05,
+      "loss": 1.4956,
+      "step": 61640
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9368686868686868e-05,
+      "loss": 1.5019,
+      "step": 61650
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9363636363636364e-05,
+      "loss": 1.5084,
+      "step": 61660
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.935858585858586e-05,
+      "loss": 1.4827,
+      "step": 61670
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9353535353535353e-05,
+      "loss": 1.5006,
+      "step": 61680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.934848484848485e-05,
+      "loss": 1.4959,
+      "step": 61690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9343434343434345e-05,
+      "loss": 1.4975,
+      "step": 61700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9338383838383837e-05,
+      "loss": 1.487,
+      "step": 61710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9333333333333333e-05,
+      "loss": 1.4943,
+      "step": 61720
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.932828282828283e-05,
+      "loss": 1.509,
+      "step": 61730
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9323232323232322e-05,
+      "loss": 1.4945,
+      "step": 61740
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9318181818181818e-05,
+      "loss": 1.482,
+      "step": 61750
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9313131313131314e-05,
+      "loss": 1.4996,
+      "step": 61760
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.930808080808081e-05,
+      "loss": 1.4994,
+      "step": 61770
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9303030303030303e-05,
+      "loss": 1.4997,
+      "step": 61780
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.92979797979798e-05,
+      "loss": 1.4836,
+      "step": 61790
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9292929292929295e-05,
+      "loss": 1.4836,
+      "step": 61800
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9287878787878787e-05,
+      "loss": 1.4822,
+      "step": 61810
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9282828282828283e-05,
+      "loss": 1.4854,
+      "step": 61820
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.927777777777778e-05,
+      "loss": 1.4884,
+      "step": 61830
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9272727272727272e-05,
+      "loss": 1.488,
+      "step": 61840
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9267676767676768e-05,
+      "loss": 1.4975,
+      "step": 61850
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9262626262626264e-05,
+      "loss": 1.4896,
+      "step": 61860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9257575757575756e-05,
+      "loss": 1.476,
+      "step": 61870
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9252525252525252e-05,
+      "loss": 1.4866,
+      "step": 61880
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.924747474747475e-05,
+      "loss": 1.4818,
+      "step": 61890
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.924242424242424e-05,
+      "loss": 1.4996,
+      "step": 61900
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9237373737373737e-05,
+      "loss": 1.5068,
+      "step": 61910
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9232323232323233e-05,
+      "loss": 1.507,
+      "step": 61920
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9227272727272726e-05,
+      "loss": 1.4907,
+      "step": 61930
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.922222222222222e-05,
+      "loss": 1.4894,
+      "step": 61940
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9217171717171718e-05,
+      "loss": 1.5092,
+      "step": 61950
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9212121212121213e-05,
+      "loss": 1.4965,
+      "step": 61960
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9207070707070706e-05,
+      "loss": 1.4851,
+      "step": 61970
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9202020202020202e-05,
+      "loss": 1.5119,
+      "step": 61980
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9196969696969698e-05,
+      "loss": 1.4831,
+      "step": 61990
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.919191919191919e-05,
+      "loss": 1.4951,
+      "step": 62000
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9186868686868687e-05,
+      "loss": 1.5149,
+      "step": 62010
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9181818181818183e-05,
+      "loss": 1.5165,
+      "step": 62020
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.917676767676768e-05,
+      "loss": 1.5083,
+      "step": 62030
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9171717171717175e-05,
+      "loss": 1.4993,
+      "step": 62040
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9166666666666667e-05,
+      "loss": 1.4901,
+      "step": 62050
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9161616161616163e-05,
+      "loss": 1.4879,
+      "step": 62060
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.915656565656566e-05,
+      "loss": 1.4911,
+      "step": 62070
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9151515151515155e-05,
+      "loss": 1.4934,
+      "step": 62080
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9146464646464648e-05,
+      "loss": 1.4762,
+      "step": 62090
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9141414141414144e-05,
+      "loss": 1.4764,
+      "step": 62100
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.913636363636364e-05,
+      "loss": 1.5111,
+      "step": 62110
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9131313131313132e-05,
+      "loss": 1.5098,
+      "step": 62120
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.912626262626263e-05,
+      "loss": 1.5034,
+      "step": 62130
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9121212121212124e-05,
+      "loss": 1.4743,
+      "step": 62140
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9116161616161617e-05,
+      "loss": 1.4909,
+      "step": 62150
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9111111111111113e-05,
+      "loss": 1.5066,
+      "step": 62160
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.910606060606061e-05,
+      "loss": 1.4957,
+      "step": 62170
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.91010101010101e-05,
+      "loss": 1.4973,
+      "step": 62180
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9095959595959598e-05,
+      "loss": 1.4944,
+      "step": 62190
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9090909090909094e-05,
+      "loss": 1.5043,
+      "step": 62200
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9085858585858586e-05,
+      "loss": 1.4839,
+      "step": 62210
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9080808080808082e-05,
+      "loss": 1.4888,
+      "step": 62220
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9075757575757578e-05,
+      "loss": 1.4923,
+      "step": 62230
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.907070707070707e-05,
+      "loss": 1.4719,
+      "step": 62240
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9065656565656567e-05,
+      "loss": 1.5122,
+      "step": 62250
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9060606060606063e-05,
+      "loss": 1.5028,
+      "step": 62260
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.905555555555556e-05,
+      "loss": 1.4967,
+      "step": 62270
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.905050505050505e-05,
+      "loss": 1.4856,
+      "step": 62280
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9045454545454547e-05,
+      "loss": 1.4838,
+      "step": 62290
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9040404040404043e-05,
+      "loss": 1.5043,
+      "step": 62300
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9035353535353536e-05,
+      "loss": 1.4847,
+      "step": 62310
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9030303030303032e-05,
+      "loss": 1.4969,
+      "step": 62320
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9025252525252528e-05,
+      "loss": 1.4962,
+      "step": 62330
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.902020202020202e-05,
+      "loss": 1.4857,
+      "step": 62340
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9015151515151516e-05,
+      "loss": 1.4939,
+      "step": 62350
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9010101010101012e-05,
+      "loss": 1.4939,
+      "step": 62360
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9005050505050505e-05,
+      "loss": 1.4829,
+      "step": 62370
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.9e-05,
+      "loss": 1.5054,
+      "step": 62380
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8994949494949497e-05,
+      "loss": 1.5021,
+      "step": 62390
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.898989898989899e-05,
+      "loss": 1.4861,
+      "step": 62400
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8984848484848486e-05,
+      "loss": 1.5064,
+      "step": 62410
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.897979797979798e-05,
+      "loss": 1.5138,
+      "step": 62420
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8974747474747474e-05,
+      "loss": 1.4969,
+      "step": 62430
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.896969696969697e-05,
+      "loss": 1.519,
+      "step": 62440
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8964646464646466e-05,
+      "loss": 1.5182,
+      "step": 62450
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.895959595959596e-05,
+      "loss": 1.4996,
+      "step": 62460
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8954545454545455e-05,
+      "loss": 1.5042,
+      "step": 62470
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.894949494949495e-05,
+      "loss": 1.5021,
+      "step": 62480
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8944444444444447e-05,
+      "loss": 1.5028,
+      "step": 62490
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.893939393939394e-05,
+      "loss": 1.5039,
+      "step": 62500
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8934343434343435e-05,
+      "loss": 1.501,
+      "step": 62510
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.892929292929293e-05,
+      "loss": 1.4817,
+      "step": 62520
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8924242424242424e-05,
+      "loss": 1.4861,
+      "step": 62530
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.891919191919192e-05,
+      "loss": 1.5048,
+      "step": 62540
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8914141414141416e-05,
+      "loss": 1.4839,
+      "step": 62550
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.890909090909091e-05,
+      "loss": 1.4765,
+      "step": 62560
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8904040404040405e-05,
+      "loss": 1.5114,
+      "step": 62570
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.88989898989899e-05,
+      "loss": 1.5179,
+      "step": 62580
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8893939393939393e-05,
+      "loss": 1.5124,
+      "step": 62590
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.888888888888889e-05,
+      "loss": 1.4932,
+      "step": 62600
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8883838383838385e-05,
+      "loss": 1.5081,
+      "step": 62610
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8878787878787878e-05,
+      "loss": 1.5163,
+      "step": 62620
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8873737373737374e-05,
+      "loss": 1.506,
+      "step": 62630
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.886868686868687e-05,
+      "loss": 1.4838,
+      "step": 62640
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8863636363636362e-05,
+      "loss": 1.5013,
+      "step": 62650
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8858585858585858e-05,
+      "loss": 1.5027,
+      "step": 62660
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8853535353535354e-05,
+      "loss": 1.5028,
+      "step": 62670
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.884848484848485e-05,
+      "loss": 1.4971,
+      "step": 62680
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8843434343434343e-05,
+      "loss": 1.5105,
+      "step": 62690
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.883838383838384e-05,
+      "loss": 1.4843,
+      "step": 62700
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8833333333333335e-05,
+      "loss": 1.4974,
+      "step": 62710
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8828282828282827e-05,
+      "loss": 1.4895,
+      "step": 62720
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8823232323232323e-05,
+      "loss": 1.5105,
+      "step": 62730
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.881818181818182e-05,
+      "loss": 1.4892,
+      "step": 62740
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8813131313131312e-05,
+      "loss": 1.5211,
+      "step": 62750
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8808080808080808e-05,
+      "loss": 1.5002,
+      "step": 62760
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8803030303030304e-05,
+      "loss": 1.4826,
+      "step": 62770
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8797979797979797e-05,
+      "loss": 1.4899,
+      "step": 62780
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8792929292929293e-05,
+      "loss": 1.5015,
+      "step": 62790
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.878787878787879e-05,
+      "loss": 1.4799,
+      "step": 62800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.878282828282828e-05,
+      "loss": 1.5026,
+      "step": 62810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8777777777777777e-05,
+      "loss": 1.4892,
+      "step": 62820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8772727272727273e-05,
+      "loss": 1.4765,
+      "step": 62830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8767676767676766e-05,
+      "loss": 1.4955,
+      "step": 62840
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8762626262626262e-05,
+      "loss": 1.4979,
+      "step": 62850
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.8757575757575758e-05,
+      "loss": 1.5113,
+      "step": 62860
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.875252525252525e-05,
+      "loss": 1.4751,
+      "step": 62870
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.874747474747475e-05,
+      "loss": 1.4933,
+      "step": 62880
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8742424242424246e-05,
+      "loss": 1.5076,
+      "step": 62890
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.873737373737374e-05,
+      "loss": 1.4858,
+      "step": 62900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8732323232323234e-05,
+      "loss": 1.4683,
+      "step": 62910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.872727272727273e-05,
+      "loss": 1.5085,
+      "step": 62920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8722222222222223e-05,
+      "loss": 1.4965,
+      "step": 62930
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.871717171717172e-05,
+      "loss": 1.4891,
+      "step": 62940
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8712121212121215e-05,
+      "loss": 1.4863,
+      "step": 62950
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8707070707070707e-05,
+      "loss": 1.5126,
+      "step": 62960
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8702020202020203e-05,
+      "loss": 1.4907,
+      "step": 62970
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.86969696969697e-05,
+      "loss": 1.5153,
+      "step": 62980
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8691919191919195e-05,
+      "loss": 1.5013,
+      "step": 62990
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8686868686868688e-05,
+      "loss": 1.505,
+      "step": 63000
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8681818181818184e-05,
+      "loss": 1.4932,
+      "step": 63010
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.867676767676768e-05,
+      "loss": 1.4853,
+      "step": 63020
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8671717171717173e-05,
+      "loss": 1.5096,
+      "step": 63030
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.866666666666667e-05,
+      "loss": 1.4793,
+      "step": 63040
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8661616161616165e-05,
+      "loss": 1.512,
+      "step": 63050
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8656565656565657e-05,
+      "loss": 1.493,
+      "step": 63060
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8651515151515153e-05,
+      "loss": 1.5126,
+      "step": 63070
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.864646464646465e-05,
+      "loss": 1.4915,
+      "step": 63080
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8641414141414142e-05,
+      "loss": 1.5074,
+      "step": 63090
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8636363636363638e-05,
+      "loss": 1.5092,
+      "step": 63100
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8631313131313134e-05,
+      "loss": 1.4904,
+      "step": 63110
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8626262626262626e-05,
+      "loss": 1.5121,
+      "step": 63120
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8621212121212122e-05,
+      "loss": 1.4799,
+      "step": 63130
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.861616161616162e-05,
+      "loss": 1.4928,
+      "step": 63140
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.861111111111111e-05,
+      "loss": 1.4704,
+      "step": 63150
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8606060606060607e-05,
+      "loss": 1.4834,
+      "step": 63160
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8601010101010103e-05,
+      "loss": 1.5131,
+      "step": 63170
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8595959595959596e-05,
+      "loss": 1.474,
+      "step": 63180
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.859090909090909e-05,
+      "loss": 1.5027,
+      "step": 63190
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8585858585858588e-05,
+      "loss": 1.517,
+      "step": 63200
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8580808080808084e-05,
+      "loss": 1.5019,
+      "step": 63210
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8575757575757576e-05,
+      "loss": 1.4787,
+      "step": 63220
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8570707070707072e-05,
+      "loss": 1.4966,
+      "step": 63230
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8565656565656568e-05,
+      "loss": 1.5104,
+      "step": 63240
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.856060606060606e-05,
+      "loss": 1.4926,
+      "step": 63250
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8555555555555557e-05,
+      "loss": 1.4869,
+      "step": 63260
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8550505050505053e-05,
+      "loss": 1.4965,
+      "step": 63270
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8545454545454545e-05,
+      "loss": 1.5102,
+      "step": 63280
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.854040404040404e-05,
+      "loss": 1.4938,
+      "step": 63290
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8535353535353537e-05,
+      "loss": 1.4755,
+      "step": 63300
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.853030303030303e-05,
+      "loss": 1.5002,
+      "step": 63310
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8525252525252526e-05,
+      "loss": 1.4853,
+      "step": 63320
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8520202020202022e-05,
+      "loss": 1.4857,
+      "step": 63330
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8515151515151514e-05,
+      "loss": 1.4987,
+      "step": 63340
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.851010101010101e-05,
+      "loss": 1.5082,
+      "step": 63350
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8505050505050506e-05,
+      "loss": 1.511,
+      "step": 63360
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.85e-05,
+      "loss": 1.4875,
+      "step": 63370
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8494949494949495e-05,
+      "loss": 1.4897,
+      "step": 63380
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.848989898989899e-05,
+      "loss": 1.4982,
+      "step": 63390
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8484848484848487e-05,
+      "loss": 1.4849,
+      "step": 63400
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.847979797979798e-05,
+      "loss": 1.5083,
+      "step": 63410
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8474747474747476e-05,
+      "loss": 1.5144,
+      "step": 63420
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.846969696969697e-05,
+      "loss": 1.4932,
+      "step": 63430
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8464646464646464e-05,
+      "loss": 1.4917,
+      "step": 63440
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.845959595959596e-05,
+      "loss": 1.49,
+      "step": 63450
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8454545454545456e-05,
+      "loss": 1.4895,
+      "step": 63460
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.844949494949495e-05,
+      "loss": 1.4809,
+      "step": 63470
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8444444444444445e-05,
+      "loss": 1.4928,
+      "step": 63480
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.843939393939394e-05,
+      "loss": 1.4983,
+      "step": 63490
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8434343434343433e-05,
+      "loss": 1.4943,
+      "step": 63500
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.842929292929293e-05,
+      "loss": 1.4818,
+      "step": 63510
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.8424242424242425e-05,
+      "loss": 1.4829,
+      "step": 63520
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8419191919191918e-05,
+      "loss": 1.4966,
+      "step": 63530
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8414141414141414e-05,
+      "loss": 1.4884,
+      "step": 63540
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.840909090909091e-05,
+      "loss": 1.5083,
+      "step": 63550
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8404040404040403e-05,
+      "loss": 1.4995,
+      "step": 63560
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.83989898989899e-05,
+      "loss": 1.4911,
+      "step": 63570
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8393939393939395e-05,
+      "loss": 1.4648,
+      "step": 63580
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.838888888888889e-05,
+      "loss": 1.5158,
+      "step": 63590
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8383838383838383e-05,
+      "loss": 1.4921,
+      "step": 63600
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.837878787878788e-05,
+      "loss": 1.4828,
+      "step": 63610
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8373737373737375e-05,
+      "loss": 1.4953,
+      "step": 63620
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8368686868686868e-05,
+      "loss": 1.4833,
+      "step": 63630
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8363636363636364e-05,
+      "loss": 1.4941,
+      "step": 63640
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.835858585858586e-05,
+      "loss": 1.4822,
+      "step": 63650
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8353535353535352e-05,
+      "loss": 1.4827,
+      "step": 63660
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8348484848484848e-05,
+      "loss": 1.4899,
+      "step": 63670
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8343434343434344e-05,
+      "loss": 1.4932,
+      "step": 63680
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8338383838383837e-05,
+      "loss": 1.4962,
+      "step": 63690
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8333333333333333e-05,
+      "loss": 1.4939,
+      "step": 63700
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.832828282828283e-05,
+      "loss": 1.4825,
+      "step": 63710
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.832323232323232e-05,
+      "loss": 1.5076,
+      "step": 63720
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.831818181818182e-05,
+      "loss": 1.4963,
+      "step": 63730
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8313131313131317e-05,
+      "loss": 1.4994,
+      "step": 63740
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.830808080808081e-05,
+      "loss": 1.5093,
+      "step": 63750
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8303030303030305e-05,
+      "loss": 1.4839,
+      "step": 63760
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.82979797979798e-05,
+      "loss": 1.4718,
+      "step": 63770
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8292929292929294e-05,
+      "loss": 1.499,
+      "step": 63780
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.828787878787879e-05,
+      "loss": 1.4958,
+      "step": 63790
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8282828282828286e-05,
+      "loss": 1.5014,
+      "step": 63800
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.827777777777778e-05,
+      "loss": 1.5108,
+      "step": 63810
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8272727272727275e-05,
+      "loss": 1.4749,
+      "step": 63820
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.826767676767677e-05,
+      "loss": 1.4943,
+      "step": 63830
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8262626262626263e-05,
+      "loss": 1.4648,
+      "step": 63840
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.825757575757576e-05,
+      "loss": 1.4928,
+      "step": 63850
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8252525252525255e-05,
+      "loss": 1.5129,
+      "step": 63860
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8247474747474748e-05,
+      "loss": 1.5129,
+      "step": 63870
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8242424242424244e-05,
+      "loss": 1.4934,
+      "step": 63880
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.823737373737374e-05,
+      "loss": 1.5074,
+      "step": 63890
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8232323232323232e-05,
+      "loss": 1.495,
+      "step": 63900
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.822727272727273e-05,
+      "loss": 1.5147,
+      "step": 63910
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8222222222222224e-05,
+      "loss": 1.5117,
+      "step": 63920
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.821717171717172e-05,
+      "loss": 1.4764,
+      "step": 63930
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8212121212121213e-05,
+      "loss": 1.4806,
+      "step": 63940
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.820707070707071e-05,
+      "loss": 1.4872,
+      "step": 63950
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8202020202020205e-05,
+      "loss": 1.5133,
+      "step": 63960
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8196969696969697e-05,
+      "loss": 1.5,
+      "step": 63970
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8191919191919193e-05,
+      "loss": 1.4839,
+      "step": 63980
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.818686868686869e-05,
+      "loss": 1.4795,
+      "step": 63990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8181818181818182e-05,
+      "loss": 1.4954,
+      "step": 64000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8176767676767678e-05,
+      "loss": 1.4936,
+      "step": 64010
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8171717171717174e-05,
+      "loss": 1.5277,
+      "step": 64020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8166666666666667e-05,
+      "loss": 1.4867,
+      "step": 64030
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8161616161616163e-05,
+      "loss": 1.4942,
+      "step": 64040
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.815656565656566e-05,
+      "loss": 1.4769,
+      "step": 64050
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.815151515151515e-05,
+      "loss": 1.5003,
+      "step": 64060
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8146464646464647e-05,
+      "loss": 1.5075,
+      "step": 64070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8141414141414143e-05,
+      "loss": 1.49,
+      "step": 64080
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8136363636363636e-05,
+      "loss": 1.4883,
+      "step": 64090
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8131313131313132e-05,
+      "loss": 1.491,
+      "step": 64100
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8126262626262628e-05,
+      "loss": 1.4958,
+      "step": 64110
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8121212121212124e-05,
+      "loss": 1.4924,
+      "step": 64120
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8116161616161616e-05,
+      "loss": 1.4877,
+      "step": 64130
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8111111111111112e-05,
+      "loss": 1.4827,
+      "step": 64140
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.810606060606061e-05,
+      "loss": 1.506,
+      "step": 64150
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.81010101010101e-05,
+      "loss": 1.4937,
+      "step": 64160
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8095959595959597e-05,
+      "loss": 1.509,
+      "step": 64170
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8090909090909093e-05,
+      "loss": 1.5008,
+      "step": 64180
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.8085858585858586e-05,
+      "loss": 1.4873,
+      "step": 64190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.808080808080808e-05,
+      "loss": 1.5016,
+      "step": 64200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8075757575757578e-05,
+      "loss": 1.4863,
+      "step": 64210
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.807070707070707e-05,
+      "loss": 1.4992,
+      "step": 64220
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8065656565656566e-05,
+      "loss": 1.488,
+      "step": 64230
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8060606060606062e-05,
+      "loss": 1.4994,
+      "step": 64240
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8055555555555555e-05,
+      "loss": 1.5004,
+      "step": 64250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.805050505050505e-05,
+      "loss": 1.4961,
+      "step": 64260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8045454545454547e-05,
+      "loss": 1.4849,
+      "step": 64270
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.804040404040404e-05,
+      "loss": 1.4877,
+      "step": 64280
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8035353535353535e-05,
+      "loss": 1.4921,
+      "step": 64290
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.803030303030303e-05,
+      "loss": 1.4737,
+      "step": 64300
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8025252525252527e-05,
+      "loss": 1.4785,
+      "step": 64310
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.802020202020202e-05,
+      "loss": 1.5058,
+      "step": 64320
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8015151515151516e-05,
+      "loss": 1.493,
+      "step": 64330
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8010101010101012e-05,
+      "loss": 1.5015,
+      "step": 64340
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8005050505050504e-05,
+      "loss": 1.4817,
+      "step": 64350
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.8e-05,
+      "loss": 1.483,
+      "step": 64360
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7994949494949496e-05,
+      "loss": 1.4817,
+      "step": 64370
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.798989898989899e-05,
+      "loss": 1.5165,
+      "step": 64380
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7984848484848485e-05,
+      "loss": 1.477,
+      "step": 64390
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.797979797979798e-05,
+      "loss": 1.4971,
+      "step": 64400
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7974747474747474e-05,
+      "loss": 1.484,
+      "step": 64410
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.796969696969697e-05,
+      "loss": 1.4915,
+      "step": 64420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7964646464646466e-05,
+      "loss": 1.4958,
+      "step": 64430
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7959595959595958e-05,
+      "loss": 1.4858,
+      "step": 64440
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7954545454545454e-05,
+      "loss": 1.4988,
+      "step": 64450
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.794949494949495e-05,
+      "loss": 1.4991,
+      "step": 64460
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7944444444444443e-05,
+      "loss": 1.5104,
+      "step": 64470
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.793939393939394e-05,
+      "loss": 1.4882,
+      "step": 64480
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7934343434343435e-05,
+      "loss": 1.4974,
+      "step": 64490
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7929292929292927e-05,
+      "loss": 1.4934,
+      "step": 64500
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7924242424242423e-05,
+      "loss": 1.5017,
+      "step": 64510
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.791919191919192e-05,
+      "loss": 1.4984,
+      "step": 64520
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7914141414141415e-05,
+      "loss": 1.5152,
+      "step": 64530
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7909090909090908e-05,
+      "loss": 1.48,
+      "step": 64540
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7904040404040404e-05,
+      "loss": 1.5176,
+      "step": 64550
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.78989898989899e-05,
+      "loss": 1.5098,
+      "step": 64560
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7893939393939393e-05,
+      "loss": 1.5157,
+      "step": 64570
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.788888888888889e-05,
+      "loss": 1.5081,
+      "step": 64580
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7883838383838385e-05,
+      "loss": 1.5022,
+      "step": 64590
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.787878787878788e-05,
+      "loss": 1.5138,
+      "step": 64600
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7873737373737376e-05,
+      "loss": 1.4915,
+      "step": 64610
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7868686868686872e-05,
+      "loss": 1.5303,
+      "step": 64620
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7863636363636365e-05,
+      "loss": 1.5094,
+      "step": 64630
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.785858585858586e-05,
+      "loss": 1.5108,
+      "step": 64640
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7853535353535357e-05,
+      "loss": 1.508,
+      "step": 64650
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.784848484848485e-05,
+      "loss": 1.4875,
+      "step": 64660
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7843434343434346e-05,
+      "loss": 1.4919,
+      "step": 64670
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.783838383838384e-05,
+      "loss": 1.5058,
+      "step": 64680
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7833333333333334e-05,
+      "loss": 1.4884,
+      "step": 64690
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.782828282828283e-05,
+      "loss": 1.477,
+      "step": 64700
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7823232323232326e-05,
+      "loss": 1.492,
+      "step": 64710
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.781818181818182e-05,
+      "loss": 1.5091,
+      "step": 64720
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7813131313131315e-05,
+      "loss": 1.5058,
+      "step": 64730
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.780808080808081e-05,
+      "loss": 1.5186,
+      "step": 64740
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7803030303030303e-05,
+      "loss": 1.4907,
+      "step": 64750
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.77979797979798e-05,
+      "loss": 1.5018,
+      "step": 64760
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7792929292929295e-05,
+      "loss": 1.5079,
+      "step": 64770
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7787878787878788e-05,
+      "loss": 1.5137,
+      "step": 64780
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7782828282828284e-05,
+      "loss": 1.4941,
+      "step": 64790
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.777777777777778e-05,
+      "loss": 1.5069,
+      "step": 64800
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7772727272727273e-05,
+      "loss": 1.4896,
+      "step": 64810
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.776767676767677e-05,
+      "loss": 1.4972,
+      "step": 64820
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7762626262626265e-05,
+      "loss": 1.4969,
+      "step": 64830
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.775757575757576e-05,
+      "loss": 1.499,
+      "step": 64840
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.7752525252525253e-05,
+      "loss": 1.4985,
+      "step": 64850
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.774747474747475e-05,
+      "loss": 1.482,
+      "step": 64860
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7742424242424245e-05,
+      "loss": 1.5002,
+      "step": 64870
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7737373737373738e-05,
+      "loss": 1.485,
+      "step": 64880
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7732323232323234e-05,
+      "loss": 1.492,
+      "step": 64890
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.772727272727273e-05,
+      "loss": 1.504,
+      "step": 64900
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7722222222222222e-05,
+      "loss": 1.4805,
+      "step": 64910
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.771717171717172e-05,
+      "loss": 1.4796,
+      "step": 64920
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7712121212121214e-05,
+      "loss": 1.4834,
+      "step": 64930
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7707070707070707e-05,
+      "loss": 1.4851,
+      "step": 64940
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7702020202020203e-05,
+      "loss": 1.4988,
+      "step": 64950
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.76969696969697e-05,
+      "loss": 1.4847,
+      "step": 64960
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.769191919191919e-05,
+      "loss": 1.4902,
+      "step": 64970
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7686868686868687e-05,
+      "loss": 1.5023,
+      "step": 64980
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7681818181818183e-05,
+      "loss": 1.4754,
+      "step": 64990
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7676767676767676e-05,
+      "loss": 1.5095,
+      "step": 65000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7671717171717172e-05,
+      "loss": 1.4807,
+      "step": 65010
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7666666666666668e-05,
+      "loss": 1.4851,
+      "step": 65020
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7661616161616164e-05,
+      "loss": 1.4766,
+      "step": 65030
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7656565656565657e-05,
+      "loss": 1.4851,
+      "step": 65040
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7651515151515153e-05,
+      "loss": 1.5236,
+      "step": 65050
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.764646464646465e-05,
+      "loss": 1.5062,
+      "step": 65060
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.764141414141414e-05,
+      "loss": 1.487,
+      "step": 65070
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7636363636363637e-05,
+      "loss": 1.5012,
+      "step": 65080
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7631313131313133e-05,
+      "loss": 1.4948,
+      "step": 65090
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7626262626262626e-05,
+      "loss": 1.4963,
+      "step": 65100
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7621212121212122e-05,
+      "loss": 1.4881,
+      "step": 65110
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7616161616161618e-05,
+      "loss": 1.5033,
+      "step": 65120
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.761111111111111e-05,
+      "loss": 1.4854,
+      "step": 65130
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7606060606060606e-05,
+      "loss": 1.4999,
+      "step": 65140
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7601010101010102e-05,
+      "loss": 1.5028,
+      "step": 65150
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7595959595959595e-05,
+      "loss": 1.4971,
+      "step": 65160
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.759090909090909e-05,
+      "loss": 1.4943,
+      "step": 65170
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7585858585858587e-05,
+      "loss": 1.5209,
+      "step": 65180
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.758080808080808e-05,
+      "loss": 1.5064,
+      "step": 65190
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7575757575757576e-05,
+      "loss": 1.4978,
+      "step": 65200
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.757070707070707e-05,
+      "loss": 1.4693,
+      "step": 65210
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7565656565656564e-05,
+      "loss": 1.4715,
+      "step": 65220
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.756060606060606e-05,
+      "loss": 1.4887,
+      "step": 65230
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7555555555555556e-05,
+      "loss": 1.4853,
+      "step": 65240
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7550505050505052e-05,
+      "loss": 1.484,
+      "step": 65250
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7545454545454545e-05,
+      "loss": 1.4975,
+      "step": 65260
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.754040404040404e-05,
+      "loss": 1.4948,
+      "step": 65270
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7535353535353537e-05,
+      "loss": 1.503,
+      "step": 65280
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.753030303030303e-05,
+      "loss": 1.5007,
+      "step": 65290
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7525252525252525e-05,
+      "loss": 1.478,
+      "step": 65300
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.752020202020202e-05,
+      "loss": 1.5106,
+      "step": 65310
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7515151515151514e-05,
+      "loss": 1.4856,
+      "step": 65320
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.751010101010101e-05,
+      "loss": 1.5142,
+      "step": 65330
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7505050505050506e-05,
+      "loss": 1.4839,
+      "step": 65340
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.75e-05,
+      "loss": 1.4937,
+      "step": 65350
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7494949494949494e-05,
+      "loss": 1.482,
+      "step": 65360
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.748989898989899e-05,
+      "loss": 1.4998,
+      "step": 65370
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7484848484848483e-05,
+      "loss": 1.4853,
+      "step": 65380
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.747979797979798e-05,
+      "loss": 1.4956,
+      "step": 65390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7474747474747475e-05,
+      "loss": 1.5154,
+      "step": 65400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7469696969696968e-05,
+      "loss": 1.4887,
+      "step": 65410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7464646464646464e-05,
+      "loss": 1.498,
+      "step": 65420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.745959595959596e-05,
+      "loss": 1.4863,
+      "step": 65430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7454545454545456e-05,
+      "loss": 1.491,
+      "step": 65440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.744949494949495e-05,
+      "loss": 1.5043,
+      "step": 65450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7444444444444448e-05,
+      "loss": 1.4922,
+      "step": 65460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.743939393939394e-05,
+      "loss": 1.4865,
+      "step": 65470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7434343434343436e-05,
+      "loss": 1.5007,
+      "step": 65480
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7429292929292932e-05,
+      "loss": 1.4936,
+      "step": 65490
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7424242424242425e-05,
+      "loss": 1.5061,
+      "step": 65500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.741919191919192e-05,
+      "loss": 1.4986,
+      "step": 65510
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.7414141414141417e-05,
+      "loss": 1.4979,
+      "step": 65520
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.740909090909091e-05,
+      "loss": 1.4952,
+      "step": 65530
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7404040404040405e-05,
+      "loss": 1.5014,
+      "step": 65540
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.73989898989899e-05,
+      "loss": 1.5098,
+      "step": 65550
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7393939393939397e-05,
+      "loss": 1.5142,
+      "step": 65560
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.738888888888889e-05,
+      "loss": 1.4874,
+      "step": 65570
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7383838383838386e-05,
+      "loss": 1.492,
+      "step": 65580
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7378787878787882e-05,
+      "loss": 1.5076,
+      "step": 65590
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7373737373737375e-05,
+      "loss": 1.4796,
+      "step": 65600
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.736868686868687e-05,
+      "loss": 1.4811,
+      "step": 65610
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7363636363636366e-05,
+      "loss": 1.5157,
+      "step": 65620
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.735858585858586e-05,
+      "loss": 1.5093,
+      "step": 65630
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7353535353535355e-05,
+      "loss": 1.4844,
+      "step": 65640
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.734848484848485e-05,
+      "loss": 1.4969,
+      "step": 65650
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7343434343434344e-05,
+      "loss": 1.5096,
+      "step": 65660
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.733838383838384e-05,
+      "loss": 1.4921,
+      "step": 65670
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7333333333333336e-05,
+      "loss": 1.484,
+      "step": 65680
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7328282828282828e-05,
+      "loss": 1.5278,
+      "step": 65690
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7323232323232324e-05,
+      "loss": 1.499,
+      "step": 65700
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.731818181818182e-05,
+      "loss": 1.5027,
+      "step": 65710
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7313131313131313e-05,
+      "loss": 1.4796,
+      "step": 65720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.730808080808081e-05,
+      "loss": 1.4989,
+      "step": 65730
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7303030303030305e-05,
+      "loss": 1.5004,
+      "step": 65740
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.72979797979798e-05,
+      "loss": 1.4835,
+      "step": 65750
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7292929292929293e-05,
+      "loss": 1.4996,
+      "step": 65760
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.728787878787879e-05,
+      "loss": 1.4839,
+      "step": 65770
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7282828282828285e-05,
+      "loss": 1.492,
+      "step": 65780
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7277777777777778e-05,
+      "loss": 1.5175,
+      "step": 65790
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7272727272727274e-05,
+      "loss": 1.4964,
+      "step": 65800
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.726767676767677e-05,
+      "loss": 1.48,
+      "step": 65810
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7262626262626263e-05,
+      "loss": 1.5026,
+      "step": 65820
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.725757575757576e-05,
+      "loss": 1.4984,
+      "step": 65830
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7252525252525255e-05,
+      "loss": 1.4905,
+      "step": 65840
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7247474747474747e-05,
+      "loss": 1.4746,
+      "step": 65850
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7242424242424243e-05,
+      "loss": 1.4714,
+      "step": 65860
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.723737373737374e-05,
+      "loss": 1.4931,
+      "step": 65870
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7232323232323232e-05,
+      "loss": 1.5058,
+      "step": 65880
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7227272727272728e-05,
+      "loss": 1.5061,
+      "step": 65890
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7222222222222224e-05,
+      "loss": 1.4922,
+      "step": 65900
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7217171717171716e-05,
+      "loss": 1.5055,
+      "step": 65910
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7212121212121212e-05,
+      "loss": 1.4919,
+      "step": 65920
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.720707070707071e-05,
+      "loss": 1.4946,
+      "step": 65930
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.72020202020202e-05,
+      "loss": 1.4657,
+      "step": 65940
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7196969696969697e-05,
+      "loss": 1.5041,
+      "step": 65950
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7191919191919193e-05,
+      "loss": 1.4917,
+      "step": 65960
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.718686868686869e-05,
+      "loss": 1.5011,
+      "step": 65970
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.718181818181818e-05,
+      "loss": 1.4796,
+      "step": 65980
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7176767676767677e-05,
+      "loss": 1.4843,
+      "step": 65990
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7171717171717173e-05,
+      "loss": 1.5186,
+      "step": 66000
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7166666666666666e-05,
+      "loss": 1.4785,
+      "step": 66010
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7161616161616162e-05,
+      "loss": 1.4977,
+      "step": 66020
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7156565656565658e-05,
+      "loss": 1.4785,
+      "step": 66030
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.715151515151515e-05,
+      "loss": 1.4933,
+      "step": 66040
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7146464646464647e-05,
+      "loss": 1.4969,
+      "step": 66050
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7141414141414143e-05,
+      "loss": 1.5125,
+      "step": 66060
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7136363636363635e-05,
+      "loss": 1.492,
+      "step": 66070
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.713131313131313e-05,
+      "loss": 1.4971,
+      "step": 66080
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7126262626262627e-05,
+      "loss": 1.4786,
+      "step": 66090
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.712121212121212e-05,
+      "loss": 1.4901,
+      "step": 66100
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7116161616161616e-05,
+      "loss": 1.4726,
+      "step": 66110
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7111111111111112e-05,
+      "loss": 1.4724,
+      "step": 66120
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7106060606060604e-05,
+      "loss": 1.5083,
+      "step": 66130
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.71010101010101e-05,
+      "loss": 1.4959,
+      "step": 66140
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7095959595959596e-05,
+      "loss": 1.4722,
+      "step": 66150
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7090909090909092e-05,
+      "loss": 1.4975,
+      "step": 66160
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.7085858585858585e-05,
+      "loss": 1.4976,
+      "step": 66170
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.708080808080808e-05,
+      "loss": 1.4801,
+      "step": 66180
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7075757575757577e-05,
+      "loss": 1.4996,
+      "step": 66190
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.707070707070707e-05,
+      "loss": 1.4839,
+      "step": 66200
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7065656565656566e-05,
+      "loss": 1.4994,
+      "step": 66210
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.706060606060606e-05,
+      "loss": 1.4905,
+      "step": 66220
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7055555555555554e-05,
+      "loss": 1.4819,
+      "step": 66230
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.705050505050505e-05,
+      "loss": 1.5016,
+      "step": 66240
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7045454545454546e-05,
+      "loss": 1.4879,
+      "step": 66250
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.704040404040404e-05,
+      "loss": 1.484,
+      "step": 66260
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7035353535353535e-05,
+      "loss": 1.517,
+      "step": 66270
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.703030303030303e-05,
+      "loss": 1.4637,
+      "step": 66280
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7025252525252527e-05,
+      "loss": 1.4949,
+      "step": 66290
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7020202020202023e-05,
+      "loss": 1.4922,
+      "step": 66300
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.701515151515152e-05,
+      "loss": 1.4915,
+      "step": 66310
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.701010101010101e-05,
+      "loss": 1.4949,
+      "step": 66320
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7005050505050507e-05,
+      "loss": 1.4819,
+      "step": 66330
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 1.493,
+      "step": 66340
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6994949494949496e-05,
+      "loss": 1.4742,
+      "step": 66350
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6989898989898992e-05,
+      "loss": 1.4791,
+      "step": 66360
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6984848484848488e-05,
+      "loss": 1.4709,
+      "step": 66370
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.697979797979798e-05,
+      "loss": 1.493,
+      "step": 66380
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6974747474747476e-05,
+      "loss": 1.4864,
+      "step": 66390
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6969696969696972e-05,
+      "loss": 1.4786,
+      "step": 66400
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6964646464646465e-05,
+      "loss": 1.4772,
+      "step": 66410
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.695959595959596e-05,
+      "loss": 1.4807,
+      "step": 66420
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6954545454545457e-05,
+      "loss": 1.5071,
+      "step": 66430
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.694949494949495e-05,
+      "loss": 1.4867,
+      "step": 66440
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6944444444444446e-05,
+      "loss": 1.4939,
+      "step": 66450
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.693939393939394e-05,
+      "loss": 1.4698,
+      "step": 66460
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6934343434343438e-05,
+      "loss": 1.4862,
+      "step": 66470
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.692929292929293e-05,
+      "loss": 1.4828,
+      "step": 66480
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6924242424242426e-05,
+      "loss": 1.5254,
+      "step": 66490
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6919191919191922e-05,
+      "loss": 1.4956,
+      "step": 66500
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6914141414141415e-05,
+      "loss": 1.5255,
+      "step": 66510
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.690909090909091e-05,
+      "loss": 1.4849,
+      "step": 66520
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6904040404040407e-05,
+      "loss": 1.5033,
+      "step": 66530
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.68989898989899e-05,
+      "loss": 1.4728,
+      "step": 66540
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6893939393939395e-05,
+      "loss": 1.4884,
+      "step": 66550
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.688888888888889e-05,
+      "loss": 1.4869,
+      "step": 66560
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6883838383838384e-05,
+      "loss": 1.4638,
+      "step": 66570
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.687878787878788e-05,
+      "loss": 1.4798,
+      "step": 66580
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6873737373737376e-05,
+      "loss": 1.5046,
+      "step": 66590
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.686868686868687e-05,
+      "loss": 1.4921,
+      "step": 66600
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6863636363636364e-05,
+      "loss": 1.4877,
+      "step": 66610
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.685858585858586e-05,
+      "loss": 1.4896,
+      "step": 66620
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6853535353535353e-05,
+      "loss": 1.4849,
+      "step": 66630
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.684848484848485e-05,
+      "loss": 1.4963,
+      "step": 66640
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6843434343434345e-05,
+      "loss": 1.472,
+      "step": 66650
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.683838383838384e-05,
+      "loss": 1.5083,
+      "step": 66660
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6833333333333334e-05,
+      "loss": 1.4922,
+      "step": 66670
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.682828282828283e-05,
+      "loss": 1.4764,
+      "step": 66680
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6823232323232326e-05,
+      "loss": 1.4763,
+      "step": 66690
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6818181818181818e-05,
+      "loss": 1.5016,
+      "step": 66700
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6813131313131314e-05,
+      "loss": 1.4874,
+      "step": 66710
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.680808080808081e-05,
+      "loss": 1.4831,
+      "step": 66720
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6803030303030303e-05,
+      "loss": 1.486,
+      "step": 66730
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.67979797979798e-05,
+      "loss": 1.4783,
+      "step": 66740
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6792929292929295e-05,
+      "loss": 1.4771,
+      "step": 66750
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6787878787878787e-05,
+      "loss": 1.4927,
+      "step": 66760
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6782828282828283e-05,
+      "loss": 1.473,
+      "step": 66770
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.677777777777778e-05,
+      "loss": 1.4802,
+      "step": 66780
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6772727272727272e-05,
+      "loss": 1.4879,
+      "step": 66790
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6767676767676768e-05,
+      "loss": 1.4912,
+      "step": 66800
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6762626262626264e-05,
+      "loss": 1.4928,
+      "step": 66810
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6757575757575757e-05,
+      "loss": 1.4703,
+      "step": 66820
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.6752525252525253e-05,
+      "loss": 1.4991,
+      "step": 66830
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.674747474747475e-05,
+      "loss": 1.4772,
+      "step": 66840
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.674242424242424e-05,
+      "loss": 1.4829,
+      "step": 66850
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6737373737373737e-05,
+      "loss": 1.4958,
+      "step": 66860
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6732323232323233e-05,
+      "loss": 1.4849,
+      "step": 66870
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.672727272727273e-05,
+      "loss": 1.505,
+      "step": 66880
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6722222222222222e-05,
+      "loss": 1.4732,
+      "step": 66890
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6717171717171718e-05,
+      "loss": 1.4988,
+      "step": 66900
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6712121212121214e-05,
+      "loss": 1.4758,
+      "step": 66910
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6707070707070706e-05,
+      "loss": 1.4999,
+      "step": 66920
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6702020202020202e-05,
+      "loss": 1.4814,
+      "step": 66930
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6696969696969698e-05,
+      "loss": 1.4689,
+      "step": 66940
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.669191919191919e-05,
+      "loss": 1.4918,
+      "step": 66950
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6686868686868687e-05,
+      "loss": 1.4846,
+      "step": 66960
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6681818181818183e-05,
+      "loss": 1.5103,
+      "step": 66970
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6676767676767675e-05,
+      "loss": 1.4748,
+      "step": 66980
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.667171717171717e-05,
+      "loss": 1.4818,
+      "step": 66990
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6666666666666667e-05,
+      "loss": 1.4906,
+      "step": 67000
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.666161616161616e-05,
+      "loss": 1.469,
+      "step": 67010
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6656565656565656e-05,
+      "loss": 1.4705,
+      "step": 67020
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6651515151515152e-05,
+      "loss": 1.4739,
+      "step": 67030
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6646464646464645e-05,
+      "loss": 1.4826,
+      "step": 67040
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.664141414141414e-05,
+      "loss": 1.4682,
+      "step": 67050
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6636363636363637e-05,
+      "loss": 1.4759,
+      "step": 67060
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6631313131313133e-05,
+      "loss": 1.4931,
+      "step": 67070
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6626262626262625e-05,
+      "loss": 1.4845,
+      "step": 67080
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.662121212121212e-05,
+      "loss": 1.4561,
+      "step": 67090
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6616161616161617e-05,
+      "loss": 1.4663,
+      "step": 67100
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.661111111111111e-05,
+      "loss": 1.4692,
+      "step": 67110
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6606060606060606e-05,
+      "loss": 1.4997,
+      "step": 67120
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6601010101010102e-05,
+      "loss": 1.46,
+      "step": 67130
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6595959595959594e-05,
+      "loss": 1.4866,
+      "step": 67140
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6590909090909094e-05,
+      "loss": 1.475,
+      "step": 67150
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6585858585858586e-05,
+      "loss": 1.5057,
+      "step": 67160
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6580808080808082e-05,
+      "loss": 1.4996,
+      "step": 67170
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.657575757575758e-05,
+      "loss": 1.4938,
+      "step": 67180
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6570707070707074e-05,
+      "loss": 1.4909,
+      "step": 67190
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6565656565656567e-05,
+      "loss": 1.4894,
+      "step": 67200
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6560606060606063e-05,
+      "loss": 1.504,
+      "step": 67210
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.655555555555556e-05,
+      "loss": 1.494,
+      "step": 67220
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.655050505050505e-05,
+      "loss": 1.4958,
+      "step": 67230
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6545454545454548e-05,
+      "loss": 1.4817,
+      "step": 67240
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6540404040404044e-05,
+      "loss": 1.5053,
+      "step": 67250
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6535353535353536e-05,
+      "loss": 1.5056,
+      "step": 67260
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6530303030303032e-05,
+      "loss": 1.4961,
+      "step": 67270
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6525252525252528e-05,
+      "loss": 1.4939,
+      "step": 67280
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.652020202020202e-05,
+      "loss": 1.4999,
+      "step": 67290
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6515151515151517e-05,
+      "loss": 1.501,
+      "step": 67300
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6510101010101013e-05,
+      "loss": 1.4624,
+      "step": 67310
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6505050505050505e-05,
+      "loss": 1.4803,
+      "step": 67320
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.65e-05,
+      "loss": 1.4758,
+      "step": 67330
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6494949494949497e-05,
+      "loss": 1.5045,
+      "step": 67340
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.648989898989899e-05,
+      "loss": 1.4744,
+      "step": 67350
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6484848484848486e-05,
+      "loss": 1.4742,
+      "step": 67360
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6479797979797982e-05,
+      "loss": 1.4881,
+      "step": 67370
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6474747474747478e-05,
+      "loss": 1.4693,
+      "step": 67380
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.646969696969697e-05,
+      "loss": 1.4862,
+      "step": 67390
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6464646464646466e-05,
+      "loss": 1.478,
+      "step": 67400
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6459595959595962e-05,
+      "loss": 1.4828,
+      "step": 67410
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6454545454545455e-05,
+      "loss": 1.4885,
+      "step": 67420
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.644949494949495e-05,
+      "loss": 1.501,
+      "step": 67430
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6444444444444447e-05,
+      "loss": 1.5025,
+      "step": 67440
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.643939393939394e-05,
+      "loss": 1.4927,
+      "step": 67450
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6434343434343436e-05,
+      "loss": 1.5056,
+      "step": 67460
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.642929292929293e-05,
+      "loss": 1.4867,
+      "step": 67470
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6424242424242424e-05,
+      "loss": 1.4771,
+      "step": 67480
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.641919191919192e-05,
+      "loss": 1.4938,
+      "step": 67490
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.6414141414141416e-05,
+      "loss": 1.5169,
+      "step": 67500
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.640909090909091e-05,
+      "loss": 1.4638,
+      "step": 67510
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6404040404040405e-05,
+      "loss": 1.4832,
+      "step": 67520
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.63989898989899e-05,
+      "loss": 1.4779,
+      "step": 67530
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6393939393939393e-05,
+      "loss": 1.4826,
+      "step": 67540
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.638888888888889e-05,
+      "loss": 1.5109,
+      "step": 67550
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6383838383838385e-05,
+      "loss": 1.4868,
+      "step": 67560
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6378787878787878e-05,
+      "loss": 1.4887,
+      "step": 67570
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6373737373737374e-05,
+      "loss": 1.4776,
+      "step": 67580
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.636868686868687e-05,
+      "loss": 1.489,
+      "step": 67590
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6363636363636366e-05,
+      "loss": 1.4786,
+      "step": 67600
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.635858585858586e-05,
+      "loss": 1.4934,
+      "step": 67610
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6353535353535354e-05,
+      "loss": 1.4736,
+      "step": 67620
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.634848484848485e-05,
+      "loss": 1.4648,
+      "step": 67630
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6343434343434343e-05,
+      "loss": 1.5058,
+      "step": 67640
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.633838383838384e-05,
+      "loss": 1.4845,
+      "step": 67650
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6333333333333335e-05,
+      "loss": 1.4834,
+      "step": 67660
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6328282828282828e-05,
+      "loss": 1.4875,
+      "step": 67670
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6323232323232324e-05,
+      "loss": 1.4925,
+      "step": 67680
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.631818181818182e-05,
+      "loss": 1.4823,
+      "step": 67690
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6313131313131312e-05,
+      "loss": 1.504,
+      "step": 67700
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6308080808080808e-05,
+      "loss": 1.4868,
+      "step": 67710
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6303030303030304e-05,
+      "loss": 1.4804,
+      "step": 67720
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6297979797979797e-05,
+      "loss": 1.4949,
+      "step": 67730
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6292929292929293e-05,
+      "loss": 1.4981,
+      "step": 67740
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.628787878787879e-05,
+      "loss": 1.4857,
+      "step": 67750
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.628282828282828e-05,
+      "loss": 1.5147,
+      "step": 67760
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6277777777777777e-05,
+      "loss": 1.4898,
+      "step": 67770
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6272727272727273e-05,
+      "loss": 1.4742,
+      "step": 67780
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.626767676767677e-05,
+      "loss": 1.485,
+      "step": 67790
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6262626262626262e-05,
+      "loss": 1.4996,
+      "step": 67800
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6257575757575758e-05,
+      "loss": 1.5238,
+      "step": 67810
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6252525252525254e-05,
+      "loss": 1.4837,
+      "step": 67820
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6247474747474747e-05,
+      "loss": 1.4945,
+      "step": 67830
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6242424242424243e-05,
+      "loss": 1.4964,
+      "step": 67840
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.623737373737374e-05,
+      "loss": 1.4795,
+      "step": 67850
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.623232323232323e-05,
+      "loss": 1.4926,
+      "step": 67860
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6227272727272727e-05,
+      "loss": 1.4743,
+      "step": 67870
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6222222222222223e-05,
+      "loss": 1.4855,
+      "step": 67880
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6217171717171716e-05,
+      "loss": 1.4916,
+      "step": 67890
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6212121212121212e-05,
+      "loss": 1.5003,
+      "step": 67900
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6207070707070708e-05,
+      "loss": 1.4863,
+      "step": 67910
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.62020202020202e-05,
+      "loss": 1.5038,
+      "step": 67920
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6196969696969696e-05,
+      "loss": 1.506,
+      "step": 67930
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6191919191919192e-05,
+      "loss": 1.4773,
+      "step": 67940
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6186868686868685e-05,
+      "loss": 1.4546,
+      "step": 67950
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.618181818181818e-05,
+      "loss": 1.4852,
+      "step": 67960
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6176767676767677e-05,
+      "loss": 1.5068,
+      "step": 67970
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.617171717171717e-05,
+      "loss": 1.4724,
+      "step": 67980
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6166666666666665e-05,
+      "loss": 1.4902,
+      "step": 67990
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6161616161616165e-05,
+      "loss": 1.5012,
+      "step": 68000
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6156565656565657e-05,
+      "loss": 1.503,
+      "step": 68010
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6151515151515153e-05,
+      "loss": 1.4924,
+      "step": 68020
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.614646464646465e-05,
+      "loss": 1.4757,
+      "step": 68030
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6141414141414142e-05,
+      "loss": 1.485,
+      "step": 68040
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6136363636363638e-05,
+      "loss": 1.4838,
+      "step": 68050
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6131313131313134e-05,
+      "loss": 1.4899,
+      "step": 68060
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6126262626262627e-05,
+      "loss": 1.4685,
+      "step": 68070
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6121212121212123e-05,
+      "loss": 1.4777,
+      "step": 68080
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.611616161616162e-05,
+      "loss": 1.4818,
+      "step": 68090
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6111111111111115e-05,
+      "loss": 1.4759,
+      "step": 68100
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6106060606060607e-05,
+      "loss": 1.4844,
+      "step": 68110
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6101010101010103e-05,
+      "loss": 1.4902,
+      "step": 68120
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.60959595959596e-05,
+      "loss": 1.4745,
+      "step": 68130
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6090909090909092e-05,
+      "loss": 1.4939,
+      "step": 68140
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6085858585858588e-05,
+      "loss": 1.4709,
+      "step": 68150
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6080808080808084e-05,
+      "loss": 1.4962,
+      "step": 68160
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6075757575757576e-05,
+      "loss": 1.501,
+      "step": 68170
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.6070707070707072e-05,
+      "loss": 1.5106,
+      "step": 68180
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.606565656565657e-05,
+      "loss": 1.4788,
+      "step": 68190
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.606060606060606e-05,
+      "loss": 1.4888,
+      "step": 68200
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.6055555555555557e-05,
+      "loss": 1.5053,
+      "step": 68210
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.6050505050505053e-05,
+      "loss": 1.4747,
+      "step": 68220
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.6045454545454546e-05,
+      "loss": 1.4917,
+      "step": 68230
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.604040404040404e-05,
+      "loss": 1.4851,
+      "step": 68240
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.6035353535353538e-05,
+      "loss": 1.4801,
+      "step": 68250
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.603030303030303e-05,
+      "loss": 1.4893,
+      "step": 68260
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.6025252525252526e-05,
+      "loss": 1.4806,
+      "step": 68270
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.6020202020202022e-05,
+      "loss": 1.4857,
+      "step": 68280
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.6015151515151515e-05,
+      "loss": 1.4662,
+      "step": 68290
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.601010101010101e-05,
+      "loss": 1.4752,
+      "step": 68300
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.6005050505050507e-05,
+      "loss": 1.4781,
+      "step": 68310
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 1.4912,
+      "step": 68320
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5994949494949495e-05,
+      "loss": 1.4811,
+      "step": 68330
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.598989898989899e-05,
+      "loss": 1.4965,
+      "step": 68340
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5984848484848487e-05,
+      "loss": 1.4816,
+      "step": 68350
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.597979797979798e-05,
+      "loss": 1.4842,
+      "step": 68360
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5974747474747476e-05,
+      "loss": 1.492,
+      "step": 68370
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5969696969696972e-05,
+      "loss": 1.4566,
+      "step": 68380
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5964646464646464e-05,
+      "loss": 1.5115,
+      "step": 68390
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.595959595959596e-05,
+      "loss": 1.496,
+      "step": 68400
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5954545454545456e-05,
+      "loss": 1.4789,
+      "step": 68410
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.594949494949495e-05,
+      "loss": 1.4927,
+      "step": 68420
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5944444444444445e-05,
+      "loss": 1.4824,
+      "step": 68430
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.593939393939394e-05,
+      "loss": 1.4644,
+      "step": 68440
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5934343434343434e-05,
+      "loss": 1.5089,
+      "step": 68450
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.592929292929293e-05,
+      "loss": 1.5005,
+      "step": 68460
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5924242424242426e-05,
+      "loss": 1.4711,
+      "step": 68470
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5919191919191918e-05,
+      "loss": 1.5075,
+      "step": 68480
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5914141414141414e-05,
+      "loss": 1.4936,
+      "step": 68490
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.590909090909091e-05,
+      "loss": 1.4756,
+      "step": 68500
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5904040404040406e-05,
+      "loss": 1.4903,
+      "step": 68510
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.58989898989899e-05,
+      "loss": 1.4718,
+      "step": 68520
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5893939393939395e-05,
+      "loss": 1.4848,
+      "step": 68530
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.588888888888889e-05,
+      "loss": 1.4725,
+      "step": 68540
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5883838383838383e-05,
+      "loss": 1.4649,
+      "step": 68550
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.587878787878788e-05,
+      "loss": 1.4622,
+      "step": 68560
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5873737373737375e-05,
+      "loss": 1.5,
+      "step": 68570
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5868686868686868e-05,
+      "loss": 1.4704,
+      "step": 68580
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5863636363636364e-05,
+      "loss": 1.4883,
+      "step": 68590
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.585858585858586e-05,
+      "loss": 1.4719,
+      "step": 68600
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5853535353535353e-05,
+      "loss": 1.4872,
+      "step": 68610
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.584848484848485e-05,
+      "loss": 1.4726,
+      "step": 68620
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5843434343434344e-05,
+      "loss": 1.4765,
+      "step": 68630
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5838383838383837e-05,
+      "loss": 1.4781,
+      "step": 68640
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5833333333333333e-05,
+      "loss": 1.49,
+      "step": 68650
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.582828282828283e-05,
+      "loss": 1.4819,
+      "step": 68660
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.582323232323232e-05,
+      "loss": 1.5187,
+      "step": 68670
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5818181818181818e-05,
+      "loss": 1.4964,
+      "step": 68680
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5813131313131314e-05,
+      "loss": 1.5101,
+      "step": 68690
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5808080808080806e-05,
+      "loss": 1.4868,
+      "step": 68700
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5803030303030302e-05,
+      "loss": 1.4891,
+      "step": 68710
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5797979797979798e-05,
+      "loss": 1.497,
+      "step": 68720
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5792929292929294e-05,
+      "loss": 1.5023,
+      "step": 68730
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5787878787878787e-05,
+      "loss": 1.5002,
+      "step": 68740
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5782828282828283e-05,
+      "loss": 1.47,
+      "step": 68750
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.577777777777778e-05,
+      "loss": 1.4993,
+      "step": 68760
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.577272727272727e-05,
+      "loss": 1.4895,
+      "step": 68770
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5767676767676767e-05,
+      "loss": 1.4818,
+      "step": 68780
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5762626262626263e-05,
+      "loss": 1.4878,
+      "step": 68790
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5757575757575756e-05,
+      "loss": 1.4715,
+      "step": 68800
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5752525252525252e-05,
+      "loss": 1.4897,
+      "step": 68810
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5747474747474748e-05,
+      "loss": 1.4982,
+      "step": 68820
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.574242424242424e-05,
+      "loss": 1.493,
+      "step": 68830
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5737373737373737e-05,
+      "loss": 1.4948,
+      "step": 68840
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.5732323232323236e-05,
+      "loss": 1.4552,
+      "step": 68850
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.572727272727273e-05,
+      "loss": 1.5022,
+      "step": 68860
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5722222222222225e-05,
+      "loss": 1.4752,
+      "step": 68870
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.571717171717172e-05,
+      "loss": 1.4879,
+      "step": 68880
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5712121212121213e-05,
+      "loss": 1.4961,
+      "step": 68890
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.570707070707071e-05,
+      "loss": 1.4896,
+      "step": 68900
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5702020202020205e-05,
+      "loss": 1.4594,
+      "step": 68910
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5696969696969698e-05,
+      "loss": 1.472,
+      "step": 68920
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5691919191919194e-05,
+      "loss": 1.4777,
+      "step": 68930
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.568686868686869e-05,
+      "loss": 1.5107,
+      "step": 68940
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5681818181818182e-05,
+      "loss": 1.4798,
+      "step": 68950
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5676767676767678e-05,
+      "loss": 1.4794,
+      "step": 68960
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5671717171717174e-05,
+      "loss": 1.4657,
+      "step": 68970
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5666666666666667e-05,
+      "loss": 1.5015,
+      "step": 68980
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5661616161616163e-05,
+      "loss": 1.5,
+      "step": 68990
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.565656565656566e-05,
+      "loss": 1.4759,
+      "step": 69000
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.565151515151515e-05,
+      "loss": 1.4745,
+      "step": 69010
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5646464646464647e-05,
+      "loss": 1.4944,
+      "step": 69020
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5641414141414143e-05,
+      "loss": 1.4818,
+      "step": 69030
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.563636363636364e-05,
+      "loss": 1.4765,
+      "step": 69040
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5631313131313132e-05,
+      "loss": 1.4719,
+      "step": 69050
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5626262626262628e-05,
+      "loss": 1.4714,
+      "step": 69060
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5621212121212124e-05,
+      "loss": 1.4881,
+      "step": 69070
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5616161616161617e-05,
+      "loss": 1.497,
+      "step": 69080
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5611111111111113e-05,
+      "loss": 1.5038,
+      "step": 69090
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.560606060606061e-05,
+      "loss": 1.484,
+      "step": 69100
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.56010101010101e-05,
+      "loss": 1.4755,
+      "step": 69110
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5595959595959597e-05,
+      "loss": 1.5023,
+      "step": 69120
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5590909090909093e-05,
+      "loss": 1.4747,
+      "step": 69130
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5585858585858586e-05,
+      "loss": 1.4866,
+      "step": 69140
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5580808080808082e-05,
+      "loss": 1.4967,
+      "step": 69150
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5575757575757578e-05,
+      "loss": 1.4765,
+      "step": 69160
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.557070707070707e-05,
+      "loss": 1.4914,
+      "step": 69170
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5565656565656566e-05,
+      "loss": 1.4997,
+      "step": 69180
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5560606060606062e-05,
+      "loss": 1.4956,
+      "step": 69190
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5555555555555555e-05,
+      "loss": 1.4767,
+      "step": 69200
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.555050505050505e-05,
+      "loss": 1.4547,
+      "step": 69210
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5545454545454547e-05,
+      "loss": 1.4754,
+      "step": 69220
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5540404040404043e-05,
+      "loss": 1.466,
+      "step": 69230
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5535353535353536e-05,
+      "loss": 1.4872,
+      "step": 69240
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.553030303030303e-05,
+      "loss": 1.5026,
+      "step": 69250
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5525252525252527e-05,
+      "loss": 1.4876,
+      "step": 69260
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.552020202020202e-05,
+      "loss": 1.5011,
+      "step": 69270
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5515151515151516e-05,
+      "loss": 1.4781,
+      "step": 69280
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5510101010101012e-05,
+      "loss": 1.5016,
+      "step": 69290
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5505050505050505e-05,
+      "loss": 1.497,
+      "step": 69300
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.55e-05,
+      "loss": 1.4828,
+      "step": 69310
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5494949494949497e-05,
+      "loss": 1.4785,
+      "step": 69320
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.548989898989899e-05,
+      "loss": 1.4964,
+      "step": 69330
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5484848484848485e-05,
+      "loss": 1.4878,
+      "step": 69340
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.547979797979798e-05,
+      "loss": 1.497,
+      "step": 69350
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5474747474747474e-05,
+      "loss": 1.4831,
+      "step": 69360
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.546969696969697e-05,
+      "loss": 1.5042,
+      "step": 69370
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5464646464646466e-05,
+      "loss": 1.4906,
+      "step": 69380
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.545959595959596e-05,
+      "loss": 1.4877,
+      "step": 69390
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5454545454545454e-05,
+      "loss": 1.465,
+      "step": 69400
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.544949494949495e-05,
+      "loss": 1.4902,
+      "step": 69410
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5444444444444446e-05,
+      "loss": 1.4751,
+      "step": 69420
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.543939393939394e-05,
+      "loss": 1.4754,
+      "step": 69430
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5434343434343435e-05,
+      "loss": 1.4799,
+      "step": 69440
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.542929292929293e-05,
+      "loss": 1.4785,
+      "step": 69450
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5424242424242424e-05,
+      "loss": 1.4913,
+      "step": 69460
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.541919191919192e-05,
+      "loss": 1.4853,
+      "step": 69470
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5414141414141416e-05,
+      "loss": 1.4845,
+      "step": 69480
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5409090909090908e-05,
+      "loss": 1.4784,
+      "step": 69490
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.5404040404040404e-05,
+      "loss": 1.4849,
+      "step": 69500
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.53989898989899e-05,
+      "loss": 1.5013,
+      "step": 69510
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5393939393939393e-05,
+      "loss": 1.4754,
+      "step": 69520
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.538888888888889e-05,
+      "loss": 1.4731,
+      "step": 69530
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5383838383838385e-05,
+      "loss": 1.5265,
+      "step": 69540
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5378787878787877e-05,
+      "loss": 1.4893,
+      "step": 69550
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5373737373737373e-05,
+      "loss": 1.5235,
+      "step": 69560
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.536868686868687e-05,
+      "loss": 1.5066,
+      "step": 69570
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5363636363636362e-05,
+      "loss": 1.4704,
+      "step": 69580
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5358585858585858e-05,
+      "loss": 1.434,
+      "step": 69590
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5353535353535354e-05,
+      "loss": 1.4909,
+      "step": 69600
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5348484848484847e-05,
+      "loss": 1.4845,
+      "step": 69610
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5343434343434342e-05,
+      "loss": 1.4743,
+      "step": 69620
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.533838383838384e-05,
+      "loss": 1.4888,
+      "step": 69630
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5333333333333334e-05,
+      "loss": 1.4883,
+      "step": 69640
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5328282828282827e-05,
+      "loss": 1.478,
+      "step": 69650
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5323232323232323e-05,
+      "loss": 1.4783,
+      "step": 69660
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.531818181818182e-05,
+      "loss": 1.5129,
+      "step": 69670
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.531313131313131e-05,
+      "loss": 1.4698,
+      "step": 69680
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5308080808080808e-05,
+      "loss": 1.47,
+      "step": 69690
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5303030303030304e-05,
+      "loss": 1.4987,
+      "step": 69700
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.52979797979798e-05,
+      "loss": 1.4986,
+      "step": 69710
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5292929292929296e-05,
+      "loss": 1.4942,
+      "step": 69720
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5287878787878788e-05,
+      "loss": 1.4869,
+      "step": 69730
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5282828282828284e-05,
+      "loss": 1.5013,
+      "step": 69740
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.527777777777778e-05,
+      "loss": 1.505,
+      "step": 69750
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5272727272727276e-05,
+      "loss": 1.4731,
+      "step": 69760
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.526767676767677e-05,
+      "loss": 1.4853,
+      "step": 69770
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5262626262626265e-05,
+      "loss": 1.47,
+      "step": 69780
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5257575757575759e-05,
+      "loss": 1.5083,
+      "step": 69790
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5252525252525255e-05,
+      "loss": 1.5199,
+      "step": 69800
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.524747474747475e-05,
+      "loss": 1.4823,
+      "step": 69810
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5242424242424244e-05,
+      "loss": 1.4893,
+      "step": 69820
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.523737373737374e-05,
+      "loss": 1.4908,
+      "step": 69830
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5232323232323234e-05,
+      "loss": 1.4777,
+      "step": 69840
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5227272727272728e-05,
+      "loss": 1.5048,
+      "step": 69850
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5222222222222224e-05,
+      "loss": 1.4827,
+      "step": 69860
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5217171717171719e-05,
+      "loss": 1.4822,
+      "step": 69870
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5212121212121213e-05,
+      "loss": 1.4842,
+      "step": 69880
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5207070707070709e-05,
+      "loss": 1.4816,
+      "step": 69890
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5202020202020203e-05,
+      "loss": 1.4931,
+      "step": 69900
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5196969696969699e-05,
+      "loss": 1.4793,
+      "step": 69910
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5191919191919193e-05,
+      "loss": 1.4784,
+      "step": 69920
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5186868686868688e-05,
+      "loss": 1.493,
+      "step": 69930
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5181818181818184e-05,
+      "loss": 1.51,
+      "step": 69940
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5176767676767678e-05,
+      "loss": 1.4824,
+      "step": 69950
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5171717171717172e-05,
+      "loss": 1.4762,
+      "step": 69960
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5166666666666668e-05,
+      "loss": 1.4982,
+      "step": 69970
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5161616161616163e-05,
+      "loss": 1.4828,
+      "step": 69980
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5156565656565659e-05,
+      "loss": 1.4842,
+      "step": 69990
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5151515151515153e-05,
+      "loss": 1.4875,
+      "step": 70000
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5146464646464647e-05,
+      "loss": 1.488,
+      "step": 70010
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5141414141414143e-05,
+      "loss": 1.4758,
+      "step": 70020
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5136363636363637e-05,
+      "loss": 1.4799,
+      "step": 70030
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5131313131313132e-05,
+      "loss": 1.4895,
+      "step": 70040
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5126262626262628e-05,
+      "loss": 1.4723,
+      "step": 70050
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5121212121212122e-05,
+      "loss": 1.4977,
+      "step": 70060
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5116161616161616e-05,
+      "loss": 1.4816,
+      "step": 70070
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5111111111111112e-05,
+      "loss": 1.4832,
+      "step": 70080
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5106060606060607e-05,
+      "loss": 1.4903,
+      "step": 70090
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5101010101010103e-05,
+      "loss": 1.4987,
+      "step": 70100
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5095959595959597e-05,
+      "loss": 1.4718,
+      "step": 70110
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5090909090909091e-05,
+      "loss": 1.4968,
+      "step": 70120
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5085858585858587e-05,
+      "loss": 1.4764,
+      "step": 70130
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5080808080808081e-05,
+      "loss": 1.4626,
+      "step": 70140
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5075757575757576e-05,
+      "loss": 1.4665,
+      "step": 70150
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5070707070707072e-05,
+      "loss": 1.4829,
+      "step": 70160
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.5065656565656566e-05,
+      "loss": 1.4803,
+      "step": 70170
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.506060606060606e-05,
+      "loss": 1.494,
+      "step": 70180
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.5055555555555556e-05,
+      "loss": 1.5188,
+      "step": 70190
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.505050505050505e-05,
+      "loss": 1.501,
+      "step": 70200
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.5045454545454547e-05,
+      "loss": 1.4871,
+      "step": 70210
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.5040404040404041e-05,
+      "loss": 1.4892,
+      "step": 70220
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.5035353535353535e-05,
+      "loss": 1.4937,
+      "step": 70230
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.5030303030303031e-05,
+      "loss": 1.5023,
+      "step": 70240
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.5025252525252526e-05,
+      "loss": 1.4939,
+      "step": 70250
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.502020202020202e-05,
+      "loss": 1.4589,
+      "step": 70260
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.5015151515151516e-05,
+      "loss": 1.4839,
+      "step": 70270
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.501010101010101e-05,
+      "loss": 1.4684,
+      "step": 70280
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.5005050505050506e-05,
+      "loss": 1.4906,
+      "step": 70290
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.5e-05,
+      "loss": 1.493,
+      "step": 70300
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4994949494949495e-05,
+      "loss": 1.5049,
+      "step": 70310
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.498989898989899e-05,
+      "loss": 1.4858,
+      "step": 70320
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4984848484848485e-05,
+      "loss": 1.4848,
+      "step": 70330
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.497979797979798e-05,
+      "loss": 1.4727,
+      "step": 70340
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4974747474747475e-05,
+      "loss": 1.461,
+      "step": 70350
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.496969696969697e-05,
+      "loss": 1.4739,
+      "step": 70360
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4964646464646464e-05,
+      "loss": 1.4882,
+      "step": 70370
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.495959595959596e-05,
+      "loss": 1.4969,
+      "step": 70380
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4954545454545454e-05,
+      "loss": 1.4689,
+      "step": 70390
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.494949494949495e-05,
+      "loss": 1.4817,
+      "step": 70400
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4944444444444444e-05,
+      "loss": 1.5039,
+      "step": 70410
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4939393939393939e-05,
+      "loss": 1.4891,
+      "step": 70420
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4934343434343435e-05,
+      "loss": 1.496,
+      "step": 70430
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4929292929292929e-05,
+      "loss": 1.479,
+      "step": 70440
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4924242424242423e-05,
+      "loss": 1.4898,
+      "step": 70450
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.491919191919192e-05,
+      "loss": 1.4863,
+      "step": 70460
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4914141414141414e-05,
+      "loss": 1.5009,
+      "step": 70470
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4909090909090908e-05,
+      "loss": 1.4851,
+      "step": 70480
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4904040404040404e-05,
+      "loss": 1.4928,
+      "step": 70490
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4898989898989898e-05,
+      "loss": 1.4716,
+      "step": 70500
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4893939393939394e-05,
+      "loss": 1.495,
+      "step": 70510
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4888888888888888e-05,
+      "loss": 1.4797,
+      "step": 70520
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4883838383838383e-05,
+      "loss": 1.4804,
+      "step": 70530
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4878787878787879e-05,
+      "loss": 1.4978,
+      "step": 70540
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4873737373737373e-05,
+      "loss": 1.4596,
+      "step": 70550
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.486868686868687e-05,
+      "loss": 1.4952,
+      "step": 70560
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4863636363636365e-05,
+      "loss": 1.4725,
+      "step": 70570
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4858585858585861e-05,
+      "loss": 1.5129,
+      "step": 70580
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4853535353535355e-05,
+      "loss": 1.4952,
+      "step": 70590
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.484848484848485e-05,
+      "loss": 1.4858,
+      "step": 70600
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4843434343434346e-05,
+      "loss": 1.4731,
+      "step": 70610
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.483838383838384e-05,
+      "loss": 1.4901,
+      "step": 70620
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4833333333333336e-05,
+      "loss": 1.4747,
+      "step": 70630
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.482828282828283e-05,
+      "loss": 1.4743,
+      "step": 70640
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4823232323232324e-05,
+      "loss": 1.4887,
+      "step": 70650
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.481818181818182e-05,
+      "loss": 1.4932,
+      "step": 70660
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4813131313131315e-05,
+      "loss": 1.4738,
+      "step": 70670
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4808080808080809e-05,
+      "loss": 1.4935,
+      "step": 70680
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4803030303030305e-05,
+      "loss": 1.4813,
+      "step": 70690
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.47979797979798e-05,
+      "loss": 1.4813,
+      "step": 70700
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4792929292929295e-05,
+      "loss": 1.4801,
+      "step": 70710
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.478787878787879e-05,
+      "loss": 1.4956,
+      "step": 70720
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4782828282828284e-05,
+      "loss": 1.4914,
+      "step": 70730
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.477777777777778e-05,
+      "loss": 1.5004,
+      "step": 70740
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4772727272727274e-05,
+      "loss": 1.487,
+      "step": 70750
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4767676767676769e-05,
+      "loss": 1.4924,
+      "step": 70760
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4762626262626264e-05,
+      "loss": 1.4883,
+      "step": 70770
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4757575757575759e-05,
+      "loss": 1.4805,
+      "step": 70780
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4752525252525253e-05,
+      "loss": 1.4947,
+      "step": 70790
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4747474747474749e-05,
+      "loss": 1.4844,
+      "step": 70800
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4742424242424243e-05,
+      "loss": 1.4756,
+      "step": 70810
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.473737373737374e-05,
+      "loss": 1.4795,
+      "step": 70820
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4732323232323234e-05,
+      "loss": 1.5103,
+      "step": 70830
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.4727272727272728e-05,
+      "loss": 1.4828,
+      "step": 70840
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4722222222222224e-05,
+      "loss": 1.4847,
+      "step": 70850
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4717171717171718e-05,
+      "loss": 1.4955,
+      "step": 70860
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4712121212121213e-05,
+      "loss": 1.4839,
+      "step": 70870
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4707070707070709e-05,
+      "loss": 1.4888,
+      "step": 70880
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4702020202020203e-05,
+      "loss": 1.4986,
+      "step": 70890
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4696969696969697e-05,
+      "loss": 1.4999,
+      "step": 70900
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4691919191919193e-05,
+      "loss": 1.4833,
+      "step": 70910
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4686868686868687e-05,
+      "loss": 1.4799,
+      "step": 70920
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4681818181818183e-05,
+      "loss": 1.4888,
+      "step": 70930
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4676767676767678e-05,
+      "loss": 1.4687,
+      "step": 70940
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4671717171717172e-05,
+      "loss": 1.4767,
+      "step": 70950
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4666666666666668e-05,
+      "loss": 1.443,
+      "step": 70960
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4661616161616162e-05,
+      "loss": 1.4768,
+      "step": 70970
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4656565656565657e-05,
+      "loss": 1.4733,
+      "step": 70980
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4651515151515153e-05,
+      "loss": 1.4868,
+      "step": 70990
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4646464646464647e-05,
+      "loss": 1.4907,
+      "step": 71000
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4641414141414143e-05,
+      "loss": 1.4815,
+      "step": 71010
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4636363636363637e-05,
+      "loss": 1.4899,
+      "step": 71020
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4631313131313131e-05,
+      "loss": 1.4719,
+      "step": 71030
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4626262626262627e-05,
+      "loss": 1.4882,
+      "step": 71040
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4621212121212122e-05,
+      "loss": 1.4787,
+      "step": 71050
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4616161616161616e-05,
+      "loss": 1.4986,
+      "step": 71060
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4611111111111112e-05,
+      "loss": 1.4787,
+      "step": 71070
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4606060606060606e-05,
+      "loss": 1.4642,
+      "step": 71080
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.46010101010101e-05,
+      "loss": 1.484,
+      "step": 71090
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4595959595959597e-05,
+      "loss": 1.4892,
+      "step": 71100
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4590909090909091e-05,
+      "loss": 1.4865,
+      "step": 71110
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4585858585858587e-05,
+      "loss": 1.4958,
+      "step": 71120
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4580808080808081e-05,
+      "loss": 1.473,
+      "step": 71130
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4575757575757575e-05,
+      "loss": 1.4836,
+      "step": 71140
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4570707070707071e-05,
+      "loss": 1.5073,
+      "step": 71150
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4565656565656566e-05,
+      "loss": 1.4781,
+      "step": 71160
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.456060606060606e-05,
+      "loss": 1.4738,
+      "step": 71170
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4555555555555556e-05,
+      "loss": 1.4769,
+      "step": 71180
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.455050505050505e-05,
+      "loss": 1.5027,
+      "step": 71190
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4545454545454545e-05,
+      "loss": 1.4894,
+      "step": 71200
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.454040404040404e-05,
+      "loss": 1.4723,
+      "step": 71210
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4535353535353535e-05,
+      "loss": 1.4544,
+      "step": 71220
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4530303030303031e-05,
+      "loss": 1.4828,
+      "step": 71230
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4525252525252525e-05,
+      "loss": 1.4615,
+      "step": 71240
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.452020202020202e-05,
+      "loss": 1.4963,
+      "step": 71250
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4515151515151516e-05,
+      "loss": 1.4611,
+      "step": 71260
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.451010101010101e-05,
+      "loss": 1.498,
+      "step": 71270
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4505050505050504e-05,
+      "loss": 1.4877,
+      "step": 71280
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.45e-05,
+      "loss": 1.4969,
+      "step": 71290
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4494949494949494e-05,
+      "loss": 1.4869,
+      "step": 71300
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.448989898989899e-05,
+      "loss": 1.4796,
+      "step": 71310
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4484848484848485e-05,
+      "loss": 1.4766,
+      "step": 71320
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4479797979797979e-05,
+      "loss": 1.4693,
+      "step": 71330
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4474747474747475e-05,
+      "loss": 1.4847,
+      "step": 71340
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.446969696969697e-05,
+      "loss": 1.4868,
+      "step": 71350
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4464646464646464e-05,
+      "loss": 1.4825,
+      "step": 71360
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.445959595959596e-05,
+      "loss": 1.4937,
+      "step": 71370
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4454545454545454e-05,
+      "loss": 1.4847,
+      "step": 71380
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4449494949494948e-05,
+      "loss": 1.4797,
+      "step": 71390
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4444444444444444e-05,
+      "loss": 1.4771,
+      "step": 71400
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4439393939393942e-05,
+      "loss": 1.4959,
+      "step": 71410
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4434343434343436e-05,
+      "loss": 1.489,
+      "step": 71420
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4429292929292932e-05,
+      "loss": 1.4794,
+      "step": 71430
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4424242424242426e-05,
+      "loss": 1.4841,
+      "step": 71440
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.441919191919192e-05,
+      "loss": 1.4836,
+      "step": 71450
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4414141414141417e-05,
+      "loss": 1.5034,
+      "step": 71460
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4409090909090911e-05,
+      "loss": 1.467,
+      "step": 71470
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4404040404040405e-05,
+      "loss": 1.4504,
+      "step": 71480
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4398989898989901e-05,
+      "loss": 1.4835,
+      "step": 71490
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.4393939393939396e-05,
+      "loss": 1.5022,
+      "step": 71500
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.438888888888889e-05,
+      "loss": 1.4858,
+      "step": 71510
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4383838383838386e-05,
+      "loss": 1.4956,
+      "step": 71520
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.437878787878788e-05,
+      "loss": 1.4813,
+      "step": 71530
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4373737373737376e-05,
+      "loss": 1.4784,
+      "step": 71540
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.436868686868687e-05,
+      "loss": 1.4762,
+      "step": 71550
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4363636363636365e-05,
+      "loss": 1.5024,
+      "step": 71560
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.435858585858586e-05,
+      "loss": 1.479,
+      "step": 71570
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4353535353535355e-05,
+      "loss": 1.4679,
+      "step": 71580
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.434848484848485e-05,
+      "loss": 1.4755,
+      "step": 71590
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4343434343434345e-05,
+      "loss": 1.4558,
+      "step": 71600
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.433838383838384e-05,
+      "loss": 1.476,
+      "step": 71610
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4333333333333334e-05,
+      "loss": 1.4983,
+      "step": 71620
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.432828282828283e-05,
+      "loss": 1.4881,
+      "step": 71630
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4323232323232324e-05,
+      "loss": 1.4862,
+      "step": 71640
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.431818181818182e-05,
+      "loss": 1.4696,
+      "step": 71650
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4313131313131314e-05,
+      "loss": 1.4661,
+      "step": 71660
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4308080808080809e-05,
+      "loss": 1.4775,
+      "step": 71670
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4303030303030305e-05,
+      "loss": 1.4811,
+      "step": 71680
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4297979797979799e-05,
+      "loss": 1.4892,
+      "step": 71690
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4292929292929293e-05,
+      "loss": 1.4961,
+      "step": 71700
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.428787878787879e-05,
+      "loss": 1.4789,
+      "step": 71710
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4282828282828284e-05,
+      "loss": 1.47,
+      "step": 71720
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.427777777777778e-05,
+      "loss": 1.4799,
+      "step": 71730
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4272727272727274e-05,
+      "loss": 1.4929,
+      "step": 71740
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4267676767676768e-05,
+      "loss": 1.4853,
+      "step": 71750
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4262626262626264e-05,
+      "loss": 1.5049,
+      "step": 71760
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4257575757575758e-05,
+      "loss": 1.4911,
+      "step": 71770
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4252525252525253e-05,
+      "loss": 1.5067,
+      "step": 71780
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4247474747474749e-05,
+      "loss": 1.4925,
+      "step": 71790
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4242424242424243e-05,
+      "loss": 1.4825,
+      "step": 71800
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4237373737373737e-05,
+      "loss": 1.4663,
+      "step": 71810
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4232323232323233e-05,
+      "loss": 1.4962,
+      "step": 71820
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4227272727272728e-05,
+      "loss": 1.4843,
+      "step": 71830
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4222222222222224e-05,
+      "loss": 1.4628,
+      "step": 71840
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4217171717171718e-05,
+      "loss": 1.4748,
+      "step": 71850
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4212121212121212e-05,
+      "loss": 1.5004,
+      "step": 71860
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4207070707070708e-05,
+      "loss": 1.4848,
+      "step": 71870
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4202020202020203e-05,
+      "loss": 1.5005,
+      "step": 71880
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4196969696969697e-05,
+      "loss": 1.4675,
+      "step": 71890
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4191919191919193e-05,
+      "loss": 1.4767,
+      "step": 71900
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4186868686868687e-05,
+      "loss": 1.5051,
+      "step": 71910
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4181818181818181e-05,
+      "loss": 1.4867,
+      "step": 71920
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4176767676767677e-05,
+      "loss": 1.4769,
+      "step": 71930
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4171717171717172e-05,
+      "loss": 1.4816,
+      "step": 71940
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4166666666666668e-05,
+      "loss": 1.4975,
+      "step": 71950
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4161616161616162e-05,
+      "loss": 1.4818,
+      "step": 71960
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4156565656565656e-05,
+      "loss": 1.4533,
+      "step": 71970
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4151515151515152e-05,
+      "loss": 1.4737,
+      "step": 71980
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4146464646464647e-05,
+      "loss": 1.4752,
+      "step": 71990
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4141414141414141e-05,
+      "loss": 1.4788,
+      "step": 72000
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4136363636363637e-05,
+      "loss": 1.4919,
+      "step": 72010
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4131313131313131e-05,
+      "loss": 1.5006,
+      "step": 72020
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4126262626262627e-05,
+      "loss": 1.4843,
+      "step": 72030
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4121212121212121e-05,
+      "loss": 1.4961,
+      "step": 72040
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4116161616161616e-05,
+      "loss": 1.4748,
+      "step": 72050
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4111111111111112e-05,
+      "loss": 1.4646,
+      "step": 72060
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4106060606060606e-05,
+      "loss": 1.4764,
+      "step": 72070
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.41010101010101e-05,
+      "loss": 1.4807,
+      "step": 72080
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4095959595959596e-05,
+      "loss": 1.4928,
+      "step": 72090
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.409090909090909e-05,
+      "loss": 1.4987,
+      "step": 72100
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4085858585858585e-05,
+      "loss": 1.4838,
+      "step": 72110
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4080808080808081e-05,
+      "loss": 1.4801,
+      "step": 72120
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4075757575757575e-05,
+      "loss": 1.4943,
+      "step": 72130
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4070707070707071e-05,
+      "loss": 1.4945,
+      "step": 72140
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4065656565656565e-05,
+      "loss": 1.4707,
+      "step": 72150
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.406060606060606e-05,
+      "loss": 1.4639,
+      "step": 72160
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.4055555555555556e-05,
+      "loss": 1.4584,
+      "step": 72170
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.405050505050505e-05,
+      "loss": 1.4988,
+      "step": 72180
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4045454545454544e-05,
+      "loss": 1.4901,
+      "step": 72190
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.404040404040404e-05,
+      "loss": 1.4957,
+      "step": 72200
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4035353535353535e-05,
+      "loss": 1.4711,
+      "step": 72210
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4030303030303029e-05,
+      "loss": 1.4899,
+      "step": 72220
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4025252525252525e-05,
+      "loss": 1.4574,
+      "step": 72230
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.402020202020202e-05,
+      "loss": 1.4956,
+      "step": 72240
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4015151515151515e-05,
+      "loss": 1.4927,
+      "step": 72250
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.401010101010101e-05,
+      "loss": 1.4902,
+      "step": 72260
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4005050505050507e-05,
+      "loss": 1.4831,
+      "step": 72270
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.4000000000000001e-05,
+      "loss": 1.4859,
+      "step": 72280
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3994949494949497e-05,
+      "loss": 1.4764,
+      "step": 72290
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3989898989898992e-05,
+      "loss": 1.4814,
+      "step": 72300
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3984848484848486e-05,
+      "loss": 1.4968,
+      "step": 72310
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3979797979797982e-05,
+      "loss": 1.4831,
+      "step": 72320
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3974747474747476e-05,
+      "loss": 1.4693,
+      "step": 72330
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3969696969696972e-05,
+      "loss": 1.4599,
+      "step": 72340
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3964646464646467e-05,
+      "loss": 1.4668,
+      "step": 72350
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3959595959595961e-05,
+      "loss": 1.5001,
+      "step": 72360
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3954545454545457e-05,
+      "loss": 1.4716,
+      "step": 72370
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3949494949494951e-05,
+      "loss": 1.4838,
+      "step": 72380
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3944444444444446e-05,
+      "loss": 1.4765,
+      "step": 72390
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3939393939393942e-05,
+      "loss": 1.5267,
+      "step": 72400
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3934343434343436e-05,
+      "loss": 1.4768,
+      "step": 72410
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.392929292929293e-05,
+      "loss": 1.4954,
+      "step": 72420
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3924242424242426e-05,
+      "loss": 1.4705,
+      "step": 72430
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.391919191919192e-05,
+      "loss": 1.4773,
+      "step": 72440
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3914141414141416e-05,
+      "loss": 1.4834,
+      "step": 72450
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.390909090909091e-05,
+      "loss": 1.4916,
+      "step": 72460
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3904040404040405e-05,
+      "loss": 1.4748,
+      "step": 72470
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3898989898989901e-05,
+      "loss": 1.4829,
+      "step": 72480
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3893939393939395e-05,
+      "loss": 1.4959,
+      "step": 72490
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.388888888888889e-05,
+      "loss": 1.4841,
+      "step": 72500
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3883838383838386e-05,
+      "loss": 1.4943,
+      "step": 72510
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.387878787878788e-05,
+      "loss": 1.4807,
+      "step": 72520
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3873737373737374e-05,
+      "loss": 1.467,
+      "step": 72530
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.386868686868687e-05,
+      "loss": 1.4675,
+      "step": 72540
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3863636363636364e-05,
+      "loss": 1.4967,
+      "step": 72550
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.385858585858586e-05,
+      "loss": 1.4885,
+      "step": 72560
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3853535353535355e-05,
+      "loss": 1.4758,
+      "step": 72570
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3848484848484849e-05,
+      "loss": 1.4779,
+      "step": 72580
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3843434343434345e-05,
+      "loss": 1.5001,
+      "step": 72590
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.383838383838384e-05,
+      "loss": 1.4926,
+      "step": 72600
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3833333333333334e-05,
+      "loss": 1.4713,
+      "step": 72610
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.382828282828283e-05,
+      "loss": 1.4694,
+      "step": 72620
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3823232323232324e-05,
+      "loss": 1.4794,
+      "step": 72630
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3818181818181818e-05,
+      "loss": 1.4971,
+      "step": 72640
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3813131313131314e-05,
+      "loss": 1.4794,
+      "step": 72650
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3808080808080808e-05,
+      "loss": 1.4605,
+      "step": 72660
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3803030303030304e-05,
+      "loss": 1.4729,
+      "step": 72670
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3797979797979799e-05,
+      "loss": 1.4717,
+      "step": 72680
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3792929292929293e-05,
+      "loss": 1.4979,
+      "step": 72690
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3787878787878789e-05,
+      "loss": 1.4812,
+      "step": 72700
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3782828282828283e-05,
+      "loss": 1.4727,
+      "step": 72710
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3777777777777778e-05,
+      "loss": 1.4625,
+      "step": 72720
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3772727272727274e-05,
+      "loss": 1.4758,
+      "step": 72730
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3767676767676768e-05,
+      "loss": 1.5007,
+      "step": 72740
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3762626262626264e-05,
+      "loss": 1.4645,
+      "step": 72750
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3757575757575758e-05,
+      "loss": 1.4748,
+      "step": 72760
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3752525252525252e-05,
+      "loss": 1.4734,
+      "step": 72770
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3747474747474748e-05,
+      "loss": 1.478,
+      "step": 72780
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3742424242424243e-05,
+      "loss": 1.4714,
+      "step": 72790
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3737373737373737e-05,
+      "loss": 1.4784,
+      "step": 72800
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3732323232323233e-05,
+      "loss": 1.4762,
+      "step": 72810
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3727272727272727e-05,
+      "loss": 1.4781,
+      "step": 72820
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3722222222222222e-05,
+      "loss": 1.4843,
+      "step": 72830
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.3717171717171718e-05,
+      "loss": 1.4813,
+      "step": 72840
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3712121212121212e-05,
+      "loss": 1.4844,
+      "step": 72850
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3707070707070708e-05,
+      "loss": 1.4751,
+      "step": 72860
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3702020202020202e-05,
+      "loss": 1.4871,
+      "step": 72870
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3696969696969697e-05,
+      "loss": 1.4736,
+      "step": 72880
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3691919191919193e-05,
+      "loss": 1.4941,
+      "step": 72890
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3686868686868687e-05,
+      "loss": 1.4913,
+      "step": 72900
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3681818181818181e-05,
+      "loss": 1.489,
+      "step": 72910
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3676767676767677e-05,
+      "loss": 1.4704,
+      "step": 72920
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3671717171717171e-05,
+      "loss": 1.4818,
+      "step": 72930
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3666666666666666e-05,
+      "loss": 1.47,
+      "step": 72940
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3661616161616162e-05,
+      "loss": 1.487,
+      "step": 72950
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3656565656565656e-05,
+      "loss": 1.4778,
+      "step": 72960
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3651515151515152e-05,
+      "loss": 1.4625,
+      "step": 72970
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3646464646464646e-05,
+      "loss": 1.4617,
+      "step": 72980
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.364141414141414e-05,
+      "loss": 1.4981,
+      "step": 72990
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3636363636363637e-05,
+      "loss": 1.5098,
+      "step": 73000
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3631313131313131e-05,
+      "loss": 1.4779,
+      "step": 73010
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3626262626262625e-05,
+      "loss": 1.4822,
+      "step": 73020
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3621212121212121e-05,
+      "loss": 1.477,
+      "step": 73030
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3616161616161615e-05,
+      "loss": 1.4766,
+      "step": 73040
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3611111111111111e-05,
+      "loss": 1.4811,
+      "step": 73050
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3606060606060606e-05,
+      "loss": 1.4672,
+      "step": 73060
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.36010101010101e-05,
+      "loss": 1.5002,
+      "step": 73070
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3595959595959596e-05,
+      "loss": 1.4787,
+      "step": 73080
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.359090909090909e-05,
+      "loss": 1.4875,
+      "step": 73090
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3585858585858585e-05,
+      "loss": 1.4655,
+      "step": 73100
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.358080808080808e-05,
+      "loss": 1.4867,
+      "step": 73110
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3575757575757578e-05,
+      "loss": 1.4908,
+      "step": 73120
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3570707070707073e-05,
+      "loss": 1.4719,
+      "step": 73130
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3565656565656567e-05,
+      "loss": 1.4807,
+      "step": 73140
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3560606060606063e-05,
+      "loss": 1.4638,
+      "step": 73150
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3555555555555557e-05,
+      "loss": 1.4664,
+      "step": 73160
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3550505050505053e-05,
+      "loss": 1.4848,
+      "step": 73170
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3545454545454547e-05,
+      "loss": 1.4763,
+      "step": 73180
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3540404040404042e-05,
+      "loss": 1.471,
+      "step": 73190
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3535353535353538e-05,
+      "loss": 1.4724,
+      "step": 73200
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3530303030303032e-05,
+      "loss": 1.4949,
+      "step": 73210
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3525252525252526e-05,
+      "loss": 1.4876,
+      "step": 73220
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3520202020202022e-05,
+      "loss": 1.4649,
+      "step": 73230
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3515151515151517e-05,
+      "loss": 1.4682,
+      "step": 73240
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3510101010101011e-05,
+      "loss": 1.493,
+      "step": 73250
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3505050505050507e-05,
+      "loss": 1.4908,
+      "step": 73260
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3500000000000001e-05,
+      "loss": 1.4701,
+      "step": 73270
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3494949494949497e-05,
+      "loss": 1.4844,
+      "step": 73280
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3489898989898991e-05,
+      "loss": 1.495,
+      "step": 73290
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3484848484848486e-05,
+      "loss": 1.4714,
+      "step": 73300
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3479797979797982e-05,
+      "loss": 1.4756,
+      "step": 73310
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3474747474747476e-05,
+      "loss": 1.4751,
+      "step": 73320
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.346969696969697e-05,
+      "loss": 1.4711,
+      "step": 73330
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3464646464646466e-05,
+      "loss": 1.4797,
+      "step": 73340
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.345959595959596e-05,
+      "loss": 1.4831,
+      "step": 73350
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3454545454545457e-05,
+      "loss": 1.4781,
+      "step": 73360
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3449494949494951e-05,
+      "loss": 1.4847,
+      "step": 73370
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3444444444444445e-05,
+      "loss": 1.4899,
+      "step": 73380
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3439393939393941e-05,
+      "loss": 1.4775,
+      "step": 73390
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3434343434343436e-05,
+      "loss": 1.4983,
+      "step": 73400
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.342929292929293e-05,
+      "loss": 1.4814,
+      "step": 73410
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3424242424242426e-05,
+      "loss": 1.456,
+      "step": 73420
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.341919191919192e-05,
+      "loss": 1.4737,
+      "step": 73430
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3414141414141414e-05,
+      "loss": 1.4948,
+      "step": 73440
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.340909090909091e-05,
+      "loss": 1.4675,
+      "step": 73450
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3404040404040405e-05,
+      "loss": 1.4855,
+      "step": 73460
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.33989898989899e-05,
+      "loss": 1.495,
+      "step": 73470
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3393939393939395e-05,
+      "loss": 1.4908,
+      "step": 73480
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.338888888888889e-05,
+      "loss": 1.4817,
+      "step": 73490
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.3383838383838385e-05,
+      "loss": 1.4931,
+      "step": 73500
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.337878787878788e-05,
+      "loss": 1.4726,
+      "step": 73510
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3373737373737374e-05,
+      "loss": 1.4767,
+      "step": 73520
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.336868686868687e-05,
+      "loss": 1.4554,
+      "step": 73530
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3363636363636364e-05,
+      "loss": 1.5036,
+      "step": 73540
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3358585858585858e-05,
+      "loss": 1.4652,
+      "step": 73550
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3353535353535354e-05,
+      "loss": 1.4879,
+      "step": 73560
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3348484848484849e-05,
+      "loss": 1.4738,
+      "step": 73570
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3343434343434345e-05,
+      "loss": 1.4827,
+      "step": 73580
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3338383838383839e-05,
+      "loss": 1.4792,
+      "step": 73590
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 1.4719,
+      "step": 73600
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.332828282828283e-05,
+      "loss": 1.4824,
+      "step": 73610
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3323232323232324e-05,
+      "loss": 1.4856,
+      "step": 73620
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3318181818181818e-05,
+      "loss": 1.4977,
+      "step": 73630
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3313131313131314e-05,
+      "loss": 1.4817,
+      "step": 73640
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3308080808080808e-05,
+      "loss": 1.4627,
+      "step": 73650
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3303030303030302e-05,
+      "loss": 1.4802,
+      "step": 73660
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3297979797979798e-05,
+      "loss": 1.475,
+      "step": 73670
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3292929292929293e-05,
+      "loss": 1.4845,
+      "step": 73680
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3287878787878789e-05,
+      "loss": 1.4919,
+      "step": 73690
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3282828282828283e-05,
+      "loss": 1.4985,
+      "step": 73700
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3277777777777777e-05,
+      "loss": 1.4608,
+      "step": 73710
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3272727272727273e-05,
+      "loss": 1.47,
+      "step": 73720
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3267676767676768e-05,
+      "loss": 1.4702,
+      "step": 73730
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3262626262626262e-05,
+      "loss": 1.4862,
+      "step": 73740
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3257575757575758e-05,
+      "loss": 1.483,
+      "step": 73750
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3252525252525252e-05,
+      "loss": 1.4739,
+      "step": 73760
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3247474747474748e-05,
+      "loss": 1.4756,
+      "step": 73770
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3242424242424242e-05,
+      "loss": 1.4837,
+      "step": 73780
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3237373737373737e-05,
+      "loss": 1.4872,
+      "step": 73790
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3232323232323233e-05,
+      "loss": 1.4897,
+      "step": 73800
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3227272727272727e-05,
+      "loss": 1.4629,
+      "step": 73810
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3222222222222221e-05,
+      "loss": 1.4798,
+      "step": 73820
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3217171717171717e-05,
+      "loss": 1.4863,
+      "step": 73830
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3212121212121212e-05,
+      "loss": 1.4909,
+      "step": 73840
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3207070707070706e-05,
+      "loss": 1.4562,
+      "step": 73850
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3202020202020202e-05,
+      "loss": 1.4441,
+      "step": 73860
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3196969696969696e-05,
+      "loss": 1.4847,
+      "step": 73870
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3191919191919192e-05,
+      "loss": 1.4962,
+      "step": 73880
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3186868686868687e-05,
+      "loss": 1.4926,
+      "step": 73890
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.318181818181818e-05,
+      "loss": 1.4807,
+      "step": 73900
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3176767676767677e-05,
+      "loss": 1.4925,
+      "step": 73910
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3171717171717171e-05,
+      "loss": 1.4555,
+      "step": 73920
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3166666666666665e-05,
+      "loss": 1.4736,
+      "step": 73930
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3161616161616161e-05,
+      "loss": 1.4819,
+      "step": 73940
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3156565656565656e-05,
+      "loss": 1.4695,
+      "step": 73950
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.315151515151515e-05,
+      "loss": 1.477,
+      "step": 73960
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3146464646464648e-05,
+      "loss": 1.4799,
+      "step": 73970
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3141414141414144e-05,
+      "loss": 1.4789,
+      "step": 73980
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3136363636363638e-05,
+      "loss": 1.486,
+      "step": 73990
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3131313131313134e-05,
+      "loss": 1.486,
+      "step": 74000
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3126262626262628e-05,
+      "loss": 1.4869,
+      "step": 74010
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3121212121212123e-05,
+      "loss": 1.4841,
+      "step": 74020
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3116161616161619e-05,
+      "loss": 1.4636,
+      "step": 74030
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3111111111111113e-05,
+      "loss": 1.4869,
+      "step": 74040
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3106060606060607e-05,
+      "loss": 1.4798,
+      "step": 74050
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3101010101010103e-05,
+      "loss": 1.4735,
+      "step": 74060
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3095959595959597e-05,
+      "loss": 1.4951,
+      "step": 74070
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3090909090909093e-05,
+      "loss": 1.4735,
+      "step": 74080
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3085858585858588e-05,
+      "loss": 1.4663,
+      "step": 74090
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3080808080808082e-05,
+      "loss": 1.4861,
+      "step": 74100
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3075757575757578e-05,
+      "loss": 1.4711,
+      "step": 74110
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3070707070707072e-05,
+      "loss": 1.475,
+      "step": 74120
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3065656565656567e-05,
+      "loss": 1.4877,
+      "step": 74130
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3060606060606063e-05,
+      "loss": 1.4838,
+      "step": 74140
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3055555555555557e-05,
+      "loss": 1.499,
+      "step": 74150
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3050505050505051e-05,
+      "loss": 1.4762,
+      "step": 74160
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.3045454545454547e-05,
+      "loss": 1.4907,
+      "step": 74170
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3040404040404041e-05,
+      "loss": 1.4929,
+      "step": 74180
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3035353535353537e-05,
+      "loss": 1.4641,
+      "step": 74190
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3030303030303032e-05,
+      "loss": 1.481,
+      "step": 74200
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3025252525252526e-05,
+      "loss": 1.4762,
+      "step": 74210
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3020202020202022e-05,
+      "loss": 1.4876,
+      "step": 74220
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3015151515151516e-05,
+      "loss": 1.4847,
+      "step": 74230
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.301010101010101e-05,
+      "loss": 1.474,
+      "step": 74240
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3005050505050507e-05,
+      "loss": 1.472,
+      "step": 74250
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 1.4801,
+      "step": 74260
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2994949494949495e-05,
+      "loss": 1.4598,
+      "step": 74270
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2989898989898991e-05,
+      "loss": 1.4697,
+      "step": 74280
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2984848484848485e-05,
+      "loss": 1.4698,
+      "step": 74290
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2979797979797981e-05,
+      "loss": 1.4621,
+      "step": 74300
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2974747474747476e-05,
+      "loss": 1.4959,
+      "step": 74310
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.296969696969697e-05,
+      "loss": 1.4757,
+      "step": 74320
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2964646464646466e-05,
+      "loss": 1.4853,
+      "step": 74330
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.295959595959596e-05,
+      "loss": 1.4844,
+      "step": 74340
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2954545454545455e-05,
+      "loss": 1.4659,
+      "step": 74350
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.294949494949495e-05,
+      "loss": 1.4962,
+      "step": 74360
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2944444444444445e-05,
+      "loss": 1.4718,
+      "step": 74370
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.293939393939394e-05,
+      "loss": 1.4692,
+      "step": 74380
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2934343434343435e-05,
+      "loss": 1.4598,
+      "step": 74390
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.292929292929293e-05,
+      "loss": 1.4844,
+      "step": 74400
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2924242424242426e-05,
+      "loss": 1.4786,
+      "step": 74410
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.291919191919192e-05,
+      "loss": 1.4968,
+      "step": 74420
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2914141414141414e-05,
+      "loss": 1.4765,
+      "step": 74430
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.290909090909091e-05,
+      "loss": 1.5122,
+      "step": 74440
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2904040404040404e-05,
+      "loss": 1.4543,
+      "step": 74450
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2898989898989899e-05,
+      "loss": 1.4959,
+      "step": 74460
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2893939393939395e-05,
+      "loss": 1.4943,
+      "step": 74470
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2888888888888889e-05,
+      "loss": 1.4949,
+      "step": 74480
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2883838383838385e-05,
+      "loss": 1.4848,
+      "step": 74490
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.287878787878788e-05,
+      "loss": 1.4822,
+      "step": 74500
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2873737373737374e-05,
+      "loss": 1.4569,
+      "step": 74510
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.286868686868687e-05,
+      "loss": 1.4776,
+      "step": 74520
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2863636363636364e-05,
+      "loss": 1.4788,
+      "step": 74530
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2858585858585858e-05,
+      "loss": 1.4708,
+      "step": 74540
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2853535353535354e-05,
+      "loss": 1.5001,
+      "step": 74550
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2848484848484848e-05,
+      "loss": 1.4809,
+      "step": 74560
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2843434343434343e-05,
+      "loss": 1.4804,
+      "step": 74570
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2838383838383839e-05,
+      "loss": 1.4546,
+      "step": 74580
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2833333333333333e-05,
+      "loss": 1.4679,
+      "step": 74590
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2828282828282829e-05,
+      "loss": 1.4937,
+      "step": 74600
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2823232323232323e-05,
+      "loss": 1.481,
+      "step": 74610
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2818181818181818e-05,
+      "loss": 1.4736,
+      "step": 74620
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2813131313131314e-05,
+      "loss": 1.4962,
+      "step": 74630
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2808080808080808e-05,
+      "loss": 1.4836,
+      "step": 74640
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2803030303030302e-05,
+      "loss": 1.4772,
+      "step": 74650
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2797979797979798e-05,
+      "loss": 1.4814,
+      "step": 74660
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2792929292929292e-05,
+      "loss": 1.4873,
+      "step": 74670
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2787878787878787e-05,
+      "loss": 1.4897,
+      "step": 74680
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2782828282828283e-05,
+      "loss": 1.4879,
+      "step": 74690
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2777777777777777e-05,
+      "loss": 1.4758,
+      "step": 74700
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2772727272727273e-05,
+      "loss": 1.4654,
+      "step": 74710
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2767676767676767e-05,
+      "loss": 1.4745,
+      "step": 74720
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2762626262626262e-05,
+      "loss": 1.4791,
+      "step": 74730
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2757575757575758e-05,
+      "loss": 1.4855,
+      "step": 74740
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2752525252525252e-05,
+      "loss": 1.4843,
+      "step": 74750
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2747474747474746e-05,
+      "loss": 1.4735,
+      "step": 74760
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2742424242424242e-05,
+      "loss": 1.4541,
+      "step": 74770
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2737373737373736e-05,
+      "loss": 1.4821,
+      "step": 74780
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2732323232323232e-05,
+      "loss": 1.4693,
+      "step": 74790
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2727272727272727e-05,
+      "loss": 1.4863,
+      "step": 74800
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2722222222222221e-05,
+      "loss": 1.4742,
+      "step": 74810
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2717171717171717e-05,
+      "loss": 1.4877,
+      "step": 74820
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2712121212121215e-05,
+      "loss": 1.4913,
+      "step": 74830
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2707070707070709e-05,
+      "loss": 1.4697,
+      "step": 74840
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2702020202020203e-05,
+      "loss": 1.497,
+      "step": 74850
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.26969696969697e-05,
+      "loss": 1.4728,
+      "step": 74860
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2691919191919194e-05,
+      "loss": 1.4904,
+      "step": 74870
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2686868686868688e-05,
+      "loss": 1.4907,
+      "step": 74880
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2681818181818184e-05,
+      "loss": 1.4826,
+      "step": 74890
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2676767676767678e-05,
+      "loss": 1.4925,
+      "step": 74900
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2671717171717174e-05,
+      "loss": 1.4742,
+      "step": 74910
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2666666666666668e-05,
+      "loss": 1.4919,
+      "step": 74920
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2661616161616163e-05,
+      "loss": 1.4876,
+      "step": 74930
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2656565656565659e-05,
+      "loss": 1.4932,
+      "step": 74940
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2651515151515153e-05,
+      "loss": 1.4915,
+      "step": 74950
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2646464646464647e-05,
+      "loss": 1.4677,
+      "step": 74960
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2641414141414143e-05,
+      "loss": 1.4867,
+      "step": 74970
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2636363636363638e-05,
+      "loss": 1.4787,
+      "step": 74980
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2631313131313132e-05,
+      "loss": 1.4837,
+      "step": 74990
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2626262626262628e-05,
+      "loss": 1.4923,
+      "step": 75000
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2621212121212122e-05,
+      "loss": 1.4737,
+      "step": 75010
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2616161616161618e-05,
+      "loss": 1.494,
+      "step": 75020
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2611111111111113e-05,
+      "loss": 1.4759,
+      "step": 75030
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2606060606060607e-05,
+      "loss": 1.4647,
+      "step": 75040
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2601010101010103e-05,
+      "loss": 1.4544,
+      "step": 75050
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2595959595959597e-05,
+      "loss": 1.4832,
+      "step": 75060
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2590909090909091e-05,
+      "loss": 1.5091,
+      "step": 75070
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2585858585858587e-05,
+      "loss": 1.4636,
+      "step": 75080
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2580808080808082e-05,
+      "loss": 1.4934,
+      "step": 75090
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2575757575757578e-05,
+      "loss": 1.4804,
+      "step": 75100
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2570707070707072e-05,
+      "loss": 1.4862,
+      "step": 75110
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2565656565656566e-05,
+      "loss": 1.4857,
+      "step": 75120
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2560606060606062e-05,
+      "loss": 1.4827,
+      "step": 75130
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2555555555555557e-05,
+      "loss": 1.4724,
+      "step": 75140
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2550505050505051e-05,
+      "loss": 1.4748,
+      "step": 75150
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2545454545454547e-05,
+      "loss": 1.4791,
+      "step": 75160
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2540404040404041e-05,
+      "loss": 1.4665,
+      "step": 75170
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2535353535353535e-05,
+      "loss": 1.4782,
+      "step": 75180
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2530303030303031e-05,
+      "loss": 1.4797,
+      "step": 75190
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2525252525252526e-05,
+      "loss": 1.4858,
+      "step": 75200
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2520202020202022e-05,
+      "loss": 1.4953,
+      "step": 75210
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2515151515151516e-05,
+      "loss": 1.4996,
+      "step": 75220
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.251010101010101e-05,
+      "loss": 1.4658,
+      "step": 75230
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2505050505050506e-05,
+      "loss": 1.5238,
+      "step": 75240
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.25e-05,
+      "loss": 1.4813,
+      "step": 75250
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2494949494949495e-05,
+      "loss": 1.4855,
+      "step": 75260
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2489898989898991e-05,
+      "loss": 1.4741,
+      "step": 75270
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2484848484848485e-05,
+      "loss": 1.4709,
+      "step": 75280
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.247979797979798e-05,
+      "loss": 1.4763,
+      "step": 75290
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2474747474747475e-05,
+      "loss": 1.481,
+      "step": 75300
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.246969696969697e-05,
+      "loss": 1.4849,
+      "step": 75310
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2464646464646466e-05,
+      "loss": 1.4623,
+      "step": 75320
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.245959595959596e-05,
+      "loss": 1.4721,
+      "step": 75330
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2454545454545454e-05,
+      "loss": 1.4541,
+      "step": 75340
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.244949494949495e-05,
+      "loss": 1.4737,
+      "step": 75350
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2444444444444445e-05,
+      "loss": 1.4602,
+      "step": 75360
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2439393939393939e-05,
+      "loss": 1.4959,
+      "step": 75370
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2434343434343435e-05,
+      "loss": 1.4936,
+      "step": 75380
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.242929292929293e-05,
+      "loss": 1.49,
+      "step": 75390
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2424242424242424e-05,
+      "loss": 1.4772,
+      "step": 75400
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.241919191919192e-05,
+      "loss": 1.4686,
+      "step": 75410
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2414141414141414e-05,
+      "loss": 1.4858,
+      "step": 75420
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.240909090909091e-05,
+      "loss": 1.4663,
+      "step": 75430
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2404040404040404e-05,
+      "loss": 1.4898,
+      "step": 75440
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2398989898989898e-05,
+      "loss": 1.4748,
+      "step": 75450
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2393939393939394e-05,
+      "loss": 1.4831,
+      "step": 75460
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.238888888888889e-05,
+      "loss": 1.4669,
+      "step": 75470
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2383838383838385e-05,
+      "loss": 1.4881,
+      "step": 75480
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.237878787878788e-05,
+      "loss": 1.4641,
+      "step": 75490
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2373737373737375e-05,
+      "loss": 1.4815,
+      "step": 75500
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.236868686868687e-05,
+      "loss": 1.4796,
+      "step": 75510
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2363636363636365e-05,
+      "loss": 1.4555,
+      "step": 75520
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.235858585858586e-05,
+      "loss": 1.4885,
+      "step": 75530
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2353535353535354e-05,
+      "loss": 1.4855,
+      "step": 75540
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.234848484848485e-05,
+      "loss": 1.4931,
+      "step": 75550
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2343434343434344e-05,
+      "loss": 1.4598,
+      "step": 75560
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.233838383838384e-05,
+      "loss": 1.4742,
+      "step": 75570
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2333333333333334e-05,
+      "loss": 1.494,
+      "step": 75580
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2328282828282829e-05,
+      "loss": 1.4805,
+      "step": 75590
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2323232323232325e-05,
+      "loss": 1.4827,
+      "step": 75600
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2318181818181819e-05,
+      "loss": 1.4776,
+      "step": 75610
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2313131313131313e-05,
+      "loss": 1.4708,
+      "step": 75620
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.230808080808081e-05,
+      "loss": 1.4743,
+      "step": 75630
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2303030303030304e-05,
+      "loss": 1.4884,
+      "step": 75640
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2297979797979798e-05,
+      "loss": 1.4683,
+      "step": 75650
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2292929292929294e-05,
+      "loss": 1.455,
+      "step": 75660
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2287878787878788e-05,
+      "loss": 1.4895,
+      "step": 75670
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2282828282828284e-05,
+      "loss": 1.4686,
+      "step": 75680
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2277777777777778e-05,
+      "loss": 1.4749,
+      "step": 75690
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2272727272727273e-05,
+      "loss": 1.4709,
+      "step": 75700
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2267676767676769e-05,
+      "loss": 1.4716,
+      "step": 75710
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2262626262626263e-05,
+      "loss": 1.4849,
+      "step": 75720
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2257575757575757e-05,
+      "loss": 1.4961,
+      "step": 75730
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2252525252525253e-05,
+      "loss": 1.4914,
+      "step": 75740
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2247474747474748e-05,
+      "loss": 1.4773,
+      "step": 75750
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2242424242424242e-05,
+      "loss": 1.4778,
+      "step": 75760
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2237373737373738e-05,
+      "loss": 1.4883,
+      "step": 75770
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2232323232323232e-05,
+      "loss": 1.4895,
+      "step": 75780
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2227272727272728e-05,
+      "loss": 1.4864,
+      "step": 75790
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2222222222222222e-05,
+      "loss": 1.4654,
+      "step": 75800
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2217171717171717e-05,
+      "loss": 1.4794,
+      "step": 75810
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2212121212121213e-05,
+      "loss": 1.4907,
+      "step": 75820
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2207070707070707e-05,
+      "loss": 1.4723,
+      "step": 75830
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2202020202020201e-05,
+      "loss": 1.4943,
+      "step": 75840
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2196969696969697e-05,
+      "loss": 1.4841,
+      "step": 75850
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2191919191919192e-05,
+      "loss": 1.4895,
+      "step": 75860
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2186868686868688e-05,
+      "loss": 1.4779,
+      "step": 75870
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2181818181818182e-05,
+      "loss": 1.4765,
+      "step": 75880
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2176767676767676e-05,
+      "loss": 1.4854,
+      "step": 75890
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2171717171717172e-05,
+      "loss": 1.4693,
+      "step": 75900
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2166666666666668e-05,
+      "loss": 1.4788,
+      "step": 75910
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2161616161616162e-05,
+      "loss": 1.4691,
+      "step": 75920
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2156565656565658e-05,
+      "loss": 1.4877,
+      "step": 75930
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2151515151515153e-05,
+      "loss": 1.4946,
+      "step": 75940
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2146464646464647e-05,
+      "loss": 1.4695,
+      "step": 75950
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2141414141414143e-05,
+      "loss": 1.5068,
+      "step": 75960
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2136363636363637e-05,
+      "loss": 1.4752,
+      "step": 75970
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2131313131313132e-05,
+      "loss": 1.4821,
+      "step": 75980
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2126262626262628e-05,
+      "loss": 1.4573,
+      "step": 75990
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2121212121212122e-05,
+      "loss": 1.4791,
+      "step": 76000
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2116161616161616e-05,
+      "loss": 1.4632,
+      "step": 76010
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2111111111111112e-05,
+      "loss": 1.4799,
+      "step": 76020
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2106060606060607e-05,
+      "loss": 1.4929,
+      "step": 76030
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2101010101010103e-05,
+      "loss": 1.4694,
+      "step": 76040
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2095959595959597e-05,
+      "loss": 1.4654,
+      "step": 76050
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2090909090909091e-05,
+      "loss": 1.4836,
+      "step": 76060
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2085858585858587e-05,
+      "loss": 1.4748,
+      "step": 76070
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2080808080808081e-05,
+      "loss": 1.4799,
+      "step": 76080
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2075757575757576e-05,
+      "loss": 1.4732,
+      "step": 76090
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2070707070707072e-05,
+      "loss": 1.4601,
+      "step": 76100
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2065656565656566e-05,
+      "loss": 1.4726,
+      "step": 76110
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2060606060606062e-05,
+      "loss": 1.4663,
+      "step": 76120
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2055555555555556e-05,
+      "loss": 1.4943,
+      "step": 76130
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.205050505050505e-05,
+      "loss": 1.4773,
+      "step": 76140
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2045454545454547e-05,
+      "loss": 1.4364,
+      "step": 76150
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2040404040404041e-05,
+      "loss": 1.4876,
+      "step": 76160
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2035353535353535e-05,
+      "loss": 1.5189,
+      "step": 76170
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2030303030303031e-05,
+      "loss": 1.4916,
+      "step": 76180
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2025252525252525e-05,
+      "loss": 1.4854,
+      "step": 76190
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.202020202020202e-05,
+      "loss": 1.4937,
+      "step": 76200
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2015151515151516e-05,
+      "loss": 1.4888,
+      "step": 76210
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.201010101010101e-05,
+      "loss": 1.4793,
+      "step": 76220
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2005050505050506e-05,
+      "loss": 1.4885,
+      "step": 76230
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2e-05,
+      "loss": 1.4918,
+      "step": 76240
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1994949494949495e-05,
+      "loss": 1.4895,
+      "step": 76250
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.198989898989899e-05,
+      "loss": 1.508,
+      "step": 76260
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1984848484848485e-05,
+      "loss": 1.484,
+      "step": 76270
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.197979797979798e-05,
+      "loss": 1.4703,
+      "step": 76280
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1974747474747475e-05,
+      "loss": 1.4562,
+      "step": 76290
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.196969696969697e-05,
+      "loss": 1.4672,
+      "step": 76300
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1964646464646464e-05,
+      "loss": 1.4696,
+      "step": 76310
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1959595959595961e-05,
+      "loss": 1.4997,
+      "step": 76320
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1954545454545456e-05,
+      "loss": 1.4944,
+      "step": 76330
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.194949494949495e-05,
+      "loss": 1.4687,
+      "step": 76340
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1944444444444446e-05,
+      "loss": 1.4971,
+      "step": 76350
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.193939393939394e-05,
+      "loss": 1.4828,
+      "step": 76360
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1934343434343435e-05,
+      "loss": 1.4912,
+      "step": 76370
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.192929292929293e-05,
+      "loss": 1.496,
+      "step": 76380
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1924242424242425e-05,
+      "loss": 1.4716,
+      "step": 76390
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1919191919191921e-05,
+      "loss": 1.4781,
+      "step": 76400
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1914141414141415e-05,
+      "loss": 1.4646,
+      "step": 76410
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.190909090909091e-05,
+      "loss": 1.4749,
+      "step": 76420
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1904040404040405e-05,
+      "loss": 1.4822,
+      "step": 76430
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.18989898989899e-05,
+      "loss": 1.4701,
+      "step": 76440
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1893939393939394e-05,
+      "loss": 1.4694,
+      "step": 76450
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.188888888888889e-05,
+      "loss": 1.4925,
+      "step": 76460
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1883838383838384e-05,
+      "loss": 1.4904,
+      "step": 76470
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.187878787878788e-05,
+      "loss": 1.4654,
+      "step": 76480
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1873737373737375e-05,
+      "loss": 1.5075,
+      "step": 76490
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1868686868686869e-05,
+      "loss": 1.482,
+      "step": 76500
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1863636363636365e-05,
+      "loss": 1.4758,
+      "step": 76510
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.185858585858586e-05,
+      "loss": 1.4692,
+      "step": 76520
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1853535353535354e-05,
+      "loss": 1.4732,
+      "step": 76530
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.184848484848485e-05,
+      "loss": 1.4817,
+      "step": 76540
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1843434343434344e-05,
+      "loss": 1.457,
+      "step": 76550
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1838383838383838e-05,
+      "loss": 1.5034,
+      "step": 76560
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1833333333333334e-05,
+      "loss": 1.494,
+      "step": 76570
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1828282828282828e-05,
+      "loss": 1.4994,
+      "step": 76580
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1823232323232324e-05,
+      "loss": 1.4888,
+      "step": 76590
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1818181818181819e-05,
+      "loss": 1.4978,
+      "step": 76600
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1813131313131313e-05,
+      "loss": 1.4809,
+      "step": 76610
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1808080808080809e-05,
+      "loss": 1.4861,
+      "step": 76620
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1803030303030303e-05,
+      "loss": 1.4971,
+      "step": 76630
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1797979797979798e-05,
+      "loss": 1.4863,
+      "step": 76640
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1792929292929294e-05,
+      "loss": 1.4761,
+      "step": 76650
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1787878787878788e-05,
+      "loss": 1.4745,
+      "step": 76660
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1782828282828282e-05,
+      "loss": 1.4741,
+      "step": 76670
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1777777777777778e-05,
+      "loss": 1.463,
+      "step": 76680
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1772727272727272e-05,
+      "loss": 1.4745,
+      "step": 76690
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1767676767676768e-05,
+      "loss": 1.4693,
+      "step": 76700
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1762626262626263e-05,
+      "loss": 1.467,
+      "step": 76710
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1757575757575757e-05,
+      "loss": 1.4812,
+      "step": 76720
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1752525252525253e-05,
+      "loss": 1.4475,
+      "step": 76730
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1747474747474747e-05,
+      "loss": 1.4812,
+      "step": 76740
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1742424242424243e-05,
+      "loss": 1.4798,
+      "step": 76750
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.173737373737374e-05,
+      "loss": 1.4726,
+      "step": 76760
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1732323232323234e-05,
+      "loss": 1.4885,
+      "step": 76770
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1727272727272728e-05,
+      "loss": 1.4637,
+      "step": 76780
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1722222222222224e-05,
+      "loss": 1.4789,
+      "step": 76790
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1717171717171718e-05,
+      "loss": 1.4867,
+      "step": 76800
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1712121212121212e-05,
+      "loss": 1.4914,
+      "step": 76810
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1707070707070708e-05,
+      "loss": 1.482,
+      "step": 76820
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.1702020202020203e-05,
+      "loss": 1.491,
+      "step": 76830
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1696969696969699e-05,
+      "loss": 1.4917,
+      "step": 76840
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1691919191919193e-05,
+      "loss": 1.5031,
+      "step": 76850
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1686868686868687e-05,
+      "loss": 1.479,
+      "step": 76860
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1681818181818183e-05,
+      "loss": 1.4857,
+      "step": 76870
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1676767676767678e-05,
+      "loss": 1.4806,
+      "step": 76880
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1671717171717172e-05,
+      "loss": 1.4885,
+      "step": 76890
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1666666666666668e-05,
+      "loss": 1.4682,
+      "step": 76900
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1661616161616162e-05,
+      "loss": 1.4654,
+      "step": 76910
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1656565656565656e-05,
+      "loss": 1.4818,
+      "step": 76920
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1651515151515152e-05,
+      "loss": 1.4863,
+      "step": 76930
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1646464646464647e-05,
+      "loss": 1.4585,
+      "step": 76940
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1641414141414143e-05,
+      "loss": 1.4697,
+      "step": 76950
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1636363636363637e-05,
+      "loss": 1.4484,
+      "step": 76960
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1631313131313131e-05,
+      "loss": 1.4536,
+      "step": 76970
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1626262626262627e-05,
+      "loss": 1.4559,
+      "step": 76980
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1621212121212122e-05,
+      "loss": 1.4682,
+      "step": 76990
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1616161616161616e-05,
+      "loss": 1.4741,
+      "step": 77000
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1611111111111112e-05,
+      "loss": 1.4636,
+      "step": 77010
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1606060606060606e-05,
+      "loss": 1.4645,
+      "step": 77020
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.16010101010101e-05,
+      "loss": 1.4789,
+      "step": 77030
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1595959595959597e-05,
+      "loss": 1.5042,
+      "step": 77040
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.159090909090909e-05,
+      "loss": 1.4859,
+      "step": 77050
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1585858585858587e-05,
+      "loss": 1.4602,
+      "step": 77060
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1580808080808081e-05,
+      "loss": 1.4765,
+      "step": 77070
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1575757575757575e-05,
+      "loss": 1.4923,
+      "step": 77080
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1570707070707071e-05,
+      "loss": 1.4686,
+      "step": 77090
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1565656565656566e-05,
+      "loss": 1.4829,
+      "step": 77100
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.156060606060606e-05,
+      "loss": 1.486,
+      "step": 77110
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1555555555555556e-05,
+      "loss": 1.4851,
+      "step": 77120
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.155050505050505e-05,
+      "loss": 1.4653,
+      "step": 77130
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1545454545454545e-05,
+      "loss": 1.4811,
+      "step": 77140
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.154040404040404e-05,
+      "loss": 1.4578,
+      "step": 77150
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1535353535353535e-05,
+      "loss": 1.4843,
+      "step": 77160
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.153030303030303e-05,
+      "loss": 1.479,
+      "step": 77170
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1525252525252527e-05,
+      "loss": 1.4809,
+      "step": 77180
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1520202020202021e-05,
+      "loss": 1.462,
+      "step": 77190
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1515151515151517e-05,
+      "loss": 1.4634,
+      "step": 77200
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1510101010101011e-05,
+      "loss": 1.5114,
+      "step": 77210
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1505050505050506e-05,
+      "loss": 1.4811,
+      "step": 77220
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1500000000000002e-05,
+      "loss": 1.4709,
+      "step": 77230
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1494949494949496e-05,
+      "loss": 1.4848,
+      "step": 77240
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.148989898989899e-05,
+      "loss": 1.4895,
+      "step": 77250
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1484848484848486e-05,
+      "loss": 1.4817,
+      "step": 77260
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.147979797979798e-05,
+      "loss": 1.485,
+      "step": 77270
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1474747474747475e-05,
+      "loss": 1.5035,
+      "step": 77280
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1469696969696971e-05,
+      "loss": 1.4635,
+      "step": 77290
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1464646464646465e-05,
+      "loss": 1.4787,
+      "step": 77300
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1459595959595961e-05,
+      "loss": 1.4905,
+      "step": 77310
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1454545454545455e-05,
+      "loss": 1.4766,
+      "step": 77320
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.144949494949495e-05,
+      "loss": 1.4748,
+      "step": 77330
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1444444444444446e-05,
+      "loss": 1.4761,
+      "step": 77340
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.143939393939394e-05,
+      "loss": 1.4722,
+      "step": 77350
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1434343434343434e-05,
+      "loss": 1.4843,
+      "step": 77360
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.142929292929293e-05,
+      "loss": 1.4557,
+      "step": 77370
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1424242424242425e-05,
+      "loss": 1.4958,
+      "step": 77380
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1419191919191919e-05,
+      "loss": 1.4738,
+      "step": 77390
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1414141414141415e-05,
+      "loss": 1.4853,
+      "step": 77400
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.140909090909091e-05,
+      "loss": 1.4822,
+      "step": 77410
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1404040404040405e-05,
+      "loss": 1.4839,
+      "step": 77420
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.13989898989899e-05,
+      "loss": 1.4667,
+      "step": 77430
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1393939393939394e-05,
+      "loss": 1.4567,
+      "step": 77440
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.138888888888889e-05,
+      "loss": 1.4846,
+      "step": 77450
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1383838383838384e-05,
+      "loss": 1.4711,
+      "step": 77460
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1378787878787878e-05,
+      "loss": 1.4826,
+      "step": 77470
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1373737373737374e-05,
+      "loss": 1.4639,
+      "step": 77480
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.1368686868686869e-05,
+      "loss": 1.4854,
+      "step": 77490
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1363636363636365e-05,
+      "loss": 1.4898,
+      "step": 77500
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1358585858585859e-05,
+      "loss": 1.4977,
+      "step": 77510
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1353535353535353e-05,
+      "loss": 1.4756,
+      "step": 77520
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.134848484848485e-05,
+      "loss": 1.4672,
+      "step": 77530
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1343434343434344e-05,
+      "loss": 1.4985,
+      "step": 77540
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1338383838383838e-05,
+      "loss": 1.4929,
+      "step": 77550
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1333333333333334e-05,
+      "loss": 1.4743,
+      "step": 77560
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1328282828282828e-05,
+      "loss": 1.4655,
+      "step": 77570
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1323232323232322e-05,
+      "loss": 1.473,
+      "step": 77580
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1318181818181818e-05,
+      "loss": 1.473,
+      "step": 77590
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1313131313131314e-05,
+      "loss": 1.4635,
+      "step": 77600
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1308080808080809e-05,
+      "loss": 1.4908,
+      "step": 77610
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1303030303030305e-05,
+      "loss": 1.4847,
+      "step": 77620
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1297979797979799e-05,
+      "loss": 1.4589,
+      "step": 77630
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1292929292929293e-05,
+      "loss": 1.4744,
+      "step": 77640
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.128787878787879e-05,
+      "loss": 1.4833,
+      "step": 77650
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1282828282828284e-05,
+      "loss": 1.4689,
+      "step": 77660
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.127777777777778e-05,
+      "loss": 1.4827,
+      "step": 77670
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1272727272727274e-05,
+      "loss": 1.4715,
+      "step": 77680
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1267676767676768e-05,
+      "loss": 1.4733,
+      "step": 77690
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1262626262626264e-05,
+      "loss": 1.4771,
+      "step": 77700
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1257575757575758e-05,
+      "loss": 1.4696,
+      "step": 77710
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1252525252525253e-05,
+      "loss": 1.4835,
+      "step": 77720
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1247474747474749e-05,
+      "loss": 1.4678,
+      "step": 77730
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1242424242424243e-05,
+      "loss": 1.4589,
+      "step": 77740
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1237373737373737e-05,
+      "loss": 1.4586,
+      "step": 77750
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1232323232323233e-05,
+      "loss": 1.4959,
+      "step": 77760
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1227272727272728e-05,
+      "loss": 1.4667,
+      "step": 77770
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1222222222222224e-05,
+      "loss": 1.4787,
+      "step": 77780
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1217171717171718e-05,
+      "loss": 1.4854,
+      "step": 77790
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1212121212121212e-05,
+      "loss": 1.4751,
+      "step": 77800
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1207070707070708e-05,
+      "loss": 1.493,
+      "step": 77810
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1202020202020202e-05,
+      "loss": 1.4866,
+      "step": 77820
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1196969696969697e-05,
+      "loss": 1.4925,
+      "step": 77830
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1191919191919193e-05,
+      "loss": 1.4846,
+      "step": 77840
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1186868686868687e-05,
+      "loss": 1.5029,
+      "step": 77850
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1181818181818183e-05,
+      "loss": 1.483,
+      "step": 77860
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1176767676767677e-05,
+      "loss": 1.4754,
+      "step": 77870
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1171717171717172e-05,
+      "loss": 1.4943,
+      "step": 77880
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1166666666666668e-05,
+      "loss": 1.4803,
+      "step": 77890
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1161616161616162e-05,
+      "loss": 1.4878,
+      "step": 77900
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1156565656565656e-05,
+      "loss": 1.4967,
+      "step": 77910
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1151515151515152e-05,
+      "loss": 1.4844,
+      "step": 77920
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1146464646464646e-05,
+      "loss": 1.473,
+      "step": 77930
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.114141414141414e-05,
+      "loss": 1.4728,
+      "step": 77940
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1136363636363637e-05,
+      "loss": 1.478,
+      "step": 77950
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1131313131313131e-05,
+      "loss": 1.4792,
+      "step": 77960
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1126262626262627e-05,
+      "loss": 1.4848,
+      "step": 77970
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1121212121212121e-05,
+      "loss": 1.4726,
+      "step": 77980
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1116161616161616e-05,
+      "loss": 1.4816,
+      "step": 77990
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1111111111111112e-05,
+      "loss": 1.4889,
+      "step": 78000
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1106060606060606e-05,
+      "loss": 1.4907,
+      "step": 78010
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.11010101010101e-05,
+      "loss": 1.4491,
+      "step": 78020
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1095959595959598e-05,
+      "loss": 1.5003,
+      "step": 78030
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1090909090909092e-05,
+      "loss": 1.4718,
+      "step": 78040
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1085858585858587e-05,
+      "loss": 1.4881,
+      "step": 78050
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1080808080808083e-05,
+      "loss": 1.4814,
+      "step": 78060
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1075757575757577e-05,
+      "loss": 1.4565,
+      "step": 78070
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1070707070707071e-05,
+      "loss": 1.4972,
+      "step": 78080
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1065656565656567e-05,
+      "loss": 1.4913,
+      "step": 78090
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1060606060606061e-05,
+      "loss": 1.4743,
+      "step": 78100
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1055555555555556e-05,
+      "loss": 1.4724,
+      "step": 78110
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1050505050505052e-05,
+      "loss": 1.4784,
+      "step": 78120
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1045454545454546e-05,
+      "loss": 1.4947,
+      "step": 78130
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1040404040404042e-05,
+      "loss": 1.4871,
+      "step": 78140
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.1035353535353536e-05,
+      "loss": 1.4919,
+      "step": 78150
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.103030303030303e-05,
+      "loss": 1.453,
+      "step": 78160
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.1025252525252527e-05,
+      "loss": 1.5025,
+      "step": 78170
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.102020202020202e-05,
+      "loss": 1.4604,
+      "step": 78180
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.1015151515151515e-05,
+      "loss": 1.4741,
+      "step": 78190
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.1010101010101011e-05,
+      "loss": 1.4603,
+      "step": 78200
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.1005050505050505e-05,
+      "loss": 1.4859,
+      "step": 78210
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 1.4803,
+      "step": 78220
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0994949494949496e-05,
+      "loss": 1.4674,
+      "step": 78230
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.098989898989899e-05,
+      "loss": 1.4707,
+      "step": 78240
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0984848484848486e-05,
+      "loss": 1.4533,
+      "step": 78250
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.097979797979798e-05,
+      "loss": 1.4697,
+      "step": 78260
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0974747474747475e-05,
+      "loss": 1.4745,
+      "step": 78270
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.096969696969697e-05,
+      "loss": 1.4796,
+      "step": 78280
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0964646464646465e-05,
+      "loss": 1.4794,
+      "step": 78290
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.095959595959596e-05,
+      "loss": 1.4564,
+      "step": 78300
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0954545454545455e-05,
+      "loss": 1.4674,
+      "step": 78310
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.094949494949495e-05,
+      "loss": 1.4882,
+      "step": 78320
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0944444444444445e-05,
+      "loss": 1.4529,
+      "step": 78330
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.093939393939394e-05,
+      "loss": 1.48,
+      "step": 78340
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0934343434343434e-05,
+      "loss": 1.4869,
+      "step": 78350
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.092929292929293e-05,
+      "loss": 1.4767,
+      "step": 78360
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0924242424242424e-05,
+      "loss": 1.465,
+      "step": 78370
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0919191919191919e-05,
+      "loss": 1.4778,
+      "step": 78380
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0914141414141415e-05,
+      "loss": 1.469,
+      "step": 78390
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0909090909090909e-05,
+      "loss": 1.4865,
+      "step": 78400
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0904040404040403e-05,
+      "loss": 1.4815,
+      "step": 78410
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.08989898989899e-05,
+      "loss": 1.4769,
+      "step": 78420
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0893939393939393e-05,
+      "loss": 1.4605,
+      "step": 78430
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.088888888888889e-05,
+      "loss": 1.4816,
+      "step": 78440
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0883838383838384e-05,
+      "loss": 1.476,
+      "step": 78450
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.087878787878788e-05,
+      "loss": 1.4719,
+      "step": 78460
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0873737373737374e-05,
+      "loss": 1.4762,
+      "step": 78470
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.086868686868687e-05,
+      "loss": 1.4624,
+      "step": 78480
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0863636363636364e-05,
+      "loss": 1.453,
+      "step": 78490
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.085858585858586e-05,
+      "loss": 1.4915,
+      "step": 78500
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0853535353535355e-05,
+      "loss": 1.4895,
+      "step": 78510
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0848484848484849e-05,
+      "loss": 1.505,
+      "step": 78520
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0843434343434345e-05,
+      "loss": 1.4931,
+      "step": 78530
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.083838383838384e-05,
+      "loss": 1.4702,
+      "step": 78540
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0833333333333334e-05,
+      "loss": 1.4761,
+      "step": 78550
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.082828282828283e-05,
+      "loss": 1.4613,
+      "step": 78560
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0823232323232324e-05,
+      "loss": 1.4581,
+      "step": 78570
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.081818181818182e-05,
+      "loss": 1.4718,
+      "step": 78580
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0813131313131314e-05,
+      "loss": 1.4894,
+      "step": 78590
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0808080808080808e-05,
+      "loss": 1.4953,
+      "step": 78600
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0803030303030304e-05,
+      "loss": 1.4831,
+      "step": 78610
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0797979797979799e-05,
+      "loss": 1.464,
+      "step": 78620
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0792929292929293e-05,
+      "loss": 1.4731,
+      "step": 78630
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0787878787878789e-05,
+      "loss": 1.4777,
+      "step": 78640
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0782828282828283e-05,
+      "loss": 1.4702,
+      "step": 78650
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0777777777777778e-05,
+      "loss": 1.4722,
+      "step": 78660
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0772727272727274e-05,
+      "loss": 1.4711,
+      "step": 78670
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0767676767676768e-05,
+      "loss": 1.4763,
+      "step": 78680
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0762626262626264e-05,
+      "loss": 1.4672,
+      "step": 78690
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0757575757575758e-05,
+      "loss": 1.4813,
+      "step": 78700
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0752525252525252e-05,
+      "loss": 1.4601,
+      "step": 78710
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0747474747474748e-05,
+      "loss": 1.4665,
+      "step": 78720
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0742424242424243e-05,
+      "loss": 1.4734,
+      "step": 78730
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0737373737373737e-05,
+      "loss": 1.487,
+      "step": 78740
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0732323232323233e-05,
+      "loss": 1.4776,
+      "step": 78750
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0727272727272727e-05,
+      "loss": 1.4809,
+      "step": 78760
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0722222222222222e-05,
+      "loss": 1.4876,
+      "step": 78770
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0717171717171718e-05,
+      "loss": 1.4737,
+      "step": 78780
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0712121212121212e-05,
+      "loss": 1.4939,
+      "step": 78790
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0707070707070708e-05,
+      "loss": 1.4632,
+      "step": 78800
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0702020202020202e-05,
+      "loss": 1.4787,
+      "step": 78810
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.0696969696969696e-05,
+      "loss": 1.4702,
+      "step": 78820
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0691919191919192e-05,
+      "loss": 1.4755,
+      "step": 78830
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0686868686868687e-05,
+      "loss": 1.4841,
+      "step": 78840
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0681818181818181e-05,
+      "loss": 1.4778,
+      "step": 78850
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0676767676767677e-05,
+      "loss": 1.4716,
+      "step": 78860
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0671717171717171e-05,
+      "loss": 1.4525,
+      "step": 78870
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0666666666666667e-05,
+      "loss": 1.4683,
+      "step": 78880
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0661616161616163e-05,
+      "loss": 1.4926,
+      "step": 78890
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0656565656565658e-05,
+      "loss": 1.465,
+      "step": 78900
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0651515151515152e-05,
+      "loss": 1.4854,
+      "step": 78910
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0646464646464648e-05,
+      "loss": 1.4829,
+      "step": 78920
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0641414141414142e-05,
+      "loss": 1.4675,
+      "step": 78930
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0636363636363638e-05,
+      "loss": 1.4733,
+      "step": 78940
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0631313131313132e-05,
+      "loss": 1.4744,
+      "step": 78950
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0626262626262627e-05,
+      "loss": 1.4761,
+      "step": 78960
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0621212121212123e-05,
+      "loss": 1.4815,
+      "step": 78970
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0616161616161617e-05,
+      "loss": 1.4684,
+      "step": 78980
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0611111111111111e-05,
+      "loss": 1.4813,
+      "step": 78990
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0606060606060607e-05,
+      "loss": 1.4577,
+      "step": 79000
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0601010101010102e-05,
+      "loss": 1.4766,
+      "step": 79010
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0595959595959596e-05,
+      "loss": 1.5036,
+      "step": 79020
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0590909090909092e-05,
+      "loss": 1.4901,
+      "step": 79030
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0585858585858586e-05,
+      "loss": 1.4842,
+      "step": 79040
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0580808080808082e-05,
+      "loss": 1.4808,
+      "step": 79050
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0575757575757577e-05,
+      "loss": 1.4817,
+      "step": 79060
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.057070707070707e-05,
+      "loss": 1.4748,
+      "step": 79070
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0565656565656567e-05,
+      "loss": 1.4792,
+      "step": 79080
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0560606060606061e-05,
+      "loss": 1.487,
+      "step": 79090
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0555555555555555e-05,
+      "loss": 1.484,
+      "step": 79100
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0550505050505051e-05,
+      "loss": 1.4973,
+      "step": 79110
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0545454545454546e-05,
+      "loss": 1.4607,
+      "step": 79120
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.054040404040404e-05,
+      "loss": 1.4602,
+      "step": 79130
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0535353535353536e-05,
+      "loss": 1.488,
+      "step": 79140
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.053030303030303e-05,
+      "loss": 1.4852,
+      "step": 79150
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0525252525252526e-05,
+      "loss": 1.4875,
+      "step": 79160
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.052020202020202e-05,
+      "loss": 1.4623,
+      "step": 79170
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0515151515151515e-05,
+      "loss": 1.4874,
+      "step": 79180
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.051010101010101e-05,
+      "loss": 1.474,
+      "step": 79190
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0505050505050505e-05,
+      "loss": 1.4812,
+      "step": 79200
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.05e-05,
+      "loss": 1.4751,
+      "step": 79210
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0494949494949495e-05,
+      "loss": 1.4804,
+      "step": 79220
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.048989898989899e-05,
+      "loss": 1.4934,
+      "step": 79230
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0484848484848486e-05,
+      "loss": 1.4826,
+      "step": 79240
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.047979797979798e-05,
+      "loss": 1.4874,
+      "step": 79250
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0474747474747474e-05,
+      "loss": 1.4831,
+      "step": 79260
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.046969696969697e-05,
+      "loss": 1.4745,
+      "step": 79270
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0464646464646465e-05,
+      "loss": 1.4834,
+      "step": 79280
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0459595959595959e-05,
+      "loss": 1.4831,
+      "step": 79290
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0454545454545455e-05,
+      "loss": 1.4737,
+      "step": 79300
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0449494949494951e-05,
+      "loss": 1.4765,
+      "step": 79310
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0444444444444445e-05,
+      "loss": 1.4581,
+      "step": 79320
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0439393939393941e-05,
+      "loss": 1.4743,
+      "step": 79330
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0434343434343435e-05,
+      "loss": 1.4657,
+      "step": 79340
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.042929292929293e-05,
+      "loss": 1.4945,
+      "step": 79350
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0424242424242426e-05,
+      "loss": 1.4708,
+      "step": 79360
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.041919191919192e-05,
+      "loss": 1.469,
+      "step": 79370
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0414141414141414e-05,
+      "loss": 1.4906,
+      "step": 79380
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.040909090909091e-05,
+      "loss": 1.4679,
+      "step": 79390
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0404040404040405e-05,
+      "loss": 1.4745,
+      "step": 79400
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.03989898989899e-05,
+      "loss": 1.4505,
+      "step": 79410
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0393939393939395e-05,
+      "loss": 1.4491,
+      "step": 79420
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.038888888888889e-05,
+      "loss": 1.483,
+      "step": 79430
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0383838383838385e-05,
+      "loss": 1.5027,
+      "step": 79440
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.037878787878788e-05,
+      "loss": 1.4782,
+      "step": 79450
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0373737373737374e-05,
+      "loss": 1.4657,
+      "step": 79460
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.036868686868687e-05,
+      "loss": 1.4834,
+      "step": 79470
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0363636363636364e-05,
+      "loss": 1.4818,
+      "step": 79480
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.0358585858585858e-05,
+      "loss": 1.4744,
+      "step": 79490
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0353535353535354e-05,
+      "loss": 1.4807,
+      "step": 79500
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0348484848484849e-05,
+      "loss": 1.4882,
+      "step": 79510
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0343434343434345e-05,
+      "loss": 1.4827,
+      "step": 79520
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0338383838383839e-05,
+      "loss": 1.4754,
+      "step": 79530
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0333333333333333e-05,
+      "loss": 1.4906,
+      "step": 79540
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.032828282828283e-05,
+      "loss": 1.488,
+      "step": 79550
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0323232323232324e-05,
+      "loss": 1.4777,
+      "step": 79560
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0318181818181818e-05,
+      "loss": 1.4894,
+      "step": 79570
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0313131313131314e-05,
+      "loss": 1.4937,
+      "step": 79580
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0308080808080808e-05,
+      "loss": 1.4775,
+      "step": 79590
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0303030303030304e-05,
+      "loss": 1.4921,
+      "step": 79600
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0297979797979798e-05,
+      "loss": 1.4887,
+      "step": 79610
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0292929292929293e-05,
+      "loss": 1.4815,
+      "step": 79620
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0287878787878789e-05,
+      "loss": 1.5202,
+      "step": 79630
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0282828282828283e-05,
+      "loss": 1.4956,
+      "step": 79640
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0277777777777777e-05,
+      "loss": 1.4606,
+      "step": 79650
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0272727272727273e-05,
+      "loss": 1.4679,
+      "step": 79660
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0267676767676768e-05,
+      "loss": 1.4663,
+      "step": 79670
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0262626262626262e-05,
+      "loss": 1.4938,
+      "step": 79680
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0257575757575758e-05,
+      "loss": 1.4911,
+      "step": 79690
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0252525252525252e-05,
+      "loss": 1.4581,
+      "step": 79700
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0247474747474748e-05,
+      "loss": 1.4615,
+      "step": 79710
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0242424242424242e-05,
+      "loss": 1.482,
+      "step": 79720
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0237373737373737e-05,
+      "loss": 1.4776,
+      "step": 79730
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0232323232323233e-05,
+      "loss": 1.4663,
+      "step": 79740
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0227272727272729e-05,
+      "loss": 1.4728,
+      "step": 79750
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0222222222222223e-05,
+      "loss": 1.4575,
+      "step": 79760
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0217171717171719e-05,
+      "loss": 1.4928,
+      "step": 79770
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0212121212121213e-05,
+      "loss": 1.4861,
+      "step": 79780
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0207070707070708e-05,
+      "loss": 1.4493,
+      "step": 79790
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0202020202020204e-05,
+      "loss": 1.4888,
+      "step": 79800
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0196969696969698e-05,
+      "loss": 1.4629,
+      "step": 79810
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0191919191919192e-05,
+      "loss": 1.4913,
+      "step": 79820
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0186868686868688e-05,
+      "loss": 1.4756,
+      "step": 79830
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0181818181818182e-05,
+      "loss": 1.4796,
+      "step": 79840
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0176767676767677e-05,
+      "loss": 1.485,
+      "step": 79850
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0171717171717173e-05,
+      "loss": 1.4762,
+      "step": 79860
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0166666666666667e-05,
+      "loss": 1.479,
+      "step": 79870
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0161616161616163e-05,
+      "loss": 1.469,
+      "step": 79880
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0156565656565657e-05,
+      "loss": 1.481,
+      "step": 79890
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0151515151515152e-05,
+      "loss": 1.4797,
+      "step": 79900
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0146464646464648e-05,
+      "loss": 1.4679,
+      "step": 79910
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0141414141414142e-05,
+      "loss": 1.4703,
+      "step": 79920
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0136363636363636e-05,
+      "loss": 1.4769,
+      "step": 79930
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0131313131313132e-05,
+      "loss": 1.47,
+      "step": 79940
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0126262626262626e-05,
+      "loss": 1.4719,
+      "step": 79950
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0121212121212122e-05,
+      "loss": 1.4584,
+      "step": 79960
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0116161616161617e-05,
+      "loss": 1.4603,
+      "step": 79970
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0111111111111111e-05,
+      "loss": 1.4744,
+      "step": 79980
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0106060606060607e-05,
+      "loss": 1.4738,
+      "step": 79990
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0101010101010101e-05,
+      "loss": 1.488,
+      "step": 80000
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0095959595959596e-05,
+      "loss": 1.4917,
+      "step": 80010
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0090909090909092e-05,
+      "loss": 1.4792,
+      "step": 80020
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0085858585858586e-05,
+      "loss": 1.4852,
+      "step": 80030
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.008080808080808e-05,
+      "loss": 1.4595,
+      "step": 80040
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0075757575757576e-05,
+      "loss": 1.4836,
+      "step": 80050
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.007070707070707e-05,
+      "loss": 1.4765,
+      "step": 80060
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0065656565656566e-05,
+      "loss": 1.4962,
+      "step": 80070
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.006060606060606e-05,
+      "loss": 1.4796,
+      "step": 80080
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0055555555555555e-05,
+      "loss": 1.4942,
+      "step": 80090
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0050505050505051e-05,
+      "loss": 1.47,
+      "step": 80100
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0045454545454545e-05,
+      "loss": 1.4702,
+      "step": 80110
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.004040404040404e-05,
+      "loss": 1.4911,
+      "step": 80120
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0035353535353536e-05,
+      "loss": 1.4747,
+      "step": 80130
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.003030303030303e-05,
+      "loss": 1.4789,
+      "step": 80140
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.0025252525252524e-05,
+      "loss": 1.4573,
+      "step": 80150
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.0020202020202022e-05,
+      "loss": 1.4813,
+      "step": 80160
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.0015151515151516e-05,
+      "loss": 1.449,
+      "step": 80170
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.001010101010101e-05,
+      "loss": 1.4696,
+      "step": 80180
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.0005050505050507e-05,
+      "loss": 1.495,
+      "step": 80190
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1e-05,
+      "loss": 1.4682,
+      "step": 80200
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.994949494949495e-06,
+      "loss": 1.4545,
+      "step": 80210
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.989898989898991e-06,
+      "loss": 1.4832,
+      "step": 80220
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.984848484848485e-06,
+      "loss": 1.4977,
+      "step": 80230
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.979797979797981e-06,
+      "loss": 1.4888,
+      "step": 80240
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.974747474747476e-06,
+      "loss": 1.4845,
+      "step": 80250
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.96969696969697e-06,
+      "loss": 1.4634,
+      "step": 80260
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.964646464646466e-06,
+      "loss": 1.4548,
+      "step": 80270
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.95959595959596e-06,
+      "loss": 1.4742,
+      "step": 80280
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.954545454545455e-06,
+      "loss": 1.4917,
+      "step": 80290
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.94949494949495e-06,
+      "loss": 1.4759,
+      "step": 80300
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.944444444444445e-06,
+      "loss": 1.4693,
+      "step": 80310
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.93939393939394e-06,
+      "loss": 1.4534,
+      "step": 80320
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.934343434343435e-06,
+      "loss": 1.4484,
+      "step": 80330
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.92929292929293e-06,
+      "loss": 1.474,
+      "step": 80340
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.924242424242425e-06,
+      "loss": 1.5007,
+      "step": 80350
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.91919191919192e-06,
+      "loss": 1.475,
+      "step": 80360
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.914141414141414e-06,
+      "loss": 1.486,
+      "step": 80370
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.90909090909091e-06,
+      "loss": 1.4897,
+      "step": 80380
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.904040404040404e-06,
+      "loss": 1.4797,
+      "step": 80390
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.898989898989899e-06,
+      "loss": 1.4637,
+      "step": 80400
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.893939393939395e-06,
+      "loss": 1.4664,
+      "step": 80410
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.888888888888889e-06,
+      "loss": 1.4704,
+      "step": 80420
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.883838383838385e-06,
+      "loss": 1.4773,
+      "step": 80430
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.87878787878788e-06,
+      "loss": 1.4737,
+      "step": 80440
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.873737373737373e-06,
+      "loss": 1.4797,
+      "step": 80450
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.86868686868687e-06,
+      "loss": 1.4743,
+      "step": 80460
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.863636363636364e-06,
+      "loss": 1.489,
+      "step": 80470
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.858585858585858e-06,
+      "loss": 1.462,
+      "step": 80480
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.853535353535354e-06,
+      "loss": 1.4643,
+      "step": 80490
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.848484848484848e-06,
+      "loss": 1.4605,
+      "step": 80500
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.843434343434343e-06,
+      "loss": 1.4906,
+      "step": 80510
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.838383838383839e-06,
+      "loss": 1.4831,
+      "step": 80520
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.833333333333333e-06,
+      "loss": 1.4846,
+      "step": 80530
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.828282828282829e-06,
+      "loss": 1.4787,
+      "step": 80540
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.823232323232323e-06,
+      "loss": 1.4655,
+      "step": 80550
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.818181818181818e-06,
+      "loss": 1.4809,
+      "step": 80560
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.813131313131313e-06,
+      "loss": 1.4834,
+      "step": 80570
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.808080808080808e-06,
+      "loss": 1.4772,
+      "step": 80580
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.803030303030304e-06,
+      "loss": 1.4698,
+      "step": 80590
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.7979797979798e-06,
+      "loss": 1.4997,
+      "step": 80600
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.792929292929294e-06,
+      "loss": 1.5073,
+      "step": 80610
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.787878787878788e-06,
+      "loss": 1.4831,
+      "step": 80620
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.782828282828284e-06,
+      "loss": 1.4602,
+      "step": 80630
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.777777777777779e-06,
+      "loss": 1.4949,
+      "step": 80640
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.772727272727273e-06,
+      "loss": 1.4854,
+      "step": 80650
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.767676767676769e-06,
+      "loss": 1.4528,
+      "step": 80660
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.762626262626263e-06,
+      "loss": 1.4715,
+      "step": 80670
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.75757575757576e-06,
+      "loss": 1.4878,
+      "step": 80680
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.752525252525254e-06,
+      "loss": 1.4996,
+      "step": 80690
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.747474747474748e-06,
+      "loss": 1.4672,
+      "step": 80700
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.742424242424244e-06,
+      "loss": 1.4698,
+      "step": 80710
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.737373737373738e-06,
+      "loss": 1.4737,
+      "step": 80720
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.732323232323232e-06,
+      "loss": 1.4589,
+      "step": 80730
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.727272727272728e-06,
+      "loss": 1.464,
+      "step": 80740
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.722222222222223e-06,
+      "loss": 1.4699,
+      "step": 80750
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.717171717171717e-06,
+      "loss": 1.4606,
+      "step": 80760
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.712121212121213e-06,
+      "loss": 1.4495,
+      "step": 80770
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.707070707070707e-06,
+      "loss": 1.4704,
+      "step": 80780
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.702020202020203e-06,
+      "loss": 1.4819,
+      "step": 80790
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.696969696969698e-06,
+      "loss": 1.4852,
+      "step": 80800
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.691919191919192e-06,
+      "loss": 1.4803,
+      "step": 80810
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 9.686868686868688e-06,
+      "loss": 1.4727,
+      "step": 80820
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.681818181818182e-06,
+      "loss": 1.4549,
+      "step": 80830
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.676767676767676e-06,
+      "loss": 1.4679,
+      "step": 80840
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.671717171717172e-06,
+      "loss": 1.4654,
+      "step": 80850
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.666666666666667e-06,
+      "loss": 1.4703,
+      "step": 80860
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.661616161616161e-06,
+      "loss": 1.4682,
+      "step": 80870
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.656565656565657e-06,
+      "loss": 1.4736,
+      "step": 80880
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.651515151515151e-06,
+      "loss": 1.4695,
+      "step": 80890
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.646464646464647e-06,
+      "loss": 1.4623,
+      "step": 80900
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.641414141414142e-06,
+      "loss": 1.4844,
+      "step": 80910
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.636363636363636e-06,
+      "loss": 1.4863,
+      "step": 80920
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.631313131313132e-06,
+      "loss": 1.4832,
+      "step": 80930
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.626262626262626e-06,
+      "loss": 1.4772,
+      "step": 80940
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.62121212121212e-06,
+      "loss": 1.4752,
+      "step": 80950
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.616161616161616e-06,
+      "loss": 1.4692,
+      "step": 80960
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.61111111111111e-06,
+      "loss": 1.4776,
+      "step": 80970
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.606060606060607e-06,
+      "loss": 1.4636,
+      "step": 80980
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.601010101010101e-06,
+      "loss": 1.4838,
+      "step": 80990
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.595959595959595e-06,
+      "loss": 1.4759,
+      "step": 81000
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.590909090909091e-06,
+      "loss": 1.4794,
+      "step": 81010
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.585858585858587e-06,
+      "loss": 1.498,
+      "step": 81020
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.580808080808082e-06,
+      "loss": 1.4644,
+      "step": 81030
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.575757575757578e-06,
+      "loss": 1.4739,
+      "step": 81040
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.570707070707072e-06,
+      "loss": 1.47,
+      "step": 81050
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.565656565656566e-06,
+      "loss": 1.4663,
+      "step": 81060
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.560606060606062e-06,
+      "loss": 1.4895,
+      "step": 81070
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.555555555555556e-06,
+      "loss": 1.4483,
+      "step": 81080
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.55050505050505e-06,
+      "loss": 1.4919,
+      "step": 81090
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.545454545454547e-06,
+      "loss": 1.4818,
+      "step": 81100
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.540404040404041e-06,
+      "loss": 1.4619,
+      "step": 81110
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.535353535353535e-06,
+      "loss": 1.4774,
+      "step": 81120
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.530303030303031e-06,
+      "loss": 1.4882,
+      "step": 81130
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.525252525252526e-06,
+      "loss": 1.4918,
+      "step": 81140
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.520202020202022e-06,
+      "loss": 1.4749,
+      "step": 81150
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.515151515151516e-06,
+      "loss": 1.4566,
+      "step": 81160
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.51010101010101e-06,
+      "loss": 1.4794,
+      "step": 81170
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.505050505050506e-06,
+      "loss": 1.4776,
+      "step": 81180
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.5e-06,
+      "loss": 1.475,
+      "step": 81190
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.494949494949495e-06,
+      "loss": 1.468,
+      "step": 81200
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.48989898989899e-06,
+      "loss": 1.4796,
+      "step": 81210
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.484848484848485e-06,
+      "loss": 1.4717,
+      "step": 81220
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.47979797979798e-06,
+      "loss": 1.4853,
+      "step": 81230
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.474747474747475e-06,
+      "loss": 1.4861,
+      "step": 81240
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.46969696969697e-06,
+      "loss": 1.4898,
+      "step": 81250
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.464646464646466e-06,
+      "loss": 1.4695,
+      "step": 81260
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.45959595959596e-06,
+      "loss": 1.4891,
+      "step": 81270
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.454545454545454e-06,
+      "loss": 1.4595,
+      "step": 81280
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.44949494949495e-06,
+      "loss": 1.4594,
+      "step": 81290
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.444444444444445e-06,
+      "loss": 1.4953,
+      "step": 81300
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.439393939393939e-06,
+      "loss": 1.4712,
+      "step": 81310
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.434343434343435e-06,
+      "loss": 1.4646,
+      "step": 81320
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.429292929292929e-06,
+      "loss": 1.4922,
+      "step": 81330
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.424242424242425e-06,
+      "loss": 1.4712,
+      "step": 81340
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.41919191919192e-06,
+      "loss": 1.4636,
+      "step": 81350
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.414141414141414e-06,
+      "loss": 1.476,
+      "step": 81360
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.40909090909091e-06,
+      "loss": 1.4803,
+      "step": 81370
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.404040404040404e-06,
+      "loss": 1.4752,
+      "step": 81380
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.398989898989898e-06,
+      "loss": 1.4652,
+      "step": 81390
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.393939393939394e-06,
+      "loss": 1.4577,
+      "step": 81400
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.388888888888889e-06,
+      "loss": 1.4979,
+      "step": 81410
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.383838383838383e-06,
+      "loss": 1.4723,
+      "step": 81420
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.378787878787879e-06,
+      "loss": 1.4826,
+      "step": 81430
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.373737373737375e-06,
+      "loss": 1.4716,
+      "step": 81440
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.36868686868687e-06,
+      "loss": 1.4656,
+      "step": 81450
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.363636363636365e-06,
+      "loss": 1.4906,
+      "step": 81460
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.35858585858586e-06,
+      "loss": 1.4664,
+      "step": 81470
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.353535353535354e-06,
+      "loss": 1.4753,
+      "step": 81480
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.34848484848485e-06,
+      "loss": 1.4865,
+      "step": 81490
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.343434343434344e-06,
+      "loss": 1.4802,
+      "step": 81500
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.33838383838384e-06,
+      "loss": 1.4657,
+      "step": 81510
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.333333333333334e-06,
+      "loss": 1.4804,
+      "step": 81520
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.328282828282829e-06,
+      "loss": 1.4613,
+      "step": 81530
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.323232323232325e-06,
+      "loss": 1.4645,
+      "step": 81540
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.318181818181819e-06,
+      "loss": 1.4603,
+      "step": 81550
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.313131313131313e-06,
+      "loss": 1.4866,
+      "step": 81560
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.30808080808081e-06,
+      "loss": 1.4947,
+      "step": 81570
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.303030303030303e-06,
+      "loss": 1.4688,
+      "step": 81580
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.297979797979798e-06,
+      "loss": 1.4873,
+      "step": 81590
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.292929292929294e-06,
+      "loss": 1.4712,
+      "step": 81600
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.287878787878788e-06,
+      "loss": 1.4919,
+      "step": 81610
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.282828282828284e-06,
+      "loss": 1.4619,
+      "step": 81620
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.277777777777778e-06,
+      "loss": 1.4626,
+      "step": 81630
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.272727272727273e-06,
+      "loss": 1.4801,
+      "step": 81640
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.267676767676769e-06,
+      "loss": 1.4703,
+      "step": 81650
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.262626262626263e-06,
+      "loss": 1.493,
+      "step": 81660
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.257575757575757e-06,
+      "loss": 1.4869,
+      "step": 81670
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.252525252525253e-06,
+      "loss": 1.4908,
+      "step": 81680
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.247474747474748e-06,
+      "loss": 1.4825,
+      "step": 81690
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.242424242424244e-06,
+      "loss": 1.4761,
+      "step": 81700
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.237373737373738e-06,
+      "loss": 1.4628,
+      "step": 81710
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.232323232323232e-06,
+      "loss": 1.4868,
+      "step": 81720
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.227272727272728e-06,
+      "loss": 1.4813,
+      "step": 81730
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.222222222222222e-06,
+      "loss": 1.4754,
+      "step": 81740
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.217171717171717e-06,
+      "loss": 1.4909,
+      "step": 81750
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.212121212121213e-06,
+      "loss": 1.4928,
+      "step": 81760
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.207070707070707e-06,
+      "loss": 1.4689,
+      "step": 81770
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.202020202020201e-06,
+      "loss": 1.472,
+      "step": 81780
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.196969696969697e-06,
+      "loss": 1.4687,
+      "step": 81790
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.191919191919192e-06,
+      "loss": 1.4815,
+      "step": 81800
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.186868686868688e-06,
+      "loss": 1.4621,
+      "step": 81810
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.181818181818182e-06,
+      "loss": 1.4637,
+      "step": 81820
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.176767676767676e-06,
+      "loss": 1.4723,
+      "step": 81830
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.171717171717172e-06,
+      "loss": 1.4685,
+      "step": 81840
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.166666666666666e-06,
+      "loss": 1.4766,
+      "step": 81850
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.16161616161616e-06,
+      "loss": 1.4834,
+      "step": 81860
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.156565656565658e-06,
+      "loss": 1.4868,
+      "step": 81870
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.151515151515153e-06,
+      "loss": 1.474,
+      "step": 81880
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.146464646464647e-06,
+      "loss": 1.4813,
+      "step": 81890
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.141414141414143e-06,
+      "loss": 1.4758,
+      "step": 81900
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.136363636363637e-06,
+      "loss": 1.4559,
+      "step": 81910
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.131313131313132e-06,
+      "loss": 1.4519,
+      "step": 81920
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.126262626262628e-06,
+      "loss": 1.4776,
+      "step": 81930
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.121212121212122e-06,
+      "loss": 1.4728,
+      "step": 81940
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.116161616161616e-06,
+      "loss": 1.4899,
+      "step": 81950
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.111111111111112e-06,
+      "loss": 1.4913,
+      "step": 81960
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.106060606060606e-06,
+      "loss": 1.4877,
+      "step": 81970
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.101010101010102e-06,
+      "loss": 1.4673,
+      "step": 81980
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.095959595959597e-06,
+      "loss": 1.455,
+      "step": 81990
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.090909090909091e-06,
+      "loss": 1.453,
+      "step": 82000
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.085858585858587e-06,
+      "loss": 1.477,
+      "step": 82010
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.080808080808081e-06,
+      "loss": 1.4678,
+      "step": 82020
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.075757575757576e-06,
+      "loss": 1.4841,
+      "step": 82030
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.070707070707072e-06,
+      "loss": 1.4858,
+      "step": 82040
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.065656565656566e-06,
+      "loss": 1.4804,
+      "step": 82050
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.060606060606062e-06,
+      "loss": 1.4724,
+      "step": 82060
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.055555555555556e-06,
+      "loss": 1.4876,
+      "step": 82070
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.05050505050505e-06,
+      "loss": 1.4726,
+      "step": 82080
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.045454545454546e-06,
+      "loss": 1.4782,
+      "step": 82090
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.04040404040404e-06,
+      "loss": 1.4662,
+      "step": 82100
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.035353535353535e-06,
+      "loss": 1.4702,
+      "step": 82110
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.030303030303031e-06,
+      "loss": 1.5005,
+      "step": 82120
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.025252525252525e-06,
+      "loss": 1.4741,
+      "step": 82130
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.02020202020202e-06,
+      "loss": 1.4714,
+      "step": 82140
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.015151515151516e-06,
+      "loss": 1.4667,
+      "step": 82150
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 9.01010101010101e-06,
+      "loss": 1.4811,
+      "step": 82160
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 9.005050505050506e-06,
+      "loss": 1.489,
+      "step": 82170
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 9e-06,
+      "loss": 1.5011,
+      "step": 82180
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.994949494949495e-06,
+      "loss": 1.467,
+      "step": 82190
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.98989898989899e-06,
+      "loss": 1.4705,
+      "step": 82200
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.984848484848485e-06,
+      "loss": 1.4742,
+      "step": 82210
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.979797979797979e-06,
+      "loss": 1.4645,
+      "step": 82220
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.974747474747475e-06,
+      "loss": 1.469,
+      "step": 82230
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.96969696969697e-06,
+      "loss": 1.4704,
+      "step": 82240
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.964646464646464e-06,
+      "loss": 1.462,
+      "step": 82250
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.95959595959596e-06,
+      "loss": 1.4735,
+      "step": 82260
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.954545454545454e-06,
+      "loss": 1.4661,
+      "step": 82270
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.94949494949495e-06,
+      "loss": 1.4678,
+      "step": 82280
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.944444444444444e-06,
+      "loss": 1.4713,
+      "step": 82290
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.93939393939394e-06,
+      "loss": 1.4557,
+      "step": 82300
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.934343434343436e-06,
+      "loss": 1.4543,
+      "step": 82310
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.92929292929293e-06,
+      "loss": 1.4899,
+      "step": 82320
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.924242424242425e-06,
+      "loss": 1.46,
+      "step": 82330
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.91919191919192e-06,
+      "loss": 1.4621,
+      "step": 82340
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.914141414141415e-06,
+      "loss": 1.4666,
+      "step": 82350
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.90909090909091e-06,
+      "loss": 1.4755,
+      "step": 82360
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.904040404040405e-06,
+      "loss": 1.4645,
+      "step": 82370
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.8989898989899e-06,
+      "loss": 1.458,
+      "step": 82380
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.893939393939394e-06,
+      "loss": 1.4535,
+      "step": 82390
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.88888888888889e-06,
+      "loss": 1.4712,
+      "step": 82400
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.883838383838384e-06,
+      "loss": 1.4678,
+      "step": 82410
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.87878787878788e-06,
+      "loss": 1.4859,
+      "step": 82420
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.873737373737375e-06,
+      "loss": 1.4841,
+      "step": 82430
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.868686868686869e-06,
+      "loss": 1.4921,
+      "step": 82440
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.863636363636365e-06,
+      "loss": 1.4945,
+      "step": 82450
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.85858585858586e-06,
+      "loss": 1.4681,
+      "step": 82460
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.853535353535353e-06,
+      "loss": 1.4779,
+      "step": 82470
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.84848484848485e-06,
+      "loss": 1.4699,
+      "step": 82480
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.843434343434344e-06,
+      "loss": 1.4743,
+      "step": 82490
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.838383838383838e-06,
+      "loss": 1.4727,
+      "step": 82500
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.833333333333334e-06,
+      "loss": 1.4656,
+      "step": 82510
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.828282828282828e-06,
+      "loss": 1.4613,
+      "step": 82520
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.823232323232324e-06,
+      "loss": 1.4924,
+      "step": 82530
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.818181818181819e-06,
+      "loss": 1.49,
+      "step": 82540
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.813131313131313e-06,
+      "loss": 1.5009,
+      "step": 82550
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.808080808080809e-06,
+      "loss": 1.4498,
+      "step": 82560
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.803030303030303e-06,
+      "loss": 1.5037,
+      "step": 82570
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.797979797979797e-06,
+      "loss": 1.4805,
+      "step": 82580
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.792929292929293e-06,
+      "loss": 1.4676,
+      "step": 82590
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.787878787878788e-06,
+      "loss": 1.4768,
+      "step": 82600
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.782828282828282e-06,
+      "loss": 1.4617,
+      "step": 82610
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.777777777777778e-06,
+      "loss": 1.466,
+      "step": 82620
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.772727272727272e-06,
+      "loss": 1.4609,
+      "step": 82630
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.767676767676768e-06,
+      "loss": 1.4695,
+      "step": 82640
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.762626262626263e-06,
+      "loss": 1.4614,
+      "step": 82650
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.757575757575757e-06,
+      "loss": 1.4755,
+      "step": 82660
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.752525252525253e-06,
+      "loss": 1.4792,
+      "step": 82670
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.747474747474747e-06,
+      "loss": 1.4811,
+      "step": 82680
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.742424242424242e-06,
+      "loss": 1.4677,
+      "step": 82690
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.737373737373738e-06,
+      "loss": 1.4805,
+      "step": 82700
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.732323232323232e-06,
+      "loss": 1.4629,
+      "step": 82710
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.727272727272728e-06,
+      "loss": 1.4862,
+      "step": 82720
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.722222222222224e-06,
+      "loss": 1.4584,
+      "step": 82730
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.717171717171718e-06,
+      "loss": 1.4695,
+      "step": 82740
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.712121212121212e-06,
+      "loss": 1.4713,
+      "step": 82750
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.707070707070708e-06,
+      "loss": 1.4537,
+      "step": 82760
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.702020202020203e-06,
+      "loss": 1.4779,
+      "step": 82770
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.696969696969699e-06,
+      "loss": 1.4903,
+      "step": 82780
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.691919191919193e-06,
+      "loss": 1.4596,
+      "step": 82790
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.686868686868687e-06,
+      "loss": 1.4843,
+      "step": 82800
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 8.681818181818183e-06,
+      "loss": 1.4939,
+      "step": 82810
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.676767676767678e-06,
+      "loss": 1.4754,
+      "step": 82820
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.671717171717172e-06,
+      "loss": 1.4745,
+      "step": 82830
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.666666666666668e-06,
+      "loss": 1.487,
+      "step": 82840
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.661616161616162e-06,
+      "loss": 1.4526,
+      "step": 82850
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.656565656565656e-06,
+      "loss": 1.4804,
+      "step": 82860
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.651515151515152e-06,
+      "loss": 1.471,
+      "step": 82870
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.646464646464647e-06,
+      "loss": 1.4764,
+      "step": 82880
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.641414141414143e-06,
+      "loss": 1.4747,
+      "step": 82890
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.636363636363637e-06,
+      "loss": 1.4671,
+      "step": 82900
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.631313131313131e-06,
+      "loss": 1.4595,
+      "step": 82910
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.626262626262627e-06,
+      "loss": 1.512,
+      "step": 82920
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.621212121212122e-06,
+      "loss": 1.465,
+      "step": 82930
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.616161616161616e-06,
+      "loss": 1.4731,
+      "step": 82940
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.611111111111112e-06,
+      "loss": 1.4852,
+      "step": 82950
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.606060606060606e-06,
+      "loss": 1.502,
+      "step": 82960
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.6010101010101e-06,
+      "loss": 1.4842,
+      "step": 82970
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.595959595959596e-06,
+      "loss": 1.4731,
+      "step": 82980
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.59090909090909e-06,
+      "loss": 1.4906,
+      "step": 82990
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.585858585858587e-06,
+      "loss": 1.4796,
+      "step": 83000
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.580808080808081e-06,
+      "loss": 1.4715,
+      "step": 83010
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.575757575757575e-06,
+      "loss": 1.4676,
+      "step": 83020
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.570707070707071e-06,
+      "loss": 1.4579,
+      "step": 83030
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.565656565656566e-06,
+      "loss": 1.4789,
+      "step": 83040
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.56060606060606e-06,
+      "loss": 1.4708,
+      "step": 83050
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.555555555555556e-06,
+      "loss": 1.464,
+      "step": 83060
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.55050505050505e-06,
+      "loss": 1.488,
+      "step": 83070
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.545454545454546e-06,
+      "loss": 1.484,
+      "step": 83080
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.54040404040404e-06,
+      "loss": 1.4797,
+      "step": 83090
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.535353535353535e-06,
+      "loss": 1.4625,
+      "step": 83100
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.53030303030303e-06,
+      "loss": 1.4662,
+      "step": 83110
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.525252525252525e-06,
+      "loss": 1.4598,
+      "step": 83120
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.52020202020202e-06,
+      "loss": 1.4787,
+      "step": 83130
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.515151515151515e-06,
+      "loss": 1.4869,
+      "step": 83140
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.510101010101011e-06,
+      "loss": 1.4778,
+      "step": 83150
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.505050505050506e-06,
+      "loss": 1.4663,
+      "step": 83160
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.500000000000002e-06,
+      "loss": 1.4725,
+      "step": 83170
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.494949494949496e-06,
+      "loss": 1.4631,
+      "step": 83180
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.48989898989899e-06,
+      "loss": 1.4579,
+      "step": 83190
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.484848484848486e-06,
+      "loss": 1.4849,
+      "step": 83200
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.47979797979798e-06,
+      "loss": 1.4765,
+      "step": 83210
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.474747474747475e-06,
+      "loss": 1.4551,
+      "step": 83220
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.46969696969697e-06,
+      "loss": 1.4684,
+      "step": 83230
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.464646464646465e-06,
+      "loss": 1.485,
+      "step": 83240
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.459595959595961e-06,
+      "loss": 1.4686,
+      "step": 83250
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.454545454545455e-06,
+      "loss": 1.4879,
+      "step": 83260
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.44949494949495e-06,
+      "loss": 1.4713,
+      "step": 83270
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.444444444444446e-06,
+      "loss": 1.479,
+      "step": 83280
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.43939393939394e-06,
+      "loss": 1.4535,
+      "step": 83290
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.434343434343434e-06,
+      "loss": 1.4626,
+      "step": 83300
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.42929292929293e-06,
+      "loss": 1.4793,
+      "step": 83310
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.424242424242425e-06,
+      "loss": 1.482,
+      "step": 83320
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.41919191919192e-06,
+      "loss": 1.4771,
+      "step": 83330
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.414141414141415e-06,
+      "loss": 1.4605,
+      "step": 83340
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.409090909090909e-06,
+      "loss": 1.4564,
+      "step": 83350
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.404040404040405e-06,
+      "loss": 1.4655,
+      "step": 83360
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.3989898989899e-06,
+      "loss": 1.4584,
+      "step": 83370
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.393939393939394e-06,
+      "loss": 1.4728,
+      "step": 83380
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.38888888888889e-06,
+      "loss": 1.4717,
+      "step": 83390
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.383838383838384e-06,
+      "loss": 1.4959,
+      "step": 83400
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.378787878787878e-06,
+      "loss": 1.4908,
+      "step": 83410
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.373737373737374e-06,
+      "loss": 1.4592,
+      "step": 83420
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.368686868686869e-06,
+      "loss": 1.4805,
+      "step": 83430
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.363636363636365e-06,
+      "loss": 1.4508,
+      "step": 83440
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.358585858585859e-06,
+      "loss": 1.4861,
+      "step": 83450
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.353535353535353e-06,
+      "loss": 1.5018,
+      "step": 83460
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.348484848484849e-06,
+      "loss": 1.4734,
+      "step": 83470
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 8.343434343434343e-06,
+      "loss": 1.4659,
+      "step": 83480
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.338383838383838e-06,
+      "loss": 1.477,
+      "step": 83490
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.333333333333334e-06,
+      "loss": 1.4765,
+      "step": 83500
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.328282828282828e-06,
+      "loss": 1.478,
+      "step": 83510
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.323232323232322e-06,
+      "loss": 1.4481,
+      "step": 83520
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.318181818181818e-06,
+      "loss": 1.4792,
+      "step": 83530
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.313131313131313e-06,
+      "loss": 1.4585,
+      "step": 83540
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.308080808080809e-06,
+      "loss": 1.4562,
+      "step": 83550
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.303030303030303e-06,
+      "loss": 1.4786,
+      "step": 83560
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.297979797979797e-06,
+      "loss": 1.486,
+      "step": 83570
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.292929292929293e-06,
+      "loss": 1.4768,
+      "step": 83580
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.28787878787879e-06,
+      "loss": 1.4841,
+      "step": 83590
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.282828282828283e-06,
+      "loss": 1.4856,
+      "step": 83600
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.27777777777778e-06,
+      "loss": 1.4936,
+      "step": 83610
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.272727272727274e-06,
+      "loss": 1.492,
+      "step": 83620
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.267676767676768e-06,
+      "loss": 1.4595,
+      "step": 83630
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.262626262626264e-06,
+      "loss": 1.4664,
+      "step": 83640
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.257575757575758e-06,
+      "loss": 1.4592,
+      "step": 83650
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.252525252525253e-06,
+      "loss": 1.4965,
+      "step": 83660
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.247474747474749e-06,
+      "loss": 1.4655,
+      "step": 83670
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.242424242424243e-06,
+      "loss": 1.4794,
+      "step": 83680
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.237373737373739e-06,
+      "loss": 1.4554,
+      "step": 83690
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.232323232323233e-06,
+      "loss": 1.4613,
+      "step": 83700
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.227272727272728e-06,
+      "loss": 1.4801,
+      "step": 83710
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.222222222222223e-06,
+      "loss": 1.4696,
+      "step": 83720
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.217171717171718e-06,
+      "loss": 1.4741,
+      "step": 83730
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.212121212121212e-06,
+      "loss": 1.4944,
+      "step": 83740
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.207070707070708e-06,
+      "loss": 1.4832,
+      "step": 83750
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.202020202020202e-06,
+      "loss": 1.4878,
+      "step": 83760
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.196969696969697e-06,
+      "loss": 1.455,
+      "step": 83770
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.191919191919193e-06,
+      "loss": 1.466,
+      "step": 83780
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.186868686868687e-06,
+      "loss": 1.4617,
+      "step": 83790
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.181818181818183e-06,
+      "loss": 1.4904,
+      "step": 83800
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.176767676767677e-06,
+      "loss": 1.4854,
+      "step": 83810
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.171717171717172e-06,
+      "loss": 1.4838,
+      "step": 83820
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.166666666666668e-06,
+      "loss": 1.4593,
+      "step": 83830
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.161616161616162e-06,
+      "loss": 1.4643,
+      "step": 83840
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.156565656565656e-06,
+      "loss": 1.5,
+      "step": 83850
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.151515151515152e-06,
+      "loss": 1.4922,
+      "step": 83860
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.146464646464646e-06,
+      "loss": 1.4662,
+      "step": 83870
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.14141414141414e-06,
+      "loss": 1.4762,
+      "step": 83880
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.136363636363637e-06,
+      "loss": 1.4846,
+      "step": 83890
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.131313131313131e-06,
+      "loss": 1.4796,
+      "step": 83900
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.126262626262627e-06,
+      "loss": 1.4732,
+      "step": 83910
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.121212121212121e-06,
+      "loss": 1.4552,
+      "step": 83920
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.116161616161616e-06,
+      "loss": 1.4871,
+      "step": 83930
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.111111111111112e-06,
+      "loss": 1.4631,
+      "step": 83940
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.106060606060606e-06,
+      "loss": 1.4679,
+      "step": 83950
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.1010101010101e-06,
+      "loss": 1.4509,
+      "step": 83960
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.095959595959596e-06,
+      "loss": 1.4677,
+      "step": 83970
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.09090909090909e-06,
+      "loss": 1.4663,
+      "step": 83980
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.085858585858585e-06,
+      "loss": 1.4685,
+      "step": 83990
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.080808080808082e-06,
+      "loss": 1.4483,
+      "step": 84000
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.075757575757577e-06,
+      "loss": 1.4767,
+      "step": 84010
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.070707070707071e-06,
+      "loss": 1.4756,
+      "step": 84020
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.065656565656567e-06,
+      "loss": 1.4792,
+      "step": 84030
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.060606060606061e-06,
+      "loss": 1.4594,
+      "step": 84040
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.055555555555557e-06,
+      "loss": 1.4616,
+      "step": 84050
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.050505050505052e-06,
+      "loss": 1.4742,
+      "step": 84060
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.045454545454546e-06,
+      "loss": 1.473,
+      "step": 84070
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.040404040404042e-06,
+      "loss": 1.4741,
+      "step": 84080
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.035353535353536e-06,
+      "loss": 1.4651,
+      "step": 84090
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.03030303030303e-06,
+      "loss": 1.4724,
+      "step": 84100
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.025252525252526e-06,
+      "loss": 1.4734,
+      "step": 84110
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.02020202020202e-06,
+      "loss": 1.4752,
+      "step": 84120
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.015151515151515e-06,
+      "loss": 1.4628,
+      "step": 84130
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.010101010101011e-06,
+      "loss": 1.4847,
+      "step": 84140
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 8.005050505050505e-06,
+      "loss": 1.4408,
+      "step": 84150
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 1.4569,
+      "step": 84160
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.994949494949496e-06,
+      "loss": 1.4641,
+      "step": 84170
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.98989898989899e-06,
+      "loss": 1.4798,
+      "step": 84180
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.984848484848486e-06,
+      "loss": 1.4826,
+      "step": 84190
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.97979797979798e-06,
+      "loss": 1.488,
+      "step": 84200
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.974747474747475e-06,
+      "loss": 1.4721,
+      "step": 84210
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.96969696969697e-06,
+      "loss": 1.4527,
+      "step": 84220
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.964646464646465e-06,
+      "loss": 1.4642,
+      "step": 84230
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.959595959595959e-06,
+      "loss": 1.4675,
+      "step": 84240
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.954545454545455e-06,
+      "loss": 1.4767,
+      "step": 84250
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.94949494949495e-06,
+      "loss": 1.4766,
+      "step": 84260
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.944444444444445e-06,
+      "loss": 1.4577,
+      "step": 84270
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.93939393939394e-06,
+      "loss": 1.488,
+      "step": 84280
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.934343434343434e-06,
+      "loss": 1.4705,
+      "step": 84290
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.92929292929293e-06,
+      "loss": 1.5026,
+      "step": 84300
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.924242424242424e-06,
+      "loss": 1.4834,
+      "step": 84310
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.919191919191919e-06,
+      "loss": 1.454,
+      "step": 84320
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.914141414141415e-06,
+      "loss": 1.5045,
+      "step": 84330
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.909090909090909e-06,
+      "loss": 1.4697,
+      "step": 84340
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.904040404040403e-06,
+      "loss": 1.4621,
+      "step": 84350
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.898989898989899e-06,
+      "loss": 1.4715,
+      "step": 84360
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.893939393939393e-06,
+      "loss": 1.4742,
+      "step": 84370
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.88888888888889e-06,
+      "loss": 1.4751,
+      "step": 84380
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.883838383838384e-06,
+      "loss": 1.4744,
+      "step": 84390
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.878787878787878e-06,
+      "loss": 1.4434,
+      "step": 84400
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.873737373737374e-06,
+      "loss": 1.4516,
+      "step": 84410
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.868686868686868e-06,
+      "loss": 1.4674,
+      "step": 84420
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.863636363636364e-06,
+      "loss": 1.4867,
+      "step": 84430
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.85858585858586e-06,
+      "loss": 1.4729,
+      "step": 84440
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.853535353535355e-06,
+      "loss": 1.4703,
+      "step": 84450
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.848484848484849e-06,
+      "loss": 1.4611,
+      "step": 84460
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.843434343434345e-06,
+      "loss": 1.4872,
+      "step": 84470
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.838383838383839e-06,
+      "loss": 1.4604,
+      "step": 84480
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.833333333333333e-06,
+      "loss": 1.4889,
+      "step": 84490
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.82828282828283e-06,
+      "loss": 1.4712,
+      "step": 84500
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.823232323232324e-06,
+      "loss": 1.486,
+      "step": 84510
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.81818181818182e-06,
+      "loss": 1.4743,
+      "step": 84520
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.813131313131314e-06,
+      "loss": 1.4554,
+      "step": 84530
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.808080808080808e-06,
+      "loss": 1.4705,
+      "step": 84540
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.803030303030304e-06,
+      "loss": 1.4737,
+      "step": 84550
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.797979797979799e-06,
+      "loss": 1.4728,
+      "step": 84560
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.792929292929293e-06,
+      "loss": 1.4418,
+      "step": 84570
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.787878787878789e-06,
+      "loss": 1.4522,
+      "step": 84580
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.782828282828283e-06,
+      "loss": 1.4716,
+      "step": 84590
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.777777777777777e-06,
+      "loss": 1.4923,
+      "step": 84600
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.772727272727273e-06,
+      "loss": 1.4731,
+      "step": 84610
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.767676767676768e-06,
+      "loss": 1.4751,
+      "step": 84620
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.762626262626264e-06,
+      "loss": 1.4578,
+      "step": 84630
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.757575757575758e-06,
+      "loss": 1.4686,
+      "step": 84640
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.752525252525252e-06,
+      "loss": 1.4495,
+      "step": 84650
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.747474747474748e-06,
+      "loss": 1.4994,
+      "step": 84660
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.742424242424243e-06,
+      "loss": 1.4575,
+      "step": 84670
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.737373737373737e-06,
+      "loss": 1.4998,
+      "step": 84680
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.732323232323233e-06,
+      "loss": 1.4726,
+      "step": 84690
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.727272727272727e-06,
+      "loss": 1.4487,
+      "step": 84700
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.722222222222223e-06,
+      "loss": 1.454,
+      "step": 84710
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.717171717171717e-06,
+      "loss": 1.4676,
+      "step": 84720
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.712121212121212e-06,
+      "loss": 1.4615,
+      "step": 84730
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.707070707070708e-06,
+      "loss": 1.4757,
+      "step": 84740
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.702020202020202e-06,
+      "loss": 1.4685,
+      "step": 84750
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.696969696969696e-06,
+      "loss": 1.4733,
+      "step": 84760
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.691919191919192e-06,
+      "loss": 1.4579,
+      "step": 84770
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.686868686868687e-06,
+      "loss": 1.4664,
+      "step": 84780
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.681818181818181e-06,
+      "loss": 1.4694,
+      "step": 84790
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.676767676767677e-06,
+      "loss": 1.4767,
+      "step": 84800
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.671717171717171e-06,
+      "loss": 1.4812,
+      "step": 84810
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.666666666666667e-06,
+      "loss": 1.4577,
+      "step": 84820
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.661616161616162e-06,
+      "loss": 1.4759,
+      "step": 84830
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.656565656565656e-06,
+      "loss": 1.4711,
+      "step": 84840
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.651515151515152e-06,
+      "loss": 1.4509,
+      "step": 84850
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.646464646464648e-06,
+      "loss": 1.439,
+      "step": 84860
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.641414141414142e-06,
+      "loss": 1.4677,
+      "step": 84870
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.636363636363638e-06,
+      "loss": 1.4557,
+      "step": 84880
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.631313131313132e-06,
+      "loss": 1.4787,
+      "step": 84890
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.6262626262626275e-06,
+      "loss": 1.4584,
+      "step": 84900
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.621212121212122e-06,
+      "loss": 1.4737,
+      "step": 84910
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.616161616161617e-06,
+      "loss": 1.4687,
+      "step": 84920
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.611111111111112e-06,
+      "loss": 1.4827,
+      "step": 84930
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.606060606060606e-06,
+      "loss": 1.4719,
+      "step": 84940
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.6010101010101016e-06,
+      "loss": 1.4604,
+      "step": 84950
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.595959595959597e-06,
+      "loss": 1.4937,
+      "step": 84960
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.590909090909092e-06,
+      "loss": 1.4638,
+      "step": 84970
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.585858585858586e-06,
+      "loss": 1.4825,
+      "step": 84980
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.580808080808081e-06,
+      "loss": 1.4748,
+      "step": 84990
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.5757575757575764e-06,
+      "loss": 1.4398,
+      "step": 85000
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.5707070707070716e-06,
+      "loss": 1.4576,
+      "step": 85010
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.565656565656566e-06,
+      "loss": 1.4665,
+      "step": 85020
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.560606060606061e-06,
+      "loss": 1.4795,
+      "step": 85030
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.555555555555556e-06,
+      "loss": 1.476,
+      "step": 85040
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.550505050505051e-06,
+      "loss": 1.4651,
+      "step": 85050
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.545454545454546e-06,
+      "loss": 1.4779,
+      "step": 85060
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.540404040404041e-06,
+      "loss": 1.4522,
+      "step": 85070
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.535353535353536e-06,
+      "loss": 1.4685,
+      "step": 85080
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.53030303030303e-06,
+      "loss": 1.4788,
+      "step": 85090
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.525252525252525e-06,
+      "loss": 1.4694,
+      "step": 85100
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.5202020202020205e-06,
+      "loss": 1.4822,
+      "step": 85110
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.515151515151516e-06,
+      "loss": 1.487,
+      "step": 85120
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.51010101010101e-06,
+      "loss": 1.4733,
+      "step": 85130
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.505050505050505e-06,
+      "loss": 1.4755,
+      "step": 85140
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.5e-06,
+      "loss": 1.4696,
+      "step": 85150
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.494949494949495e-06,
+      "loss": 1.4779,
+      "step": 85160
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.48989898989899e-06,
+      "loss": 1.4702,
+      "step": 85170
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.484848484848485e-06,
+      "loss": 1.4537,
+      "step": 85180
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.47979797979798e-06,
+      "loss": 1.4574,
+      "step": 85190
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.474747474747475e-06,
+      "loss": 1.4624,
+      "step": 85200
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.469696969696969e-06,
+      "loss": 1.4658,
+      "step": 85210
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4646464646464645e-06,
+      "loss": 1.4865,
+      "step": 85220
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.45959595959596e-06,
+      "loss": 1.4885,
+      "step": 85230
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.454545454545454e-06,
+      "loss": 1.4641,
+      "step": 85240
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.449494949494949e-06,
+      "loss": 1.4664,
+      "step": 85250
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.444444444444444e-06,
+      "loss": 1.4563,
+      "step": 85260
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.439393939393939e-06,
+      "loss": 1.4834,
+      "step": 85270
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.434343434343435e-06,
+      "loss": 1.4805,
+      "step": 85280
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4292929292929305e-06,
+      "loss": 1.4785,
+      "step": 85290
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.424242424242425e-06,
+      "loss": 1.4768,
+      "step": 85300
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.41919191919192e-06,
+      "loss": 1.4842,
+      "step": 85310
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.414141414141415e-06,
+      "loss": 1.4586,
+      "step": 85320
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.40909090909091e-06,
+      "loss": 1.4722,
+      "step": 85330
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4040404040404045e-06,
+      "loss": 1.4576,
+      "step": 85340
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.3989898989899e-06,
+      "loss": 1.4781,
+      "step": 85350
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.393939393939395e-06,
+      "loss": 1.4805,
+      "step": 85360
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.38888888888889e-06,
+      "loss": 1.4744,
+      "step": 85370
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.383838383838384e-06,
+      "loss": 1.4646,
+      "step": 85380
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.378787878787879e-06,
+      "loss": 1.471,
+      "step": 85390
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.3737373737373745e-06,
+      "loss": 1.4972,
+      "step": 85400
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.36868686868687e-06,
+      "loss": 1.4544,
+      "step": 85410
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.363636363636364e-06,
+      "loss": 1.499,
+      "step": 85420
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.358585858585859e-06,
+      "loss": 1.4822,
+      "step": 85430
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.353535353535354e-06,
+      "loss": 1.4696,
+      "step": 85440
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.3484848484848486e-06,
+      "loss": 1.4647,
+      "step": 85450
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.343434343434344e-06,
+      "loss": 1.4719,
+      "step": 85460
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.338383838383839e-06,
+      "loss": 1.4833,
+      "step": 85470
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.333333333333334e-06,
+      "loss": 1.4426,
+      "step": 85480
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.328282828282828e-06,
+      "loss": 1.4666,
+      "step": 85490
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3232323232323234e-06,
+      "loss": 1.4685,
+      "step": 85500
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3181818181818186e-06,
+      "loss": 1.4831,
+      "step": 85510
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.313131313131314e-06,
+      "loss": 1.473,
+      "step": 85520
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.308080808080808e-06,
+      "loss": 1.4736,
+      "step": 85530
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.303030303030303e-06,
+      "loss": 1.4485,
+      "step": 85540
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.297979797979798e-06,
+      "loss": 1.4585,
+      "step": 85550
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.2929292929292934e-06,
+      "loss": 1.4864,
+      "step": 85560
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.287878787878788e-06,
+      "loss": 1.4754,
+      "step": 85570
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.282828282828283e-06,
+      "loss": 1.4671,
+      "step": 85580
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.277777777777778e-06,
+      "loss": 1.4535,
+      "step": 85590
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.272727272727272e-06,
+      "loss": 1.4807,
+      "step": 85600
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.2676767676767675e-06,
+      "loss": 1.4788,
+      "step": 85610
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.262626262626263e-06,
+      "loss": 1.4734,
+      "step": 85620
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.257575757575758e-06,
+      "loss": 1.4587,
+      "step": 85630
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.252525252525252e-06,
+      "loss": 1.4703,
+      "step": 85640
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.247474747474747e-06,
+      "loss": 1.4577,
+      "step": 85650
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.242424242424242e-06,
+      "loss": 1.4743,
+      "step": 85660
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.2373737373737375e-06,
+      "loss": 1.4548,
+      "step": 85670
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.232323232323232e-06,
+      "loss": 1.4707,
+      "step": 85680
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.227272727272727e-06,
+      "loss": 1.4794,
+      "step": 85690
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.222222222222222e-06,
+      "loss": 1.4868,
+      "step": 85700
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.217171717171718e-06,
+      "loss": 1.4623,
+      "step": 85710
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.212121212121213e-06,
+      "loss": 1.4605,
+      "step": 85720
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.207070707070708e-06,
+      "loss": 1.4548,
+      "step": 85730
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.202020202020203e-06,
+      "loss": 1.4564,
+      "step": 85740
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.196969696969698e-06,
+      "loss": 1.4785,
+      "step": 85750
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.191919191919193e-06,
+      "loss": 1.4519,
+      "step": 85760
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.186868686868688e-06,
+      "loss": 1.4831,
+      "step": 85770
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.181818181818182e-06,
+      "loss": 1.4782,
+      "step": 85780
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.1767676767676775e-06,
+      "loss": 1.471,
+      "step": 85790
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.171717171717173e-06,
+      "loss": 1.4639,
+      "step": 85800
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.166666666666667e-06,
+      "loss": 1.4761,
+      "step": 85810
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.161616161616162e-06,
+      "loss": 1.457,
+      "step": 85820
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.156565656565657e-06,
+      "loss": 1.4774,
+      "step": 85830
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.151515151515152e-06,
+      "loss": 1.4854,
+      "step": 85840
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.146464646464647e-06,
+      "loss": 1.4671,
+      "step": 85850
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.141414141414142e-06,
+      "loss": 1.4847,
+      "step": 85860
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.136363636363637e-06,
+      "loss": 1.4694,
+      "step": 85870
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.131313131313132e-06,
+      "loss": 1.4533,
+      "step": 85880
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.126262626262626e-06,
+      "loss": 1.4644,
+      "step": 85890
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.1212121212121215e-06,
+      "loss": 1.4867,
+      "step": 85900
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.116161616161617e-06,
+      "loss": 1.4856,
+      "step": 85910
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.111111111111112e-06,
+      "loss": 1.4929,
+      "step": 85920
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.106060606060606e-06,
+      "loss": 1.4808,
+      "step": 85930
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.101010101010101e-06,
+      "loss": 1.4948,
+      "step": 85940
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.095959595959596e-06,
+      "loss": 1.4823,
+      "step": 85950
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.090909090909091e-06,
+      "loss": 1.4902,
+      "step": 85960
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.085858585858586e-06,
+      "loss": 1.4735,
+      "step": 85970
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.080808080808081e-06,
+      "loss": 1.478,
+      "step": 85980
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.075757575757576e-06,
+      "loss": 1.4829,
+      "step": 85990
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.0707070707070704e-06,
+      "loss": 1.4797,
+      "step": 86000
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.0656565656565656e-06,
+      "loss": 1.4619,
+      "step": 86010
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.060606060606061e-06,
+      "loss": 1.4653,
+      "step": 86020
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.055555555555556e-06,
+      "loss": 1.4877,
+      "step": 86030
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.05050505050505e-06,
+      "loss": 1.4575,
+      "step": 86040
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.045454545454545e-06,
+      "loss": 1.4671,
+      "step": 86050
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.0404040404040404e-06,
+      "loss": 1.4617,
+      "step": 86060
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.035353535353536e-06,
+      "loss": 1.4681,
+      "step": 86070
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.03030303030303e-06,
+      "loss": 1.4734,
+      "step": 86080
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.025252525252525e-06,
+      "loss": 1.4691,
+      "step": 86090
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.02020202020202e-06,
+      "loss": 1.4687,
+      "step": 86100
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.0151515151515145e-06,
+      "loss": 1.4973,
+      "step": 86110
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.01010101010101e-06,
+      "loss": 1.4848,
+      "step": 86120
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.005050505050505e-06,
+      "loss": 1.4684,
+      "step": 86130
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.000000000000001e-06,
+      "loss": 1.4813,
+      "step": 86140
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.994949494949496e-06,
+      "loss": 1.4565,
+      "step": 86150
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.989898989898991e-06,
+      "loss": 1.4582,
+      "step": 86160
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.984848484848486e-06,
+      "loss": 1.4915,
+      "step": 86170
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.9797979797979805e-06,
+      "loss": 1.4718,
+      "step": 86180
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.974747474747476e-06,
+      "loss": 1.4619,
+      "step": 86190
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.969696969696971e-06,
+      "loss": 1.4571,
+      "step": 86200
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.964646464646465e-06,
+      "loss": 1.4823,
+      "step": 86210
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.95959595959596e-06,
+      "loss": 1.4812,
+      "step": 86220
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.954545454545455e-06,
+      "loss": 1.4529,
+      "step": 86230
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.9494949494949505e-06,
+      "loss": 1.4642,
+      "step": 86240
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.944444444444445e-06,
+      "loss": 1.4793,
+      "step": 86250
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.93939393939394e-06,
+      "loss": 1.4529,
+      "step": 86260
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.934343434343435e-06,
+      "loss": 1.4471,
+      "step": 86270
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.92929292929293e-06,
+      "loss": 1.459,
+      "step": 86280
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.9242424242424245e-06,
+      "loss": 1.4481,
+      "step": 86290
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.91919191919192e-06,
+      "loss": 1.4617,
+      "step": 86300
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.914141414141415e-06,
+      "loss": 1.468,
+      "step": 86310
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.909090909090909e-06,
+      "loss": 1.4669,
+      "step": 86320
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.904040404040404e-06,
+      "loss": 1.4515,
+      "step": 86330
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.898989898989899e-06,
+      "loss": 1.4807,
+      "step": 86340
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.8939393939393945e-06,
+      "loss": 1.4567,
+      "step": 86350
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.888888888888889e-06,
+      "loss": 1.465,
+      "step": 86360
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.883838383838384e-06,
+      "loss": 1.4633,
+      "step": 86370
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.878787878787879e-06,
+      "loss": 1.4847,
+      "step": 86380
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.873737373737374e-06,
+      "loss": 1.4611,
+      "step": 86390
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.8686868686868685e-06,
+      "loss": 1.4759,
+      "step": 86400
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.863636363636364e-06,
+      "loss": 1.4629,
+      "step": 86410
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.858585858585859e-06,
+      "loss": 1.4715,
+      "step": 86420
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.853535353535354e-06,
+      "loss": 1.4825,
+      "step": 86430
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.848484848484848e-06,
+      "loss": 1.4823,
+      "step": 86440
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.843434343434343e-06,
+      "loss": 1.4734,
+      "step": 86450
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.8383838383838386e-06,
+      "loss": 1.4649,
+      "step": 86460
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.833333333333333e-06,
+      "loss": 1.4531,
+      "step": 86470
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.828282828282828e-06,
+      "loss": 1.4609,
+      "step": 86480
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.823232323232323e-06,
+      "loss": 1.4456,
+      "step": 86490
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.818181818181818e-06,
+      "loss": 1.4846,
+      "step": 86500
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.8131313131313126e-06,
+      "loss": 1.4866,
+      "step": 86510
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.808080808080808e-06,
+      "loss": 1.4772,
+      "step": 86520
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.803030303030303e-06,
+      "loss": 1.4799,
+      "step": 86530
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.797979797979798e-06,
+      "loss": 1.4564,
+      "step": 86540
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.792929292929292e-06,
+      "loss": 1.4735,
+      "step": 86550
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.787878787878789e-06,
+      "loss": 1.4687,
+      "step": 86560
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.7828282828282834e-06,
+      "loss": 1.4862,
+      "step": 86570
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.777777777777779e-06,
+      "loss": 1.4504,
+      "step": 86580
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.772727272727274e-06,
+      "loss": 1.4868,
+      "step": 86590
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.767676767676769e-06,
+      "loss": 1.4627,
+      "step": 86600
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.762626262626263e-06,
+      "loss": 1.4597,
+      "step": 86610
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.757575757575758e-06,
+      "loss": 1.4763,
+      "step": 86620
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.7525252525252534e-06,
+      "loss": 1.4925,
+      "step": 86630
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.747474747474749e-06,
+      "loss": 1.4621,
+      "step": 86640
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.742424242424243e-06,
+      "loss": 1.4784,
+      "step": 86650
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.737373737373738e-06,
+      "loss": 1.4599,
+      "step": 86660
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.732323232323233e-06,
+      "loss": 1.4858,
+      "step": 86670
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.727272727272728e-06,
+      "loss": 1.4705,
+      "step": 86680
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.722222222222223e-06,
+      "loss": 1.4798,
+      "step": 86690
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.717171717171718e-06,
+      "loss": 1.4882,
+      "step": 86700
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.712121212121213e-06,
+      "loss": 1.4717,
+      "step": 86710
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.707070707070707e-06,
+      "loss": 1.4746,
+      "step": 86720
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.702020202020202e-06,
+      "loss": 1.4569,
+      "step": 86730
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.6969696969696975e-06,
+      "loss": 1.4672,
+      "step": 86740
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.691919191919193e-06,
+      "loss": 1.4643,
+      "step": 86750
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.686868686868687e-06,
+      "loss": 1.4658,
+      "step": 86760
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.681818181818182e-06,
+      "loss": 1.474,
+      "step": 86770
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.676767676767677e-06,
+      "loss": 1.4786,
+      "step": 86780
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.671717171717172e-06,
+      "loss": 1.4355,
+      "step": 86790
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 1.4723,
+      "step": 86800
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 6.661616161616162e-06,
+      "loss": 1.4963,
+      "step": 86810
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.656565656565657e-06,
+      "loss": 1.454,
+      "step": 86820
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.651515151515151e-06,
+      "loss": 1.4653,
+      "step": 86830
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.646464646464646e-06,
+      "loss": 1.4617,
+      "step": 86840
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.6414141414141415e-06,
+      "loss": 1.4713,
+      "step": 86850
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.636363636363637e-06,
+      "loss": 1.4657,
+      "step": 86860
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.631313131313131e-06,
+      "loss": 1.4828,
+      "step": 86870
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.626262626262626e-06,
+      "loss": 1.4483,
+      "step": 86880
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.621212121212121e-06,
+      "loss": 1.4673,
+      "step": 86890
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.616161616161616e-06,
+      "loss": 1.4586,
+      "step": 86900
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.611111111111111e-06,
+      "loss": 1.4656,
+      "step": 86910
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.606060606060606e-06,
+      "loss": 1.4802,
+      "step": 86920
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.601010101010101e-06,
+      "loss": 1.4627,
+      "step": 86930
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.595959595959596e-06,
+      "loss": 1.4813,
+      "step": 86940
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.59090909090909e-06,
+      "loss": 1.4923,
+      "step": 86950
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.5858585858585856e-06,
+      "loss": 1.4916,
+      "step": 86960
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.580808080808081e-06,
+      "loss": 1.4481,
+      "step": 86970
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.575757575757575e-06,
+      "loss": 1.4812,
+      "step": 86980
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.570707070707072e-06,
+      "loss": 1.4886,
+      "step": 86990
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.565656565656567e-06,
+      "loss": 1.4599,
+      "step": 87000
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.560606060606061e-06,
+      "loss": 1.4872,
+      "step": 87010
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.555555555555556e-06,
+      "loss": 1.4905,
+      "step": 87020
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.5505050505050516e-06,
+      "loss": 1.466,
+      "step": 87030
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.545454545454547e-06,
+      "loss": 1.4691,
+      "step": 87040
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.540404040404041e-06,
+      "loss": 1.4554,
+      "step": 87050
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.535353535353536e-06,
+      "loss": 1.4592,
+      "step": 87060
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.530303030303031e-06,
+      "loss": 1.471,
+      "step": 87070
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.525252525252526e-06,
+      "loss": 1.4601,
+      "step": 87080
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.520202020202021e-06,
+      "loss": 1.4523,
+      "step": 87090
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.515151515151516e-06,
+      "loss": 1.4788,
+      "step": 87100
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.510101010101011e-06,
+      "loss": 1.4498,
+      "step": 87110
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.505050505050505e-06,
+      "loss": 1.4781,
+      "step": 87120
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.5000000000000004e-06,
+      "loss": 1.4755,
+      "step": 87130
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.494949494949496e-06,
+      "loss": 1.4589,
+      "step": 87140
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.489898989898991e-06,
+      "loss": 1.4546,
+      "step": 87150
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.484848484848485e-06,
+      "loss": 1.4825,
+      "step": 87160
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.47979797979798e-06,
+      "loss": 1.4654,
+      "step": 87170
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.474747474747475e-06,
+      "loss": 1.4834,
+      "step": 87180
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.46969696969697e-06,
+      "loss": 1.4843,
+      "step": 87190
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.464646464646465e-06,
+      "loss": 1.4756,
+      "step": 87200
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.45959595959596e-06,
+      "loss": 1.4671,
+      "step": 87210
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.454545454545455e-06,
+      "loss": 1.4624,
+      "step": 87220
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.449494949494949e-06,
+      "loss": 1.4693,
+      "step": 87230
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.4444444444444445e-06,
+      "loss": 1.4891,
+      "step": 87240
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.43939393939394e-06,
+      "loss": 1.4814,
+      "step": 87250
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.434343434343435e-06,
+      "loss": 1.4666,
+      "step": 87260
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.429292929292929e-06,
+      "loss": 1.4799,
+      "step": 87270
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.424242424242424e-06,
+      "loss": 1.48,
+      "step": 87280
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.419191919191919e-06,
+      "loss": 1.4667,
+      "step": 87290
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.4141414141414145e-06,
+      "loss": 1.4661,
+      "step": 87300
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.409090909090909e-06,
+      "loss": 1.4717,
+      "step": 87310
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.404040404040404e-06,
+      "loss": 1.4727,
+      "step": 87320
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.398989898989899e-06,
+      "loss": 1.46,
+      "step": 87330
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.393939393939393e-06,
+      "loss": 1.4709,
+      "step": 87340
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.3888888888888885e-06,
+      "loss": 1.4693,
+      "step": 87350
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.383838383838384e-06,
+      "loss": 1.4782,
+      "step": 87360
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.378787878787879e-06,
+      "loss": 1.4525,
+      "step": 87370
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.373737373737373e-06,
+      "loss": 1.4746,
+      "step": 87380
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.368686868686868e-06,
+      "loss": 1.4716,
+      "step": 87390
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.363636363636363e-06,
+      "loss": 1.454,
+      "step": 87400
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.3585858585858585e-06,
+      "loss": 1.4552,
+      "step": 87410
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.3535353535353545e-06,
+      "loss": 1.4967,
+      "step": 87420
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.34848484848485e-06,
+      "loss": 1.4849,
+      "step": 87430
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.343434343434344e-06,
+      "loss": 1.4857,
+      "step": 87440
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.338383838383839e-06,
+      "loss": 1.4727,
+      "step": 87450
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.333333333333334e-06,
+      "loss": 1.4631,
+      "step": 87460
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 6.328282828282829e-06,
+      "loss": 1.4575,
+      "step": 87470
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.323232323232324e-06,
+      "loss": 1.4942,
+      "step": 87480
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.318181818181819e-06,
+      "loss": 1.4813,
+      "step": 87490
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.313131313131314e-06,
+      "loss": 1.4858,
+      "step": 87500
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.308080808080809e-06,
+      "loss": 1.4536,
+      "step": 87510
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.303030303030303e-06,
+      "loss": 1.4846,
+      "step": 87520
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.2979797979797986e-06,
+      "loss": 1.4716,
+      "step": 87530
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.292929292929294e-06,
+      "loss": 1.4701,
+      "step": 87540
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.287878787878789e-06,
+      "loss": 1.4853,
+      "step": 87550
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.282828282828283e-06,
+      "loss": 1.4435,
+      "step": 87560
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.277777777777778e-06,
+      "loss": 1.4835,
+      "step": 87570
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.2727272727272734e-06,
+      "loss": 1.4816,
+      "step": 87580
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.267676767676768e-06,
+      "loss": 1.4545,
+      "step": 87590
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.262626262626263e-06,
+      "loss": 1.4794,
+      "step": 87600
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.257575757575758e-06,
+      "loss": 1.4506,
+      "step": 87610
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.252525252525253e-06,
+      "loss": 1.4651,
+      "step": 87620
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.2474747474747474e-06,
+      "loss": 1.462,
+      "step": 87630
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.242424242424243e-06,
+      "loss": 1.472,
+      "step": 87640
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.237373737373738e-06,
+      "loss": 1.4621,
+      "step": 87650
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.232323232323233e-06,
+      "loss": 1.4752,
+      "step": 87660
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.227272727272727e-06,
+      "loss": 1.4803,
+      "step": 87670
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.222222222222222e-06,
+      "loss": 1.4933,
+      "step": 87680
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.2171717171717175e-06,
+      "loss": 1.4791,
+      "step": 87690
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.212121212121212e-06,
+      "loss": 1.4789,
+      "step": 87700
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.207070707070707e-06,
+      "loss": 1.4693,
+      "step": 87710
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.202020202020202e-06,
+      "loss": 1.4622,
+      "step": 87720
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.196969696969697e-06,
+      "loss": 1.4769,
+      "step": 87730
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.191919191919192e-06,
+      "loss": 1.456,
+      "step": 87740
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.1868686868686875e-06,
+      "loss": 1.459,
+      "step": 87750
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.181818181818183e-06,
+      "loss": 1.4665,
+      "step": 87760
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.176767676767677e-06,
+      "loss": 1.455,
+      "step": 87770
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.171717171717172e-06,
+      "loss": 1.4587,
+      "step": 87780
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.166666666666667e-06,
+      "loss": 1.4706,
+      "step": 87790
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.161616161616162e-06,
+      "loss": 1.4564,
+      "step": 87800
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.156565656565657e-06,
+      "loss": 1.4631,
+      "step": 87810
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.151515151515152e-06,
+      "loss": 1.4684,
+      "step": 87820
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.146464646464647e-06,
+      "loss": 1.4639,
+      "step": 87830
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.141414141414142e-06,
+      "loss": 1.4658,
+      "step": 87840
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.136363636363636e-06,
+      "loss": 1.464,
+      "step": 87850
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.1313131313131315e-06,
+      "loss": 1.4797,
+      "step": 87860
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.126262626262627e-06,
+      "loss": 1.4773,
+      "step": 87870
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.121212121212121e-06,
+      "loss": 1.463,
+      "step": 87880
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.116161616161616e-06,
+      "loss": 1.4695,
+      "step": 87890
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.111111111111111e-06,
+      "loss": 1.4591,
+      "step": 87900
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.106060606060606e-06,
+      "loss": 1.4663,
+      "step": 87910
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.101010101010101e-06,
+      "loss": 1.4891,
+      "step": 87920
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.095959595959596e-06,
+      "loss": 1.487,
+      "step": 87930
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.090909090909091e-06,
+      "loss": 1.4799,
+      "step": 87940
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.085858585858586e-06,
+      "loss": 1.455,
+      "step": 87950
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.080808080808081e-06,
+      "loss": 1.4884,
+      "step": 87960
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.075757575757576e-06,
+      "loss": 1.4765,
+      "step": 87970
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.0707070707070715e-06,
+      "loss": 1.4778,
+      "step": 87980
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.065656565656566e-06,
+      "loss": 1.4516,
+      "step": 87990
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.060606060606061e-06,
+      "loss": 1.4588,
+      "step": 88000
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.055555555555556e-06,
+      "loss": 1.4521,
+      "step": 88010
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.050505050505051e-06,
+      "loss": 1.4638,
+      "step": 88020
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.0454545454545456e-06,
+      "loss": 1.4767,
+      "step": 88030
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.040404040404041e-06,
+      "loss": 1.4821,
+      "step": 88040
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.035353535353536e-06,
+      "loss": 1.4879,
+      "step": 88050
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.030303030303031e-06,
+      "loss": 1.4593,
+      "step": 88060
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.025252525252525e-06,
+      "loss": 1.4758,
+      "step": 88070
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.0202020202020204e-06,
+      "loss": 1.4784,
+      "step": 88080
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.0151515151515156e-06,
+      "loss": 1.4712,
+      "step": 88090
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.01010101010101e-06,
+      "loss": 1.4812,
+      "step": 88100
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.005050505050505e-06,
+      "loss": 1.4541,
+      "step": 88110
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6e-06,
+      "loss": 1.4851,
+      "step": 88120
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.994949494949495e-06,
+      "loss": 1.4954,
+      "step": 88130
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.98989898989899e-06,
+      "loss": 1.4595,
+      "step": 88140
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.984848484848485e-06,
+      "loss": 1.4629,
+      "step": 88150
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.979797979797981e-06,
+      "loss": 1.4796,
+      "step": 88160
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.974747474747475e-06,
+      "loss": 1.4744,
+      "step": 88170
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.96969696969697e-06,
+      "loss": 1.4524,
+      "step": 88180
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.964646464646465e-06,
+      "loss": 1.4413,
+      "step": 88190
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.9595959595959605e-06,
+      "loss": 1.473,
+      "step": 88200
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.954545454545455e-06,
+      "loss": 1.4584,
+      "step": 88210
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.94949494949495e-06,
+      "loss": 1.4581,
+      "step": 88220
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.944444444444445e-06,
+      "loss": 1.4527,
+      "step": 88230
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.93939393939394e-06,
+      "loss": 1.4713,
+      "step": 88240
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.9343434343434345e-06,
+      "loss": 1.4552,
+      "step": 88250
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.92929292929293e-06,
+      "loss": 1.4752,
+      "step": 88260
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.924242424242425e-06,
+      "loss": 1.4741,
+      "step": 88270
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.919191919191919e-06,
+      "loss": 1.4609,
+      "step": 88280
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.914141414141414e-06,
+      "loss": 1.4672,
+      "step": 88290
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.909090909090909e-06,
+      "loss": 1.4727,
+      "step": 88300
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.9040404040404045e-06,
+      "loss": 1.4881,
+      "step": 88310
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.898989898989899e-06,
+      "loss": 1.4619,
+      "step": 88320
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.893939393939394e-06,
+      "loss": 1.4691,
+      "step": 88330
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.888888888888889e-06,
+      "loss": 1.4796,
+      "step": 88340
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.883838383838384e-06,
+      "loss": 1.468,
+      "step": 88350
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.8787878787878785e-06,
+      "loss": 1.4691,
+      "step": 88360
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.873737373737374e-06,
+      "loss": 1.4508,
+      "step": 88370
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.86868686868687e-06,
+      "loss": 1.4604,
+      "step": 88380
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.863636363636364e-06,
+      "loss": 1.4694,
+      "step": 88390
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.858585858585859e-06,
+      "loss": 1.482,
+      "step": 88400
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.853535353535354e-06,
+      "loss": 1.4516,
+      "step": 88410
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.848484848484849e-06,
+      "loss": 1.4779,
+      "step": 88420
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.843434343434344e-06,
+      "loss": 1.4674,
+      "step": 88430
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.838383838383839e-06,
+      "loss": 1.4795,
+      "step": 88440
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.833333333333334e-06,
+      "loss": 1.4618,
+      "step": 88450
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.828282828282828e-06,
+      "loss": 1.4627,
+      "step": 88460
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.823232323232323e-06,
+      "loss": 1.4852,
+      "step": 88470
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.8181818181818185e-06,
+      "loss": 1.4548,
+      "step": 88480
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.813131313131314e-06,
+      "loss": 1.4792,
+      "step": 88490
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.808080808080808e-06,
+      "loss": 1.4708,
+      "step": 88500
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.803030303030303e-06,
+      "loss": 1.4604,
+      "step": 88510
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.797979797979798e-06,
+      "loss": 1.4637,
+      "step": 88520
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.792929292929293e-06,
+      "loss": 1.4658,
+      "step": 88530
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.787878787878788e-06,
+      "loss": 1.4705,
+      "step": 88540
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.782828282828283e-06,
+      "loss": 1.4746,
+      "step": 88550
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.777777777777778e-06,
+      "loss": 1.4751,
+      "step": 88560
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.772727272727272e-06,
+      "loss": 1.4463,
+      "step": 88570
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.7676767676767674e-06,
+      "loss": 1.4662,
+      "step": 88580
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.762626262626263e-06,
+      "loss": 1.4638,
+      "step": 88590
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.7575757575757586e-06,
+      "loss": 1.4841,
+      "step": 88600
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.752525252525253e-06,
+      "loss": 1.4767,
+      "step": 88610
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.747474747474748e-06,
+      "loss": 1.4789,
+      "step": 88620
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.742424242424243e-06,
+      "loss": 1.4711,
+      "step": 88630
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.7373737373737374e-06,
+      "loss": 1.4681,
+      "step": 88640
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.732323232323233e-06,
+      "loss": 1.4513,
+      "step": 88650
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.727272727272728e-06,
+      "loss": 1.4667,
+      "step": 88660
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.722222222222223e-06,
+      "loss": 1.4817,
+      "step": 88670
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.717171717171717e-06,
+      "loss": 1.4705,
+      "step": 88680
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.712121212121212e-06,
+      "loss": 1.443,
+      "step": 88690
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.7070707070707075e-06,
+      "loss": 1.4701,
+      "step": 88700
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.702020202020203e-06,
+      "loss": 1.4748,
+      "step": 88710
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.696969696969697e-06,
+      "loss": 1.4674,
+      "step": 88720
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.691919191919192e-06,
+      "loss": 1.4619,
+      "step": 88730
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.686868686868687e-06,
+      "loss": 1.4821,
+      "step": 88740
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.681818181818182e-06,
+      "loss": 1.4838,
+      "step": 88750
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.676767676767677e-06,
+      "loss": 1.4784,
+      "step": 88760
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.671717171717172e-06,
+      "loss": 1.4669,
+      "step": 88770
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.666666666666667e-06,
+      "loss": 1.4667,
+      "step": 88780
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.661616161616161e-06,
+      "loss": 1.465,
+      "step": 88790
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.656565656565657e-06,
+      "loss": 1.4569,
+      "step": 88800
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.651515151515152e-06,
+      "loss": 1.4611,
+      "step": 88810
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.646464646464647e-06,
+      "loss": 1.4963,
+      "step": 88820
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.641414141414142e-06,
+      "loss": 1.4689,
+      "step": 88830
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.636363636363637e-06,
+      "loss": 1.4632,
+      "step": 88840
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.631313131313132e-06,
+      "loss": 1.4785,
+      "step": 88850
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.626262626262626e-06,
+      "loss": 1.4787,
+      "step": 88860
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.6212121212121215e-06,
+      "loss": 1.4932,
+      "step": 88870
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.616161616161617e-06,
+      "loss": 1.4868,
+      "step": 88880
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.611111111111112e-06,
+      "loss": 1.4738,
+      "step": 88890
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.606060606060606e-06,
+      "loss": 1.4709,
+      "step": 88900
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.601010101010101e-06,
+      "loss": 1.4823,
+      "step": 88910
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.595959595959596e-06,
+      "loss": 1.481,
+      "step": 88920
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.5909090909090915e-06,
+      "loss": 1.4636,
+      "step": 88930
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.585858585858586e-06,
+      "loss": 1.4698,
+      "step": 88940
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.580808080808081e-06,
+      "loss": 1.4566,
+      "step": 88950
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.575757575757576e-06,
+      "loss": 1.478,
+      "step": 88960
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.57070707070707e-06,
+      "loss": 1.463,
+      "step": 88970
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.5656565656565655e-06,
+      "loss": 1.4479,
+      "step": 88980
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.560606060606061e-06,
+      "loss": 1.479,
+      "step": 88990
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.555555555555556e-06,
+      "loss": 1.4861,
+      "step": 89000
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.55050505050505e-06,
+      "loss": 1.4758,
+      "step": 89010
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.545454545454546e-06,
+      "loss": 1.4736,
+      "step": 89020
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.540404040404041e-06,
+      "loss": 1.4611,
+      "step": 89030
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.5353535353535355e-06,
+      "loss": 1.4686,
+      "step": 89040
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.530303030303031e-06,
+      "loss": 1.4911,
+      "step": 89050
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.525252525252526e-06,
+      "loss": 1.4699,
+      "step": 89060
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.520202020202021e-06,
+      "loss": 1.4816,
+      "step": 89070
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.515151515151515e-06,
+      "loss": 1.4738,
+      "step": 89080
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.51010101010101e-06,
+      "loss": 1.4878,
+      "step": 89090
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.5050505050505056e-06,
+      "loss": 1.4915,
+      "step": 89100
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.500000000000001e-06,
+      "loss": 1.4782,
+      "step": 89110
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.494949494949495e-06,
+      "loss": 1.479,
+      "step": 89120
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.48989898989899e-06,
+      "loss": 1.4779,
+      "step": 89130
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.484848484848485e-06,
+      "loss": 1.4546,
+      "step": 89140
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.47979797979798e-06,
+      "loss": 1.4609,
+      "step": 89150
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.474747474747475e-06,
+      "loss": 1.4524,
+      "step": 89160
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.46969696969697e-06,
+      "loss": 1.4586,
+      "step": 89170
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.464646464646465e-06,
+      "loss": 1.4896,
+      "step": 89180
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.459595959595959e-06,
+      "loss": 1.4756,
+      "step": 89190
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.4545454545454545e-06,
+      "loss": 1.4721,
+      "step": 89200
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.44949494949495e-06,
+      "loss": 1.4816,
+      "step": 89210
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.444444444444445e-06,
+      "loss": 1.452,
+      "step": 89220
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.43939393939394e-06,
+      "loss": 1.4685,
+      "step": 89230
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.434343434343435e-06,
+      "loss": 1.4732,
+      "step": 89240
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.42929292929293e-06,
+      "loss": 1.483,
+      "step": 89250
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.4242424242424245e-06,
+      "loss": 1.4632,
+      "step": 89260
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.41919191919192e-06,
+      "loss": 1.4746,
+      "step": 89270
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.414141414141415e-06,
+      "loss": 1.4675,
+      "step": 89280
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.40909090909091e-06,
+      "loss": 1.4686,
+      "step": 89290
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.404040404040404e-06,
+      "loss": 1.4589,
+      "step": 89300
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.398989898989899e-06,
+      "loss": 1.4672,
+      "step": 89310
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.3939393939393945e-06,
+      "loss": 1.4685,
+      "step": 89320
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.388888888888889e-06,
+      "loss": 1.4571,
+      "step": 89330
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.383838383838384e-06,
+      "loss": 1.4759,
+      "step": 89340
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.378787878787879e-06,
+      "loss": 1.4766,
+      "step": 89350
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.373737373737374e-06,
+      "loss": 1.4576,
+      "step": 89360
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.3686868686868685e-06,
+      "loss": 1.4552,
+      "step": 89370
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.363636363636364e-06,
+      "loss": 1.4736,
+      "step": 89380
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.358585858585859e-06,
+      "loss": 1.4759,
+      "step": 89390
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.353535353535354e-06,
+      "loss": 1.4761,
+      "step": 89400
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.348484848484848e-06,
+      "loss": 1.4417,
+      "step": 89410
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.343434343434343e-06,
+      "loss": 1.4819,
+      "step": 89420
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.3383838383838385e-06,
+      "loss": 1.4597,
+      "step": 89430
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.333333333333334e-06,
+      "loss": 1.4587,
+      "step": 89440
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.328282828282829e-06,
+      "loss": 1.469,
+      "step": 89450
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.323232323232324e-06,
+      "loss": 1.4605,
+      "step": 89460
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.318181818181819e-06,
+      "loss": 1.4773,
+      "step": 89470
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.313131313131313e-06,
+      "loss": 1.4725,
+      "step": 89480
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.3080808080808085e-06,
+      "loss": 1.4804,
+      "step": 89490
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.303030303030304e-06,
+      "loss": 1.4877,
+      "step": 89500
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.297979797979798e-06,
+      "loss": 1.4689,
+      "step": 89510
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.292929292929293e-06,
+      "loss": 1.4726,
+      "step": 89520
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.287878787878788e-06,
+      "loss": 1.4739,
+      "step": 89530
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.282828282828283e-06,
+      "loss": 1.4639,
+      "step": 89540
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.277777777777778e-06,
+      "loss": 1.4557,
+      "step": 89550
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.272727272727273e-06,
+      "loss": 1.4581,
+      "step": 89560
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.267676767676768e-06,
+      "loss": 1.4595,
+      "step": 89570
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.262626262626263e-06,
+      "loss": 1.4702,
+      "step": 89580
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.257575757575757e-06,
+      "loss": 1.488,
+      "step": 89590
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.2525252525252526e-06,
+      "loss": 1.4491,
+      "step": 89600
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.247474747474748e-06,
+      "loss": 1.4748,
+      "step": 89610
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.242424242424243e-06,
+      "loss": 1.4739,
+      "step": 89620
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.237373737373737e-06,
+      "loss": 1.4594,
+      "step": 89630
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.232323232323232e-06,
+      "loss": 1.4654,
+      "step": 89640
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.2272727272727274e-06,
+      "loss": 1.4676,
+      "step": 89650
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.2222222222222226e-06,
+      "loss": 1.4693,
+      "step": 89660
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.217171717171718e-06,
+      "loss": 1.4715,
+      "step": 89670
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.212121212121213e-06,
+      "loss": 1.4744,
+      "step": 89680
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.207070707070707e-06,
+      "loss": 1.4637,
+      "step": 89690
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.202020202020202e-06,
+      "loss": 1.4647,
+      "step": 89700
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.1969696969696974e-06,
+      "loss": 1.4778,
+      "step": 89710
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.191919191919193e-06,
+      "loss": 1.4573,
+      "step": 89720
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.186868686868687e-06,
+      "loss": 1.4736,
+      "step": 89730
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.181818181818182e-06,
+      "loss": 1.4905,
+      "step": 89740
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.176767676767677e-06,
+      "loss": 1.472,
+      "step": 89750
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.171717171717172e-06,
+      "loss": 1.4728,
+      "step": 89760
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.166666666666667e-06,
+      "loss": 1.4583,
+      "step": 89770
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.161616161616162e-06,
+      "loss": 1.4812,
+      "step": 89780
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.156565656565657e-06,
+      "loss": 1.479,
+      "step": 89790
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.151515151515152e-06,
+      "loss": 1.4705,
+      "step": 89800
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.146464646464646e-06,
+      "loss": 1.4713,
+      "step": 89810
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.1414141414141415e-06,
+      "loss": 1.4759,
+      "step": 89820
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.136363636363637e-06,
+      "loss": 1.4687,
+      "step": 89830
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.131313131313131e-06,
+      "loss": 1.4487,
+      "step": 89840
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.126262626262626e-06,
+      "loss": 1.4765,
+      "step": 89850
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.121212121212121e-06,
+      "loss": 1.4557,
+      "step": 89860
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.116161616161616e-06,
+      "loss": 1.464,
+      "step": 89870
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.1111111111111115e-06,
+      "loss": 1.444,
+      "step": 89880
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.106060606060607e-06,
+      "loss": 1.4552,
+      "step": 89890
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.101010101010102e-06,
+      "loss": 1.479,
+      "step": 89900
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.095959595959596e-06,
+      "loss": 1.4703,
+      "step": 89910
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.090909090909091e-06,
+      "loss": 1.4631,
+      "step": 89920
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.085858585858586e-06,
+      "loss": 1.4569,
+      "step": 89930
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0808080808080815e-06,
+      "loss": 1.4731,
+      "step": 89940
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.075757575757576e-06,
+      "loss": 1.4651,
+      "step": 89950
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.070707070707071e-06,
+      "loss": 1.4802,
+      "step": 89960
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.065656565656566e-06,
+      "loss": 1.4635,
+      "step": 89970
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.060606060606061e-06,
+      "loss": 1.4531,
+      "step": 89980
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0555555555555555e-06,
+      "loss": 1.4679,
+      "step": 89990
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.050505050505051e-06,
+      "loss": 1.4568,
+      "step": 90000
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.045454545454546e-06,
+      "loss": 1.4778,
+      "step": 90010
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.04040404040404e-06,
+      "loss": 1.4658,
+      "step": 90020
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.035353535353535e-06,
+      "loss": 1.4856,
+      "step": 90030
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.03030303030303e-06,
+      "loss": 1.4687,
+      "step": 90040
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.0252525252525255e-06,
+      "loss": 1.4502,
+      "step": 90050
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.02020202020202e-06,
+      "loss": 1.4567,
+      "step": 90060
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.015151515151515e-06,
+      "loss": 1.4755,
+      "step": 90070
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.010101010101011e-06,
+      "loss": 1.4689,
+      "step": 90080
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.005050505050505e-06,
+      "loss": 1.4665,
+      "step": 90090
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5e-06,
+      "loss": 1.4606,
+      "step": 90100
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.9949494949494956e-06,
+      "loss": 1.4803,
+      "step": 90110
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.989898989898991e-06,
+      "loss": 1.4706,
+      "step": 90120
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 4.984848484848485e-06,
+      "loss": 1.4751,
+      "step": 90130
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.97979797979798e-06,
+      "loss": 1.4831,
+      "step": 90140
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.974747474747475e-06,
+      "loss": 1.4674,
+      "step": 90150
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.96969696969697e-06,
+      "loss": 1.4681,
+      "step": 90160
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.964646464646465e-06,
+      "loss": 1.4642,
+      "step": 90170
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.95959595959596e-06,
+      "loss": 1.4755,
+      "step": 90180
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.954545454545455e-06,
+      "loss": 1.4794,
+      "step": 90190
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.949494949494949e-06,
+      "loss": 1.4632,
+      "step": 90200
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9444444444444444e-06,
+      "loss": 1.472,
+      "step": 90210
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.93939393939394e-06,
+      "loss": 1.4586,
+      "step": 90220
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.934343434343435e-06,
+      "loss": 1.4595,
+      "step": 90230
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.929292929292929e-06,
+      "loss": 1.4732,
+      "step": 90240
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.924242424242424e-06,
+      "loss": 1.4638,
+      "step": 90250
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.919191919191919e-06,
+      "loss": 1.4871,
+      "step": 90260
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.9141414141414145e-06,
+      "loss": 1.465,
+      "step": 90270
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.909090909090909e-06,
+      "loss": 1.4546,
+      "step": 90280
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.904040404040404e-06,
+      "loss": 1.4819,
+      "step": 90290
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8989898989899e-06,
+      "loss": 1.4758,
+      "step": 90300
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.893939393939394e-06,
+      "loss": 1.4769,
+      "step": 90310
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.888888888888889e-06,
+      "loss": 1.4748,
+      "step": 90320
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8838383838383845e-06,
+      "loss": 1.4768,
+      "step": 90330
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.87878787878788e-06,
+      "loss": 1.4481,
+      "step": 90340
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.873737373737374e-06,
+      "loss": 1.476,
+      "step": 90350
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.868686868686869e-06,
+      "loss": 1.4607,
+      "step": 90360
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.863636363636364e-06,
+      "loss": 1.4951,
+      "step": 90370
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8585858585858585e-06,
+      "loss": 1.4733,
+      "step": 90380
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.853535353535354e-06,
+      "loss": 1.4665,
+      "step": 90390
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.848484848484849e-06,
+      "loss": 1.4698,
+      "step": 90400
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.843434343434344e-06,
+      "loss": 1.4649,
+      "step": 90410
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.838383838383838e-06,
+      "loss": 1.4709,
+      "step": 90420
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.833333333333333e-06,
+      "loss": 1.4896,
+      "step": 90430
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.8282828282828285e-06,
+      "loss": 1.4849,
+      "step": 90440
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.823232323232324e-06,
+      "loss": 1.4645,
+      "step": 90450
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.818181818181818e-06,
+      "loss": 1.4487,
+      "step": 90460
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.813131313131313e-06,
+      "loss": 1.4547,
+      "step": 90470
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.808080808080808e-06,
+      "loss": 1.4579,
+      "step": 90480
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.803030303030303e-06,
+      "loss": 1.4629,
+      "step": 90490
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.797979797979798e-06,
+      "loss": 1.486,
+      "step": 90500
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.792929292929294e-06,
+      "loss": 1.47,
+      "step": 90510
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.787878787878789e-06,
+      "loss": 1.4534,
+      "step": 90520
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.782828282828283e-06,
+      "loss": 1.4782,
+      "step": 90530
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.777777777777778e-06,
+      "loss": 1.474,
+      "step": 90540
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.772727272727273e-06,
+      "loss": 1.481,
+      "step": 90550
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.767676767676768e-06,
+      "loss": 1.4641,
+      "step": 90560
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.762626262626263e-06,
+      "loss": 1.4696,
+      "step": 90570
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.757575757575758e-06,
+      "loss": 1.4596,
+      "step": 90580
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.752525252525253e-06,
+      "loss": 1.4716,
+      "step": 90590
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.747474747474747e-06,
+      "loss": 1.4771,
+      "step": 90600
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.7424242424242426e-06,
+      "loss": 1.4673,
+      "step": 90610
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.737373737373738e-06,
+      "loss": 1.4649,
+      "step": 90620
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.732323232323233e-06,
+      "loss": 1.4668,
+      "step": 90630
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.727272727272727e-06,
+      "loss": 1.4553,
+      "step": 90640
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.722222222222222e-06,
+      "loss": 1.4719,
+      "step": 90650
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.717171717171717e-06,
+      "loss": 1.4721,
+      "step": 90660
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.7121212121212126e-06,
+      "loss": 1.4807,
+      "step": 90670
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.707070707070707e-06,
+      "loss": 1.471,
+      "step": 90680
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.702020202020202e-06,
+      "loss": 1.466,
+      "step": 90690
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.696969696969697e-06,
+      "loss": 1.4616,
+      "step": 90700
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.6919191919191914e-06,
+      "loss": 1.4512,
+      "step": 90710
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.6868686868686874e-06,
+      "loss": 1.4454,
+      "step": 90720
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.681818181818183e-06,
+      "loss": 1.4521,
+      "step": 90730
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.676767676767677e-06,
+      "loss": 1.4724,
+      "step": 90740
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.671717171717172e-06,
+      "loss": 1.4617,
+      "step": 90750
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.666666666666667e-06,
+      "loss": 1.4705,
+      "step": 90760
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.661616161616162e-06,
+      "loss": 1.464,
+      "step": 90770
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.656565656565657e-06,
+      "loss": 1.4528,
+      "step": 90780
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.651515151515152e-06,
+      "loss": 1.4726,
+      "step": 90790
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.646464646464647e-06,
+      "loss": 1.4808,
+      "step": 90800
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.641414141414142e-06,
+      "loss": 1.475,
+      "step": 90810
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.636363636363636e-06,
+      "loss": 1.46,
+      "step": 90820
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.6313131313131315e-06,
+      "loss": 1.456,
+      "step": 90830
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.626262626262627e-06,
+      "loss": 1.4773,
+      "step": 90840
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.621212121212122e-06,
+      "loss": 1.4589,
+      "step": 90850
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.616161616161616e-06,
+      "loss": 1.461,
+      "step": 90860
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.611111111111111e-06,
+      "loss": 1.4801,
+      "step": 90870
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.606060606060606e-06,
+      "loss": 1.4714,
+      "step": 90880
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.601010101010101e-06,
+      "loss": 1.4641,
+      "step": 90890
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.595959595959596e-06,
+      "loss": 1.4744,
+      "step": 90900
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.590909090909091e-06,
+      "loss": 1.4874,
+      "step": 90910
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.585858585858586e-06,
+      "loss": 1.4939,
+      "step": 90920
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.58080808080808e-06,
+      "loss": 1.4759,
+      "step": 90930
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.575757575757576e-06,
+      "loss": 1.4752,
+      "step": 90940
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.5707070707070715e-06,
+      "loss": 1.468,
+      "step": 90950
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.565656565656566e-06,
+      "loss": 1.4762,
+      "step": 90960
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.560606060606061e-06,
+      "loss": 1.4704,
+      "step": 90970
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.555555555555556e-06,
+      "loss": 1.4505,
+      "step": 90980
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.550505050505051e-06,
+      "loss": 1.4622,
+      "step": 90990
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.5454545454545455e-06,
+      "loss": 1.4585,
+      "step": 91000
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.540404040404041e-06,
+      "loss": 1.4814,
+      "step": 91010
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.535353535353536e-06,
+      "loss": 1.4637,
+      "step": 91020
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.530303030303031e-06,
+      "loss": 1.4791,
+      "step": 91030
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.525252525252525e-06,
+      "loss": 1.4613,
+      "step": 91040
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.52020202020202e-06,
+      "loss": 1.449,
+      "step": 91050
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.5151515151515155e-06,
+      "loss": 1.4617,
+      "step": 91060
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.51010101010101e-06,
+      "loss": 1.4755,
+      "step": 91070
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.505050505050505e-06,
+      "loss": 1.4894,
+      "step": 91080
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.5e-06,
+      "loss": 1.4719,
+      "step": 91090
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.494949494949495e-06,
+      "loss": 1.4535,
+      "step": 91100
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.4898989898989896e-06,
+      "loss": 1.4587,
+      "step": 91110
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.484848484848485e-06,
+      "loss": 1.4767,
+      "step": 91120
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.47979797979798e-06,
+      "loss": 1.4792,
+      "step": 91130
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.474747474747475e-06,
+      "loss": 1.4767,
+      "step": 91140
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.46969696969697e-06,
+      "loss": 1.4813,
+      "step": 91150
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.464646464646465e-06,
+      "loss": 1.473,
+      "step": 91160
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.45959595959596e-06,
+      "loss": 1.4733,
+      "step": 91170
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.454545454545455e-06,
+      "loss": 1.4479,
+      "step": 91180
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.44949494949495e-06,
+      "loss": 1.4588,
+      "step": 91190
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.444444444444445e-06,
+      "loss": 1.4512,
+      "step": 91200
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.43939393939394e-06,
+      "loss": 1.4868,
+      "step": 91210
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.4343434343434344e-06,
+      "loss": 1.4865,
+      "step": 91220
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.42929292929293e-06,
+      "loss": 1.4883,
+      "step": 91230
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.424242424242425e-06,
+      "loss": 1.4697,
+      "step": 91240
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.419191919191919e-06,
+      "loss": 1.4755,
+      "step": 91250
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.414141414141414e-06,
+      "loss": 1.473,
+      "step": 91260
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.409090909090909e-06,
+      "loss": 1.4725,
+      "step": 91270
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.4040404040404044e-06,
+      "loss": 1.4763,
+      "step": 91280
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.398989898989899e-06,
+      "loss": 1.4818,
+      "step": 91290
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.393939393939394e-06,
+      "loss": 1.467,
+      "step": 91300
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.388888888888889e-06,
+      "loss": 1.4666,
+      "step": 91310
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.383838383838384e-06,
+      "loss": 1.4732,
+      "step": 91320
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.3787878787878785e-06,
+      "loss": 1.4482,
+      "step": 91330
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.373737373737374e-06,
+      "loss": 1.4701,
+      "step": 91340
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.368686868686869e-06,
+      "loss": 1.4808,
+      "step": 91350
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.363636363636364e-06,
+      "loss": 1.4825,
+      "step": 91360
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.358585858585859e-06,
+      "loss": 1.4605,
+      "step": 91370
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.353535353535354e-06,
+      "loss": 1.4462,
+      "step": 91380
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.348484848484849e-06,
+      "loss": 1.4721,
+      "step": 91390
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.343434343434344e-06,
+      "loss": 1.4625,
+      "step": 91400
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.338383838383839e-06,
+      "loss": 1.4534,
+      "step": 91410
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.333333333333334e-06,
+      "loss": 1.4776,
+      "step": 91420
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.328282828282828e-06,
+      "loss": 1.4715,
+      "step": 91430
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.323232323232323e-06,
+      "loss": 1.4657,
+      "step": 91440
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.3181818181818185e-06,
+      "loss": 1.4836,
+      "step": 91450
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 4.313131313131314e-06,
+      "loss": 1.4783,
+      "step": 91460
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.308080808080808e-06,
+      "loss": 1.4855,
+      "step": 91470
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.303030303030303e-06,
+      "loss": 1.4586,
+      "step": 91480
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.297979797979798e-06,
+      "loss": 1.4808,
+      "step": 91490
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.292929292929293e-06,
+      "loss": 1.4592,
+      "step": 91500
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.287878787878788e-06,
+      "loss": 1.475,
+      "step": 91510
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.282828282828283e-06,
+      "loss": 1.4748,
+      "step": 91520
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.277777777777778e-06,
+      "loss": 1.4441,
+      "step": 91530
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.272727272727273e-06,
+      "loss": 1.4638,
+      "step": 91540
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.267676767676767e-06,
+      "loss": 1.4876,
+      "step": 91550
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.2626262626262625e-06,
+      "loss": 1.4608,
+      "step": 91560
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.257575757575758e-06,
+      "loss": 1.4535,
+      "step": 91570
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.252525252525253e-06,
+      "loss": 1.4693,
+      "step": 91580
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.247474747474748e-06,
+      "loss": 1.4549,
+      "step": 91590
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.242424242424243e-06,
+      "loss": 1.4816,
+      "step": 91600
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.237373737373737e-06,
+      "loss": 1.445,
+      "step": 91610
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.2323232323232325e-06,
+      "loss": 1.4646,
+      "step": 91620
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.227272727272728e-06,
+      "loss": 1.4556,
+      "step": 91630
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.222222222222223e-06,
+      "loss": 1.444,
+      "step": 91640
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.217171717171717e-06,
+      "loss": 1.4568,
+      "step": 91650
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.212121212121212e-06,
+      "loss": 1.4788,
+      "step": 91660
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.207070707070707e-06,
+      "loss": 1.4818,
+      "step": 91670
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.2020202020202026e-06,
+      "loss": 1.4492,
+      "step": 91680
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.196969696969697e-06,
+      "loss": 1.4698,
+      "step": 91690
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.191919191919192e-06,
+      "loss": 1.4625,
+      "step": 91700
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.186868686868687e-06,
+      "loss": 1.4715,
+      "step": 91710
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.181818181818182e-06,
+      "loss": 1.4642,
+      "step": 91720
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.176767676767677e-06,
+      "loss": 1.4664,
+      "step": 91730
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.171717171717172e-06,
+      "loss": 1.4533,
+      "step": 91740
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.166666666666667e-06,
+      "loss": 1.4596,
+      "step": 91750
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.161616161616161e-06,
+      "loss": 1.4621,
+      "step": 91760
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.156565656565656e-06,
+      "loss": 1.4803,
+      "step": 91770
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.1515151515151514e-06,
+      "loss": 1.4739,
+      "step": 91780
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.146464646464647e-06,
+      "loss": 1.4839,
+      "step": 91790
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.141414141414142e-06,
+      "loss": 1.4779,
+      "step": 91800
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.136363636363637e-06,
+      "loss": 1.4492,
+      "step": 91810
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.131313131313132e-06,
+      "loss": 1.4718,
+      "step": 91820
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.126262626262626e-06,
+      "loss": 1.4597,
+      "step": 91830
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.1212121212121215e-06,
+      "loss": 1.4777,
+      "step": 91840
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.116161616161617e-06,
+      "loss": 1.4685,
+      "step": 91850
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.111111111111112e-06,
+      "loss": 1.4735,
+      "step": 91860
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.106060606060606e-06,
+      "loss": 1.4763,
+      "step": 91870
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.101010101010101e-06,
+      "loss": 1.4751,
+      "step": 91880
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.095959595959596e-06,
+      "loss": 1.479,
+      "step": 91890
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.0909090909090915e-06,
+      "loss": 1.4805,
+      "step": 91900
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.085858585858586e-06,
+      "loss": 1.4722,
+      "step": 91910
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.080808080808081e-06,
+      "loss": 1.4712,
+      "step": 91920
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.075757575757576e-06,
+      "loss": 1.4535,
+      "step": 91930
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.07070707070707e-06,
+      "loss": 1.4787,
+      "step": 91940
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.0656565656565655e-06,
+      "loss": 1.4398,
+      "step": 91950
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.060606060606061e-06,
+      "loss": 1.47,
+      "step": 91960
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.055555555555556e-06,
+      "loss": 1.4752,
+      "step": 91970
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.05050505050505e-06,
+      "loss": 1.4709,
+      "step": 91980
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.045454545454545e-06,
+      "loss": 1.4594,
+      "step": 91990
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.040404040404041e-06,
+      "loss": 1.476,
+      "step": 92000
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.0353535353535355e-06,
+      "loss": 1.4715,
+      "step": 92010
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.030303030303031e-06,
+      "loss": 1.4705,
+      "step": 92020
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.025252525252526e-06,
+      "loss": 1.4614,
+      "step": 92030
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.020202020202021e-06,
+      "loss": 1.4659,
+      "step": 92040
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.015151515151515e-06,
+      "loss": 1.4633,
+      "step": 92050
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.01010101010101e-06,
+      "loss": 1.4655,
+      "step": 92060
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.0050505050505055e-06,
+      "loss": 1.4666,
+      "step": 92070
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 1.4769,
+      "step": 92080
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.994949494949495e-06,
+      "loss": 1.4623,
+      "step": 92090
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.98989898989899e-06,
+      "loss": 1.4679,
+      "step": 92100
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.984848484848485e-06,
+      "loss": 1.4643,
+      "step": 92110
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.9797979797979795e-06,
+      "loss": 1.4664,
+      "step": 92120
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 3.974747474747475e-06,
+      "loss": 1.4821,
+      "step": 92130
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.96969696969697e-06,
+      "loss": 1.4722,
+      "step": 92140
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.964646464646465e-06,
+      "loss": 1.4656,
+      "step": 92150
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.959595959595959e-06,
+      "loss": 1.4778,
+      "step": 92160
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.954545454545454e-06,
+      "loss": 1.4686,
+      "step": 92170
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.9494949494949496e-06,
+      "loss": 1.4591,
+      "step": 92180
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.944444444444445e-06,
+      "loss": 1.4649,
+      "step": 92190
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.939393939393939e-06,
+      "loss": 1.4568,
+      "step": 92200
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.934343434343434e-06,
+      "loss": 1.4735,
+      "step": 92210
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.92929292929293e-06,
+      "loss": 1.482,
+      "step": 92220
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.9242424242424244e-06,
+      "loss": 1.4699,
+      "step": 92230
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.9191919191919196e-06,
+      "loss": 1.481,
+      "step": 92240
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.914141414141415e-06,
+      "loss": 1.4855,
+      "step": 92250
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.90909090909091e-06,
+      "loss": 1.4802,
+      "step": 92260
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.904040404040404e-06,
+      "loss": 1.4943,
+      "step": 92270
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.898989898989899e-06,
+      "loss": 1.4653,
+      "step": 92280
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.8939393939393944e-06,
+      "loss": 1.4682,
+      "step": 92290
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.888888888888889e-06,
+      "loss": 1.4621,
+      "step": 92300
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.883838383838384e-06,
+      "loss": 1.4672,
+      "step": 92310
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.878787878787879e-06,
+      "loss": 1.4527,
+      "step": 92320
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.873737373737374e-06,
+      "loss": 1.4521,
+      "step": 92330
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.8686868686868685e-06,
+      "loss": 1.4826,
+      "step": 92340
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.863636363636364e-06,
+      "loss": 1.4689,
+      "step": 92350
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.858585858585859e-06,
+      "loss": 1.4705,
+      "step": 92360
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.853535353535354e-06,
+      "loss": 1.4513,
+      "step": 92370
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.848484848484848e-06,
+      "loss": 1.4592,
+      "step": 92380
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.843434343434343e-06,
+      "loss": 1.454,
+      "step": 92390
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.8383838383838385e-06,
+      "loss": 1.4823,
+      "step": 92400
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.833333333333334e-06,
+      "loss": 1.4493,
+      "step": 92410
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.828282828282828e-06,
+      "loss": 1.4686,
+      "step": 92420
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.823232323232324e-06,
+      "loss": 1.4638,
+      "step": 92430
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.818181818181819e-06,
+      "loss": 1.471,
+      "step": 92440
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.8131313131313138e-06,
+      "loss": 1.4729,
+      "step": 92450
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.8080808080808085e-06,
+      "loss": 1.4821,
+      "step": 92460
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.803030303030303e-06,
+      "loss": 1.4635,
+      "step": 92470
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.7979797979797984e-06,
+      "loss": 1.4713,
+      "step": 92480
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.792929292929293e-06,
+      "loss": 1.4721,
+      "step": 92490
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.7878787878787882e-06,
+      "loss": 1.436,
+      "step": 92500
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.782828282828283e-06,
+      "loss": 1.4797,
+      "step": 92510
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.777777777777778e-06,
+      "loss": 1.4514,
+      "step": 92520
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.772727272727273e-06,
+      "loss": 1.4693,
+      "step": 92530
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.767676767676768e-06,
+      "loss": 1.4369,
+      "step": 92540
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.7626262626262627e-06,
+      "loss": 1.4588,
+      "step": 92550
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.757575757575758e-06,
+      "loss": 1.4546,
+      "step": 92560
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.7525252525252525e-06,
+      "loss": 1.4586,
+      "step": 92570
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.7474747474747477e-06,
+      "loss": 1.4752,
+      "step": 92580
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.7424242424242424e-06,
+      "loss": 1.467,
+      "step": 92590
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.7373737373737375e-06,
+      "loss": 1.4611,
+      "step": 92600
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.7323232323232322e-06,
+      "loss": 1.4502,
+      "step": 92610
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.727272727272727e-06,
+      "loss": 1.4543,
+      "step": 92620
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.722222222222222e-06,
+      "loss": 1.468,
+      "step": 92630
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.7171717171717177e-06,
+      "loss": 1.487,
+      "step": 92640
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.7121212121212124e-06,
+      "loss": 1.4446,
+      "step": 92650
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.7070707070707075e-06,
+      "loss": 1.4634,
+      "step": 92660
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.7020202020202023e-06,
+      "loss": 1.4802,
+      "step": 92670
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.6969696969696974e-06,
+      "loss": 1.4685,
+      "step": 92680
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.691919191919192e-06,
+      "loss": 1.4766,
+      "step": 92690
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.6868686868686873e-06,
+      "loss": 1.471,
+      "step": 92700
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.681818181818182e-06,
+      "loss": 1.4677,
+      "step": 92710
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.676767676767677e-06,
+      "loss": 1.4614,
+      "step": 92720
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.671717171717172e-06,
+      "loss": 1.4538,
+      "step": 92730
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.666666666666667e-06,
+      "loss": 1.4627,
+      "step": 92740
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.6616161616161617e-06,
+      "loss": 1.437,
+      "step": 92750
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.656565656565657e-06,
+      "loss": 1.4655,
+      "step": 92760
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.6515151515151516e-06,
+      "loss": 1.4453,
+      "step": 92770
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.6464646464646467e-06,
+      "loss": 1.4448,
+      "step": 92780
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 3.6414141414141414e-06,
+      "loss": 1.4795,
+      "step": 92790
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.636363636363636e-06,
+      "loss": 1.4718,
+      "step": 92800
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.6313131313131313e-06,
+      "loss": 1.4671,
+      "step": 92810
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.626262626262626e-06,
+      "loss": 1.4649,
+      "step": 92820
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.621212121212121e-06,
+      "loss": 1.4665,
+      "step": 92830
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.616161616161616e-06,
+      "loss": 1.4887,
+      "step": 92840
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.611111111111111e-06,
+      "loss": 1.4386,
+      "step": 92850
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.6060606060606066e-06,
+      "loss": 1.4805,
+      "step": 92860
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.6010101010101013e-06,
+      "loss": 1.4315,
+      "step": 92870
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.5959595959595965e-06,
+      "loss": 1.4721,
+      "step": 92880
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.590909090909091e-06,
+      "loss": 1.4486,
+      "step": 92890
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.5858585858585863e-06,
+      "loss": 1.4816,
+      "step": 92900
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.580808080808081e-06,
+      "loss": 1.4748,
+      "step": 92910
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.575757575757576e-06,
+      "loss": 1.4718,
+      "step": 92920
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.570707070707071e-06,
+      "loss": 1.4472,
+      "step": 92930
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.565656565656566e-06,
+      "loss": 1.4624,
+      "step": 92940
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.5606060606060608e-06,
+      "loss": 1.4732,
+      "step": 92950
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.555555555555556e-06,
+      "loss": 1.4675,
+      "step": 92960
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.5505050505050506e-06,
+      "loss": 1.4491,
+      "step": 92970
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.5454545454545454e-06,
+      "loss": 1.4384,
+      "step": 92980
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.5404040404040405e-06,
+      "loss": 1.4581,
+      "step": 92990
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.5353535353535352e-06,
+      "loss": 1.4655,
+      "step": 93000
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.5303030303030304e-06,
+      "loss": 1.4587,
+      "step": 93010
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.525252525252525e-06,
+      "loss": 1.4869,
+      "step": 93020
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.5202020202020202e-06,
+      "loss": 1.4678,
+      "step": 93030
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.515151515151515e-06,
+      "loss": 1.4592,
+      "step": 93040
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.51010101010101e-06,
+      "loss": 1.4559,
+      "step": 93050
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.505050505050505e-06,
+      "loss": 1.471,
+      "step": 93060
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.5000000000000004e-06,
+      "loss": 1.4678,
+      "step": 93070
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.4949494949494955e-06,
+      "loss": 1.4769,
+      "step": 93080
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.4898989898989902e-06,
+      "loss": 1.4545,
+      "step": 93090
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.4848484848484854e-06,
+      "loss": 1.4592,
+      "step": 93100
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.47979797979798e-06,
+      "loss": 1.4783,
+      "step": 93110
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.4747474747474752e-06,
+      "loss": 1.4825,
+      "step": 93120
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.46969696969697e-06,
+      "loss": 1.4647,
+      "step": 93130
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.464646464646465e-06,
+      "loss": 1.4609,
+      "step": 93140
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.45959595959596e-06,
+      "loss": 1.4467,
+      "step": 93150
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.4545454545454545e-06,
+      "loss": 1.4799,
+      "step": 93160
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.4494949494949497e-06,
+      "loss": 1.487,
+      "step": 93170
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.4444444444444444e-06,
+      "loss": 1.468,
+      "step": 93180
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.4393939393939395e-06,
+      "loss": 1.4586,
+      "step": 93190
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.4343434343434343e-06,
+      "loss": 1.481,
+      "step": 93200
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.4292929292929294e-06,
+      "loss": 1.4721,
+      "step": 93210
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.424242424242424e-06,
+      "loss": 1.4766,
+      "step": 93220
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.4191919191919193e-06,
+      "loss": 1.4437,
+      "step": 93230
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.414141414141414e-06,
+      "loss": 1.4583,
+      "step": 93240
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.409090909090909e-06,
+      "loss": 1.4704,
+      "step": 93250
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.404040404040404e-06,
+      "loss": 1.4628,
+      "step": 93260
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.398989898989899e-06,
+      "loss": 1.4757,
+      "step": 93270
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.3939393939393946e-06,
+      "loss": 1.4657,
+      "step": 93280
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.3888888888888893e-06,
+      "loss": 1.4539,
+      "step": 93290
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.3838383838383844e-06,
+      "loss": 1.4699,
+      "step": 93300
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.378787878787879e-06,
+      "loss": 1.4578,
+      "step": 93310
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.3737373737373743e-06,
+      "loss": 1.459,
+      "step": 93320
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.368686868686869e-06,
+      "loss": 1.4694,
+      "step": 93330
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.363636363636364e-06,
+      "loss": 1.4637,
+      "step": 93340
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.358585858585859e-06,
+      "loss": 1.4664,
+      "step": 93350
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.3535353535353536e-06,
+      "loss": 1.4711,
+      "step": 93360
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.3484848484848487e-06,
+      "loss": 1.4697,
+      "step": 93370
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.3434343434343435e-06,
+      "loss": 1.4606,
+      "step": 93380
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.3383838383838386e-06,
+      "loss": 1.4839,
+      "step": 93390
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.3333333333333333e-06,
+      "loss": 1.4545,
+      "step": 93400
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.3282828282828285e-06,
+      "loss": 1.4646,
+      "step": 93410
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.323232323232323e-06,
+      "loss": 1.4515,
+      "step": 93420
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.3181818181818183e-06,
+      "loss": 1.4339,
+      "step": 93430
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.313131313131313e-06,
+      "loss": 1.4764,
+      "step": 93440
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.308080808080808e-06,
+      "loss": 1.4603,
+      "step": 93450
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 3.303030303030303e-06,
+      "loss": 1.458,
+      "step": 93460
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.297979797979798e-06,
+      "loss": 1.46,
+      "step": 93470
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.2929292929292928e-06,
+      "loss": 1.4738,
+      "step": 93480
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.2878787878787875e-06,
+      "loss": 1.4537,
+      "step": 93490
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.2828282828282835e-06,
+      "loss": 1.4551,
+      "step": 93500
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.277777777777778e-06,
+      "loss": 1.4687,
+      "step": 93510
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.2727272727272733e-06,
+      "loss": 1.4665,
+      "step": 93520
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.267676767676768e-06,
+      "loss": 1.4575,
+      "step": 93530
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.262626262626263e-06,
+      "loss": 1.4762,
+      "step": 93540
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.257575757575758e-06,
+      "loss": 1.4511,
+      "step": 93550
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.2525252525252527e-06,
+      "loss": 1.441,
+      "step": 93560
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.247474747474748e-06,
+      "loss": 1.4609,
+      "step": 93570
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.2424242424242425e-06,
+      "loss": 1.4781,
+      "step": 93580
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.2373737373737377e-06,
+      "loss": 1.4899,
+      "step": 93590
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.2323232323232324e-06,
+      "loss": 1.488,
+      "step": 93600
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.2272727272727275e-06,
+      "loss": 1.4673,
+      "step": 93610
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.2222222222222222e-06,
+      "loss": 1.4642,
+      "step": 93620
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.2171717171717174e-06,
+      "loss": 1.4828,
+      "step": 93630
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.212121212121212e-06,
+      "loss": 1.4749,
+      "step": 93640
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.2070707070707072e-06,
+      "loss": 1.4523,
+      "step": 93650
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.202020202020202e-06,
+      "loss": 1.4855,
+      "step": 93660
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.1969696969696967e-06,
+      "loss": 1.4754,
+      "step": 93670
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.191919191919192e-06,
+      "loss": 1.477,
+      "step": 93680
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.1868686868686866e-06,
+      "loss": 1.4535,
+      "step": 93690
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.1818181818181817e-06,
+      "loss": 1.4678,
+      "step": 93700
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.1767676767676773e-06,
+      "loss": 1.4742,
+      "step": 93710
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.171717171717172e-06,
+      "loss": 1.4561,
+      "step": 93720
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.166666666666667e-06,
+      "loss": 1.4851,
+      "step": 93730
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.161616161616162e-06,
+      "loss": 1.4875,
+      "step": 93740
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.156565656565657e-06,
+      "loss": 1.4786,
+      "step": 93750
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.1515151515151517e-06,
+      "loss": 1.4618,
+      "step": 93760
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.146464646464647e-06,
+      "loss": 1.4634,
+      "step": 93770
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.1414141414141416e-06,
+      "loss": 1.4705,
+      "step": 93780
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.1363636363636367e-06,
+      "loss": 1.4285,
+      "step": 93790
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.1313131313131314e-06,
+      "loss": 1.4704,
+      "step": 93800
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.1262626262626266e-06,
+      "loss": 1.4715,
+      "step": 93810
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.1212121212121213e-06,
+      "loss": 1.4665,
+      "step": 93820
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.1161616161616164e-06,
+      "loss": 1.45,
+      "step": 93830
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.111111111111111e-06,
+      "loss": 1.4848,
+      "step": 93840
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.106060606060606e-06,
+      "loss": 1.4598,
+      "step": 93850
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.101010101010101e-06,
+      "loss": 1.4596,
+      "step": 93860
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.095959595959596e-06,
+      "loss": 1.4572,
+      "step": 93870
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.0909090909090913e-06,
+      "loss": 1.4752,
+      "step": 93880
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.085858585858586e-06,
+      "loss": 1.4581,
+      "step": 93890
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.080808080808081e-06,
+      "loss": 1.4613,
+      "step": 93900
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.075757575757576e-06,
+      "loss": 1.4648,
+      "step": 93910
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.070707070707071e-06,
+      "loss": 1.4647,
+      "step": 93920
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.0656565656565658e-06,
+      "loss": 1.4785,
+      "step": 93930
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.0606060606060605e-06,
+      "loss": 1.4655,
+      "step": 93940
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.0555555555555556e-06,
+      "loss": 1.4736,
+      "step": 93950
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.0505050505050503e-06,
+      "loss": 1.4651,
+      "step": 93960
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.0454545454545455e-06,
+      "loss": 1.4675,
+      "step": 93970
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.0404040404040406e-06,
+      "loss": 1.4665,
+      "step": 93980
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.0353535353535358e-06,
+      "loss": 1.454,
+      "step": 93990
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.0303030303030305e-06,
+      "loss": 1.4739,
+      "step": 94000
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.0252525252525256e-06,
+      "loss": 1.4444,
+      "step": 94010
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.0202020202020203e-06,
+      "loss": 1.4629,
+      "step": 94020
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.0151515151515155e-06,
+      "loss": 1.4543,
+      "step": 94030
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.0101010101010102e-06,
+      "loss": 1.4556,
+      "step": 94040
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3.005050505050505e-06,
+      "loss": 1.4394,
+      "step": 94050
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 3e-06,
+      "loss": 1.471,
+      "step": 94060
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.994949494949495e-06,
+      "loss": 1.4624,
+      "step": 94070
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9898989898989904e-06,
+      "loss": 1.4524,
+      "step": 94080
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.984848484848485e-06,
+      "loss": 1.4794,
+      "step": 94090
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.9797979797979802e-06,
+      "loss": 1.4578,
+      "step": 94100
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.974747474747475e-06,
+      "loss": 1.4899,
+      "step": 94110
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.96969696969697e-06,
+      "loss": 1.4751,
+      "step": 94120
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.964646464646465e-06,
+      "loss": 1.4564,
+      "step": 94130
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9595959595959595e-06,
+      "loss": 1.4578,
+      "step": 94140
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9545454545454547e-06,
+      "loss": 1.4616,
+      "step": 94150
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9494949494949494e-06,
+      "loss": 1.4665,
+      "step": 94160
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9444444444444445e-06,
+      "loss": 1.4525,
+      "step": 94170
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9393939393939393e-06,
+      "loss": 1.4722,
+      "step": 94180
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.934343434343435e-06,
+      "loss": 1.4831,
+      "step": 94190
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9292929292929295e-06,
+      "loss": 1.479,
+      "step": 94200
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9242424242424247e-06,
+      "loss": 1.4584,
+      "step": 94210
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9191919191919194e-06,
+      "loss": 1.4744,
+      "step": 94220
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.914141414141414e-06,
+      "loss": 1.4543,
+      "step": 94230
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.9090909090909093e-06,
+      "loss": 1.4885,
+      "step": 94240
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.904040404040404e-06,
+      "loss": 1.4552,
+      "step": 94250
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.898989898989899e-06,
+      "loss": 1.4714,
+      "step": 94260
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.893939393939394e-06,
+      "loss": 1.4718,
+      "step": 94270
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.888888888888889e-06,
+      "loss": 1.4597,
+      "step": 94280
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8838383838383837e-06,
+      "loss": 1.4515,
+      "step": 94290
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8787878787878793e-06,
+      "loss": 1.4561,
+      "step": 94300
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.873737373737374e-06,
+      "loss": 1.4536,
+      "step": 94310
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8686868686868687e-06,
+      "loss": 1.4864,
+      "step": 94320
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.863636363636364e-06,
+      "loss": 1.4928,
+      "step": 94330
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8585858585858586e-06,
+      "loss": 1.4636,
+      "step": 94340
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8535353535353537e-06,
+      "loss": 1.4945,
+      "step": 94350
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8484848484848484e-06,
+      "loss": 1.4768,
+      "step": 94360
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8434343434343436e-06,
+      "loss": 1.4715,
+      "step": 94370
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8383838383838383e-06,
+      "loss": 1.4652,
+      "step": 94380
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8333333333333335e-06,
+      "loss": 1.4547,
+      "step": 94390
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8282828282828286e-06,
+      "loss": 1.453,
+      "step": 94400
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8232323232323233e-06,
+      "loss": 1.4843,
+      "step": 94410
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8181818181818185e-06,
+      "loss": 1.4797,
+      "step": 94420
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.813131313131313e-06,
+      "loss": 1.4726,
+      "step": 94430
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.8080808080808083e-06,
+      "loss": 1.4772,
+      "step": 94440
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.803030303030303e-06,
+      "loss": 1.4601,
+      "step": 94450
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.797979797979798e-06,
+      "loss": 1.4774,
+      "step": 94460
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.792929292929293e-06,
+      "loss": 1.4611,
+      "step": 94470
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.787878787878788e-06,
+      "loss": 1.4616,
+      "step": 94480
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7828282828282828e-06,
+      "loss": 1.4717,
+      "step": 94490
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.777777777777778e-06,
+      "loss": 1.4526,
+      "step": 94500
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.772727272727273e-06,
+      "loss": 1.474,
+      "step": 94510
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7676767676767678e-06,
+      "loss": 1.4813,
+      "step": 94520
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.762626262626263e-06,
+      "loss": 1.4691,
+      "step": 94530
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7575757575757576e-06,
+      "loss": 1.4697,
+      "step": 94540
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7525252525252528e-06,
+      "loss": 1.453,
+      "step": 94550
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7474747474747475e-06,
+      "loss": 1.4634,
+      "step": 94560
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7424242424242426e-06,
+      "loss": 1.4667,
+      "step": 94570
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7373737373737374e-06,
+      "loss": 1.457,
+      "step": 94580
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7323232323232325e-06,
+      "loss": 1.4724,
+      "step": 94590
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7272727272727272e-06,
+      "loss": 1.4567,
+      "step": 94600
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7222222222222224e-06,
+      "loss": 1.4544,
+      "step": 94610
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7171717171717175e-06,
+      "loss": 1.4669,
+      "step": 94620
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7121212121212122e-06,
+      "loss": 1.4783,
+      "step": 94630
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7070707070707074e-06,
+      "loss": 1.4637,
+      "step": 94640
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.702020202020202e-06,
+      "loss": 1.4607,
+      "step": 94650
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.6969696969696972e-06,
+      "loss": 1.4493,
+      "step": 94660
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.691919191919192e-06,
+      "loss": 1.4676,
+      "step": 94670
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.686868686868687e-06,
+      "loss": 1.4728,
+      "step": 94680
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.681818181818182e-06,
+      "loss": 1.4806,
+      "step": 94690
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.676767676767677e-06,
+      "loss": 1.4719,
+      "step": 94700
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.6717171717171717e-06,
+      "loss": 1.4698,
+      "step": 94710
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.666666666666667e-06,
+      "loss": 1.4709,
+      "step": 94720
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.661616161616162e-06,
+      "loss": 1.4561,
+      "step": 94730
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.6565656565656567e-06,
+      "loss": 1.457,
+      "step": 94740
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.651515151515152e-06,
+      "loss": 1.4668,
+      "step": 94750
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.6464646464646466e-06,
+      "loss": 1.466,
+      "step": 94760
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.6414141414141417e-06,
+      "loss": 1.4574,
+      "step": 94770
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.6363636363636364e-06,
+      "loss": 1.4795,
+      "step": 94780
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.6313131313131316e-06,
+      "loss": 1.4464,
+      "step": 94790
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.6262626262626263e-06,
+      "loss": 1.4583,
+      "step": 94800
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.6212121212121214e-06,
+      "loss": 1.4645,
+      "step": 94810
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.616161616161616e-06,
+      "loss": 1.4885,
+      "step": 94820
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.6111111111111113e-06,
+      "loss": 1.4737,
+      "step": 94830
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.6060606060606064e-06,
+      "loss": 1.4616,
+      "step": 94840
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.601010101010101e-06,
+      "loss": 1.457,
+      "step": 94850
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5959595959595963e-06,
+      "loss": 1.4608,
+      "step": 94860
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.590909090909091e-06,
+      "loss": 1.479,
+      "step": 94870
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.585858585858586e-06,
+      "loss": 1.496,
+      "step": 94880
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.580808080808081e-06,
+      "loss": 1.4642,
+      "step": 94890
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.575757575757576e-06,
+      "loss": 1.4742,
+      "step": 94900
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5707070707070707e-06,
+      "loss": 1.4687,
+      "step": 94910
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5656565656565655e-06,
+      "loss": 1.4573,
+      "step": 94920
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5606060606060606e-06,
+      "loss": 1.4578,
+      "step": 94930
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5555555555555557e-06,
+      "loss": 1.4583,
+      "step": 94940
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.550505050505051e-06,
+      "loss": 1.4712,
+      "step": 94950
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5454545454545456e-06,
+      "loss": 1.4574,
+      "step": 94960
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5404040404040408e-06,
+      "loss": 1.4498,
+      "step": 94970
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5353535353535355e-06,
+      "loss": 1.464,
+      "step": 94980
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5303030303030306e-06,
+      "loss": 1.4633,
+      "step": 94990
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5252525252525253e-06,
+      "loss": 1.4507,
+      "step": 95000
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.52020202020202e-06,
+      "loss": 1.4587,
+      "step": 95010
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.515151515151515e-06,
+      "loss": 1.4597,
+      "step": 95020
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.51010101010101e-06,
+      "loss": 1.4679,
+      "step": 95030
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5050505050505055e-06,
+      "loss": 1.4664,
+      "step": 95040
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.5e-06,
+      "loss": 1.465,
+      "step": 95050
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.4949494949494953e-06,
+      "loss": 1.4541,
+      "step": 95060
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.48989898989899e-06,
+      "loss": 1.4451,
+      "step": 95070
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.484848484848485e-06,
+      "loss": 1.4599,
+      "step": 95080
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.47979797979798e-06,
+      "loss": 1.4801,
+      "step": 95090
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.4747474747474747e-06,
+      "loss": 1.4601,
+      "step": 95100
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.46969696969697e-06,
+      "loss": 1.4595,
+      "step": 95110
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.4646464646464645e-06,
+      "loss": 1.4702,
+      "step": 95120
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.4595959595959597e-06,
+      "loss": 1.4816,
+      "step": 95130
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.4545454545454544e-06,
+      "loss": 1.4585,
+      "step": 95140
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.44949494949495e-06,
+      "loss": 1.4676,
+      "step": 95150
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.4444444444444447e-06,
+      "loss": 1.495,
+      "step": 95160
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.43939393939394e-06,
+      "loss": 1.472,
+      "step": 95170
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.4343434343434345e-06,
+      "loss": 1.4659,
+      "step": 95180
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.4292929292929292e-06,
+      "loss": 1.4572,
+      "step": 95190
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.4242424242424244e-06,
+      "loss": 1.4851,
+      "step": 95200
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.419191919191919e-06,
+      "loss": 1.4753,
+      "step": 95210
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.4141414141414143e-06,
+      "loss": 1.4561,
+      "step": 95220
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.409090909090909e-06,
+      "loss": 1.4658,
+      "step": 95230
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.404040404040404e-06,
+      "loss": 1.4677,
+      "step": 95240
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.398989898989899e-06,
+      "loss": 1.4802,
+      "step": 95250
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.3939393939393944e-06,
+      "loss": 1.4674,
+      "step": 95260
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.388888888888889e-06,
+      "loss": 1.46,
+      "step": 95270
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.383838383838384e-06,
+      "loss": 1.4583,
+      "step": 95280
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.378787878787879e-06,
+      "loss": 1.4673,
+      "step": 95290
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.3737373737373737e-06,
+      "loss": 1.4686,
+      "step": 95300
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.368686868686869e-06,
+      "loss": 1.4528,
+      "step": 95310
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.3636363636363636e-06,
+      "loss": 1.4662,
+      "step": 95320
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.3585858585858587e-06,
+      "loss": 1.4735,
+      "step": 95330
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.3535353535353534e-06,
+      "loss": 1.4509,
+      "step": 95340
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.3484848484848486e-06,
+      "loss": 1.4527,
+      "step": 95350
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.3434343434343437e-06,
+      "loss": 1.4827,
+      "step": 95360
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.3383838383838384e-06,
+      "loss": 1.4883,
+      "step": 95370
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.3333333333333336e-06,
+      "loss": 1.4648,
+      "step": 95380
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.3282828282828283e-06,
+      "loss": 1.4653,
+      "step": 95390
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.3232323232323234e-06,
+      "loss": 1.4703,
+      "step": 95400
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.318181818181818e-06,
+      "loss": 1.4506,
+      "step": 95410
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.3131313131313133e-06,
+      "loss": 1.4734,
+      "step": 95420
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.308080808080808e-06,
+      "loss": 1.448,
+      "step": 95430
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.303030303030303e-06,
+      "loss": 1.4652,
+      "step": 95440
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.297979797979798e-06,
+      "loss": 1.4372,
+      "step": 95450
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.292929292929293e-06,
+      "loss": 1.4653,
+      "step": 95460
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.287878787878788e-06,
+      "loss": 1.4678,
+      "step": 95470
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.282828282828283e-06,
+      "loss": 1.4705,
+      "step": 95480
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.277777777777778e-06,
+      "loss": 1.4776,
+      "step": 95490
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.2727272727272728e-06,
+      "loss": 1.4716,
+      "step": 95500
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.267676767676768e-06,
+      "loss": 1.4704,
+      "step": 95510
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.2626262626262626e-06,
+      "loss": 1.4694,
+      "step": 95520
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.2575757575757578e-06,
+      "loss": 1.4744,
+      "step": 95530
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.2525252525252525e-06,
+      "loss": 1.4848,
+      "step": 95540
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.2474747474747476e-06,
+      "loss": 1.4571,
+      "step": 95550
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.2424242424242423e-06,
+      "loss": 1.4464,
+      "step": 95560
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.2373737373737375e-06,
+      "loss": 1.4498,
+      "step": 95570
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.2323232323232326e-06,
+      "loss": 1.4685,
+      "step": 95580
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.2272727272727274e-06,
+      "loss": 1.4834,
+      "step": 95590
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.2222222222222225e-06,
+      "loss": 1.4726,
+      "step": 95600
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.2171717171717172e-06,
+      "loss": 1.4761,
+      "step": 95610
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.2121212121212124e-06,
+      "loss": 1.458,
+      "step": 95620
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.207070707070707e-06,
+      "loss": 1.4652,
+      "step": 95630
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.2020202020202022e-06,
+      "loss": 1.4778,
+      "step": 95640
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.196969696969697e-06,
+      "loss": 1.4756,
+      "step": 95650
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.191919191919192e-06,
+      "loss": 1.4685,
+      "step": 95660
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.186868686868687e-06,
+      "loss": 1.4786,
+      "step": 95670
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.181818181818182e-06,
+      "loss": 1.4664,
+      "step": 95680
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.176767676767677e-06,
+      "loss": 1.47,
+      "step": 95690
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.171717171717172e-06,
+      "loss": 1.4809,
+      "step": 95700
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.166666666666667e-06,
+      "loss": 1.4899,
+      "step": 95710
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.1616161616161617e-06,
+      "loss": 1.4601,
+      "step": 95720
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.156565656565657e-06,
+      "loss": 1.4475,
+      "step": 95730
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.1515151515151515e-06,
+      "loss": 1.4774,
+      "step": 95740
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.1464646464646467e-06,
+      "loss": 1.4789,
+      "step": 95750
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.1414141414141414e-06,
+      "loss": 1.4479,
+      "step": 95760
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.1363636363636365e-06,
+      "loss": 1.447,
+      "step": 95770
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.1313131313131313e-06,
+      "loss": 1.4512,
+      "step": 95780
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.1262626262626264e-06,
+      "loss": 1.4732,
+      "step": 95790
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.1212121212121216e-06,
+      "loss": 1.4426,
+      "step": 95800
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.1161616161616163e-06,
+      "loss": 1.4929,
+      "step": 95810
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.1111111111111114e-06,
+      "loss": 1.4496,
+      "step": 95820
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.106060606060606e-06,
+      "loss": 1.4514,
+      "step": 95830
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.1010101010101013e-06,
+      "loss": 1.4852,
+      "step": 95840
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.095959595959596e-06,
+      "loss": 1.4719,
+      "step": 95850
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.090909090909091e-06,
+      "loss": 1.4581,
+      "step": 95860
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.085858585858586e-06,
+      "loss": 1.4625,
+      "step": 95870
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.0808080808080806e-06,
+      "loss": 1.4658,
+      "step": 95880
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.0757575757575757e-06,
+      "loss": 1.4788,
+      "step": 95890
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.070707070707071e-06,
+      "loss": 1.4614,
+      "step": 95900
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.065656565656566e-06,
+      "loss": 1.4409,
+      "step": 95910
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.0606060606060607e-06,
+      "loss": 1.4701,
+      "step": 95920
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.055555555555556e-06,
+      "loss": 1.453,
+      "step": 95930
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.0505050505050506e-06,
+      "loss": 1.4693,
+      "step": 95940
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.0454545454545457e-06,
+      "loss": 1.455,
+      "step": 95950
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.0404040404040405e-06,
+      "loss": 1.4601,
+      "step": 95960
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.035353535353535e-06,
+      "loss": 1.4711,
+      "step": 95970
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.0303030303030303e-06,
+      "loss": 1.4725,
+      "step": 95980
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.025252525252525e-06,
+      "loss": 1.4757,
+      "step": 95990
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.0202020202020206e-06,
+      "loss": 1.4928,
+      "step": 96000
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.0151515151515153e-06,
+      "loss": 1.4507,
+      "step": 96010
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.0101010101010105e-06,
+      "loss": 1.4508,
+      "step": 96020
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.005050505050505e-06,
+      "loss": 1.4691,
+      "step": 96030
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 1.4533,
+      "step": 96040
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.994949494949495e-06,
+      "loss": 1.4545,
+      "step": 96050
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.9898989898989898e-06,
+      "loss": 1.4557,
+      "step": 96060
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.984848484848485e-06,
+      "loss": 1.4795,
+      "step": 96070
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.9797979797979796e-06,
+      "loss": 1.4643,
+      "step": 96080
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.9747474747474748e-06,
+      "loss": 1.473,
+      "step": 96090
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.9696969696969695e-06,
+      "loss": 1.4583,
+      "step": 96100
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.964646464646465e-06,
+      "loss": 1.4658,
+      "step": 96110
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.9595959595959598e-06,
+      "loss": 1.4646,
+      "step": 96120
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.954545454545455e-06,
+      "loss": 1.4583,
+      "step": 96130
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.9494949494949496e-06,
+      "loss": 1.4781,
+      "step": 96140
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.9444444444444444e-06,
+      "loss": 1.4621,
+      "step": 96150
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.9393939393939395e-06,
+      "loss": 1.4728,
+      "step": 96160
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.9343434343434342e-06,
+      "loss": 1.4645,
+      "step": 96170
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.9292929292929294e-06,
+      "loss": 1.4772,
+      "step": 96180
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.924242424242424e-06,
+      "loss": 1.4411,
+      "step": 96190
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.9191919191919192e-06,
+      "loss": 1.4626,
+      "step": 96200
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.914141414141414e-06,
+      "loss": 1.4635,
+      "step": 96210
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.9090909090909095e-06,
+      "loss": 1.4536,
+      "step": 96220
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.9040404040404042e-06,
+      "loss": 1.4552,
+      "step": 96230
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8989898989898992e-06,
+      "loss": 1.4388,
+      "step": 96240
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8939393939393941e-06,
+      "loss": 1.4938,
+      "step": 96250
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.888888888888889e-06,
+      "loss": 1.4547,
+      "step": 96260
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.883838383838384e-06,
+      "loss": 1.434,
+      "step": 96270
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.878787878787879e-06,
+      "loss": 1.4638,
+      "step": 96280
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8737373737373738e-06,
+      "loss": 1.4614,
+      "step": 96290
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8686868686868688e-06,
+      "loss": 1.4696,
+      "step": 96300
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8636363636363635e-06,
+      "loss": 1.4458,
+      "step": 96310
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8585858585858588e-06,
+      "loss": 1.4661,
+      "step": 96320
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8535353535353538e-06,
+      "loss": 1.4572,
+      "step": 96330
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8484848484848487e-06,
+      "loss": 1.4553,
+      "step": 96340
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8434343434343436e-06,
+      "loss": 1.4556,
+      "step": 96350
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8383838383838386e-06,
+      "loss": 1.4758,
+      "step": 96360
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8333333333333335e-06,
+      "loss": 1.452,
+      "step": 96370
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8282828282828284e-06,
+      "loss": 1.4622,
+      "step": 96380
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8232323232323234e-06,
+      "loss": 1.4744,
+      "step": 96390
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.818181818181818e-06,
+      "loss": 1.4553,
+      "step": 96400
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.813131313131313e-06,
+      "loss": 1.4556,
+      "step": 96410
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.808080808080808e-06,
+      "loss": 1.4547,
+      "step": 96420
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.8030303030303033e-06,
+      "loss": 1.4542,
+      "step": 96430
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.7979797979797982e-06,
+      "loss": 1.4593,
+      "step": 96440
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.7929292929292932e-06,
+      "loss": 1.4465,
+      "step": 96450
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.787878787878788e-06,
+      "loss": 1.4619,
+      "step": 96460
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.782828282828283e-06,
+      "loss": 1.5023,
+      "step": 96470
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.777777777777778e-06,
+      "loss": 1.462,
+      "step": 96480
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.7727272727272727e-06,
+      "loss": 1.4777,
+      "step": 96490
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.7676767676767676e-06,
+      "loss": 1.4684,
+      "step": 96500
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.7626262626262625e-06,
+      "loss": 1.4595,
+      "step": 96510
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.7575757575757575e-06,
+      "loss": 1.4541,
+      "step": 96520
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.7525252525252524e-06,
+      "loss": 1.4645,
+      "step": 96530
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.7474747474747478e-06,
+      "loss": 1.4779,
+      "step": 96540
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.7424242424242427e-06,
+      "loss": 1.46,
+      "step": 96550
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.7373737373737376e-06,
+      "loss": 1.4743,
+      "step": 96560
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.7323232323232326e-06,
+      "loss": 1.4692,
+      "step": 96570
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.7272727272727273e-06,
+      "loss": 1.4649,
+      "step": 96580
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.7222222222222222e-06,
+      "loss": 1.4715,
+      "step": 96590
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.7171717171717171e-06,
+      "loss": 1.4491,
+      "step": 96600
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.712121212121212e-06,
+      "loss": 1.463,
+      "step": 96610
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.707070707070707e-06,
+      "loss": 1.474,
+      "step": 96620
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.702020202020202e-06,
+      "loss": 1.4732,
+      "step": 96630
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.6969696969696973e-06,
+      "loss": 1.4452,
+      "step": 96640
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.6919191919191922e-06,
+      "loss": 1.4397,
+      "step": 96650
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.6868686868686871e-06,
+      "loss": 1.4626,
+      "step": 96660
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.681818181818182e-06,
+      "loss": 1.4573,
+      "step": 96670
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.6767676767676768e-06,
+      "loss": 1.4735,
+      "step": 96680
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.6717171717171717e-06,
+      "loss": 1.446,
+      "step": 96690
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.6666666666666667e-06,
+      "loss": 1.4639,
+      "step": 96700
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.6616161616161616e-06,
+      "loss": 1.4587,
+      "step": 96710
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.6565656565656565e-06,
+      "loss": 1.4577,
+      "step": 96720
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.6515151515151515e-06,
+      "loss": 1.459,
+      "step": 96730
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.6464646464646464e-06,
+      "loss": 1.4686,
+      "step": 96740
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.6414141414141417e-06,
+      "loss": 1.468,
+      "step": 96750
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.6363636363636367e-06,
+      "loss": 1.4709,
+      "step": 96760
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.6313131313131314e-06,
+      "loss": 1.464,
+      "step": 96770
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.6262626262626263e-06,
+      "loss": 1.4708,
+      "step": 96780
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.6212121212121213e-06,
+      "loss": 1.4747,
+      "step": 96790
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.6161616161616162e-06,
+      "loss": 1.4613,
+      "step": 96800
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.6111111111111111e-06,
+      "loss": 1.4506,
+      "step": 96810
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.606060606060606e-06,
+      "loss": 1.462,
+      "step": 96820
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.601010101010101e-06,
+      "loss": 1.4514,
+      "step": 96830
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.595959595959596e-06,
+      "loss": 1.4834,
+      "step": 96840
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5909090909090908e-06,
+      "loss": 1.4628,
+      "step": 96850
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.585858585858586e-06,
+      "loss": 1.4577,
+      "step": 96860
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.580808080808081e-06,
+      "loss": 1.4586,
+      "step": 96870
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5757575757575759e-06,
+      "loss": 1.4563,
+      "step": 96880
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5707070707070708e-06,
+      "loss": 1.4678,
+      "step": 96890
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5656565656565657e-06,
+      "loss": 1.4627,
+      "step": 96900
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5606060606060606e-06,
+      "loss": 1.4657,
+      "step": 96910
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5555555555555556e-06,
+      "loss": 1.4754,
+      "step": 96920
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5505050505050505e-06,
+      "loss": 1.4708,
+      "step": 96930
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5454545454545457e-06,
+      "loss": 1.4846,
+      "step": 96940
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5404040404040406e-06,
+      "loss": 1.4538,
+      "step": 96950
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5353535353535355e-06,
+      "loss": 1.4604,
+      "step": 96960
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5303030303030302e-06,
+      "loss": 1.4575,
+      "step": 96970
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5252525252525252e-06,
+      "loss": 1.4769,
+      "step": 96980
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5202020202020203e-06,
+      "loss": 1.4526,
+      "step": 96990
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5151515151515152e-06,
+      "loss": 1.4606,
+      "step": 97000
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5101010101010102e-06,
+      "loss": 1.4691,
+      "step": 97010
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5050505050505051e-06,
+      "loss": 1.4547,
+      "step": 97020
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.5e-06,
+      "loss": 1.4614,
+      "step": 97030
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4949494949494952e-06,
+      "loss": 1.4774,
+      "step": 97040
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4898989898989901e-06,
+      "loss": 1.453,
+      "step": 97050
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.484848484848485e-06,
+      "loss": 1.4716,
+      "step": 97060
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4797979797979798e-06,
+      "loss": 1.4896,
+      "step": 97070
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4747474747474747e-06,
+      "loss": 1.4659,
+      "step": 97080
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4696969696969696e-06,
+      "loss": 1.4672,
+      "step": 97090
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4646464646464648e-06,
+      "loss": 1.4622,
+      "step": 97100
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4595959595959597e-06,
+      "loss": 1.4683,
+      "step": 97110
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4545454545454546e-06,
+      "loss": 1.4578,
+      "step": 97120
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4494949494949496e-06,
+      "loss": 1.4543,
+      "step": 97130
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4444444444444445e-06,
+      "loss": 1.4665,
+      "step": 97140
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4393939393939396e-06,
+      "loss": 1.4776,
+      "step": 97150
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4343434343434344e-06,
+      "loss": 1.4643,
+      "step": 97160
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4292929292929293e-06,
+      "loss": 1.464,
+      "step": 97170
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4242424242424242e-06,
+      "loss": 1.4618,
+      "step": 97180
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4191919191919192e-06,
+      "loss": 1.4517,
+      "step": 97190
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4141414141414143e-06,
+      "loss": 1.4707,
+      "step": 97200
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4090909090909092e-06,
+      "loss": 1.4594,
+      "step": 97210
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.4040404040404042e-06,
+      "loss": 1.47,
+      "step": 97220
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.398989898989899e-06,
+      "loss": 1.4738,
+      "step": 97230
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.393939393939394e-06,
+      "loss": 1.4592,
+      "step": 97240
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.388888888888889e-06,
+      "loss": 1.4462,
+      "step": 97250
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3838383838383839e-06,
+      "loss": 1.4754,
+      "step": 97260
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3787878787878788e-06,
+      "loss": 1.4571,
+      "step": 97270
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3737373737373738e-06,
+      "loss": 1.4918,
+      "step": 97280
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3686868686868687e-06,
+      "loss": 1.461,
+      "step": 97290
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3636363636363636e-06,
+      "loss": 1.448,
+      "step": 97300
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3585858585858588e-06,
+      "loss": 1.4769,
+      "step": 97310
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3535353535353537e-06,
+      "loss": 1.4802,
+      "step": 97320
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3484848484848486e-06,
+      "loss": 1.4785,
+      "step": 97330
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3434343434343436e-06,
+      "loss": 1.4762,
+      "step": 97340
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3383838383838385e-06,
+      "loss": 1.4757,
+      "step": 97350
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3333333333333334e-06,
+      "loss": 1.4618,
+      "step": 97360
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3282828282828283e-06,
+      "loss": 1.4597,
+      "step": 97370
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3232323232323233e-06,
+      "loss": 1.4843,
+      "step": 97380
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3181818181818182e-06,
+      "loss": 1.4299,
+      "step": 97390
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3131313131313131e-06,
+      "loss": 1.4807,
+      "step": 97400
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.308080808080808e-06,
+      "loss": 1.4488,
+      "step": 97410
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.3030303030303032e-06,
+      "loss": 1.4666,
+      "step": 97420
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.2979797979797981e-06,
+      "loss": 1.469,
+      "step": 97430
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.292929292929293e-06,
+      "loss": 1.4614,
+      "step": 97440
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.287878787878788e-06,
+      "loss": 1.4676,
+      "step": 97450
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2828282828282827e-06,
+      "loss": 1.4659,
+      "step": 97460
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2777777777777779e-06,
+      "loss": 1.4699,
+      "step": 97470
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2727272727272728e-06,
+      "loss": 1.4495,
+      "step": 97480
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2676767676767677e-06,
+      "loss": 1.4578,
+      "step": 97490
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2626262626262627e-06,
+      "loss": 1.4868,
+      "step": 97500
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2575757575757576e-06,
+      "loss": 1.4622,
+      "step": 97510
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2525252525252527e-06,
+      "loss": 1.4751,
+      "step": 97520
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2474747474747477e-06,
+      "loss": 1.4626,
+      "step": 97530
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2424242424242426e-06,
+      "loss": 1.4608,
+      "step": 97540
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2373737373737373e-06,
+      "loss": 1.4515,
+      "step": 97550
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2323232323232323e-06,
+      "loss": 1.4616,
+      "step": 97560
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2272727272727272e-06,
+      "loss": 1.4762,
+      "step": 97570
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2222222222222223e-06,
+      "loss": 1.4632,
+      "step": 97580
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2171717171717173e-06,
+      "loss": 1.4611,
+      "step": 97590
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2121212121212122e-06,
+      "loss": 1.467,
+      "step": 97600
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.2070707070707071e-06,
+      "loss": 1.4579,
+      "step": 97610
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.202020202020202e-06,
+      "loss": 1.4783,
+      "step": 97620
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1969696969696972e-06,
+      "loss": 1.4547,
+      "step": 97630
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.191919191919192e-06,
+      "loss": 1.4597,
+      "step": 97640
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1868686868686869e-06,
+      "loss": 1.4668,
+      "step": 97650
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1818181818181818e-06,
+      "loss": 1.4625,
+      "step": 97660
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1767676767676767e-06,
+      "loss": 1.4744,
+      "step": 97670
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1717171717171719e-06,
+      "loss": 1.4565,
+      "step": 97680
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1666666666666668e-06,
+      "loss": 1.4771,
+      "step": 97690
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1616161616161617e-06,
+      "loss": 1.4482,
+      "step": 97700
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1565656565656567e-06,
+      "loss": 1.4832,
+      "step": 97710
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1515151515151516e-06,
+      "loss": 1.4766,
+      "step": 97720
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1464646464646465e-06,
+      "loss": 1.4463,
+      "step": 97730
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1414141414141414e-06,
+      "loss": 1.4701,
+      "step": 97740
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1363636363636364e-06,
+      "loss": 1.4669,
+      "step": 97750
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1313131313131313e-06,
+      "loss": 1.4858,
+      "step": 97760
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1262626262626262e-06,
+      "loss": 1.4673,
+      "step": 97770
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1212121212121212e-06,
+      "loss": 1.4619,
+      "step": 97780
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1161616161616163e-06,
+      "loss": 1.474,
+      "step": 97790
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1111111111111112e-06,
+      "loss": 1.4806,
+      "step": 97800
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1060606060606062e-06,
+      "loss": 1.4391,
+      "step": 97810
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.1010101010101011e-06,
+      "loss": 1.4425,
+      "step": 97820
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.095959595959596e-06,
+      "loss": 1.4652,
+      "step": 97830
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.090909090909091e-06,
+      "loss": 1.472,
+      "step": 97840
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.085858585858586e-06,
+      "loss": 1.4663,
+      "step": 97850
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0808080808080808e-06,
+      "loss": 1.4405,
+      "step": 97860
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0757575757575758e-06,
+      "loss": 1.4704,
+      "step": 97870
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0707070707070707e-06,
+      "loss": 1.472,
+      "step": 97880
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0656565656565656e-06,
+      "loss": 1.467,
+      "step": 97890
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0606060606060608e-06,
+      "loss": 1.4663,
+      "step": 97900
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0555555555555557e-06,
+      "loss": 1.4651,
+      "step": 97910
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0505050505050506e-06,
+      "loss": 1.4638,
+      "step": 97920
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0454545454545456e-06,
+      "loss": 1.4592,
+      "step": 97930
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0404040404040403e-06,
+      "loss": 1.4671,
+      "step": 97940
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0353535353535354e-06,
+      "loss": 1.4775,
+      "step": 97950
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0303030303030304e-06,
+      "loss": 1.4589,
+      "step": 97960
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0252525252525253e-06,
+      "loss": 1.4674,
+      "step": 97970
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0202020202020202e-06,
+      "loss": 1.4439,
+      "step": 97980
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0151515151515152e-06,
+      "loss": 1.462,
+      "step": 97990
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0101010101010103e-06,
+      "loss": 1.4898,
+      "step": 98000
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0050505050505052e-06,
+      "loss": 1.4717,
+      "step": 98010
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 1.4695,
+      "step": 98020
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.949494949494949e-07,
+      "loss": 1.4658,
+      "step": 98030
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.898989898989898e-07,
+      "loss": 1.4596,
+      "step": 98040
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.848484848484847e-07,
+      "loss": 1.4529,
+      "step": 98050
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.797979797979799e-07,
+      "loss": 1.465,
+      "step": 98060
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.747474747474748e-07,
+      "loss": 1.4509,
+      "step": 98070
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.696969696969698e-07,
+      "loss": 1.4807,
+      "step": 98080
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.646464646464647e-07,
+      "loss": 1.4452,
+      "step": 98090
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.595959595959596e-07,
+      "loss": 1.4781,
+      "step": 98100
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.545454545454548e-07,
+      "loss": 1.4765,
+      "step": 98110
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.494949494949496e-07,
+      "loss": 1.4818,
+      "step": 98120
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.444444444444445e-07,
+      "loss": 1.4789,
+      "step": 98130
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.393939393939395e-07,
+      "loss": 1.4786,
+      "step": 98140
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.343434343434344e-07,
+      "loss": 1.4544,
+      "step": 98150
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.292929292929294e-07,
+      "loss": 1.4531,
+      "step": 98160
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.242424242424244e-07,
+      "loss": 1.4668,
+      "step": 98170
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.191919191919193e-07,
+      "loss": 1.4399,
+      "step": 98180
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.141414141414142e-07,
+      "loss": 1.4651,
+      "step": 98190
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.09090909090909e-07,
+      "loss": 1.4712,
+      "step": 98200
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 9.04040404040404e-07,
+      "loss": 1.4826,
+      "step": 98210
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.989898989898991e-07,
+      "loss": 1.4531,
+      "step": 98220
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.93939393939394e-07,
+      "loss": 1.4843,
+      "step": 98230
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.88888888888889e-07,
+      "loss": 1.4664,
+      "step": 98240
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.838383838383838e-07,
+      "loss": 1.4625,
+      "step": 98250
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.787878787878787e-07,
+      "loss": 1.4648,
+      "step": 98260
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.737373737373739e-07,
+      "loss": 1.4803,
+      "step": 98270
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.686868686868688e-07,
+      "loss": 1.4571,
+      "step": 98280
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.636363636363636e-07,
+      "loss": 1.4787,
+      "step": 98290
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.585858585858586e-07,
+      "loss": 1.4945,
+      "step": 98300
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.535353535353535e-07,
+      "loss": 1.4536,
+      "step": 98310
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.484848484848486e-07,
+      "loss": 1.462,
+      "step": 98320
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.434343434343436e-07,
+      "loss": 1.4631,
+      "step": 98330
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.383838383838384e-07,
+      "loss": 1.4527,
+      "step": 98340
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.333333333333333e-07,
+      "loss": 1.4685,
+      "step": 98350
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.282828282828283e-07,
+      "loss": 1.4555,
+      "step": 98360
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.232323232323232e-07,
+      "loss": 1.4682,
+      "step": 98370
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.181818181818183e-07,
+      "loss": 1.4456,
+      "step": 98380
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.131313131313132e-07,
+      "loss": 1.4746,
+      "step": 98390
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.080808080808081e-07,
+      "loss": 1.4606,
+      "step": 98400
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 8.03030303030303e-07,
+      "loss": 1.4628,
+      "step": 98410
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.97979797979798e-07,
+      "loss": 1.4729,
+      "step": 98420
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.92929292929293e-07,
+      "loss": 1.476,
+      "step": 98430
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.878787878787879e-07,
+      "loss": 1.4589,
+      "step": 98440
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.828282828282829e-07,
+      "loss": 1.4508,
+      "step": 98450
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.777777777777778e-07,
+      "loss": 1.4541,
+      "step": 98460
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.727272727272728e-07,
+      "loss": 1.4695,
+      "step": 98470
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.676767676767678e-07,
+      "loss": 1.4567,
+      "step": 98480
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.626262626262626e-07,
+      "loss": 1.4529,
+      "step": 98490
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.575757575757576e-07,
+      "loss": 1.4481,
+      "step": 98500
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.525252525252526e-07,
+      "loss": 1.4752,
+      "step": 98510
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.474747474747476e-07,
+      "loss": 1.4531,
+      "step": 98520
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.424242424242425e-07,
+      "loss": 1.4778,
+      "step": 98530
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.373737373737373e-07,
+      "loss": 1.464,
+      "step": 98540
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.323232323232324e-07,
+      "loss": 1.4999,
+      "step": 98550
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.272727272727273e-07,
+      "loss": 1.4922,
+      "step": 98560
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.222222222222222e-07,
+      "loss": 1.466,
+      "step": 98570
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.171717171717172e-07,
+      "loss": 1.4693,
+      "step": 98580
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.121212121212121e-07,
+      "loss": 1.4679,
+      "step": 98590
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.070707070707071e-07,
+      "loss": 1.4708,
+      "step": 98600
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 7.020202020202021e-07,
+      "loss": 1.4633,
+      "step": 98610
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.96969696969697e-07,
+      "loss": 1.4703,
+      "step": 98620
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.919191919191919e-07,
+      "loss": 1.4592,
+      "step": 98630
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.868686868686869e-07,
+      "loss": 1.4776,
+      "step": 98640
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.818181818181818e-07,
+      "loss": 1.4651,
+      "step": 98650
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.767676767676768e-07,
+      "loss": 1.4405,
+      "step": 98660
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.717171717171718e-07,
+      "loss": 1.4632,
+      "step": 98670
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.666666666666667e-07,
+      "loss": 1.466,
+      "step": 98680
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.616161616161616e-07,
+      "loss": 1.4735,
+      "step": 98690
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.565656565656566e-07,
+      "loss": 1.4641,
+      "step": 98700
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.515151515151516e-07,
+      "loss": 1.468,
+      "step": 98710
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.464646464646465e-07,
+      "loss": 1.4497,
+      "step": 98720
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.414141414141414e-07,
+      "loss": 1.475,
+      "step": 98730
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.363636363636364e-07,
+      "loss": 1.4711,
+      "step": 98740
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.313131313131313e-07,
+      "loss": 1.4572,
+      "step": 98750
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.262626262626264e-07,
+      "loss": 1.4537,
+      "step": 98760
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.212121212121213e-07,
+      "loss": 1.453,
+      "step": 98770
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 6.161616161616161e-07,
+      "loss": 1.4522,
+      "step": 98780
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 6.111111111111112e-07,
+      "loss": 1.4631,
+      "step": 98790
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 6.060606060606061e-07,
+      "loss": 1.453,
+      "step": 98800
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 6.01010101010101e-07,
+      "loss": 1.4653,
+      "step": 98810
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.95959595959596e-07,
+      "loss": 1.459,
+      "step": 98820
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.909090909090909e-07,
+      "loss": 1.4424,
+      "step": 98830
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.858585858585859e-07,
+      "loss": 1.4702,
+      "step": 98840
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.808080808080809e-07,
+      "loss": 1.4596,
+      "step": 98850
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.757575757575758e-07,
+      "loss": 1.4767,
+      "step": 98860
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.707070707070707e-07,
+      "loss": 1.4691,
+      "step": 98870
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.656565656565657e-07,
+      "loss": 1.4702,
+      "step": 98880
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.606060606060606e-07,
+      "loss": 1.443,
+      "step": 98890
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.555555555555556e-07,
+      "loss": 1.454,
+      "step": 98900
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.505050505050506e-07,
+      "loss": 1.4424,
+      "step": 98910
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.454545454545455e-07,
+      "loss": 1.4584,
+      "step": 98920
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.404040404040404e-07,
+      "loss": 1.4854,
+      "step": 98930
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.353535353535354e-07,
+      "loss": 1.4647,
+      "step": 98940
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.303030303030304e-07,
+      "loss": 1.4503,
+      "step": 98950
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.252525252525253e-07,
+      "loss": 1.4907,
+      "step": 98960
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.202020202020201e-07,
+      "loss": 1.4504,
+      "step": 98970
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.151515151515152e-07,
+      "loss": 1.4435,
+      "step": 98980
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.101010101010101e-07,
+      "loss": 1.4555,
+      "step": 98990
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.050505050505052e-07,
+      "loss": 1.4704,
+      "step": 99000
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.000000000000001e-07,
+      "loss": 1.4754,
+      "step": 99010
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.949494949494949e-07,
+      "loss": 1.4882,
+      "step": 99020
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.898989898989899e-07,
+      "loss": 1.4382,
+      "step": 99030
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.848484848484849e-07,
+      "loss": 1.46,
+      "step": 99040
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.797979797979798e-07,
+      "loss": 1.4917,
+      "step": 99050
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.747474747474748e-07,
+      "loss": 1.4693,
+      "step": 99060
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.696969696969697e-07,
+      "loss": 1.4606,
+      "step": 99070
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.646464646464647e-07,
+      "loss": 1.4746,
+      "step": 99080
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.5959595959595964e-07,
+      "loss": 1.4674,
+      "step": 99090
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.545454545454545e-07,
+      "loss": 1.4494,
+      "step": 99100
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.4949494949494956e-07,
+      "loss": 1.4337,
+      "step": 99110
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.444444444444445e-07,
+      "loss": 1.4482,
+      "step": 99120
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.3939393939393937e-07,
+      "loss": 1.4674,
+      "step": 99130
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.343434343434344e-07,
+      "loss": 1.4505,
+      "step": 99140
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.292929292929293e-07,
+      "loss": 1.4738,
+      "step": 99150
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.242424242424243e-07,
+      "loss": 1.4636,
+      "step": 99160
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.191919191919192e-07,
+      "loss": 1.4624,
+      "step": 99170
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.1414141414141413e-07,
+      "loss": 1.4448,
+      "step": 99180
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.0909090909090917e-07,
+      "loss": 1.4483,
+      "step": 99190
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 4.0404040404040405e-07,
+      "loss": 1.4518,
+      "step": 99200
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.98989898989899e-07,
+      "loss": 1.4819,
+      "step": 99210
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.9393939393939396e-07,
+      "loss": 1.4557,
+      "step": 99220
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.888888888888889e-07,
+      "loss": 1.4491,
+      "step": 99230
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.838383838383839e-07,
+      "loss": 1.4562,
+      "step": 99240
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.787878787878788e-07,
+      "loss": 1.4609,
+      "step": 99250
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.737373737373738e-07,
+      "loss": 1.4552,
+      "step": 99260
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.686868686868687e-07,
+      "loss": 1.4601,
+      "step": 99270
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.6363636363636366e-07,
+      "loss": 1.4718,
+      "step": 99280
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.585858585858586e-07,
+      "loss": 1.4856,
+      "step": 99290
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.535353535353536e-07,
+      "loss": 1.4667,
+      "step": 99300
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.484848484848485e-07,
+      "loss": 1.4725,
+      "step": 99310
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.4343434343434344e-07,
+      "loss": 1.4693,
+      "step": 99320
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.383838383838384e-07,
+      "loss": 1.4658,
+      "step": 99330
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.3333333333333335e-07,
+      "loss": 1.4909,
+      "step": 99340
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.282828282828283e-07,
+      "loss": 1.4722,
+      "step": 99350
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.2323232323232327e-07,
+      "loss": 1.4836,
+      "step": 99360
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.181818181818182e-07,
+      "loss": 1.4656,
+      "step": 99370
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.131313131313132e-07,
+      "loss": 1.4453,
+      "step": 99380
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.0808080808080806e-07,
+      "loss": 1.4692,
+      "step": 99390
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 3.0303030303030305e-07,
+      "loss": 1.4614,
+      "step": 99400
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.97979797979798e-07,
+      "loss": 1.4568,
+      "step": 99410
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.9292929292929296e-07,
+      "loss": 1.4599,
+      "step": 99420
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.878787878787879e-07,
+      "loss": 1.4752,
+      "step": 99430
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.8282828282828283e-07,
+      "loss": 1.4456,
+      "step": 99440
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.777777777777778e-07,
+      "loss": 1.459,
+      "step": 99450
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.7272727272727274e-07,
+      "loss": 1.465,
+      "step": 99460
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.676767676767677e-07,
+      "loss": 1.4626,
+      "step": 99470
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.6262626262626266e-07,
+      "loss": 1.4539,
+      "step": 99480
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.575757575757576e-07,
+      "loss": 1.4435,
+      "step": 99490
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.525252525252526e-07,
+      "loss": 1.4585,
+      "step": 99500
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.4747474747474745e-07,
+      "loss": 1.4814,
+      "step": 99510
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.4242424242424244e-07,
+      "loss": 1.4616,
+      "step": 99520
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.373737373737374e-07,
+      "loss": 1.483,
+      "step": 99530
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.3232323232323235e-07,
+      "loss": 1.4639,
+      "step": 99540
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.2727272727272726e-07,
+      "loss": 1.4619,
+      "step": 99550
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.2222222222222224e-07,
+      "loss": 1.4684,
+      "step": 99560
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.171717171717172e-07,
+      "loss": 1.4626,
+      "step": 99570
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.1212121212121216e-07,
+      "loss": 1.4446,
+      "step": 99580
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.0707070707070707e-07,
+      "loss": 1.4556,
+      "step": 99590
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.0202020202020202e-07,
+      "loss": 1.4624,
+      "step": 99600
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.9696969696969698e-07,
+      "loss": 1.4483,
+      "step": 99610
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.9191919191919194e-07,
+      "loss": 1.4714,
+      "step": 99620
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.868686868686869e-07,
+      "loss": 1.4729,
+      "step": 99630
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.8181818181818183e-07,
+      "loss": 1.4605,
+      "step": 99640
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.767676767676768e-07,
+      "loss": 1.4632,
+      "step": 99650
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.7171717171717172e-07,
+      "loss": 1.4685,
+      "step": 99660
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.6666666666666668e-07,
+      "loss": 1.4626,
+      "step": 99670
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.6161616161616163e-07,
+      "loss": 1.4755,
+      "step": 99680
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.565656565656566e-07,
+      "loss": 1.4688,
+      "step": 99690
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.5151515151515152e-07,
+      "loss": 1.4741,
+      "step": 99700
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.4646464646464648e-07,
+      "loss": 1.4808,
+      "step": 99710
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.4141414141414141e-07,
+      "loss": 1.4423,
+      "step": 99720
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.3636363636363637e-07,
+      "loss": 1.4596,
+      "step": 99730
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.3131313131313133e-07,
+      "loss": 1.4536,
+      "step": 99740
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.262626262626263e-07,
+      "loss": 1.447,
+      "step": 99750
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.2121212121212122e-07,
+      "loss": 1.4628,
+      "step": 99760
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.1616161616161618e-07,
+      "loss": 1.4602,
+      "step": 99770
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.1111111111111112e-07,
+      "loss": 1.4803,
+      "step": 99780
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0606060606060608e-07,
+      "loss": 1.4582,
+      "step": 99790
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0101010101010101e-07,
+      "loss": 1.4731,
+      "step": 99800
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.595959595959597e-08,
+      "loss": 1.4939,
+      "step": 99810
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.090909090909091e-08,
+      "loss": 1.4719,
+      "step": 99820
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.585858585858586e-08,
+      "loss": 1.4556,
+      "step": 99830
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 8.080808080808082e-08,
+      "loss": 1.4495,
+      "step": 99840
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.575757575757576e-08,
+      "loss": 1.4604,
+      "step": 99850
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 7.070707070707071e-08,
+      "loss": 1.466,
+      "step": 99860
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 6.565656565656566e-08,
+      "loss": 1.4482,
+      "step": 99870
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 6.060606060606061e-08,
+      "loss": 1.4653,
+      "step": 99880
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.555555555555556e-08,
+      "loss": 1.4544,
+      "step": 99890
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.0505050505050506e-08,
+      "loss": 1.4722,
+      "step": 99900
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.545454545454546e-08,
+      "loss": 1.4725,
+      "step": 99910
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 4.040404040404041e-08,
+      "loss": 1.4746,
+      "step": 99920
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.5353535353535353e-08,
+      "loss": 1.4743,
+      "step": 99930
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.0303030303030305e-08,
+      "loss": 1.4515,
+      "step": 99940
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.5252525252525253e-08,
+      "loss": 1.4627,
+      "step": 99950
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.0202020202020204e-08,
+      "loss": 1.4782,
+      "step": 99960
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.5151515151515152e-08,
+      "loss": 1.4618,
+      "step": 99970
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0101010101010102e-08,
+      "loss": 1.4682,
+      "step": 99980
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.050505050505051e-09,
+      "loss": 1.4652,
+      "step": 99990
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0,
+      "loss": 1.4559,
+      "step": 100000
+    },
+    {
+      "epoch": 1.5,
+      "step": 100000,
+      "total_flos": 6.70969378743989e+18,
+      "train_loss": 1.5298656570243836,
+      "train_runtime": 127603.5055,
+      "train_samples_per_second": 200.621,
+      "train_steps_per_second": 0.784
+    }
+  ],
+  "max_steps": 100000,
+  "num_train_epochs": 2,
+  "total_flos": 6.70969378743989e+18,
+  "trial_name": null,
+  "trial_params": null
+}