diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,60019 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.7356240666769654,
+  "eval_steps": 500,
+  "global_step": 50000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 5e-06,
+      "loss": 36.8582,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1e-05,
+      "loss": 36.436,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5e-05,
+      "loss": 33.3434,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2e-05,
+      "loss": 28.1421,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5e-05,
+      "loss": 20.3659,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3e-05,
+      "loss": 12.4206,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.5e-05,
+      "loss": 8.0893,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4e-05,
+      "loss": 7.1108,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.5e-05,
+      "loss": 6.721,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5e-05,
+      "loss": 6.302,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.500000000000001e-05,
+      "loss": 6.2785,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6e-05,
+      "loss": 6.1095,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.500000000000001e-05,
+      "loss": 6.0131,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7e-05,
+      "loss": 7.8711,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.500000000000001e-05,
+      "loss": 6.2801,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8e-05,
+      "loss": 5.689,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.5e-05,
+      "loss": 5.5867,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9e-05,
+      "loss": 5.4931,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.5e-05,
+      "loss": 5.4438,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0001,
+      "loss": 5.3726,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999999866082645e-05,
+      "loss": 5.1215,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999999464330591e-05,
+      "loss": 5.0901,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999998794743857e-05,
+      "loss": 5.0367,
+      "step": 115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999997857322477e-05,
+      "loss": 5.1792,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999996652066505e-05,
+      "loss": 4.8346,
+      "step": 125
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999995178976003e-05,
+      "loss": 4.8685,
+      "step": 130
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999993438051052e-05,
+      "loss": 4.8739,
+      "step": 135
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999991429291743e-05,
+      "loss": 4.7365,
+      "step": 140
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999989152698185e-05,
+      "loss": 4.7402,
+      "step": 145
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.9999866082705e-05,
+      "loss": 4.6108,
+      "step": 150
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999983796008824e-05,
+      "loss": 4.6051,
+      "step": 155
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999980715913309e-05,
+      "loss": 4.5677,
+      "step": 160
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999977367984117e-05,
+      "loss": 4.6531,
+      "step": 165
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999973752221428e-05,
+      "loss": 4.5252,
+      "step": 170
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999969868625439e-05,
+      "loss": 4.5603,
+      "step": 175
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999965717196354e-05,
+      "loss": 4.444,
+      "step": 180
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999961297934398e-05,
+      "loss": 4.6131,
+      "step": 185
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999956610839806e-05,
+      "loss": 4.5129,
+      "step": 190
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999951655912829e-05,
+      "loss": 4.4684,
+      "step": 195
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999946433153735e-05,
+      "loss": 4.4337,
+      "step": 200
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999940942562802e-05,
+      "loss": 4.4378,
+      "step": 205
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999935184140324e-05,
+      "loss": 4.3133,
+      "step": 210
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999929157886609e-05,
+      "loss": 4.4076,
+      "step": 215
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99992286380198e-05,
+      "loss": 4.4622,
+      "step": 220
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999916301886776e-05,
+      "loss": 4.5325,
+      "step": 225
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999909472141347e-05,
+      "loss": 4.3896,
+      "step": 230
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999902374566058e-05,
+      "loss": 4.3327,
+      "step": 235
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999895009161291e-05,
+      "loss": 4.4087,
+      "step": 240
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999887375927439e-05,
+      "loss": 4.316,
+      "step": 245
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999879474864912e-05,
+      "loss": 4.2426,
+      "step": 250
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999871305974131e-05,
+      "loss": 4.2689,
+      "step": 255
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999862869255539e-05,
+      "loss": 4.2368,
+      "step": 260
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999854164709581e-05,
+      "loss": 4.2533,
+      "step": 265
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999845192336727e-05,
+      "loss": 4.053,
+      "step": 270
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99983595213746e-05,
+      "loss": 4.2694,
+      "step": 275
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999826444112269e-05,
+      "loss": 4.2806,
+      "step": 280
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999816668261668e-05,
+      "loss": 4.162,
+      "step": 285
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999806624586178e-05,
+      "loss": 4.0032,
+      "step": 290
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99979631308634e-05,
+      "loss": 4.0645,
+      "step": 295
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999785733762704e-05,
+      "loss": 4.1695,
+      "step": 300
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999774886615836e-05,
+      "loss": 4.1508,
+      "step": 305
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999763771646319e-05,
+      "loss": 4.2306,
+      "step": 310
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999752388854748e-05,
+      "loss": 4.2333,
+      "step": 315
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999740738241732e-05,
+      "loss": 6.6336,
+      "step": 320
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999728819807895e-05,
+      "loss": 4.5915,
+      "step": 325
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.999716633553878e-05,
+      "loss": 4.3166,
+      "step": 330
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.99970417948033e-05,
+      "loss": 4.2002,
+      "step": 335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999691457587919e-05,
+      "loss": 4.0727,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999678467877329e-05,
+      "loss": 3.9954,
+      "step": 345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999665210349252e-05,
+      "loss": 4.1158,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999651685004403e-05,
+      "loss": 4.1896,
+      "step": 355
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999637891843504e-05,
+      "loss": 4.156,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999623830867291e-05,
+      "loss": 4.2087,
+      "step": 365
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999609502076522e-05,
+      "loss": 4.1638,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999594905471961e-05,
+      "loss": 4.0546,
+      "step": 375
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999580041054391e-05,
+      "loss": 4.1975,
+      "step": 380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99956490882461e-05,
+      "loss": 4.0555,
+      "step": 385
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999549508783426e-05,
+      "loss": 4.1404,
+      "step": 390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999533840931665e-05,
+      "loss": 3.8646,
+      "step": 395
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999517905270168e-05,
+      "loss": 4.0442,
+      "step": 400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999501701799784e-05,
+      "loss": 3.809,
+      "step": 405
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999485230521386e-05,
+      "loss": 4.0165,
+      "step": 410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999468491435855e-05,
+      "loss": 4.0438,
+      "step": 415
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999451484544086e-05,
+      "loss": 4.0704,
+      "step": 420
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999434209846992e-05,
+      "loss": 4.0441,
+      "step": 425
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999416667345496e-05,
+      "loss": 4.0077,
+      "step": 430
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999398857040539e-05,
+      "loss": 4.057,
+      "step": 435
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999380778933075e-05,
+      "loss": 4.0793,
+      "step": 440
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999362433024073e-05,
+      "loss": 4.0209,
+      "step": 445
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999343819314516e-05,
+      "loss": 3.8796,
+      "step": 450
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999324937805398e-05,
+      "loss": 3.9225,
+      "step": 455
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999305788497733e-05,
+      "loss": 4.0506,
+      "step": 460
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999286371392548e-05,
+      "loss": 3.8821,
+      "step": 465
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999266686490881e-05,
+      "loss": 3.8889,
+      "step": 470
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999246733793785e-05,
+      "loss": 4.0313,
+      "step": 475
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999226513302334e-05,
+      "loss": 3.9746,
+      "step": 480
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999206025017605e-05,
+      "loss": 3.8541,
+      "step": 485
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999185268940701e-05,
+      "loss": 3.8769,
+      "step": 490
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99916424507273e-05,
+      "loss": 3.9535,
+      "step": 495
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99914295341482e-05,
+      "loss": 4.0212,
+      "step": 500
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99912139396811e-05,
+      "loss": 3.9744,
+      "step": 505
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999099566733756e-05,
+      "loss": 3.9146,
+      "step": 510
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999077471712927e-05,
+      "loss": 3.9776,
+      "step": 515
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999055108906808e-05,
+      "loss": 3.9252,
+      "step": 520
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999032478316595e-05,
+      "loss": 4.0011,
+      "step": 525
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.999009579943502e-05,
+      "loss": 4.0303,
+      "step": 530
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998986413788752e-05,
+      "loss": 3.8832,
+      "step": 535
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99896297985359e-05,
+      "loss": 3.9643,
+      "step": 540
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998939278139268e-05,
+      "loss": 3.9469,
+      "step": 545
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998915308647059e-05,
+      "loss": 3.7428,
+      "step": 550
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998891071378244e-05,
+      "loss": 3.8554,
+      "step": 555
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998866566334123e-05,
+      "loss": 4.0769,
+      "step": 560
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998841793516008e-05,
+      "loss": 3.89,
+      "step": 565
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998816752925226e-05,
+      "loss": 3.8947,
+      "step": 570
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998791444563119e-05,
+      "loss": 3.9495,
+      "step": 575
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998765868431043e-05,
+      "loss": 3.8987,
+      "step": 580
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998740024530366e-05,
+      "loss": 3.8747,
+      "step": 585
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998713912862473e-05,
+      "loss": 3.9176,
+      "step": 590
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998687533428765e-05,
+      "loss": 3.9462,
+      "step": 595
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998660886230653e-05,
+      "loss": 3.9947,
+      "step": 600
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998633971269566e-05,
+      "loss": 3.9659,
+      "step": 605
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998606788546944e-05,
+      "loss": 3.9316,
+      "step": 610
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998579338064242e-05,
+      "loss": 3.8443,
+      "step": 615
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998551619822934e-05,
+      "loss": 4.0227,
+      "step": 620
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998523633824502e-05,
+      "loss": 3.8771,
+      "step": 625
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998495380070447e-05,
+      "loss": 3.8291,
+      "step": 630
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998466858562282e-05,
+      "loss": 3.9013,
+      "step": 635
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998438069301532e-05,
+      "loss": 3.8337,
+      "step": 640
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998409012289742e-05,
+      "loss": 3.7509,
+      "step": 645
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99837968752847e-05,
+      "loss": 3.9153,
+      "step": 650
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998350095019283e-05,
+      "loss": 3.8688,
+      "step": 655
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998320234763769e-05,
+      "loss": 3.7805,
+      "step": 660
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998290106763527e-05,
+      "loss": 3.8153,
+      "step": 665
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998259711020168e-05,
+      "loss": 3.9248,
+      "step": 670
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998229047535325e-05,
+      "loss": 3.8821,
+      "step": 675
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998198116310636e-05,
+      "loss": 3.7846,
+      "step": 680
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998166917347762e-05,
+      "loss": 3.8495,
+      "step": 685
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998135450648372e-05,
+      "loss": 3.8341,
+      "step": 690
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99810371621415e-05,
+      "loss": 3.8882,
+      "step": 695
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9980717140468e-05,
+      "loss": 3.8065,
+      "step": 700
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.998039444148033e-05,
+      "loss": 3.8563,
+      "step": 705
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99800690651958e-05,
+      "loss": 3.979,
+      "step": 710
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997974101163179e-05,
+      "loss": 3.7937,
+      "step": 715
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997941028080594e-05,
+      "loss": 3.7536,
+      "step": 720
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997907687273592e-05,
+      "loss": 3.8701,
+      "step": 725
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997874078743961e-05,
+      "loss": 3.7025,
+      "step": 730
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9978402024935e-05,
+      "loss": 3.7635,
+      "step": 735
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997806058524026e-05,
+      "loss": 3.7789,
+      "step": 740
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997771646837367e-05,
+      "loss": 3.8438,
+      "step": 745
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997736967435362e-05,
+      "loss": 3.8539,
+      "step": 750
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997702020319876e-05,
+      "loss": 3.7831,
+      "step": 755
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997666805492776e-05,
+      "loss": 3.7172,
+      "step": 760
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99763132295595e-05,
+      "loss": 3.8207,
+      "step": 765
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997595572711297e-05,
+      "loss": 3.7689,
+      "step": 770
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997559554760735e-05,
+      "loss": 3.7719,
+      "step": 775
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99752326910619e-05,
+      "loss": 3.743,
+      "step": 780
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99748671574961e-05,
+      "loss": 3.811,
+      "step": 785
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99744989469295e-05,
+      "loss": 3.8579,
+      "step": 790
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997412805938183e-05,
+      "loss": 3.8754,
+      "step": 795
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997375449487294e-05,
+      "loss": 3.93,
+      "step": 800
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997337825342287e-05,
+      "loss": 3.8045,
+      "step": 805
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997299933505176e-05,
+      "loss": 3.7506,
+      "step": 810
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99726177397799e-05,
+      "loss": 3.7852,
+      "step": 815
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997223346762777e-05,
+      "loss": 3.7094,
+      "step": 820
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99718465186159e-05,
+      "loss": 3.8293,
+      "step": 825
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997145689276503e-05,
+      "loss": 3.7565,
+      "step": 830
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997106459009607e-05,
+      "loss": 3.7169,
+      "step": 835
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.997066961063e-05,
+      "loss": 3.7464,
+      "step": 840
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9970271954388e-05,
+      "loss": 3.9149,
+      "step": 845
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996987162139133e-05,
+      "loss": 3.7558,
+      "step": 850
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996946861166148e-05,
+      "loss": 3.8458,
+      "step": 855
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996906292522e-05,
+      "loss": 3.8145,
+      "step": 860
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996865456208866e-05,
+      "loss": 3.7821,
+      "step": 865
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99682435222893e-05,
+      "loss": 3.7187,
+      "step": 870
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996782980584397e-05,
+      "loss": 3.7453,
+      "step": 875
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99674134127748e-05,
+      "loss": 3.6195,
+      "step": 880
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996699434310412e-05,
+      "loss": 3.6799,
+      "step": 885
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996657259685437e-05,
+      "loss": 3.7416,
+      "step": 890
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996614817404813e-05,
+      "loss": 3.7252,
+      "step": 895
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996572107470816e-05,
+      "loss": 3.7968,
+      "step": 900
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99652912988573e-05,
+      "loss": 3.7682,
+      "step": 905
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996485884651862e-05,
+      "loss": 3.7256,
+      "step": 910
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996442371771525e-05,
+      "loss": 3.7599,
+      "step": 915
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996398591247051e-05,
+      "loss": 3.6372,
+      "step": 920
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996354543080786e-05,
+      "loss": 3.7332,
+      "step": 925
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996310227275087e-05,
+      "loss": 3.6448,
+      "step": 930
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.99626564383233e-05,
+      "loss": 3.8486,
+      "step": 935
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996220792754903e-05,
+      "loss": 3.8241,
+      "step": 940
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996175674045207e-05,
+      "loss": 3.7831,
+      "step": 945
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996130287705662e-05,
+      "loss": 3.7034,
+      "step": 950
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996084633738696e-05,
+      "loss": 3.8457,
+      "step": 955
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.996038712146756e-05,
+      "loss": 3.646,
+      "step": 960
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995992522932303e-05,
+      "loss": 3.8192,
+      "step": 965
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995946066097808e-05,
+      "loss": 3.6553,
+      "step": 970
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995899341645762e-05,
+      "loss": 3.7517,
+      "step": 975
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995852349578668e-05,
+      "loss": 3.7439,
+      "step": 980
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995805089899042e-05,
+      "loss": 3.6635,
+      "step": 985
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995757562609417e-05,
+      "loss": 3.7294,
+      "step": 990
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995709767712338e-05,
+      "loss": 3.642,
+      "step": 995
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995661705210364e-05,
+      "loss": 3.7906,
+      "step": 1000
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995613375106073e-05,
+      "loss": 3.7062,
+      "step": 1005
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995564777402049e-05,
+      "loss": 3.6426,
+      "step": 1010
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.995515912100899e-05,
+      "loss": 3.8575,
+      "step": 1015
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.995466779205241e-05,
+      "loss": 3.7994,
+      "step": 1020
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.995417378717702e-05,
+      "loss": 3.7,
+      "step": 1025
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.995367710640935e-05,
+      "loss": 3.7451,
+      "step": 1030
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.995317774977595e-05,
+      "loss": 3.7707,
+      "step": 1035
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99526757173036e-05,
+      "loss": 3.7414,
+      "step": 1040
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99521710090192e-05,
+      "loss": 3.6846,
+      "step": 1045
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.995166362494974e-05,
+      "loss": 3.7814,
+      "step": 1050
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.995115356512243e-05,
+      "loss": 3.7537,
+      "step": 1055
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.995064082956461e-05,
+      "loss": 3.6902,
+      "step": 1060
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.995012541830372e-05,
+      "loss": 3.7709,
+      "step": 1065
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994960733136738e-05,
+      "loss": 3.7579,
+      "step": 1070
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994908656878333e-05,
+      "loss": 3.7414,
+      "step": 1075
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994856313057948e-05,
+      "loss": 3.7037,
+      "step": 1080
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994803701678384e-05,
+      "loss": 3.6258,
+      "step": 1085
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994750822742463e-05,
+      "loss": 3.7534,
+      "step": 1090
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994697676253016e-05,
+      "loss": 3.6885,
+      "step": 1095
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994644262212891e-05,
+      "loss": 3.755,
+      "step": 1100
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994590580624946e-05,
+      "loss": 3.6189,
+      "step": 1105
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99453663149206e-05,
+      "loss": 3.7928,
+      "step": 1110
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994482414817121e-05,
+      "loss": 3.7195,
+      "step": 1115
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994427930603035e-05,
+      "loss": 3.6559,
+      "step": 1120
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994373178852718e-05,
+      "loss": 3.7571,
+      "step": 1125
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994318159569106e-05,
+      "loss": 3.5427,
+      "step": 1130
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994262872755143e-05,
+      "loss": 3.7414,
+      "step": 1135
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994207318413791e-05,
+      "loss": 3.7116,
+      "step": 1140
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994151496548028e-05,
+      "loss": 3.7035,
+      "step": 1145
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994095407160844e-05,
+      "loss": 3.7604,
+      "step": 1150
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.994039050255242e-05,
+      "loss": 3.7098,
+      "step": 1155
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993982425834241e-05,
+      "loss": 3.7415,
+      "step": 1160
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993925533900875e-05,
+      "loss": 3.7289,
+      "step": 1165
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993868374458191e-05,
+      "loss": 3.5431,
+      "step": 1170
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99381094750925e-05,
+      "loss": 3.6138,
+      "step": 1175
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99375325305713e-05,
+      "loss": 3.6664,
+      "step": 1180
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993695291104922e-05,
+      "loss": 3.8036,
+      "step": 1185
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993637061655729e-05,
+      "loss": 3.6467,
+      "step": 1190
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99357856471267e-05,
+      "loss": 3.6933,
+      "step": 1195
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99351980027888e-05,
+      "loss": 3.606,
+      "step": 1200
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993460768357506e-05,
+      "loss": 3.6791,
+      "step": 1205
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993401468951711e-05,
+      "loss": 3.7006,
+      "step": 1210
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993341902064669e-05,
+      "loss": 3.6638,
+      "step": 1215
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993282067699574e-05,
+      "loss": 3.7021,
+      "step": 1220
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993221965859629e-05,
+      "loss": 3.6783,
+      "step": 1225
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993161596548054e-05,
+      "loss": 3.6213,
+      "step": 1230
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993100959768083e-05,
+      "loss": 3.592,
+      "step": 1235
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.993040055522965e-05,
+      "loss": 3.5386,
+      "step": 1240
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992978883815963e-05,
+      "loss": 3.6455,
+      "step": 1245
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992917444650349e-05,
+      "loss": 3.6984,
+      "step": 1250
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992855738029419e-05,
+      "loss": 3.4803,
+      "step": 1255
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992793763956477e-05,
+      "loss": 3.7862,
+      "step": 1260
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992731522434842e-05,
+      "loss": 3.5789,
+      "step": 1265
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992669013467851e-05,
+      "loss": 3.697,
+      "step": 1270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992606237058846e-05,
+      "loss": 3.5824,
+      "step": 1275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992543193211197e-05,
+      "loss": 3.6335,
+      "step": 1280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992479881928279e-05,
+      "loss": 3.6293,
+      "step": 1285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992416303213479e-05,
+      "loss": 3.5455,
+      "step": 1290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992352457070209e-05,
+      "loss": 3.7804,
+      "step": 1295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992288343501885e-05,
+      "loss": 3.6653,
+      "step": 1300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992223962511941e-05,
+      "loss": 3.7346,
+      "step": 1305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992159314103829e-05,
+      "loss": 3.5424,
+      "step": 1310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99209439828101e-05,
+      "loss": 3.691,
+      "step": 1315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.992029215046961e-05,
+      "loss": 3.668,
+      "step": 1320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991963764405174e-05,
+      "loss": 3.6627,
+      "step": 1325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991898046359156e-05,
+      "loss": 3.7111,
+      "step": 1330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991832060912427e-05,
+      "loss": 3.6311,
+      "step": 1335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99176580806852e-05,
+      "loss": 3.7015,
+      "step": 1340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991699287830984e-05,
+      "loss": 3.7114,
+      "step": 1345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991632500203387e-05,
+      "loss": 3.7063,
+      "step": 1350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.9915654451893e-05,
+      "loss": 3.6582,
+      "step": 1355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991498122792319e-05,
+      "loss": 3.7561,
+      "step": 1360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991430533016048e-05,
+      "loss": 3.6336,
+      "step": 1365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991362675864111e-05,
+      "loss": 3.6364,
+      "step": 1370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99129455134014e-05,
+      "loss": 3.654,
+      "step": 1375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991226159447784e-05,
+      "loss": 3.53,
+      "step": 1380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991157500190706e-05,
+      "loss": 3.6855,
+      "step": 1385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.991088573572589e-05,
+      "loss": 3.7159,
+      "step": 1390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99101937959712e-05,
+      "loss": 3.7023,
+      "step": 1395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990949918268005e-05,
+      "loss": 3.6552,
+      "step": 1400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990880189588969e-05,
+      "loss": 3.5965,
+      "step": 1405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990810193563744e-05,
+      "loss": 3.72,
+      "step": 1410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990739930196082e-05,
+      "loss": 3.6708,
+      "step": 1415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990669399489743e-05,
+      "loss": 3.6388,
+      "step": 1420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990598601448508e-05,
+      "loss": 3.7352,
+      "step": 1425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990527536076171e-05,
+      "loss": 3.6231,
+      "step": 1430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990456203376535e-05,
+      "loss": 3.6777,
+      "step": 1435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990384603353423e-05,
+      "loss": 3.626,
+      "step": 1440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990312736010667e-05,
+      "loss": 3.6531,
+      "step": 1445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990240601352123e-05,
+      "loss": 3.4931,
+      "step": 1450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990168199381651e-05,
+      "loss": 3.6424,
+      "step": 1455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.99009553010313e-05,
+      "loss": 3.7137,
+      "step": 1460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.990022593520453e-05,
+      "loss": 3.6326,
+      "step": 1465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989949389637527e-05,
+      "loss": 3.5883,
+      "step": 1470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989875918458274e-05,
+      "loss": 3.7551,
+      "step": 1475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989802179986627e-05,
+      "loss": 3.6676,
+      "step": 1480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989728174226538e-05,
+      "loss": 3.6205,
+      "step": 1485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98965390118197e-05,
+      "loss": 3.5362,
+      "step": 1490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989579360856904e-05,
+      "loss": 3.7318,
+      "step": 1495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98950455325533e-05,
+      "loss": 3.599,
+      "step": 1500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98942947838126e-05,
+      "loss": 3.6245,
+      "step": 1505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989354136238709e-05,
+      "loss": 3.6117,
+      "step": 1510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989278526831717e-05,
+      "loss": 3.6012,
+      "step": 1515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989202650164333e-05,
+      "loss": 3.536,
+      "step": 1520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989126506240621e-05,
+      "loss": 3.4522,
+      "step": 1525
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.989050095064661e-05,
+      "loss": 3.6257,
+      "step": 1530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988973416640546e-05,
+      "loss": 3.6948,
+      "step": 1535
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988896470972382e-05,
+      "loss": 3.5979,
+      "step": 1540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988819258064292e-05,
+      "loss": 3.6327,
+      "step": 1545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988741777920414e-05,
+      "loss": 3.6886,
+      "step": 1550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988664030544895e-05,
+      "loss": 3.5627,
+      "step": 1555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988586015941899e-05,
+      "loss": 3.6867,
+      "step": 1560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988507734115607e-05,
+      "loss": 3.5822,
+      "step": 1565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988429185070214e-05,
+      "loss": 3.6509,
+      "step": 1570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988350368809925e-05,
+      "loss": 3.6293,
+      "step": 1575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988271285338962e-05,
+      "loss": 3.5471,
+      "step": 1580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988191934661564e-05,
+      "loss": 3.5963,
+      "step": 1585
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988112316781976e-05,
+      "loss": 3.6466,
+      "step": 1590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.988032431704471e-05,
+      "loss": 3.6479,
+      "step": 1595
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98795227943332e-05,
+      "loss": 3.5415,
+      "step": 1600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987871859972822e-05,
+      "loss": 3.552,
+      "step": 1605
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987791173327283e-05,
+      "loss": 3.6037,
+      "step": 1610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987710219501026e-05,
+      "loss": 3.6271,
+      "step": 1615
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987628998498384e-05,
+      "loss": 3.6115,
+      "step": 1620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987547510323711e-05,
+      "loss": 3.5685,
+      "step": 1625
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987465754981374e-05,
+      "loss": 3.6826,
+      "step": 1630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987383732475746e-05,
+      "loss": 3.506,
+      "step": 1635
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987301442811228e-05,
+      "loss": 3.6862,
+      "step": 1640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987218885992222e-05,
+      "loss": 3.5918,
+      "step": 1645
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.987136062023153e-05,
+      "loss": 3.6213,
+      "step": 1650
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.98705297090846e-05,
+      "loss": 3.67,
+      "step": 1655
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986969612652588e-05,
+      "loss": 3.6196,
+      "step": 1660
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986885987260007e-05,
+      "loss": 3.4913,
+      "step": 1665
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986802094735195e-05,
+      "loss": 3.5478,
+      "step": 1670
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986717935082648e-05,
+      "loss": 3.6092,
+      "step": 1675
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986633508306871e-05,
+      "loss": 3.722,
+      "step": 1680
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986548814412387e-05,
+      "loss": 3.5733,
+      "step": 1685
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986463853403734e-05,
+      "loss": 3.5771,
+      "step": 1690
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 9.986378625285461e-05,
+      "loss": 3.5213,
+      "step": 1695
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.986293130062138e-05,
+      "loss": 3.5425,
+      "step": 1700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.986207367738341e-05,
+      "loss": 3.5565,
+      "step": 1705
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.986121338318663e-05,
+      "loss": 3.6279,
+      "step": 1710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.986035041807715e-05,
+      "loss": 3.5245,
+      "step": 1715
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.985948478210119e-05,
+      "loss": 3.7299,
+      "step": 1720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.985861647530512e-05,
+      "loss": 3.5174,
+      "step": 1725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.985774549773543e-05,
+      "loss": 3.5689,
+      "step": 1730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.985687184943881e-05,
+      "loss": 3.6092,
+      "step": 1735
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.985599553046205e-05,
+      "loss": 3.6711,
+      "step": 1740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98551165408521e-05,
+      "loss": 3.5535,
+      "step": 1745
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.985423488065599e-05,
+      "loss": 3.627,
+      "step": 1750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.985335054992101e-05,
+      "loss": 3.5171,
+      "step": 1755
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.985246354869451e-05,
+      "loss": 3.7639,
+      "step": 1760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.9851573877024e-05,
+      "loss": 3.6664,
+      "step": 1765
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.985068153495713e-05,
+      "loss": 3.5746,
+      "step": 1770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984978652254172e-05,
+      "loss": 3.5116,
+      "step": 1775
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98488888398257e-05,
+      "loss": 3.627,
+      "step": 1780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984798848685717e-05,
+      "loss": 3.6477,
+      "step": 1785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984708546368433e-05,
+      "loss": 3.4687,
+      "step": 1790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984617977035558e-05,
+      "loss": 3.5837,
+      "step": 1795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984527140691942e-05,
+      "loss": 3.6509,
+      "step": 1800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984436037342453e-05,
+      "loss": 3.5699,
+      "step": 1805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984344666991967e-05,
+      "loss": 3.5617,
+      "step": 1810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984253029645383e-05,
+      "loss": 3.5451,
+      "step": 1815
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984161125307608e-05,
+      "loss": 3.6511,
+      "step": 1820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.984068953983564e-05,
+      "loss": 3.5686,
+      "step": 1825
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983976515678188e-05,
+      "loss": 3.642,
+      "step": 1830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983883810396432e-05,
+      "loss": 3.4821,
+      "step": 1835
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983790838143265e-05,
+      "loss": 3.6497,
+      "step": 1840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983697598923664e-05,
+      "loss": 3.5598,
+      "step": 1845
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983604092742624e-05,
+      "loss": 3.5851,
+      "step": 1850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983510319605154e-05,
+      "loss": 3.631,
+      "step": 1855
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983416279516277e-05,
+      "loss": 3.5769,
+      "step": 1860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983321972481033e-05,
+      "loss": 3.6566,
+      "step": 1865
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983227398504471e-05,
+      "loss": 3.5082,
+      "step": 1870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983132557591657e-05,
+      "loss": 3.615,
+      "step": 1875
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.983037449747672e-05,
+      "loss": 3.4946,
+      "step": 1880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982942074977611e-05,
+      "loss": 3.4628,
+      "step": 1885
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982846433286581e-05,
+      "loss": 3.5923,
+      "step": 1890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982750524679707e-05,
+      "loss": 3.5175,
+      "step": 1895
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982654349162127e-05,
+      "loss": 3.3945,
+      "step": 1900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982557906738992e-05,
+      "loss": 3.5921,
+      "step": 1905
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982461197415469e-05,
+      "loss": 3.4892,
+      "step": 1910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982364221196736e-05,
+      "loss": 3.4811,
+      "step": 1915
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982266978087991e-05,
+      "loss": 3.6099,
+      "step": 1920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.98216946809444e-05,
+      "loss": 3.5779,
+      "step": 1925
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.982071691221308e-05,
+      "loss": 3.4674,
+      "step": 1930
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981973647473835e-05,
+      "loss": 3.6066,
+      "step": 1935
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981875336857267e-05,
+      "loss": 3.6073,
+      "step": 1940
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981776759376874e-05,
+      "loss": 3.5693,
+      "step": 1945
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981677915037937e-05,
+      "loss": 3.6177,
+      "step": 1950
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981578803845748e-05,
+      "loss": 3.5681,
+      "step": 1955
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981479425805619e-05,
+      "loss": 3.5485,
+      "step": 1960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981379780922872e-05,
+      "loss": 3.5014,
+      "step": 1965
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981279869202845e-05,
+      "loss": 3.5455,
+      "step": 1970
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981179690650888e-05,
+      "loss": 3.6372,
+      "step": 1975
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.981079245272372e-05,
+      "loss": 3.5364,
+      "step": 1980
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980978533072672e-05,
+      "loss": 3.5147,
+      "step": 1985
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980877554057187e-05,
+      "loss": 3.541,
+      "step": 1990
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980776308231323e-05,
+      "loss": 3.56,
+      "step": 1995
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980674795600506e-05,
+      "loss": 3.6994,
+      "step": 2000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980573016170174e-05,
+      "loss": 3.427,
+      "step": 2005
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980470969945777e-05,
+      "loss": 3.4751,
+      "step": 2010
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980368656932782e-05,
+      "loss": 3.5724,
+      "step": 2015
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980266077136669e-05,
+      "loss": 3.5777,
+      "step": 2020
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980163230562934e-05,
+      "loss": 3.4552,
+      "step": 2025
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.980060117217086e-05,
+      "loss": 3.5417,
+      "step": 2030
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979956737104646e-05,
+      "loss": 3.5426,
+      "step": 2035
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979853090231157e-05,
+      "loss": 3.5165,
+      "step": 2040
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979749176602167e-05,
+      "loss": 3.4136,
+      "step": 2045
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979644996223242e-05,
+      "loss": 3.6283,
+      "step": 2050
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979540549099965e-05,
+      "loss": 3.5351,
+      "step": 2055
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979435835237929e-05,
+      "loss": 3.5846,
+      "step": 2060
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979330854642746e-05,
+      "loss": 3.5313,
+      "step": 2065
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979225607320035e-05,
+      "loss": 3.5679,
+      "step": 2070
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979120093275438e-05,
+      "loss": 3.7253,
+      "step": 2075
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.979014312514606e-05,
+      "loss": 3.4608,
+      "step": 2080
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978908265043203e-05,
+      "loss": 3.4302,
+      "step": 2085
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978801950866912e-05,
+      "loss": 3.4994,
+      "step": 2090
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978695369991427e-05,
+      "loss": 3.4884,
+      "step": 2095
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978588522422458e-05,
+      "loss": 3.5799,
+      "step": 2100
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978481408165727e-05,
+      "loss": 3.4848,
+      "step": 2105
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978374027226976e-05,
+      "loss": 3.4723,
+      "step": 2110
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97826637961195e-05,
+      "loss": 3.5265,
+      "step": 2115
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978158465326421e-05,
+      "loss": 3.5483,
+      "step": 2120
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.978050284376168e-05,
+      "loss": 3.4276,
+      "step": 2125
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977941836766985e-05,
+      "loss": 3.4464,
+      "step": 2130
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977833122504683e-05,
+      "loss": 3.5769,
+      "step": 2135
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977724141595083e-05,
+      "loss": 3.4552,
+      "step": 2140
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977614894044025e-05,
+      "loss": 3.5288,
+      "step": 2145
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97750537985736e-05,
+      "loss": 3.5324,
+      "step": 2150
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977395599040955e-05,
+      "loss": 3.5573,
+      "step": 2155
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97728555160069e-05,
+      "loss": 3.6184,
+      "step": 2160
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977175237542461e-05,
+      "loss": 3.5799,
+      "step": 2165
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.977064656872176e-05,
+      "loss": 3.5674,
+      "step": 2170
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976953809595759e-05,
+      "loss": 3.4998,
+      "step": 2175
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976842695719147e-05,
+      "loss": 3.6066,
+      "step": 2180
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976731315248293e-05,
+      "loss": 3.5246,
+      "step": 2185
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976619668189162e-05,
+      "loss": 3.4313,
+      "step": 2190
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976507754547737e-05,
+      "loss": 3.6008,
+      "step": 2195
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976395574330011e-05,
+      "loss": 3.5807,
+      "step": 2200
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976283127541992e-05,
+      "loss": 3.5578,
+      "step": 2205
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.976170414189706e-05,
+      "loss": 3.4462,
+      "step": 2210
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97605743427919e-05,
+      "loss": 3.5743,
+      "step": 2215
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975944187816495e-05,
+      "loss": 3.4555,
+      "step": 2220
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975830674807689e-05,
+      "loss": 3.6288,
+      "step": 2225
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.97571689525885e-05,
+      "loss": 3.6448,
+      "step": 2230
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975602849176076e-05,
+      "loss": 3.4175,
+      "step": 2235
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975488536565473e-05,
+      "loss": 3.4592,
+      "step": 2240
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975373957433166e-05,
+      "loss": 3.5217,
+      "step": 2245
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975259111785294e-05,
+      "loss": 3.5049,
+      "step": 2250
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975143999628005e-05,
+      "loss": 3.5514,
+      "step": 2255
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.975028620967468e-05,
+      "loss": 3.5651,
+      "step": 2260
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974912975809863e-05,
+      "loss": 3.544,
+      "step": 2265
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974797064161387e-05,
+      "loss": 3.4471,
+      "step": 2270
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974680886028245e-05,
+      "loss": 3.4647,
+      "step": 2275
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974564441416662e-05,
+      "loss": 3.439,
+      "step": 2280
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974447730332876e-05,
+      "loss": 3.5534,
+      "step": 2285
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974330752783139e-05,
+      "loss": 3.3947,
+      "step": 2290
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974213508773716e-05,
+      "loss": 3.3856,
+      "step": 2295
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.974095998310888e-05,
+      "loss": 3.4989,
+      "step": 2300
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973978221400951e-05,
+      "loss": 3.4952,
+      "step": 2305
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973860178050211e-05,
+      "loss": 3.5732,
+      "step": 2310
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973741868264993e-05,
+      "loss": 3.5241,
+      "step": 2315
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973623292051637e-05,
+      "loss": 3.3449,
+      "step": 2320
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973504449416491e-05,
+      "loss": 3.6237,
+      "step": 2325
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973385340365921e-05,
+      "loss": 3.4684,
+      "step": 2330
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973265964906309e-05,
+      "loss": 3.5566,
+      "step": 2335
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973146323044051e-05,
+      "loss": 3.4961,
+      "step": 2340
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.973026414785551e-05,
+      "loss": 3.5553,
+      "step": 2345
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972906240137237e-05,
+      "loss": 3.5508,
+      "step": 2350
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972785799105544e-05,
+      "loss": 3.43,
+      "step": 2355
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972665091696925e-05,
+      "loss": 3.3883,
+      "step": 2360
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972544117917843e-05,
+      "loss": 3.4162,
+      "step": 2365
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972422877774783e-05,
+      "loss": 3.5355,
+      "step": 2370
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972301371274235e-05,
+      "loss": 3.5303,
+      "step": 2375
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.97217959842271e-05,
+      "loss": 3.433,
+      "step": 2380
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.972057559226729e-05,
+      "loss": 3.4573,
+      "step": 2385
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.971935253692832e-05,
+      "loss": 3.6361,
+      "step": 2390
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.97181268182757e-05,
+      "loss": 3.431,
+      "step": 2395
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.971689843637507e-05,
+      "loss": 3.5276,
+      "step": 2400
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.971566739129224e-05,
+      "loss": 3.5809,
+      "step": 2405
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.971443368309317e-05,
+      "loss": 3.5297,
+      "step": 2410
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.971319731184392e-05,
+      "loss": 3.6008,
+      "step": 2415
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.971195827761073e-05,
+      "loss": 3.5452,
+      "step": 2420
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.971071658045996e-05,
+      "loss": 3.5742,
+      "step": 2425
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970947222045815e-05,
+      "loss": 3.6179,
+      "step": 2430
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970822519767194e-05,
+      "loss": 3.5432,
+      "step": 2435
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970697551216811e-05,
+      "loss": 3.3845,
+      "step": 2440
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970572316401362e-05,
+      "loss": 3.5806,
+      "step": 2445
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970446815327559e-05,
+      "loss": 3.4862,
+      "step": 2450
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.97032104800212e-05,
+      "loss": 3.484,
+      "step": 2455
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970195014431782e-05,
+      "loss": 3.5194,
+      "step": 2460
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.970068714623299e-05,
+      "loss": 3.4714,
+      "step": 2465
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969942148583435e-05,
+      "loss": 3.5821,
+      "step": 2470
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969815316318968e-05,
+      "loss": 3.5868,
+      "step": 2475
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969688217836697e-05,
+      "loss": 3.4322,
+      "step": 2480
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969560853143425e-05,
+      "loss": 3.5074,
+      "step": 2485
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969433222245976e-05,
+      "loss": 3.4918,
+      "step": 2490
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969305325151189e-05,
+      "loss": 3.4235,
+      "step": 2495
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969177161865912e-05,
+      "loss": 3.425,
+      "step": 2500
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.969048732397013e-05,
+      "loss": 3.4389,
+      "step": 2505
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96892003675137e-05,
+      "loss": 3.404,
+      "step": 2510
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968791074935878e-05,
+      "loss": 3.5364,
+      "step": 2515
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968661846957443e-05,
+      "loss": 3.572,
+      "step": 2520
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96853235282299e-05,
+      "loss": 3.5268,
+      "step": 2525
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968402592539454e-05,
+      "loss": 3.5079,
+      "step": 2530
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968272566113785e-05,
+      "loss": 3.5473,
+      "step": 2535
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96814227355295e-05,
+      "loss": 3.5368,
+      "step": 2540
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.968011714863927e-05,
+      "loss": 3.5,
+      "step": 2545
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967880890053712e-05,
+      "loss": 3.4411,
+      "step": 2550
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967749799129309e-05,
+      "loss": 3.6209,
+      "step": 2555
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967618442097743e-05,
+      "loss": 3.6402,
+      "step": 2560
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96748681896605e-05,
+      "loss": 3.4623,
+      "step": 2565
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967354929741279e-05,
+      "loss": 3.4334,
+      "step": 2570
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967222774430499e-05,
+      "loss": 3.4669,
+      "step": 2575
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.967090353040784e-05,
+      "loss": 3.3767,
+      "step": 2580
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96695766557923e-05,
+      "loss": 3.4785,
+      "step": 2585
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966824712052945e-05,
+      "loss": 3.3911,
+      "step": 2590
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96669149246905e-05,
+      "loss": 3.3581,
+      "step": 2595
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96655800683468e-05,
+      "loss": 3.5115,
+      "step": 2600
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96642425515699e-05,
+      "loss": 3.5209,
+      "step": 2605
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966290237443141e-05,
+      "loss": 3.4256,
+      "step": 2610
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966155953700311e-05,
+      "loss": 3.4028,
+      "step": 2615
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.966021403935696e-05,
+      "loss": 3.4976,
+      "step": 2620
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965886588156502e-05,
+      "loss": 3.4887,
+      "step": 2625
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96575150636995e-05,
+      "loss": 3.4162,
+      "step": 2630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965616158583276e-05,
+      "loss": 3.3569,
+      "step": 2635
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965480544803732e-05,
+      "loss": 3.4162,
+      "step": 2640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965344665038581e-05,
+      "loss": 3.4089,
+      "step": 2645
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965208519295103e-05,
+      "loss": 3.462,
+      "step": 2650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.965072107580589e-05,
+      "loss": 3.4648,
+      "step": 2655
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964935429902346e-05,
+      "loss": 3.485,
+      "step": 2660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964798486267698e-05,
+      "loss": 3.4881,
+      "step": 2665
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964661276683979e-05,
+      "loss": 3.4721,
+      "step": 2670
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964523801158539e-05,
+      "loss": 3.4973,
+      "step": 2675
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96438605969874e-05,
+      "loss": 3.4927,
+      "step": 2680
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964248052311965e-05,
+      "loss": 3.4409,
+      "step": 2685
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.964109779005603e-05,
+      "loss": 3.6038,
+      "step": 2690
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963971239787062e-05,
+      "loss": 3.4809,
+      "step": 2695
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963832434663763e-05,
+      "loss": 3.466,
+      "step": 2700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963693363643142e-05,
+      "loss": 3.4614,
+      "step": 2705
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963554026732647e-05,
+      "loss": 3.4954,
+      "step": 2710
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963414423939745e-05,
+      "loss": 3.646,
+      "step": 2715
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963274555271911e-05,
+      "loss": 3.3999,
+      "step": 2720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.963134420736637e-05,
+      "loss": 3.4129,
+      "step": 2725
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962994020341433e-05,
+      "loss": 3.5107,
+      "step": 2730
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962853354093817e-05,
+      "loss": 3.5422,
+      "step": 2735
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962712422001326e-05,
+      "loss": 3.4917,
+      "step": 2740
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962571224071506e-05,
+      "loss": 3.4766,
+      "step": 2745
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962429760311924e-05,
+      "loss": 3.4916,
+      "step": 2750
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962288030730156e-05,
+      "loss": 3.5731,
+      "step": 2755
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962146035333794e-05,
+      "loss": 3.4089,
+      "step": 2760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.962003774130446e-05,
+      "loss": 3.341,
+      "step": 2765
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96186124712773e-05,
+      "loss": 3.5665,
+      "step": 2770
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961718454333284e-05,
+      "loss": 3.4891,
+      "step": 2775
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961575395754753e-05,
+      "loss": 3.6351,
+      "step": 2780
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961432071399803e-05,
+      "loss": 3.5069,
+      "step": 2785
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96128848127611e-05,
+      "loss": 3.4067,
+      "step": 2790
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.961144625391367e-05,
+      "loss": 3.5011,
+      "step": 2795
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.96100050375328e-05,
+      "loss": 3.4193,
+      "step": 2800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960856116369566e-05,
+      "loss": 3.4709,
+      "step": 2805
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960711463247964e-05,
+      "loss": 3.4983,
+      "step": 2810
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960566544396219e-05,
+      "loss": 3.4031,
+      "step": 2815
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960421359822097e-05,
+      "loss": 3.3498,
+      "step": 2820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960275909533371e-05,
+      "loss": 3.5886,
+      "step": 2825
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.960130193537835e-05,
+      "loss": 3.3241,
+      "step": 2830
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959984211843296e-05,
+      "loss": 3.3983,
+      "step": 2835
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95983796445757e-05,
+      "loss": 3.5179,
+      "step": 2840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959691451388495e-05,
+      "loss": 3.4722,
+      "step": 2845
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959544672643916e-05,
+      "loss": 3.3835,
+      "step": 2850
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959397628231697e-05,
+      "loss": 3.3606,
+      "step": 2855
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959250318159714e-05,
+      "loss": 3.3338,
+      "step": 2860
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.959102742435859e-05,
+      "loss": 3.3634,
+      "step": 2865
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958954901068038e-05,
+      "loss": 3.7289,
+      "step": 2870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958806794064169e-05,
+      "loss": 3.4842,
+      "step": 2875
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958658421432184e-05,
+      "loss": 3.4533,
+      "step": 2880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958509783180032e-05,
+      "loss": 3.4905,
+      "step": 2885
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958360879315677e-05,
+      "loss": 3.4722,
+      "step": 2890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958211709847094e-05,
+      "loss": 3.4689,
+      "step": 2895
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.958062274782274e-05,
+      "loss": 3.5134,
+      "step": 2900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95791257412922e-05,
+      "loss": 3.3721,
+      "step": 2905
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957762607895954e-05,
+      "loss": 3.5434,
+      "step": 2910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957612376090504e-05,
+      "loss": 3.4534,
+      "step": 2915
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957461878720924e-05,
+      "loss": 3.4671,
+      "step": 2920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957311115795272e-05,
+      "loss": 3.2611,
+      "step": 2925
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957160087321625e-05,
+      "loss": 3.4579,
+      "step": 2930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.957008793308074e-05,
+      "loss": 3.4275,
+      "step": 2935
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95685723376272e-05,
+      "loss": 3.4817,
+      "step": 2940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956705408693684e-05,
+      "loss": 3.4235,
+      "step": 2945
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956553318109098e-05,
+      "loss": 3.4143,
+      "step": 2950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956400962017111e-05,
+      "loss": 3.4916,
+      "step": 2955
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956248340425882e-05,
+      "loss": 3.435,
+      "step": 2960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.956095453343588e-05,
+      "loss": 3.3965,
+      "step": 2965
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955942300778416e-05,
+      "loss": 3.4561,
+      "step": 2970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955788882738574e-05,
+      "loss": 3.4361,
+      "step": 2975
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955635199232276e-05,
+      "loss": 3.351,
+      "step": 2980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955481250267757e-05,
+      "loss": 3.389,
+      "step": 2985
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955327035853264e-05,
+      "loss": 3.541,
+      "step": 2990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955172555997057e-05,
+      "loss": 3.4672,
+      "step": 2995
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.955017810707409e-05,
+      "loss": 3.4076,
+      "step": 3000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954862799992611e-05,
+      "loss": 3.5537,
+      "step": 3005
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954707523860967e-05,
+      "loss": 3.2676,
+      "step": 3010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954551982320794e-05,
+      "loss": 3.4536,
+      "step": 3015
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954396175380425e-05,
+      "loss": 3.4325,
+      "step": 3020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954240103048203e-05,
+      "loss": 3.512,
+      "step": 3025
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.954083765332492e-05,
+      "loss": 3.3833,
+      "step": 3030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953927162241664e-05,
+      "loss": 3.5591,
+      "step": 3035
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953770293784109e-05,
+      "loss": 3.4443,
+      "step": 3040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953613159968229e-05,
+      "loss": 3.4367,
+      "step": 3045
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.953455760802444e-05,
+      "loss": 3.436,
+      "step": 3050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 9.95329809629518e-05,
+      "loss": 3.4944,
+      "step": 3055
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.953140166454889e-05,
+      "loss": 3.5404,
+      "step": 3060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.952981971290025e-05,
+      "loss": 3.4896,
+      "step": 3065
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.952823510809066e-05,
+      "loss": 3.4083,
+      "step": 3070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.952664785020498e-05,
+      "loss": 3.4877,
+      "step": 3075
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.952505793932824e-05,
+      "loss": 3.3453,
+      "step": 3080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.952346537554562e-05,
+      "loss": 3.3434,
+      "step": 3085
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.952187015894242e-05,
+      "loss": 3.3733,
+      "step": 3090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.95202722896041e-05,
+      "loss": 3.4415,
+      "step": 3095
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.951867176761623e-05,
+      "loss": 3.4449,
+      "step": 3100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.951706859306454e-05,
+      "loss": 3.5253,
+      "step": 3105
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.951546276603494e-05,
+      "loss": 3.4753,
+      "step": 3110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.951385428661344e-05,
+      "loss": 3.4426,
+      "step": 3115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.951224315488619e-05,
+      "loss": 3.2768,
+      "step": 3120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.951062937093951e-05,
+      "loss": 3.4269,
+      "step": 3125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950901293485982e-05,
+      "loss": 3.4409,
+      "step": 3130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950739384673372e-05,
+      "loss": 3.493,
+      "step": 3135
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950577210664795e-05,
+      "loss": 3.4889,
+      "step": 3140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950414771468937e-05,
+      "loss": 3.4767,
+      "step": 3145
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9502520670945e-05,
+      "loss": 3.2876,
+      "step": 3150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.950089097550199e-05,
+      "loss": 3.4106,
+      "step": 3155
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949925862844763e-05,
+      "loss": 3.4185,
+      "step": 3160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949762362986938e-05,
+      "loss": 3.4577,
+      "step": 3165
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94959859798548e-05,
+      "loss": 3.5254,
+      "step": 3170
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949434567849166e-05,
+      "loss": 3.4504,
+      "step": 3175
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949270272586777e-05,
+      "loss": 3.4066,
+      "step": 3180
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.949105712207117e-05,
+      "loss": 3.3599,
+      "step": 3185
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948940886718998e-05,
+      "loss": 3.3692,
+      "step": 3190
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948775796131251e-05,
+      "loss": 3.4562,
+      "step": 3195
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948610440452721e-05,
+      "loss": 3.4946,
+      "step": 3200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948444819692265e-05,
+      "loss": 3.4708,
+      "step": 3205
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948278933858753e-05,
+      "loss": 3.4336,
+      "step": 3210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.948112782961073e-05,
+      "loss": 3.455,
+      "step": 3215
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947946367008121e-05,
+      "loss": 3.4308,
+      "step": 3220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947779686008817e-05,
+      "loss": 3.4366,
+      "step": 3225
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947612739972087e-05,
+      "loss": 3.4058,
+      "step": 3230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947445528906874e-05,
+      "loss": 3.5143,
+      "step": 3235
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.947278052822133e-05,
+      "loss": 3.314,
+      "step": 3240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94711031172684e-05,
+      "loss": 3.5274,
+      "step": 3245
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946942305629976e-05,
+      "loss": 3.4424,
+      "step": 3250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946774034540543e-05,
+      "loss": 3.4843,
+      "step": 3255
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946605498467552e-05,
+      "loss": 3.474,
+      "step": 3260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946436697420034e-05,
+      "loss": 3.3482,
+      "step": 3265
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946267631407029e-05,
+      "loss": 3.3296,
+      "step": 3270
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.946098300437596e-05,
+      "loss": 3.5068,
+      "step": 3275
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945928704520802e-05,
+      "loss": 3.4767,
+      "step": 3280
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945758843665736e-05,
+      "loss": 3.4456,
+      "step": 3285
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945588717881493e-05,
+      "loss": 3.3508,
+      "step": 3290
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945418327177188e-05,
+      "loss": 3.431,
+      "step": 3295
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945247671561946e-05,
+      "loss": 3.3876,
+      "step": 3300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.945076751044912e-05,
+      "loss": 3.4163,
+      "step": 3305
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94490556563524e-05,
+      "loss": 3.406,
+      "step": 3310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944734115342101e-05,
+      "loss": 3.533,
+      "step": 3315
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944562400174675e-05,
+      "loss": 3.3778,
+      "step": 3320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944390420142167e-05,
+      "loss": 3.3114,
+      "step": 3325
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944218175253784e-05,
+      "loss": 3.2862,
+      "step": 3330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.944045665518752e-05,
+      "loss": 3.4588,
+      "step": 3335
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943872890946319e-05,
+      "loss": 3.5009,
+      "step": 3340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943699851545733e-05,
+      "loss": 3.4731,
+      "step": 3345
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943526547326265e-05,
+      "loss": 3.5258,
+      "step": 3350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943352978297199e-05,
+      "loss": 3.4813,
+      "step": 3355
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943179144467833e-05,
+      "loss": 3.2838,
+      "step": 3360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.943005045847476e-05,
+      "loss": 3.4235,
+      "step": 3365
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942830682445458e-05,
+      "loss": 3.4691,
+      "step": 3370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942656054271117e-05,
+      "loss": 3.4248,
+      "step": 3375
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942481161333808e-05,
+      "loss": 3.3871,
+      "step": 3380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.942306003642897e-05,
+      "loss": 3.5433,
+      "step": 3385
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.94213058120777e-05,
+      "loss": 3.429,
+      "step": 3390
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941954894037821e-05,
+      "loss": 3.4903,
+      "step": 3395
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941778942142463e-05,
+      "loss": 3.441,
+      "step": 3400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941602725531123e-05,
+      "loss": 3.5273,
+      "step": 3405
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941426244213235e-05,
+      "loss": 3.5748,
+      "step": 3410
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941249498198257e-05,
+      "loss": 3.5508,
+      "step": 3415
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.941072487495655e-05,
+      "loss": 3.4156,
+      "step": 3420
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940895212114912e-05,
+      "loss": 3.4585,
+      "step": 3425
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940717672065523e-05,
+      "loss": 3.4368,
+      "step": 3430
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940539867356997e-05,
+      "loss": 3.4622,
+      "step": 3435
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940361797998862e-05,
+      "loss": 3.382,
+      "step": 3440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940183464000654e-05,
+      "loss": 3.3797,
+      "step": 3445
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.940004865371927e-05,
+      "loss": 3.4458,
+      "step": 3450
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939826002122246e-05,
+      "loss": 3.2774,
+      "step": 3455
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939646874261195e-05,
+      "loss": 3.4156,
+      "step": 3460
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939467481798368e-05,
+      "loss": 3.3665,
+      "step": 3465
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939287824743375e-05,
+      "loss": 3.3799,
+      "step": 3470
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.939107903105837e-05,
+      "loss": 3.4318,
+      "step": 3475
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938927716895396e-05,
+      "loss": 3.3655,
+      "step": 3480
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938747266121701e-05,
+      "loss": 3.5167,
+      "step": 3485
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938566550794422e-05,
+      "loss": 3.2889,
+      "step": 3490
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938385570923235e-05,
+      "loss": 3.4244,
+      "step": 3495
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938204326517835e-05,
+      "loss": 3.4022,
+      "step": 3500
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.938022817587931e-05,
+      "loss": 3.5012,
+      "step": 3505
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93784104414325e-05,
+      "loss": 3.3812,
+      "step": 3510
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937659006193523e-05,
+      "loss": 3.4138,
+      "step": 3515
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937476703748506e-05,
+      "loss": 3.4636,
+      "step": 3520
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937294136817961e-05,
+      "loss": 3.396,
+      "step": 3525
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.937111305411671e-05,
+      "loss": 3.5573,
+      "step": 3530
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936928209539425e-05,
+      "loss": 3.482,
+      "step": 3535
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936744849211035e-05,
+      "loss": 3.4767,
+      "step": 3540
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936561224436321e-05,
+      "loss": 3.3864,
+      "step": 3545
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93637733522512e-05,
+      "loss": 3.5068,
+      "step": 3550
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936193181587282e-05,
+      "loss": 3.2801,
+      "step": 3555
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.936008763532671e-05,
+      "loss": 3.5536,
+      "step": 3560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935824081071168e-05,
+      "loss": 3.4482,
+      "step": 3565
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935639134212665e-05,
+      "loss": 3.2857,
+      "step": 3570
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935453922967067e-05,
+      "loss": 3.3529,
+      "step": 3575
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.935268447344297e-05,
+      "loss": 3.3987,
+      "step": 3580
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93508270735429e-05,
+      "loss": 3.4436,
+      "step": 3585
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934896703006995e-05,
+      "loss": 3.4698,
+      "step": 3590
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934710434312376e-05,
+      "loss": 3.4337,
+      "step": 3595
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934523901280411e-05,
+      "loss": 3.413,
+      "step": 3600
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934337103921093e-05,
+      "loss": 3.4446,
+      "step": 3605
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.934150042244427e-05,
+      "loss": 3.4319,
+      "step": 3610
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933962716260433e-05,
+      "loss": 3.4009,
+      "step": 3615
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933775125979145e-05,
+      "loss": 3.4214,
+      "step": 3620
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933587271410614e-05,
+      "loss": 3.3533,
+      "step": 3625
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933399152564901e-05,
+      "loss": 3.3533,
+      "step": 3630
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933210769452084e-05,
+      "loss": 3.3774,
+      "step": 3635
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.933022122082252e-05,
+      "loss": 3.4434,
+      "step": 3640
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932833210465513e-05,
+      "loss": 3.2982,
+      "step": 3645
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932644034611984e-05,
+      "loss": 3.4446,
+      "step": 3650
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9324545945318e-05,
+      "loss": 3.324,
+      "step": 3655
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932264890235109e-05,
+      "loss": 3.3172,
+      "step": 3660
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.932074921732073e-05,
+      "loss": 3.3785,
+      "step": 3665
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931884689032866e-05,
+      "loss": 3.466,
+      "step": 3670
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.93169419214768e-05,
+      "loss": 3.4787,
+      "step": 3675
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931503431086718e-05,
+      "loss": 3.3268,
+      "step": 3680
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.9313124058602e-05,
+      "loss": 3.395,
+      "step": 3685
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.931121116478359e-05,
+      "loss": 3.3367,
+      "step": 3690
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930929562951439e-05,
+      "loss": 3.4003,
+      "step": 3695
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930737745289704e-05,
+      "loss": 3.2812,
+      "step": 3700
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930545663503428e-05,
+      "loss": 3.4256,
+      "step": 3705
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930353317602898e-05,
+      "loss": 3.459,
+      "step": 3710
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.930160707598421e-05,
+      "loss": 3.4366,
+      "step": 3715
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929967833500313e-05,
+      "loss": 3.3955,
+      "step": 3720
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929774695318905e-05,
+      "loss": 3.343,
+      "step": 3725
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929581293064545e-05,
+      "loss": 3.4552,
+      "step": 3730
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.929387626747591e-05,
+      "loss": 3.5061,
+      "step": 3735
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.929193696378416e-05,
+      "loss": 3.3023,
+      "step": 3740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92899950196741e-05,
+      "loss": 3.3614,
+      "step": 3745
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.928805043524976e-05,
+      "loss": 3.4335,
+      "step": 3750
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.928610321061529e-05,
+      "loss": 3.414,
+      "step": 3755
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.928415334587501e-05,
+      "loss": 3.4719,
+      "step": 3760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.928220084113336e-05,
+      "loss": 3.2881,
+      "step": 3765
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.928024569649493e-05,
+      "loss": 3.4115,
+      "step": 3770
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.927828791206446e-05,
+      "loss": 3.3499,
+      "step": 3775
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.927632748794681e-05,
+      "loss": 3.4183,
+      "step": 3780
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9274364424247e-05,
+      "loss": 3.305,
+      "step": 3785
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.927239872107018e-05,
+      "loss": 3.3067,
+      "step": 3790
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.927043037852166e-05,
+      "loss": 3.3557,
+      "step": 3795
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.926845939670686e-05,
+      "loss": 3.4884,
+      "step": 3800
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.926648577573139e-05,
+      "loss": 3.4721,
+      "step": 3805
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.926450951570092e-05,
+      "loss": 3.2948,
+      "step": 3810
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.926253061672138e-05,
+      "loss": 3.4563,
+      "step": 3815
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92605490788987e-05,
+      "loss": 3.3842,
+      "step": 3820
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925856490233906e-05,
+      "loss": 3.4406,
+      "step": 3825
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925657808714877e-05,
+      "loss": 3.4418,
+      "step": 3830
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925458863343423e-05,
+      "loss": 3.4609,
+      "step": 3835
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925259654130201e-05,
+      "loss": 3.4373,
+      "step": 3840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.925060181085882e-05,
+      "loss": 3.3349,
+      "step": 3845
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924860444221153e-05,
+      "loss": 3.3203,
+      "step": 3850
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92466044354671e-05,
+      "loss": 3.5919,
+      "step": 3855
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92446017907327e-05,
+      "loss": 3.4,
+      "step": 3860
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924259650811558e-05,
+      "loss": 3.3052,
+      "step": 3865
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.924058858772315e-05,
+      "loss": 3.2657,
+      "step": 3870
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9238578029663e-05,
+      "loss": 3.3802,
+      "step": 3875
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92365648340428e-05,
+      "loss": 3.3185,
+      "step": 3880
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923454900097042e-05,
+      "loss": 3.4221,
+      "step": 3885
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923253053055382e-05,
+      "loss": 3.3723,
+      "step": 3890
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.923050942290113e-05,
+      "loss": 3.4131,
+      "step": 3895
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92284856781206e-05,
+      "loss": 3.4617,
+      "step": 3900
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922645929632066e-05,
+      "loss": 3.2905,
+      "step": 3905
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922443027760983e-05,
+      "loss": 3.4057,
+      "step": 3910
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922239862209684e-05,
+      "loss": 3.3285,
+      "step": 3915
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.922036432989048e-05,
+      "loss": 3.2393,
+      "step": 3920
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921832740109975e-05,
+      "loss": 3.5533,
+      "step": 3925
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921628783583372e-05,
+      "loss": 3.4596,
+      "step": 3930
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.92142456342017e-05,
+      "loss": 3.4591,
+      "step": 3935
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921220079631303e-05,
+      "loss": 3.3071,
+      "step": 3940
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.921015332227727e-05,
+      "loss": 3.3045,
+      "step": 3945
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920810321220411e-05,
+      "loss": 3.298,
+      "step": 3950
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920605046620337e-05,
+      "loss": 3.4355,
+      "step": 3955
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920399508438497e-05,
+      "loss": 3.4139,
+      "step": 3960
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.920193706685905e-05,
+      "loss": 3.3744,
+      "step": 3965
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919987641373582e-05,
+      "loss": 3.2618,
+      "step": 3970
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919781312512568e-05,
+      "loss": 3.3576,
+      "step": 3975
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919574720113918e-05,
+      "loss": 3.3915,
+      "step": 3980
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919367864188693e-05,
+      "loss": 3.4337,
+      "step": 3985
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.919160744747979e-05,
+      "loss": 3.293,
+      "step": 3990
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918953361802866e-05,
+      "loss": 3.3483,
+      "step": 3995
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918745715364466e-05,
+      "loss": 3.4237,
+      "step": 4000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918537805443901e-05,
+      "loss": 3.4919,
+      "step": 4005
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.918329632052311e-05,
+      "loss": 3.5303,
+      "step": 4010
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91812119520084e-05,
+      "loss": 3.3598,
+      "step": 4015
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91791249490066e-05,
+      "loss": 3.2344,
+      "step": 4020
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917703531162949e-05,
+      "loss": 3.463,
+      "step": 4025
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917494303998899e-05,
+      "loss": 3.4476,
+      "step": 4030
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917284813419718e-05,
+      "loss": 3.3511,
+      "step": 4035
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.917075059436628e-05,
+      "loss": 3.363,
+      "step": 4040
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916865042060867e-05,
+      "loss": 3.2727,
+      "step": 4045
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91665476130368e-05,
+      "loss": 3.4005,
+      "step": 4050
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916444217176337e-05,
+      "loss": 3.4734,
+      "step": 4055
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916233409690112e-05,
+      "loss": 3.4809,
+      "step": 4060
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.916022338856298e-05,
+      "loss": 3.3954,
+      "step": 4065
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915811004686205e-05,
+      "loss": 3.3517,
+      "step": 4070
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915599407191148e-05,
+      "loss": 3.2302,
+      "step": 4075
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915387546382465e-05,
+      "loss": 3.3732,
+      "step": 4080
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.915175422271504e-05,
+      "loss": 3.4876,
+      "step": 4085
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914963034869629e-05,
+      "loss": 3.3918,
+      "step": 4090
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914750384188213e-05,
+      "loss": 3.3412,
+      "step": 4095
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914537470238652e-05,
+      "loss": 3.4592,
+      "step": 4100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914324293032349e-05,
+      "loss": 3.1623,
+      "step": 4105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.914110852580723e-05,
+      "loss": 3.3962,
+      "step": 4110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913897148895207e-05,
+      "loss": 3.3617,
+      "step": 4115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91368318198725e-05,
+      "loss": 3.3529,
+      "step": 4120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91346895186831e-05,
+      "loss": 3.3813,
+      "step": 4125
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.913254458549867e-05,
+      "loss": 3.2716,
+      "step": 4130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91303970204341e-05,
+      "loss": 3.3777,
+      "step": 4135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912824682360442e-05,
+      "loss": 3.4082,
+      "step": 4140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912609399512479e-05,
+      "loss": 3.4043,
+      "step": 4145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912393853511056e-05,
+      "loss": 3.3647,
+      "step": 4150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.912178044367717e-05,
+      "loss": 3.2551,
+      "step": 4155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911961972094024e-05,
+      "loss": 3.3984,
+      "step": 4160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91174563670155e-05,
+      "loss": 3.3382,
+      "step": 4165
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911529038201886e-05,
+      "loss": 3.357,
+      "step": 4170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.91131217660663e-05,
+      "loss": 3.3887,
+      "step": 4175
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.911095051927401e-05,
+      "loss": 3.2626,
+      "step": 4180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910877664175831e-05,
+      "loss": 3.4158,
+      "step": 4185
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910660013363564e-05,
+      "loss": 3.3364,
+      "step": 4190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910442099502257e-05,
+      "loss": 3.4484,
+      "step": 4195
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910223922603586e-05,
+      "loss": 3.321,
+      "step": 4200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.910005482679234e-05,
+      "loss": 3.2402,
+      "step": 4205
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909786779740908e-05,
+      "loss": 3.3944,
+      "step": 4210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909567813800318e-05,
+      "loss": 3.4041,
+      "step": 4215
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909348584869196e-05,
+      "loss": 3.234,
+      "step": 4220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.909129092959284e-05,
+      "loss": 3.3064,
+      "step": 4225
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90890933808234e-05,
+      "loss": 3.2789,
+      "step": 4230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908689320250138e-05,
+      "loss": 3.3856,
+      "step": 4235
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908469039474459e-05,
+      "loss": 3.3865,
+      "step": 4240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908248495767106e-05,
+      "loss": 3.3233,
+      "step": 4245
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.908027689139892e-05,
+      "loss": 3.3631,
+      "step": 4250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907806619604646e-05,
+      "loss": 3.3774,
+      "step": 4255
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907585287173208e-05,
+      "loss": 3.3486,
+      "step": 4260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907363691857436e-05,
+      "loss": 3.5036,
+      "step": 4265
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.907141833669198e-05,
+      "loss": 3.4313,
+      "step": 4270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.906919712620381e-05,
+      "loss": 3.3474,
+      "step": 4275
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.906697328722881e-05,
+      "loss": 3.3061,
+      "step": 4280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.906474681988612e-05,
+      "loss": 3.3868,
+      "step": 4285
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.9062517724295e-05,
+      "loss": 3.3315,
+      "step": 4290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.906028600057486e-05,
+      "loss": 3.4266,
+      "step": 4295
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905805164884523e-05,
+      "loss": 3.3955,
+      "step": 4300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905581466922582e-05,
+      "loss": 3.3866,
+      "step": 4305
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905357506183644e-05,
+      "loss": 3.365,
+      "step": 4310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.905133282679706e-05,
+      "loss": 3.3911,
+      "step": 4315
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90490879642278e-05,
+      "loss": 3.2299,
+      "step": 4320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90468404742489e-05,
+      "loss": 3.4052,
+      "step": 4325
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904459035698077e-05,
+      "loss": 3.3204,
+      "step": 4330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904233761254391e-05,
+      "loss": 3.4726,
+      "step": 4335
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.904008224105903e-05,
+      "loss": 3.3115,
+      "step": 4340
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903782424264692e-05,
+      "loss": 3.3613,
+      "step": 4345
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903556361742854e-05,
+      "loss": 3.3341,
+      "step": 4350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903330036552499e-05,
+      "loss": 3.2214,
+      "step": 4355
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.903103448705748e-05,
+      "loss": 3.3062,
+      "step": 4360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902876598214741e-05,
+      "loss": 3.4319,
+      "step": 4365
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90264948509163e-05,
+      "loss": 3.3975,
+      "step": 4370
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.90242210934858e-05,
+      "loss": 3.3517,
+      "step": 4375
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.902194470997771e-05,
+      "loss": 3.3307,
+      "step": 4380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901966570051397e-05,
+      "loss": 3.3633,
+      "step": 4385
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901738406521665e-05,
+      "loss": 3.4175,
+      "step": 4390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901509980420798e-05,
+      "loss": 3.3983,
+      "step": 4395
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901281291761031e-05,
+      "loss": 3.3504,
+      "step": 4400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.901052340554615e-05,
+      "loss": 3.2634,
+      "step": 4405
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.900823126813814e-05,
+      "loss": 3.2829,
+      "step": 4410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.900593650550908e-05,
+      "loss": 3.3915,
+      "step": 4415
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.900363911778186e-05,
+      "loss": 3.2734,
+      "step": 4420
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.900133910507957e-05,
+      "loss": 3.3411,
+      "step": 4425
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89990364675254e-05,
+      "loss": 3.3714,
+      "step": 4430
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.899673120524272e-05,
+      "loss": 3.3767,
+      "step": 4435
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.899442331835497e-05,
+      "loss": 3.2648,
+      "step": 4440
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.899211280698583e-05,
+      "loss": 3.38,
+      "step": 4445
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.898979967125903e-05,
+      "loss": 3.3776,
+      "step": 4450
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.898748391129849e-05,
+      "loss": 3.3312,
+      "step": 4455
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.898516552722825e-05,
+      "loss": 3.3567,
+      "step": 4460
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89828445191725e-05,
+      "loss": 3.3419,
+      "step": 4465
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.898052088725557e-05,
+      "loss": 3.2322,
+      "step": 4470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.897819463160196e-05,
+      "loss": 3.4651,
+      "step": 4475
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.897586575233626e-05,
+      "loss": 3.337,
+      "step": 4480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.897353424958319e-05,
+      "loss": 3.3138,
+      "step": 4485
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.897120012346768e-05,
+      "loss": 3.3681,
+      "step": 4490
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896886337411473e-05,
+      "loss": 3.3878,
+      "step": 4495
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896652400164957e-05,
+      "loss": 3.3543,
+      "step": 4500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896418200619744e-05,
+      "loss": 3.3343,
+      "step": 4505
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.896183738788384e-05,
+      "loss": 3.4145,
+      "step": 4510
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895949014683436e-05,
+      "loss": 3.2833,
+      "step": 4515
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89571402831747e-05,
+      "loss": 3.4583,
+      "step": 4520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89547877970308e-05,
+      "loss": 3.4615,
+      "step": 4525
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895243268852861e-05,
+      "loss": 3.2551,
+      "step": 4530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.895007495779433e-05,
+      "loss": 3.452,
+      "step": 4535
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894771460495422e-05,
+      "loss": 3.4053,
+      "step": 4540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894535163013477e-05,
+      "loss": 3.3979,
+      "step": 4545
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89429860334625e-05,
+      "loss": 3.3345,
+      "step": 4550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.894061781506413e-05,
+      "loss": 3.3746,
+      "step": 4555
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.893824697506659e-05,
+      "loss": 3.3403,
+      "step": 4560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.89358735135968e-05,
+      "loss": 3.3938,
+      "step": 4565
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.893349743078193e-05,
+      "loss": 3.3538,
+      "step": 4570
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.893111872674926e-05,
+      "loss": 3.3469,
+      "step": 4575
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892873740162621e-05,
+      "loss": 3.419,
+      "step": 4580
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892635345554033e-05,
+      "loss": 3.3737,
+      "step": 4585
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892396688861934e-05,
+      "loss": 3.2992,
+      "step": 4590
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.892157770099106e-05,
+      "loss": 3.3269,
+      "step": 4595
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.891918589278347e-05,
+      "loss": 3.2386,
+      "step": 4600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.891679146412472e-05,
+      "loss": 3.3334,
+      "step": 4605
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.891439441514305e-05,
+      "loss": 3.3472,
+      "step": 4610
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.891199474596686e-05,
+      "loss": 3.3735,
+      "step": 4615
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890959245672471e-05,
+      "loss": 3.3808,
+      "step": 4620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890718754754526e-05,
+      "loss": 3.2001,
+      "step": 4625
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890478001855737e-05,
+      "loss": 3.4037,
+      "step": 4630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.890236986988996e-05,
+      "loss": 3.2019,
+      "step": 4635
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889995710167217e-05,
+      "loss": 3.303,
+      "step": 4640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889754171403322e-05,
+      "loss": 3.3192,
+      "step": 4645
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889512370710253e-05,
+      "loss": 3.3924,
+      "step": 4650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889270308100958e-05,
+      "loss": 3.3916,
+      "step": 4655
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.889027983588405e-05,
+      "loss": 3.3101,
+      "step": 4660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888785397185576e-05,
+      "loss": 3.3623,
+      "step": 4665
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888542548905463e-05,
+      "loss": 3.5145,
+      "step": 4670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888299438761079e-05,
+      "loss": 3.383,
+      "step": 4675
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.888056066765444e-05,
+      "loss": 3.2083,
+      "step": 4680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887812432931592e-05,
+      "loss": 3.3125,
+      "step": 4685
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88756853727258e-05,
+      "loss": 3.1498,
+      "step": 4690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887324379801465e-05,
+      "loss": 3.3812,
+      "step": 4695
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.887079960531334e-05,
+      "loss": 3.3256,
+      "step": 4700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886835279475275e-05,
+      "loss": 3.3613,
+      "step": 4705
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886590336646395e-05,
+      "loss": 3.2359,
+      "step": 4710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.886345132057815e-05,
+      "loss": 3.3136,
+      "step": 4715
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88609966572267e-05,
+      "loss": 3.4824,
+      "step": 4720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88585393765411e-05,
+      "loss": 3.4604,
+      "step": 4725
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885607947865298e-05,
+      "loss": 3.3308,
+      "step": 4730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885361696369408e-05,
+      "loss": 3.2742,
+      "step": 4735
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.885115183179634e-05,
+      "loss": 3.3338,
+      "step": 4740
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884868408309181e-05,
+      "loss": 3.3776,
+      "step": 4745
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884621371771266e-05,
+      "loss": 3.3428,
+      "step": 4750
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884374073579123e-05,
+      "loss": 3.3398,
+      "step": 4755
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.884126513745997e-05,
+      "loss": 3.4845,
+      "step": 4760
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.883878692285154e-05,
+      "loss": 3.3476,
+      "step": 4765
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.883630609209864e-05,
+      "loss": 3.3957,
+      "step": 4770
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.883382264533418e-05,
+      "loss": 3.4113,
+      "step": 4775
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.883133658269119e-05,
+      "loss": 3.289,
+      "step": 4780
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882884790430285e-05,
+      "loss": 3.2885,
+      "step": 4785
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882635661030246e-05,
+      "loss": 3.3649,
+      "step": 4790
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882386270082348e-05,
+      "loss": 3.2872,
+      "step": 4795
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.882136617599948e-05,
+      "loss": 3.3741,
+      "step": 4800
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.881886703596421e-05,
+      "loss": 3.3495,
+      "step": 4805
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.881636528085154e-05,
+      "loss": 3.2666,
+      "step": 4810
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.881386091079548e-05,
+      "loss": 3.5511,
+      "step": 4815
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.881135392593019e-05,
+      "loss": 3.3589,
+      "step": 4820
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880884432638992e-05,
+      "loss": 3.363,
+      "step": 4825
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880633211230916e-05,
+      "loss": 3.3065,
+      "step": 4830
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.880381728382244e-05,
+      "loss": 3.1909,
+      "step": 4835
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.88012998410645e-05,
+      "loss": 3.3959,
+      "step": 4840
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879877978417018e-05,
+      "loss": 3.3147,
+      "step": 4845
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879625711327444e-05,
+      "loss": 3.19,
+      "step": 4850
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.879373182851246e-05,
+      "loss": 3.3227,
+      "step": 4855
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87912039300195e-05,
+      "loss": 3.2557,
+      "step": 4860
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.878867341793096e-05,
+      "loss": 3.2014,
+      "step": 4865
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87861402923824e-05,
+      "loss": 3.3371,
+      "step": 4870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87836045535095e-05,
+      "loss": 3.3667,
+      "step": 4875
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.878106620144812e-05,
+      "loss": 3.4182,
+      "step": 4880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877852523633418e-05,
+      "loss": 3.2815,
+      "step": 4885
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877598165830384e-05,
+      "loss": 3.2704,
+      "step": 4890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877343546749333e-05,
+      "loss": 3.5058,
+      "step": 4895
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.877088666403907e-05,
+      "loss": 3.2637,
+      "step": 4900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876833524807755e-05,
+      "loss": 3.2947,
+      "step": 4905
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876578121974546e-05,
+      "loss": 3.2861,
+      "step": 4910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876322457917961e-05,
+      "loss": 3.2579,
+      "step": 4915
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.876066532651695e-05,
+      "loss": 3.2833,
+      "step": 4920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875810346189459e-05,
+      "loss": 3.3645,
+      "step": 4925
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875553898544974e-05,
+      "loss": 3.2267,
+      "step": 4930
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.875297189731978e-05,
+      "loss": 3.3948,
+      "step": 4935
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87504021976422e-05,
+      "loss": 3.3684,
+      "step": 4940
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87478298865547e-05,
+      "loss": 3.5025,
+      "step": 4945
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.874525496419501e-05,
+      "loss": 3.3599,
+      "step": 4950
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87426774307011e-05,
+      "loss": 3.4368,
+      "step": 4955
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.874009728621104e-05,
+      "loss": 3.3679,
+      "step": 4960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.873751453086301e-05,
+      "loss": 3.2992,
+      "step": 4965
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87349291647954e-05,
+      "loss": 3.2955,
+      "step": 4970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.873234118814668e-05,
+      "loss": 3.3344,
+      "step": 4975
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872975060105548e-05,
+      "loss": 3.2484,
+      "step": 4980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872715740366057e-05,
+      "loss": 3.275,
+      "step": 4985
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872456159610087e-05,
+      "loss": 3.2115,
+      "step": 4990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.872196317851541e-05,
+      "loss": 3.4357,
+      "step": 4995
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87193621510434e-05,
+      "loss": 3.3528,
+      "step": 5000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871675851382415e-05,
+      "loss": 3.4347,
+      "step": 5005
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871415226699714e-05,
+      "loss": 3.401,
+      "step": 5010
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.871154341070197e-05,
+      "loss": 3.3973,
+      "step": 5015
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.87089319450784e-05,
+      "loss": 3.3954,
+      "step": 5020
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870631787026632e-05,
+      "loss": 3.1666,
+      "step": 5025
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870370118640575e-05,
+      "loss": 3.3389,
+      "step": 5030
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.870108189363687e-05,
+      "loss": 3.3552,
+      "step": 5035
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869845999209996e-05,
+      "loss": 3.3327,
+      "step": 5040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869583548193548e-05,
+      "loss": 3.3247,
+      "step": 5045
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869320836328404e-05,
+      "loss": 3.2804,
+      "step": 5050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.869057863628635e-05,
+      "loss": 3.3718,
+      "step": 5055
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868794630108324e-05,
+      "loss": 3.2322,
+      "step": 5060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868531135781579e-05,
+      "loss": 3.4453,
+      "step": 5065
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868267380662508e-05,
+      "loss": 3.3247,
+      "step": 5070
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.868003364765244e-05,
+      "loss": 3.3043,
+      "step": 5075
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867739088103926e-05,
+      "loss": 3.3156,
+      "step": 5080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867474550692713e-05,
+      "loss": 3.2769,
+      "step": 5085
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.867209752545774e-05,
+      "loss": 3.4328,
+      "step": 5090
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.866944693677295e-05,
+      "loss": 3.3625,
+      "step": 5095
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.866679374101474e-05,
+      "loss": 3.3378,
+      "step": 5100
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.866413793832521e-05,
+      "loss": 3.1726,
+      "step": 5105
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.866147952884665e-05,
+      "loss": 3.2853,
+      "step": 5110
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.865881851272146e-05,
+      "loss": 3.3609,
+      "step": 5115
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.865615489009217e-05,
+      "loss": 3.2424,
+      "step": 5120
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.865348866110147e-05,
+      "loss": 3.2732,
+      "step": 5125
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.865081982589218e-05,
+      "loss": 3.2458,
+      "step": 5130
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.864814838460726e-05,
+      "loss": 3.414,
+      "step": 5135
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.864547433738982e-05,
+      "loss": 3.4375,
+      "step": 5140
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.864279768438309e-05,
+      "loss": 3.3272,
+      "step": 5145
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.864011842573045e-05,
+      "loss": 3.3272,
+      "step": 5150
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.863743656157542e-05,
+      "loss": 3.3442,
+      "step": 5155
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.863475209206167e-05,
+      "loss": 3.3205,
+      "step": 5160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.8632065017333e-05,
+      "loss": 3.3838,
+      "step": 5165
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862937533753333e-05,
+      "loss": 3.4572,
+      "step": 5170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862668305280675e-05,
+      "loss": 3.3705,
+      "step": 5175
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862398816329748e-05,
+      "loss": 3.4184,
+      "step": 5180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.862129066914986e-05,
+      "loss": 3.3498,
+      "step": 5185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.86185905705084e-05,
+      "loss": 3.3355,
+      "step": 5190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.861588786751774e-05,
+      "loss": 3.2361,
+      "step": 5195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.861318256032265e-05,
+      "loss": 3.3061,
+      "step": 5200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.861047464906805e-05,
+      "loss": 3.2658,
+      "step": 5205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.860776413389898e-05,
+      "loss": 3.3426,
+      "step": 5210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.860505101496064e-05,
+      "loss": 3.3834,
+      "step": 5215
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.860233529239838e-05,
+      "loss": 3.3098,
+      "step": 5220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859961696635766e-05,
+      "loss": 3.297,
+      "step": 5225
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859689603698409e-05,
+      "loss": 3.2037,
+      "step": 5230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859417250442342e-05,
+      "loss": 3.2092,
+      "step": 5235
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.859144636882155e-05,
+      "loss": 3.2145,
+      "step": 5240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858871763032451e-05,
+      "loss": 3.3515,
+      "step": 5245
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858598628907846e-05,
+      "loss": 3.3954,
+      "step": 5250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858325234522971e-05,
+      "loss": 3.2845,
+      "step": 5255
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.858051579892474e-05,
+      "loss": 3.523,
+      "step": 5260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.857777665031008e-05,
+      "loss": 3.3628,
+      "step": 5265
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.857503489953252e-05,
+      "loss": 3.474,
+      "step": 5270
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85722905467389e-05,
+      "loss": 3.3617,
+      "step": 5275
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85695435920762e-05,
+      "loss": 3.3669,
+      "step": 5280
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85667940356916e-05,
+      "loss": 3.3269,
+      "step": 5285
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.856404187773239e-05,
+      "loss": 3.3671,
+      "step": 5290
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.856128711834595e-05,
+      "loss": 3.2698,
+      "step": 5295
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85585297576799e-05,
+      "loss": 3.2562,
+      "step": 5300
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855576979588193e-05,
+      "loss": 3.4577,
+      "step": 5305
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855300723309986e-05,
+      "loss": 3.4169,
+      "step": 5310
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.855024206948167e-05,
+      "loss": 3.2697,
+      "step": 5315
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85474743051755e-05,
+      "loss": 3.3731,
+      "step": 5320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.854470394032962e-05,
+      "loss": 3.2302,
+      "step": 5325
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.854193097509242e-05,
+      "loss": 3.2399,
+      "step": 5330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.853915540961241e-05,
+      "loss": 3.2407,
+      "step": 5335
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.85363772440383e-05,
+      "loss": 3.2491,
+      "step": 5340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.853359647851892e-05,
+      "loss": 3.1775,
+      "step": 5345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.853081311320319e-05,
+      "loss": 3.4107,
+      "step": 5350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852802714824023e-05,
+      "loss": 3.2223,
+      "step": 5355
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852523858377926e-05,
+      "loss": 3.4253,
+      "step": 5360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.852244741996967e-05,
+      "loss": 3.3484,
+      "step": 5365
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.851965365696099e-05,
+      "loss": 3.3558,
+      "step": 5370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.851685729490284e-05,
+      "loss": 3.4056,
+      "step": 5375
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.8514058333945e-05,
+      "loss": 3.2742,
+      "step": 5380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.851125677423746e-05,
+      "loss": 3.3228,
+      "step": 5385
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850845261593025e-05,
+      "loss": 3.3968,
+      "step": 5390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850564585917358e-05,
+      "loss": 3.0962,
+      "step": 5395
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850283650411782e-05,
+      "loss": 3.3304,
+      "step": 5400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.850002455091342e-05,
+      "loss": 3.2595,
+      "step": 5405
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.849720999971106e-05,
+      "loss": 3.2959,
+      "step": 5410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.849439285066146e-05,
+      "loss": 3.2467,
+      "step": 5415
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.849157310391556e-05,
+      "loss": 3.3414,
+      "step": 5420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.84887507596244e-05,
+      "loss": 3.3417,
+      "step": 5425
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848592581793915e-05,
+      "loss": 3.3026,
+      "step": 5430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848309827901115e-05,
+      "loss": 3.3234,
+      "step": 5435
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.848026814299183e-05,
+      "loss": 3.3182,
+      "step": 5440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.847743541003283e-05,
+      "loss": 3.3553,
+      "step": 5445
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.847460008028586e-05,
+      "loss": 3.2451,
+      "step": 5450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.847176215390281e-05,
+      "loss": 3.4231,
+      "step": 5455
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846892163103572e-05,
+      "loss": 3.3787,
+      "step": 5460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846607851183674e-05,
+      "loss": 3.3636,
+      "step": 5465
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846323279645815e-05,
+      "loss": 3.3867,
+      "step": 5470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.846038448505239e-05,
+      "loss": 3.411,
+      "step": 5475
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.845753357777204e-05,
+      "loss": 3.3431,
+      "step": 5480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.845468007476982e-05,
+      "loss": 3.2351,
+      "step": 5485
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.845182397619857e-05,
+      "loss": 3.2709,
+      "step": 5490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.84489652822113e-05,
+      "loss": 3.385,
+      "step": 5495
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.844610399296115e-05,
+      "loss": 3.2979,
+      "step": 5500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.844324010860136e-05,
+      "loss": 3.2291,
+      "step": 5505
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.844037362928533e-05,
+      "loss": 3.3024,
+      "step": 5510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.843750455516664e-05,
+      "loss": 3.4462,
+      "step": 5515
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.8434632886399e-05,
+      "loss": 3.3663,
+      "step": 5520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.843175862313617e-05,
+      "loss": 3.4911,
+      "step": 5525
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842888176553217e-05,
+      "loss": 3.4183,
+      "step": 5530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842600231374107e-05,
+      "loss": 3.2331,
+      "step": 5535
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842312026791713e-05,
+      "loss": 3.2982,
+      "step": 5540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.842023562821476e-05,
+      "loss": 3.2442,
+      "step": 5545
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841734839478842e-05,
+      "loss": 3.3348,
+      "step": 5550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841445856779281e-05,
+      "loss": 3.197,
+      "step": 5555
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.841156614738272e-05,
+      "loss": 3.385,
+      "step": 5560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.840867113371311e-05,
+      "loss": 3.1596,
+      "step": 5565
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.840577352693901e-05,
+      "loss": 3.3825,
+      "step": 5570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.840287332721568e-05,
+      "loss": 3.2763,
+      "step": 5575
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.839997053469844e-05,
+      "loss": 3.2282,
+      "step": 5580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.839706514954284e-05,
+      "loss": 3.4097,
+      "step": 5585
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.839415717190443e-05,
+      "loss": 3.2052,
+      "step": 5590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.839124660193905e-05,
+      "loss": 3.3771,
+      "step": 5595
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838833343980258e-05,
+      "loss": 3.2393,
+      "step": 5600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838541768565107e-05,
+      "loss": 3.2544,
+      "step": 5605
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.838249933964072e-05,
+      "loss": 3.3124,
+      "step": 5610
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837957840192784e-05,
+      "loss": 3.1553,
+      "step": 5615
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837665487266892e-05,
+      "loss": 3.2402,
+      "step": 5620
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837372875202054e-05,
+      "loss": 3.3002,
+      "step": 5625
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.837080004013945e-05,
+      "loss": 3.1832,
+      "step": 5630
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836786873718255e-05,
+      "loss": 3.2039,
+      "step": 5635
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836493484330683e-05,
+      "loss": 3.2478,
+      "step": 5640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.836199835866949e-05,
+      "loss": 3.2549,
+      "step": 5645
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.835905928342777e-05,
+      "loss": 3.193,
+      "step": 5650
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.835611761773918e-05,
+      "loss": 3.3463,
+      "step": 5655
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.835317336176122e-05,
+      "loss": 3.2712,
+      "step": 5660
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.835022651565167e-05,
+      "loss": 3.2217,
+      "step": 5665
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.834727707956834e-05,
+      "loss": 3.4091,
+      "step": 5670
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.834432505366926e-05,
+      "loss": 3.274,
+      "step": 5675
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.834137043811252e-05,
+      "loss": 3.3663,
+      "step": 5680
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.833841323305641e-05,
+      "loss": 3.3541,
+      "step": 5685
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.833545343865935e-05,
+      "loss": 3.2671,
+      "step": 5690
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.833249105507987e-05,
+      "loss": 3.3894,
+      "step": 5695
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.832952608247665e-05,
+      "loss": 3.2517,
+      "step": 5700
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.832655852100854e-05,
+      "loss": 3.2508,
+      "step": 5705
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.832358837083449e-05,
+      "loss": 3.3446,
+      "step": 5710
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.832061563211359e-05,
+      "loss": 3.3419,
+      "step": 5715
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83176403050051e-05,
+      "loss": 3.2531,
+      "step": 5720
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.831466238966838e-05,
+      "loss": 3.1958,
+      "step": 5725
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.831168188626296e-05,
+      "loss": 3.3074,
+      "step": 5730
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.83086987949485e-05,
+      "loss": 3.1773,
+      "step": 5735
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.830571311588479e-05,
+      "loss": 3.3071,
+      "step": 5740
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.830272484923175e-05,
+      "loss": 3.2232,
+      "step": 5745
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.829973399514947e-05,
+      "loss": 3.2454,
+      "step": 5750
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.829674055379816e-05,
+      "loss": 3.2476,
+      "step": 5755
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.829374452533817e-05,
+      "loss": 3.4125,
+      "step": 5760
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.829074590992997e-05,
+      "loss": 3.3882,
+      "step": 5765
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.82877447077342e-05,
+      "loss": 3.2199,
+      "step": 5770
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.828474091891164e-05,
+      "loss": 3.3518,
+      "step": 5775
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.82817345436232e-05,
+      "loss": 3.3196,
+      "step": 5780
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.827872558202987e-05,
+      "loss": 3.3306,
+      "step": 5785
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.827571403429286e-05,
+      "loss": 3.2627,
+      "step": 5790
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.82726999005735e-05,
+      "loss": 3.4234,
+      "step": 5795
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.826968318103325e-05,
+      "loss": 3.4615,
+      "step": 5800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.826666387583368e-05,
+      "loss": 3.2848,
+      "step": 5805
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.826364198513655e-05,
+      "loss": 3.3712,
+      "step": 5810
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.826061750910373e-05,
+      "loss": 3.2936,
+      "step": 5815
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.825759044789722e-05,
+      "loss": 3.3516,
+      "step": 5820
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.825456080167918e-05,
+      "loss": 3.2884,
+      "step": 5825
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.82515285706119e-05,
+      "loss": 3.2724,
+      "step": 5830
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.82484937548578e-05,
+      "loss": 3.3803,
+      "step": 5835
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.824545635457944e-05,
+      "loss": 3.216,
+      "step": 5840
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.824241636993954e-05,
+      "loss": 3.3224,
+      "step": 5845
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.823937380110093e-05,
+      "loss": 3.3627,
+      "step": 5850
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.82363286482266e-05,
+      "loss": 3.2811,
+      "step": 5855
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.823328091147968e-05,
+      "loss": 3.3791,
+      "step": 5860
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.823023059102341e-05,
+      "loss": 3.4209,
+      "step": 5865
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.822717768702119e-05,
+      "loss": 3.3022,
+      "step": 5870
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.822412219963655e-05,
+      "loss": 3.2178,
+      "step": 5875
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.822106412903317e-05,
+      "loss": 3.3057,
+      "step": 5880
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.821800347537487e-05,
+      "loss": 3.2002,
+      "step": 5885
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.821494023882558e-05,
+      "loss": 3.18,
+      "step": 5890
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.82118744195494e-05,
+      "loss": 3.1029,
+      "step": 5895
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.820880601771055e-05,
+      "loss": 3.1768,
+      "step": 5900
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.820573503347342e-05,
+      "loss": 3.1996,
+      "step": 5905
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.820266146700246e-05,
+      "loss": 3.1868,
+      "step": 5910
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.819958531846236e-05,
+      "loss": 3.2009,
+      "step": 5915
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.819650658801789e-05,
+      "loss": 3.3274,
+      "step": 5920
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.819342527583397e-05,
+      "loss": 3.228,
+      "step": 5925
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.819034138207564e-05,
+      "loss": 3.4065,
+      "step": 5930
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.81872549069081e-05,
+      "loss": 3.2714,
+      "step": 5935
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.818416585049669e-05,
+      "loss": 3.3868,
+      "step": 5940
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.818107421300689e-05,
+      "loss": 3.3532,
+      "step": 5945
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.817797999460427e-05,
+      "loss": 3.2936,
+      "step": 5950
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.817488319545462e-05,
+      "loss": 3.3664,
+      "step": 5955
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.817178381572383e-05,
+      "loss": 3.2224,
+      "step": 5960
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.816868185557789e-05,
+      "loss": 3.2365,
+      "step": 5965
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.816557731518298e-05,
+      "loss": 3.3029,
+      "step": 5970
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.81624701947054e-05,
+      "loss": 3.268,
+      "step": 5975
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.815936049431159e-05,
+      "loss": 3.3175,
+      "step": 5980
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.81562482141681e-05,
+      "loss": 3.2225,
+      "step": 5985
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.81531333544417e-05,
+      "loss": 2.9792,
+      "step": 5990
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.81500159152992e-05,
+      "loss": 3.2598,
+      "step": 5995
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.814689589690762e-05,
+      "loss": 3.3487,
+      "step": 6000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.814377329943407e-05,
+      "loss": 3.2593,
+      "step": 6005
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.814064812304582e-05,
+      "loss": 3.3482,
+      "step": 6010
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.813752036791029e-05,
+      "loss": 3.3064,
+      "step": 6015
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.8134390034195e-05,
+      "loss": 3.3931,
+      "step": 6020
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.813125712206766e-05,
+      "loss": 3.2551,
+      "step": 6025
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.812812163169607e-05,
+      "loss": 3.4208,
+      "step": 6030
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.812498356324819e-05,
+      "loss": 3.2581,
+      "step": 6035
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.812184291689212e-05,
+      "loss": 3.3868,
+      "step": 6040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.811869969279612e-05,
+      "loss": 3.2907,
+      "step": 6045
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.811555389112852e-05,
+      "loss": 3.4377,
+      "step": 6050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.811240551205786e-05,
+      "loss": 3.2765,
+      "step": 6055
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.810925455575277e-05,
+      "loss": 3.3077,
+      "step": 6060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.810610102238207e-05,
+      "loss": 3.3081,
+      "step": 6065
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.810294491211465e-05,
+      "loss": 3.342,
+      "step": 6070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.809978622511957e-05,
+      "loss": 3.4163,
+      "step": 6075
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.809662496156606e-05,
+      "loss": 3.3022,
+      "step": 6080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.809346112162345e-05,
+      "loss": 3.3029,
+      "step": 6085
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.809029470546121e-05,
+      "loss": 3.2348,
+      "step": 6090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.808712571324894e-05,
+      "loss": 3.214,
+      "step": 6095
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.808395414515644e-05,
+      "loss": 3.2771,
+      "step": 6100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.808078000135357e-05,
+      "loss": 3.3961,
+      "step": 6105
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.807760328201034e-05,
+      "loss": 3.2991,
+      "step": 6110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.807442398729696e-05,
+      "loss": 3.2146,
+      "step": 6115
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80712421173837e-05,
+      "loss": 3.3272,
+      "step": 6120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.806805767244102e-05,
+      "loss": 3.266,
+      "step": 6125
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.806487065263952e-05,
+      "loss": 3.2701,
+      "step": 6130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.806168105814988e-05,
+      "loss": 3.2993,
+      "step": 6135
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.805848888914297e-05,
+      "loss": 3.2467,
+      "step": 6140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.805529414578979e-05,
+      "loss": 3.1811,
+      "step": 6145
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.805209682826148e-05,
+      "loss": 3.1487,
+      "step": 6150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80488969367293e-05,
+      "loss": 3.3643,
+      "step": 6155
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.804569447136465e-05,
+      "loss": 3.2701,
+      "step": 6160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.804248943233909e-05,
+      "loss": 3.2543,
+      "step": 6165
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80392818198243e-05,
+      "loss": 3.4081,
+      "step": 6170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.803607163399211e-05,
+      "loss": 3.36,
+      "step": 6175
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.803285887501447e-05,
+      "loss": 3.3636,
+      "step": 6180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.802964354306348e-05,
+      "loss": 3.3841,
+      "step": 6185
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.802642563831137e-05,
+      "loss": 3.3954,
+      "step": 6190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.802320516093053e-05,
+      "loss": 3.2169,
+      "step": 6195
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.801998211109344e-05,
+      "loss": 3.3083,
+      "step": 6200
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80167564889728e-05,
+      "loss": 3.2891,
+      "step": 6205
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.801352829474135e-05,
+      "loss": 3.2634,
+      "step": 6210
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.801029752857203e-05,
+      "loss": 3.2511,
+      "step": 6215
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.80070641906379e-05,
+      "loss": 3.4696,
+      "step": 6220
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.800382828111216e-05,
+      "loss": 3.3448,
+      "step": 6225
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.800058980016815e-05,
+      "loss": 3.2583,
+      "step": 6230
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.799734874797935e-05,
+      "loss": 3.3676,
+      "step": 6235
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.799410512471938e-05,
+      "loss": 3.2892,
+      "step": 6240
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.799085893056197e-05,
+      "loss": 3.2776,
+      "step": 6245
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.798761016568101e-05,
+      "loss": 3.2854,
+      "step": 6250
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.798435883025055e-05,
+      "loss": 3.2523,
+      "step": 6255
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.798110492444474e-05,
+      "loss": 3.2694,
+      "step": 6260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.797784844843786e-05,
+      "loss": 3.3979,
+      "step": 6265
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79745894024044e-05,
+      "loss": 3.3332,
+      "step": 6270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.797132778651889e-05,
+      "loss": 3.2527,
+      "step": 6275
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.796806360095607e-05,
+      "loss": 3.204,
+      "step": 6280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.796479684589077e-05,
+      "loss": 3.0896,
+      "step": 6285
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.796152752149801e-05,
+      "loss": 3.1498,
+      "step": 6290
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79582556279529e-05,
+      "loss": 3.4112,
+      "step": 6295
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79549811654307e-05,
+      "loss": 3.2251,
+      "step": 6300
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.795170413410684e-05,
+      "loss": 3.3795,
+      "step": 6305
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.794842453415683e-05,
+      "loss": 3.1733,
+      "step": 6310
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.794514236575635e-05,
+      "loss": 3.3062,
+      "step": 6315
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.794185762908124e-05,
+      "loss": 3.1747,
+      "step": 6320
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.793857032430742e-05,
+      "loss": 3.1752,
+      "step": 6325
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.793528045161102e-05,
+      "loss": 3.2879,
+      "step": 6330
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.793198801116824e-05,
+      "loss": 3.379,
+      "step": 6335
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.792869300315546e-05,
+      "loss": 3.4034,
+      "step": 6340
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.792539542774916e-05,
+      "loss": 3.3353,
+      "step": 6345
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.792209528512601e-05,
+      "loss": 3.2603,
+      "step": 6350
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.791879257546279e-05,
+      "loss": 3.2391,
+      "step": 6355
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.791548729893638e-05,
+      "loss": 3.2733,
+      "step": 6360
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.791217945572386e-05,
+      "loss": 3.2233,
+      "step": 6365
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.790886904600242e-05,
+      "loss": 3.2681,
+      "step": 6370
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.79055560699494e-05,
+      "loss": 3.2534,
+      "step": 6375
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.790224052774224e-05,
+      "loss": 3.1874,
+      "step": 6380
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.789892241955856e-05,
+      "loss": 3.1782,
+      "step": 6385
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.789560174557611e-05,
+      "loss": 3.3792,
+      "step": 6390
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.789227850597274e-05,
+      "loss": 3.2174,
+      "step": 6395
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.788895270092646e-05,
+      "loss": 3.2504,
+      "step": 6400
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.788562433061547e-05,
+      "loss": 3.252,
+      "step": 6405
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.788229339521803e-05,
+      "loss": 3.3025,
+      "step": 6410
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.787895989491257e-05,
+      "loss": 3.263,
+      "step": 6415
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.787562382987766e-05,
+      "loss": 3.2085,
+      "step": 6420
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.7872285200292e-05,
+      "loss": 3.2059,
+      "step": 6425
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.786894400633443e-05,
+      "loss": 3.2466,
+      "step": 6430
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.786560024818393e-05,
+      "loss": 3.1523,
+      "step": 6435
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.786225392601962e-05,
+      "loss": 3.2794,
+      "step": 6440
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.785890504002073e-05,
+      "loss": 3.4396,
+      "step": 6445
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.785555359036668e-05,
+      "loss": 3.3074,
+      "step": 6450
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.785219957723696e-05,
+      "loss": 3.2025,
+      "step": 6455
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.784884300081128e-05,
+      "loss": 3.377,
+      "step": 6460
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.784548386126942e-05,
+      "loss": 3.3473,
+      "step": 6465
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.784212215879129e-05,
+      "loss": 3.1966,
+      "step": 6470
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.7838757893557e-05,
+      "loss": 3.4038,
+      "step": 6475
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.783539106574679e-05,
+      "loss": 3.3406,
+      "step": 6480
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.783202167554095e-05,
+      "loss": 3.2793,
+      "step": 6485
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.782864972311999e-05,
+      "loss": 3.4175,
+      "step": 6490
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.782527520866455e-05,
+      "loss": 3.2029,
+      "step": 6495
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.782189813235538e-05,
+      "loss": 3.3178,
+      "step": 6500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.781851849437338e-05,
+      "loss": 3.3465,
+      "step": 6505
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.781513629489959e-05,
+      "loss": 3.2722,
+      "step": 6510
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.781175153411518e-05,
+      "loss": 3.2393,
+      "step": 6515
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.780836421220148e-05,
+      "loss": 3.3001,
+      "step": 6520
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.78049743293399e-05,
+      "loss": 3.2075,
+      "step": 6525
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.780158188571207e-05,
+      "loss": 3.3853,
+      "step": 6530
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.779818688149967e-05,
+      "loss": 3.2698,
+      "step": 6535
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.779478931688459e-05,
+      "loss": 3.2239,
+      "step": 6540
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.779138919204883e-05,
+      "loss": 3.2906,
+      "step": 6545
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.778798650717448e-05,
+      "loss": 3.1925,
+      "step": 6550
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.778458126244387e-05,
+      "loss": 3.2603,
+      "step": 6555
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.778117345803939e-05,
+      "loss": 3.3053,
+      "step": 6560
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.777776309414357e-05,
+      "loss": 3.2803,
+      "step": 6565
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77743501709391e-05,
+      "loss": 3.2304,
+      "step": 6570
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77709346886088e-05,
+      "loss": 3.2461,
+      "step": 6575
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.776751664733562e-05,
+      "loss": 3.3655,
+      "step": 6580
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.776409604730267e-05,
+      "loss": 3.2841,
+      "step": 6585
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.776067288869317e-05,
+      "loss": 3.2596,
+      "step": 6590
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77572471716905e-05,
+      "loss": 3.3748,
+      "step": 6595
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.775381889647814e-05,
+      "loss": 3.3131,
+      "step": 6600
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.775038806323978e-05,
+      "loss": 3.2583,
+      "step": 6605
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.774695467215912e-05,
+      "loss": 3.2763,
+      "step": 6610
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.774351872342016e-05,
+      "loss": 3.162,
+      "step": 6615
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.774008021720691e-05,
+      "loss": 3.388,
+      "step": 6620
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.773663915370357e-05,
+      "loss": 3.2887,
+      "step": 6625
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.773319553309446e-05,
+      "loss": 3.3739,
+      "step": 6630
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.772974935556404e-05,
+      "loss": 3.2807,
+      "step": 6635
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.772630062129693e-05,
+      "loss": 3.3513,
+      "step": 6640
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.772284933047787e-05,
+      "loss": 3.378,
+      "step": 6645
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77193954832917e-05,
+      "loss": 3.2168,
+      "step": 6650
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.771593907992346e-05,
+      "loss": 3.2105,
+      "step": 6655
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.77124801205583e-05,
+      "loss": 3.317,
+      "step": 6660
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.770901860538151e-05,
+      "loss": 3.2669,
+      "step": 6665
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.770555453457848e-05,
+      "loss": 3.243,
+      "step": 6670
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.770208790833482e-05,
+      "loss": 3.1562,
+      "step": 6675
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.769861872683618e-05,
+      "loss": 3.2308,
+      "step": 6680
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.769514699026842e-05,
+      "loss": 3.35,
+      "step": 6685
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.769167269881749e-05,
+      "loss": 3.2599,
+      "step": 6690
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.768819585266952e-05,
+      "loss": 3.3793,
+      "step": 6695
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.768471645201076e-05,
+      "loss": 3.264,
+      "step": 6700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.768123449702755e-05,
+      "loss": 3.3279,
+      "step": 6705
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.767774998790645e-05,
+      "loss": 3.1656,
+      "step": 6710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.767426292483409e-05,
+      "loss": 3.2608,
+      "step": 6715
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.767077330799727e-05,
+      "loss": 3.2897,
+      "step": 6720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.766728113758293e-05,
+      "loss": 3.258,
+      "step": 6725
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76637864137781e-05,
+      "loss": 3.2447,
+      "step": 6730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.766028913677001e-05,
+      "loss": 3.261,
+      "step": 6735
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.7656789306746e-05,
+      "loss": 3.2445,
+      "step": 6740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.765328692389353e-05,
+      "loss": 3.2369,
+      "step": 6745
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.764978198840021e-05,
+      "loss": 3.2379,
+      "step": 6750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76462745004538e-05,
+      "loss": 3.3056,
+      "step": 6755
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76427644602422e-05,
+      "loss": 3.3424,
+      "step": 6760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.763925186795339e-05,
+      "loss": 3.2702,
+      "step": 6765
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.763573672377557e-05,
+      "loss": 3.2821,
+      "step": 6770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.763221902789703e-05,
+      "loss": 3.2842,
+      "step": 6775
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.762869878050618e-05,
+      "loss": 3.2085,
+      "step": 6780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76251759817916e-05,
+      "loss": 3.3365,
+      "step": 6785
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.762165063194199e-05,
+      "loss": 3.2566,
+      "step": 6790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.761812273114621e-05,
+      "loss": 3.1512,
+      "step": 6795
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.761459227959323e-05,
+      "loss": 3.2496,
+      "step": 6800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.761105927747214e-05,
+      "loss": 3.2834,
+      "step": 6805
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.760752372497223e-05,
+      "loss": 3.3332,
+      "step": 6810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.760398562228287e-05,
+      "loss": 3.1329,
+      "step": 6815
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.76004449695936e-05,
+      "loss": 3.2076,
+      "step": 6820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.759690176709405e-05,
+      "loss": 3.3671,
+      "step": 6825
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.759335601497405e-05,
+      "loss": 3.1451,
+      "step": 6830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.758980771342352e-05,
+      "loss": 3.269,
+      "step": 6835
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.758625686263255e-05,
+      "loss": 3.1334,
+      "step": 6840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.758270346279133e-05,
+      "loss": 3.2098,
+      "step": 6845
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75791475140902e-05,
+      "loss": 3.3745,
+      "step": 6850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.757558901671965e-05,
+      "loss": 3.2332,
+      "step": 6855
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.757202797087031e-05,
+      "loss": 3.2799,
+      "step": 6860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.756846437673291e-05,
+      "loss": 3.2991,
+      "step": 6865
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.756489823449835e-05,
+      "loss": 3.3347,
+      "step": 6870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.756132954435767e-05,
+      "loss": 3.281,
+      "step": 6875
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.755775830650202e-05,
+      "loss": 3.205,
+      "step": 6880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.75541845211227e-05,
+      "loss": 3.196,
+      "step": 6885
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.755060818841117e-05,
+      "loss": 3.2814,
+      "step": 6890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.754702930855896e-05,
+      "loss": 3.2376,
+      "step": 6895
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.754344788175783e-05,
+      "loss": 3.1394,
+      "step": 6900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.753986390819959e-05,
+      "loss": 3.2659,
+      "step": 6905
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.753627738807622e-05,
+      "loss": 3.3685,
+      "step": 6910
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.753268832157987e-05,
+      "loss": 3.1142,
+      "step": 6915
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.752909670890278e-05,
+      "loss": 3.2651,
+      "step": 6920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.752550255023732e-05,
+      "loss": 3.2975,
+      "step": 6925
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.752190584577605e-05,
+      "loss": 3.2748,
+      "step": 6930
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.751830659571161e-05,
+      "loss": 3.2998,
+      "step": 6935
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.751470480023682e-05,
+      "loss": 3.1289,
+      "step": 6940
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.751110045954461e-05,
+      "loss": 3.3877,
+      "step": 6945
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.750749357382805e-05,
+      "loss": 3.3537,
+      "step": 6950
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.750388414328035e-05,
+      "loss": 3.1941,
+      "step": 6955
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.750027216809488e-05,
+      "loss": 3.3269,
+      "step": 6960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.749665764846507e-05,
+      "loss": 3.2491,
+      "step": 6965
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.74930405845846e-05,
+      "loss": 3.2194,
+      "step": 6970
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.748942097664717e-05,
+      "loss": 3.2709,
+      "step": 6975
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.74857988248467e-05,
+      "loss": 3.2486,
+      "step": 6980
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.748217412937722e-05,
+      "loss": 3.1083,
+      "step": 6985
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.74785468904329e-05,
+      "loss": 3.2725,
+      "step": 6990
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.7474917108208e-05,
+      "loss": 3.1469,
+      "step": 6995
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.747128478289701e-05,
+      "loss": 3.4008,
+      "step": 7000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.746764991469446e-05,
+      "loss": 3.1304,
+      "step": 7005
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.746401250379509e-05,
+      "loss": 3.1908,
+      "step": 7010
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.746037255039372e-05,
+      "loss": 3.3044,
+      "step": 7015
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.745673005468534e-05,
+      "loss": 3.2627,
+      "step": 7020
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.745308501686508e-05,
+      "loss": 3.306,
+      "step": 7025
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.744943743712818e-05,
+      "loss": 3.2451,
+      "step": 7030
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.744578731567003e-05,
+      "loss": 3.3595,
+      "step": 7035
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.744213465268616e-05,
+      "loss": 3.3433,
+      "step": 7040
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.743847944837222e-05,
+      "loss": 3.2643,
+      "step": 7045
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.743482170292402e-05,
+      "loss": 3.1604,
+      "step": 7050
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.743116141653751e-05,
+      "loss": 3.1945,
+      "step": 7055
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.742749858940873e-05,
+      "loss": 3.1644,
+      "step": 7060
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.74238332217339e-05,
+      "loss": 3.1915,
+      "step": 7065
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.742016531370936e-05,
+      "loss": 3.1786,
+      "step": 7070
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.74164948655316e-05,
+      "loss": 3.2166,
+      "step": 7075
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.741282187739722e-05,
+      "loss": 3.3161,
+      "step": 7080
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.740914634950298e-05,
+      "loss": 3.267,
+      "step": 7085
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.740546828204576e-05,
+      "loss": 3.3271,
+      "step": 7090
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.740178767522257e-05,
+      "loss": 3.3565,
+      "step": 7095
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.73981045292306e-05,
+      "loss": 3.2495,
+      "step": 7100
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.739441884426713e-05,
+      "loss": 3.2728,
+      "step": 7105
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.739073062052959e-05,
+      "loss": 3.2828,
+      "step": 7110
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.738703985821553e-05,
+      "loss": 3.2346,
+      "step": 7115
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.738334655752268e-05,
+      "loss": 3.2525,
+      "step": 7120
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.737965071864888e-05,
+      "loss": 3.1563,
+      "step": 7125
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.737595234179209e-05,
+      "loss": 3.2015,
+      "step": 7130
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.737225142715042e-05,
+      "loss": 3.4297,
+      "step": 7135
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.736854797492212e-05,
+      "loss": 3.3028,
+      "step": 7140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.736484198530557e-05,
+      "loss": 3.2155,
+      "step": 7145
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.73611334584993e-05,
+      "loss": 3.1629,
+      "step": 7150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.735742239470195e-05,
+      "loss": 3.3837,
+      "step": 7155
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.735370879411233e-05,
+      "loss": 3.1388,
+      "step": 7160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.734999265692933e-05,
+      "loss": 3.3518,
+      "step": 7165
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.734627398335205e-05,
+      "loss": 3.2847,
+      "step": 7170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.734255277357967e-05,
+      "loss": 3.2905,
+      "step": 7175
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.733882902781152e-05,
+      "loss": 3.2435,
+      "step": 7180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.733510274624709e-05,
+      "loss": 3.3573,
+      "step": 7185
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.733137392908597e-05,
+      "loss": 3.2757,
+      "step": 7190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.73276425765279e-05,
+      "loss": 3.4034,
+      "step": 7195
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.732390868877276e-05,
+      "loss": 3.2189,
+      "step": 7200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.732017226602058e-05,
+      "loss": 3.2446,
+      "step": 7205
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.731643330847146e-05,
+      "loss": 3.2496,
+      "step": 7210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.731269181632574e-05,
+      "loss": 3.3255,
+      "step": 7215
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.73089477897838e-05,
+      "loss": 3.3638,
+      "step": 7220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.730520122904623e-05,
+      "loss": 3.0677,
+      "step": 7225
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.73014521343137e-05,
+      "loss": 3.3251,
+      "step": 7230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.729770050578703e-05,
+      "loss": 3.2238,
+      "step": 7235
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.729394634366721e-05,
+      "loss": 3.2912,
+      "step": 7240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.729018964815533e-05,
+      "loss": 3.1568,
+      "step": 7245
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72864304194526e-05,
+      "loss": 3.1953,
+      "step": 7250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.728266865776042e-05,
+      "loss": 3.2674,
+      "step": 7255
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72789043632803e-05,
+      "loss": 3.4111,
+      "step": 7260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.727513753621384e-05,
+      "loss": 3.1312,
+      "step": 7265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.727136817676285e-05,
+      "loss": 3.2487,
+      "step": 7270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.726759628512925e-05,
+      "loss": 3.3054,
+      "step": 7275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.726382186151508e-05,
+      "loss": 3.3169,
+      "step": 7280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.726004490612252e-05,
+      "loss": 3.251,
+      "step": 7285
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.725626541915388e-05,
+      "loss": 3.1748,
+      "step": 7290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.725248340081163e-05,
+      "loss": 3.2471,
+      "step": 7295
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.724869885129837e-05,
+      "loss": 3.2279,
+      "step": 7300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72449117708168e-05,
+      "loss": 3.3635,
+      "step": 7305
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72411221595698e-05,
+      "loss": 3.2402,
+      "step": 7310
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.723733001776037e-05,
+      "loss": 3.2872,
+      "step": 7315
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.723353534559164e-05,
+      "loss": 3.2142,
+      "step": 7320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.722973814326686e-05,
+      "loss": 3.2892,
+      "step": 7325
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.722593841098946e-05,
+      "loss": 3.3057,
+      "step": 7330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.722213614896298e-05,
+      "loss": 3.2027,
+      "step": 7335
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.721833135739107e-05,
+      "loss": 3.246,
+      "step": 7340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.721452403647758e-05,
+      "loss": 3.1727,
+      "step": 7345
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.721071418642643e-05,
+      "loss": 3.1492,
+      "step": 7350
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.72069018074417e-05,
+      "loss": 3.2242,
+      "step": 7355
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.720308689972762e-05,
+      "loss": 3.2429,
+      "step": 7360
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.719926946348853e-05,
+      "loss": 3.2524,
+      "step": 7365
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.719544949892893e-05,
+      "loss": 3.3384,
+      "step": 7370
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.719162700625343e-05,
+      "loss": 3.2465,
+      "step": 7375
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.718780198566682e-05,
+      "loss": 3.3088,
+      "step": 7380
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.718397443737395e-05,
+      "loss": 3.1997,
+      "step": 7385
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.718014436157989e-05,
+      "loss": 3.2445,
+      "step": 7390
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.717631175848978e-05,
+      "loss": 3.1735,
+      "step": 7395
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.717247662830893e-05,
+      "loss": 3.2078,
+      "step": 7400
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.716863897124278e-05,
+      "loss": 3.0723,
+      "step": 7405
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.716479878749689e-05,
+      "loss": 3.2583,
+      "step": 7410
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.716095607727698e-05,
+      "loss": 3.1436,
+      "step": 7415
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.715711084078889e-05,
+      "loss": 3.2478,
+      "step": 7420
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.715326307823859e-05,
+      "loss": 3.2616,
+      "step": 7425
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.714941278983221e-05,
+      "loss": 3.1916,
+      "step": 7430
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.714555997577597e-05,
+      "loss": 3.1828,
+      "step": 7435
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.714170463627626e-05,
+      "loss": 3.2572,
+      "step": 7440
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.713784677153963e-05,
+      "loss": 3.2618,
+      "step": 7445
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.71339863817727e-05,
+      "loss": 3.2733,
+      "step": 7450
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.713012346718227e-05,
+      "loss": 3.2509,
+      "step": 7455
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.712625802797525e-05,
+      "loss": 3.2529,
+      "step": 7460
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.712239006435875e-05,
+      "loss": 3.2256,
+      "step": 7465
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.71185195765399e-05,
+      "loss": 3.2847,
+      "step": 7470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.711464656472608e-05,
+      "loss": 3.1418,
+      "step": 7475
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.711077102912471e-05,
+      "loss": 3.2553,
+      "step": 7480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.710689296994345e-05,
+      "loss": 3.3022,
+      "step": 7485
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.710301238738998e-05,
+      "loss": 3.2269,
+      "step": 7490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70991292816722e-05,
+      "loss": 3.3335,
+      "step": 7495
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70952436529981e-05,
+      "loss": 3.204,
+      "step": 7500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.709135550157585e-05,
+      "loss": 3.0815,
+      "step": 7505
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.708746482761368e-05,
+      "loss": 3.1357,
+      "step": 7510
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.708357163132004e-05,
+      "loss": 3.3397,
+      "step": 7515
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.707967591290348e-05,
+      "loss": 3.1944,
+      "step": 7520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.707577767257265e-05,
+      "loss": 3.137,
+      "step": 7525
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.707187691053639e-05,
+      "loss": 3.186,
+      "step": 7530
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.706797362700363e-05,
+      "loss": 3.1766,
+      "step": 7535
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.706406782218347e-05,
+      "loss": 3.1286,
+      "step": 7540
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.706015949628515e-05,
+      "loss": 3.3041,
+      "step": 7545
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.705624864951801e-05,
+      "loss": 3.3509,
+      "step": 7550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.705233528209154e-05,
+      "loss": 3.3339,
+      "step": 7555
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.704841939421538e-05,
+      "loss": 3.3551,
+      "step": 7560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.704450098609927e-05,
+      "loss": 3.2253,
+      "step": 7565
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.704058005795312e-05,
+      "loss": 3.1635,
+      "step": 7570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.703665660998697e-05,
+      "loss": 3.148,
+      "step": 7575
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.703273064241097e-05,
+      "loss": 3.1226,
+      "step": 7580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.702880215543544e-05,
+      "loss": 3.2321,
+      "step": 7585
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70248711492708e-05,
+      "loss": 3.1307,
+      "step": 7590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.702093762412763e-05,
+      "loss": 3.1404,
+      "step": 7595
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.701700158021665e-05,
+      "loss": 3.2579,
+      "step": 7600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.701306301774867e-05,
+      "loss": 3.2686,
+      "step": 7605
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.70091219369347e-05,
+      "loss": 3.3773,
+      "step": 7610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.700517833798584e-05,
+      "loss": 3.3202,
+      "step": 7615
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.700123222111331e-05,
+      "loss": 3.2765,
+      "step": 7620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.699728358652853e-05,
+      "loss": 3.1592,
+      "step": 7625
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.699333243444301e-05,
+      "loss": 3.2185,
+      "step": 7630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.698937876506837e-05,
+      "loss": 3.2579,
+      "step": 7635
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.698542257861645e-05,
+      "loss": 3.2039,
+      "step": 7640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.698146387529912e-05,
+      "loss": 3.3157,
+      "step": 7645
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.697750265532846e-05,
+      "loss": 3.1584,
+      "step": 7650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.697353891891664e-05,
+      "loss": 3.3074,
+      "step": 7655
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.696957266627603e-05,
+      "loss": 3.2895,
+      "step": 7660
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.696560389761902e-05,
+      "loss": 3.3115,
+      "step": 7665
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.696163261315826e-05,
+      "loss": 3.3099,
+      "step": 7670
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.695765881310649e-05,
+      "loss": 3.2525,
+      "step": 7675
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.695368249767653e-05,
+      "loss": 3.3125,
+      "step": 7680
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.69497036670814e-05,
+      "loss": 3.2175,
+      "step": 7685
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.694572232153422e-05,
+      "loss": 3.1813,
+      "step": 7690
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.694173846124829e-05,
+      "loss": 3.215,
+      "step": 7695
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.693775208643699e-05,
+      "loss": 3.1403,
+      "step": 7700
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.693376319731386e-05,
+      "loss": 3.2198,
+      "step": 7705
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.692977179409258e-05,
+      "loss": 3.2749,
+      "step": 7710
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.692577787698693e-05,
+      "loss": 3.3919,
+      "step": 7715
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.69217814462109e-05,
+      "loss": 3.321,
+      "step": 7720
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.691778250197853e-05,
+      "loss": 3.1631,
+      "step": 7725
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.691378104450404e-05,
+      "loss": 3.1989,
+      "step": 7730
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.690977707400177e-05,
+      "loss": 3.1842,
+      "step": 7735
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.690577059068621e-05,
+      "loss": 3.1432,
+      "step": 7740
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.690176159477197e-05,
+      "loss": 3.3377,
+      "step": 7745
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.689775008647381e-05,
+      "loss": 3.2993,
+      "step": 7750
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.689373606600663e-05,
+      "loss": 3.1598,
+      "step": 7755
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.688971953358538e-05,
+      "loss": 3.1095,
+      "step": 7760
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.688570048942529e-05,
+      "loss": 3.3578,
+      "step": 7765
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.688167893374161e-05,
+      "loss": 3.2149,
+      "step": 7770
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.687765486674975e-05,
+      "loss": 3.2384,
+      "step": 7775
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.687362828866531e-05,
+      "loss": 3.2259,
+      "step": 7780
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.686959919970396e-05,
+      "loss": 3.1899,
+      "step": 7785
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.68655676000815e-05,
+      "loss": 3.0709,
+      "step": 7790
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.686153349001392e-05,
+      "loss": 3.3656,
+      "step": 7795
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.685749686971732e-05,
+      "loss": 3.2587,
+      "step": 7800
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.685345773940792e-05,
+      "loss": 3.1391,
+      "step": 7805
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.684941609930208e-05,
+      "loss": 3.0704,
+      "step": 7810
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.684537194961629e-05,
+      "loss": 3.2732,
+      "step": 7815
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.684132529056719e-05,
+      "loss": 3.3172,
+      "step": 7820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.683727612237157e-05,
+      "loss": 3.0803,
+      "step": 7825
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.68332244452463e-05,
+      "loss": 3.3653,
+      "step": 7830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.682917025940844e-05,
+      "loss": 3.2175,
+      "step": 7835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.682511356507512e-05,
+      "loss": 3.2022,
+      "step": 7840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.68210543624637e-05,
+      "loss": 3.2471,
+      "step": 7845
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.681699265179158e-05,
+      "loss": 3.27,
+      "step": 7850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.681292843327634e-05,
+      "loss": 3.1682,
+      "step": 7855
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.68088617071357e-05,
+      "loss": 3.29,
+      "step": 7860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.680479247358749e-05,
+      "loss": 3.1335,
+      "step": 7865
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.68007207328497e-05,
+      "loss": 3.1872,
+      "step": 7870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.679664648514043e-05,
+      "loss": 3.3978,
+      "step": 7875
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.679256973067792e-05,
+      "loss": 3.1959,
+      "step": 7880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.678849046968057e-05,
+      "loss": 3.2753,
+      "step": 7885
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.678440870236687e-05,
+      "loss": 3.2287,
+      "step": 7890
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.678032442895548e-05,
+      "loss": 3.2864,
+      "step": 7895
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.677623764966517e-05,
+      "loss": 3.2517,
+      "step": 7900
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.677214836471488e-05,
+      "loss": 3.1264,
+      "step": 7905
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.676805657432366e-05,
+      "loss": 3.2963,
+      "step": 7910
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.676396227871065e-05,
+      "loss": 3.1791,
+      "step": 7915
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.675986547809521e-05,
+      "loss": 3.233,
+      "step": 7920
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67557661726968e-05,
+      "loss": 3.191,
+      "step": 7925
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.675166436273497e-05,
+      "loss": 3.1826,
+      "step": 7930
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.674756004842948e-05,
+      "loss": 3.2482,
+      "step": 7935
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.674345323000015e-05,
+      "loss": 3.0719,
+      "step": 7940
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.673934390766699e-05,
+      "loss": 3.231,
+      "step": 7945
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.673523208165014e-05,
+      "loss": 3.1746,
+      "step": 7950
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.673111775216982e-05,
+      "loss": 3.3258,
+      "step": 7955
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.672700091944644e-05,
+      "loss": 3.2553,
+      "step": 7960
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.672288158370053e-05,
+      "loss": 3.3082,
+      "step": 7965
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.671875974515277e-05,
+      "loss": 3.2271,
+      "step": 7970
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.671463540402391e-05,
+      "loss": 3.1963,
+      "step": 7975
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67105085605349e-05,
+      "loss": 3.1497,
+      "step": 7980
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67063792149068e-05,
+      "loss": 3.1582,
+      "step": 7985
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.67022473673608e-05,
+      "loss": 3.1487,
+      "step": 7990
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.669811301811826e-05,
+      "loss": 3.2297,
+      "step": 7995
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.669397616740062e-05,
+      "loss": 3.1368,
+      "step": 8000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.668983681542949e-05,
+      "loss": 3.3086,
+      "step": 8005
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.668569496242658e-05,
+      "loss": 3.3245,
+      "step": 8010
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.668155060861378e-05,
+      "loss": 3.1546,
+      "step": 8015
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.667740375421306e-05,
+      "loss": 3.2925,
+      "step": 8020
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.667325439944659e-05,
+      "loss": 3.2639,
+      "step": 8025
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.666910254453663e-05,
+      "loss": 3.2269,
+      "step": 8030
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.666494818970558e-05,
+      "loss": 3.2357,
+      "step": 8035
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.666079133517596e-05,
+      "loss": 3.124,
+      "step": 8040
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.665663198117046e-05,
+      "loss": 3.2939,
+      "step": 8045
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.665247012791187e-05,
+      "loss": 3.2515,
+      "step": 8050
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.664830577562312e-05,
+      "loss": 3.2342,
+      "step": 8055
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.664413892452732e-05,
+      "loss": 3.1892,
+      "step": 8060
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.663996957484765e-05,
+      "loss": 3.102,
+      "step": 8065
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.663579772680743e-05,
+      "loss": 3.1225,
+      "step": 8070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.663162338063016e-05,
+      "loss": 3.3701,
+      "step": 8075
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.662744653653946e-05,
+      "loss": 3.2052,
+      "step": 8080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.662326719475902e-05,
+      "loss": 3.28,
+      "step": 8085
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.661908535551277e-05,
+      "loss": 3.1377,
+      "step": 8090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.661490101902468e-05,
+      "loss": 3.1592,
+      "step": 8095
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.661071418551892e-05,
+      "loss": 3.1826,
+      "step": 8100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.660652485521975e-05,
+      "loss": 3.1462,
+      "step": 8105
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.660233302835157e-05,
+      "loss": 3.2492,
+      "step": 8110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.659813870513894e-05,
+      "loss": 3.213,
+      "step": 8115
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.659394188580653e-05,
+      "loss": 3.1834,
+      "step": 8120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.658974257057916e-05,
+      "loss": 3.2429,
+      "step": 8125
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.658554075968175e-05,
+      "loss": 3.2367,
+      "step": 8130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.658133645333942e-05,
+      "loss": 3.1683,
+      "step": 8135
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.657712965177734e-05,
+      "loss": 3.1664,
+      "step": 8140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.657292035522088e-05,
+      "loss": 3.1645,
+      "step": 8145
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.65687085638955e-05,
+      "loss": 3.213,
+      "step": 8150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.656449427802683e-05,
+      "loss": 3.2297,
+      "step": 8155
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.656027749784062e-05,
+      "loss": 3.3124,
+      "step": 8160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.655605822356273e-05,
+      "loss": 3.1798,
+      "step": 8165
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.65518364554192e-05,
+      "loss": 3.2579,
+      "step": 8170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.654761219363615e-05,
+      "loss": 3.2958,
+      "step": 8175
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.654338543843987e-05,
+      "loss": 3.2147,
+      "step": 8180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.653915619005678e-05,
+      "loss": 3.1763,
+      "step": 8185
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.653492444871343e-05,
+      "loss": 3.3526,
+      "step": 8190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.65306902146365e-05,
+      "loss": 3.1872,
+      "step": 8195
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.652645348805278e-05,
+      "loss": 3.2188,
+      "step": 8200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.652221426918926e-05,
+      "loss": 3.2905,
+      "step": 8205
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.6517972558273e-05,
+      "loss": 3.2597,
+      "step": 8210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.65137283555312e-05,
+      "loss": 3.1922,
+      "step": 8215
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.650948166119124e-05,
+      "loss": 3.2745,
+      "step": 8220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.650523247548058e-05,
+      "loss": 3.2563,
+      "step": 8225
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.650098079862686e-05,
+      "loss": 3.3716,
+      "step": 8230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.64967266308578e-05,
+      "loss": 3.3042,
+      "step": 8235
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.649246997240131e-05,
+      "loss": 3.2726,
+      "step": 8240
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.648821082348538e-05,
+      "loss": 3.1934,
+      "step": 8245
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.648394918433819e-05,
+      "loss": 3.2696,
+      "step": 8250
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.647968505518799e-05,
+      "loss": 3.2345,
+      "step": 8255
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.64754184362632e-05,
+      "loss": 3.1941,
+      "step": 8260
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.647114932779239e-05,
+      "loss": 3.4017,
+      "step": 8265
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.646687773000425e-05,
+      "loss": 3.2306,
+      "step": 8270
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.646260364312758e-05,
+      "loss": 3.254,
+      "step": 8275
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.645832706739131e-05,
+      "loss": 3.1357,
+      "step": 8280
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.645404800302455e-05,
+      "loss": 3.0908,
+      "step": 8285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.644976645025653e-05,
+      "loss": 3.2536,
+      "step": 8290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.644548240931657e-05,
+      "loss": 3.2097,
+      "step": 8295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.644119588043414e-05,
+      "loss": 3.2357,
+      "step": 8300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.64369068638389e-05,
+      "loss": 3.2573,
+      "step": 8305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.643261535976058e-05,
+      "loss": 3.2766,
+      "step": 8310
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.642832136842905e-05,
+      "loss": 3.201,
+      "step": 8315
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.642402489007435e-05,
+      "loss": 3.2111,
+      "step": 8320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.64197259249266e-05,
+      "loss": 3.2677,
+      "step": 8325
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.641542447321612e-05,
+      "loss": 3.2044,
+      "step": 8330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.64111205351733e-05,
+      "loss": 3.1524,
+      "step": 8335
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.640681411102868e-05,
+      "loss": 3.3116,
+      "step": 8340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.640250520101297e-05,
+      "loss": 3.2017,
+      "step": 8345
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.639819380535696e-05,
+      "loss": 3.424,
+      "step": 8350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.63938799242916e-05,
+      "loss": 3.1269,
+      "step": 8355
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.638956355804801e-05,
+      "loss": 3.2319,
+      "step": 8360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.638524470685735e-05,
+      "loss": 3.1691,
+      "step": 8365
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.6380923370951e-05,
+      "loss": 3.2155,
+      "step": 8370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.637659955056043e-05,
+      "loss": 3.2131,
+      "step": 8375
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.637227324591727e-05,
+      "loss": 3.2628,
+      "step": 8380
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.636794445725323e-05,
+      "loss": 3.2942,
+      "step": 8385
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.636361318480023e-05,
+      "loss": 3.3053,
+      "step": 8390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.635927942879026e-05,
+      "loss": 3.2512,
+      "step": 8395
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.635494318945549e-05,
+      "loss": 3.1941,
+      "step": 8400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.635060446702815e-05,
+      "loss": 3.1677,
+      "step": 8405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.63462632617407e-05,
+      "loss": 3.1875,
+      "step": 8410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.634191957382567e-05,
+      "loss": 3.1736,
+      "step": 8415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.633757340351575e-05,
+      "loss": 3.2173,
+      "step": 8420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.633322475104372e-05,
+      "loss": 3.1083,
+      "step": 8425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.632887361664256e-05,
+      "loss": 3.2212,
+      "step": 8430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.63245200005453e-05,
+      "loss": 3.2655,
+      "step": 8435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.63201639029852e-05,
+      "loss": 3.2163,
+      "step": 8440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.631580532419558e-05,
+      "loss": 3.189,
+      "step": 8445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.631144426440992e-05,
+      "loss": 3.2151,
+      "step": 8450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.630708072386182e-05,
+      "loss": 3.3274,
+      "step": 8455
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.630271470278503e-05,
+      "loss": 3.1812,
+      "step": 8460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.629834620141342e-05,
+      "loss": 3.2273,
+      "step": 8465
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.6293975219981e-05,
+      "loss": 3.2112,
+      "step": 8470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628960175872193e-05,
+      "loss": 3.1794,
+      "step": 8475
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628522581787043e-05,
+      "loss": 3.0626,
+      "step": 8480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.628084739766096e-05,
+      "loss": 3.2628,
+      "step": 8485
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.627646649832802e-05,
+      "loss": 3.2623,
+      "step": 8490
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.627208312010631e-05,
+      "loss": 3.2921,
+      "step": 8495
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.626769726323061e-05,
+      "loss": 3.258,
+      "step": 8500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.62633089279359e-05,
+      "loss": 3.1675,
+      "step": 8505
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.62589181144572e-05,
+      "loss": 3.0997,
+      "step": 8510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.625452482302972e-05,
+      "loss": 3.2197,
+      "step": 8515
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.625012905388881e-05,
+      "loss": 3.2235,
+      "step": 8520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.624573080726995e-05,
+      "loss": 3.1072,
+      "step": 8525
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.624133008340871e-05,
+      "loss": 3.2245,
+      "step": 8530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.623692688254084e-05,
+      "loss": 3.2487,
+      "step": 8535
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.623252120490223e-05,
+      "loss": 3.2156,
+      "step": 8540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.622811305072883e-05,
+      "loss": 3.2146,
+      "step": 8545
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.622370242025678e-05,
+      "loss": 3.3379,
+      "step": 8550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.621928931372239e-05,
+      "loss": 3.2566,
+      "step": 8555
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.6214873731362e-05,
+      "loss": 3.1414,
+      "step": 8560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.621045567341219e-05,
+      "loss": 3.2289,
+      "step": 8565
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.620603514010958e-05,
+      "loss": 3.1922,
+      "step": 8570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.620161213169098e-05,
+      "loss": 3.1398,
+      "step": 8575
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.619718664839333e-05,
+      "loss": 3.2609,
+      "step": 8580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.619275869045366e-05,
+      "loss": 3.1928,
+      "step": 8585
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.618832825810918e-05,
+      "loss": 3.2851,
+      "step": 8590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.618389535159722e-05,
+      "loss": 3.0511,
+      "step": 8595
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.617945997115523e-05,
+      "loss": 3.2039,
+      "step": 8600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.617502211702081e-05,
+      "loss": 3.164,
+      "step": 8605
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.617058178943166e-05,
+      "loss": 3.0388,
+      "step": 8610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.616613898862565e-05,
+      "loss": 3.2577,
+      "step": 8615
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.616169371484078e-05,
+      "loss": 3.0994,
+      "step": 8620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.615724596831514e-05,
+      "loss": 3.1989,
+      "step": 8625
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.615279574928702e-05,
+      "loss": 3.156,
+      "step": 8630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.614834305799476e-05,
+      "loss": 3.1162,
+      "step": 8635
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61438878946769e-05,
+      "loss": 3.1668,
+      "step": 8640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61394302595721e-05,
+      "loss": 3.409,
+      "step": 8645
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.613497015291913e-05,
+      "loss": 3.306,
+      "step": 8650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61305075749569e-05,
+      "loss": 3.257,
+      "step": 8655
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.612604252592445e-05,
+      "loss": 3.113,
+      "step": 8660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.612157500606098e-05,
+      "loss": 3.2449,
+      "step": 8665
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.61171050156058e-05,
+      "loss": 3.1239,
+      "step": 8670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.611263255479833e-05,
+      "loss": 3.1753,
+      "step": 8675
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.610815762387816e-05,
+      "loss": 3.2155,
+      "step": 8680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.610368022308502e-05,
+      "loss": 3.2423,
+      "step": 8685
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.609920035265871e-05,
+      "loss": 3.1813,
+      "step": 8690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.609471801283924e-05,
+      "loss": 3.1327,
+      "step": 8695
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.60902332038667e-05,
+      "loss": 3.2479,
+      "step": 8700
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.608574592598131e-05,
+      "loss": 3.1986,
+      "step": 8705
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.608125617942348e-05,
+      "loss": 3.2884,
+      "step": 8710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.607676396443366e-05,
+      "loss": 3.2552,
+      "step": 8715
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.607226928125253e-05,
+      "loss": 3.2104,
+      "step": 8720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.606777213012084e-05,
+      "loss": 3.2524,
+      "step": 8725
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.606327251127947e-05,
+      "loss": 3.2516,
+      "step": 8730
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.605877042496947e-05,
+      "loss": 3.2092,
+      "step": 8735
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.605426587143202e-05,
+      "loss": 3.2117,
+      "step": 8740
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.60497588509084e-05,
+      "loss": 3.1618,
+      "step": 8745
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.604524936364001e-05,
+      "loss": 3.2176,
+      "step": 8750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.604073740986843e-05,
+      "loss": 3.2011,
+      "step": 8755
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.603622298983536e-05,
+      "loss": 3.2471,
+      "step": 8760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.603170610378265e-05,
+      "loss": 3.262,
+      "step": 8765
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.60271867519522e-05,
+      "loss": 3.1948,
+      "step": 8770
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.602266493458612e-05,
+      "loss": 3.1873,
+      "step": 8775
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.601814065192663e-05,
+      "loss": 3.1634,
+      "step": 8780
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.601361390421608e-05,
+      "loss": 3.0276,
+      "step": 8785
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.600908469169696e-05,
+      "loss": 3.2862,
+      "step": 8790
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.60045530146119e-05,
+      "loss": 3.2787,
+      "step": 8795
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.600001887320362e-05,
+      "loss": 3.154,
+      "step": 8800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.599548226771501e-05,
+      "loss": 3.1733,
+      "step": 8805
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.599094319838908e-05,
+      "loss": 3.3918,
+      "step": 8810
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.598640166546899e-05,
+      "loss": 3.1317,
+      "step": 8815
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.5981857669198e-05,
+      "loss": 3.2254,
+      "step": 8820
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.597731120981952e-05,
+      "loss": 3.2574,
+      "step": 8825
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59727622875771e-05,
+      "loss": 3.1262,
+      "step": 8830
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59682109027144e-05,
+      "loss": 3.083,
+      "step": 8835
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.596365705547522e-05,
+      "loss": 3.0535,
+      "step": 8840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59591007461035e-05,
+      "loss": 3.2864,
+      "step": 8845
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.595454197484332e-05,
+      "loss": 3.1581,
+      "step": 8850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.594998074193889e-05,
+      "loss": 3.1011,
+      "step": 8855
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59454170476345e-05,
+      "loss": 3.2873,
+      "step": 8860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.594085089217462e-05,
+      "loss": 3.2008,
+      "step": 8865
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.593628227580388e-05,
+      "loss": 3.1216,
+      "step": 8870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.593171119876698e-05,
+      "loss": 3.1559,
+      "step": 8875
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59271376613088e-05,
+      "loss": 3.2167,
+      "step": 8880
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.59225616636743e-05,
+      "loss": 3.2709,
+      "step": 8885
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.591798320610863e-05,
+      "loss": 3.2248,
+      "step": 8890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.591340228885701e-05,
+      "loss": 3.1386,
+      "step": 8895
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.590881891216486e-05,
+      "loss": 3.2918,
+      "step": 8900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.590423307627769e-05,
+      "loss": 3.2697,
+      "step": 8905
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589964478144114e-05,
+      "loss": 3.1804,
+      "step": 8910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.5895054027901e-05,
+      "loss": 3.1439,
+      "step": 8915
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.589046081590318e-05,
+      "loss": 3.3343,
+      "step": 8920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.588586514569371e-05,
+      "loss": 3.1331,
+      "step": 8925
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.588126701751879e-05,
+      "loss": 3.1814,
+      "step": 8930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58766664316247e-05,
+      "loss": 3.1384,
+      "step": 8935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.587206338825792e-05,
+      "loss": 3.217,
+      "step": 8940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.586745788766498e-05,
+      "loss": 3.1849,
+      "step": 8945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.586284993009259e-05,
+      "loss": 3.2201,
+      "step": 8950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58582395157876e-05,
+      "loss": 2.9876,
+      "step": 8955
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.585362664499698e-05,
+      "loss": 3.1251,
+      "step": 8960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.58490113179678e-05,
+      "loss": 3.1427,
+      "step": 8965
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.584439353494732e-05,
+      "loss": 3.2738,
+      "step": 8970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.583977329618288e-05,
+      "loss": 3.3441,
+      "step": 8975
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.583515060192196e-05,
+      "loss": 3.2444,
+      "step": 8980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.583052545241222e-05,
+      "loss": 3.1918,
+      "step": 8985
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582589784790139e-05,
+      "loss": 3.2837,
+      "step": 8990
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.582126778863738e-05,
+      "loss": 3.1947,
+      "step": 8995
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.581663527486817e-05,
+      "loss": 3.0834,
+      "step": 9000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.581200030684194e-05,
+      "loss": 3.2427,
+      "step": 9005
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580736288480696e-05,
+      "loss": 3.2714,
+      "step": 9010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.580272300901163e-05,
+      "loss": 3.1816,
+      "step": 9015
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.579808067970453e-05,
+      "loss": 3.228,
+      "step": 9020
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57934358971343e-05,
+      "loss": 3.3195,
+      "step": 9025
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578878866154976e-05,
+      "loss": 3.1912,
+      "step": 9030
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.578413897319987e-05,
+      "loss": 3.1085,
+      "step": 9035
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577948683233365e-05,
+      "loss": 3.3157,
+      "step": 9040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577483223920035e-05,
+      "loss": 3.1671,
+      "step": 9045
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.577017519404928e-05,
+      "loss": 3.2664,
+      "step": 9050
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.576551569712989e-05,
+      "loss": 3.2894,
+      "step": 9055
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57608537486918e-05,
+      "loss": 3.2923,
+      "step": 9060
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.575618934898472e-05,
+      "loss": 3.1757,
+      "step": 9065
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.575152249825851e-05,
+      "loss": 3.1415,
+      "step": 9070
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574685319676317e-05,
+      "loss": 3.3179,
+      "step": 9075
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.574218144474881e-05,
+      "loss": 3.1292,
+      "step": 9080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573750724246569e-05,
+      "loss": 3.2277,
+      "step": 9085
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.573283059016418e-05,
+      "loss": 3.1073,
+      "step": 9090
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57281514880948e-05,
+      "loss": 3.1453,
+      "step": 9095
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.572346993650821e-05,
+      "loss": 3.1831,
+      "step": 9100
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.571878593565517e-05,
+      "loss": 3.2741,
+      "step": 9105
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.571409948578658e-05,
+      "loss": 3.2552,
+      "step": 9110
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.57094105871535e-05,
+      "loss": 3.1554,
+      "step": 9115
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.570471924000707e-05,
+      "loss": 3.2886,
+      "step": 9120
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.570002544459864e-05,
+      "loss": 3.1255,
+      "step": 9125
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.569532920117959e-05,
+      "loss": 3.3279,
+      "step": 9130
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.569063051000151e-05,
+      "loss": 3.1547,
+      "step": 9135
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.568592937131611e-05,
+      "loss": 3.2373,
+      "step": 9140
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.568122578537517e-05,
+      "loss": 3.2019,
+      "step": 9145
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567651975243068e-05,
+      "loss": 3.247,
+      "step": 9150
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.567181127273473e-05,
+      "loss": 3.1998,
+      "step": 9155
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566710034653951e-05,
+      "loss": 3.2026,
+      "step": 9160
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.566238697409741e-05,
+      "loss": 3.2606,
+      "step": 9165
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.565767115566088e-05,
+      "loss": 3.1522,
+      "step": 9170
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.565295289148256e-05,
+      "loss": 3.254,
+      "step": 9175
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.564823218181513e-05,
+      "loss": 3.1414,
+      "step": 9180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.564350902691154e-05,
+      "loss": 3.332,
+      "step": 9185
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.563878342702477e-05,
+      "loss": 3.1733,
+      "step": 9190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.563405538240796e-05,
+      "loss": 3.1493,
+      "step": 9195
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.562932489331433e-05,
+      "loss": 3.1695,
+      "step": 9200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.562459195999733e-05,
+      "loss": 3.2231,
+      "step": 9205
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.561985658271049e-05,
+      "loss": 3.2477,
+      "step": 9210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.561511876170743e-05,
+      "loss": 3.2964,
+      "step": 9215
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.561037849724199e-05,
+      "loss": 3.2173,
+      "step": 9220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.560563578956805e-05,
+      "loss": 3.2187,
+      "step": 9225
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.560089063893968e-05,
+      "loss": 3.3162,
+      "step": 9230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.559614304561107e-05,
+      "loss": 3.1232,
+      "step": 9235
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.559139300983651e-05,
+      "loss": 3.1531,
+      "step": 9240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.558664053187045e-05,
+      "loss": 3.2253,
+      "step": 9245
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55818856119675e-05,
+      "loss": 3.1712,
+      "step": 9250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.557712825038232e-05,
+      "loss": 3.0659,
+      "step": 9255
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.557236844736978e-05,
+      "loss": 3.1691,
+      "step": 9260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.556760620318486e-05,
+      "loss": 3.1506,
+      "step": 9265
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55628415180826e-05,
+      "loss": 3.2332,
+      "step": 9270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.555807439231828e-05,
+      "loss": 3.1632,
+      "step": 9275
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.555330482614725e-05,
+      "loss": 3.1352,
+      "step": 9280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.5548532819825e-05,
+      "loss": 3.144,
+      "step": 9285
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.554375837360713e-05,
+      "loss": 3.2551,
+      "step": 9290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.553898148774944e-05,
+      "loss": 3.0904,
+      "step": 9295
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.553420216250777e-05,
+      "loss": 3.3831,
+      "step": 9300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.552942039813815e-05,
+      "loss": 3.2737,
+      "step": 9305
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55246361948967e-05,
+      "loss": 3.1152,
+      "step": 9310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.551984955303977e-05,
+      "loss": 3.1832,
+      "step": 9315
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.551506047282367e-05,
+      "loss": 3.2322,
+      "step": 9320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.5510268954505e-05,
+      "loss": 3.2124,
+      "step": 9325
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.55054749983404e-05,
+      "loss": 3.2329,
+      "step": 9330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.550067860458668e-05,
+      "loss": 3.099,
+      "step": 9335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549587977350076e-05,
+      "loss": 3.1429,
+      "step": 9340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.549107850533969e-05,
+      "loss": 3.2193,
+      "step": 9345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.54862748003607e-05,
+      "loss": 3.195,
+      "step": 9350
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.548146865882106e-05,
+      "loss": 3.0163,
+      "step": 9355
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547666008097824e-05,
+      "loss": 3.1305,
+      "step": 9360
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.547184906708984e-05,
+      "loss": 3.1581,
+      "step": 9365
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546703561741353e-05,
+      "loss": 3.1238,
+      "step": 9370
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.546221973220719e-05,
+      "loss": 3.11,
+      "step": 9375
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.545740141172876e-05,
+      "loss": 3.1631,
+      "step": 9380
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.545258065623638e-05,
+      "loss": 3.3853,
+      "step": 9385
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544775746598826e-05,
+      "loss": 3.3461,
+      "step": 9390
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.544293184124276e-05,
+      "loss": 3.1871,
+      "step": 9395
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.543810378225838e-05,
+      "loss": 3.2542,
+      "step": 9400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.543327328929375e-05,
+      "loss": 3.1825,
+      "step": 9405
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542844036260762e-05,
+      "loss": 3.3906,
+      "step": 9410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.542360500245885e-05,
+      "loss": 3.2462,
+      "step": 9415
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.541876720910652e-05,
+      "loss": 3.2732,
+      "step": 9420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.541392698280972e-05,
+      "loss": 3.1754,
+      "step": 9425
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540908432382774e-05,
+      "loss": 3.2354,
+      "step": 9430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.540423923241999e-05,
+      "loss": 3.1843,
+      "step": 9435
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.5399391708846e-05,
+      "loss": 3.1853,
+      "step": 9440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.539454175336545e-05,
+      "loss": 3.1725,
+      "step": 9445
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538968936623813e-05,
+      "loss": 3.262,
+      "step": 9450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.538483454772396e-05,
+      "loss": 3.2544,
+      "step": 9455
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.5379977298083e-05,
+      "loss": 3.1385,
+      "step": 9460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.537511761757546e-05,
+      "loss": 3.1921,
+      "step": 9465
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.537025550646164e-05,
+      "loss": 3.29,
+      "step": 9470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.536539096500198e-05,
+      "loss": 3.2885,
+      "step": 9475
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.536052399345707e-05,
+      "loss": 3.2187,
+      "step": 9480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535565459208762e-05,
+      "loss": 3.2303,
+      "step": 9485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.535078276115445e-05,
+      "loss": 3.2621,
+      "step": 9490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.534590850091857e-05,
+      "loss": 3.3346,
+      "step": 9495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.534103181164104e-05,
+      "loss": 3.1984,
+      "step": 9500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533615269358312e-05,
+      "loss": 3.1731,
+      "step": 9505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.533127114700613e-05,
+      "loss": 3.2743,
+      "step": 9510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532638717217159e-05,
+      "loss": 3.2726,
+      "step": 9515
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.532150076934112e-05,
+      "loss": 3.0906,
+      "step": 9520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.531661193877646e-05,
+      "loss": 3.1501,
+      "step": 9525
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.531172068073947e-05,
+      "loss": 3.2787,
+      "step": 9530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.53068269954922e-05,
+      "loss": 3.2654,
+      "step": 9535
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.530193088329677e-05,
+      "loss": 3.0241,
+      "step": 9540
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.529703234441545e-05,
+      "loss": 3.2419,
+      "step": 9545
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.529213137911064e-05,
+      "loss": 3.2415,
+      "step": 9550
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.528722798764487e-05,
+      "loss": 3.1685,
+      "step": 9555
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.528232217028081e-05,
+      "loss": 3.1845,
+      "step": 9560
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.527741392728122e-05,
+      "loss": 3.1447,
+      "step": 9565
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.527250325890907e-05,
+      "loss": 3.3321,
+      "step": 9570
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.526759016542735e-05,
+      "loss": 3.1111,
+      "step": 9575
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.526267464709926e-05,
+      "loss": 3.1426,
+      "step": 9580
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.525775670418813e-05,
+      "loss": 3.0952,
+      "step": 9585
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.525283633695741e-05,
+      "loss": 3.1481,
+      "step": 9590
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.524791354567062e-05,
+      "loss": 3.2116,
+      "step": 9595
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.524298833059149e-05,
+      "loss": 3.2304,
+      "step": 9600
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.523806069198384e-05,
+      "loss": 3.2157,
+      "step": 9605
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.523313063011163e-05,
+      "loss": 3.2908,
+      "step": 9610
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.522819814523896e-05,
+      "loss": 3.1691,
+      "step": 9615
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.522326323763002e-05,
+      "loss": 3.3104,
+      "step": 9620
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.52183259075492e-05,
+      "loss": 3.2319,
+      "step": 9625
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.521338615526094e-05,
+      "loss": 3.203,
+      "step": 9630
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.520844398102985e-05,
+      "loss": 3.3231,
+      "step": 9635
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.52034993851207e-05,
+      "loss": 3.2182,
+      "step": 9640
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.519855236779832e-05,
+      "loss": 3.1891,
+      "step": 9645
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.519360292932772e-05,
+      "loss": 3.2547,
+      "step": 9650
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.518865106997403e-05,
+      "loss": 3.1262,
+      "step": 9655
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.518369679000251e-05,
+      "loss": 3.2918,
+      "step": 9660
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.517874008967854e-05,
+      "loss": 3.2451,
+      "step": 9665
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.517378096926764e-05,
+      "loss": 3.2026,
+      "step": 9670
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.516881942903544e-05,
+      "loss": 3.2687,
+      "step": 9675
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.516385546924772e-05,
+      "loss": 3.3413,
+      "step": 9680
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.51588890901704e-05,
+      "loss": 3.1463,
+      "step": 9685
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.515392029206948e-05,
+      "loss": 3.2448,
+      "step": 9690
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.514894907521117e-05,
+      "loss": 3.2364,
+      "step": 9695
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.514397543986173e-05,
+      "loss": 3.1014,
+      "step": 9700
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.51389993862876e-05,
+      "loss": 3.1562,
+      "step": 9705
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.513402091475531e-05,
+      "loss": 3.2521,
+      "step": 9710
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.512904002553156e-05,
+      "loss": 3.1611,
+      "step": 9715
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.512405671888315e-05,
+      "loss": 3.206,
+      "step": 9720
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.511907099507701e-05,
+      "loss": 3.2732,
+      "step": 9725
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.511408285438026e-05,
+      "loss": 3.0915,
+      "step": 9730
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.510909229706005e-05,
+      "loss": 3.1405,
+      "step": 9735
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.510409932338374e-05,
+      "loss": 3.218,
+      "step": 9740
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.509910393361874e-05,
+      "loss": 3.1784,
+      "step": 9745
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.50941061280327e-05,
+      "loss": 3.1465,
+      "step": 9750
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.508910590689328e-05,
+      "loss": 3.0287,
+      "step": 9755
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.508410327046838e-05,
+      "loss": 3.0616,
+      "step": 9760
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.507909821902592e-05,
+      "loss": 3.2812,
+      "step": 9765
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.507409075283406e-05,
+      "loss": 3.2265,
+      "step": 9770
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.5069080872161e-05,
+      "loss": 3.3055,
+      "step": 9775
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.506406857727511e-05,
+      "loss": 3.2841,
+      "step": 9780
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.505905386844489e-05,
+      "loss": 3.2566,
+      "step": 9785
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.505403674593896e-05,
+      "loss": 3.0677,
+      "step": 9790
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.504901721002608e-05,
+      "loss": 3.2523,
+      "step": 9795
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.504399526097512e-05,
+      "loss": 3.1102,
+      "step": 9800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.503897089905509e-05,
+      "loss": 3.2463,
+      "step": 9805
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.503394412453513e-05,
+      "loss": 3.1479,
+      "step": 9810
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.50289149376845e-05,
+      "loss": 3.2265,
+      "step": 9815
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.502388333877264e-05,
+      "loss": 3.2966,
+      "step": 9820
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.501884932806902e-05,
+      "loss": 3.1988,
+      "step": 9825
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.501381290584334e-05,
+      "loss": 3.0817,
+      "step": 9830
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.500877407236537e-05,
+      "loss": 3.2079,
+      "step": 9835
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.5003732827905e-05,
+      "loss": 3.2448,
+      "step": 9840
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.499868917273232e-05,
+      "loss": 3.1921,
+      "step": 9845
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.499364310711749e-05,
+      "loss": 3.2129,
+      "step": 9850
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.498859463133078e-05,
+      "loss": 3.2493,
+      "step": 9855
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.498354374564266e-05,
+      "loss": 3.146,
+      "step": 9860
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.497849045032368e-05,
+      "loss": 3.2518,
+      "step": 9865
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.497343474564451e-05,
+      "loss": 3.2355,
+      "step": 9870
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.496837663187601e-05,
+      "loss": 3.1457,
+      "step": 9875
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.496331610928909e-05,
+      "loss": 3.0948,
+      "step": 9880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.495825317815483e-05,
+      "loss": 3.289,
+      "step": 9885
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.495318783874443e-05,
+      "loss": 3.3239,
+      "step": 9890
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.494812009132927e-05,
+      "loss": 3.2277,
+      "step": 9895
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.494304993618078e-05,
+      "loss": 3.2394,
+      "step": 9900
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.493797737357053e-05,
+      "loss": 3.2025,
+      "step": 9905
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.493290240377028e-05,
+      "loss": 3.2037,
+      "step": 9910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.492782502705189e-05,
+      "loss": 3.2456,
+      "step": 9915
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.49227452436873e-05,
+      "loss": 3.1374,
+      "step": 9920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.491766305394862e-05,
+      "loss": 3.0698,
+      "step": 9925
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.491257845810813e-05,
+      "loss": 3.1733,
+      "step": 9930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.490749145643813e-05,
+      "loss": 3.1261,
+      "step": 9935
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.490240204921119e-05,
+      "loss": 3.2587,
+      "step": 9940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.48973102366999e-05,
+      "loss": 3.2826,
+      "step": 9945
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.489221601917699e-05,
+      "loss": 3.1513,
+      "step": 9950
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.488711939691537e-05,
+      "loss": 3.291,
+      "step": 9955
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.488202037018806e-05,
+      "loss": 3.1635,
+      "step": 9960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.487691893926815e-05,
+      "loss": 3.1682,
+      "step": 9965
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.487181510442897e-05,
+      "loss": 3.0849,
+      "step": 9970
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.486670886594387e-05,
+      "loss": 3.2895,
+      "step": 9975
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.486160022408642e-05,
+      "loss": 3.0897,
+      "step": 9980
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.485648917913021e-05,
+      "loss": 3.3051,
+      "step": 9985
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.485137573134907e-05,
+      "loss": 3.1988,
+      "step": 9990
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.484625988101691e-05,
+      "loss": 3.3076,
+      "step": 9995
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.484114162840778e-05,
+      "loss": 3.1551,
+      "step": 10000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.483602097379582e-05,
+      "loss": 3.2604,
+      "step": 10005
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.483089791745533e-05,
+      "loss": 3.2369,
+      "step": 10010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.482577245966076e-05,
+      "loss": 3.2101,
+      "step": 10015
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.482064460068663e-05,
+      "loss": 3.2406,
+      "step": 10020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.481551434080766e-05,
+      "loss": 3.2238,
+      "step": 10025
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.481038168029865e-05,
+      "loss": 3.2845,
+      "step": 10030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.480524661943453e-05,
+      "loss": 3.1305,
+      "step": 10035
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.480010915849038e-05,
+      "loss": 3.1469,
+      "step": 10040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.479496929774138e-05,
+      "loss": 3.2071,
+      "step": 10045
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47898270374629e-05,
+      "loss": 3.2709,
+      "step": 10050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.478468237793036e-05,
+      "loss": 3.1651,
+      "step": 10055
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.477953531941935e-05,
+      "loss": 3.1693,
+      "step": 10060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.477438586220557e-05,
+      "loss": 3.1181,
+      "step": 10065
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.476923400656489e-05,
+      "loss": 3.3772,
+      "step": 10070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.476407975277326e-05,
+      "loss": 3.2061,
+      "step": 10075
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.475892310110677e-05,
+      "loss": 3.2168,
+      "step": 10080
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.475376405184167e-05,
+      "loss": 3.2817,
+      "step": 10085
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47486026052543e-05,
+      "loss": 3.3413,
+      "step": 10090
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.474343876162114e-05,
+      "loss": 3.261,
+      "step": 10095
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.473827252121881e-05,
+      "loss": 3.1539,
+      "step": 10100
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.473310388432404e-05,
+      "loss": 3.269,
+      "step": 10105
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47279328512137e-05,
+      "loss": 3.2244,
+      "step": 10110
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47227594221648e-05,
+      "loss": 3.213,
+      "step": 10115
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.471758359745445e-05,
+      "loss": 3.2533,
+      "step": 10120
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.47124053773599e-05,
+      "loss": 3.1002,
+      "step": 10125
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.470722476215855e-05,
+      "loss": 3.1939,
+      "step": 10130
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.470204175212791e-05,
+      "loss": 3.092,
+      "step": 10135
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.469685634754558e-05,
+      "loss": 3.2245,
+      "step": 10140
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.469166854868936e-05,
+      "loss": 3.0703,
+      "step": 10145
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.468647835583715e-05,
+      "loss": 3.2126,
+      "step": 10150
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.468128576926696e-05,
+      "loss": 3.2284,
+      "step": 10155
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.467609078925692e-05,
+      "loss": 3.0968,
+      "step": 10160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.467089341608535e-05,
+      "loss": 3.0222,
+      "step": 10165
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.466569365003065e-05,
+      "loss": 3.1703,
+      "step": 10170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.46604914913713e-05,
+      "loss": 3.2043,
+      "step": 10175
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.465528694038605e-05,
+      "loss": 3.2188,
+      "step": 10180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.465007999735363e-05,
+      "loss": 3.1666,
+      "step": 10185
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.464487066255298e-05,
+      "loss": 3.2149,
+      "step": 10190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.463965893626314e-05,
+      "loss": 3.1996,
+      "step": 10195
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.463444481876331e-05,
+      "loss": 3.1175,
+      "step": 10200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.462922831033278e-05,
+      "loss": 3.201,
+      "step": 10205
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.462400941125096e-05,
+      "loss": 3.3403,
+      "step": 10210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.461878812179745e-05,
+      "loss": 3.18,
+      "step": 10215
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.461356444225191e-05,
+      "loss": 3.239,
+      "step": 10220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.460833837289416e-05,
+      "loss": 3.1588,
+      "step": 10225
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.460310991400416e-05,
+      "loss": 3.1431,
+      "step": 10230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.459787906586198e-05,
+      "loss": 3.1485,
+      "step": 10235
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.459264582874779e-05,
+      "loss": 3.1467,
+      "step": 10240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.458741020294195e-05,
+      "loss": 3.2861,
+      "step": 10245
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.458217218872492e-05,
+      "loss": 3.1618,
+      "step": 10250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.457693178637726e-05,
+      "loss": 3.1702,
+      "step": 10255
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.45716889961797e-05,
+      "loss": 3.2412,
+      "step": 10260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.456644381841308e-05,
+      "loss": 3.3409,
+      "step": 10265
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.456119625335835e-05,
+      "loss": 3.18,
+      "step": 10270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.455594630129663e-05,
+      "loss": 3.2114,
+      "step": 10275
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.455069396250913e-05,
+      "loss": 3.1988,
+      "step": 10280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.454543923727722e-05,
+      "loss": 3.3,
+      "step": 10285
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.454018212588234e-05,
+      "loss": 3.1811,
+      "step": 10290
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.453492262860612e-05,
+      "loss": 3.2058,
+      "step": 10295
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.452966074573034e-05,
+      "loss": 3.1364,
+      "step": 10300
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.452439647753677e-05,
+      "loss": 3.175,
+      "step": 10305
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.451912982430747e-05,
+      "loss": 3.3041,
+      "step": 10310
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.451386078632455e-05,
+      "loss": 3.1716,
+      "step": 10315
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.450858936387023e-05,
+      "loss": 3.1325,
+      "step": 10320
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.450331555722691e-05,
+      "loss": 3.2238,
+      "step": 10325
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.449803936667708e-05,
+      "loss": 3.265,
+      "step": 10330
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.449276079250335e-05,
+      "loss": 3.3097,
+      "step": 10335
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.448747983498853e-05,
+      "loss": 3.1855,
+      "step": 10340
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.448219649441546e-05,
+      "loss": 3.2504,
+      "step": 10345
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.447691077106716e-05,
+      "loss": 3.2192,
+      "step": 10350
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.447162266522678e-05,
+      "loss": 3.2598,
+      "step": 10355
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.446633217717757e-05,
+      "loss": 3.277,
+      "step": 10360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.446103930720296e-05,
+      "loss": 3.2488,
+      "step": 10365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.445574405558643e-05,
+      "loss": 3.2508,
+      "step": 10370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.445044642261166e-05,
+      "loss": 3.1843,
+      "step": 10375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.444514640856242e-05,
+      "loss": 3.2362,
+      "step": 10380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.443984401372261e-05,
+      "loss": 3.2894,
+      "step": 10385
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.443453923837628e-05,
+      "loss": 3.2023,
+      "step": 10390
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.442923208280756e-05,
+      "loss": 3.1164,
+      "step": 10395
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.442392254730077e-05,
+      "loss": 3.1913,
+      "step": 10400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.441861063214031e-05,
+      "loss": 3.2694,
+      "step": 10405
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.441329633761073e-05,
+      "loss": 3.1656,
+      "step": 10410
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.440797966399669e-05,
+      "loss": 3.2851,
+      "step": 10415
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.440266061158298e-05,
+      "loss": 3.1535,
+      "step": 10420
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.439733918065456e-05,
+      "loss": 3.1466,
+      "step": 10425
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.439201537149646e-05,
+      "loss": 3.3068,
+      "step": 10430
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.438668918439385e-05,
+      "loss": 3.0219,
+      "step": 10435
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.438136061963206e-05,
+      "loss": 3.2256,
+      "step": 10440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.43760296774965e-05,
+      "loss": 3.003,
+      "step": 10445
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.437069635827275e-05,
+      "loss": 3.3688,
+      "step": 10450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.436536066224651e-05,
+      "loss": 3.2576,
+      "step": 10455
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.436002258970358e-05,
+      "loss": 3.1979,
+      "step": 10460
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.43546821409299e-05,
+      "loss": 3.1634,
+      "step": 10465
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.434933931621155e-05,
+      "loss": 3.1795,
+      "step": 10470
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.434399411583472e-05,
+      "loss": 3.1387,
+      "step": 10475
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.433864654008576e-05,
+      "loss": 3.2157,
+      "step": 10480
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.43332965892511e-05,
+      "loss": 3.1302,
+      "step": 10485
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.432794426361732e-05,
+      "loss": 3.1815,
+      "step": 10490
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.432258956347115e-05,
+      "loss": 3.1593,
+      "step": 10495
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.43172324890994e-05,
+      "loss": 3.1853,
+      "step": 10500
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.431187304078903e-05,
+      "loss": 3.1736,
+      "step": 10505
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.430651121882716e-05,
+      "loss": 3.332,
+      "step": 10510
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.430114702350099e-05,
+      "loss": 3.1088,
+      "step": 10515
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.429578045509786e-05,
+      "loss": 3.1794,
+      "step": 10520
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.429041151390523e-05,
+      "loss": 3.08,
+      "step": 10525
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.428504020021072e-05,
+      "loss": 3.0959,
+      "step": 10530
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.427966651430204e-05,
+      "loss": 3.1775,
+      "step": 10535
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.427429045646704e-05,
+      "loss": 3.2315,
+      "step": 10540
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.426891202699372e-05,
+      "loss": 3.1827,
+      "step": 10545
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.426353122617015e-05,
+      "loss": 3.1253,
+      "step": 10550
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.425814805428462e-05,
+      "loss": 3.1965,
+      "step": 10555
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.425276251162542e-05,
+      "loss": 3.1442,
+      "step": 10560
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.424737459848108e-05,
+      "loss": 3.2014,
+      "step": 10565
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.424198431514021e-05,
+      "loss": 3.154,
+      "step": 10570
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.423659166189156e-05,
+      "loss": 3.1593,
+      "step": 10575
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.423119663902397e-05,
+      "loss": 3.1928,
+      "step": 10580
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.422579924682644e-05,
+      "loss": 3.1487,
+      "step": 10585
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.422039948558811e-05,
+      "loss": 2.9727,
+      "step": 10590
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.421499735559823e-05,
+      "loss": 3.174,
+      "step": 10595
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.420959285714617e-05,
+      "loss": 3.165,
+      "step": 10600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.420418599052141e-05,
+      "loss": 3.2789,
+      "step": 10605
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.41987767560136e-05,
+      "loss": 3.0922,
+      "step": 10610
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.41933651539125e-05,
+      "loss": 3.2444,
+      "step": 10615
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.418795118450799e-05,
+      "loss": 3.3518,
+      "step": 10620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.418253484809007e-05,
+      "loss": 3.1737,
+      "step": 10625
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.417711614494887e-05,
+      "loss": 3.1982,
+      "step": 10630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.417169507537469e-05,
+      "loss": 3.1144,
+      "step": 10635
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.416627163965788e-05,
+      "loss": 3.1625,
+      "step": 10640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.416084583808898e-05,
+      "loss": 3.2276,
+      "step": 10645
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.415541767095862e-05,
+      "loss": 3.2956,
+      "step": 10650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.414998713855758e-05,
+      "loss": 3.1359,
+      "step": 10655
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.414455424117677e-05,
+      "loss": 3.2332,
+      "step": 10660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.413911897910718e-05,
+      "loss": 3.117,
+      "step": 10665
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.413368135263997e-05,
+      "loss": 3.2056,
+      "step": 10670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.412824136206643e-05,
+      "loss": 3.2273,
+      "step": 10675
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.412279900767798e-05,
+      "loss": 3.1583,
+      "step": 10680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.411735428976611e-05,
+      "loss": 3.1107,
+      "step": 10685
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.411190720862251e-05,
+      "loss": 3.3497,
+      "step": 10690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.410645776453894e-05,
+      "loss": 3.3216,
+      "step": 10695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.410100595780733e-05,
+      "loss": 3.2659,
+      "step": 10700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.40955517887197e-05,
+      "loss": 3.1466,
+      "step": 10705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.40900952575682e-05,
+      "loss": 3.155,
+      "step": 10710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.408463636464514e-05,
+      "loss": 3.2419,
+      "step": 10715
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.407917511024296e-05,
+      "loss": 3.1759,
+      "step": 10720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.407371149465417e-05,
+      "loss": 3.1452,
+      "step": 10725
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.406824551817145e-05,
+      "loss": 3.1331,
+      "step": 10730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.406277718108757e-05,
+      "loss": 3.1422,
+      "step": 10735
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.40573064836955e-05,
+      "loss": 3.1716,
+      "step": 10740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.405183342628827e-05,
+      "loss": 3.1739,
+      "step": 10745
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.404635800915902e-05,
+      "loss": 3.221,
+      "step": 10750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.40408802326011e-05,
+      "loss": 3.1404,
+      "step": 10755
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.40354000969079e-05,
+      "loss": 3.1306,
+      "step": 10760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.402991760237301e-05,
+      "loss": 3.2507,
+      "step": 10765
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.402443274929009e-05,
+      "loss": 3.1038,
+      "step": 10770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.401894553795294e-05,
+      "loss": 3.3133,
+      "step": 10775
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.40134559686555e-05,
+      "loss": 3.1716,
+      "step": 10780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.400796404169185e-05,
+      "loss": 3.1162,
+      "step": 10785
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.400246975735614e-05,
+      "loss": 3.2734,
+      "step": 10790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.39969731159427e-05,
+      "loss": 3.188,
+      "step": 10795
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.399147411774596e-05,
+      "loss": 3.2006,
+      "step": 10800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.398597276306052e-05,
+      "loss": 3.1795,
+      "step": 10805
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.398046905218101e-05,
+      "loss": 3.2133,
+      "step": 10810
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.39749629854023e-05,
+      "loss": 3.2879,
+      "step": 10815
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.396945456301931e-05,
+      "loss": 3.2911,
+      "step": 10820
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.396394378532709e-05,
+      "loss": 3.2129,
+      "step": 10825
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.395843065262089e-05,
+      "loss": 3.177,
+      "step": 10830
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.395291516519598e-05,
+      "loss": 3.1939,
+      "step": 10835
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.394739732334783e-05,
+      "loss": 3.0506,
+      "step": 10840
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.394187712737201e-05,
+      "loss": 3.0591,
+      "step": 10845
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.393635457756423e-05,
+      "loss": 3.1478,
+      "step": 10850
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.39308296742203e-05,
+      "loss": 3.1314,
+      "step": 10855
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.392530241763617e-05,
+      "loss": 3.0794,
+      "step": 10860
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.391977280810795e-05,
+      "loss": 3.1787,
+      "step": 10865
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.391424084593182e-05,
+      "loss": 3.1424,
+      "step": 10870
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.390870653140408e-05,
+      "loss": 3.2964,
+      "step": 10875
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.390316986482124e-05,
+      "loss": 3.125,
+      "step": 10880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.389763084647987e-05,
+      "loss": 3.2799,
+      "step": 10885
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.389208947667665e-05,
+      "loss": 3.2227,
+      "step": 10890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.388654575570845e-05,
+      "loss": 3.1887,
+      "step": 10895
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.388099968387221e-05,
+      "loss": 3.1643,
+      "step": 10900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.3875451261465e-05,
+      "loss": 3.2382,
+      "step": 10905
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.386990048878406e-05,
+      "loss": 3.203,
+      "step": 10910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.386434736612674e-05,
+      "loss": 3.2416,
+      "step": 10915
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.385879189379046e-05,
+      "loss": 3.1912,
+      "step": 10920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.385323407207285e-05,
+      "loss": 3.2149,
+      "step": 10925
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.384767390127159e-05,
+      "loss": 3.3415,
+      "step": 10930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.384211138168456e-05,
+      "loss": 3.2601,
+      "step": 10935
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.383654651360969e-05,
+      "loss": 3.3162,
+      "step": 10940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.38309792973451e-05,
+      "loss": 3.2576,
+      "step": 10945
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.382540973318899e-05,
+      "loss": 3.1355,
+      "step": 10950
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.381983782143971e-05,
+      "loss": 3.1795,
+      "step": 10955
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.381426356239574e-05,
+      "loss": 3.1106,
+      "step": 10960
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.380868695635567e-05,
+      "loss": 3.1844,
+      "step": 10965
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.380310800361822e-05,
+      "loss": 3.1683,
+      "step": 10970
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.379752670448224e-05,
+      "loss": 3.2391,
+      "step": 10975
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.37919430592467e-05,
+      "loss": 3.2296,
+      "step": 10980
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.37863570682107e-05,
+      "loss": 3.2572,
+      "step": 10985
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.378076873167348e-05,
+      "loss": 3.1569,
+      "step": 10990
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.377517804993434e-05,
+      "loss": 3.1809,
+      "step": 10995
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.376958502329283e-05,
+      "loss": 3.172,
+      "step": 11000
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.376398965204849e-05,
+      "loss": 3.2952,
+      "step": 11005
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.375839193650107e-05,
+      "loss": 3.2253,
+      "step": 11010
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.375279187695043e-05,
+      "loss": 3.2732,
+      "step": 11015
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.374718947369652e-05,
+      "loss": 3.1155,
+      "step": 11020
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.374158472703949e-05,
+      "loss": 3.1757,
+      "step": 11025
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.373597763727953e-05,
+      "loss": 3.2807,
+      "step": 11030
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.3730368204717e-05,
+      "loss": 3.0936,
+      "step": 11035
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.37247564296524e-05,
+      "loss": 3.1328,
+      "step": 11040
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.37191423123863e-05,
+      "loss": 3.4596,
+      "step": 11045
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.371352585321948e-05,
+      "loss": 3.175,
+      "step": 11050
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.370790705245276e-05,
+      "loss": 3.1264,
+      "step": 11055
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.370228591038713e-05,
+      "loss": 3.2213,
+      "step": 11060
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.36966624273237e-05,
+      "loss": 3.1638,
+      "step": 11065
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.36910366035637e-05,
+      "loss": 3.1152,
+      "step": 11070
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.36854084394085e-05,
+      "loss": 3.0393,
+      "step": 11075
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.367977793515956e-05,
+      "loss": 3.1805,
+      "step": 11080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.367414509111852e-05,
+      "loss": 3.1523,
+      "step": 11085
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.366850990758709e-05,
+      "loss": 3.1376,
+      "step": 11090
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.366287238486713e-05,
+      "loss": 3.1997,
+      "step": 11095
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.365723252326064e-05,
+      "loss": 3.2429,
+      "step": 11100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.365159032306971e-05,
+      "loss": 3.0749,
+      "step": 11105
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.36459457845966e-05,
+      "loss": 3.1492,
+      "step": 11110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.364029890814366e-05,
+      "loss": 3.3838,
+      "step": 11115
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.363464969401337e-05,
+      "loss": 3.1633,
+      "step": 11120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.362899814250834e-05,
+      "loss": 3.2988,
+      "step": 11125
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.36233442539313e-05,
+      "loss": 3.0692,
+      "step": 11130
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.361768802858515e-05,
+      "loss": 3.2303,
+      "step": 11135
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.361202946677284e-05,
+      "loss": 3.3134,
+      "step": 11140
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.360636856879748e-05,
+      "loss": 3.1828,
+      "step": 11145
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.360070533496232e-05,
+      "loss": 3.2186,
+      "step": 11150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.359503976557073e-05,
+      "loss": 3.2142,
+      "step": 11155
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.35893718609262e-05,
+      "loss": 3.1636,
+      "step": 11160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.35837016213323e-05,
+      "loss": 3.0616,
+      "step": 11165
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.35780290470928e-05,
+      "loss": 3.0942,
+      "step": 11170
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.357235413851158e-05,
+      "loss": 3.0339,
+      "step": 11175
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.35666768958926e-05,
+      "loss": 3.159,
+      "step": 11180
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.356099731953997e-05,
+      "loss": 3.2619,
+      "step": 11185
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.355531540975796e-05,
+      "loss": 3.0411,
+      "step": 11190
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.354963116685089e-05,
+      "loss": 3.048,
+      "step": 11195
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.354394459112328e-05,
+      "loss": 3.113,
+      "step": 11200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.353825568287973e-05,
+      "loss": 3.2689,
+      "step": 11205
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.353256444242495e-05,
+      "loss": 3.194,
+      "step": 11210
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.352687087006386e-05,
+      "loss": 3.1105,
+      "step": 11215
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.35211749661014e-05,
+      "loss": 3.2601,
+      "step": 11220
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.351547673084271e-05,
+      "loss": 3.3613,
+      "step": 11225
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.350977616459302e-05,
+      "loss": 3.1335,
+      "step": 11230
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.350407326765767e-05,
+      "loss": 3.0603,
+      "step": 11235
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.349836804034217e-05,
+      "loss": 3.1792,
+      "step": 11240
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.349266048295214e-05,
+      "loss": 3.0698,
+      "step": 11245
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.348695059579329e-05,
+      "loss": 3.1493,
+      "step": 11250
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.348123837917151e-05,
+      "loss": 3.0789,
+      "step": 11255
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.347552383339275e-05,
+      "loss": 3.1081,
+      "step": 11260
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.346980695876315e-05,
+      "loss": 3.2772,
+      "step": 11265
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.346408775558893e-05,
+      "loss": 3.1471,
+      "step": 11270
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.345836622417646e-05,
+      "loss": 3.1529,
+      "step": 11275
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.345264236483222e-05,
+      "loss": 3.2781,
+      "step": 11280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.344691617786282e-05,
+      "loss": 3.2223,
+      "step": 11285
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.3441187663575e-05,
+      "loss": 3.2007,
+      "step": 11290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.343545682227561e-05,
+      "loss": 3.1829,
+      "step": 11295
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.342972365427164e-05,
+      "loss": 3.197,
+      "step": 11300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.34239881598702e-05,
+      "loss": 3.148,
+      "step": 11305
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.341825033937852e-05,
+      "loss": 3.0734,
+      "step": 11310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.341251019310396e-05,
+      "loss": 3.1022,
+      "step": 11315
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.340676772135397e-05,
+      "loss": 3.2241,
+      "step": 11320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.340102292443622e-05,
+      "loss": 3.211,
+      "step": 11325
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.339527580265839e-05,
+      "loss": 3.1503,
+      "step": 11330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.338952635632834e-05,
+      "loss": 3.2529,
+      "step": 11335
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.33837745857541e-05,
+      "loss": 3.1577,
+      "step": 11340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.337802049124368e-05,
+      "loss": 3.1989,
+      "step": 11345
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.337226407310542e-05,
+      "loss": 3.2069,
+      "step": 11350
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.33665053316476e-05,
+      "loss": 2.9499,
+      "step": 11355
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.336074426717872e-05,
+      "loss": 3.2399,
+      "step": 11360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.335498088000739e-05,
+      "loss": 3.2425,
+      "step": 11365
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.33492151704423e-05,
+      "loss": 3.2154,
+      "step": 11370
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.334344713879235e-05,
+      "loss": 3.1952,
+      "step": 11375
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.33376767853665e-05,
+      "loss": 3.2182,
+      "step": 11380
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.333190411047383e-05,
+      "loss": 3.2266,
+      "step": 11385
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.332612911442359e-05,
+      "loss": 3.1459,
+      "step": 11390
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.332035179752512e-05,
+      "loss": 3.2465,
+      "step": 11395
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.331457216008787e-05,
+      "loss": 3.1905,
+      "step": 11400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.330879020242149e-05,
+      "loss": 3.2045,
+      "step": 11405
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.330300592483565e-05,
+      "loss": 3.2071,
+      "step": 11410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.329721932764024e-05,
+      "loss": 3.2815,
+      "step": 11415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.329143041114519e-05,
+      "loss": 3.1132,
+      "step": 11420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.328563917566063e-05,
+      "loss": 3.2158,
+      "step": 11425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.327984562149673e-05,
+      "loss": 3.1861,
+      "step": 11430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.327404974896387e-05,
+      "loss": 3.1602,
+      "step": 11435
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.326825155837254e-05,
+      "loss": 3.0944,
+      "step": 11440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.326245105003329e-05,
+      "loss": 3.1239,
+      "step": 11445
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.325664822425684e-05,
+      "loss": 3.2608,
+      "step": 11450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.325084308135403e-05,
+      "loss": 3.2732,
+      "step": 11455
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.324503562163584e-05,
+      "loss": 3.1915,
+      "step": 11460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.323922584541335e-05,
+      "loss": 3.1817,
+      "step": 11465
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.323341375299777e-05,
+      "loss": 3.2185,
+      "step": 11470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.322759934470045e-05,
+      "loss": 3.1709,
+      "step": 11475
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.322178262083283e-05,
+      "loss": 3.1387,
+      "step": 11480
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.32159635817065e-05,
+      "loss": 3.2153,
+      "step": 11485
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.321014222763315e-05,
+      "loss": 3.0675,
+      "step": 11490
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.320431855892467e-05,
+      "loss": 3.1413,
+      "step": 11495
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.319849257589295e-05,
+      "loss": 3.1408,
+      "step": 11500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.319266427885009e-05,
+      "loss": 3.4393,
+      "step": 11505
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.318683366810831e-05,
+      "loss": 3.2335,
+      "step": 11510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.318100074397995e-05,
+      "loss": 3.1748,
+      "step": 11515
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.317516550677743e-05,
+      "loss": 3.1433,
+      "step": 11520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.316932795681333e-05,
+      "loss": 3.2044,
+      "step": 11525
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.316348809440036e-05,
+      "loss": 3.1334,
+      "step": 11530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.315764591985135e-05,
+      "loss": 3.2151,
+      "step": 11535
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.315180143347924e-05,
+      "loss": 3.2988,
+      "step": 11540
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.314595463559708e-05,
+      "loss": 3.1536,
+      "step": 11545
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.314010552651811e-05,
+      "loss": 3.0641,
+      "step": 11550
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.31342541065556e-05,
+      "loss": 3.0803,
+      "step": 11555
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.312840037602303e-05,
+      "loss": 3.2276,
+      "step": 11560
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.312254433523396e-05,
+      "loss": 3.1193,
+      "step": 11565
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.311668598450205e-05,
+      "loss": 3.2609,
+      "step": 11570
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.311082532414115e-05,
+      "loss": 3.034,
+      "step": 11575
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.310496235446517e-05,
+      "loss": 3.1105,
+      "step": 11580
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.309909707578821e-05,
+      "loss": 3.1358,
+      "step": 11585
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.309322948842441e-05,
+      "loss": 3.0501,
+      "step": 11590
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.30873595926881e-05,
+      "loss": 3.246,
+      "step": 11595
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.308148738889373e-05,
+      "loss": 3.2746,
+      "step": 11600
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.307561287735583e-05,
+      "loss": 3.2525,
+      "step": 11605
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.306973605838908e-05,
+      "loss": 3.1875,
+      "step": 11610
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.30638569323083e-05,
+      "loss": 3.1745,
+      "step": 11615
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.30579754994284e-05,
+      "loss": 3.2305,
+      "step": 11620
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.305209176006445e-05,
+      "loss": 3.174,
+      "step": 11625
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.30462057145316e-05,
+      "loss": 3.3184,
+      "step": 11630
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.304031736314515e-05,
+      "loss": 3.2121,
+      "step": 11635
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.303442670622056e-05,
+      "loss": 3.2367,
+      "step": 11640
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.302853374407332e-05,
+      "loss": 3.1984,
+      "step": 11645
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.302263847701913e-05,
+      "loss": 3.1518,
+      "step": 11650
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.301674090537377e-05,
+      "loss": 3.253,
+      "step": 11655
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.301084102945317e-05,
+      "loss": 3.1761,
+      "step": 11660
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.300493884957335e-05,
+      "loss": 3.0898,
+      "step": 11665
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.299903436605049e-05,
+      "loss": 3.2059,
+      "step": 11670
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.299312757920085e-05,
+      "loss": 3.241,
+      "step": 11675
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.298721848934085e-05,
+      "loss": 3.2058,
+      "step": 11680
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.298130709678703e-05,
+      "loss": 3.1413,
+      "step": 11685
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.297539340185604e-05,
+      "loss": 3.1593,
+      "step": 11690
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.296947740486466e-05,
+      "loss": 3.219,
+      "step": 11695
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.29635591061298e-05,
+      "loss": 3.0901,
+      "step": 11700
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.295763850596845e-05,
+      "loss": 3.2039,
+      "step": 11705
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.29517156046978e-05,
+      "loss": 3.3325,
+      "step": 11710
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.29457904026351e-05,
+      "loss": 3.2524,
+      "step": 11715
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.293986290009776e-05,
+      "loss": 3.1733,
+      "step": 11720
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.293393309740328e-05,
+      "loss": 3.2337,
+      "step": 11725
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.292800099486931e-05,
+      "loss": 3.2875,
+      "step": 11730
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.292206659281362e-05,
+      "loss": 3.3694,
+      "step": 11735
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.29161298915541e-05,
+      "loss": 3.1628,
+      "step": 11740
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.291019089140874e-05,
+      "loss": 3.1935,
+      "step": 11745
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.290424959269568e-05,
+      "loss": 3.1925,
+      "step": 11750
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.289830599573319e-05,
+      "loss": 3.2131,
+      "step": 11755
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.289236010083965e-05,
+      "loss": 3.1687,
+      "step": 11760
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.288641190833356e-05,
+      "loss": 3.1597,
+      "step": 11765
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.288046141853354e-05,
+      "loss": 3.164,
+      "step": 11770
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.287450863175835e-05,
+      "loss": 3.0825,
+      "step": 11775
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.286855354832684e-05,
+      "loss": 3.1902,
+      "step": 11780
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.286259616855804e-05,
+      "loss": 3.2027,
+      "step": 11785
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.285663649277105e-05,
+      "loss": 3.1082,
+      "step": 11790
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.28506745212851e-05,
+      "loss": 3.1316,
+      "step": 11795
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.284471025441959e-05,
+      "loss": 3.0985,
+      "step": 11800
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.283874369249396e-05,
+      "loss": 3.1496,
+      "step": 11805
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.283277483582786e-05,
+      "loss": 3.3027,
+      "step": 11810
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.282680368474101e-05,
+      "loss": 3.23,
+      "step": 11815
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.282083023955326e-05,
+      "loss": 3.2685,
+      "step": 11820
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.28148545005846e-05,
+      "loss": 3.1573,
+      "step": 11825
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.280887646815512e-05,
+      "loss": 3.1784,
+      "step": 11830
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.280289614258505e-05,
+      "loss": 3.2825,
+      "step": 11835
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.279691352419474e-05,
+      "loss": 3.1901,
+      "step": 11840
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.279092861330466e-05,
+      "loss": 3.2749,
+      "step": 11845
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.278494141023541e-05,
+      "loss": 3.2569,
+      "step": 11850
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.277895191530767e-05,
+      "loss": 3.1145,
+      "step": 11855
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.277296012884234e-05,
+      "loss": 3.1588,
+      "step": 11860
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.276696605116034e-05,
+      "loss": 3.2155,
+      "step": 11865
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.276096968258276e-05,
+      "loss": 3.1521,
+      "step": 11870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.275497102343082e-05,
+      "loss": 3.1794,
+      "step": 11875
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.274897007402583e-05,
+      "loss": 3.2704,
+      "step": 11880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.274296683468925e-05,
+      "loss": 3.2922,
+      "step": 11885
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.273696130574268e-05,
+      "loss": 2.9417,
+      "step": 11890
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.273095348750778e-05,
+      "loss": 3.2367,
+      "step": 11895
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.27249433803064e-05,
+      "loss": 3.0911,
+      "step": 11900
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.271893098446045e-05,
+      "loss": 3.2096,
+      "step": 11905
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.271291630029205e-05,
+      "loss": 3.1477,
+      "step": 11910
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.270689932812331e-05,
+      "loss": 3.266,
+      "step": 11915
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.270088006827661e-05,
+      "loss": 3.1014,
+      "step": 11920
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.269485852107436e-05,
+      "loss": 3.1926,
+      "step": 11925
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.26888346868391e-05,
+      "loss": 3.247,
+      "step": 11930
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.268280856589353e-05,
+      "loss": 3.2351,
+      "step": 11935
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.267678015856045e-05,
+      "loss": 3.2524,
+      "step": 11940
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.267074946516276e-05,
+      "loss": 3.1214,
+      "step": 11945
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.266471648602353e-05,
+      "loss": 3.1277,
+      "step": 11950
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.265868122146592e-05,
+      "loss": 3.14,
+      "step": 11955
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.265264367181323e-05,
+      "loss": 3.342,
+      "step": 11960
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.264660383738886e-05,
+      "loss": 3.1209,
+      "step": 11965
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.264056171851634e-05,
+      "loss": 3.0629,
+      "step": 11970
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.263451731551935e-05,
+      "loss": 3.1468,
+      "step": 11975
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.262847062872164e-05,
+      "loss": 3.2558,
+      "step": 11980
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.262242165844714e-05,
+      "loss": 3.1727,
+      "step": 11985
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.261637040501987e-05,
+      "loss": 3.2141,
+      "step": 11990
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.261031686876398e-05,
+      "loss": 3.2461,
+      "step": 11995
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.260426105000371e-05,
+      "loss": 3.1402,
+      "step": 12000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.259820294906349e-05,
+      "loss": 3.1071,
+      "step": 12005
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.259214256626782e-05,
+      "loss": 3.2636,
+      "step": 12010
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.258607990194133e-05,
+      "loss": 3.1986,
+      "step": 12015
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.258001495640875e-05,
+      "loss": 3.1291,
+      "step": 12020
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.257394772999504e-05,
+      "loss": 3.1259,
+      "step": 12025
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.256787822302513e-05,
+      "loss": 3.1436,
+      "step": 12030
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.256180643582418e-05,
+      "loss": 3.2441,
+      "step": 12035
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.255573236871741e-05,
+      "loss": 3.1968,
+      "step": 12040
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.254965602203023e-05,
+      "loss": 3.1817,
+      "step": 12045
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.254357739608809e-05,
+      "loss": 3.1183,
+      "step": 12050
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.253749649121663e-05,
+      "loss": 3.188,
+      "step": 12055
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.253141330774156e-05,
+      "loss": 3.1013,
+      "step": 12060
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.252532784598877e-05,
+      "loss": 3.3074,
+      "step": 12065
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.251924010628422e-05,
+      "loss": 3.1537,
+      "step": 12070
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.251315008895403e-05,
+      "loss": 3.1244,
+      "step": 12075
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.250705779432438e-05,
+      "loss": 3.1574,
+      "step": 12080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.250096322272166e-05,
+      "loss": 3.237,
+      "step": 12085
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.249486637447234e-05,
+      "loss": 3.2685,
+      "step": 12090
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.248876724990296e-05,
+      "loss": 3.1316,
+      "step": 12095
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.248266584934028e-05,
+      "loss": 3.2262,
+      "step": 12100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.247656217311111e-05,
+      "loss": 3.1589,
+      "step": 12105
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.24704562215424e-05,
+      "loss": 3.1537,
+      "step": 12110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.246434799496126e-05,
+      "loss": 3.226,
+      "step": 12115
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.245823749369486e-05,
+      "loss": 3.1008,
+      "step": 12120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.245212471807054e-05,
+      "loss": 3.2052,
+      "step": 12125
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.244600966841571e-05,
+      "loss": 3.38,
+      "step": 12130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.243989234505797e-05,
+      "loss": 3.2401,
+      "step": 12135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.243377274832499e-05,
+      "loss": 3.0714,
+      "step": 12140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.242765087854457e-05,
+      "loss": 3.1833,
+      "step": 12145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.242152673604466e-05,
+      "loss": 3.1689,
+      "step": 12150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.24154003211533e-05,
+      "loss": 3.2152,
+      "step": 12155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.240927163419867e-05,
+      "loss": 3.1976,
+      "step": 12160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.240314067550906e-05,
+      "loss": 3.1752,
+      "step": 12165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.239700744541287e-05,
+      "loss": 3.3463,
+      "step": 12170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.239087194423868e-05,
+      "loss": 3.2752,
+      "step": 12175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.238473417231511e-05,
+      "loss": 3.1797,
+      "step": 12180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.237859412997096e-05,
+      "loss": 3.2087,
+      "step": 12185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.237245181753512e-05,
+      "loss": 3.2551,
+      "step": 12190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.236630723533663e-05,
+      "loss": 2.9731,
+      "step": 12195
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.236016038370464e-05,
+      "loss": 3.256,
+      "step": 12200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.235401126296841e-05,
+      "loss": 3.1829,
+      "step": 12205
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.234785987345731e-05,
+      "loss": 3.23,
+      "step": 12210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.234170621550089e-05,
+      "loss": 3.2042,
+      "step": 12215
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.233555028942875e-05,
+      "loss": 3.1692,
+      "step": 12220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.232939209557068e-05,
+      "loss": 3.2406,
+      "step": 12225
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.232323163425651e-05,
+      "loss": 3.186,
+      "step": 12230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.231706890581626e-05,
+      "loss": 3.3039,
+      "step": 12235
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.231090391058005e-05,
+      "loss": 3.0573,
+      "step": 12240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.230473664887813e-05,
+      "loss": 3.1665,
+      "step": 12245
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.229856712104083e-05,
+      "loss": 3.0744,
+      "step": 12250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.229239532739867e-05,
+      "loss": 3.1611,
+      "step": 12255
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.228622126828224e-05,
+      "loss": 3.1051,
+      "step": 12260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.228004494402225e-05,
+      "loss": 3.2276,
+      "step": 12265
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.227386635494956e-05,
+      "loss": 3.1423,
+      "step": 12270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.226768550139513e-05,
+      "loss": 3.0684,
+      "step": 12275
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.226150238369007e-05,
+      "loss": 3.2299,
+      "step": 12280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.225531700216557e-05,
+      "loss": 3.1694,
+      "step": 12285
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.224912935715296e-05,
+      "loss": 3.0442,
+      "step": 12290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.224293944898372e-05,
+      "loss": 3.2403,
+      "step": 12295
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.223674727798938e-05,
+      "loss": 3.2086,
+      "step": 12300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.22305528445017e-05,
+      "loss": 3.1157,
+      "step": 12305
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.222435614885244e-05,
+      "loss": 3.078,
+      "step": 12310
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.221815719137354e-05,
+      "loss": 3.1922,
+      "step": 12315
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.221195597239708e-05,
+      "loss": 3.2164,
+      "step": 12320
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.220575249225526e-05,
+      "loss": 3.0984,
+      "step": 12325
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.219954675128035e-05,
+      "loss": 3.15,
+      "step": 12330
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.219333874980478e-05,
+      "loss": 3.0629,
+      "step": 12335
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.218712848816109e-05,
+      "loss": 3.1013,
+      "step": 12340
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.218091596668195e-05,
+      "loss": 3.1987,
+      "step": 12345
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.217470118570014e-05,
+      "loss": 3.1455,
+      "step": 12350
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.216848414554858e-05,
+      "loss": 3.2936,
+      "step": 12355
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.216226484656028e-05,
+      "loss": 3.1993,
+      "step": 12360
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.215604328906839e-05,
+      "loss": 3.1609,
+      "step": 12365
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.214981947340622e-05,
+      "loss": 3.1895,
+      "step": 12370
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.214359339990711e-05,
+      "loss": 3.1618,
+      "step": 12375
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.213736506890458e-05,
+      "loss": 3.1531,
+      "step": 12380
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.213113448073228e-05,
+      "loss": 3.0957,
+      "step": 12385
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.212490163572396e-05,
+      "loss": 3.3018,
+      "step": 12390
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.21186665342135e-05,
+      "loss": 3.121,
+      "step": 12395
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.211242917653486e-05,
+      "loss": 3.2867,
+      "step": 12400
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.210618956302221e-05,
+      "loss": 3.2774,
+      "step": 12405
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.209994769400976e-05,
+      "loss": 3.146,
+      "step": 12410
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.209370356983184e-05,
+      "loss": 3.0648,
+      "step": 12415
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.208745719082299e-05,
+      "loss": 3.1437,
+      "step": 12420
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.208120855731774e-05,
+      "loss": 3.1613,
+      "step": 12425
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.207495766965087e-05,
+      "loss": 3.2727,
+      "step": 12430
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.20687045281572e-05,
+      "loss": 3.1471,
+      "step": 12435
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.206244913317168e-05,
+      "loss": 3.2746,
+      "step": 12440
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.20561914850294e-05,
+      "loss": 3.1144,
+      "step": 12445
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.204993158406556e-05,
+      "loss": 3.1459,
+      "step": 12450
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.204366943061549e-05,
+      "loss": 3.1822,
+      "step": 12455
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.203740502501464e-05,
+      "loss": 3.2058,
+      "step": 12460
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.203113836759856e-05,
+      "loss": 3.2432,
+      "step": 12465
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.202486945870294e-05,
+      "loss": 3.1842,
+      "step": 12470
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.201859829866358e-05,
+      "loss": 3.1859,
+      "step": 12475
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.201232488781644e-05,
+      "loss": 3.1626,
+      "step": 12480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.200604922649753e-05,
+      "loss": 3.1598,
+      "step": 12485
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.199977131504303e-05,
+      "loss": 3.2228,
+      "step": 12490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.199349115378923e-05,
+      "loss": 3.1886,
+      "step": 12495
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.198720874307254e-05,
+      "loss": 3.1003,
+      "step": 12500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.198092408322948e-05,
+      "loss": 3.2718,
+      "step": 12505
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.19746371745967e-05,
+      "loss": 3.2299,
+      "step": 12510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.196834801751102e-05,
+      "loss": 3.1207,
+      "step": 12515
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.196205661230925e-05,
+      "loss": 3.1091,
+      "step": 12520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.195576295932846e-05,
+      "loss": 3.1893,
+      "step": 12525
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.194946705890576e-05,
+      "loss": 3.177,
+      "step": 12530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.194316891137841e-05,
+      "loss": 3.1807,
+      "step": 12535
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.193686851708377e-05,
+      "loss": 3.3119,
+      "step": 12540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.193056587635933e-05,
+      "loss": 3.1253,
+      "step": 12545
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.192426098954273e-05,
+      "loss": 3.2002,
+      "step": 12550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.191795385697169e-05,
+      "loss": 3.2056,
+      "step": 12555
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.191164447898404e-05,
+      "loss": 3.1383,
+      "step": 12560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.190533285591779e-05,
+      "loss": 3.1465,
+      "step": 12565
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.189901898811102e-05,
+      "loss": 3.2332,
+      "step": 12570
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.189270287590194e-05,
+      "loss": 3.3051,
+      "step": 12575
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.188638451962887e-05,
+      "loss": 3.2628,
+      "step": 12580
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.188006391963031e-05,
+      "loss": 3.0057,
+      "step": 12585
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.18737410762448e-05,
+      "loss": 3.0868,
+      "step": 12590
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.186741598981106e-05,
+      "loss": 3.1973,
+      "step": 12595
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.186108866066786e-05,
+      "loss": 3.1987,
+      "step": 12600
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.185475908915417e-05,
+      "loss": 3.072,
+      "step": 12605
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.184842727560906e-05,
+      "loss": 3.2032,
+      "step": 12610
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.184209322037167e-05,
+      "loss": 3.1957,
+      "step": 12615
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.183575692378132e-05,
+      "loss": 3.1275,
+      "step": 12620
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.182941838617742e-05,
+      "loss": 3.24,
+      "step": 12625
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.182307760789951e-05,
+      "loss": 3.1435,
+      "step": 12630
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.181673458928723e-05,
+      "loss": 3.0986,
+      "step": 12635
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.181038933068038e-05,
+      "loss": 3.0957,
+      "step": 12640
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.180404183241884e-05,
+      "loss": 3.1369,
+      "step": 12645
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.179769209484262e-05,
+      "loss": 3.2334,
+      "step": 12650
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.179134011829186e-05,
+      "loss": 3.0723,
+      "step": 12655
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.178498590310685e-05,
+      "loss": 3.1059,
+      "step": 12660
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.177862944962792e-05,
+      "loss": 3.1311,
+      "step": 12665
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.177227075819558e-05,
+      "loss": 3.1498,
+      "step": 12670
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.176590982915046e-05,
+      "loss": 3.1359,
+      "step": 12675
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.175954666283326e-05,
+      "loss": 3.2422,
+      "step": 12680
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.175318125958488e-05,
+      "loss": 3.0175,
+      "step": 12685
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.174681361974627e-05,
+      "loss": 3.1651,
+      "step": 12690
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.174044374365853e-05,
+      "loss": 3.1278,
+      "step": 12695
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.173407163166288e-05,
+      "loss": 3.1645,
+      "step": 12700
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.172769728410064e-05,
+      "loss": 3.098,
+      "step": 12705
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.172132070131328e-05,
+      "loss": 3.1502,
+      "step": 12710
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.171494188364235e-05,
+      "loss": 3.0798,
+      "step": 12715
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.170856083142959e-05,
+      "loss": 3.2042,
+      "step": 12720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.170217754501677e-05,
+      "loss": 3.1853,
+      "step": 12725
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.169579202474585e-05,
+      "loss": 3.2055,
+      "step": 12730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.168940427095884e-05,
+      "loss": 3.1373,
+      "step": 12735
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.168301428399797e-05,
+      "loss": 3.2043,
+      "step": 12740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.167662206420549e-05,
+      "loss": 3.068,
+      "step": 12745
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.167022761192385e-05,
+      "loss": 3.2203,
+      "step": 12750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.166383092749553e-05,
+      "loss": 3.2311,
+      "step": 12755
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.165743201126322e-05,
+      "loss": 3.2072,
+      "step": 12760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.165103086356968e-05,
+      "loss": 3.2168,
+      "step": 12765
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.16446274847578e-05,
+      "loss": 3.1869,
+      "step": 12770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.16382218751706e-05,
+      "loss": 3.0816,
+      "step": 12775
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.163181403515118e-05,
+      "loss": 3.2431,
+      "step": 12780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.16254039650428e-05,
+      "loss": 3.1988,
+      "step": 12785
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.161899166518885e-05,
+      "loss": 3.1691,
+      "step": 12790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.161257713593281e-05,
+      "loss": 3.2476,
+      "step": 12795
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.160616037761825e-05,
+      "loss": 3.1494,
+      "step": 12800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.159974139058895e-05,
+      "loss": 3.2,
+      "step": 12805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.159332017518872e-05,
+      "loss": 3.1937,
+      "step": 12810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.158689673176152e-05,
+      "loss": 3.1695,
+      "step": 12815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.158047106065146e-05,
+      "loss": 3.0722,
+      "step": 12820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.157404316220272e-05,
+      "loss": 3.055,
+      "step": 12825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.156761303675965e-05,
+      "loss": 3.1851,
+      "step": 12830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.156118068466668e-05,
+      "loss": 3.1795,
+      "step": 12835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.155474610626835e-05,
+      "loss": 3.153,
+      "step": 12840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.154830930190939e-05,
+      "loss": 3.1798,
+      "step": 12845
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.154187027193454e-05,
+      "loss": 3.1957,
+      "step": 12850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.153542901668876e-05,
+      "loss": 3.1924,
+      "step": 12855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.152898553651708e-05,
+      "loss": 3.0678,
+      "step": 12860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.152253983176464e-05,
+      "loss": 3.1894,
+      "step": 12865
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.151609190277674e-05,
+      "loss": 3.142,
+      "step": 12870
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.150964174989877e-05,
+      "loss": 3.1666,
+      "step": 12875
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.150318937347623e-05,
+      "loss": 3.075,
+      "step": 12880
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.149673477385476e-05,
+      "loss": 3.0448,
+      "step": 12885
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.149027795138014e-05,
+      "loss": 3.3364,
+      "step": 12890
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.148381890639821e-05,
+      "loss": 3.2442,
+      "step": 12895
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.147735763925498e-05,
+      "loss": 3.1761,
+      "step": 12900
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.147089415029655e-05,
+      "loss": 3.163,
+      "step": 12905
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.146442843986914e-05,
+      "loss": 3.0371,
+      "step": 12910
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.145796050831912e-05,
+      "loss": 3.1758,
+      "step": 12915
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.145149035599295e-05,
+      "loss": 3.1541,
+      "step": 12920
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.144501798323722e-05,
+      "loss": 3.1886,
+      "step": 12925
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.143854339039861e-05,
+      "loss": 3.1493,
+      "step": 12930
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.143206657782397e-05,
+      "loss": 3.0523,
+      "step": 12935
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.142558754586024e-05,
+      "loss": 3.235,
+      "step": 12940
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.141910629485448e-05,
+      "loss": 3.2276,
+      "step": 12945
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.141262282515386e-05,
+      "loss": 3.1471,
+      "step": 12950
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.14061371371057e-05,
+      "loss": 3.1906,
+      "step": 12955
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.139964923105741e-05,
+      "loss": 3.2437,
+      "step": 12960
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.13931591073565e-05,
+      "loss": 3.2329,
+      "step": 12965
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.138666676635068e-05,
+      "loss": 3.2697,
+      "step": 12970
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.138017220838768e-05,
+      "loss": 3.2555,
+      "step": 12975
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.137367543381541e-05,
+      "loss": 3.0707,
+      "step": 12980
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.136717644298187e-05,
+      "loss": 3.1753,
+      "step": 12985
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.136067523623523e-05,
+      "loss": 3.1905,
+      "step": 12990
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.135417181392369e-05,
+      "loss": 3.1668,
+      "step": 12995
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.134766617639565e-05,
+      "loss": 3.1088,
+      "step": 13000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.134115832399959e-05,
+      "loss": 3.1061,
+      "step": 13005
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.13346482570841e-05,
+      "loss": 3.1796,
+      "step": 13010
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.132813597599792e-05,
+      "loss": 3.1907,
+      "step": 13015
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.13216214810899e-05,
+      "loss": 3.1443,
+      "step": 13020
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.1315104772709e-05,
+      "loss": 3.1726,
+      "step": 13025
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.130858585120427e-05,
+      "loss": 3.3164,
+      "step": 13030
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.130206471692494e-05,
+      "loss": 3.1047,
+      "step": 13035
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.129554137022033e-05,
+      "loss": 3.1916,
+      "step": 13040
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.128901581143984e-05,
+      "loss": 3.132,
+      "step": 13045
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.128248804093306e-05,
+      "loss": 3.111,
+      "step": 13050
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.127595805904967e-05,
+      "loss": 3.1955,
+      "step": 13055
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.126942586613943e-05,
+      "loss": 3.2584,
+      "step": 13060
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.126289146255225e-05,
+      "loss": 3.1837,
+      "step": 13065
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.125635484863817e-05,
+      "loss": 3.2282,
+      "step": 13070
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.124981602474735e-05,
+      "loss": 3.2083,
+      "step": 13075
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.124327499123004e-05,
+      "loss": 3.1386,
+      "step": 13080
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.123673174843662e-05,
+      "loss": 3.2478,
+      "step": 13085
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.123018629671759e-05,
+      "loss": 3.1624,
+      "step": 13090
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.122363863642357e-05,
+      "loss": 3.183,
+      "step": 13095
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.121708876790532e-05,
+      "loss": 3.0989,
+      "step": 13100
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.121053669151368e-05,
+      "loss": 3.1316,
+      "step": 13105
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.120398240759962e-05,
+      "loss": 3.2023,
+      "step": 13110
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.119742591651423e-05,
+      "loss": 3.0062,
+      "step": 13115
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.119086721860873e-05,
+      "loss": 3.1842,
+      "step": 13120
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.118430631423446e-05,
+      "loss": 3.3393,
+      "step": 13125
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.117774320374284e-05,
+      "loss": 3.1893,
+      "step": 13130
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.117117788748546e-05,
+      "loss": 3.1538,
+      "step": 13135
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.1164610365814e-05,
+      "loss": 3.0948,
+      "step": 13140
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.115804063908026e-05,
+      "loss": 3.156,
+      "step": 13145
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.115146870763615e-05,
+      "loss": 3.173,
+      "step": 13150
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.11448945718337e-05,
+      "loss": 3.1155,
+      "step": 13155
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.113831823202512e-05,
+      "loss": 3.1183,
+      "step": 13160
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.113173968856262e-05,
+      "loss": 3.1147,
+      "step": 13165
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.112515894179863e-05,
+      "loss": 3.1721,
+      "step": 13170
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.111857599208564e-05,
+      "loss": 3.1488,
+      "step": 13175
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.111199083977629e-05,
+      "loss": 3.1293,
+      "step": 13180
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.110540348522333e-05,
+      "loss": 3.1407,
+      "step": 13185
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.109881392877961e-05,
+      "loss": 3.1425,
+      "step": 13190
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.109222217079813e-05,
+      "loss": 3.1654,
+      "step": 13195
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.108562821163198e-05,
+      "loss": 3.0816,
+      "step": 13200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.107903205163438e-05,
+      "loss": 3.2016,
+      "step": 13205
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.107243369115868e-05,
+      "loss": 3.0213,
+      "step": 13210
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.10658331305583e-05,
+      "loss": 3.0876,
+      "step": 13215
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.105923037018684e-05,
+      "loss": 3.118,
+      "step": 13220
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.105262541039797e-05,
+      "loss": 3.101,
+      "step": 13225
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.104601825154553e-05,
+      "loss": 3.2303,
+      "step": 13230
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.103940889398342e-05,
+      "loss": 3.1633,
+      "step": 13235
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.103279733806567e-05,
+      "loss": 3.2149,
+      "step": 13240
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.102618358414648e-05,
+      "loss": 3.1078,
+      "step": 13245
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.101956763258008e-05,
+      "loss": 3.2421,
+      "step": 13250
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.101294948372091e-05,
+      "loss": 3.1559,
+      "step": 13255
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.100632913792346e-05,
+      "loss": 3.1289,
+      "step": 13260
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.099970659554237e-05,
+      "loss": 3.1864,
+      "step": 13265
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.099308185693241e-05,
+      "loss": 2.9943,
+      "step": 13270
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.098645492244841e-05,
+      "loss": 3.0761,
+      "step": 13275
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.097982579244537e-05,
+      "loss": 3.2372,
+      "step": 13280
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.097319446727839e-05,
+      "loss": 3.1734,
+      "step": 13285
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.096656094730268e-05,
+      "loss": 3.1729,
+      "step": 13290
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.09599252328736e-05,
+      "loss": 3.1413,
+      "step": 13295
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.09532873243466e-05,
+      "loss": 3.0994,
+      "step": 13300
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.094664722207724e-05,
+      "loss": 3.0629,
+      "step": 13305
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.094000492642122e-05,
+      "loss": 2.9223,
+      "step": 13310
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.093336043773435e-05,
+      "loss": 3.0772,
+      "step": 13315
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.092671375637254e-05,
+      "loss": 3.0503,
+      "step": 13320
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.092006488269183e-05,
+      "loss": 3.1039,
+      "step": 13325
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.09134138170484e-05,
+      "loss": 3.1894,
+      "step": 13330
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.090676055979853e-05,
+      "loss": 3.233,
+      "step": 13335
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.090010511129861e-05,
+      "loss": 3.2699,
+      "step": 13340
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.089344747190511e-05,
+      "loss": 3.1615,
+      "step": 13345
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.088678764197472e-05,
+      "loss": 3.1542,
+      "step": 13350
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.088012562186416e-05,
+      "loss": 3.1957,
+      "step": 13355
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.087346141193029e-05,
+      "loss": 3.1846,
+      "step": 13360
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.08667950125301e-05,
+      "loss": 3.1523,
+      "step": 13365
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.086012642402069e-05,
+      "loss": 3.1114,
+      "step": 13370
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.085345564675926e-05,
+      "loss": 3.0552,
+      "step": 13375
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.084678268110316e-05,
+      "loss": 3.0996,
+      "step": 13380
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.084010752740983e-05,
+      "loss": 3.1254,
+      "step": 13385
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.083343018603685e-05,
+      "loss": 3.355,
+      "step": 13390
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.08267506573419e-05,
+      "loss": 3.1491,
+      "step": 13395
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.082006894168275e-05,
+      "loss": 3.0534,
+      "step": 13400
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.081338503941738e-05,
+      "loss": 3.4148,
+      "step": 13405
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.080669895090378e-05,
+      "loss": 3.2259,
+      "step": 13410
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.080001067650013e-05,
+      "loss": 3.2727,
+      "step": 13415
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.079332021656467e-05,
+      "loss": 3.1488,
+      "step": 13420
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.078662757145582e-05,
+      "loss": 3.2914,
+      "step": 13425
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.077993274153205e-05,
+      "loss": 3.1659,
+      "step": 13430
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.077323572715201e-05,
+      "loss": 3.0865,
+      "step": 13435
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.076653652867444e-05,
+      "loss": 3.1898,
+      "step": 13440
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.075983514645818e-05,
+      "loss": 3.0644,
+      "step": 13445
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.07531315808622e-05,
+      "loss": 3.2102,
+      "step": 13450
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.074642583224562e-05,
+      "loss": 3.2071,
+      "step": 13455
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.073971790096761e-05,
+      "loss": 3.1279,
+      "step": 13460
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.073300778738752e-05,
+      "loss": 3.2279,
+      "step": 13465
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.072629549186475e-05,
+      "loss": 3.0532,
+      "step": 13470
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.07195810147589e-05,
+      "loss": 3.1392,
+      "step": 13475
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.071286435642963e-05,
+      "loss": 3.0843,
+      "step": 13480
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.070614551723672e-05,
+      "loss": 3.2195,
+      "step": 13485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.06994244975401e-05,
+      "loss": 3.1225,
+      "step": 13490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.069270129769978e-05,
+      "loss": 3.0033,
+      "step": 13495
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.06859759180759e-05,
+      "loss": 3.1562,
+      "step": 13500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.067924835902873e-05,
+      "loss": 3.1545,
+      "step": 13505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.067251862091862e-05,
+      "loss": 3.236,
+      "step": 13510
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.066578670410608e-05,
+      "loss": 3.1588,
+      "step": 13515
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.065905260895173e-05,
+      "loss": 3.0693,
+      "step": 13520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.065231633581628e-05,
+      "loss": 3.2141,
+      "step": 13525
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.064557788506055e-05,
+      "loss": 3.0379,
+      "step": 13530
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.063883725704554e-05,
+      "loss": 3.1132,
+      "step": 13535
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.06320944521323e-05,
+      "loss": 3.3204,
+      "step": 13540
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.062534947068201e-05,
+      "loss": 3.1146,
+      "step": 13545
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.061860231305602e-05,
+      "loss": 3.101,
+      "step": 13550
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.061185297961572e-05,
+      "loss": 3.1753,
+      "step": 13555
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.060510147072266e-05,
+      "loss": 3.1588,
+      "step": 13560
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.05983477867385e-05,
+      "loss": 3.2222,
+      "step": 13565
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.0591591928025e-05,
+      "loss": 3.2729,
+      "step": 13570
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.058483389494408e-05,
+      "loss": 3.0588,
+      "step": 13575
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.057807368785771e-05,
+      "loss": 3.1867,
+      "step": 13580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.057131130712804e-05,
+      "loss": 3.1553,
+      "step": 13585
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.056454675311732e-05,
+      "loss": 3.0128,
+      "step": 13590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.055778002618787e-05,
+      "loss": 3.2062,
+      "step": 13595
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.055101112670219e-05,
+      "loss": 3.231,
+      "step": 13600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.054424005502286e-05,
+      "loss": 3.1139,
+      "step": 13605
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.05374668115126e-05,
+      "loss": 3.1908,
+      "step": 13610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.053069139653421e-05,
+      "loss": 3.2695,
+      "step": 13615
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.052391381045064e-05,
+      "loss": 3.201,
+      "step": 13620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.051713405362495e-05,
+      "loss": 3.1028,
+      "step": 13625
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.051035212642028e-05,
+      "loss": 3.1792,
+      "step": 13630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.050356802919995e-05,
+      "loss": 3.2245,
+      "step": 13635
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.049678176232737e-05,
+      "loss": 2.997,
+      "step": 13640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.048999332616604e-05,
+      "loss": 3.1159,
+      "step": 13645
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.048320272107957e-05,
+      "loss": 3.2879,
+      "step": 13650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.047640994743177e-05,
+      "loss": 3.2372,
+      "step": 13655
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.046961500558647e-05,
+      "loss": 3.2735,
+      "step": 13660
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.046281789590767e-05,
+      "loss": 3.1823,
+      "step": 13665
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.045601861875945e-05,
+      "loss": 3.0461,
+      "step": 13670
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.044921717450607e-05,
+      "loss": 3.2066,
+      "step": 13675
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.044241356351179e-05,
+      "loss": 3.0291,
+      "step": 13680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.043560778614113e-05,
+      "loss": 3.0333,
+      "step": 13685
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.042879984275863e-05,
+      "loss": 3.1381,
+      "step": 13690
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.042198973372895e-05,
+      "loss": 3.3353,
+      "step": 13695
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.041517745941692e-05,
+      "loss": 3.2258,
+      "step": 13700
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.040836302018743e-05,
+      "loss": 3.3662,
+      "step": 13705
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.040154641640551e-05,
+      "loss": 3.1441,
+      "step": 13710
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.039472764843631e-05,
+      "loss": 3.0824,
+      "step": 13715
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.038790671664509e-05,
+      "loss": 3.1909,
+      "step": 13720
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.038108362139723e-05,
+      "loss": 3.0921,
+      "step": 13725
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.037425836305821e-05,
+      "loss": 3.1004,
+      "step": 13730
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.036743094199365e-05,
+      "loss": 3.0544,
+      "step": 13735
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.036060135856927e-05,
+      "loss": 3.0452,
+      "step": 13740
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.035376961315091e-05,
+      "loss": 3.0742,
+      "step": 13745
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.034693570610455e-05,
+      "loss": 3.0865,
+      "step": 13750
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.034009963779622e-05,
+      "loss": 3.1258,
+      "step": 13755
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.033326140859212e-05,
+      "loss": 3.1621,
+      "step": 13760
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.032642101885857e-05,
+      "loss": 3.1298,
+      "step": 13765
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.031957846896197e-05,
+      "loss": 3.1534,
+      "step": 13770
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.031273375926888e-05,
+      "loss": 3.1953,
+      "step": 13775
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.030588689014592e-05,
+      "loss": 3.1238,
+      "step": 13780
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.029903786195986e-05,
+      "loss": 3.2912,
+      "step": 13785
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.02921866750776e-05,
+      "loss": 3.1015,
+      "step": 13790
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.028533332986612e-05,
+      "loss": 3.1963,
+      "step": 13795
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.027847782669254e-05,
+      "loss": 3.1595,
+      "step": 13800
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.027162016592408e-05,
+      "loss": 3.1526,
+      "step": 13805
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.026476034792811e-05,
+      "loss": 3.1917,
+      "step": 13810
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.025789837307206e-05,
+      "loss": 3.0979,
+      "step": 13815
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.025103424172351e-05,
+      "loss": 3.3543,
+      "step": 13820
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.024416795425017e-05,
+      "loss": 3.1162,
+      "step": 13825
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.023729951101982e-05,
+      "loss": 3.2038,
+      "step": 13830
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.02304289124004e-05,
+      "loss": 3.1558,
+      "step": 13835
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.022355615875995e-05,
+      "loss": 3.2786,
+      "step": 13840
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.02166812504666e-05,
+      "loss": 3.2359,
+      "step": 13845
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.020980418788862e-05,
+      "loss": 3.1537,
+      "step": 13850
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.020292497139443e-05,
+      "loss": 3.1784,
+      "step": 13855
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.01960436013525e-05,
+      "loss": 3.2082,
+      "step": 13860
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.018916007813143e-05,
+      "loss": 3.1436,
+      "step": 13865
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.018227440209997e-05,
+      "loss": 3.1275,
+      "step": 13870
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.017538657362697e-05,
+      "loss": 3.1091,
+      "step": 13875
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.016849659308138e-05,
+      "loss": 3.1301,
+      "step": 13880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.016160446083228e-05,
+      "loss": 3.1598,
+      "step": 13885
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.015471017724884e-05,
+      "loss": 3.0852,
+      "step": 13890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.01478137427004e-05,
+      "loss": 3.2063,
+      "step": 13895
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.014091515755634e-05,
+      "loss": 3.083,
+      "step": 13900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.013401442218624e-05,
+      "loss": 3.1694,
+      "step": 13905
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.012711153695972e-05,
+      "loss": 3.0721,
+      "step": 13910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.012020650224657e-05,
+      "loss": 3.2064,
+      "step": 13915
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.011329931841665e-05,
+      "loss": 3.0418,
+      "step": 13920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.010638998583996e-05,
+      "loss": 3.236,
+      "step": 13925
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.009947850488663e-05,
+      "loss": 3.0877,
+      "step": 13930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.009256487592685e-05,
+      "loss": 3.0622,
+      "step": 13935
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.008564909933101e-05,
+      "loss": 3.1485,
+      "step": 13940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.007873117546952e-05,
+      "loss": 3.1761,
+      "step": 13945
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.007181110471299e-05,
+      "loss": 3.1609,
+      "step": 13950
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.00648888874321e-05,
+      "loss": 3.1786,
+      "step": 13955
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.005796452399763e-05,
+      "loss": 3.1904,
+      "step": 13960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.005103801478051e-05,
+      "loss": 3.0921,
+      "step": 13965
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.004410936015177e-05,
+      "loss": 3.0136,
+      "step": 13970
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.003717856048257e-05,
+      "loss": 2.9849,
+      "step": 13975
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.003024561614417e-05,
+      "loss": 3.0531,
+      "step": 13980
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.002331052750792e-05,
+      "loss": 3.0929,
+      "step": 13985
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.001637329494533e-05,
+      "loss": 3.2056,
+      "step": 13990
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.000943391882802e-05,
+      "loss": 3.1931,
+      "step": 13995
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.00024923995277e-05,
+      "loss": 3.1919,
+      "step": 14000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.999554873741618e-05,
+      "loss": 3.0966,
+      "step": 14005
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.998860293286544e-05,
+      "loss": 3.1006,
+      "step": 14010
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.998165498624755e-05,
+      "loss": 3.2497,
+      "step": 14015
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.997470489793468e-05,
+      "loss": 3.0586,
+      "step": 14020
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.996775266829911e-05,
+      "loss": 3.2807,
+      "step": 14025
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.996079829771327e-05,
+      "loss": 3.0574,
+      "step": 14030
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.995384178654967e-05,
+      "loss": 3.1766,
+      "step": 14035
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.994688313518099e-05,
+      "loss": 3.1952,
+      "step": 14040
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.993992234397991e-05,
+      "loss": 3.3005,
+      "step": 14045
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.993295941331936e-05,
+      "loss": 3.146,
+      "step": 14050
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.99259943435723e-05,
+      "loss": 3.204,
+      "step": 14055
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.991902713511182e-05,
+      "loss": 3.1466,
+      "step": 14060
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.991205778831115e-05,
+      "loss": 3.1241,
+      "step": 14065
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.990508630354362e-05,
+      "loss": 3.1122,
+      "step": 14070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.989811268118265e-05,
+      "loss": 3.0168,
+      "step": 14075
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.989113692160181e-05,
+      "loss": 3.1159,
+      "step": 14080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.988415902517475e-05,
+      "loss": 3.1199,
+      "step": 14085
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.987717899227527e-05,
+      "loss": 3.1689,
+      "step": 14090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.987019682327729e-05,
+      "loss": 3.19,
+      "step": 14095
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.986321251855479e-05,
+      "loss": 3.1468,
+      "step": 14100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.98562260784819e-05,
+      "loss": 3.1892,
+      "step": 14105
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.98492375034329e-05,
+      "loss": 3.2953,
+      "step": 14110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.98422467937821e-05,
+      "loss": 3.1414,
+      "step": 14115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.9835253949904e-05,
+      "loss": 3.0654,
+      "step": 14120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.982825897217315e-05,
+      "loss": 3.1681,
+      "step": 14125
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.982126186096429e-05,
+      "loss": 3.2366,
+      "step": 14130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.981426261665222e-05,
+      "loss": 3.1783,
+      "step": 14135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.980726123961187e-05,
+      "loss": 3.282,
+      "step": 14140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.980025773021827e-05,
+      "loss": 3.0466,
+      "step": 14145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.979325208884661e-05,
+      "loss": 3.0677,
+      "step": 14150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.978624431587211e-05,
+      "loss": 3.0849,
+      "step": 14155
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.977923441167019e-05,
+      "loss": 3.1154,
+      "step": 14160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.977222237661633e-05,
+      "loss": 3.1658,
+      "step": 14165
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.976520821108617e-05,
+      "loss": 3.0963,
+      "step": 14170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.975819191545543e-05,
+      "loss": 3.1255,
+      "step": 14175
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.975117349009992e-05,
+      "loss": 3.0314,
+      "step": 14180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.974415293539562e-05,
+      "loss": 3.1734,
+      "step": 14185
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.97371302517186e-05,
+      "loss": 3.2081,
+      "step": 14190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.973010543944504e-05,
+      "loss": 3.1809,
+      "step": 14195
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.972307849895126e-05,
+      "loss": 3.1258,
+      "step": 14200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.971604943061362e-05,
+      "loss": 3.1084,
+      "step": 14205
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.97090182348087e-05,
+      "loss": 3.1599,
+      "step": 14210
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.970198491191311e-05,
+      "loss": 3.1623,
+      "step": 14215
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.96949494623036e-05,
+      "loss": 3.2142,
+      "step": 14220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.968791188635706e-05,
+      "loss": 3.0776,
+      "step": 14225
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.968087218445044e-05,
+      "loss": 3.1465,
+      "step": 14230
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.967383035696089e-05,
+      "loss": 3.2111,
+      "step": 14235
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.966678640426555e-05,
+      "loss": 3.2122,
+      "step": 14240
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.96597403267418e-05,
+      "loss": 3.2703,
+      "step": 14245
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.965269212476703e-05,
+      "loss": 3.2125,
+      "step": 14250
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.964564179871882e-05,
+      "loss": 3.1247,
+      "step": 14255
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.963858934897484e-05,
+      "loss": 3.1791,
+      "step": 14260
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.963153477591285e-05,
+      "loss": 3.1877,
+      "step": 14265
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.962447807991075e-05,
+      "loss": 3.1947,
+      "step": 14270
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.961741926134655e-05,
+      "loss": 3.0562,
+      "step": 14275
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.961035832059835e-05,
+      "loss": 3.1316,
+      "step": 14280
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.960329525804439e-05,
+      "loss": 3.1089,
+      "step": 14285
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.959623007406303e-05,
+      "loss": 3.2776,
+      "step": 14290
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.958916276903273e-05,
+      "loss": 3.1577,
+      "step": 14295
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.958209334333206e-05,
+      "loss": 3.0687,
+      "step": 14300
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.95750217973397e-05,
+      "loss": 3.1822,
+      "step": 14305
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.956794813143445e-05,
+      "loss": 3.1329,
+      "step": 14310
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.956087234599523e-05,
+      "loss": 3.1496,
+      "step": 14315
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.955379444140108e-05,
+      "loss": 3.1488,
+      "step": 14320
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.954671441803114e-05,
+      "loss": 2.9593,
+      "step": 14325
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.953963227626463e-05,
+      "loss": 3.2086,
+      "step": 14330
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.953254801648095e-05,
+      "loss": 3.0098,
+      "step": 14335
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.952546163905957e-05,
+      "loss": 3.0692,
+      "step": 14340
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.95183731443801e-05,
+      "loss": 3.1651,
+      "step": 14345
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.951128253282224e-05,
+      "loss": 3.1112,
+      "step": 14350
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.950418980476582e-05,
+      "loss": 3.1961,
+      "step": 14355
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.949709496059077e-05,
+      "loss": 3.2352,
+      "step": 14360
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.948999800067712e-05,
+      "loss": 3.1279,
+      "step": 14365
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.948289892540507e-05,
+      "loss": 3.039,
+      "step": 14370
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.947579773515486e-05,
+      "loss": 3.0509,
+      "step": 14375
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.946869443030691e-05,
+      "loss": 3.3307,
+      "step": 14380
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.946158901124171e-05,
+      "loss": 3.1597,
+      "step": 14385
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.945448147833985e-05,
+      "loss": 3.1098,
+      "step": 14390
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.94473718319821e-05,
+      "loss": 3.1112,
+      "step": 14395
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.944026007254928e-05,
+      "loss": 3.1519,
+      "step": 14400
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.943314620042234e-05,
+      "loss": 3.1443,
+      "step": 14405
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.942603021598237e-05,
+      "loss": 3.0891,
+      "step": 14410
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.941891211961054e-05,
+      "loss": 3.0261,
+      "step": 14415
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.941179191168814e-05,
+      "loss": 3.141,
+      "step": 14420
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.940466959259659e-05,
+      "loss": 3.2329,
+      "step": 14425
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.939754516271739e-05,
+      "loss": 3.2377,
+      "step": 14430
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.939041862243218e-05,
+      "loss": 3.227,
+      "step": 14435
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.938328997212272e-05,
+      "loss": 3.1683,
+      "step": 14440
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.937615921217086e-05,
+      "loss": 3.0683,
+      "step": 14445
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.93690263429586e-05,
+      "loss": 3.136,
+      "step": 14450
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.936189136486798e-05,
+      "loss": 3.07,
+      "step": 14455
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.935475427828122e-05,
+      "loss": 3.2374,
+      "step": 14460
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.934761508358064e-05,
+      "loss": 3.2021,
+      "step": 14465
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.934047378114865e-05,
+      "loss": 3.1688,
+      "step": 14470
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.93333303713678e-05,
+      "loss": 3.0761,
+      "step": 14475
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.932618485462074e-05,
+      "loss": 3.0631,
+      "step": 14480
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.931903723129022e-05,
+      "loss": 3.1265,
+      "step": 14485
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.931188750175914e-05,
+      "loss": 3.1213,
+      "step": 14490
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.930473566641046e-05,
+      "loss": 3.2958,
+      "step": 14495
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.929758172562733e-05,
+      "loss": 3.0922,
+      "step": 14500
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.92904256797929e-05,
+      "loss": 3.0504,
+      "step": 14505
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.928326752929053e-05,
+      "loss": 3.0952,
+      "step": 14510
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.927610727450366e-05,
+      "loss": 3.0809,
+      "step": 14515
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.926894491581585e-05,
+      "loss": 3.1134,
+      "step": 14520
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.926178045361075e-05,
+      "loss": 3.2284,
+      "step": 14525
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.925461388827216e-05,
+      "loss": 3.175,
+      "step": 14530
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.924744522018396e-05,
+      "loss": 3.1823,
+      "step": 14535
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.924027444973013e-05,
+      "loss": 3.103,
+      "step": 14540
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.923310157729482e-05,
+      "loss": 3.1841,
+      "step": 14545
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.922592660326226e-05,
+      "loss": 3.162,
+      "step": 14550
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.921874952801676e-05,
+      "loss": 3.252,
+      "step": 14555
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.921157035194278e-05,
+      "loss": 3.2366,
+      "step": 14560
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.920438907542492e-05,
+      "loss": 3.1522,
+      "step": 14565
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.919720569884783e-05,
+      "loss": 3.1314,
+      "step": 14570
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.919002022259631e-05,
+      "loss": 3.1438,
+      "step": 14575
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.918283264705526e-05,
+      "loss": 3.1422,
+      "step": 14580
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.91756429726097e-05,
+      "loss": 3.1481,
+      "step": 14585
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.916845119964477e-05,
+      "loss": 3.176,
+      "step": 14590
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.916125732854567e-05,
+      "loss": 3.1987,
+      "step": 14595
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.91540613596978e-05,
+      "loss": 3.1874,
+      "step": 14600
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.91468632934866e-05,
+      "loss": 3.1943,
+      "step": 14605
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 8.913966313029766e-05,
+      "loss": 3.2071,
+      "step": 14610
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.91324608705167e-05,
+      "loss": 3.1552,
+      "step": 14615
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.912525651452945e-05,
+      "loss": 3.2049,
+      "step": 14620
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.911805006272188e-05,
+      "loss": 3.0897,
+      "step": 14625
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.911084151548001e-05,
+      "loss": 3.1076,
+      "step": 14630
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.910363087318997e-05,
+      "loss": 3.1931,
+      "step": 14635
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.909641813623801e-05,
+      "loss": 3.1137,
+      "step": 14640
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.908920330501051e-05,
+      "loss": 3.1612,
+      "step": 14645
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.908198637989394e-05,
+      "loss": 3.0687,
+      "step": 14650
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.907476736127487e-05,
+      "loss": 3.2163,
+      "step": 14655
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.906754624954002e-05,
+      "loss": 3.1841,
+      "step": 14660
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.90603230450762e-05,
+      "loss": 3.0262,
+      "step": 14665
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.905309774827034e-05,
+      "loss": 3.1664,
+      "step": 14670
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.904587035950945e-05,
+      "loss": 3.2181,
+      "step": 14675
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.903864087918072e-05,
+      "loss": 3.1181,
+      "step": 14680
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.903140930767137e-05,
+      "loss": 3.0053,
+      "step": 14685
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.902417564536882e-05,
+      "loss": 3.2067,
+      "step": 14690
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.901693989266052e-05,
+      "loss": 3.1976,
+      "step": 14695
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.900970204993408e-05,
+      "loss": 3.154,
+      "step": 14700
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.900246211757721e-05,
+      "loss": 3.0934,
+      "step": 14705
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.899522009597772e-05,
+      "loss": 3.1668,
+      "step": 14710
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.898797598552354e-05,
+      "loss": 3.1045,
+      "step": 14715
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.898072978660274e-05,
+      "loss": 3.1163,
+      "step": 14720
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.897348149960346e-05,
+      "loss": 3.1955,
+      "step": 14725
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.896623112491395e-05,
+      "loss": 3.1125,
+      "step": 14730
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.895897866292262e-05,
+      "loss": 3.2273,
+      "step": 14735
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.895172411401796e-05,
+      "loss": 3.1112,
+      "step": 14740
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.894446747858856e-05,
+      "loss": 3.2283,
+      "step": 14745
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.893720875702313e-05,
+      "loss": 3.1903,
+      "step": 14750
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.892994794971055e-05,
+      "loss": 3.0033,
+      "step": 14755
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.892268505703967e-05,
+      "loss": 3.1287,
+      "step": 14760
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.891542007939961e-05,
+      "loss": 3.2693,
+      "step": 14765
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.890815301717953e-05,
+      "loss": 3.2824,
+      "step": 14770
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.890088387076867e-05,
+      "loss": 3.1159,
+      "step": 14775
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.889361264055644e-05,
+      "loss": 3.0386,
+      "step": 14780
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.88863393269323e-05,
+      "loss": 3.2047,
+      "step": 14785
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.887906393028594e-05,
+      "loss": 3.035,
+      "step": 14790
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.8871786451007e-05,
+      "loss": 3.1702,
+      "step": 14795
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.886450688948535e-05,
+      "loss": 3.1119,
+      "step": 14800
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.885722524611091e-05,
+      "loss": 3.2628,
+      "step": 14805
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.884994152127378e-05,
+      "loss": 3.1195,
+      "step": 14810
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.884265571536407e-05,
+      "loss": 3.1736,
+      "step": 14815
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.883536782877209e-05,
+      "loss": 3.1109,
+      "step": 14820
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.882807786188825e-05,
+      "loss": 3.2484,
+      "step": 14825
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.882078581510302e-05,
+      "loss": 3.2293,
+      "step": 14830
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.881349168880701e-05,
+      "loss": 3.2117,
+      "step": 14835
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.880619548339095e-05,
+      "loss": 3.0624,
+      "step": 14840
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.87988971992457e-05,
+      "loss": 3.1319,
+      "step": 14845
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.879159683676217e-05,
+      "loss": 3.2506,
+      "step": 14850
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.878429439633145e-05,
+      "loss": 3.251,
+      "step": 14855
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.877698987834468e-05,
+      "loss": 3.0883,
+      "step": 14860
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.876968328319314e-05,
+      "loss": 3.1292,
+      "step": 14865
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.876237461126828e-05,
+      "loss": 3.2183,
+      "step": 14870
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.875506386296154e-05,
+      "loss": 3.1071,
+      "step": 14875
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.874775103866455e-05,
+      "loss": 3.1481,
+      "step": 14880
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.874043613876903e-05,
+      "loss": 3.0636,
+      "step": 14885
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.873311916366684e-05,
+      "loss": 3.1119,
+      "step": 14890
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.872580011374991e-05,
+      "loss": 3.0377,
+      "step": 14895
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.871847898941031e-05,
+      "loss": 3.1334,
+      "step": 14900
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.871115579104021e-05,
+      "loss": 3.1483,
+      "step": 14905
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.870383051903188e-05,
+      "loss": 3.1949,
+      "step": 14910
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.869650317377771e-05,
+      "loss": 2.9866,
+      "step": 14915
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.868917375567022e-05,
+      "loss": 3.0791,
+      "step": 14920
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.868184226510201e-05,
+      "loss": 3.1673,
+      "step": 14925
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.867450870246581e-05,
+      "loss": 3.1918,
+      "step": 14930
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.866717306815446e-05,
+      "loss": 3.3324,
+      "step": 14935
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.865983536256092e-05,
+      "loss": 3.2103,
+      "step": 14940
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.865249558607822e-05,
+      "loss": 2.9528,
+      "step": 14945
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.864515373909954e-05,
+      "loss": 2.966,
+      "step": 14950
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.863780982201817e-05,
+      "loss": 2.9945,
+      "step": 14955
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.863046383522751e-05,
+      "loss": 3.0508,
+      "step": 14960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.862311577912103e-05,
+      "loss": 3.1282,
+      "step": 14965
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.861576565409235e-05,
+      "loss": 3.0827,
+      "step": 14970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.860841346053521e-05,
+      "loss": 3.2132,
+      "step": 14975
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.860105919884345e-05,
+      "loss": 3.1937,
+      "step": 14980
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.8593702869411e-05,
+      "loss": 3.1956,
+      "step": 14985
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.85863444726319e-05,
+      "loss": 3.0375,
+      "step": 14990
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.857898400890037e-05,
+      "loss": 3.1812,
+      "step": 14995
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.857162147861063e-05,
+      "loss": 2.9875,
+      "step": 15000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.856425688215711e-05,
+      "loss": 3.1124,
+      "step": 15005
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.855689021993428e-05,
+      "loss": 3.1822,
+      "step": 15010
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.854952149233675e-05,
+      "loss": 2.9891,
+      "step": 15015
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.854215069975926e-05,
+      "loss": 3.1671,
+      "step": 15020
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.853477784259662e-05,
+      "loss": 3.0189,
+      "step": 15025
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.85274029212438e-05,
+      "loss": 3.1119,
+      "step": 15030
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.852002593609583e-05,
+      "loss": 3.1746,
+      "step": 15035
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.851264688754788e-05,
+      "loss": 3.1741,
+      "step": 15040
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.850526577599521e-05,
+      "loss": 3.1598,
+      "step": 15045
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.849788260183323e-05,
+      "loss": 3.1595,
+      "step": 15050
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.84904973654574e-05,
+      "loss": 3.1666,
+      "step": 15055
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.848311006726335e-05,
+      "loss": 3.2753,
+      "step": 15060
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.84757207076468e-05,
+      "loss": 3.1386,
+      "step": 15065
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.846832928700354e-05,
+      "loss": 3.2427,
+      "step": 15070
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.846093580572956e-05,
+      "loss": 3.0536,
+      "step": 15075
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.845354026422085e-05,
+      "loss": 3.08,
+      "step": 15080
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.84461426628736e-05,
+      "loss": 3.1132,
+      "step": 15085
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.843874300208406e-05,
+      "loss": 3.1861,
+      "step": 15090
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.843134128224863e-05,
+      "loss": 3.1121,
+      "step": 15095
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.842393750376378e-05,
+      "loss": 3.1887,
+      "step": 15100
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.841653166702611e-05,
+      "loss": 3.1744,
+      "step": 15105
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.840912377243231e-05,
+      "loss": 3.0424,
+      "step": 15110
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.840171382037924e-05,
+      "loss": 3.1915,
+      "step": 15115
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.83943018112638e-05,
+      "loss": 3.1921,
+      "step": 15120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.838688774548303e-05,
+      "loss": 3.154,
+      "step": 15125
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.837947162343409e-05,
+      "loss": 3.0967,
+      "step": 15130
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.837205344551422e-05,
+      "loss": 3.1372,
+      "step": 15135
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.836463321212081e-05,
+      "loss": 3.2003,
+      "step": 15140
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.835721092365133e-05,
+      "loss": 3.1037,
+      "step": 15145
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.834978658050337e-05,
+      "loss": 3.0892,
+      "step": 15150
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.834236018307463e-05,
+      "loss": 3.0962,
+      "step": 15155
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.833493173176291e-05,
+      "loss": 3.1365,
+      "step": 15160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.832750122696614e-05,
+      "loss": 3.2621,
+      "step": 15165
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.832006866908235e-05,
+      "loss": 3.167,
+      "step": 15170
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.83126340585097e-05,
+      "loss": 3.046,
+      "step": 15175
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.830519739564638e-05,
+      "loss": 3.1535,
+      "step": 15180
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.82977586808908e-05,
+      "loss": 3.0912,
+      "step": 15185
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.829031791464141e-05,
+      "loss": 3.2134,
+      "step": 15190
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.82828750972968e-05,
+      "loss": 3.1247,
+      "step": 15195
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.827543022925565e-05,
+      "loss": 3.2428,
+      "step": 15200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.826798331091677e-05,
+      "loss": 3.113,
+      "step": 15205
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.826053434267904e-05,
+      "loss": 2.9983,
+      "step": 15210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.825308332494151e-05,
+      "loss": 2.964,
+      "step": 15215
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.824563025810331e-05,
+      "loss": 3.1756,
+      "step": 15220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.823817514256365e-05,
+      "loss": 3.1226,
+      "step": 15225
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.823071797872189e-05,
+      "loss": 3.1224,
+      "step": 15230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.822325876697749e-05,
+      "loss": 3.0958,
+      "step": 15235
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.821579750773003e-05,
+      "loss": 3.2365,
+      "step": 15240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.820833420137916e-05,
+      "loss": 3.2126,
+      "step": 15245
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.820086884832468e-05,
+      "loss": 3.1905,
+      "step": 15250
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.819340144896649e-05,
+      "loss": 3.1306,
+      "step": 15255
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.81859320037046e-05,
+      "loss": 3.114,
+      "step": 15260
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.817846051293911e-05,
+      "loss": 3.2969,
+      "step": 15265
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.817098697707025e-05,
+      "loss": 3.0068,
+      "step": 15270
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.816351139649837e-05,
+      "loss": 3.2092,
+      "step": 15275
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.815603377162389e-05,
+      "loss": 2.987,
+      "step": 15280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.814855410284737e-05,
+      "loss": 3.113,
+      "step": 15285
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.814107239056949e-05,
+      "loss": 3.0848,
+      "step": 15290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.8133588635191e-05,
+      "loss": 3.1369,
+      "step": 15295
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.81261028371128e-05,
+      "loss": 3.1933,
+      "step": 15300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.811861499673585e-05,
+      "loss": 3.1606,
+      "step": 15305
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.811112511446131e-05,
+      "loss": 3.1954,
+      "step": 15310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.810363319069034e-05,
+      "loss": 3.015,
+      "step": 15315
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.809613922582427e-05,
+      "loss": 3.0489,
+      "step": 15320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.808864322026453e-05,
+      "loss": 3.1091,
+      "step": 15325
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.808114517441267e-05,
+      "loss": 3.0524,
+      "step": 15330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.807364508867032e-05,
+      "loss": 2.9373,
+      "step": 15335
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.806614296343923e-05,
+      "loss": 3.05,
+      "step": 15340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.805863879912131e-05,
+      "loss": 3.0799,
+      "step": 15345
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.805113259611848e-05,
+      "loss": 3.2647,
+      "step": 15350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.804362435483286e-05,
+      "loss": 3.1435,
+      "step": 15355
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.803611407566663e-05,
+      "loss": 3.1484,
+      "step": 15360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.80286017590221e-05,
+      "loss": 3.0909,
+      "step": 15365
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.802108740530168e-05,
+      "loss": 3.2364,
+      "step": 15370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.801357101490788e-05,
+      "loss": 3.1489,
+      "step": 15375
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.800605258824335e-05,
+      "loss": 3.0227,
+      "step": 15380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.799853212571081e-05,
+      "loss": 3.1785,
+      "step": 15385
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.799100962771311e-05,
+      "loss": 3.1823,
+      "step": 15390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.798348509465322e-05,
+      "loss": 3.1702,
+      "step": 15395
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.797595852693421e-05,
+      "loss": 3.1741,
+      "step": 15400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.796842992495922e-05,
+      "loss": 3.1171,
+      "step": 15405
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.796089928913158e-05,
+      "loss": 2.9984,
+      "step": 15410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.795336661985466e-05,
+      "loss": 3.0366,
+      "step": 15415
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.794583191753196e-05,
+      "loss": 3.0478,
+      "step": 15420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.79382951825671e-05,
+      "loss": 3.1095,
+      "step": 15425
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.793075641536378e-05,
+      "loss": 3.1746,
+      "step": 15430
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.792321561632585e-05,
+      "loss": 3.1267,
+      "step": 15435
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.791567278585725e-05,
+      "loss": 3.108,
+      "step": 15440
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.790812792436201e-05,
+      "loss": 3.1297,
+      "step": 15445
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.790058103224427e-05,
+      "loss": 3.087,
+      "step": 15450
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.789303210990835e-05,
+      "loss": 3.1694,
+      "step": 15455
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.788548115775859e-05,
+      "loss": 3.0765,
+      "step": 15460
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.787792817619946e-05,
+      "loss": 3.1085,
+      "step": 15465
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.787037316563556e-05,
+      "loss": 3.137,
+      "step": 15470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.786281612647159e-05,
+      "loss": 3.1643,
+      "step": 15475
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.785525705911235e-05,
+      "loss": 3.021,
+      "step": 15480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.784769596396276e-05,
+      "loss": 3.1912,
+      "step": 15485
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.784013284142787e-05,
+      "loss": 3.1613,
+      "step": 15490
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.783256769191277e-05,
+      "loss": 3.1463,
+      "step": 15495
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.782500051582274e-05,
+      "loss": 3.18,
+      "step": 15500
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.78174313135631e-05,
+      "loss": 3.2386,
+      "step": 15505
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.780986008553933e-05,
+      "loss": 3.131,
+      "step": 15510
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.780228683215698e-05,
+      "loss": 3.1293,
+      "step": 15515
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.779471155382176e-05,
+      "loss": 3.1631,
+      "step": 15520
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.77871342509394e-05,
+      "loss": 3.2214,
+      "step": 15525
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.777955492391585e-05,
+      "loss": 3.0457,
+      "step": 15530
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.777197357315707e-05,
+      "loss": 3.0205,
+      "step": 15535
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.776439019906919e-05,
+      "loss": 3.0395,
+      "step": 15540
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.775680480205842e-05,
+      "loss": 3.0778,
+      "step": 15545
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.774921738253109e-05,
+      "loss": 2.8555,
+      "step": 15550
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.774162794089363e-05,
+      "loss": 3.1277,
+      "step": 15555
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.77340364775526e-05,
+      "loss": 3.0982,
+      "step": 15560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.772644299291464e-05,
+      "loss": 3.192,
+      "step": 15565
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.771884748738648e-05,
+      "loss": 3.1539,
+      "step": 15570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.771124996137505e-05,
+      "loss": 3.1126,
+      "step": 15575
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.770365041528727e-05,
+      "loss": 3.2138,
+      "step": 15580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.769604884953026e-05,
+      "loss": 3.2369,
+      "step": 15585
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.76884452645112e-05,
+      "loss": 3.2055,
+      "step": 15590
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.768083966063738e-05,
+      "loss": 3.238,
+      "step": 15595
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.767323203831623e-05,
+      "loss": 3.1635,
+      "step": 15600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.766562239795525e-05,
+      "loss": 3.078,
+      "step": 15605
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.765801073996207e-05,
+      "loss": 3.0443,
+      "step": 15610
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.765039706474444e-05,
+      "loss": 3.0841,
+      "step": 15615
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.764278137271019e-05,
+      "loss": 3.0427,
+      "step": 15620
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.763516366426723e-05,
+      "loss": 3.1132,
+      "step": 15625
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.762754393982369e-05,
+      "loss": 3.1703,
+      "step": 15630
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.761992219978767e-05,
+      "loss": 3.1339,
+      "step": 15635
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.761229844456748e-05,
+      "loss": 3.2309,
+      "step": 15640
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.76046726745715e-05,
+      "loss": 3.0972,
+      "step": 15645
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.759704489020822e-05,
+      "loss": 3.0879,
+      "step": 15650
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.758941509188623e-05,
+      "loss": 3.0574,
+      "step": 15655
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.758178328001423e-05,
+      "loss": 3.2373,
+      "step": 15660
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.757414945500101e-05,
+      "loss": 3.0821,
+      "step": 15665
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.756651361725556e-05,
+      "loss": 3.015,
+      "step": 15670
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.755887576718684e-05,
+      "loss": 3.0959,
+      "step": 15675
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.755123590520402e-05,
+      "loss": 3.02,
+      "step": 15680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.754359403171634e-05,
+      "loss": 3.1302,
+      "step": 15685
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.753595014713314e-05,
+      "loss": 3.2364,
+      "step": 15690
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.752830425186389e-05,
+      "loss": 3.2586,
+      "step": 15695
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.752065634631816e-05,
+      "loss": 3.1424,
+      "step": 15700
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.751300643090561e-05,
+      "loss": 3.1565,
+      "step": 15705
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.750535450603605e-05,
+      "loss": 3.0599,
+      "step": 15710
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.749770057211934e-05,
+      "loss": 3.1335,
+      "step": 15715
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.74900446295655e-05,
+      "loss": 3.0864,
+      "step": 15720
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.748238667878461e-05,
+      "loss": 3.1252,
+      "step": 15725
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.74747267201869e-05,
+      "loss": 3.1042,
+      "step": 15730
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.746706475418271e-05,
+      "loss": 3.131,
+      "step": 15735
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.745940078118242e-05,
+      "loss": 3.1264,
+      "step": 15740
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.74517348015966e-05,
+      "loss": 3.1652,
+      "step": 15745
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.74440668158359e-05,
+      "loss": 3.1474,
+      "step": 15750
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.743639682431106e-05,
+      "loss": 3.1873,
+      "step": 15755
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.742872482743291e-05,
+      "loss": 3.2669,
+      "step": 15760
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.742105082561247e-05,
+      "loss": 3.0604,
+      "step": 15765
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.741337481926078e-05,
+      "loss": 3.1451,
+      "step": 15770
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.740569680878901e-05,
+      "loss": 3.1505,
+      "step": 15775
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.739801679460845e-05,
+      "loss": 3.1852,
+      "step": 15780
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.739033477713053e-05,
+      "loss": 3.2702,
+      "step": 15785
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.738265075676671e-05,
+      "loss": 3.1084,
+      "step": 15790
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.737496473392862e-05,
+      "loss": 3.0537,
+      "step": 15795
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.736727670902798e-05,
+      "loss": 3.1341,
+      "step": 15800
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.735958668247661e-05,
+      "loss": 3.1093,
+      "step": 15805
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.735189465468643e-05,
+      "loss": 3.1998,
+      "step": 15810
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.734420062606949e-05,
+      "loss": 3.0865,
+      "step": 15815
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.733650459703795e-05,
+      "loss": 3.1103,
+      "step": 15820
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.732880656800402e-05,
+      "loss": 3.1569,
+      "step": 15825
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.73211065393801e-05,
+      "loss": 3.1755,
+      "step": 15830
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.731340451157865e-05,
+      "loss": 3.2323,
+      "step": 15835
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.730570048501222e-05,
+      "loss": 2.9895,
+      "step": 15840
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.729799446009352e-05,
+      "loss": 2.9975,
+      "step": 15845
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.729028643723531e-05,
+      "loss": 3.1418,
+      "step": 15850
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.728257641685051e-05,
+      "loss": 3.0753,
+      "step": 15855
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.727486439935214e-05,
+      "loss": 3.0517,
+      "step": 15860
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.726715038515325e-05,
+      "loss": 3.0379,
+      "step": 15865
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.725943437466711e-05,
+      "loss": 3.0974,
+      "step": 15870
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.7251716368307e-05,
+      "loss": 3.1182,
+      "step": 15875
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.724399636648639e-05,
+      "loss": 3.0094,
+      "step": 15880
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.723627436961878e-05,
+      "loss": 3.2197,
+      "step": 15885
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.722855037811786e-05,
+      "loss": 3.1863,
+      "step": 15890
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.722082439239733e-05,
+      "loss": 3.142,
+      "step": 15895
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.721309641287107e-05,
+      "loss": 3.2071,
+      "step": 15900
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.720536643995306e-05,
+      "loss": 3.0811,
+      "step": 15905
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.719763447405736e-05,
+      "loss": 3.0791,
+      "step": 15910
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.718990051559812e-05,
+      "loss": 3.2444,
+      "step": 15915
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.718216456498967e-05,
+      "loss": 2.9429,
+      "step": 15920
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.717442662264637e-05,
+      "loss": 3.2147,
+      "step": 15925
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.716668668898272e-05,
+      "loss": 3.1561,
+      "step": 15930
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.715894476441336e-05,
+      "loss": 3.1608,
+      "step": 15935
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.715120084935294e-05,
+      "loss": 3.099,
+      "step": 15940
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.714345494421632e-05,
+      "loss": 3.1323,
+      "step": 15945
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.713570704941843e-05,
+      "loss": 3.1344,
+      "step": 15950
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.712795716537427e-05,
+      "loss": 3.2597,
+      "step": 15955
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.7120205292499e-05,
+      "loss": 3.0449,
+      "step": 15960
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.711245143120786e-05,
+      "loss": 3.0844,
+      "step": 15965
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 8.71046955819162e-05,
+      "loss": 3.1387,
+      "step": 15970
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.709693774503948e-05,
+      "loss": 3.1474,
+      "step": 15975
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.708917792099325e-05,
+      "loss": 3.0084,
+      "step": 15980
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.708141611019319e-05,
+      "loss": 3.1413,
+      "step": 15985
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.707365231305509e-05,
+      "loss": 3.1354,
+      "step": 15990
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.70658865299948e-05,
+      "loss": 3.1096,
+      "step": 15995
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.705811876142834e-05,
+      "loss": 3.1025,
+      "step": 16000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.705034900777178e-05,
+      "loss": 3.1684,
+      "step": 16005
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.704257726944137e-05,
+      "loss": 3.1365,
+      "step": 16010
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.703480354685334e-05,
+      "loss": 3.1835,
+      "step": 16015
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.702702784042417e-05,
+      "loss": 3.0147,
+      "step": 16020
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.701925015057036e-05,
+      "loss": 3.2129,
+      "step": 16025
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.701147047770853e-05,
+      "loss": 3.0954,
+      "step": 16030
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.700368882225543e-05,
+      "loss": 3.156,
+      "step": 16035
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.699590518462789e-05,
+      "loss": 3.0894,
+      "step": 16040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.698811956524284e-05,
+      "loss": 3.1867,
+      "step": 16045
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.698033196451736e-05,
+      "loss": 3.0979,
+      "step": 16050
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.697254238286858e-05,
+      "loss": 3.1647,
+      "step": 16055
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.696475082071379e-05,
+      "loss": 3.0779,
+      "step": 16060
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.695695727847036e-05,
+      "loss": 3.0642,
+      "step": 16065
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.694916175655572e-05,
+      "loss": 3.0688,
+      "step": 16070
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.694136425538751e-05,
+      "loss": 3.025,
+      "step": 16075
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.693356477538339e-05,
+      "loss": 3.2126,
+      "step": 16080
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.692576331696118e-05,
+      "loss": 3.0334,
+      "step": 16085
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.691795988053875e-05,
+      "loss": 3.1987,
+      "step": 16090
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.691015446653409e-05,
+      "loss": 3.1785,
+      "step": 16095
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.690234707536537e-05,
+      "loss": 3.1145,
+      "step": 16100
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.689453770745076e-05,
+      "loss": 3.1015,
+      "step": 16105
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.68867263632086e-05,
+      "loss": 3.1258,
+      "step": 16110
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.687891304305734e-05,
+      "loss": 3.035,
+      "step": 16115
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.687109774741548e-05,
+      "loss": 3.0388,
+      "step": 16120
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.686328047670169e-05,
+      "loss": 3.274,
+      "step": 16125
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.68554612313347e-05,
+      "loss": 3.1065,
+      "step": 16130
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.684764001173335e-05,
+      "loss": 3.107,
+      "step": 16135
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.683981681831664e-05,
+      "loss": 3.066,
+      "step": 16140
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.68319916515036e-05,
+      "loss": 3.082,
+      "step": 16145
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.68241645117134e-05,
+      "loss": 3.1176,
+      "step": 16150
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.681633539936536e-05,
+      "loss": 3.2663,
+      "step": 16155
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.68085043148788e-05,
+      "loss": 3.0767,
+      "step": 16160
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.680067125867323e-05,
+      "loss": 3.0783,
+      "step": 16165
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.679283623116827e-05,
+      "loss": 3.1967,
+      "step": 16170
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.678499923278361e-05,
+      "loss": 3.0437,
+      "step": 16175
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.677716026393901e-05,
+      "loss": 3.0954,
+      "step": 16180
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.676931932505443e-05,
+      "loss": 3.1052,
+      "step": 16185
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.676147641654985e-05,
+      "loss": 3.0518,
+      "step": 16190
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.675363153884543e-05,
+      "loss": 3.0966,
+      "step": 16195
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.674578469236137e-05,
+      "loss": 3.035,
+      "step": 16200
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.673793587751799e-05,
+      "loss": 3.1318,
+      "step": 16205
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.673008509473575e-05,
+      "loss": 3.1483,
+      "step": 16210
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.672223234443517e-05,
+      "loss": 3.0417,
+      "step": 16215
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.671437762703694e-05,
+      "loss": 3.0706,
+      "step": 16220
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.670652094296176e-05,
+      "loss": 3.1225,
+      "step": 16225
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.669866229263053e-05,
+      "loss": 2.9923,
+      "step": 16230
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.669080167646418e-05,
+      "loss": 3.1489,
+      "step": 16235
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.668293909488381e-05,
+      "loss": 3.1993,
+      "step": 16240
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.667507454831058e-05,
+      "loss": 3.1849,
+      "step": 16245
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.666720803716577e-05,
+      "loss": 3.2268,
+      "step": 16250
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.665933956187075e-05,
+      "loss": 3.1364,
+      "step": 16255
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.665146912284705e-05,
+      "loss": 3.0651,
+      "step": 16260
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.664359672051622e-05,
+      "loss": 3.1741,
+      "step": 16265
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.66357223553e-05,
+      "loss": 3.1358,
+      "step": 16270
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.662784602762018e-05,
+      "loss": 3.0817,
+      "step": 16275
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.661996773789866e-05,
+      "loss": 3.114,
+      "step": 16280
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.661208748655746e-05,
+      "loss": 3.062,
+      "step": 16285
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.660420527401872e-05,
+      "loss": 3.2113,
+      "step": 16290
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.659632110070464e-05,
+      "loss": 3.1405,
+      "step": 16295
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.658843496703755e-05,
+      "loss": 2.9713,
+      "step": 16300
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.65805468734399e-05,
+      "loss": 3.1608,
+      "step": 16305
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.657265682033426e-05,
+      "loss": 3.1659,
+      "step": 16310
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.656476480814321e-05,
+      "loss": 3.0807,
+      "step": 16315
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.655687083728956e-05,
+      "loss": 3.1626,
+      "step": 16320
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.654897490819614e-05,
+      "loss": 3.1292,
+      "step": 16325
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.654107702128588e-05,
+      "loss": 2.9699,
+      "step": 16330
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.653317717698191e-05,
+      "loss": 3.2386,
+      "step": 16335
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.652527537570737e-05,
+      "loss": 3.1223,
+      "step": 16340
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.651737161788554e-05,
+      "loss": 3.216,
+      "step": 16345
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.650946590393977e-05,
+      "loss": 3.1364,
+      "step": 16350
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.650155823429358e-05,
+      "loss": 3.1291,
+      "step": 16355
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.649364860937054e-05,
+      "loss": 3.0189,
+      "step": 16360
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.648573702959437e-05,
+      "loss": 3.0991,
+      "step": 16365
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.647782349538884e-05,
+      "loss": 3.1672,
+      "step": 16370
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.646990800717788e-05,
+      "loss": 3.1712,
+      "step": 16375
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.646199056538549e-05,
+      "loss": 3.0306,
+      "step": 16380
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.645407117043575e-05,
+      "loss": 3.1269,
+      "step": 16385
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.644614982275294e-05,
+      "loss": 3.0917,
+      "step": 16390
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.643822652276132e-05,
+      "loss": 3.0893,
+      "step": 16395
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.643030127088537e-05,
+      "loss": 3.2728,
+      "step": 16400
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.642237406754957e-05,
+      "loss": 3.0431,
+      "step": 16405
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.64144449131786e-05,
+      "loss": 3.2519,
+      "step": 16410
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.640651380819717e-05,
+      "loss": 3.0658,
+      "step": 16415
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.639858075303016e-05,
+      "loss": 3.2254,
+      "step": 16420
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.639064574810247e-05,
+      "loss": 3.1115,
+      "step": 16425
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.63827087938392e-05,
+      "loss": 3.2189,
+      "step": 16430
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.637476989066548e-05,
+      "loss": 3.1206,
+      "step": 16435
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.636682903900659e-05,
+      "loss": 3.113,
+      "step": 16440
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.635888623928789e-05,
+      "loss": 3.0153,
+      "step": 16445
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.635094149193485e-05,
+      "loss": 3.2769,
+      "step": 16450
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.634299479737305e-05,
+      "loss": 3.1504,
+      "step": 16455
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.633504615602817e-05,
+      "loss": 3.1519,
+      "step": 16460
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.632709556832599e-05,
+      "loss": 3.1155,
+      "step": 16465
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.631914303469239e-05,
+      "loss": 3.0,
+      "step": 16470
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.631118855555338e-05,
+      "loss": 3.0934,
+      "step": 16475
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.630323213133507e-05,
+      "loss": 3.1266,
+      "step": 16480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.629527376246361e-05,
+      "loss": 3.0651,
+      "step": 16485
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.628731344936536e-05,
+      "loss": 3.0464,
+      "step": 16490
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.627935119246669e-05,
+      "loss": 3.1991,
+      "step": 16495
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.627138699219414e-05,
+      "loss": 3.1495,
+      "step": 16500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.626342084897432e-05,
+      "loss": 3.1732,
+      "step": 16505
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.625545276323395e-05,
+      "loss": 3.1631,
+      "step": 16510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.624748273539985e-05,
+      "loss": 3.0626,
+      "step": 16515
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.623951076589896e-05,
+      "loss": 3.0491,
+      "step": 16520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.62315368551583e-05,
+      "loss": 3.2385,
+      "step": 16525
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.622356100360505e-05,
+      "loss": 3.0814,
+      "step": 16530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.621558321166638e-05,
+      "loss": 3.011,
+      "step": 16535
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.62076034797697e-05,
+      "loss": 3.1209,
+      "step": 16540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.619962180834245e-05,
+      "loss": 3.0421,
+      "step": 16545
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.619163819781214e-05,
+      "loss": 2.8672,
+      "step": 16550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.618365264860649e-05,
+      "loss": 3.1604,
+      "step": 16555
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.61756651611532e-05,
+      "loss": 3.1918,
+      "step": 16560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.616767573588016e-05,
+      "loss": 3.0754,
+      "step": 16565
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.615968437321537e-05,
+      "loss": 3.0203,
+      "step": 16570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.615169107358686e-05,
+      "loss": 3.0684,
+      "step": 16575
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.614369583742283e-05,
+      "loss": 3.0216,
+      "step": 16580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.613569866515155e-05,
+      "loss": 3.1492,
+      "step": 16585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.61276995572014e-05,
+      "loss": 3.0986,
+      "step": 16590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.611969851400089e-05,
+      "loss": 3.0608,
+      "step": 16595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.611169553597859e-05,
+      "loss": 3.0796,
+      "step": 16600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.610369062356319e-05,
+      "loss": 3.1504,
+      "step": 16605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.609568377718351e-05,
+      "loss": 3.0042,
+      "step": 16610
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.608767499726844e-05,
+      "loss": 3.309,
+      "step": 16615
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.607966428424701e-05,
+      "loss": 3.2772,
+      "step": 16620
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.607165163854828e-05,
+      "loss": 3.1409,
+      "step": 16625
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.60636370606015e-05,
+      "loss": 3.0724,
+      "step": 16630
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.605562055083598e-05,
+      "loss": 3.1008,
+      "step": 16635
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.604760210968112e-05,
+      "loss": 2.9748,
+      "step": 16640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.603958173756649e-05,
+      "loss": 3.0054,
+      "step": 16645
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 8.603155943492168e-05,
+      "loss": 3.1097,
+      "step": 16650
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.602353520217641e-05,
+      "loss": 3.1716,
+      "step": 16655
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.601550903976053e-05,
+      "loss": 3.1987,
+      "step": 16660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.600748094810399e-05,
+      "loss": 3.2416,
+      "step": 16665
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.599945092763681e-05,
+      "loss": 3.1459,
+      "step": 16670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.599141897878915e-05,
+      "loss": 3.1299,
+      "step": 16675
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.598338510199124e-05,
+      "loss": 3.0216,
+      "step": 16680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.597534929767346e-05,
+      "loss": 3.1698,
+      "step": 16685
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.596731156626622e-05,
+      "loss": 3.0582,
+      "step": 16690
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.59592719082001e-05,
+      "loss": 3.1427,
+      "step": 16695
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.595123032390577e-05,
+      "loss": 2.9865,
+      "step": 16700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.594318681381398e-05,
+      "loss": 3.0051,
+      "step": 16705
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.593514137835559e-05,
+      "loss": 3.1642,
+      "step": 16710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.59270940179616e-05,
+      "loss": 3.14,
+      "step": 16715
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.591904473306305e-05,
+      "loss": 3.0705,
+      "step": 16720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.59109935240911e-05,
+      "loss": 3.1042,
+      "step": 16725
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.590294039147708e-05,
+      "loss": 3.0548,
+      "step": 16730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.589488533565235e-05,
+      "loss": 3.1014,
+      "step": 16735
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.588682835704837e-05,
+      "loss": 3.067,
+      "step": 16740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.587876945609676e-05,
+      "loss": 3.1437,
+      "step": 16745
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.58707086332292e-05,
+      "loss": 3.1014,
+      "step": 16750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.58626458888775e-05,
+      "loss": 3.2396,
+      "step": 16755
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.585458122347352e-05,
+      "loss": 3.1189,
+      "step": 16760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.584651463744928e-05,
+      "loss": 3.2104,
+      "step": 16765
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.583844613123688e-05,
+      "loss": 3.1289,
+      "step": 16770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.583037570526853e-05,
+      "loss": 3.1774,
+      "step": 16775
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.582230335997654e-05,
+      "loss": 3.1309,
+      "step": 16780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.581422909579331e-05,
+      "loss": 3.0677,
+      "step": 16785
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.580615291315135e-05,
+      "loss": 3.099,
+      "step": 16790
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.579807481248328e-05,
+      "loss": 3.1199,
+      "step": 16795
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.578999479422185e-05,
+      "loss": 3.0348,
+      "step": 16800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.578191285879985e-05,
+      "loss": 3.198,
+      "step": 16805
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.577382900665019e-05,
+      "loss": 3.1694,
+      "step": 16810
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.576574323820594e-05,
+      "loss": 3.0533,
+      "step": 16815
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.57576555539002e-05,
+      "loss": 3.0747,
+      "step": 16820
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.574956595416622e-05,
+      "loss": 2.9872,
+      "step": 16825
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.57414744394373e-05,
+      "loss": 3.1089,
+      "step": 16830
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.573338101014694e-05,
+      "loss": 2.9946,
+      "step": 16835
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.572528566672862e-05,
+      "loss": 3.2379,
+      "step": 16840
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.5717188409616e-05,
+      "loss": 3.0487,
+      "step": 16845
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.570908923924284e-05,
+      "loss": 3.0353,
+      "step": 16850
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.570098815604298e-05,
+      "loss": 3.1054,
+      "step": 16855
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.569288516045038e-05,
+      "loss": 3.0995,
+      "step": 16860
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.568478025289908e-05,
+      "loss": 3.1414,
+      "step": 16865
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.567667343382322e-05,
+      "loss": 3.2408,
+      "step": 16870
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.566856470365709e-05,
+      "loss": 3.1547,
+      "step": 16875
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.566045406283502e-05,
+      "loss": 3.0674,
+      "step": 16880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.565234151179152e-05,
+      "loss": 3.1747,
+      "step": 16885
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.564422705096109e-05,
+      "loss": 3.1724,
+      "step": 16890
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.563611068077843e-05,
+      "loss": 3.1325,
+      "step": 16895
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.562799240167833e-05,
+      "loss": 3.106,
+      "step": 16900
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.561987221409562e-05,
+      "loss": 3.1386,
+      "step": 16905
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.561175011846531e-05,
+      "loss": 3.1424,
+      "step": 16910
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.560362611522244e-05,
+      "loss": 3.1305,
+      "step": 16915
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.559550020480222e-05,
+      "loss": 3.0194,
+      "step": 16920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.558737238763991e-05,
+      "loss": 2.958,
+      "step": 16925
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.55792426641709e-05,
+      "loss": 3.1048,
+      "step": 16930
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.557111103483066e-05,
+      "loss": 3.2016,
+      "step": 16935
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.556297750005482e-05,
+      "loss": 3.042,
+      "step": 16940
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.555484206027902e-05,
+      "loss": 3.1537,
+      "step": 16945
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.554670471593906e-05,
+      "loss": 3.1326,
+      "step": 16950
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.553856546747086e-05,
+      "loss": 3.2127,
+      "step": 16955
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.553042431531038e-05,
+      "loss": 3.0397,
+      "step": 16960
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.552228125989374e-05,
+      "loss": 3.0843,
+      "step": 16965
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.551413630165714e-05,
+      "loss": 3.1034,
+      "step": 16970
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.550598944103688e-05,
+      "loss": 3.0307,
+      "step": 16975
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.549784067846931e-05,
+      "loss": 3.0187,
+      "step": 16980
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.548969001439102e-05,
+      "loss": 3.066,
+      "step": 16985
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.548153744923857e-05,
+      "loss": 3.0248,
+      "step": 16990
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.547338298344867e-05,
+      "loss": 3.1571,
+      "step": 16995
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.546522661745813e-05,
+      "loss": 3.1328,
+      "step": 17000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.545706835170386e-05,
+      "loss": 3.1674,
+      "step": 17005
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.544890818662288e-05,
+      "loss": 3.16,
+      "step": 17010
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.54407461226523e-05,
+      "loss": 3.166,
+      "step": 17015
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.543258216022936e-05,
+      "loss": 3.0675,
+      "step": 17020
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.542441629979134e-05,
+      "loss": 2.9696,
+      "step": 17025
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.541624854177569e-05,
+      "loss": 3.0853,
+      "step": 17030
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.540807888661989e-05,
+      "loss": 3.1335,
+      "step": 17035
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.539990733476162e-05,
+      "loss": 3.1212,
+      "step": 17040
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.539173388663858e-05,
+      "loss": 3.0491,
+      "step": 17045
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.538355854268857e-05,
+      "loss": 3.112,
+      "step": 17050
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.537538130334956e-05,
+      "loss": 3.2396,
+      "step": 17055
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.536720216905958e-05,
+      "loss": 3.0703,
+      "step": 17060
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.535902114025673e-05,
+      "loss": 3.1604,
+      "step": 17065
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.535083821737926e-05,
+      "loss": 3.1761,
+      "step": 17070
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.534265340086549e-05,
+      "loss": 3.182,
+      "step": 17075
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.533446669115388e-05,
+      "loss": 3.0333,
+      "step": 17080
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.532627808868294e-05,
+      "loss": 3.0836,
+      "step": 17085
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.531808759389133e-05,
+      "loss": 3.2239,
+      "step": 17090
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.530989520721778e-05,
+      "loss": 3.1179,
+      "step": 17095
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.530170092910114e-05,
+      "loss": 2.9763,
+      "step": 17100
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.529350475998034e-05,
+      "loss": 3.0528,
+      "step": 17105
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.528530670029441e-05,
+      "loss": 3.195,
+      "step": 17110
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.527710675048255e-05,
+      "loss": 3.1292,
+      "step": 17115
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.526890491098395e-05,
+      "loss": 2.923,
+      "step": 17120
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.526070118223798e-05,
+      "loss": 3.0615,
+      "step": 17125
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.525249556468409e-05,
+      "loss": 3.1478,
+      "step": 17130
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.524428805876182e-05,
+      "loss": 3.1481,
+      "step": 17135
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.523607866491083e-05,
+      "loss": 3.0685,
+      "step": 17140
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.522786738357087e-05,
+      "loss": 3.0782,
+      "step": 17145
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.521965421518178e-05,
+      "loss": 3.0676,
+      "step": 17150
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.521143916018355e-05,
+      "loss": 3.1434,
+      "step": 17155
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.52032222190162e-05,
+      "loss": 2.9839,
+      "step": 17160
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.51950033921199e-05,
+      "loss": 3.088,
+      "step": 17165
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.518678267993491e-05,
+      "loss": 3.0178,
+      "step": 17170
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.51785600829016e-05,
+      "loss": 3.1416,
+      "step": 17175
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.517033560146037e-05,
+      "loss": 3.1301,
+      "step": 17180
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.516210923605185e-05,
+      "loss": 3.1611,
+      "step": 17185
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.515388098711668e-05,
+      "loss": 3.0413,
+      "step": 17190
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.51456508550956e-05,
+      "loss": 3.1299,
+      "step": 17195
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.513741884042951e-05,
+      "loss": 3.1065,
+      "step": 17200
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.512918494355935e-05,
+      "loss": 3.1298,
+      "step": 17205
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.512094916492619e-05,
+      "loss": 3.0629,
+      "step": 17210
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.51127115049712e-05,
+      "loss": 3.019,
+      "step": 17215
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.510447196413562e-05,
+      "loss": 3.1362,
+      "step": 17220
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.509623054286086e-05,
+      "loss": 3.1105,
+      "step": 17225
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.508798724158835e-05,
+      "loss": 3.2157,
+      "step": 17230
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.507974206075969e-05,
+      "loss": 3.078,
+      "step": 17235
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.50714950008165e-05,
+      "loss": 3.0621,
+      "step": 17240
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.506324606220061e-05,
+      "loss": 3.0678,
+      "step": 17245
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.505499524535388e-05,
+      "loss": 3.2027,
+      "step": 17250
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.504674255071823e-05,
+      "loss": 3.1155,
+      "step": 17255
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.503848797873577e-05,
+      "loss": 3.1391,
+      "step": 17260
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.503023152984868e-05,
+      "loss": 3.2058,
+      "step": 17265
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.502197320449923e-05,
+      "loss": 2.9143,
+      "step": 17270
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.501371300312977e-05,
+      "loss": 3.2991,
+      "step": 17275
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.500545092618279e-05,
+      "loss": 3.1758,
+      "step": 17280
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.499718697410085e-05,
+      "loss": 3.08,
+      "step": 17285
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.498892114732666e-05,
+      "loss": 3.1302,
+      "step": 17290
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.498065344630294e-05,
+      "loss": 2.9751,
+      "step": 17295
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.497238387147264e-05,
+      "loss": 3.1331,
+      "step": 17300
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.496411242327866e-05,
+      "loss": 3.2206,
+      "step": 17305
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.495583910216411e-05,
+      "loss": 3.2216,
+      "step": 17310
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.494756390857218e-05,
+      "loss": 3.1132,
+      "step": 17315
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.493928684294613e-05,
+      "loss": 3.1538,
+      "step": 17320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.493100790572934e-05,
+      "loss": 3.1336,
+      "step": 17325
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 8.492272709736528e-05,
+      "loss": 3.1047,
+      "step": 17330
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.491444441829754e-05,
+      "loss": 3.1513,
+      "step": 17335
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.490615986896979e-05,
+      "loss": 3.1848,
+      "step": 17340
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.489787344982581e-05,
+      "loss": 3.2808,
+      "step": 17345
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.48895851613095e-05,
+      "loss": 3.1386,
+      "step": 17350
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.488129500386478e-05,
+      "loss": 3.1678,
+      "step": 17355
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.48730029779358e-05,
+      "loss": 3.1141,
+      "step": 17360
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.486470908396667e-05,
+      "loss": 3.214,
+      "step": 17365
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.485641332240171e-05,
+      "loss": 3.0844,
+      "step": 17370
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.48481156936853e-05,
+      "loss": 3.0601,
+      "step": 17375
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.483981619826191e-05,
+      "loss": 3.1251,
+      "step": 17380
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.483151483657612e-05,
+      "loss": 3.1602,
+      "step": 17385
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.48232116090726e-05,
+      "loss": 3.1589,
+      "step": 17390
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.481490651619612e-05,
+      "loss": 3.0487,
+      "step": 17395
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.480659955839161e-05,
+      "loss": 3.1533,
+      "step": 17400
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.479829073610398e-05,
+      "loss": 3.204,
+      "step": 17405
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.478998004977836e-05,
+      "loss": 3.1666,
+      "step": 17410
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.47816674998599e-05,
+      "loss": 3.1824,
+      "step": 17415
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.477335308679388e-05,
+      "loss": 3.1078,
+      "step": 17420
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.476503681102568e-05,
+      "loss": 3.108,
+      "step": 17425
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.475671867300079e-05,
+      "loss": 3.231,
+      "step": 17430
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.474839867316478e-05,
+      "loss": 3.1161,
+      "step": 17435
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.474007681196332e-05,
+      "loss": 3.0833,
+      "step": 17440
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.473175308984218e-05,
+      "loss": 3.1703,
+      "step": 17445
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.472342750724726e-05,
+      "loss": 3.0638,
+      "step": 17450
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.471510006462453e-05,
+      "loss": 2.9786,
+      "step": 17455
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.470677076242006e-05,
+      "loss": 3.0644,
+      "step": 17460
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.469843960108002e-05,
+      "loss": 3.0162,
+      "step": 17465
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.469010658105068e-05,
+      "loss": 3.0634,
+      "step": 17470
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.468177170277845e-05,
+      "loss": 3.0468,
+      "step": 17475
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.467343496670977e-05,
+      "loss": 3.0857,
+      "step": 17480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.46650963732912e-05,
+      "loss": 3.1502,
+      "step": 17485
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.465675592296944e-05,
+      "loss": 3.1093,
+      "step": 17490
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.464841361619127e-05,
+      "loss": 3.0237,
+      "step": 17495
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.464006945340355e-05,
+      "loss": 3.1885,
+      "step": 17500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.463172343505323e-05,
+      "loss": 3.1468,
+      "step": 17505
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.462337556158741e-05,
+      "loss": 3.0957,
+      "step": 17510
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.461502583345327e-05,
+      "loss": 3.0744,
+      "step": 17515
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.460667425109804e-05,
+      "loss": 3.0219,
+      "step": 17520
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.45983208149691e-05,
+      "loss": 3.1706,
+      "step": 17525
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.458996552551396e-05,
+      "loss": 3.1608,
+      "step": 17530
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.458160838318013e-05,
+      "loss": 3.2127,
+      "step": 17535
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.45732493884153e-05,
+      "loss": 3.1078,
+      "step": 17540
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.456488854166724e-05,
+      "loss": 3.1694,
+      "step": 17545
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.455652584338383e-05,
+      "loss": 3.2065,
+      "step": 17550
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.4548161294013e-05,
+      "loss": 3.0503,
+      "step": 17555
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.453979489400283e-05,
+      "loss": 3.0387,
+      "step": 17560
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.453142664380147e-05,
+      "loss": 3.1638,
+      "step": 17565
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.45230565438572e-05,
+      "loss": 3.2002,
+      "step": 17570
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.451468459461838e-05,
+      "loss": 2.969,
+      "step": 17575
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.450631079653345e-05,
+      "loss": 3.1401,
+      "step": 17580
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.4497935150051e-05,
+      "loss": 3.1178,
+      "step": 17585
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.448955765561968e-05,
+      "loss": 3.1529,
+      "step": 17590
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.44811783136882e-05,
+      "loss": 3.0772,
+      "step": 17595
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.447279712470548e-05,
+      "loss": 3.0381,
+      "step": 17600
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.446441408912045e-05,
+      "loss": 3.0054,
+      "step": 17605
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.445602920738214e-05,
+      "loss": 3.0291,
+      "step": 17610
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.444764247993972e-05,
+      "loss": 3.1615,
+      "step": 17615
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.443925390724246e-05,
+      "loss": 3.0859,
+      "step": 17620
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.44308634897397e-05,
+      "loss": 3.0484,
+      "step": 17625
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.442247122788088e-05,
+      "loss": 3.1323,
+      "step": 17630
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.441407712211556e-05,
+      "loss": 3.2957,
+      "step": 17635
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.440568117289336e-05,
+      "loss": 3.1641,
+      "step": 17640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.439728338066404e-05,
+      "loss": 3.0549,
+      "step": 17645
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.438888374587745e-05,
+      "loss": 3.204,
+      "step": 17650
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.438048226898354e-05,
+      "loss": 3.1323,
+      "step": 17655
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.437207895043234e-05,
+      "loss": 3.071,
+      "step": 17660
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.436367379067399e-05,
+      "loss": 3.0864,
+      "step": 17665
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.435526679015873e-05,
+      "loss": 3.1499,
+      "step": 17670
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.43468579493369e-05,
+      "loss": 3.1061,
+      "step": 17675
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.433844726865894e-05,
+      "loss": 3.2557,
+      "step": 17680
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.433003474857536e-05,
+      "loss": 3.0523,
+      "step": 17685
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.432162038953682e-05,
+      "loss": 3.1433,
+      "step": 17690
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.431320419199405e-05,
+      "loss": 3.1153,
+      "step": 17695
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.430478615639786e-05,
+      "loss": 3.1762,
+      "step": 17700
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.429636628319919e-05,
+      "loss": 3.2101,
+      "step": 17705
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.428794457284907e-05,
+      "loss": 2.9982,
+      "step": 17710
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.427952102579863e-05,
+      "loss": 3.1362,
+      "step": 17715
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.427109564249908e-05,
+      "loss": 3.1056,
+      "step": 17720
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.426266842340175e-05,
+      "loss": 3.1196,
+      "step": 17725
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.425423936895807e-05,
+      "loss": 3.1356,
+      "step": 17730
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.424580847961953e-05,
+      "loss": 3.03,
+      "step": 17735
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.42373757558378e-05,
+      "loss": 3.0893,
+      "step": 17740
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.422894119806452e-05,
+      "loss": 3.0661,
+      "step": 17745
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.422050480675155e-05,
+      "loss": 3.0778,
+      "step": 17750
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.421206658235081e-05,
+      "loss": 3.1815,
+      "step": 17755
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.420362652531429e-05,
+      "loss": 3.2358,
+      "step": 17760
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.419518463609408e-05,
+      "loss": 3.103,
+      "step": 17765
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.418674091514244e-05,
+      "loss": 3.097,
+      "step": 17770
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.417829536291163e-05,
+      "loss": 3.0409,
+      "step": 17775
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.416984797985407e-05,
+      "loss": 3.1215,
+      "step": 17780
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.416139876642225e-05,
+      "loss": 2.9811,
+      "step": 17785
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.415294772306878e-05,
+      "loss": 3.1476,
+      "step": 17790
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.414449485024635e-05,
+      "loss": 3.0932,
+      "step": 17795
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.413604014840775e-05,
+      "loss": 3.1584,
+      "step": 17800
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.41275836180059e-05,
+      "loss": 3.1117,
+      "step": 17805
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.411912525949375e-05,
+      "loss": 3.2552,
+      "step": 17810
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.411066507332441e-05,
+      "loss": 3.1371,
+      "step": 17815
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.410220305995106e-05,
+      "loss": 3.138,
+      "step": 17820
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.409373921982699e-05,
+      "loss": 2.9326,
+      "step": 17825
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.408527355340559e-05,
+      "loss": 3.1295,
+      "step": 17830
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.407680606114032e-05,
+      "loss": 3.2417,
+      "step": 17835
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.406833674348475e-05,
+      "loss": 3.029,
+      "step": 17840
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.40598656008926e-05,
+      "loss": 3.2645,
+      "step": 17845
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.40513926338176e-05,
+      "loss": 3.0913,
+      "step": 17850
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.404291784271365e-05,
+      "loss": 3.1098,
+      "step": 17855
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.403444122803472e-05,
+      "loss": 3.137,
+      "step": 17860
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.402596279023482e-05,
+      "loss": 3.0659,
+      "step": 17865
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.401748252976818e-05,
+      "loss": 3.1528,
+      "step": 17870
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.400900044708904e-05,
+      "loss": 3.0822,
+      "step": 17875
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.400051654265175e-05,
+      "loss": 3.1546,
+      "step": 17880
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.399203081691078e-05,
+      "loss": 3.1363,
+      "step": 17885
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.398354327032067e-05,
+      "loss": 3.1434,
+      "step": 17890
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.397505390333608e-05,
+      "loss": 3.0446,
+      "step": 17895
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.396656271641177e-05,
+      "loss": 3.2299,
+      "step": 17900
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.395806971000257e-05,
+      "loss": 3.1317,
+      "step": 17905
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.394957488456342e-05,
+      "loss": 3.2153,
+      "step": 17910
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.394107824054938e-05,
+      "loss": 3.1009,
+      "step": 17915
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.393257977841558e-05,
+      "loss": 3.0954,
+      "step": 17920
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.392407949861726e-05,
+      "loss": 3.1738,
+      "step": 17925
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.391557740160976e-05,
+      "loss": 3.1279,
+      "step": 17930
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.390707348784849e-05,
+      "loss": 3.0725,
+      "step": 17935
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.3898567757789e-05,
+      "loss": 3.0637,
+      "step": 17940
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.389006021188691e-05,
+      "loss": 3.1506,
+      "step": 17945
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.388155085059793e-05,
+      "loss": 3.2448,
+      "step": 17950
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.387303967437789e-05,
+      "loss": 2.9932,
+      "step": 17955
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.386452668368275e-05,
+      "loss": 3.1752,
+      "step": 17960
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.385601187896845e-05,
+      "loss": 3.175,
+      "step": 17965
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.384749526069114e-05,
+      "loss": 3.0272,
+      "step": 17970
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.383897682930701e-05,
+      "loss": 3.2597,
+      "step": 17975
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.38304565852724e-05,
+      "loss": 2.9593,
+      "step": 17980
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.38219345290437e-05,
+      "loss": 3.0903,
+      "step": 17985
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.38134106610774e-05,
+      "loss": 3.056,
+      "step": 17990
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.38048849818301e-05,
+      "loss": 3.042,
+      "step": 17995
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.37963574917585e-05,
+      "loss": 3.1182,
+      "step": 18000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.378782819131939e-05,
+      "loss": 3.1055,
+      "step": 18005
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 8.377929708096966e-05,
+      "loss": 2.8882,
+      "step": 18010
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.377076416116629e-05,
+      "loss": 3.193,
+      "step": 18015
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.376222943236636e-05,
+      "loss": 3.0047,
+      "step": 18020
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.375369289502706e-05,
+      "loss": 3.1228,
+      "step": 18025
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.374515454960566e-05,
+      "loss": 3.1153,
+      "step": 18030
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.373661439655955e-05,
+      "loss": 3.0854,
+      "step": 18035
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.372807243634617e-05,
+      "loss": 3.2023,
+      "step": 18040
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.37195286694231e-05,
+      "loss": 3.0954,
+      "step": 18045
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.371098309624801e-05,
+      "loss": 3.1994,
+      "step": 18050
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.370243571727864e-05,
+      "loss": 3.111,
+      "step": 18055
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.369388653297289e-05,
+      "loss": 3.0557,
+      "step": 18060
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.368533554378867e-05,
+      "loss": 3.1247,
+      "step": 18065
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.367678275018405e-05,
+      "loss": 2.9743,
+      "step": 18070
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.366822815261716e-05,
+      "loss": 2.9952,
+      "step": 18075
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.365967175154627e-05,
+      "loss": 3.1352,
+      "step": 18080
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.36511135474297e-05,
+      "loss": 3.0902,
+      "step": 18085
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.36425535407259e-05,
+      "loss": 3.1208,
+      "step": 18090
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.363399173189339e-05,
+      "loss": 3.0848,
+      "step": 18095
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.362542812139081e-05,
+      "loss": 3.095,
+      "step": 18100
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.361686270967687e-05,
+      "loss": 3.0804,
+      "step": 18105
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.360829549721042e-05,
+      "loss": 3.188,
+      "step": 18110
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.359972648445035e-05,
+      "loss": 3.1708,
+      "step": 18115
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.359115567185572e-05,
+      "loss": 3.0385,
+      "step": 18120
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.358258305988558e-05,
+      "loss": 3.229,
+      "step": 18125
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.35740086489992e-05,
+      "loss": 3.0131,
+      "step": 18130
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.356543243965584e-05,
+      "loss": 3.248,
+      "step": 18135
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.355685443231494e-05,
+      "loss": 3.0723,
+      "step": 18140
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.354827462743595e-05,
+      "loss": 3.0507,
+      "step": 18145
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.35396930254785e-05,
+      "loss": 3.2123,
+      "step": 18150
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.353110962690227e-05,
+      "loss": 3.0943,
+      "step": 18155
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.352252443216705e-05,
+      "loss": 3.0424,
+      "step": 18160
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.351393744173272e-05,
+      "loss": 3.0282,
+      "step": 18165
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.350534865605925e-05,
+      "loss": 3.05,
+      "step": 18170
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.349675807560671e-05,
+      "loss": 3.215,
+      "step": 18175
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.348816570083531e-05,
+      "loss": 3.1357,
+      "step": 18180
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.347957153220527e-05,
+      "loss": 3.1926,
+      "step": 18185
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.3470975570177e-05,
+      "loss": 3.1555,
+      "step": 18190
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.346237781521091e-05,
+      "loss": 3.1066,
+      "step": 18195
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.345377826776759e-05,
+      "loss": 3.061,
+      "step": 18200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.344517692830768e-05,
+      "loss": 3.2473,
+      "step": 18205
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.343657379729193e-05,
+      "loss": 3.17,
+      "step": 18210
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.342796887518119e-05,
+      "loss": 3.0124,
+      "step": 18215
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.341936216243637e-05,
+      "loss": 3.0841,
+      "step": 18220
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.341075365951853e-05,
+      "loss": 3.2239,
+      "step": 18225
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.340214336688881e-05,
+      "loss": 3.1009,
+      "step": 18230
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.339353128500842e-05,
+      "loss": 3.1023,
+      "step": 18235
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.33849174143387e-05,
+      "loss": 3.0951,
+      "step": 18240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.337630175534103e-05,
+      "loss": 3.2295,
+      "step": 18245
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.336768430847695e-05,
+      "loss": 3.1375,
+      "step": 18250
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.335906507420809e-05,
+      "loss": 3.0662,
+      "step": 18255
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.335044405299612e-05,
+      "loss": 2.9748,
+      "step": 18260
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.334182124530288e-05,
+      "loss": 3.018,
+      "step": 18265
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.333319665159024e-05,
+      "loss": 3.0982,
+      "step": 18270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.332457027232019e-05,
+      "loss": 3.0507,
+      "step": 18275
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.331594210795484e-05,
+      "loss": 2.975,
+      "step": 18280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.330731215895635e-05,
+      "loss": 3.2545,
+      "step": 18285
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.329868042578702e-05,
+      "loss": 3.1306,
+      "step": 18290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.329004690890921e-05,
+      "loss": 3.1472,
+      "step": 18295
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.32814116087854e-05,
+      "loss": 3.1066,
+      "step": 18300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.327277452587816e-05,
+      "loss": 2.9767,
+      "step": 18305
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.326413566065015e-05,
+      "loss": 3.2058,
+      "step": 18310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.325549501356413e-05,
+      "loss": 3.0635,
+      "step": 18315
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.324685258508293e-05,
+      "loss": 3.2304,
+      "step": 18320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.323820837566953e-05,
+      "loss": 3.2384,
+      "step": 18325
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.322956238578696e-05,
+      "loss": 3.2054,
+      "step": 18330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.322091461589836e-05,
+      "loss": 3.095,
+      "step": 18335
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.321226506646696e-05,
+      "loss": 3.1122,
+      "step": 18340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.32036137379561e-05,
+      "loss": 3.1383,
+      "step": 18345
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.319496063082919e-05,
+      "loss": 3.1633,
+      "step": 18350
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.318630574554978e-05,
+      "loss": 3.0748,
+      "step": 18355
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.317764908258143e-05,
+      "loss": 3.1902,
+      "step": 18360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.316899064238792e-05,
+      "loss": 3.025,
+      "step": 18365
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.316033042543301e-05,
+      "loss": 3.2137,
+      "step": 18370
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.315166843218061e-05,
+      "loss": 3.0638,
+      "step": 18375
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.314300466309473e-05,
+      "loss": 3.0757,
+      "step": 18380
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.313433911863945e-05,
+      "loss": 3.0472,
+      "step": 18385
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.312567179927895e-05,
+      "loss": 3.1732,
+      "step": 18390
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.311700270547754e-05,
+      "loss": 3.1583,
+      "step": 18395
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.310833183769958e-05,
+      "loss": 3.2602,
+      "step": 18400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.30996591964095e-05,
+      "loss": 3.0719,
+      "step": 18405
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.309098478207193e-05,
+      "loss": 3.1719,
+      "step": 18410
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.308230859515152e-05,
+      "loss": 2.9966,
+      "step": 18415
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.307363063611301e-05,
+      "loss": 3.0768,
+      "step": 18420
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.306495090542126e-05,
+      "loss": 2.9882,
+      "step": 18425
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.30562694035412e-05,
+      "loss": 2.9928,
+      "step": 18430
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.304758613093789e-05,
+      "loss": 3.1309,
+      "step": 18435
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.303890108807646e-05,
+      "loss": 3.2261,
+      "step": 18440
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.303021427542215e-05,
+      "loss": 3.1285,
+      "step": 18445
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.302152569344027e-05,
+      "loss": 3.1213,
+      "step": 18450
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.301283534259625e-05,
+      "loss": 2.969,
+      "step": 18455
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.300414322335561e-05,
+      "loss": 2.9785,
+      "step": 18460
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.299544933618394e-05,
+      "loss": 3.0393,
+      "step": 18465
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.298675368154698e-05,
+      "loss": 3.1107,
+      "step": 18470
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.297805625991051e-05,
+      "loss": 3.1053,
+      "step": 18475
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.29693570717404e-05,
+      "loss": 3.0992,
+      "step": 18480
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.296065611750269e-05,
+      "loss": 3.1834,
+      "step": 18485
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.295195339766341e-05,
+      "loss": 2.9605,
+      "step": 18490
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.29432489126888e-05,
+      "loss": 3.0437,
+      "step": 18495
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.293454266304506e-05,
+      "loss": 3.1082,
+      "step": 18500
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.292583464919862e-05,
+      "loss": 3.1402,
+      "step": 18505
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.29171248716159e-05,
+      "loss": 3.081,
+      "step": 18510
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.290841333076348e-05,
+      "loss": 3.2143,
+      "step": 18515
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.289970002710801e-05,
+      "loss": 3.1106,
+      "step": 18520
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.289098496111621e-05,
+      "loss": 3.0652,
+      "step": 18525
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.288226813325495e-05,
+      "loss": 3.3098,
+      "step": 18530
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.287354954399114e-05,
+      "loss": 3.1222,
+      "step": 18535
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.286482919379182e-05,
+      "loss": 3.1643,
+      "step": 18540
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.285610708312413e-05,
+      "loss": 3.1077,
+      "step": 18545
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.284738321245525e-05,
+      "loss": 3.1931,
+      "step": 18550
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.283865758225251e-05,
+      "loss": 3.0801,
+      "step": 18555
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.282993019298331e-05,
+      "loss": 3.0939,
+      "step": 18560
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.282120104511517e-05,
+      "loss": 3.1073,
+      "step": 18565
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.281247013911566e-05,
+      "loss": 3.2426,
+      "step": 18570
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.280373747545246e-05,
+      "loss": 3.0378,
+      "step": 18575
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.27950030545934e-05,
+      "loss": 2.9451,
+      "step": 18580
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.27862668770063e-05,
+      "loss": 3.1099,
+      "step": 18585
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.277752894315917e-05,
+      "loss": 3.1426,
+      "step": 18590
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.276878925352005e-05,
+      "loss": 3.1629,
+      "step": 18595
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.276004780855711e-05,
+      "loss": 3.0264,
+      "step": 18600
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.27513046087386e-05,
+      "loss": 3.125,
+      "step": 18605
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.274255965453286e-05,
+      "loss": 3.0824,
+      "step": 18610
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.273381294640833e-05,
+      "loss": 3.041,
+      "step": 18615
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.272506448483359e-05,
+      "loss": 3.1831,
+      "step": 18620
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.27163142702772e-05,
+      "loss": 3.0431,
+      "step": 18625
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.270756230320791e-05,
+      "loss": 3.0851,
+      "step": 18630
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.269880858409455e-05,
+      "loss": 3.1591,
+      "step": 18635
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.2690053113406e-05,
+      "loss": 2.9881,
+      "step": 18640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.26812958916113e-05,
+      "loss": 2.9784,
+      "step": 18645
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.267253691917952e-05,
+      "loss": 3.0425,
+      "step": 18650
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.266377619657986e-05,
+      "loss": 2.9618,
+      "step": 18655
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.26550137242816e-05,
+      "loss": 3.0802,
+      "step": 18660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.264624950275413e-05,
+      "loss": 2.9666,
+      "step": 18665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.263748353246693e-05,
+      "loss": 3.182,
+      "step": 18670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.262871581388954e-05,
+      "loss": 3.0734,
+      "step": 18675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.261994634749164e-05,
+      "loss": 3.0408,
+      "step": 18680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.261117513374297e-05,
+      "loss": 2.9979,
+      "step": 18685
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 8.260240217311338e-05,
+      "loss": 3.1951,
+      "step": 18690
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.259362746607282e-05,
+      "loss": 3.0481,
+      "step": 18695
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.258485101309133e-05,
+      "loss": 3.0553,
+      "step": 18700
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.2576072814639e-05,
+      "loss": 3.1234,
+      "step": 18705
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.25672928711861e-05,
+      "loss": 3.1556,
+      "step": 18710
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.255851118320291e-05,
+      "loss": 3.075,
+      "step": 18715
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.254972775115986e-05,
+      "loss": 2.9703,
+      "step": 18720
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.254094257552744e-05,
+      "loss": 3.1647,
+      "step": 18725
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.253215565677623e-05,
+      "loss": 3.0375,
+      "step": 18730
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.252336699537695e-05,
+      "loss": 3.1611,
+      "step": 18735
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.251457659180038e-05,
+      "loss": 3.1613,
+      "step": 18740
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.250578444651737e-05,
+      "loss": 3.1752,
+      "step": 18745
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.24969905599989e-05,
+      "loss": 3.0548,
+      "step": 18750
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.248819493271602e-05,
+      "loss": 3.0699,
+      "step": 18755
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.247939756513992e-05,
+      "loss": 2.994,
+      "step": 18760
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.24705984577418e-05,
+      "loss": 3.0897,
+      "step": 18765
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.246179761099306e-05,
+      "loss": 3.1389,
+      "step": 18770
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.245299502536507e-05,
+      "loss": 3.0475,
+      "step": 18775
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.24441907013294e-05,
+      "loss": 3.0241,
+      "step": 18780
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.243538463935766e-05,
+      "loss": 3.1497,
+      "step": 18785
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.242657683992156e-05,
+      "loss": 3.1607,
+      "step": 18790
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.241776730349291e-05,
+      "loss": 3.0968,
+      "step": 18795
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.240895603054362e-05,
+      "loss": 3.117,
+      "step": 18800
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.240014302154567e-05,
+      "loss": 3.0887,
+      "step": 18805
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.239132827697114e-05,
+      "loss": 3.0695,
+      "step": 18810
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.238251179729221e-05,
+      "loss": 3.1498,
+      "step": 18815
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.237369358298117e-05,
+      "loss": 3.0775,
+      "step": 18820
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.236487363451037e-05,
+      "loss": 3.108,
+      "step": 18825
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.235605195235227e-05,
+      "loss": 3.0779,
+      "step": 18830
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.234722853697942e-05,
+      "loss": 3.171,
+      "step": 18835
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.233840338886448e-05,
+      "loss": 3.1391,
+      "step": 18840
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.232957650848016e-05,
+      "loss": 3.0853,
+      "step": 18845
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.23207478962993e-05,
+      "loss": 3.1828,
+      "step": 18850
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.231191755279484e-05,
+      "loss": 3.0849,
+      "step": 18855
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.230308547843977e-05,
+      "loss": 3.0959,
+      "step": 18860
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.229425167370719e-05,
+      "loss": 3.0232,
+      "step": 18865
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.228541613907032e-05,
+      "loss": 3.159,
+      "step": 18870
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.227657887500247e-05,
+      "loss": 3.213,
+      "step": 18875
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.226773988197698e-05,
+      "loss": 2.9553,
+      "step": 18880
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.225889916046734e-05,
+      "loss": 3.0582,
+      "step": 18885
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.225005671094714e-05,
+      "loss": 3.0237,
+      "step": 18890
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.224121253389004e-05,
+      "loss": 3.0115,
+      "step": 18895
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.223236662976979e-05,
+      "loss": 3.102,
+      "step": 18900
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.222351899906023e-05,
+      "loss": 3.014,
+      "step": 18905
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.22146696422353e-05,
+      "loss": 3.0924,
+      "step": 18910
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.220581855976906e-05,
+      "loss": 3.0561,
+      "step": 18915
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.219696575213561e-05,
+      "loss": 3.1176,
+      "step": 18920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.218811121980917e-05,
+      "loss": 3.1521,
+      "step": 18925
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.217925496326404e-05,
+      "loss": 3.1208,
+      "step": 18930
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.217039698297464e-05,
+      "loss": 3.1079,
+      "step": 18935
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.216153727941547e-05,
+      "loss": 3.0981,
+      "step": 18940
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.21526758530611e-05,
+      "loss": 3.0857,
+      "step": 18945
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.214381270438625e-05,
+      "loss": 3.0193,
+      "step": 18950
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.213494783386563e-05,
+      "loss": 3.1355,
+      "step": 18955
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.212608124197415e-05,
+      "loss": 3.1294,
+      "step": 18960
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.211721292918672e-05,
+      "loss": 3.1248,
+      "step": 18965
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.210834289597844e-05,
+      "loss": 3.1437,
+      "step": 18970
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.209947114282442e-05,
+      "loss": 3.1253,
+      "step": 18975
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.20905976701999e-05,
+      "loss": 3.0057,
+      "step": 18980
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.208172247858022e-05,
+      "loss": 3.0232,
+      "step": 18985
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.207284556844078e-05,
+      "loss": 3.0277,
+      "step": 18990
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.206396694025709e-05,
+      "loss": 3.0577,
+      "step": 18995
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.205508659450476e-05,
+      "loss": 3.0094,
+      "step": 19000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.204620453165947e-05,
+      "loss": 3.0011,
+      "step": 19005
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.2037320752197e-05,
+      "loss": 3.0233,
+      "step": 19010
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.202843525659325e-05,
+      "loss": 2.9968,
+      "step": 19015
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.201954804532417e-05,
+      "loss": 3.1093,
+      "step": 19020
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.201065911886585e-05,
+      "loss": 3.0963,
+      "step": 19025
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.200176847769438e-05,
+      "loss": 3.0381,
+      "step": 19030
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.199287612228607e-05,
+      "loss": 3.0024,
+      "step": 19035
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.198398205311723e-05,
+      "loss": 2.9804,
+      "step": 19040
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.197508627066428e-05,
+      "loss": 3.1326,
+      "step": 19045
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.196618877540376e-05,
+      "loss": 3.1212,
+      "step": 19050
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.195728956781226e-05,
+      "loss": 3.1452,
+      "step": 19055
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.194838864836651e-05,
+      "loss": 3.1411,
+      "step": 19060
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.193948601754327e-05,
+      "loss": 3.1764,
+      "step": 19065
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.193058167581945e-05,
+      "loss": 3.1549,
+      "step": 19070
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.192167562367203e-05,
+      "loss": 3.0656,
+      "step": 19075
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.191276786157807e-05,
+      "loss": 3.1024,
+      "step": 19080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.190385839001473e-05,
+      "loss": 3.0733,
+      "step": 19085
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.189494720945928e-05,
+      "loss": 3.02,
+      "step": 19090
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.188603432038904e-05,
+      "loss": 3.081,
+      "step": 19095
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.187711972328148e-05,
+      "loss": 3.1033,
+      "step": 19100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.18682034186141e-05,
+      "loss": 3.1473,
+      "step": 19105
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.185928540686451e-05,
+      "loss": 2.9928,
+      "step": 19110
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.185036568851045e-05,
+      "loss": 3.0858,
+      "step": 19115
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.18414442640297e-05,
+      "loss": 3.0952,
+      "step": 19120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.183252113390016e-05,
+      "loss": 3.0857,
+      "step": 19125
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.182359629859983e-05,
+      "loss": 3.1213,
+      "step": 19130
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.181466975860677e-05,
+      "loss": 3.0591,
+      "step": 19135
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.180574151439915e-05,
+      "loss": 3.1896,
+      "step": 19140
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.179681156645522e-05,
+      "loss": 3.0561,
+      "step": 19145
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.178787991525337e-05,
+      "loss": 3.19,
+      "step": 19150
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.177894656127198e-05,
+      "loss": 3.1696,
+      "step": 19155
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.177001150498961e-05,
+      "loss": 3.1167,
+      "step": 19160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.17610747468849e-05,
+      "loss": 3.1489,
+      "step": 19165
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.175213628743655e-05,
+      "loss": 3.0955,
+      "step": 19170
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.174319612712336e-05,
+      "loss": 3.1297,
+      "step": 19175
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.173425426642425e-05,
+      "loss": 3.1187,
+      "step": 19180
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.172531070581819e-05,
+      "loss": 3.0765,
+      "step": 19185
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.171636544578424e-05,
+      "loss": 3.0821,
+      "step": 19190
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.170741848680162e-05,
+      "loss": 3.1667,
+      "step": 19195
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.169846982934955e-05,
+      "loss": 3.0917,
+      "step": 19200
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.168951947390741e-05,
+      "loss": 3.087,
+      "step": 19205
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.168056742095461e-05,
+      "loss": 3.0904,
+      "step": 19210
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.167161367097071e-05,
+      "loss": 3.2034,
+      "step": 19215
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.166265822443533e-05,
+      "loss": 3.0667,
+      "step": 19220
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.165370108182817e-05,
+      "loss": 3.1422,
+      "step": 19225
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.164474224362907e-05,
+      "loss": 3.1026,
+      "step": 19230
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.16357817103179e-05,
+      "loss": 3.177,
+      "step": 19235
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.162681948237465e-05,
+      "loss": 3.0979,
+      "step": 19240
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.161785556027942e-05,
+      "loss": 3.0736,
+      "step": 19245
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.160888994451235e-05,
+      "loss": 3.1035,
+      "step": 19250
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.159992263555372e-05,
+      "loss": 3.0223,
+      "step": 19255
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.159095363388388e-05,
+      "loss": 3.1693,
+      "step": 19260
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.158198293998328e-05,
+      "loss": 3.128,
+      "step": 19265
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.157301055433242e-05,
+      "loss": 3.1037,
+      "step": 19270
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.156403647741196e-05,
+      "loss": 3.1081,
+      "step": 19275
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.155506070970259e-05,
+      "loss": 3.075,
+      "step": 19280
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.154608325168514e-05,
+      "loss": 3.0478,
+      "step": 19285
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.153710410384046e-05,
+      "loss": 3.1303,
+      "step": 19290
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.152812326664958e-05,
+      "loss": 3.0739,
+      "step": 19295
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.151914074059357e-05,
+      "loss": 3.0798,
+      "step": 19300
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.151015652615357e-05,
+      "loss": 3.1419,
+      "step": 19305
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.150117062381086e-05,
+      "loss": 3.0802,
+      "step": 19310
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.149218303404678e-05,
+      "loss": 2.9025,
+      "step": 19315
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.148319375734276e-05,
+      "loss": 3.1071,
+      "step": 19320
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.147420279418036e-05,
+      "loss": 3.0688,
+      "step": 19325
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.146521014504116e-05,
+      "loss": 3.139,
+      "step": 19330
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.145621581040687e-05,
+      "loss": 2.9834,
+      "step": 19335
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.144721979075933e-05,
+      "loss": 3.169,
+      "step": 19340
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.143822208658038e-05,
+      "loss": 3.133,
+      "step": 19345
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.142922269835202e-05,
+      "loss": 3.0914,
+      "step": 19350
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.142022162655634e-05,
+      "loss": 3.1755,
+      "step": 19355
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.141121887167547e-05,
+      "loss": 3.1082,
+      "step": 19360
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.140221443419167e-05,
+      "loss": 3.1127,
+      "step": 19365
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 8.139320831458728e-05,
+      "loss": 2.9791,
+      "step": 19370
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.138420051334474e-05,
+      "loss": 3.0873,
+      "step": 19375
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.137519103094657e-05,
+      "loss": 3.1736,
+      "step": 19380
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.136617986787535e-05,
+      "loss": 3.059,
+      "step": 19385
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.13571670246138e-05,
+      "loss": 3.0395,
+      "step": 19390
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.134815250164473e-05,
+      "loss": 3.2845,
+      "step": 19395
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.133913629945099e-05,
+      "loss": 3.0228,
+      "step": 19400
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.133011841851558e-05,
+      "loss": 3.0335,
+      "step": 19405
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.132109885932153e-05,
+      "loss": 3.1124,
+      "step": 19410
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.1312077622352e-05,
+      "loss": 2.9805,
+      "step": 19415
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.130305470809025e-05,
+      "loss": 3.1195,
+      "step": 19420
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.129403011701958e-05,
+      "loss": 3.0629,
+      "step": 19425
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.128500384962342e-05,
+      "loss": 3.1178,
+      "step": 19430
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.127597590638531e-05,
+      "loss": 3.1629,
+      "step": 19435
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.12669462877888e-05,
+      "loss": 3.2113,
+      "step": 19440
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.125791499431762e-05,
+      "loss": 3.0295,
+      "step": 19445
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.124888202645551e-05,
+      "loss": 3.1677,
+      "step": 19450
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.123984738468637e-05,
+      "loss": 3.153,
+      "step": 19455
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.123081106949415e-05,
+      "loss": 2.9473,
+      "step": 19460
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.12217730813629e-05,
+      "loss": 3.0997,
+      "step": 19465
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.121273342077673e-05,
+      "loss": 3.2174,
+      "step": 19470
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.12036920882199e-05,
+      "loss": 3.001,
+      "step": 19475
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.119464908417673e-05,
+      "loss": 3.0771,
+      "step": 19480
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.118560440913159e-05,
+      "loss": 3.0169,
+      "step": 19485
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.117655806356901e-05,
+      "loss": 3.1457,
+      "step": 19490
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.116751004797356e-05,
+      "loss": 3.0998,
+      "step": 19495
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.115846036282993e-05,
+      "loss": 3.041,
+      "step": 19500
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.114940900862285e-05,
+      "loss": 3.0244,
+      "step": 19505
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.114035598583722e-05,
+      "loss": 3.248,
+      "step": 19510
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.113130129495794e-05,
+      "loss": 3.0418,
+      "step": 19515
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.112224493647005e-05,
+      "loss": 3.1459,
+      "step": 19520
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.111318691085869e-05,
+      "loss": 3.1561,
+      "step": 19525
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.110412721860906e-05,
+      "loss": 3.2115,
+      "step": 19530
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.109506586020647e-05,
+      "loss": 3.0693,
+      "step": 19535
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.10860028361363e-05,
+      "loss": 3.1357,
+      "step": 19540
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.107693814688402e-05,
+      "loss": 3.0661,
+      "step": 19545
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.106787179293521e-05,
+      "loss": 3.0414,
+      "step": 19550
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.105880377477553e-05,
+      "loss": 2.9898,
+      "step": 19555
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.104973409289071e-05,
+      "loss": 3.0478,
+      "step": 19560
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.10406627477666e-05,
+      "loss": 3.1293,
+      "step": 19565
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.103158973988912e-05,
+      "loss": 3.0145,
+      "step": 19570
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.102251506974427e-05,
+      "loss": 3.0534,
+      "step": 19575
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.101343873781818e-05,
+      "loss": 3.1575,
+      "step": 19580
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.100436074459704e-05,
+      "loss": 3.1484,
+      "step": 19585
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.099528109056709e-05,
+      "loss": 3.1135,
+      "step": 19590
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.098619977621474e-05,
+      "loss": 2.7772,
+      "step": 19595
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.097711680202642e-05,
+      "loss": 3.0481,
+      "step": 19600
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.09680321684887e-05,
+      "loss": 2.9611,
+      "step": 19605
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.095894587608821e-05,
+      "loss": 3.1126,
+      "step": 19610
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.094985792531166e-05,
+      "loss": 3.2027,
+      "step": 19615
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.094076831664588e-05,
+      "loss": 2.9176,
+      "step": 19620
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.093167705057777e-05,
+      "loss": 3.1318,
+      "step": 19625
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.092258412759433e-05,
+      "loss": 3.1434,
+      "step": 19630
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.091348954818262e-05,
+      "loss": 3.0184,
+      "step": 19635
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.090439331282982e-05,
+      "loss": 3.0658,
+      "step": 19640
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.089529542202319e-05,
+      "loss": 3.1559,
+      "step": 19645
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.088619587625006e-05,
+      "loss": 3.1565,
+      "step": 19650
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.087709467599788e-05,
+      "loss": 3.2006,
+      "step": 19655
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.086799182175417e-05,
+      "loss": 3.088,
+      "step": 19660
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.085888731400656e-05,
+      "loss": 3.0407,
+      "step": 19665
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.084978115324271e-05,
+      "loss": 3.0303,
+      "step": 19670
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.084067333995045e-05,
+      "loss": 2.9071,
+      "step": 19675
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.083156387461764e-05,
+      "loss": 3.0769,
+      "step": 19680
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.082245275773223e-05,
+      "loss": 3.1518,
+      "step": 19685
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.081333998978231e-05,
+      "loss": 3.0798,
+      "step": 19690
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.080422557125599e-05,
+      "loss": 3.1869,
+      "step": 19695
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.079510950264152e-05,
+      "loss": 3.1415,
+      "step": 19700
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.07859917844272e-05,
+      "loss": 3.113,
+      "step": 19705
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.077687241710149e-05,
+      "loss": 2.9998,
+      "step": 19710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.076775140115283e-05,
+      "loss": 3.1162,
+      "step": 19715
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.075862873706981e-05,
+      "loss": 3.0402,
+      "step": 19720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.074950442534113e-05,
+      "loss": 3.1132,
+      "step": 19725
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.074037846645555e-05,
+      "loss": 3.079,
+      "step": 19730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.073125086090191e-05,
+      "loss": 2.9761,
+      "step": 19735
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.072212160916915e-05,
+      "loss": 3.0526,
+      "step": 19740
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.071299071174628e-05,
+      "loss": 3.0696,
+      "step": 19745
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.070385816912242e-05,
+      "loss": 3.0397,
+      "step": 19750
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.069472398178679e-05,
+      "loss": 3.0944,
+      "step": 19755
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.068558815022869e-05,
+      "loss": 3.0485,
+      "step": 19760
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.067645067493746e-05,
+      "loss": 2.9977,
+      "step": 19765
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.066731155640258e-05,
+      "loss": 3.0711,
+      "step": 19770
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.065817079511362e-05,
+      "loss": 3.0192,
+      "step": 19775
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.064902839156021e-05,
+      "loss": 3.1629,
+      "step": 19780
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.063988434623209e-05,
+      "loss": 3.1574,
+      "step": 19785
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.063073865961906e-05,
+      "loss": 3.1102,
+      "step": 19790
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.062159133221106e-05,
+      "loss": 3.1616,
+      "step": 19795
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.061244236449804e-05,
+      "loss": 3.1513,
+      "step": 19800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.060329175697012e-05,
+      "loss": 3.0226,
+      "step": 19805
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.059413951011746e-05,
+      "loss": 3.1575,
+      "step": 19810
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.05849856244303e-05,
+      "loss": 3.2058,
+      "step": 19815
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.0575830100399e-05,
+      "loss": 3.0244,
+      "step": 19820
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.056667293851398e-05,
+      "loss": 3.133,
+      "step": 19825
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.05575141392658e-05,
+      "loss": 3.2488,
+      "step": 19830
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.054835370314504e-05,
+      "loss": 3.1098,
+      "step": 19835
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.053919163064238e-05,
+      "loss": 2.9555,
+      "step": 19840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.053002792224864e-05,
+      "loss": 3.041,
+      "step": 19845
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.052086257845468e-05,
+      "loss": 3.0688,
+      "step": 19850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.051169559975143e-05,
+      "loss": 3.0953,
+      "step": 19855
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.050252698662998e-05,
+      "loss": 3.093,
+      "step": 19860
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.049335673958145e-05,
+      "loss": 3.2103,
+      "step": 19865
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.048418485909703e-05,
+      "loss": 2.9145,
+      "step": 19870
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.047501134566809e-05,
+      "loss": 2.9641,
+      "step": 19875
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.046583619978598e-05,
+      "loss": 2.9833,
+      "step": 19880
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.04566594219422e-05,
+      "loss": 3.0943,
+      "step": 19885
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.044748101262833e-05,
+      "loss": 3.2402,
+      "step": 19890
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.043830097233601e-05,
+      "loss": 3.1914,
+      "step": 19895
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.042911930155698e-05,
+      "loss": 3.0905,
+      "step": 19900
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.041993600078312e-05,
+      "loss": 3.1272,
+      "step": 19905
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.041075107050632e-05,
+      "loss": 3.0555,
+      "step": 19910
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.040156451121859e-05,
+      "loss": 3.128,
+      "step": 19915
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.0392376323412e-05,
+      "loss": 3.1389,
+      "step": 19920
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.038318650757878e-05,
+      "loss": 3.0639,
+      "step": 19925
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.037399506421119e-05,
+      "loss": 3.1116,
+      "step": 19930
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.036480199380157e-05,
+      "loss": 3.2228,
+      "step": 19935
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.035560729684238e-05,
+      "loss": 3.0704,
+      "step": 19940
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.034641097382613e-05,
+      "loss": 3.1508,
+      "step": 19945
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.033721302524547e-05,
+      "loss": 2.9602,
+      "step": 19950
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.032801345159309e-05,
+      "loss": 2.9806,
+      "step": 19955
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.031881225336175e-05,
+      "loss": 3.1332,
+      "step": 19960
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.03096094310444e-05,
+      "loss": 3.1769,
+      "step": 19965
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.030040498513396e-05,
+      "loss": 3.0925,
+      "step": 19970
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.029119891612349e-05,
+      "loss": 3.1431,
+      "step": 19975
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.028199122450612e-05,
+      "loss": 2.9594,
+      "step": 19980
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.027278191077511e-05,
+      "loss": 3.1007,
+      "step": 19985
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.026357097542376e-05,
+      "loss": 3.1405,
+      "step": 19990
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.025435841894546e-05,
+      "loss": 3.1624,
+      "step": 19995
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.02451442418337e-05,
+      "loss": 3.0712,
+      "step": 20000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.023592844458207e-05,
+      "loss": 3.0585,
+      "step": 20005
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.022671102768421e-05,
+      "loss": 3.0445,
+      "step": 20010
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.021749199163392e-05,
+      "loss": 3.0109,
+      "step": 20015
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.020827133692498e-05,
+      "loss": 2.8868,
+      "step": 20020
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.019904906405132e-05,
+      "loss": 3.0573,
+      "step": 20025
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.018982517350697e-05,
+      "loss": 3.0616,
+      "step": 20030
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.018059966578602e-05,
+      "loss": 2.9909,
+      "step": 20035
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.017137254138263e-05,
+      "loss": 3.0087,
+      "step": 20040
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.01621438007911e-05,
+      "loss": 3.1498,
+      "step": 20045
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.015291344450579e-05,
+      "loss": 3.2025,
+      "step": 20050
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.014368147302109e-05,
+      "loss": 3.0332,
+      "step": 20055
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.013444788683157e-05,
+      "loss": 3.1302,
+      "step": 20060
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.012521268643186e-05,
+      "loss": 3.2634,
+      "step": 20065
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.011597587231663e-05,
+      "loss": 2.9636,
+      "step": 20070
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.010673744498067e-05,
+      "loss": 3.0862,
+      "step": 20075
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.009749740491885e-05,
+      "loss": 3.0698,
+      "step": 20080
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.008825575262617e-05,
+      "loss": 3.0123,
+      "step": 20085
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.007901248859764e-05,
+      "loss": 3.0274,
+      "step": 20090
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.006976761332839e-05,
+      "loss": 3.0259,
+      "step": 20095
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.006052112731365e-05,
+      "loss": 2.9735,
+      "step": 20100
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.005127303104875e-05,
+      "loss": 2.9916,
+      "step": 20105
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.004202332502905e-05,
+      "loss": 3.1463,
+      "step": 20110
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.003277200975004e-05,
+      "loss": 3.1642,
+      "step": 20115
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.002351908570728e-05,
+      "loss": 2.9128,
+      "step": 20120
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.001426455339643e-05,
+      "loss": 3.0947,
+      "step": 20125
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.00050084133132e-05,
+      "loss": 3.0099,
+      "step": 20130
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.999575066595347e-05,
+      "loss": 3.1863,
+      "step": 20135
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.998649131181308e-05,
+      "loss": 3.1572,
+      "step": 20140
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.997723035138808e-05,
+      "loss": 3.0862,
+      "step": 20145
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.996796778517452e-05,
+      "loss": 3.1731,
+      "step": 20150
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.995870361366857e-05,
+      "loss": 3.1785,
+      "step": 20155
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.99494378373665e-05,
+      "loss": 3.107,
+      "step": 20160
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.994017045676465e-05,
+      "loss": 3.161,
+      "step": 20165
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.993090147235944e-05,
+      "loss": 3.0952,
+      "step": 20170
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.992163088464736e-05,
+      "loss": 3.0473,
+      "step": 20175
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.991235869412502e-05,
+      "loss": 3.1162,
+      "step": 20180
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.990308490128912e-05,
+      "loss": 2.9421,
+      "step": 20185
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.989380950663641e-05,
+      "loss": 3.22,
+      "step": 20190
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.988453251066375e-05,
+      "loss": 3.131,
+      "step": 20195
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.987525391386807e-05,
+      "loss": 2.998,
+      "step": 20200
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.986597371674643e-05,
+      "loss": 3.1249,
+      "step": 20205
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.98566919197959e-05,
+      "loss": 3.1402,
+      "step": 20210
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.984740852351369e-05,
+      "loss": 3.1333,
+      "step": 20215
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.98381235283971e-05,
+      "loss": 3.1088,
+      "step": 20220
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.982883693494349e-05,
+      "loss": 3.1226,
+      "step": 20225
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.98195487436503e-05,
+      "loss": 3.0494,
+      "step": 20230
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.981025895501509e-05,
+      "loss": 3.1474,
+      "step": 20235
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.980096756953548e-05,
+      "loss": 3.0731,
+      "step": 20240
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.979167458770917e-05,
+      "loss": 2.9618,
+      "step": 20245
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.978238001003397e-05,
+      "loss": 2.9456,
+      "step": 20250
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.977308383700774e-05,
+      "loss": 3.0471,
+      "step": 20255
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.976378606912849e-05,
+      "loss": 3.0,
+      "step": 20260
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.975448670689424e-05,
+      "loss": 3.161,
+      "step": 20265
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.974518575080314e-05,
+      "loss": 2.8679,
+      "step": 20270
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.97358832013534e-05,
+      "loss": 3.1187,
+      "step": 20275
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.972657905904335e-05,
+      "loss": 3.0397,
+      "step": 20280
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.971727332437135e-05,
+      "loss": 2.9768,
+      "step": 20285
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.970796599783592e-05,
+      "loss": 3.0054,
+      "step": 20290
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.969865707993562e-05,
+      "loss": 3.0054,
+      "step": 20295
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.968934657116908e-05,
+      "loss": 3.0186,
+      "step": 20300
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.968003447203503e-05,
+      "loss": 3.1445,
+      "step": 20305
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.967072078303231e-05,
+      "loss": 3.1091,
+      "step": 20310
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.966140550465983e-05,
+      "loss": 3.0782,
+      "step": 20315
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.965208863741658e-05,
+      "loss": 2.9001,
+      "step": 20320
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.96427701818016e-05,
+      "loss": 3.1566,
+      "step": 20325
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.963345013831409e-05,
+      "loss": 3.0635,
+      "step": 20330
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.962412850745329e-05,
+      "loss": 3.0744,
+      "step": 20335
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.961480528971852e-05,
+      "loss": 3.0706,
+      "step": 20340
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.960548048560921e-05,
+      "loss": 3.1091,
+      "step": 20345
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.959615409562487e-05,
+      "loss": 3.2196,
+      "step": 20350
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.958682612026503e-05,
+      "loss": 3.0713,
+      "step": 20355
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.957749656002942e-05,
+      "loss": 3.0732,
+      "step": 20360
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.956816541541778e-05,
+      "loss": 3.0552,
+      "step": 20365
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.955883268692995e-05,
+      "loss": 3.0385,
+      "step": 20370
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.954949837506585e-05,
+      "loss": 3.0345,
+      "step": 20375
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.954016248032549e-05,
+      "loss": 2.966,
+      "step": 20380
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.953082500320897e-05,
+      "loss": 3.0781,
+      "step": 20385
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.952148594421647e-05,
+      "loss": 3.115,
+      "step": 20390
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.951214530384827e-05,
+      "loss": 3.0264,
+      "step": 20395
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.950280308260467e-05,
+      "loss": 3.0723,
+      "step": 20400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.949345928098617e-05,
+      "loss": 2.9919,
+      "step": 20405
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.948411389949324e-05,
+      "loss": 3.0419,
+      "step": 20410
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.94747669386265e-05,
+      "loss": 3.1068,
+      "step": 20415
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.946541839888666e-05,
+      "loss": 3.0327,
+      "step": 20420
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.945606828077446e-05,
+      "loss": 3.0792,
+      "step": 20425
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.944671658479074e-05,
+      "loss": 3.0556,
+      "step": 20430
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.94373633114365e-05,
+      "loss": 3.1663,
+      "step": 20435
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.942800846121273e-05,
+      "loss": 3.0536,
+      "step": 20440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.941865203462056e-05,
+      "loss": 3.035,
+      "step": 20445
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.940929403216115e-05,
+      "loss": 3.2332,
+      "step": 20450
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.939993445433581e-05,
+      "loss": 2.9594,
+      "step": 20455
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.93905733016459e-05,
+      "loss": 3.0639,
+      "step": 20460
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.938121057459286e-05,
+      "loss": 3.0129,
+      "step": 20465
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.937184627367825e-05,
+      "loss": 3.1379,
+      "step": 20470
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.936248039940362e-05,
+      "loss": 3.1102,
+      "step": 20475
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.935311295227074e-05,
+      "loss": 3.1246,
+      "step": 20480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.934374393278136e-05,
+      "loss": 2.9195,
+      "step": 20485
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.933437334143738e-05,
+      "loss": 3.2044,
+      "step": 20490
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.932500117874072e-05,
+      "loss": 2.998,
+      "step": 20495
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.931562744519343e-05,
+      "loss": 3.158,
+      "step": 20500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.930625214129766e-05,
+      "loss": 3.0618,
+      "step": 20505
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.929687526755556e-05,
+      "loss": 3.0669,
+      "step": 20510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.928749682446945e-05,
+      "loss": 3.1302,
+      "step": 20515
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.927811681254173e-05,
+      "loss": 3.099,
+      "step": 20520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.926873523227482e-05,
+      "loss": 3.1863,
+      "step": 20525
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.925935208417127e-05,
+      "loss": 3.1438,
+      "step": 20530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.924996736873373e-05,
+      "loss": 3.1376,
+      "step": 20535
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.924058108646488e-05,
+      "loss": 3.1095,
+      "step": 20540
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.923119323786753e-05,
+      "loss": 3.0513,
+      "step": 20545
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.922180382344458e-05,
+      "loss": 3.1833,
+      "step": 20550
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.921241284369894e-05,
+      "loss": 3.0484,
+      "step": 20555
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.920302029913369e-05,
+      "loss": 3.0238,
+      "step": 20560
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.919362619025197e-05,
+      "loss": 3.0553,
+      "step": 20565
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.918423051755697e-05,
+      "loss": 3.1126,
+      "step": 20570
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.917483328155199e-05,
+      "loss": 3.013,
+      "step": 20575
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.916543448274043e-05,
+      "loss": 2.9698,
+      "step": 20580
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.915603412162573e-05,
+      "loss": 3.1305,
+      "step": 20585
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.914663219871146e-05,
+      "loss": 3.0795,
+      "step": 20590
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.913722871450123e-05,
+      "loss": 2.9205,
+      "step": 20595
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.91278236694988e-05,
+      "loss": 3.1637,
+      "step": 20600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.91184170642079e-05,
+      "loss": 3.0521,
+      "step": 20605
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.910900889913247e-05,
+      "loss": 3.092,
+      "step": 20610
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.909959917477645e-05,
+      "loss": 3.1386,
+      "step": 20615
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.909018789164391e-05,
+      "loss": 3.1058,
+      "step": 20620
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.908077505023896e-05,
+      "loss": 2.918,
+      "step": 20625
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.907136065106584e-05,
+      "loss": 3.1928,
+      "step": 20630
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.906194469462885e-05,
+      "loss": 3.1219,
+      "step": 20635
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.905252718143234e-05,
+      "loss": 3.1647,
+      "step": 20640
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.90431081119808e-05,
+      "loss": 3.1098,
+      "step": 20645
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.90336874867788e-05,
+      "loss": 3.1429,
+      "step": 20650
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.902426530633095e-05,
+      "loss": 3.0968,
+      "step": 20655
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.901484157114196e-05,
+      "loss": 3.1382,
+      "step": 20660
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.900541628171666e-05,
+      "loss": 3.0365,
+      "step": 20665
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.89959894385599e-05,
+      "loss": 3.0347,
+      "step": 20670
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.898656104217668e-05,
+      "loss": 3.0381,
+      "step": 20675
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.897713109307203e-05,
+      "loss": 3.2382,
+      "step": 20680
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.89676995917511e-05,
+      "loss": 3.1911,
+      "step": 20685
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.895826653871909e-05,
+      "loss": 2.9701,
+      "step": 20690
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.894883193448131e-05,
+      "loss": 3.0492,
+      "step": 20695
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.893939577954313e-05,
+      "loss": 3.1642,
+      "step": 20700
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.892995807441001e-05,
+      "loss": 3.0979,
+      "step": 20705
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.892051881958754e-05,
+      "loss": 3.1623,
+      "step": 20710
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.89110780155813e-05,
+      "loss": 3.2047,
+      "step": 20715
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.890163566289704e-05,
+      "loss": 3.0168,
+      "step": 20720
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.889219176204056e-05,
+      "loss": 2.9674,
+      "step": 20725
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.888274631351773e-05,
+      "loss": 3.094,
+      "step": 20730
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.887329931783449e-05,
+      "loss": 3.1133,
+      "step": 20735
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.886385077549693e-05,
+      "loss": 3.029,
+      "step": 20740
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.885440068701116e-05,
+      "loss": 3.0162,
+      "step": 20745
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.884494905288338e-05,
+      "loss": 2.9226,
+      "step": 20750
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.883549587361989e-05,
+      "loss": 3.1479,
+      "step": 20755
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.882604114972709e-05,
+      "loss": 3.048,
+      "step": 20760
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.881658488171142e-05,
+      "loss": 3.0813,
+      "step": 20765
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.880712707007943e-05,
+      "loss": 3.0825,
+      "step": 20770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.879766771533774e-05,
+      "loss": 3.06,
+      "step": 20775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.878820681799308e-05,
+      "loss": 3.0499,
+      "step": 20780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.877874437855219e-05,
+      "loss": 3.0323,
+      "step": 20785
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.8769280397522e-05,
+      "loss": 3.1045,
+      "step": 20790
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.875981487540944e-05,
+      "loss": 3.0201,
+      "step": 20795
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.875034781272157e-05,
+      "loss": 3.1598,
+      "step": 20800
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.874087920996549e-05,
+      "loss": 3.2272,
+      "step": 20805
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.87314090676484e-05,
+      "loss": 3.0193,
+      "step": 20810
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.872193738627759e-05,
+      "loss": 3.1278,
+      "step": 20815
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.871246416636045e-05,
+      "loss": 2.9622,
+      "step": 20820
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.87029894084044e-05,
+      "loss": 3.0102,
+      "step": 20825
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.869351311291701e-05,
+      "loss": 3.139,
+      "step": 20830
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.868403528040587e-05,
+      "loss": 3.1559,
+      "step": 20835
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.867455591137868e-05,
+      "loss": 2.9374,
+      "step": 20840
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.866507500634324e-05,
+      "loss": 3.139,
+      "step": 20845
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.865559256580737e-05,
+      "loss": 3.1453,
+      "step": 20850
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.864610859027908e-05,
+      "loss": 3.1436,
+      "step": 20855
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.863662308026633e-05,
+      "loss": 3.077,
+      "step": 20860
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.862713603627729e-05,
+      "loss": 3.071,
+      "step": 20865
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.86176474588201e-05,
+      "loss": 3.0549,
+      "step": 20870
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.860815734840307e-05,
+      "loss": 3.1972,
+      "step": 20875
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.859866570553454e-05,
+      "loss": 2.997,
+      "step": 20880
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.858917253072296e-05,
+      "loss": 3.1463,
+      "step": 20885
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.857967782447684e-05,
+      "loss": 3.1309,
+      "step": 20890
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.857018158730478e-05,
+      "loss": 2.9753,
+      "step": 20895
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.856068381971546e-05,
+      "loss": 2.9626,
+      "step": 20900
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.855118452221768e-05,
+      "loss": 3.1106,
+      "step": 20905
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.854168369532024e-05,
+      "loss": 3.1804,
+      "step": 20910
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.853218133953211e-05,
+      "loss": 3.1462,
+      "step": 20915
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.852267745536228e-05,
+      "loss": 2.9258,
+      "step": 20920
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.851317204331986e-05,
+      "loss": 2.9682,
+      "step": 20925
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.8503665103914e-05,
+      "loss": 3.0741,
+      "step": 20930
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.849415663765398e-05,
+      "loss": 3.0304,
+      "step": 20935
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.848464664504914e-05,
+      "loss": 2.9723,
+      "step": 20940
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.847513512660889e-05,
+      "loss": 3.0295,
+      "step": 20945
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.846562208284273e-05,
+      "loss": 3.1716,
+      "step": 20950
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.845610751426027e-05,
+      "loss": 3.1621,
+      "step": 20955
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.844659142137115e-05,
+      "loss": 3.0929,
+      "step": 20960
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.843707380468512e-05,
+      "loss": 3.0708,
+      "step": 20965
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.842755466471203e-05,
+      "loss": 3.1331,
+      "step": 20970
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.841803400196176e-05,
+      "loss": 3.0766,
+      "step": 20975
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.840851181694433e-05,
+      "loss": 3.1954,
+      "step": 20980
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.839898811016981e-05,
+      "loss": 3.0916,
+      "step": 20985
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.838946288214834e-05,
+      "loss": 3.1179,
+      "step": 20990
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.837993613339017e-05,
+      "loss": 3.0274,
+      "step": 20995
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.837040786440562e-05,
+      "loss": 3.0682,
+      "step": 21000
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.836087807570508e-05,
+      "loss": 3.0785,
+      "step": 21005
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.835134676779906e-05,
+      "loss": 3.0164,
+      "step": 21010
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.834181394119807e-05,
+      "loss": 3.1498,
+      "step": 21015
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.833227959641279e-05,
+      "loss": 3.2319,
+      "step": 21020
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.832274373395395e-05,
+      "loss": 3.1816,
+      "step": 21025
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.831320635433235e-05,
+      "loss": 3.0877,
+      "step": 21030
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.830366745805889e-05,
+      "loss": 3.0587,
+      "step": 21035
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.82941270456445e-05,
+      "loss": 3.0385,
+      "step": 21040
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.828458511760027e-05,
+      "loss": 3.0634,
+      "step": 21045
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.827504167443732e-05,
+      "loss": 2.9865,
+      "step": 21050
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.826549671666686e-05,
+      "loss": 3.151,
+      "step": 21055
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.82559502448002e-05,
+      "loss": 3.1033,
+      "step": 21060
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.824640225934867e-05,
+      "loss": 3.2161,
+      "step": 21065
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.823685276082378e-05,
+      "loss": 3.06,
+      "step": 21070
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.822730174973704e-05,
+      "loss": 3.0608,
+      "step": 21075
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.821774922660007e-05,
+      "loss": 3.0293,
+      "step": 21080
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.820819519192457e-05,
+      "loss": 3.0051,
+      "step": 21085
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.819863964622234e-05,
+      "loss": 3.0256,
+      "step": 21090
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.818908259000521e-05,
+      "loss": 3.0368,
+      "step": 21095
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.817952402378515e-05,
+      "loss": 3.0517,
+      "step": 21100
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.816996394807416e-05,
+      "loss": 2.9668,
+      "step": 21105
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.816040236338436e-05,
+      "loss": 3.0678,
+      "step": 21110
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.815083927022792e-05,
+      "loss": 3.1025,
+      "step": 21115
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.814127466911712e-05,
+      "loss": 3.0877,
+      "step": 21120
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.81317085605643e-05,
+      "loss": 3.1889,
+      "step": 21125
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.81221409450819e-05,
+      "loss": 3.0058,
+      "step": 21130
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.811257182318239e-05,
+      "loss": 3.1181,
+      "step": 21135
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.81030011953784e-05,
+      "loss": 3.0862,
+      "step": 21140
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.809342906218258e-05,
+      "loss": 3.2071,
+      "step": 21145
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.80838554241077e-05,
+      "loss": 3.1191,
+      "step": 21150
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.807428028166655e-05,
+      "loss": 3.0803,
+      "step": 21155
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.806470363537209e-05,
+      "loss": 3.168,
+      "step": 21160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.805512548573728e-05,
+      "loss": 3.0503,
+      "step": 21165
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.804554583327519e-05,
+      "loss": 3.1161,
+      "step": 21170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.803596467849898e-05,
+      "loss": 3.0539,
+      "step": 21175
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.80263820219219e-05,
+      "loss": 2.9435,
+      "step": 21180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.801679786405725e-05,
+      "loss": 2.9792,
+      "step": 21185
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.800721220541841e-05,
+      "loss": 3.1011,
+      "step": 21190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.799762504651888e-05,
+      "loss": 2.9307,
+      "step": 21195
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.79880363878722e-05,
+      "loss": 3.1084,
+      "step": 21200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.797844622999198e-05,
+      "loss": 3.1723,
+      "step": 21205
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.7968854573392e-05,
+      "loss": 3.0649,
+      "step": 21210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.7959261418586e-05,
+      "loss": 2.9992,
+      "step": 21215
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.794966676608788e-05,
+      "loss": 3.0252,
+      "step": 21220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.794007061641158e-05,
+      "loss": 3.019,
+      "step": 21225
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.793047297007116e-05,
+      "loss": 3.0049,
+      "step": 21230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.792087382758071e-05,
+      "loss": 3.008,
+      "step": 21235
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.791127318945446e-05,
+      "loss": 3.1869,
+      "step": 21240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.790167105620665e-05,
+      "loss": 3.1248,
+      "step": 21245
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.789206742835167e-05,
+      "loss": 3.0054,
+      "step": 21250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.788246230640393e-05,
+      "loss": 3.0796,
+      "step": 21255
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.787285569087795e-05,
+      "loss": 3.017,
+      "step": 21260
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.786324758228834e-05,
+      "loss": 3.0487,
+      "step": 21265
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.785363798114978e-05,
+      "loss": 3.1772,
+      "step": 21270
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.7844026887977e-05,
+      "loss": 3.1429,
+      "step": 21275
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.783441430328489e-05,
+      "loss": 3.1934,
+      "step": 21280
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.782480022758831e-05,
+      "loss": 3.2294,
+      "step": 21285
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.781518466140227e-05,
+      "loss": 3.0793,
+      "step": 21290
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.780556760524187e-05,
+      "loss": 3.0934,
+      "step": 21295
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.779594905962226e-05,
+      "loss": 3.0699,
+      "step": 21300
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.778632902505864e-05,
+      "loss": 3.156,
+      "step": 21305
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.777670750206638e-05,
+      "loss": 3.1178,
+      "step": 21310
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.776708449116083e-05,
+      "loss": 3.0412,
+      "step": 21315
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.77574599928575e-05,
+      "loss": 2.9979,
+      "step": 21320
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.774783400767192e-05,
+      "loss": 3.1834,
+      "step": 21325
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.773820653611974e-05,
+      "loss": 3.2499,
+      "step": 21330
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.772857757871668e-05,
+      "loss": 3.2645,
+      "step": 21335
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.771894713597851e-05,
+      "loss": 3.1088,
+      "step": 21340
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.77093152084211e-05,
+      "loss": 3.1515,
+      "step": 21345
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.769968179656044e-05,
+      "loss": 2.9819,
+      "step": 21350
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.769004690091253e-05,
+      "loss": 3.1179,
+      "step": 21355
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.76804105219935e-05,
+      "loss": 3.2043,
+      "step": 21360
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.767077266031955e-05,
+      "loss": 2.9709,
+      "step": 21365
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.766113331640689e-05,
+      "loss": 3.0821,
+      "step": 21370
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.765149249077195e-05,
+      "loss": 3.0061,
+      "step": 21375
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.76418501839311e-05,
+      "loss": 3.1164,
+      "step": 21380
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.763220639640091e-05,
+      "loss": 3.1789,
+      "step": 21385
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.76225611286979e-05,
+      "loss": 2.9965,
+      "step": 21390
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.761291438133878e-05,
+      "loss": 3.1809,
+      "step": 21395
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.76032661548403e-05,
+      "loss": 2.963,
+      "step": 21400
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.759361644971925e-05,
+      "loss": 3.0112,
+      "step": 21405
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.758396526649257e-05,
+      "loss": 3.1382,
+      "step": 21410
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.757431260567722e-05,
+      "loss": 2.9124,
+      "step": 21415
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.756465846779028e-05,
+      "loss": 3.0244,
+      "step": 21420
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.755500285334889e-05,
+      "loss": 3.1078,
+      "step": 21425
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.754534576287027e-05,
+      "loss": 3.0384,
+      "step": 21430
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.753568719687172e-05,
+      "loss": 3.236,
+      "step": 21435
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.75260271558706e-05,
+      "loss": 3.0519,
+      "step": 21440
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.751636564038442e-05,
+      "loss": 3.0109,
+      "step": 21445
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.750670265093067e-05,
+      "loss": 3.1848,
+      "step": 21450
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.749703818802698e-05,
+      "loss": 3.0572,
+      "step": 21455
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.748737225219107e-05,
+      "loss": 3.0672,
+      "step": 21460
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.747770484394068e-05,
+      "loss": 3.0272,
+      "step": 21465
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.746803596379367e-05,
+      "loss": 3.0877,
+      "step": 21470
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.745836561226799e-05,
+      "loss": 3.1375,
+      "step": 21475
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.744869378988165e-05,
+      "loss": 2.9483,
+      "step": 21480
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.743902049715271e-05,
+      "loss": 3.1363,
+      "step": 21485
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.742934573459936e-05,
+      "loss": 3.1228,
+      "step": 21490
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.741966950273987e-05,
+      "loss": 3.143,
+      "step": 21495
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.740999180209252e-05,
+      "loss": 3.0538,
+      "step": 21500
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.740031263317575e-05,
+      "loss": 3.1506,
+      "step": 21505
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.739063199650803e-05,
+      "loss": 3.1849,
+      "step": 21510
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.738094989260791e-05,
+      "loss": 3.0798,
+      "step": 21515
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.737126632199405e-05,
+      "loss": 3.0763,
+      "step": 21520
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.736158128518518e-05,
+      "loss": 3.1519,
+      "step": 21525
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.735189478270007e-05,
+      "loss": 2.8481,
+      "step": 21530
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.73422068150576e-05,
+      "loss": 3.0771,
+      "step": 21535
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.733251738277674e-05,
+      "loss": 2.9961,
+      "step": 21540
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.732282648637652e-05,
+      "loss": 3.1538,
+      "step": 21545
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.731313412637604e-05,
+      "loss": 2.9253,
+      "step": 21550
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.730344030329449e-05,
+      "loss": 3.1388,
+      "step": 21555
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.729374501765116e-05,
+      "loss": 3.0971,
+      "step": 21560
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.728404826996538e-05,
+      "loss": 3.1431,
+      "step": 21565
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.727435006075658e-05,
+      "loss": 3.0277,
+      "step": 21570
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.726465039054426e-05,
+      "loss": 3.0474,
+      "step": 21575
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.7254949259848e-05,
+      "loss": 3.2275,
+      "step": 21580
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.724524666918746e-05,
+      "loss": 3.1425,
+      "step": 21585
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.723554261908239e-05,
+      "loss": 3.0303,
+      "step": 21590
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.722583711005257e-05,
+      "loss": 3.1455,
+      "step": 21595
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.721613014261794e-05,
+      "loss": 3.1149,
+      "step": 21600
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.720642171729846e-05,
+      "loss": 3.0543,
+      "step": 21605
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.719671183461417e-05,
+      "loss": 3.0382,
+      "step": 21610
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.71870004950852e-05,
+      "loss": 2.9691,
+      "step": 21615
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.717728769923178e-05,
+      "loss": 2.9718,
+      "step": 21620
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.716757344757415e-05,
+      "loss": 3.0245,
+      "step": 21625
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.71578577406327e-05,
+      "loss": 3.0355,
+      "step": 21630
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.714814057892788e-05,
+      "loss": 3.0209,
+      "step": 21635
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.71384219629802e-05,
+      "loss": 3.0351,
+      "step": 21640
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.712870189331025e-05,
+      "loss": 3.0009,
+      "step": 21645
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.71189803704387e-05,
+      "loss": 3.0436,
+      "step": 21650
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.71092573948863e-05,
+      "loss": 3.0165,
+      "step": 21655
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.709953296717391e-05,
+      "loss": 3.1244,
+      "step": 21660
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.708980708782243e-05,
+      "loss": 3.0501,
+      "step": 21665
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.708007975735281e-05,
+      "loss": 3.0895,
+      "step": 21670
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.707035097628614e-05,
+      "loss": 3.1098,
+      "step": 21675
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.706062074514356e-05,
+      "loss": 3.0059,
+      "step": 21680
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.705088906444628e-05,
+      "loss": 3.0282,
+      "step": 21685
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.704115593471563e-05,
+      "loss": 3.1193,
+      "step": 21690
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.703142135647293e-05,
+      "loss": 3.0696,
+      "step": 21695
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.702168533023968e-05,
+      "loss": 3.1733,
+      "step": 21700
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.701194785653736e-05,
+      "loss": 3.0425,
+      "step": 21705
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.700220893588763e-05,
+      "loss": 3.1135,
+      "step": 21710
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.699246856881215e-05,
+      "loss": 3.235,
+      "step": 21715
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.698272675583267e-05,
+      "loss": 3.0896,
+      "step": 21720
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.697298349747103e-05,
+      "loss": 3.1526,
+      "step": 21725
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.696323879424916e-05,
+      "loss": 3.0354,
+      "step": 21730
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.695349264668905e-05,
+      "loss": 3.152,
+      "step": 21735
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.694374505531279e-05,
+      "loss": 3.0231,
+      "step": 21740
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.693399602064251e-05,
+      "loss": 3.114,
+      "step": 21745
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.692424554320041e-05,
+      "loss": 3.0553,
+      "step": 21750
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.691449362350884e-05,
+      "loss": 2.9467,
+      "step": 21755
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.690474026209015e-05,
+      "loss": 2.9969,
+      "step": 21760
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.689498545946682e-05,
+      "loss": 3.1397,
+      "step": 21765
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.688522921616136e-05,
+      "loss": 3.0741,
+      "step": 21770
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.687547153269639e-05,
+      "loss": 2.9844,
+      "step": 21775
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.686571240959462e-05,
+      "loss": 2.9814,
+      "step": 21780
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.68559518473788e-05,
+      "loss": 3.1545,
+      "step": 21785
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.684618984657176e-05,
+      "loss": 2.937,
+      "step": 21790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.683642640769643e-05,
+      "loss": 3.0658,
+      "step": 21795
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.682666153127583e-05,
+      "loss": 3.0122,
+      "step": 21800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.6816895217833e-05,
+      "loss": 3.0816,
+      "step": 21805
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.680712746789113e-05,
+      "loss": 3.0343,
+      "step": 21810
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.67973582819734e-05,
+      "loss": 2.9716,
+      "step": 21815
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.678758766060317e-05,
+      "loss": 2.9775,
+      "step": 21820
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.677781560430377e-05,
+      "loss": 2.9512,
+      "step": 21825
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.676804211359871e-05,
+      "loss": 3.0051,
+      "step": 21830
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.675826718901146e-05,
+      "loss": 3.1647,
+      "step": 21835
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.67484908310657e-05,
+      "loss": 3.0259,
+      "step": 21840
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.67387130402851e-05,
+      "loss": 3.0608,
+      "step": 21845
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.67289338171934e-05,
+      "loss": 3.084,
+      "step": 21850
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.671915316231446e-05,
+      "loss": 3.0407,
+      "step": 21855
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.67093710761722e-05,
+      "loss": 3.0521,
+      "step": 21860
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.669958755929063e-05,
+      "loss": 2.9456,
+      "step": 21865
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.668980261219381e-05,
+      "loss": 3.0994,
+      "step": 21870
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.668001623540589e-05,
+      "loss": 3.1583,
+      "step": 21875
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.667022842945109e-05,
+      "loss": 3.0719,
+      "step": 21880
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.66604391948537e-05,
+      "loss": 3.1724,
+      "step": 21885
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.665064853213814e-05,
+      "loss": 2.9569,
+      "step": 21890
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.664085644182886e-05,
+      "loss": 3.1132,
+      "step": 21895
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.663106292445036e-05,
+      "loss": 3.1161,
+      "step": 21900
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.662126798052727e-05,
+      "loss": 3.0063,
+      "step": 21905
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.661147161058426e-05,
+      "loss": 3.1588,
+      "step": 21910
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.660167381514612e-05,
+      "loss": 2.986,
+      "step": 21915
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.659187459473767e-05,
+      "loss": 2.9995,
+      "step": 21920
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.658207394988381e-05,
+      "loss": 3.0293,
+      "step": 21925
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.657227188110954e-05,
+      "loss": 3.107,
+      "step": 21930
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.656246838893994e-05,
+      "loss": 2.9003,
+      "step": 21935
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.655266347390015e-05,
+      "loss": 3.1141,
+      "step": 21940
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.654285713651538e-05,
+      "loss": 3.1364,
+      "step": 21945
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.653304937731094e-05,
+      "loss": 3.1782,
+      "step": 21950
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.652324019681219e-05,
+      "loss": 3.1515,
+      "step": 21955
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.651342959554457e-05,
+      "loss": 3.1391,
+      "step": 21960
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.65036175740336e-05,
+      "loss": 3.2109,
+      "step": 21965
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.649380413280493e-05,
+      "loss": 3.058,
+      "step": 21970
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.648398927238418e-05,
+      "loss": 3.0907,
+      "step": 21975
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.647417299329711e-05,
+      "loss": 2.9524,
+      "step": 21980
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.646435529606956e-05,
+      "loss": 3.1233,
+      "step": 21985
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.645453618122744e-05,
+      "loss": 3.1091,
+      "step": 21990
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.644471564929674e-05,
+      "loss": 2.9178,
+      "step": 21995
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.643489370080347e-05,
+      "loss": 3.1735,
+      "step": 22000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.642507033627382e-05,
+      "loss": 3.0891,
+      "step": 22005
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.641524555623396e-05,
+      "loss": 3.0216,
+      "step": 22010
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.640541936121017e-05,
+      "loss": 3.1997,
+      "step": 22015
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.639559175172884e-05,
+      "loss": 3.1342,
+      "step": 22020
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.638576272831637e-05,
+      "loss": 3.1433,
+      "step": 22025
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.637593229149929e-05,
+      "loss": 2.9865,
+      "step": 22030
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.636610044180417e-05,
+      "loss": 2.8752,
+      "step": 22035
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.63562671797577e-05,
+      "loss": 3.109,
+      "step": 22040
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.634643250588662e-05,
+      "loss": 3.0411,
+      "step": 22045
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.633659642071771e-05,
+      "loss": 3.0948,
+      "step": 22050
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.632675892477788e-05,
+      "loss": 3.14,
+      "step": 22055
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.631692001859407e-05,
+      "loss": 3.086,
+      "step": 22060
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.630707970269337e-05,
+      "loss": 3.1016,
+      "step": 22065
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.629723797760285e-05,
+      "loss": 3.1364,
+      "step": 22070
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.628739484384972e-05,
+      "loss": 2.9407,
+      "step": 22075
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.627755030196124e-05,
+      "loss": 3.0352,
+      "step": 22080
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.626770435246475e-05,
+      "loss": 3.0089,
+      "step": 22085
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.625785699588769e-05,
+      "loss": 3.1887,
+      "step": 22090
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.624800823275753e-05,
+      "loss": 3.0278,
+      "step": 22095
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.623815806360184e-05,
+      "loss": 3.1152,
+      "step": 22100
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.622830648894828e-05,
+      "loss": 3.1013,
+      "step": 22105
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.621845350932453e-05,
+      "loss": 3.0025,
+      "step": 22110
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.620859912525842e-05,
+      "loss": 3.0733,
+      "step": 22115
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.619874333727782e-05,
+      "loss": 2.936,
+      "step": 22120
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.618888614591066e-05,
+      "loss": 3.2152,
+      "step": 22125
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.617902755168496e-05,
+      "loss": 3.1023,
+      "step": 22130
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.616916755512882e-05,
+      "loss": 3.1617,
+      "step": 22135
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.61593061567704e-05,
+      "loss": 3.0767,
+      "step": 22140
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.614944335713797e-05,
+      "loss": 3.1101,
+      "step": 22145
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.613957915675982e-05,
+      "loss": 3.1514,
+      "step": 22150
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.612971355616437e-05,
+      "loss": 3.1018,
+      "step": 22155
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.611984655588006e-05,
+      "loss": 3.0577,
+      "step": 22160
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.610997815643546e-05,
+      "loss": 3.0594,
+      "step": 22165
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.610010835835918e-05,
+      "loss": 3.0784,
+      "step": 22170
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.609023716217993e-05,
+      "loss": 3.0208,
+      "step": 22175
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.608036456842646e-05,
+      "loss": 3.0175,
+      "step": 22180
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.607049057762763e-05,
+      "loss": 3.06,
+      "step": 22185
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.606061519031233e-05,
+      "loss": 3.141,
+      "step": 22190
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.60507384070096e-05,
+      "loss": 3.0034,
+      "step": 22195
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.60408602282485e-05,
+      "loss": 2.9506,
+      "step": 22200
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.603098065455812e-05,
+      "loss": 3.1615,
+      "step": 22205
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.602109968646774e-05,
+      "loss": 3.0367,
+      "step": 22210
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.601121732450662e-05,
+      "loss": 3.0341,
+      "step": 22215
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.600133356920414e-05,
+      "loss": 3.0282,
+      "step": 22220
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.599144842108975e-05,
+      "loss": 3.0115,
+      "step": 22225
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.598156188069294e-05,
+      "loss": 3.1561,
+      "step": 22230
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.597167394854334e-05,
+      "loss": 3.1559,
+      "step": 22235
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.596178462517058e-05,
+      "loss": 2.9776,
+      "step": 22240
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.595189391110443e-05,
+      "loss": 3.0575,
+      "step": 22245
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.594200180687469e-05,
+      "loss": 3.0114,
+      "step": 22250
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.593210831301125e-05,
+      "loss": 3.1583,
+      "step": 22255
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.592221343004409e-05,
+      "loss": 2.9444,
+      "step": 22260
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.591231715850322e-05,
+      "loss": 3.1364,
+      "step": 22265
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.590241949891876e-05,
+      "loss": 2.9656,
+      "step": 22270
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.589252045182093e-05,
+      "loss": 3.0161,
+      "step": 22275
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.588262001773995e-05,
+      "loss": 3.0002,
+      "step": 22280
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.587271819720618e-05,
+      "loss": 3.1396,
+      "step": 22285
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.586281499075e-05,
+      "loss": 3.1026,
+      "step": 22290
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.585291039890195e-05,
+      "loss": 3.012,
+      "step": 22295
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.584300442219255e-05,
+      "loss": 2.9474,
+      "step": 22300
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.583309706115244e-05,
+      "loss": 3.1681,
+      "step": 22305
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.582318831631233e-05,
+      "loss": 2.9352,
+      "step": 22310
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.581327818820297e-05,
+      "loss": 3.0287,
+      "step": 22315
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.580336667735526e-05,
+      "loss": 2.993,
+      "step": 22320
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.579345378430013e-05,
+      "loss": 3.0129,
+      "step": 22325
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.578353950956853e-05,
+      "loss": 3.0829,
+      "step": 22330
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.57736238536916e-05,
+      "loss": 3.0631,
+      "step": 22335
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.576370681720046e-05,
+      "loss": 2.9162,
+      "step": 22340
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.575378840062634e-05,
+      "loss": 3.051,
+      "step": 22345
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.574386860450054e-05,
+      "loss": 3.0594,
+      "step": 22350
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.573394742935443e-05,
+      "loss": 3.0436,
+      "step": 22355
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.572402487571944e-05,
+      "loss": 3.0153,
+      "step": 22360
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.571410094412714e-05,
+      "loss": 3.0606,
+      "step": 22365
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.570417563510908e-05,
+      "loss": 2.9391,
+      "step": 22370
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.569424894919695e-05,
+      "loss": 3.0836,
+      "step": 22375
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.568432088692246e-05,
+      "loss": 3.1724,
+      "step": 22380
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.567439144881747e-05,
+      "loss": 3.0025,
+      "step": 22385
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.566446063541386e-05,
+      "loss": 3.0595,
+      "step": 22390
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.565452844724357e-05,
+      "loss": 3.0373,
+      "step": 22395
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.564459488483866e-05,
+      "loss": 3.1622,
+      "step": 22400
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.563465994873123e-05,
+      "loss": 3.0902,
+      "step": 22405
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.562472363945346e-05,
+      "loss": 3.0624,
+      "step": 22410
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.561478595753762e-05,
+      "loss": 3.1155,
+      "step": 22415
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.560484690351603e-05,
+      "loss": 3.1271,
+      "step": 22420
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.55949064779211e-05,
+      "loss": 3.1007,
+      "step": 22425
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.558496468128529e-05,
+      "loss": 2.988,
+      "step": 22430
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.55750215141412e-05,
+      "loss": 2.9374,
+      "step": 22435
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.556507697702142e-05,
+      "loss": 3.011,
+      "step": 22440
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.555513107045864e-05,
+      "loss": 2.9782,
+      "step": 22445
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.554518379498564e-05,
+      "loss": 2.9222,
+      "step": 22450
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.553523515113527e-05,
+      "loss": 2.984,
+      "step": 22455
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.552528513944047e-05,
+      "loss": 2.9624,
+      "step": 22460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.551533376043418e-05,
+      "loss": 3.146,
+      "step": 22465
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.550538101464952e-05,
+      "loss": 3.1369,
+      "step": 22470
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.549542690261959e-05,
+      "loss": 3.0388,
+      "step": 22475
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.548547142487762e-05,
+      "loss": 3.044,
+      "step": 22480
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.54755145819569e-05,
+      "loss": 3.1047,
+      "step": 22485
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.546555637439074e-05,
+      "loss": 2.9294,
+      "step": 22490
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.545559680271264e-05,
+      "loss": 3.0516,
+      "step": 22495
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.544563586745605e-05,
+      "loss": 3.1602,
+      "step": 22500
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.543567356915459e-05,
+      "loss": 2.9412,
+      "step": 22505
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.542570990834187e-05,
+      "loss": 3.1266,
+      "step": 22510
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.541574488555165e-05,
+      "loss": 3.0685,
+      "step": 22515
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.540577850131768e-05,
+      "loss": 2.964,
+      "step": 22520
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.539581075617387e-05,
+      "loss": 3.0551,
+      "step": 22525
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.538584165065414e-05,
+      "loss": 3.0109,
+      "step": 22530
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.537587118529253e-05,
+      "loss": 3.0251,
+      "step": 22535
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.536589936062308e-05,
+      "loss": 3.1421,
+      "step": 22540
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.535592617717999e-05,
+      "loss": 3.0995,
+      "step": 22545
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.534595163549749e-05,
+      "loss": 3.1197,
+      "step": 22550
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.533597573610987e-05,
+      "loss": 3.0596,
+      "step": 22555
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.532599847955152e-05,
+      "loss": 3.2466,
+      "step": 22560
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.531601986635687e-05,
+      "loss": 3.1844,
+      "step": 22565
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.530603989706048e-05,
+      "loss": 2.8997,
+      "step": 22570
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.529605857219692e-05,
+      "loss": 2.9624,
+      "step": 22575
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.528607589230086e-05,
+      "loss": 3.0523,
+      "step": 22580
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.527609185790707e-05,
+      "loss": 2.938,
+      "step": 22585
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.52661064695503e-05,
+      "loss": 3.0534,
+      "step": 22590
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.525611972776552e-05,
+      "loss": 3.1028,
+      "step": 22595
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.524613163308763e-05,
+      "loss": 3.1242,
+      "step": 22600
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.523614218605168e-05,
+      "loss": 3.0878,
+      "step": 22605
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.522615138719278e-05,
+      "loss": 3.1542,
+      "step": 22610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.521615923704608e-05,
+      "loss": 3.081,
+      "step": 22615
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.520616573614686e-05,
+      "loss": 2.9216,
+      "step": 22620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.519617088503044e-05,
+      "loss": 3.1284,
+      "step": 22625
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.518617468423219e-05,
+      "loss": 3.0748,
+      "step": 22630
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.51761771342876e-05,
+      "loss": 2.9688,
+      "step": 22635
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.51661782357322e-05,
+      "loss": 3.1289,
+      "step": 22640
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.515617798910159e-05,
+      "loss": 3.0937,
+      "step": 22645
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.514617639493146e-05,
+      "loss": 3.1497,
+      "step": 22650
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.513617345375758e-05,
+      "loss": 3.0521,
+      "step": 22655
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.512616916611574e-05,
+      "loss": 3.0249,
+      "step": 22660
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.511616353254187e-05,
+      "loss": 3.1298,
+      "step": 22665
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.510615655357194e-05,
+      "loss": 3.0236,
+      "step": 22670
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.509614822974199e-05,
+      "loss": 2.9645,
+      "step": 22675
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.508613856158813e-05,
+      "loss": 3.1173,
+      "step": 22680
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.507612754964654e-05,
+      "loss": 3.0642,
+      "step": 22685
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.506611519445348e-05,
+      "loss": 3.0522,
+      "step": 22690
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.505610149654531e-05,
+      "loss": 3.2242,
+      "step": 22695
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.504608645645839e-05,
+      "loss": 3.1268,
+      "step": 22700
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.503607007472924e-05,
+      "loss": 3.0264,
+      "step": 22705
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.502605235189437e-05,
+      "loss": 3.0079,
+      "step": 22710
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.501603328849043e-05,
+      "loss": 3.008,
+      "step": 22715
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.500601288505407e-05,
+      "loss": 2.8861,
+      "step": 22720
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.499599114212209e-05,
+      "loss": 3.0102,
+      "step": 22725
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.498596806023132e-05,
+      "loss": 3.0839,
+      "step": 22730
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.497594363991864e-05,
+      "loss": 3.0121,
+      "step": 22735
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.496591788172105e-05,
+      "loss": 3.0704,
+      "step": 22740
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.495589078617559e-05,
+      "loss": 3.1231,
+      "step": 22745
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.494586235381938e-05,
+      "loss": 3.0276,
+      "step": 22750
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.493583258518963e-05,
+      "loss": 3.0074,
+      "step": 22755
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.492580148082358e-05,
+      "loss": 3.0264,
+      "step": 22760
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.491576904125858e-05,
+      "loss": 3.0867,
+      "step": 22765
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.490573526703204e-05,
+      "loss": 3.0429,
+      "step": 22770
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.489570015868142e-05,
+      "loss": 3.1383,
+      "step": 22775
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.488566371674428e-05,
+      "loss": 3.0813,
+      "step": 22780
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.487562594175825e-05,
+      "loss": 3.0973,
+      "step": 22785
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.486558683426102e-05,
+      "loss": 3.0106,
+      "step": 22790
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.485554639479033e-05,
+      "loss": 3.1046,
+      "step": 22795
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.484550462388405e-05,
+      "loss": 2.9522,
+      "step": 22800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.483546152208007e-05,
+      "loss": 3.0949,
+      "step": 22805
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.48254170899164e-05,
+      "loss": 3.1365,
+      "step": 22810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.481537132793103e-05,
+      "loss": 2.9106,
+      "step": 22815
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.480532423666211e-05,
+      "loss": 3.0271,
+      "step": 22820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.479527581664785e-05,
+      "loss": 3.1021,
+      "step": 22825
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.478522606842648e-05,
+      "loss": 3.1209,
+      "step": 22830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.477517499253637e-05,
+      "loss": 2.9705,
+      "step": 22835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.476512258951588e-05,
+      "loss": 3.1081,
+      "step": 22840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.475506885990354e-05,
+      "loss": 3.1301,
+      "step": 22845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.474501380423785e-05,
+      "loss": 3.0075,
+      "step": 22850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.473495742305745e-05,
+      "loss": 3.0396,
+      "step": 22855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.472489971690104e-05,
+      "loss": 3.1143,
+      "step": 22860
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.471484068630737e-05,
+      "loss": 3.1411,
+      "step": 22865
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.470478033181526e-05,
+      "loss": 2.9975,
+      "step": 22870
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.469471865396362e-05,
+      "loss": 2.8894,
+      "step": 22875
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.468465565329143e-05,
+      "loss": 3.0605,
+      "step": 22880
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.467459133033775e-05,
+      "loss": 3.0209,
+      "step": 22885
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.466452568564167e-05,
+      "loss": 3.0968,
+      "step": 22890
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.465445871974236e-05,
+      "loss": 2.9949,
+      "step": 22895
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.46443904331791e-05,
+      "loss": 3.0324,
+      "step": 22900
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.463432082649122e-05,
+      "loss": 3.1493,
+      "step": 22905
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.462424990021812e-05,
+      "loss": 3.1035,
+      "step": 22910
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.461417765489927e-05,
+      "loss": 3.1406,
+      "step": 22915
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.460410409107419e-05,
+      "loss": 3.062,
+      "step": 22920
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.459402920928249e-05,
+      "loss": 3.1531,
+      "step": 22925
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.458395301006386e-05,
+      "loss": 3.0004,
+      "step": 22930
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.457387549395808e-05,
+      "loss": 3.048,
+      "step": 22935
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.456379666150494e-05,
+      "loss": 3.1117,
+      "step": 22940
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.455371651324431e-05,
+      "loss": 3.0988,
+      "step": 22945
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.45436350497162e-05,
+      "loss": 3.0603,
+      "step": 22950
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.453355227146063e-05,
+      "loss": 3.0578,
+      "step": 22955
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.452346817901768e-05,
+      "loss": 3.1179,
+      "step": 22960
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.451338277292755e-05,
+      "loss": 3.1632,
+      "step": 22965
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.450329605373048e-05,
+      "loss": 3.1512,
+      "step": 22970
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.449320802196676e-05,
+      "loss": 2.9085,
+      "step": 22975
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.448311867817681e-05,
+      "loss": 3.2289,
+      "step": 22980
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.447302802290107e-05,
+      "loss": 2.9549,
+      "step": 22985
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.446293605668006e-05,
+      "loss": 3.057,
+      "step": 22990
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.445284278005437e-05,
+      "loss": 3.0636,
+      "step": 22995
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.44427481935647e-05,
+      "loss": 3.1609,
+      "step": 23000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.443265229775175e-05,
+      "loss": 3.1325,
+      "step": 23005
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.442255509315635e-05,
+      "loss": 3.0965,
+      "step": 23010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.441245658031936e-05,
+      "loss": 3.1376,
+      "step": 23015
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.440235675978172e-05,
+      "loss": 3.0483,
+      "step": 23020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.439225563208447e-05,
+      "loss": 3.0545,
+      "step": 23025
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.438215319776867e-05,
+      "loss": 2.9146,
+      "step": 23030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.437204945737552e-05,
+      "loss": 3.0508,
+      "step": 23035
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.436194441144621e-05,
+      "loss": 3.1136,
+      "step": 23040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.435183806052203e-05,
+      "loss": 2.9361,
+      "step": 23045
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.434173040514438e-05,
+      "loss": 3.2292,
+      "step": 23050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.433162144585466e-05,
+      "loss": 3.0967,
+      "step": 23055
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.432151118319439e-05,
+      "loss": 2.9326,
+      "step": 23060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.431139961770515e-05,
+      "loss": 3.0712,
+      "step": 23065
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.43012867499286e-05,
+      "loss": 3.0723,
+      "step": 23070
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.429117258040644e-05,
+      "loss": 3.0669,
+      "step": 23075
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.428105710968046e-05,
+      "loss": 3.2118,
+      "step": 23080
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.427094033829248e-05,
+      "loss": 3.0292,
+      "step": 23085
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.426082226678448e-05,
+      "loss": 3.0528,
+      "step": 23090
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.425070289569842e-05,
+      "loss": 3.0305,
+      "step": 23095
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.424058222557637e-05,
+      "loss": 3.1205,
+      "step": 23100
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.423046025696046e-05,
+      "loss": 3.0137,
+      "step": 23105
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.422033699039292e-05,
+      "loss": 3.0711,
+      "step": 23110
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.4210212426416e-05,
+      "loss": 3.0023,
+      "step": 23115
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.420008656557203e-05,
+      "loss": 2.946,
+      "step": 23120
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.418995940840343e-05,
+      "loss": 3.0809,
+      "step": 23125
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.417983095545269e-05,
+      "loss": 2.9304,
+      "step": 23130
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.416970120726237e-05,
+      "loss": 3.0555,
+      "step": 23135
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.415957016437506e-05,
+      "loss": 2.9485,
+      "step": 23140
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.414943782733347e-05,
+      "loss": 3.1747,
+      "step": 23145
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.413930419668035e-05,
+      "loss": 3.1964,
+      "step": 23150
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.412916927295852e-05,
+      "loss": 3.1008,
+      "step": 23155
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.411903305671089e-05,
+      "loss": 3.1554,
+      "step": 23160
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.410889554848045e-05,
+      "loss": 3.0491,
+      "step": 23165
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.409875674881019e-05,
+      "loss": 3.0589,
+      "step": 23170
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.408861665824323e-05,
+      "loss": 2.9356,
+      "step": 23175
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.407847527732275e-05,
+      "loss": 2.9924,
+      "step": 23180
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.406833260659198e-05,
+      "loss": 3.0358,
+      "step": 23185
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.405818864659425e-05,
+      "loss": 3.1457,
+      "step": 23190
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.404804339787293e-05,
+      "loss": 3.0177,
+      "step": 23195
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.403789686097147e-05,
+      "loss": 3.1249,
+      "step": 23200
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.40277490364334e-05,
+      "loss": 3.0379,
+      "step": 23205
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.401759992480228e-05,
+      "loss": 3.1415,
+      "step": 23210
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.40074495266218e-05,
+      "loss": 3.1837,
+      "step": 23215
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.399729784243568e-05,
+      "loss": 3.0938,
+      "step": 23220
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.398714487278768e-05,
+      "loss": 3.1155,
+      "step": 23225
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.397699061822171e-05,
+      "loss": 3.1464,
+      "step": 23230
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.396683507928169e-05,
+      "loss": 2.9267,
+      "step": 23235
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.39566782565116e-05,
+      "loss": 2.9494,
+      "step": 23240
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.394652015045552e-05,
+      "loss": 3.059,
+      "step": 23245
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.39363607616576e-05,
+      "loss": 3.0291,
+      "step": 23250
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.392620009066203e-05,
+      "loss": 2.9613,
+      "step": 23255
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.39160381380131e-05,
+      "loss": 3.0974,
+      "step": 23260
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.390587490425515e-05,
+      "loss": 2.9782,
+      "step": 23265
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.38957103899326e-05,
+      "loss": 2.9904,
+      "step": 23270
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.388554459558991e-05,
+      "loss": 3.0842,
+      "step": 23275
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.387537752177165e-05,
+      "loss": 3.1981,
+      "step": 23280
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.386520916902245e-05,
+      "loss": 3.0704,
+      "step": 23285
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.385503953788697e-05,
+      "loss": 3.0252,
+      "step": 23290
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.384486862890998e-05,
+      "loss": 3.1348,
+      "step": 23295
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.383469644263629e-05,
+      "loss": 2.9717,
+      "step": 23300
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.382452297961082e-05,
+      "loss": 2.9199,
+      "step": 23305
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.381434824037852e-05,
+      "loss": 3.0242,
+      "step": 23310
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.380417222548442e-05,
+      "loss": 3.0843,
+      "step": 23315
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.379399493547361e-05,
+      "loss": 3.1666,
+      "step": 23320
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.378381637089125e-05,
+      "loss": 2.9285,
+      "step": 23325
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.37736365322826e-05,
+      "loss": 3.1377,
+      "step": 23330
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.376345542019294e-05,
+      "loss": 3.0293,
+      "step": 23335
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.375327303516766e-05,
+      "loss": 3.1242,
+      "step": 23340
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.374308937775219e-05,
+      "loss": 2.9561,
+      "step": 23345
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.373290444849202e-05,
+      "loss": 2.9246,
+      "step": 23350
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.372271824793274e-05,
+      "loss": 3.0961,
+      "step": 23355
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.371253077662e-05,
+      "loss": 3.0761,
+      "step": 23360
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.370234203509951e-05,
+      "loss": 2.9955,
+      "step": 23365
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.369215202391705e-05,
+      "loss": 3.1243,
+      "step": 23370
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.368196074361847e-05,
+      "loss": 2.895,
+      "step": 23375
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.367176819474967e-05,
+      "loss": 3.1948,
+      "step": 23380
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.366157437785665e-05,
+      "loss": 3.0011,
+      "step": 23385
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.365137929348544e-05,
+      "loss": 3.0552,
+      "step": 23390
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.364118294218219e-05,
+      "loss": 3.0089,
+      "step": 23395
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.363098532449306e-05,
+      "loss": 3.0685,
+      "step": 23400
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.362078644096433e-05,
+      "loss": 3.0731,
+      "step": 23405
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.36105862921423e-05,
+      "loss": 3.0457,
+      "step": 23410
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.360038487857339e-05,
+      "loss": 2.9771,
+      "step": 23415
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.359018220080402e-05,
+      "loss": 3.0595,
+      "step": 23420
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.357997825938075e-05,
+      "loss": 3.0288,
+      "step": 23425
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.356977305485015e-05,
+      "loss": 3.1527,
+      "step": 23430
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.355956658775889e-05,
+      "loss": 2.9455,
+      "step": 23435
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.354935885865373e-05,
+      "loss": 2.9786,
+      "step": 23440
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.353914986808141e-05,
+      "loss": 3.1059,
+      "step": 23445
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.352893961658882e-05,
+      "loss": 3.1017,
+      "step": 23450
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.35187281047229e-05,
+      "loss": 3.1399,
+      "step": 23455
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.350851533303063e-05,
+      "loss": 3.0695,
+      "step": 23460
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.349830130205911e-05,
+      "loss": 2.8889,
+      "step": 23465
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.348808601235547e-05,
+      "loss": 3.1296,
+      "step": 23470
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.347786946446687e-05,
+      "loss": 2.9101,
+      "step": 23475
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.34676516589406e-05,
+      "loss": 3.0287,
+      "step": 23480
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.345743259632401e-05,
+      "loss": 3.0513,
+      "step": 23485
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.344721227716449e-05,
+      "loss": 2.9714,
+      "step": 23490
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.343699070200953e-05,
+      "loss": 3.0821,
+      "step": 23495
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.342676787140666e-05,
+      "loss": 3.1051,
+      "step": 23500
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.341654378590348e-05,
+      "loss": 3.0458,
+      "step": 23505
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.340631844604766e-05,
+      "loss": 3.1133,
+      "step": 23510
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.339609185238694e-05,
+      "loss": 3.0225,
+      "step": 23515
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.338586400546914e-05,
+      "loss": 3.023,
+      "step": 23520
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.337563490584212e-05,
+      "loss": 2.9926,
+      "step": 23525
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.336540455405383e-05,
+      "loss": 2.903,
+      "step": 23530
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.335517295065229e-05,
+      "loss": 3.0073,
+      "step": 23535
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.334494009618553e-05,
+      "loss": 3.0283,
+      "step": 23540
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.333470599120176e-05,
+      "loss": 2.9094,
+      "step": 23545
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.332447063624915e-05,
+      "loss": 3.067,
+      "step": 23550
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.331423403187597e-05,
+      "loss": 3.0412,
+      "step": 23555
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.330399617863058e-05,
+      "loss": 2.9455,
+      "step": 23560
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.329375707706139e-05,
+      "loss": 3.0342,
+      "step": 23565
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.328351672771687e-05,
+      "loss": 3.064,
+      "step": 23570
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.327327513114557e-05,
+      "loss": 3.06,
+      "step": 23575
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.326303228789608e-05,
+      "loss": 3.0749,
+      "step": 23580
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.325278819851712e-05,
+      "loss": 3.0996,
+      "step": 23585
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.32425428635574e-05,
+      "loss": 3.0512,
+      "step": 23590
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.323229628356575e-05,
+      "loss": 3.0738,
+      "step": 23595
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.322204845909103e-05,
+      "loss": 3.1184,
+      "step": 23600
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.32117993906822e-05,
+      "loss": 3.0553,
+      "step": 23605
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.320154907888827e-05,
+      "loss": 3.199,
+      "step": 23610
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.319129752425831e-05,
+      "loss": 3.0668,
+      "step": 23615
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.318104472734146e-05,
+      "loss": 3.0902,
+      "step": 23620
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.317079068868695e-05,
+      "loss": 3.1503,
+      "step": 23625
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.316053540884405e-05,
+      "loss": 3.1759,
+      "step": 23630
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.315027888836207e-05,
+      "loss": 3.0852,
+      "step": 23635
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.314002112779048e-05,
+      "loss": 3.1908,
+      "step": 23640
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.312976212767874e-05,
+      "loss": 3.1501,
+      "step": 23645
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.311950188857636e-05,
+      "loss": 2.9645,
+      "step": 23650
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.310924041103298e-05,
+      "loss": 3.092,
+      "step": 23655
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.309897769559825e-05,
+      "loss": 3.1737,
+      "step": 23660
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.308871374282194e-05,
+      "loss": 3.0601,
+      "step": 23665
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.307844855325385e-05,
+      "loss": 3.0161,
+      "step": 23670
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.306818212744384e-05,
+      "loss": 3.2185,
+      "step": 23675
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.305791446594188e-05,
+      "loss": 3.0037,
+      "step": 23680
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.304764556929796e-05,
+      "loss": 3.1417,
+      "step": 23685
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.303737543806215e-05,
+      "loss": 3.0896,
+      "step": 23690
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.302710407278459e-05,
+      "loss": 3.002,
+      "step": 23695
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.301683147401549e-05,
+      "loss": 3.055,
+      "step": 23700
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.300655764230512e-05,
+      "loss": 3.1268,
+      "step": 23705
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.299628257820381e-05,
+      "loss": 3.06,
+      "step": 23710
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.298600628226199e-05,
+      "loss": 3.2017,
+      "step": 23715
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.297572875503012e-05,
+      "loss": 2.942,
+      "step": 23720
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.29654499970587e-05,
+      "loss": 3.0472,
+      "step": 23725
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.295517000889838e-05,
+      "loss": 3.16,
+      "step": 23730
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.29448887910998e-05,
+      "loss": 3.111,
+      "step": 23735
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.293460634421369e-05,
+      "loss": 3.1001,
+      "step": 23740
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.292432266879088e-05,
+      "loss": 3.0844,
+      "step": 23745
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.29140377653822e-05,
+      "loss": 2.9951,
+      "step": 23750
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.29037516345386e-05,
+      "loss": 3.0623,
+      "step": 23755
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.289346427681107e-05,
+      "loss": 3.1078,
+      "step": 23760
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.288317569275069e-05,
+      "loss": 2.9356,
+      "step": 23765
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.287288588290856e-05,
+      "loss": 3.111,
+      "step": 23770
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.286259484783589e-05,
+      "loss": 3.1596,
+      "step": 23775
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.285230258808392e-05,
+      "loss": 3.1264,
+      "step": 23780
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.2842009104204e-05,
+      "loss": 2.8929,
+      "step": 23785
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.283171439674751e-05,
+      "loss": 2.9161,
+      "step": 23790
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.282141846626591e-05,
+      "loss": 2.9618,
+      "step": 23795
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.281112131331072e-05,
+      "loss": 3.1332,
+      "step": 23800
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.280082293843352e-05,
+      "loss": 3.2461,
+      "step": 23805
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.279052334218597e-05,
+      "loss": 3.0883,
+      "step": 23810
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.278022252511978e-05,
+      "loss": 3.0616,
+      "step": 23815
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.276992048778675e-05,
+      "loss": 2.9857,
+      "step": 23820
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.275961723073871e-05,
+      "loss": 3.0952,
+      "step": 23825
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.274931275452758e-05,
+      "loss": 3.0061,
+      "step": 23830
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.273900705970534e-05,
+      "loss": 3.0129,
+      "step": 23835
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.272870014682403e-05,
+      "loss": 2.9413,
+      "step": 23840
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.271839201643578e-05,
+      "loss": 3.0342,
+      "step": 23845
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.270808266909274e-05,
+      "loss": 3.2564,
+      "step": 23850
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.269777210534716e-05,
+      "loss": 2.9613,
+      "step": 23855
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.268746032575136e-05,
+      "loss": 3.0374,
+      "step": 23860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.267714733085767e-05,
+      "loss": 2.9488,
+      "step": 23865
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.266683312121856e-05,
+      "loss": 2.8877,
+      "step": 23870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.265651769738655e-05,
+      "loss": 3.062,
+      "step": 23875
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.264620105991416e-05,
+      "loss": 2.9779,
+      "step": 23880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.263588320935401e-05,
+      "loss": 3.0609,
+      "step": 23885
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.262556414625886e-05,
+      "loss": 3.0966,
+      "step": 23890
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.261524387118141e-05,
+      "loss": 3.0874,
+      "step": 23895
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.260492238467455e-05,
+      "loss": 3.0136,
+      "step": 23900
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.25945996872911e-05,
+      "loss": 3.0093,
+      "step": 23905
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.258427577958405e-05,
+      "loss": 2.8734,
+      "step": 23910
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.257395066210642e-05,
+      "loss": 2.9582,
+      "step": 23915
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.256362433541128e-05,
+      "loss": 3.0823,
+      "step": 23920
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.255329680005182e-05,
+      "loss": 2.9957,
+      "step": 23925
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.254296805658121e-05,
+      "loss": 2.9702,
+      "step": 23930
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.253263810555273e-05,
+      "loss": 3.0805,
+      "step": 23935
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.252230694751975e-05,
+      "loss": 3.1849,
+      "step": 23940
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.251197458303566e-05,
+      "loss": 3.1227,
+      "step": 23945
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.250164101265397e-05,
+      "loss": 3.0702,
+      "step": 23950
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.249130623692816e-05,
+      "loss": 2.8992,
+      "step": 23955
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.248097025641187e-05,
+      "loss": 3.1129,
+      "step": 23960
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.247063307165875e-05,
+      "loss": 3.0124,
+      "step": 23965
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.246029468322255e-05,
+      "loss": 2.8568,
+      "step": 23970
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.244995509165705e-05,
+      "loss": 3.0897,
+      "step": 23975
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.24396142975161e-05,
+      "loss": 3.0038,
+      "step": 23980
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.242927230135367e-05,
+      "loss": 3.0962,
+      "step": 23985
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.24189291037237e-05,
+      "loss": 3.0014,
+      "step": 23990
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.240858470518026e-05,
+      "loss": 3.0285,
+      "step": 23995
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.239823910627748e-05,
+      "loss": 3.0783,
+      "step": 24000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.238789230756953e-05,
+      "loss": 3.0228,
+      "step": 24005
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.237754430961065e-05,
+      "loss": 3.1316,
+      "step": 24010
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.236719511295515e-05,
+      "loss": 3.1316,
+      "step": 24015
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.235684471815744e-05,
+      "loss": 3.155,
+      "step": 24020
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.234649312577193e-05,
+      "loss": 3.1123,
+      "step": 24025
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.233614033635311e-05,
+      "loss": 2.8641,
+      "step": 24030
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.232578635045557e-05,
+      "loss": 3.0872,
+      "step": 24035
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.231543116863395e-05,
+      "loss": 3.0303,
+      "step": 24040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.23050747914429e-05,
+      "loss": 3.0468,
+      "step": 24045
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.229471721943724e-05,
+      "loss": 3.0496,
+      "step": 24050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.228435845317174e-05,
+      "loss": 3.0692,
+      "step": 24055
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.227399849320132e-05,
+      "loss": 3.1361,
+      "step": 24060
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.226363734008092e-05,
+      "loss": 2.9798,
+      "step": 24065
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.225327499436556e-05,
+      "loss": 2.9172,
+      "step": 24070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.224291145661033e-05,
+      "loss": 3.2584,
+      "step": 24075
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.223254672737035e-05,
+      "loss": 3.1387,
+      "step": 24080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.222218080720082e-05,
+      "loss": 3.041,
+      "step": 24085
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.221181369665704e-05,
+      "loss": 3.1084,
+      "step": 24090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.220144539629432e-05,
+      "loss": 3.0944,
+      "step": 24095
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.219107590666809e-05,
+      "loss": 3.0275,
+      "step": 24100
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.218070522833378e-05,
+      "loss": 3.0072,
+      "step": 24105
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.217033336184692e-05,
+      "loss": 2.9563,
+      "step": 24110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.21599603077631e-05,
+      "loss": 3.0429,
+      "step": 24115
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.2149586066638e-05,
+      "loss": 2.9644,
+      "step": 24120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.21392106390273e-05,
+      "loss": 3.0494,
+      "step": 24125
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.212883402548681e-05,
+      "loss": 3.0887,
+      "step": 24130
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.211845622657235e-05,
+      "loss": 3.0863,
+      "step": 24135
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.210807724283983e-05,
+      "loss": 2.9791,
+      "step": 24140
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.209769707484525e-05,
+      "loss": 2.9837,
+      "step": 24145
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.208731572314458e-05,
+      "loss": 3.116,
+      "step": 24150
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.207693318829399e-05,
+      "loss": 3.1317,
+      "step": 24155
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.20665494708496e-05,
+      "loss": 3.035,
+      "step": 24160
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.205616457136763e-05,
+      "loss": 3.0595,
+      "step": 24165
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.204577849040438e-05,
+      "loss": 3.009,
+      "step": 24170
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.20353912285162e-05,
+      "loss": 3.0449,
+      "step": 24175
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.202500278625952e-05,
+      "loss": 2.9996,
+      "step": 24180
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.201461316419079e-05,
+      "loss": 2.954,
+      "step": 24185
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.200422236286656e-05,
+      "loss": 3.1238,
+      "step": 24190
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.199383038284343e-05,
+      "loss": 2.894,
+      "step": 24195
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.198343722467807e-05,
+      "loss": 2.9638,
+      "step": 24200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.197304288892723e-05,
+      "loss": 3.0909,
+      "step": 24205
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.196264737614766e-05,
+      "loss": 2.9647,
+      "step": 24210
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.195225068689625e-05,
+      "loss": 3.1809,
+      "step": 24215
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.19418528217299e-05,
+      "loss": 3.1119,
+      "step": 24220
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.19314537812056e-05,
+      "loss": 3.0386,
+      "step": 24225
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.192105356588041e-05,
+      "loss": 2.9725,
+      "step": 24230
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.191065217631142e-05,
+      "loss": 2.9414,
+      "step": 24235
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.190024961305578e-05,
+      "loss": 3.2369,
+      "step": 24240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.188984587667075e-05,
+      "loss": 3.0374,
+      "step": 24245
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.187944096771365e-05,
+      "loss": 3.074,
+      "step": 24250
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.18690348867418e-05,
+      "loss": 3.0811,
+      "step": 24255
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.185862763431262e-05,
+      "loss": 2.9664,
+      "step": 24260
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.184821921098361e-05,
+      "loss": 3.0849,
+      "step": 24265
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.183780961731233e-05,
+      "loss": 3.1122,
+      "step": 24270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.182739885385637e-05,
+      "loss": 2.9884,
+      "step": 24275
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.18169869211734e-05,
+      "loss": 2.9734,
+      "step": 24280
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.180657381982118e-05,
+      "loss": 3.0364,
+      "step": 24285
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.179615955035749e-05,
+      "loss": 2.9856,
+      "step": 24290
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.178574411334018e-05,
+      "loss": 3.1643,
+      "step": 24295
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.177532750932719e-05,
+      "loss": 3.0624,
+      "step": 24300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.17649097388765e-05,
+      "loss": 3.0161,
+      "step": 24305
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.175449080254617e-05,
+      "loss": 2.9671,
+      "step": 24310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.174407070089429e-05,
+      "loss": 2.9659,
+      "step": 24315
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.173364943447904e-05,
+      "loss": 3.1239,
+      "step": 24320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.172322700385867e-05,
+      "loss": 2.9926,
+      "step": 24325
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.171280340959145e-05,
+      "loss": 2.9718,
+      "step": 24330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.170237865223579e-05,
+      "loss": 3.1722,
+      "step": 24335
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.169195273235005e-05,
+      "loss": 3.1922,
+      "step": 24340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.168152565049274e-05,
+      "loss": 3.0611,
+      "step": 24345
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.167109740722242e-05,
+      "loss": 3.1491,
+      "step": 24350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.16606680030977e-05,
+      "loss": 3.0917,
+      "step": 24355
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.165023743867722e-05,
+      "loss": 3.1837,
+      "step": 24360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.163980571451973e-05,
+      "loss": 3.1132,
+      "step": 24365
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.162937283118405e-05,
+      "loss": 3.0279,
+      "step": 24370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.1618938789229e-05,
+      "loss": 2.8517,
+      "step": 24375
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.160850358921354e-05,
+      "loss": 3.192,
+      "step": 24380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.159806723169662e-05,
+      "loss": 3.1783,
+      "step": 24385
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.158762971723728e-05,
+      "loss": 3.0754,
+      "step": 24390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.157719104639464e-05,
+      "loss": 2.8895,
+      "step": 24395
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.156675121972786e-05,
+      "loss": 3.1422,
+      "step": 24400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.15563102377962e-05,
+      "loss": 3.0001,
+      "step": 24405
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.154586810115892e-05,
+      "loss": 3.0521,
+      "step": 24410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.153542481037538e-05,
+      "loss": 2.9761,
+      "step": 24415
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.152498036600499e-05,
+      "loss": 3.0515,
+      "step": 24420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.151453476860721e-05,
+      "loss": 3.0972,
+      "step": 24425
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.150408801874163e-05,
+      "loss": 3.1894,
+      "step": 24430
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.149364011696782e-05,
+      "loss": 3.1157,
+      "step": 24435
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.148319106384543e-05,
+      "loss": 3.0658,
+      "step": 24440
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.14727408599342e-05,
+      "loss": 2.9816,
+      "step": 24445
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.146228950579392e-05,
+      "loss": 3.12,
+      "step": 24450
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.145183700198444e-05,
+      "loss": 3.0398,
+      "step": 24455
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.144138334906564e-05,
+      "loss": 3.0434,
+      "step": 24460
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.143092854759751e-05,
+      "loss": 3.0385,
+      "step": 24465
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.142047259814008e-05,
+      "loss": 2.9564,
+      "step": 24470
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.141001550125345e-05,
+      "loss": 2.9851,
+      "step": 24475
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.139955725749777e-05,
+      "loss": 3.0997,
+      "step": 24480
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.138909786743325e-05,
+      "loss": 3.0664,
+      "step": 24485
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.137863733162017e-05,
+      "loss": 3.0152,
+      "step": 24490
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.136817565061887e-05,
+      "loss": 2.948,
+      "step": 24495
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.135771282498975e-05,
+      "loss": 3.0789,
+      "step": 24500
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.134724885529329e-05,
+      "loss": 2.9825,
+      "step": 24505
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.133678374208998e-05,
+      "loss": 2.9011,
+      "step": 24510
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.132631748594043e-05,
+      "loss": 3.0237,
+      "step": 24515
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.131585008740527e-05,
+      "loss": 2.9619,
+      "step": 24520
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.13053815470452e-05,
+      "loss": 3.0071,
+      "step": 24525
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.1294911865421e-05,
+      "loss": 2.9946,
+      "step": 24530
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.128444104309353e-05,
+      "loss": 3.0534,
+      "step": 24535
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.127396908062363e-05,
+      "loss": 2.9372,
+      "step": 24540
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.126349597857224e-05,
+      "loss": 2.9692,
+      "step": 24545
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.125302173750043e-05,
+      "loss": 3.0011,
+      "step": 24550
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.124254635796925e-05,
+      "loss": 2.9071,
+      "step": 24555
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.123206984053983e-05,
+      "loss": 3.0548,
+      "step": 24560
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.122159218577335e-05,
+      "loss": 3.0405,
+      "step": 24565
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.121111339423108e-05,
+      "loss": 3.0586,
+      "step": 24570
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.120063346647433e-05,
+      "loss": 2.9439,
+      "step": 24575
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.11901524030645e-05,
+      "loss": 3.1908,
+      "step": 24580
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.1179670204563e-05,
+      "loss": 3.1355,
+      "step": 24585
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.116918687153135e-05,
+      "loss": 3.1219,
+      "step": 24590
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.115870240453109e-05,
+      "loss": 3.0333,
+      "step": 24595
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.114821680412385e-05,
+      "loss": 3.0103,
+      "step": 24600
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.113773007087132e-05,
+      "loss": 3.0648,
+      "step": 24605
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.112724220533524e-05,
+      "loss": 3.1979,
+      "step": 24610
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.11167532080774e-05,
+      "loss": 3.1013,
+      "step": 24615
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.110626307965968e-05,
+      "loss": 3.0967,
+      "step": 24620
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.1095771820644e-05,
+      "loss": 2.9702,
+      "step": 24625
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.108527943159233e-05,
+      "loss": 2.9748,
+      "step": 24630
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.107478591306675e-05,
+      "loss": 3.0973,
+      "step": 24635
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.106429126562933e-05,
+      "loss": 3.0171,
+      "step": 24640
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.105379548984224e-05,
+      "loss": 3.0593,
+      "step": 24645
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.104329858626772e-05,
+      "loss": 3.0552,
+      "step": 24650
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.103280055546805e-05,
+      "loss": 3.0506,
+      "step": 24655
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.102230139800559e-05,
+      "loss": 2.965,
+      "step": 24660
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.101180111444272e-05,
+      "loss": 3.0356,
+      "step": 24665
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.100129970534194e-05,
+      "loss": 3.1547,
+      "step": 24670
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.099079717126575e-05,
+      "loss": 3.006,
+      "step": 24675
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.098029351277676e-05,
+      "loss": 3.352,
+      "step": 24680
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.096978873043763e-05,
+      "loss": 3.0079,
+      "step": 24685
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.095928282481103e-05,
+      "loss": 2.9895,
+      "step": 24690
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.094877579645975e-05,
+      "loss": 3.0827,
+      "step": 24695
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.093826764594662e-05,
+      "loss": 3.0657,
+      "step": 24700
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.092775837383452e-05,
+      "loss": 3.0374,
+      "step": 24705
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.091724798068641e-05,
+      "loss": 3.0831,
+      "step": 24710
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.090673646706531e-05,
+      "loss": 2.8989,
+      "step": 24715
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.089622383353427e-05,
+      "loss": 2.9893,
+      "step": 24720
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.088571008065642e-05,
+      "loss": 3.1282,
+      "step": 24725
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.087519520899497e-05,
+      "loss": 3.1009,
+      "step": 24730
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.086467921911315e-05,
+      "loss": 2.917,
+      "step": 24735
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.085416211157427e-05,
+      "loss": 3.0589,
+      "step": 24740
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.084364388694172e-05,
+      "loss": 3.0445,
+      "step": 24745
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.08331245457789e-05,
+      "loss": 3.1302,
+      "step": 24750
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.082260408864931e-05,
+      "loss": 3.0026,
+      "step": 24755
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.081208251611651e-05,
+      "loss": 3.0559,
+      "step": 24760
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.08015598287441e-05,
+      "loss": 3.0506,
+      "step": 24765
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.079103602709574e-05,
+      "loss": 3.1043,
+      "step": 24770
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.078051111173517e-05,
+      "loss": 3.019,
+      "step": 24775
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.076998508322616e-05,
+      "loss": 2.9739,
+      "step": 24780
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.07594579421326e-05,
+      "loss": 3.157,
+      "step": 24785
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.074892968901835e-05,
+      "loss": 2.9806,
+      "step": 24790
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.073840032444739e-05,
+      "loss": 3.0672,
+      "step": 24795
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.072786984898375e-05,
+      "loss": 3.1137,
+      "step": 24800
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.071733826319153e-05,
+      "loss": 2.9871,
+      "step": 24805
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.070680556763484e-05,
+      "loss": 3.0986,
+      "step": 24810
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.069627176287792e-05,
+      "loss": 3.162,
+      "step": 24815
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.068573684948501e-05,
+      "loss": 3.1641,
+      "step": 24820
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.067520082802044e-05,
+      "loss": 3.0504,
+      "step": 24825
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.066466369904859e-05,
+      "loss": 3.0428,
+      "step": 24830
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.06541254631339e-05,
+      "loss": 3.1477,
+      "step": 24835
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.06435861208409e-05,
+      "loss": 3.1232,
+      "step": 24840
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.063304567273411e-05,
+      "loss": 3.0084,
+      "step": 24845
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.062250411937816e-05,
+      "loss": 3.1078,
+      "step": 24850
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.061196146133776e-05,
+      "loss": 3.0989,
+      "step": 24855
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.060141769917758e-05,
+      "loss": 2.8219,
+      "step": 24860
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.05908728334625e-05,
+      "loss": 3.0247,
+      "step": 24865
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.058032686475731e-05,
+      "loss": 2.9465,
+      "step": 24870
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.056977979362696e-05,
+      "loss": 3.1091,
+      "step": 24875
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.055923162063642e-05,
+      "loss": 3.0845,
+      "step": 24880
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.054868234635072e-05,
+      "loss": 3.0729,
+      "step": 24885
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.053813197133494e-05,
+      "loss": 3.0523,
+      "step": 24890
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.052758049615426e-05,
+      "loss": 2.9059,
+      "step": 24895
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.051702792137386e-05,
+      "loss": 2.8846,
+      "step": 24900
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.050647424755904e-05,
+      "loss": 3.0505,
+      "step": 24905
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.049591947527507e-05,
+      "loss": 3.0404,
+      "step": 24910
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.04853636050874e-05,
+      "loss": 3.0899,
+      "step": 24915
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.047480663756145e-05,
+      "loss": 3.1127,
+      "step": 24920
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.046424857326273e-05,
+      "loss": 3.0008,
+      "step": 24925
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.04536894127568e-05,
+      "loss": 3.0592,
+      "step": 24930
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.04431291566093e-05,
+      "loss": 3.0527,
+      "step": 24935
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.043256780538587e-05,
+      "loss": 3.087,
+      "step": 24940
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.042200535965228e-05,
+      "loss": 2.9107,
+      "step": 24945
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.041144181997432e-05,
+      "loss": 3.0984,
+      "step": 24950
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.040087718691785e-05,
+      "loss": 3.0576,
+      "step": 24955
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.039031146104877e-05,
+      "loss": 3.002,
+      "step": 24960
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.037974464293307e-05,
+      "loss": 2.9793,
+      "step": 24965
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.036917673313678e-05,
+      "loss": 3.016,
+      "step": 24970
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.035860773222599e-05,
+      "loss": 3.0177,
+      "step": 24975
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.034803764076684e-05,
+      "loss": 3.1715,
+      "step": 24980
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.033746645932555e-05,
+      "loss": 3.0084,
+      "step": 24985
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.032689418846839e-05,
+      "loss": 3.1574,
+      "step": 24990
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.031632082876166e-05,
+      "loss": 3.184,
+      "step": 24995
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.030574638077176e-05,
+      "loss": 3.0206,
+      "step": 25000
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.029517084506514e-05,
+      "loss": 3.0326,
+      "step": 25005
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.028459422220827e-05,
+      "loss": 3.0357,
+      "step": 25010
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.027401651276774e-05,
+      "loss": 3.0867,
+      "step": 25015
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.026343771731013e-05,
+      "loss": 3.135,
+      "step": 25020
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.025285783640215e-05,
+      "loss": 3.025,
+      "step": 25025
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.024227687061051e-05,
+      "loss": 3.0518,
+      "step": 25030
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.023169482050201e-05,
+      "loss": 3.1116,
+      "step": 25035
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.02211116866435e-05,
+      "loss": 2.9893,
+      "step": 25040
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.021052746960186e-05,
+      "loss": 2.9393,
+      "step": 25045
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.019994216994408e-05,
+      "loss": 3.0881,
+      "step": 25050
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.018935578823717e-05,
+      "loss": 2.9008,
+      "step": 25055
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.017876832504823e-05,
+      "loss": 2.9316,
+      "step": 25060
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.016817978094439e-05,
+      "loss": 2.988,
+      "step": 25065
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.015759015649282e-05,
+      "loss": 2.9712,
+      "step": 25070
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.014699945226081e-05,
+      "loss": 2.896,
+      "step": 25075
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.013640766881564e-05,
+      "loss": 3.0641,
+      "step": 25080
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.01258148067247e-05,
+      "loss": 3.0528,
+      "step": 25085
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.011522086655543e-05,
+      "loss": 3.0115,
+      "step": 25090
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.010462584887527e-05,
+      "loss": 3.2173,
+      "step": 25095
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.009402975425181e-05,
+      "loss": 2.9622,
+      "step": 25100
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.008343258325261e-05,
+      "loss": 3.1617,
+      "step": 25105
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.007283433644537e-05,
+      "loss": 3.0792,
+      "step": 25110
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.006223501439778e-05,
+      "loss": 3.0898,
+      "step": 25115
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.00516346176776e-05,
+      "loss": 3.0009,
+      "step": 25120
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.00410331468527e-05,
+      "loss": 3.0398,
+      "step": 25125
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.003043060249094e-05,
+      "loss": 2.9844,
+      "step": 25130
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.001982698516026e-05,
+      "loss": 3.0696,
+      "step": 25135
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.000922229542867e-05,
+      "loss": 2.9607,
+      "step": 25140
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.999861653386427e-05,
+      "loss": 3.0608,
+      "step": 25145
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.998800970103512e-05,
+      "loss": 3.0079,
+      "step": 25150
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.997740179750942e-05,
+      "loss": 3.0064,
+      "step": 25155
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.996679282385541e-05,
+      "loss": 3.1839,
+      "step": 25160
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.995618278064138e-05,
+      "loss": 3.0581,
+      "step": 25165
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.994557166843567e-05,
+      "loss": 2.989,
+      "step": 25170
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.993495948780667e-05,
+      "loss": 3.0278,
+      "step": 25175
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.992434623932287e-05,
+      "loss": 3.0,
+      "step": 25180
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.991373192355278e-05,
+      "loss": 3.048,
+      "step": 25185
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.990311654106495e-05,
+      "loss": 3.0057,
+      "step": 25190
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.989250009242807e-05,
+      "loss": 3.0535,
+      "step": 25195
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.988188257821078e-05,
+      "loss": 2.8955,
+      "step": 25200
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.987126399898184e-05,
+      "loss": 3.0253,
+      "step": 25205
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.986064435531007e-05,
+      "loss": 3.0823,
+      "step": 25210
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.985002364776432e-05,
+      "loss": 2.8496,
+      "step": 25215
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.98394018769135e-05,
+      "loss": 2.9589,
+      "step": 25220
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.98287790433266e-05,
+      "loss": 3.0788,
+      "step": 25225
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.981815514757266e-05,
+      "loss": 3.0423,
+      "step": 25230
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.980753019022076e-05,
+      "loss": 3.1415,
+      "step": 25235
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.979690417184005e-05,
+      "loss": 3.1551,
+      "step": 25240
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.978627709299972e-05,
+      "loss": 3.1017,
+      "step": 25245
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.977564895426904e-05,
+      "loss": 3.0812,
+      "step": 25250
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.976501975621733e-05,
+      "loss": 2.9703,
+      "step": 25255
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.975438949941395e-05,
+      "loss": 3.0676,
+      "step": 25260
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.974375818442834e-05,
+      "loss": 2.9701,
+      "step": 25265
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.973312581183001e-05,
+      "loss": 3.0822,
+      "step": 25270
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.972249238218846e-05,
+      "loss": 3.2312,
+      "step": 25275
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.971185789607331e-05,
+      "loss": 3.0459,
+      "step": 25280
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.970122235405423e-05,
+      "loss": 3.0172,
+      "step": 25285
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.96905857567009e-05,
+      "loss": 3.1689,
+      "step": 25290
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.967994810458313e-05,
+      "loss": 3.0041,
+      "step": 25295
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.966930939827072e-05,
+      "loss": 3.049,
+      "step": 25300
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.965866963833356e-05,
+      "loss": 3.0106,
+      "step": 25305
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.964802882534159e-05,
+      "loss": 2.9818,
+      "step": 25310
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.963738695986481e-05,
+      "loss": 3.067,
+      "step": 25315
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.962674404247328e-05,
+      "loss": 3.1863,
+      "step": 25320
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.961610007373708e-05,
+      "loss": 3.0926,
+      "step": 25325
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.960545505422639e-05,
+      "loss": 3.0495,
+      "step": 25330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.959480898451144e-05,
+      "loss": 2.9662,
+      "step": 25335
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.95841618651625e-05,
+      "loss": 2.9377,
+      "step": 25340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.95735136967499e-05,
+      "loss": 3.1628,
+      "step": 25345
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.956286447984404e-05,
+      "loss": 3.0872,
+      "step": 25350
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.955221421501536e-05,
+      "loss": 3.0729,
+      "step": 25355
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.954156290283435e-05,
+      "loss": 3.0801,
+      "step": 25360
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.953091054387158e-05,
+      "loss": 3.0338,
+      "step": 25365
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.952025713869768e-05,
+      "loss": 3.0549,
+      "step": 25370
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.950960268788328e-05,
+      "loss": 2.9628,
+      "step": 25375
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.949894719199916e-05,
+      "loss": 2.8867,
+      "step": 25380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.948829065161605e-05,
+      "loss": 2.9059,
+      "step": 25385
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.947763306730481e-05,
+      "loss": 2.9394,
+      "step": 25390
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.946697443963635e-05,
+      "loss": 2.9141,
+      "step": 25395
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.94563147691816e-05,
+      "loss": 3.0179,
+      "step": 25400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.944565405651158e-05,
+      "loss": 3.1019,
+      "step": 25405
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.943499230219733e-05,
+      "loss": 3.0575,
+      "step": 25410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.942432950681e-05,
+      "loss": 2.9925,
+      "step": 25415
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.941366567092072e-05,
+      "loss": 3.0565,
+      "step": 25420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.940300079510076e-05,
+      "loss": 3.0424,
+      "step": 25425
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.939233487992139e-05,
+      "loss": 2.9783,
+      "step": 25430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.938166792595393e-05,
+      "loss": 3.0159,
+      "step": 25435
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.937099993376981e-05,
+      "loss": 3.1141,
+      "step": 25440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.936033090394047e-05,
+      "loss": 3.0424,
+      "step": 25445
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.93496608370374e-05,
+      "loss": 3.1009,
+      "step": 25450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.933898973363217e-05,
+      "loss": 3.0716,
+      "step": 25455
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.932831759429643e-05,
+      "loss": 3.1368,
+      "step": 25460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.93176444196018e-05,
+      "loss": 3.0,
+      "step": 25465
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.930697021012007e-05,
+      "loss": 3.1261,
+      "step": 25470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.929629496642299e-05,
+      "loss": 2.9616,
+      "step": 25475
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.928561868908237e-05,
+      "loss": 3.0328,
+      "step": 25480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 6.927494137867015e-05,
+      "loss": 3.1082,
+      "step": 25485
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.926426303575829e-05,
+      "loss": 3.1473,
+      "step": 25490
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.925358366091876e-05,
+      "loss": 2.9616,
+      "step": 25495
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.924290325472365e-05,
+      "loss": 3.1196,
+      "step": 25500
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.923222181774504e-05,
+      "loss": 3.1501,
+      "step": 25505
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.922153935055514e-05,
+      "loss": 2.958,
+      "step": 25510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.921085585372617e-05,
+      "loss": 2.9257,
+      "step": 25515
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.920017132783038e-05,
+      "loss": 3.0175,
+      "step": 25520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.918948577344015e-05,
+      "loss": 2.993,
+      "step": 25525
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.917879919112786e-05,
+      "loss": 3.2239,
+      "step": 25530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.916811158146595e-05,
+      "loss": 3.0465,
+      "step": 25535
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.915742294502693e-05,
+      "loss": 3.1217,
+      "step": 25540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.914673328238335e-05,
+      "loss": 3.0878,
+      "step": 25545
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.91360425941078e-05,
+      "loss": 2.9964,
+      "step": 25550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.9125350880773e-05,
+      "loss": 3.0713,
+      "step": 25555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.911465814295164e-05,
+      "loss": 2.9962,
+      "step": 25560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.910396438121649e-05,
+      "loss": 2.9112,
+      "step": 25565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.909326959614041e-05,
+      "loss": 3.0688,
+      "step": 25570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.908257378829627e-05,
+      "loss": 3.0429,
+      "step": 25575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.9071876958257e-05,
+      "loss": 3.0185,
+      "step": 25580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.906117910659562e-05,
+      "loss": 3.1298,
+      "step": 25585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.905048023388517e-05,
+      "loss": 2.9407,
+      "step": 25590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.903978034069874e-05,
+      "loss": 3.0685,
+      "step": 25595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.902907942760953e-05,
+      "loss": 2.8455,
+      "step": 25600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.901837749519073e-05,
+      "loss": 2.893,
+      "step": 25605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.90076745440156e-05,
+      "loss": 3.0886,
+      "step": 25610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.899697057465748e-05,
+      "loss": 3.0935,
+      "step": 25615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.898626558768975e-05,
+      "loss": 2.9553,
+      "step": 25620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.897555958368585e-05,
+      "loss": 2.9979,
+      "step": 25625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.896485256321923e-05,
+      "loss": 2.9805,
+      "step": 25630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.895414452686346e-05,
+      "loss": 2.9231,
+      "step": 25635
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.894343547519216e-05,
+      "loss": 3.0879,
+      "step": 25640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.893272540877893e-05,
+      "loss": 2.9963,
+      "step": 25645
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.892201432819753e-05,
+      "loss": 3.0256,
+      "step": 25650
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.891130223402168e-05,
+      "loss": 2.9133,
+      "step": 25655
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.89005891268252e-05,
+      "loss": 3.0576,
+      "step": 25660
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.888987500718196e-05,
+      "loss": 2.9902,
+      "step": 25665
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.88791598756659e-05,
+      "loss": 3.053,
+      "step": 25670
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.886844373285098e-05,
+      "loss": 2.994,
+      "step": 25675
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.885772657931124e-05,
+      "loss": 3.0834,
+      "step": 25680
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.884700841562075e-05,
+      "loss": 3.0528,
+      "step": 25685
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.883628924235366e-05,
+      "loss": 3.1848,
+      "step": 25690
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.882556906008418e-05,
+      "loss": 3.0604,
+      "step": 25695
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.881484786938654e-05,
+      "loss": 3.0288,
+      "step": 25700
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.880412567083501e-05,
+      "loss": 3.0205,
+      "step": 25705
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.879340246500401e-05,
+      "loss": 3.0825,
+      "step": 25710
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.87826782524679e-05,
+      "loss": 3.0412,
+      "step": 25715
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.877195303380115e-05,
+      "loss": 2.9951,
+      "step": 25720
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.876122680957832e-05,
+      "loss": 3.0428,
+      "step": 25725
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.875049958037391e-05,
+      "loss": 3.087,
+      "step": 25730
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.87397713467626e-05,
+      "loss": 2.8961,
+      "step": 25735
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.872904210931906e-05,
+      "loss": 3.0864,
+      "step": 25740
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.8718311868618e-05,
+      "loss": 3.097,
+      "step": 25745
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.870758062523423e-05,
+      "loss": 2.9651,
+      "step": 25750
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.869684837974258e-05,
+      "loss": 3.0356,
+      "step": 25755
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.868611513271793e-05,
+      "loss": 2.9854,
+      "step": 25760
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.867538088473524e-05,
+      "loss": 3.0429,
+      "step": 25765
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.866464563636953e-05,
+      "loss": 3.1302,
+      "step": 25770
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.865390938819583e-05,
+      "loss": 3.1499,
+      "step": 25775
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.864317214078925e-05,
+      "loss": 3.1197,
+      "step": 25780
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.863243389472496e-05,
+      "loss": 3.0176,
+      "step": 25785
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.862169465057815e-05,
+      "loss": 3.0434,
+      "step": 25790
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.861095440892412e-05,
+      "loss": 3.0503,
+      "step": 25795
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.86002131703382e-05,
+      "loss": 3.0817,
+      "step": 25800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.858947093539573e-05,
+      "loss": 3.061,
+      "step": 25805
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.857872770467215e-05,
+      "loss": 2.9961,
+      "step": 25810
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.856798347874295e-05,
+      "loss": 3.0258,
+      "step": 25815
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.855723825818366e-05,
+      "loss": 3.0697,
+      "step": 25820
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.854649204356987e-05,
+      "loss": 3.0374,
+      "step": 25825
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.853574483547724e-05,
+      "loss": 2.93,
+      "step": 25830
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.852499663448141e-05,
+      "loss": 3.052,
+      "step": 25835
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.851424744115822e-05,
+      "loss": 2.8765,
+      "step": 25840
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.850349725608339e-05,
+      "loss": 3.1511,
+      "step": 25845
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.849274607983278e-05,
+      "loss": 3.0289,
+      "step": 25850
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.848199391298235e-05,
+      "loss": 2.9896,
+      "step": 25855
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.847124075610802e-05,
+      "loss": 2.9902,
+      "step": 25860
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.84604866097858e-05,
+      "loss": 3.0369,
+      "step": 25865
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.844973147459179e-05,
+      "loss": 3.0681,
+      "step": 25870
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.843897535110207e-05,
+      "loss": 2.983,
+      "step": 25875
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.842821823989286e-05,
+      "loss": 3.0735,
+      "step": 25880
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.841746014154034e-05,
+      "loss": 3.121,
+      "step": 25885
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.840670105662082e-05,
+      "loss": 2.904,
+      "step": 25890
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.83959409857106e-05,
+      "loss": 3.0563,
+      "step": 25895
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.838517992938609e-05,
+      "loss": 3.06,
+      "step": 25900
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.837441788822373e-05,
+      "loss": 2.9941,
+      "step": 25905
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.836365486279998e-05,
+      "loss": 3.1487,
+      "step": 25910
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.83528908536914e-05,
+      "loss": 3.1187,
+      "step": 25915
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.834212586147462e-05,
+      "loss": 3.1314,
+      "step": 25920
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.833135988672622e-05,
+      "loss": 2.9675,
+      "step": 25925
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.832059293002295e-05,
+      "loss": 2.9923,
+      "step": 25930
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.830982499194154e-05,
+      "loss": 2.9606,
+      "step": 25935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.82990560730588e-05,
+      "loss": 3.0385,
+      "step": 25940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.82882861739516e-05,
+      "loss": 3.0306,
+      "step": 25945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.827751529519684e-05,
+      "loss": 3.2166,
+      "step": 25950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.826674343737148e-05,
+      "loss": 3.1034,
+      "step": 25955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.825597060105254e-05,
+      "loss": 3.0706,
+      "step": 25960
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.824519678681709e-05,
+      "loss": 2.977,
+      "step": 25965
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.823442199524224e-05,
+      "loss": 3.0443,
+      "step": 25970
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.822364622690517e-05,
+      "loss": 2.9299,
+      "step": 25975
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.821286948238312e-05,
+      "loss": 2.9669,
+      "step": 25980
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.820209176225334e-05,
+      "loss": 3.0535,
+      "step": 25985
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.819131306709319e-05,
+      "loss": 3.0766,
+      "step": 25990
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.818053339748001e-05,
+      "loss": 3.0758,
+      "step": 25995
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.816975275399126e-05,
+      "loss": 2.9443,
+      "step": 26000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.815897113720445e-05,
+      "loss": 2.7828,
+      "step": 26005
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.814818854769707e-05,
+      "loss": 3.1479,
+      "step": 26010
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.813740498604673e-05,
+      "loss": 3.045,
+      "step": 26015
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.812662045283109e-05,
+      "loss": 2.9563,
+      "step": 26020
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.811583494862782e-05,
+      "loss": 3.0155,
+      "step": 26025
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.810504847401469e-05,
+      "loss": 3.1607,
+      "step": 26030
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.809426102956946e-05,
+      "loss": 3.0811,
+      "step": 26035
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.808347261587003e-05,
+      "loss": 3.0012,
+      "step": 26040
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.807268323349424e-05,
+      "loss": 3.1309,
+      "step": 26045
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.806189288302012e-05,
+      "loss": 2.9064,
+      "step": 26050
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.805110156502564e-05,
+      "loss": 3.0939,
+      "step": 26055
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.804030928008882e-05,
+      "loss": 2.9782,
+      "step": 26060
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.802951602878782e-05,
+      "loss": 3.0743,
+      "step": 26065
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.80187218117008e-05,
+      "loss": 2.9997,
+      "step": 26070
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.800792662940594e-05,
+      "loss": 2.9874,
+      "step": 26075
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.799713048248154e-05,
+      "loss": 3.1101,
+      "step": 26080
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.798633337150589e-05,
+      "loss": 3.1342,
+      "step": 26085
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.797553529705738e-05,
+      "loss": 3.1833,
+      "step": 26090
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.796473625971441e-05,
+      "loss": 3.0463,
+      "step": 26095
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.795393626005546e-05,
+      "loss": 3.1378,
+      "step": 26100
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.794313529865907e-05,
+      "loss": 3.0444,
+      "step": 26105
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.793233337610378e-05,
+      "loss": 3.0344,
+      "step": 26110
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.792153049296824e-05,
+      "loss": 3.0655,
+      "step": 26115
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.791072664983112e-05,
+      "loss": 3.0715,
+      "step": 26120
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.789992184727115e-05,
+      "loss": 2.9892,
+      "step": 26125
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.788911608586712e-05,
+      "loss": 3.0557,
+      "step": 26130
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.787830936619784e-05,
+      "loss": 2.9853,
+      "step": 26135
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.786750168884221e-05,
+      "loss": 3.0716,
+      "step": 26140
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.785669305437917e-05,
+      "loss": 3.0277,
+      "step": 26145
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.784588346338768e-05,
+      "loss": 3.1026,
+      "step": 26150
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.783507291644678e-05,
+      "loss": 3.0289,
+      "step": 26155
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.782426141413558e-05,
+      "loss": 2.9744,
+      "step": 26160
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 6.781344895703322e-05,
+      "loss": 3.0273,
+      "step": 26165
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.780263554571886e-05,
+      "loss": 3.1529,
+      "step": 26170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.779182118077178e-05,
+      "loss": 3.0802,
+      "step": 26175
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.778100586277124e-05,
+      "loss": 3.1235,
+      "step": 26180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.777018959229661e-05,
+      "loss": 2.9954,
+      "step": 26185
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.775937236992725e-05,
+      "loss": 3.0664,
+      "step": 26190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.774855419624264e-05,
+      "loss": 3.0494,
+      "step": 26195
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.773773507182226e-05,
+      "loss": 3.1164,
+      "step": 26200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.772691499724565e-05,
+      "loss": 3.1456,
+      "step": 26205
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.771609397309244e-05,
+      "loss": 3.0952,
+      "step": 26210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.770527199994224e-05,
+      "loss": 3.046,
+      "step": 26215
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.769444907837477e-05,
+      "loss": 3.0738,
+      "step": 26220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.768362520896978e-05,
+      "loss": 2.9823,
+      "step": 26225
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.767280039230706e-05,
+      "loss": 3.1185,
+      "step": 26230
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.766197462896651e-05,
+      "loss": 3.0414,
+      "step": 26235
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.765114791952795e-05,
+      "loss": 3.107,
+      "step": 26240
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.76403202645714e-05,
+      "loss": 3.0669,
+      "step": 26245
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.762949166467684e-05,
+      "loss": 2.9073,
+      "step": 26250
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.761866212042433e-05,
+      "loss": 3.1603,
+      "step": 26255
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.760783163239398e-05,
+      "loss": 2.9477,
+      "step": 26260
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.759700020116593e-05,
+      "loss": 3.0439,
+      "step": 26265
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.758616782732039e-05,
+      "loss": 2.8746,
+      "step": 26270
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.757533451143764e-05,
+      "loss": 2.8401,
+      "step": 26275
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.756450025409794e-05,
+      "loss": 2.9615,
+      "step": 26280
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.755366505588173e-05,
+      "loss": 2.9869,
+      "step": 26285
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.754282891736934e-05,
+      "loss": 3.1435,
+      "step": 26290
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.753199183914126e-05,
+      "loss": 3.1393,
+      "step": 26295
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.752115382177798e-05,
+      "loss": 3.04,
+      "step": 26300
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.751031486586009e-05,
+      "loss": 3.0362,
+      "step": 26305
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.74994749719682e-05,
+      "loss": 3.0793,
+      "step": 26310
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.748863414068293e-05,
+      "loss": 2.9952,
+      "step": 26315
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.747779237258503e-05,
+      "loss": 3.2207,
+      "step": 26320
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.746694966825523e-05,
+      "loss": 2.8921,
+      "step": 26325
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.745610602827436e-05,
+      "loss": 3.024,
+      "step": 26330
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.744526145322329e-05,
+      "loss": 2.994,
+      "step": 26335
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.74344159436829e-05,
+      "loss": 3.1267,
+      "step": 26340
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.742356950023418e-05,
+      "loss": 3.132,
+      "step": 26345
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.741272212345813e-05,
+      "loss": 2.9934,
+      "step": 26350
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.74018738139358e-05,
+      "loss": 3.0393,
+      "step": 26355
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.739102457224832e-05,
+      "loss": 3.0214,
+      "step": 26360
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.738017439897684e-05,
+      "loss": 3.0877,
+      "step": 26365
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.736932329470256e-05,
+      "loss": 3.0611,
+      "step": 26370
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.735847126000677e-05,
+      "loss": 2.9148,
+      "step": 26375
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.734761829547073e-05,
+      "loss": 3.0668,
+      "step": 26380
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.733676440167588e-05,
+      "loss": 2.8618,
+      "step": 26385
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.732590957920355e-05,
+      "loss": 3.091,
+      "step": 26390
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.731505382863525e-05,
+      "loss": 3.0884,
+      "step": 26395
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.730419715055245e-05,
+      "loss": 3.0321,
+      "step": 26400
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.729333954553674e-05,
+      "loss": 3.087,
+      "step": 26405
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.728248101416974e-05,
+      "loss": 3.0971,
+      "step": 26410
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.727162155703306e-05,
+      "loss": 3.156,
+      "step": 26415
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.726076117470844e-05,
+      "loss": 3.1685,
+      "step": 26420
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.724989986777763e-05,
+      "loss": 3.0929,
+      "step": 26425
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.723903763682245e-05,
+      "loss": 2.9052,
+      "step": 26430
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.722817448242474e-05,
+      "loss": 3.1199,
+      "step": 26435
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.721731040516642e-05,
+      "loss": 3.0348,
+      "step": 26440
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.720644540562943e-05,
+      "loss": 2.9968,
+      "step": 26445
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.719557948439578e-05,
+      "loss": 2.9339,
+      "step": 26450
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.718471264204755e-05,
+      "loss": 3.14,
+      "step": 26455
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.717384487916679e-05,
+      "loss": 3.0707,
+      "step": 26460
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.71629761963357e-05,
+      "loss": 3.0424,
+      "step": 26465
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.715210659413645e-05,
+      "loss": 2.9777,
+      "step": 26470
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.714123607315133e-05,
+      "loss": 3.2336,
+      "step": 26475
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.71303646339626e-05,
+      "loss": 3.1713,
+      "step": 26480
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.711949227715263e-05,
+      "loss": 3.0513,
+      "step": 26485
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.710861900330382e-05,
+      "loss": 2.9809,
+      "step": 26490
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.709774481299861e-05,
+      "loss": 3.0113,
+      "step": 26495
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.70868697068195e-05,
+      "loss": 2.9589,
+      "step": 26500
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.707599368534904e-05,
+      "loss": 2.8688,
+      "step": 26505
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.706511674916982e-05,
+      "loss": 3.147,
+      "step": 26510
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.70542388988645e-05,
+      "loss": 3.1229,
+      "step": 26515
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.704336013501574e-05,
+      "loss": 3.0855,
+      "step": 26520
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.703248045820632e-05,
+      "loss": 3.0218,
+      "step": 26525
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.7021599869019e-05,
+      "loss": 3.0322,
+      "step": 26530
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.701071836803663e-05,
+      "loss": 3.0117,
+      "step": 26535
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.69998359558421e-05,
+      "loss": 2.9774,
+      "step": 26540
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.698895263301837e-05,
+      "loss": 2.9245,
+      "step": 26545
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.697806840014837e-05,
+      "loss": 2.9143,
+      "step": 26550
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.69671832578152e-05,
+      "loss": 3.0846,
+      "step": 26555
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.69562972066019e-05,
+      "loss": 2.9431,
+      "step": 26560
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.694541024709162e-05,
+      "loss": 3.1009,
+      "step": 26565
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.693452237986755e-05,
+      "loss": 3.1076,
+      "step": 26570
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.692363360551289e-05,
+      "loss": 3.0148,
+      "step": 26575
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.691274392461094e-05,
+      "loss": 2.9178,
+      "step": 26580
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.690185333774502e-05,
+      "loss": 2.8794,
+      "step": 26585
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.689096184549852e-05,
+      "loss": 3.0641,
+      "step": 26590
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.688006944845484e-05,
+      "loss": 3.1628,
+      "step": 26595
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.686917614719748e-05,
+      "loss": 3.096,
+      "step": 26600
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.685828194230991e-05,
+      "loss": 2.9206,
+      "step": 26605
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.684738683437576e-05,
+      "loss": 2.901,
+      "step": 26610
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.683649082397865e-05,
+      "loss": 2.9969,
+      "step": 26615
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.682559391170218e-05,
+      "loss": 3.0546,
+      "step": 26620
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.681469609813012e-05,
+      "loss": 2.8945,
+      "step": 26625
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.680379738384618e-05,
+      "loss": 3.0824,
+      "step": 26630
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.679289776943425e-05,
+      "loss": 3.0399,
+      "step": 26635
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.678199725547812e-05,
+      "loss": 3.01,
+      "step": 26640
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.677109584256174e-05,
+      "loss": 2.9553,
+      "step": 26645
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.676019353126903e-05,
+      "loss": 3.0378,
+      "step": 26650
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.674929032218401e-05,
+      "loss": 2.9907,
+      "step": 26655
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.673838621589074e-05,
+      "loss": 3.0883,
+      "step": 26660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.672748121297332e-05,
+      "loss": 3.0381,
+      "step": 26665
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.671657531401587e-05,
+      "loss": 3.0351,
+      "step": 26670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.670566851960261e-05,
+      "loss": 2.9663,
+      "step": 26675
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.66947608303178e-05,
+      "loss": 3.23,
+      "step": 26680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.668385224674568e-05,
+      "loss": 2.9828,
+      "step": 26685
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.667294276947065e-05,
+      "loss": 3.0951,
+      "step": 26690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.666203239907704e-05,
+      "loss": 3.0705,
+      "step": 26695
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.665112113614933e-05,
+      "loss": 3.0088,
+      "step": 26700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.664020898127197e-05,
+      "loss": 3.0067,
+      "step": 26705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.662929593502951e-05,
+      "loss": 2.9854,
+      "step": 26710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.661838199800654e-05,
+      "loss": 2.887,
+      "step": 26715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.660746717078766e-05,
+      "loss": 3.0321,
+      "step": 26720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.659655145395755e-05,
+      "loss": 3.0917,
+      "step": 26725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.658563484810094e-05,
+      "loss": 3.1944,
+      "step": 26730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.65747173538026e-05,
+      "loss": 2.9979,
+      "step": 26735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.656379897164733e-05,
+      "loss": 3.0367,
+      "step": 26740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.655287970222001e-05,
+      "loss": 2.9513,
+      "step": 26745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.654195954610558e-05,
+      "loss": 3.0565,
+      "step": 26750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.653103850388894e-05,
+      "loss": 2.8759,
+      "step": 26755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.652011657615511e-05,
+      "loss": 3.022,
+      "step": 26760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.650919376348919e-05,
+      "loss": 2.9947,
+      "step": 26765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.649827006647622e-05,
+      "loss": 3.0803,
+      "step": 26770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.64873454857014e-05,
+      "loss": 3.0697,
+      "step": 26775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.64764200217499e-05,
+      "loss": 2.9836,
+      "step": 26780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.646549367520696e-05,
+      "loss": 2.9325,
+      "step": 26785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.64545664466579e-05,
+      "loss": 2.9748,
+      "step": 26790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.644363833668803e-05,
+      "loss": 3.0112,
+      "step": 26795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.643270934588275e-05,
+      "loss": 3.1733,
+      "step": 26800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.642177947482748e-05,
+      "loss": 3.1251,
+      "step": 26805
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.641084872410771e-05,
+      "loss": 3.0029,
+      "step": 26810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.639991709430897e-05,
+      "loss": 2.9176,
+      "step": 26815
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.638898458601682e-05,
+      "loss": 3.0437,
+      "step": 26820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.637805119981688e-05,
+      "loss": 3.0597,
+      "step": 26825
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.636711693629483e-05,
+      "loss": 3.0757,
+      "step": 26830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.63561817960364e-05,
+      "loss": 3.1403,
+      "step": 26835
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.634524577962731e-05,
+      "loss": 3.1007,
+      "step": 26840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.633430888765342e-05,
+      "loss": 3.1239,
+      "step": 26845
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.632337112070053e-05,
+      "loss": 3.1046,
+      "step": 26850
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.631243247935459e-05,
+      "loss": 2.9895,
+      "step": 26855
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.630149296420151e-05,
+      "loss": 3.0125,
+      "step": 26860
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.629055257582733e-05,
+      "loss": 2.9796,
+      "step": 26865
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.627961131481807e-05,
+      "loss": 3.0316,
+      "step": 26870
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.62686691817598e-05,
+      "loss": 3.0277,
+      "step": 26875
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.625772617723869e-05,
+      "loss": 3.0117,
+      "step": 26880
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.624678230184092e-05,
+      "loss": 3.1012,
+      "step": 26885
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.62358375561527e-05,
+      "loss": 2.8393,
+      "step": 26890
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.622489194076031e-05,
+      "loss": 3.0484,
+      "step": 26895
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.621394545625009e-05,
+      "loss": 3.1148,
+      "step": 26900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.620299810320839e-05,
+      "loss": 3.0232,
+      "step": 26905
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.619204988222164e-05,
+      "loss": 3.0546,
+      "step": 26910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.618110079387629e-05,
+      "loss": 3.0795,
+      "step": 26915
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.617015083875888e-05,
+      "loss": 3.073,
+      "step": 26920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.615920001745593e-05,
+      "loss": 3.0886,
+      "step": 26925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.614824833055405e-05,
+      "loss": 3.0715,
+      "step": 26930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.61372957786399e-05,
+      "loss": 3.0101,
+      "step": 26935
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.612634236230017e-05,
+      "loss": 3.0541,
+      "step": 26940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.61153880821216e-05,
+      "loss": 3.0477,
+      "step": 26945
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.610443293869096e-05,
+      "loss": 3.105,
+      "step": 26950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.609347693259511e-05,
+      "loss": 3.0213,
+      "step": 26955
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.608252006442093e-05,
+      "loss": 3.0439,
+      "step": 26960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.607156233475532e-05,
+      "loss": 3.1446,
+      "step": 26965
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.606060374418529e-05,
+      "loss": 3.0431,
+      "step": 26970
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.604964429329782e-05,
+      "loss": 3.0172,
+      "step": 26975
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.603868398267998e-05,
+      "loss": 3.0097,
+      "step": 26980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.602772281291892e-05,
+      "loss": 3.0519,
+      "step": 26985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.601676078460174e-05,
+      "loss": 2.9758,
+      "step": 26990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.600579789831569e-05,
+      "loss": 2.9197,
+      "step": 26995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.599483415464798e-05,
+      "loss": 2.9566,
+      "step": 27000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.598386955418593e-05,
+      "loss": 2.9556,
+      "step": 27005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.597290409751686e-05,
+      "loss": 3.0601,
+      "step": 27010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.596193778522818e-05,
+      "loss": 2.9378,
+      "step": 27015
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.595097061790731e-05,
+      "loss": 3.0698,
+      "step": 27020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.594000259614172e-05,
+      "loss": 3.0742,
+      "step": 27025
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.592903372051896e-05,
+      "loss": 3.0545,
+      "step": 27030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.591806399162656e-05,
+      "loss": 3.1963,
+      "step": 27035
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.590709341005216e-05,
+      "loss": 3.0018,
+      "step": 27040
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.58961219763834e-05,
+      "loss": 3.0596,
+      "step": 27045
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.588514969120803e-05,
+      "loss": 2.9486,
+      "step": 27050
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.587417655511376e-05,
+      "loss": 3.1314,
+      "step": 27055
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.58632025686884e-05,
+      "loss": 3.0154,
+      "step": 27060
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.585222773251979e-05,
+      "loss": 3.0524,
+      "step": 27065
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.584125204719583e-05,
+      "loss": 3.1378,
+      "step": 27070
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.583027551330443e-05,
+      "loss": 3.1141,
+      "step": 27075
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.581929813143359e-05,
+      "loss": 3.0795,
+      "step": 27080
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.580831990217134e-05,
+      "loss": 3.0728,
+      "step": 27085
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.579734082610573e-05,
+      "loss": 3.0448,
+      "step": 27090
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.57863609038249e-05,
+      "loss": 3.0745,
+      "step": 27095
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.577538013591699e-05,
+      "loss": 3.0397,
+      "step": 27100
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.57643985229702e-05,
+      "loss": 2.936,
+      "step": 27105
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.57534160655728e-05,
+      "loss": 3.0717,
+      "step": 27110
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.57424327643131e-05,
+      "loss": 3.0379,
+      "step": 27115
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.573144861977938e-05,
+      "loss": 3.0885,
+      "step": 27120
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.572046363256012e-05,
+      "loss": 3.0396,
+      "step": 27125
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.570947780324367e-05,
+      "loss": 3.0377,
+      "step": 27130
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.569849113241854e-05,
+      "loss": 2.9646,
+      "step": 27135
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.568750362067325e-05,
+      "loss": 3.1374,
+      "step": 27140
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.567651526859637e-05,
+      "loss": 3.0177,
+      "step": 27145
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.566552607677653e-05,
+      "loss": 3.133,
+      "step": 27150
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.565453604580236e-05,
+      "loss": 3.165,
+      "step": 27155
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.564354517626255e-05,
+      "loss": 2.9991,
+      "step": 27160
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.563255346874589e-05,
+      "loss": 3.1383,
+      "step": 27165
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.562156092384114e-05,
+      "loss": 3.1094,
+      "step": 27170
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.561056754213716e-05,
+      "loss": 3.0747,
+      "step": 27175
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.559957332422279e-05,
+      "loss": 2.9775,
+      "step": 27180
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.558857827068701e-05,
+      "loss": 3.0576,
+      "step": 27185
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.557758238211877e-05,
+      "loss": 2.9596,
+      "step": 27190
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.556658565910707e-05,
+      "loss": 3.02,
+      "step": 27195
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.5555588102241e-05,
+      "loss": 3.0063,
+      "step": 27200
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.554458971210962e-05,
+      "loss": 3.1872,
+      "step": 27205
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.553359048930215e-05,
+      "loss": 3.0331,
+      "step": 27210
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.55225904344077e-05,
+      "loss": 2.9599,
+      "step": 27215
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.551158954801558e-05,
+      "loss": 3.1187,
+      "step": 27220
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.550058783071506e-05,
+      "loss": 3.046,
+      "step": 27225
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.548958528309544e-05,
+      "loss": 3.0001,
+      "step": 27230
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.54785819057461e-05,
+      "loss": 3.0684,
+      "step": 27235
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.546757769925646e-05,
+      "loss": 3.0959,
+      "step": 27240
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.5456572664216e-05,
+      "loss": 3.0043,
+      "step": 27245
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.544556680121423e-05,
+      "loss": 2.9346,
+      "step": 27250
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.543456011084065e-05,
+      "loss": 3.0398,
+      "step": 27255
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.542355259368492e-05,
+      "loss": 2.99,
+      "step": 27260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.541254425033663e-05,
+      "loss": 2.9816,
+      "step": 27265
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.540153508138548e-05,
+      "loss": 2.9315,
+      "step": 27270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.539052508742123e-05,
+      "loss": 2.8953,
+      "step": 27275
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.537951426903358e-05,
+      "loss": 2.9977,
+      "step": 27280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.53685026268124e-05,
+      "loss": 2.9221,
+      "step": 27285
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.535749016134754e-05,
+      "loss": 3.0033,
+      "step": 27290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.53464768732289e-05,
+      "loss": 2.8802,
+      "step": 27295
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.533546276304643e-05,
+      "loss": 3.0128,
+      "step": 27300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.532444783139011e-05,
+      "loss": 3.0513,
+      "step": 27305
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.531343207885e-05,
+      "loss": 3.0324,
+      "step": 27310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.530241550601616e-05,
+      "loss": 2.9653,
+      "step": 27315
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.529139811347872e-05,
+      "loss": 2.9792,
+      "step": 27320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.528037990182786e-05,
+      "loss": 3.1732,
+      "step": 27325
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.526936087165377e-05,
+      "loss": 3.0424,
+      "step": 27330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.525834102354671e-05,
+      "loss": 2.9399,
+      "step": 27335
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.524732035809699e-05,
+      "loss": 3.0932,
+      "step": 27340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.523629887589496e-05,
+      "loss": 3.1479,
+      "step": 27345
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.522527657753099e-05,
+      "loss": 3.018,
+      "step": 27350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.521425346359553e-05,
+      "loss": 2.896,
+      "step": 27355
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.520322953467903e-05,
+      "loss": 3.0142,
+      "step": 27360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.519220479137202e-05,
+      "loss": 3.1341,
+      "step": 27365
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.518117923426506e-05,
+      "loss": 2.7908,
+      "step": 27370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.517015286394877e-05,
+      "loss": 2.9418,
+      "step": 27375
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.515912568101376e-05,
+      "loss": 3.0741,
+      "step": 27380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.514809768605077e-05,
+      "loss": 3.0879,
+      "step": 27385
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.513706887965051e-05,
+      "loss": 3.0581,
+      "step": 27390
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.512603926240377e-05,
+      "loss": 2.8984,
+      "step": 27395
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.511500883490136e-05,
+      "loss": 3.097,
+      "step": 27400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.510397759773416e-05,
+      "loss": 3.1763,
+      "step": 27405
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.509294555149307e-05,
+      "loss": 3.0306,
+      "step": 27410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.508191269676904e-05,
+      "loss": 3.007,
+      "step": 27415
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.507087903415307e-05,
+      "loss": 3.0946,
+      "step": 27420
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.505984456423622e-05,
+      "loss": 3.026,
+      "step": 27425
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.504880928760955e-05,
+      "loss": 2.9798,
+      "step": 27430
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.503777320486417e-05,
+      "loss": 2.9228,
+      "step": 27435
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.502673631659129e-05,
+      "loss": 3.0722,
+      "step": 27440
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.50156986233821e-05,
+      "loss": 3.1159,
+      "step": 27445
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.500466012582784e-05,
+      "loss": 3.0771,
+      "step": 27450
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.499362082451985e-05,
+      "loss": 3.0508,
+      "step": 27455
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.498258072004944e-05,
+      "loss": 3.2188,
+      "step": 27460
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.4971539813008e-05,
+      "loss": 3.1126,
+      "step": 27465
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.496049810398696e-05,
+      "loss": 3.063,
+      "step": 27470
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.494945559357779e-05,
+      "loss": 3.1155,
+      "step": 27475
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.493841228237201e-05,
+      "loss": 3.0245,
+      "step": 27480
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.492736817096116e-05,
+      "loss": 3.0675,
+      "step": 27485
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.491632325993686e-05,
+      "loss": 3.0657,
+      "step": 27490
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.490527754989073e-05,
+      "loss": 3.1304,
+      "step": 27495
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.489423104141448e-05,
+      "loss": 2.9246,
+      "step": 27500
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.488318373509983e-05,
+      "loss": 2.9585,
+      "step": 27505
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.487213563153852e-05,
+      "loss": 2.9961,
+      "step": 27510
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.48610867313224e-05,
+      "loss": 3.0597,
+      "step": 27515
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.485003703504332e-05,
+      "loss": 3.1043,
+      "step": 27520
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.483898654329316e-05,
+      "loss": 2.9637,
+      "step": 27525
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.48279352566639e-05,
+      "loss": 3.0227,
+      "step": 27530
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.481688317574748e-05,
+      "loss": 3.0458,
+      "step": 27535
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.480583030113594e-05,
+      "loss": 2.9748,
+      "step": 27540
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.479477663342135e-05,
+      "loss": 3.0292,
+      "step": 27545
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.478372217319582e-05,
+      "loss": 3.0692,
+      "step": 27550
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.477266692105153e-05,
+      "loss": 3.1412,
+      "step": 27555
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.476161087758063e-05,
+      "loss": 3.0403,
+      "step": 27560
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.475055404337538e-05,
+      "loss": 2.9981,
+      "step": 27565
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.473949641902807e-05,
+      "loss": 2.7523,
+      "step": 27570
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.4728438005131e-05,
+      "loss": 2.9935,
+      "step": 27575
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.471737880227658e-05,
+      "loss": 2.9499,
+      "step": 27580
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.470631881105717e-05,
+      "loss": 3.1296,
+      "step": 27585
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.469525803206524e-05,
+      "loss": 3.1069,
+      "step": 27590
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.468419646589327e-05,
+      "loss": 3.1673,
+      "step": 27595
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.467313411313383e-05,
+      "loss": 2.9595,
+      "step": 27600
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.466207097437946e-05,
+      "loss": 3.0402,
+      "step": 27605
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.465100705022278e-05,
+      "loss": 3.0567,
+      "step": 27610
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.463994234125646e-05,
+      "loss": 3.06,
+      "step": 27615
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.462887684807321e-05,
+      "loss": 3.0137,
+      "step": 27620
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.461781057126577e-05,
+      "loss": 3.0639,
+      "step": 27625
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.460674351142693e-05,
+      "loss": 2.9239,
+      "step": 27630
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.45956756691495e-05,
+      "loss": 3.1323,
+      "step": 27635
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.458460704502635e-05,
+      "loss": 3.0214,
+      "step": 27640
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.457353763965042e-05,
+      "loss": 3.0432,
+      "step": 27645
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.456246745361464e-05,
+      "loss": 3.0009,
+      "step": 27650
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.455139648751204e-05,
+      "loss": 2.9859,
+      "step": 27655
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.454032474193561e-05,
+      "loss": 2.9633,
+      "step": 27660
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.452925221747845e-05,
+      "loss": 2.9556,
+      "step": 27665
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.45181789147337e-05,
+      "loss": 3.005,
+      "step": 27670
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.45071048342945e-05,
+      "loss": 3.0275,
+      "step": 27675
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.449602997675406e-05,
+      "loss": 2.9841,
+      "step": 27680
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.448495434270562e-05,
+      "loss": 2.9935,
+      "step": 27685
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.447387793274249e-05,
+      "loss": 2.9834,
+      "step": 27690
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.446280074745797e-05,
+      "loss": 3.0733,
+      "step": 27695
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.445172278744547e-05,
+      "loss": 2.9599,
+      "step": 27700
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.444064405329836e-05,
+      "loss": 3.0284,
+      "step": 27705
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.44295645456101e-05,
+      "loss": 3.1664,
+      "step": 27710
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.441848426497424e-05,
+      "loss": 3.0552,
+      "step": 27715
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.440740321198424e-05,
+      "loss": 2.9497,
+      "step": 27720
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.439632138723372e-05,
+      "loss": 3.1084,
+      "step": 27725
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.43852387913163e-05,
+      "loss": 3.1026,
+      "step": 27730
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.437415542482562e-05,
+      "loss": 3.0711,
+      "step": 27735
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.43630712883554e-05,
+      "loss": 2.9901,
+      "step": 27740
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.435198638249937e-05,
+      "loss": 3.1002,
+      "step": 27745
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.434090070785133e-05,
+      "loss": 3.0528,
+      "step": 27750
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.43298142650051e-05,
+      "loss": 3.0605,
+      "step": 27755
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.431872705455454e-05,
+      "loss": 3.0831,
+      "step": 27760
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.430763907709358e-05,
+      "loss": 2.8946,
+      "step": 27765
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.429655033321613e-05,
+      "loss": 3.1487,
+      "step": 27770
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.428546082351621e-05,
+      "loss": 3.145,
+      "step": 27775
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.427437054858784e-05,
+      "loss": 2.9145,
+      "step": 27780
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.426327950902511e-05,
+      "loss": 2.9487,
+      "step": 27785
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.425218770542211e-05,
+      "loss": 3.0842,
+      "step": 27790
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.424109513837299e-05,
+      "loss": 3.0845,
+      "step": 27795
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.423000180847198e-05,
+      "loss": 3.0606,
+      "step": 27800
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.421890771631327e-05,
+      "loss": 3.1405,
+      "step": 27805
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.420781286249117e-05,
+      "loss": 3.027,
+      "step": 27810
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.41967172476e-05,
+      "loss": 3.0613,
+      "step": 27815
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.41856208722341e-05,
+      "loss": 2.9333,
+      "step": 27820
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.417452373698786e-05,
+      "loss": 2.959,
+      "step": 27825
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.416342584245574e-05,
+      "loss": 3.0266,
+      "step": 27830
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.415232718923222e-05,
+      "loss": 2.9777,
+      "step": 27835
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.41412277779118e-05,
+      "loss": 3.0861,
+      "step": 27840
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.413012760908908e-05,
+      "loss": 2.9276,
+      "step": 27845
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.411902668335862e-05,
+      "loss": 3.197,
+      "step": 27850
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.410792500131509e-05,
+      "loss": 3.0481,
+      "step": 27855
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.409682256355316e-05,
+      "loss": 3.1234,
+      "step": 27860
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.408571937066756e-05,
+      "loss": 3.1651,
+      "step": 27865
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.407461542325305e-05,
+      "loss": 2.9377,
+      "step": 27870
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.406351072190442e-05,
+      "loss": 3.0586,
+      "step": 27875
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.405240526721656e-05,
+      "loss": 2.9299,
+      "step": 27880
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.404129905978432e-05,
+      "loss": 3.1194,
+      "step": 27885
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.403019210020263e-05,
+      "loss": 3.0319,
+      "step": 27890
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.401908438906645e-05,
+      "loss": 3.0729,
+      "step": 27895
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.40079759269708e-05,
+      "loss": 3.0464,
+      "step": 27900
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.399686671451072e-05,
+      "loss": 2.9088,
+      "step": 27905
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.398575675228131e-05,
+      "loss": 2.962,
+      "step": 27910
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.397464604087766e-05,
+      "loss": 2.7798,
+      "step": 27915
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.396353458089497e-05,
+      "loss": 3.0646,
+      "step": 27920
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.395242237292845e-05,
+      "loss": 3.0001,
+      "step": 27925
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.394130941757331e-05,
+      "loss": 3.0565,
+      "step": 27930
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.393019571542488e-05,
+      "loss": 2.9796,
+      "step": 27935
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.391908126707847e-05,
+      "loss": 3.0481,
+      "step": 27940
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.390796607312944e-05,
+      "loss": 2.9596,
+      "step": 27945
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.389685013417321e-05,
+      "loss": 2.9615,
+      "step": 27950
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.388573345080519e-05,
+      "loss": 2.9875,
+      "step": 27955
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.387461602362094e-05,
+      "loss": 3.0436,
+      "step": 27960
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.386349785321591e-05,
+      "loss": 3.0751,
+      "step": 27965
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.385237894018572e-05,
+      "loss": 3.0681,
+      "step": 27970
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.384125928512594e-05,
+      "loss": 2.9691,
+      "step": 27975
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.383013888863223e-05,
+      "loss": 3.0771,
+      "step": 27980
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.381901775130029e-05,
+      "loss": 3.1335,
+      "step": 27985
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.380789587372582e-05,
+      "loss": 3.017,
+      "step": 27990
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.37967732565046e-05,
+      "loss": 2.8484,
+      "step": 27995
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.378564990023244e-05,
+      "loss": 2.9196,
+      "step": 28000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.377452580550516e-05,
+      "loss": 3.0142,
+      "step": 28005
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.376340097291868e-05,
+      "loss": 3.0744,
+      "step": 28010
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.375227540306889e-05,
+      "loss": 3.0397,
+      "step": 28015
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.374114909655177e-05,
+      "loss": 2.8648,
+      "step": 28020
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.37300220539633e-05,
+      "loss": 2.848,
+      "step": 28025
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.371889427589956e-05,
+      "loss": 2.9838,
+      "step": 28030
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.370776576295662e-05,
+      "loss": 3.1496,
+      "step": 28035
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.369663651573057e-05,
+      "loss": 3.1536,
+      "step": 28040
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.36855065348176e-05,
+      "loss": 3.0625,
+      "step": 28045
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.36743758208139e-05,
+      "loss": 2.9567,
+      "step": 28050
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.36632443743157e-05,
+      "loss": 3.1042,
+      "step": 28055
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.365211219591928e-05,
+      "loss": 3.1566,
+      "step": 28060
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.364097928622098e-05,
+      "loss": 3.0267,
+      "step": 28065
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.362984564581713e-05,
+      "loss": 3.0052,
+      "step": 28070
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.361871127530413e-05,
+      "loss": 2.9091,
+      "step": 28075
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.360757617527843e-05,
+      "loss": 3.1666,
+      "step": 28080
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.359644034633648e-05,
+      "loss": 3.0831,
+      "step": 28085
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.358530378907481e-05,
+      "loss": 3.0296,
+      "step": 28090
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.357416650408996e-05,
+      "loss": 2.9454,
+      "step": 28095
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.356302849197853e-05,
+      "loss": 3.0756,
+      "step": 28100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.355188975333712e-05,
+      "loss": 3.0178,
+      "step": 28105
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.354075028876245e-05,
+      "loss": 3.0003,
+      "step": 28110
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.35296100988512e-05,
+      "loss": 2.9028,
+      "step": 28115
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.35184691842001e-05,
+      "loss": 3.0261,
+      "step": 28120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.350732754540595e-05,
+      "loss": 2.9085,
+      "step": 28125
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.349618518306557e-05,
+      "loss": 2.9539,
+      "step": 28130
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.348504209777583e-05,
+      "loss": 2.9898,
+      "step": 28135
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.347389829013364e-05,
+      "loss": 3.0236,
+      "step": 28140
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.346275376073591e-05,
+      "loss": 2.973,
+      "step": 28145
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.345160851017965e-05,
+      "loss": 2.9637,
+      "step": 28150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.344046253906184e-05,
+      "loss": 3.1037,
+      "step": 28155
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.342931584797957e-05,
+      "loss": 2.9688,
+      "step": 28160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.341816843752994e-05,
+      "loss": 3.1286,
+      "step": 28165
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.340702030831003e-05,
+      "loss": 3.0708,
+      "step": 28170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.339587146091706e-05,
+      "loss": 3.0874,
+      "step": 28175
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.338472189594821e-05,
+      "loss": 3.026,
+      "step": 28180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.337357161400077e-05,
+      "loss": 3.0042,
+      "step": 28185
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.336242061567199e-05,
+      "loss": 3.0723,
+      "step": 28190
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.33512689015592e-05,
+      "loss": 3.0725,
+      "step": 28195
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.334011647225977e-05,
+      "loss": 3.06,
+      "step": 28200
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.33289633283711e-05,
+      "loss": 3.0581,
+      "step": 28205
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.331780947049061e-05,
+      "loss": 3.0322,
+      "step": 28210
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.330665489921584e-05,
+      "loss": 2.9948,
+      "step": 28215
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.329549961514423e-05,
+      "loss": 3.0758,
+      "step": 28220
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.328434361887337e-05,
+      "loss": 3.0341,
+      "step": 28225
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.327318691100085e-05,
+      "loss": 2.9027,
+      "step": 28230
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.32620294921243e-05,
+      "loss": 3.0023,
+      "step": 28235
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.32508713628414e-05,
+      "loss": 3.1171,
+      "step": 28240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.323971252374984e-05,
+      "loss": 3.0102,
+      "step": 28245
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.322855297544736e-05,
+      "loss": 2.9294,
+      "step": 28250
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.321739271853178e-05,
+      "loss": 3.0164,
+      "step": 28255
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.320623175360088e-05,
+      "loss": 3.007,
+      "step": 28260
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.319507008125254e-05,
+      "loss": 3.1245,
+      "step": 28265
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.318390770208465e-05,
+      "loss": 3.1999,
+      "step": 28270
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.317274461669514e-05,
+      "loss": 3.0789,
+      "step": 28275
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.316158082568197e-05,
+      "loss": 3.1208,
+      "step": 28280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.31504163296432e-05,
+      "loss": 2.8835,
+      "step": 28285
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.313925112917683e-05,
+      "loss": 2.986,
+      "step": 28290
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.312808522488095e-05,
+      "loss": 3.0978,
+      "step": 28295
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.311691861735372e-05,
+      "loss": 3.0211,
+      "step": 28300
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.310575130719326e-05,
+      "loss": 3.0577,
+      "step": 28305
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.30945832949978e-05,
+      "loss": 3.083,
+      "step": 28310
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.308341458136553e-05,
+      "loss": 3.0026,
+      "step": 28315
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.307224516689477e-05,
+      "loss": 3.1394,
+      "step": 28320
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.306107505218381e-05,
+      "loss": 3.0483,
+      "step": 28325
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.304990423783101e-05,
+      "loss": 3.0321,
+      "step": 28330
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.303873272443476e-05,
+      "loss": 3.0119,
+      "step": 28335
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.302756051259347e-05,
+      "loss": 3.0669,
+      "step": 28340
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.30163876029056e-05,
+      "loss": 3.0474,
+      "step": 28345
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.300521399596966e-05,
+      "loss": 3.114,
+      "step": 28350
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.299403969238417e-05,
+      "loss": 2.9039,
+      "step": 28355
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.298286469274772e-05,
+      "loss": 3.1017,
+      "step": 28360
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.297168899765893e-05,
+      "loss": 3.1151,
+      "step": 28365
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.296051260771643e-05,
+      "loss": 2.8708,
+      "step": 28370
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.294933552351888e-05,
+      "loss": 3.0849,
+      "step": 28375
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.293815774566505e-05,
+      "loss": 2.918,
+      "step": 28380
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.292697927475367e-05,
+      "loss": 2.9941,
+      "step": 28385
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.291580011138356e-05,
+      "loss": 3.1,
+      "step": 28390
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.290462025615352e-05,
+      "loss": 2.9578,
+      "step": 28395
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.289343970966245e-05,
+      "loss": 3.1104,
+      "step": 28400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.288225847250924e-05,
+      "loss": 3.1028,
+      "step": 28405
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.287107654529284e-05,
+      "loss": 2.9579,
+      "step": 28410
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.285989392861225e-05,
+      "loss": 3.1434,
+      "step": 28415
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.284871062306646e-05,
+      "loss": 3.0361,
+      "step": 28420
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.283752662925452e-05,
+      "loss": 2.9297,
+      "step": 28425
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.282634194777557e-05,
+      "loss": 2.9906,
+      "step": 28430
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.28151565792287e-05,
+      "loss": 3.0414,
+      "step": 28435
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.280397052421308e-05,
+      "loss": 3.0523,
+      "step": 28440
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.279278378332792e-05,
+      "loss": 3.0015,
+      "step": 28445
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.278159635717247e-05,
+      "loss": 2.9737,
+      "step": 28450
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.277040824634597e-05,
+      "loss": 3.0801,
+      "step": 28455
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.275921945144776e-05,
+      "loss": 3.0369,
+      "step": 28460
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.274802997307719e-05,
+      "loss": 3.1312,
+      "step": 28465
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.273683981183366e-05,
+      "loss": 2.9791,
+      "step": 28470
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.272564896831657e-05,
+      "loss": 2.9924,
+      "step": 28475
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.271445744312537e-05,
+      "loss": 3.0501,
+      "step": 28480
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.270326523685956e-05,
+      "loss": 3.0555,
+      "step": 28485
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.26920723501187e-05,
+      "loss": 2.9774,
+      "step": 28490
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.268087878350234e-05,
+      "loss": 2.8985,
+      "step": 28495
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.266968453761009e-05,
+      "loss": 2.9476,
+      "step": 28500
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.265848961304157e-05,
+      "loss": 2.9423,
+      "step": 28505
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.264729401039649e-05,
+      "loss": 3.0275,
+      "step": 28510
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.263609773027453e-05,
+      "loss": 3.1593,
+      "step": 28515
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.262490077327548e-05,
+      "loss": 2.9682,
+      "step": 28520
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.26137031399991e-05,
+      "loss": 2.9518,
+      "step": 28525
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.260250483104522e-05,
+      "loss": 2.9455,
+      "step": 28530
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.259130584701369e-05,
+      "loss": 3.0985,
+      "step": 28535
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.258010618850443e-05,
+      "loss": 3.0929,
+      "step": 28540
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.256890585611736e-05,
+      "loss": 2.9774,
+      "step": 28545
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.255770485045242e-05,
+      "loss": 3.0557,
+      "step": 28550
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.254650317210964e-05,
+      "loss": 3.0127,
+      "step": 28555
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.253530082168908e-05,
+      "loss": 2.8293,
+      "step": 28560
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.252409779979077e-05,
+      "loss": 2.9684,
+      "step": 28565
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.251289410701487e-05,
+      "loss": 3.0589,
+      "step": 28570
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.250168974396147e-05,
+      "loss": 3.057,
+      "step": 28575
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.24904847112308e-05,
+      "loss": 3.0239,
+      "step": 28580
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.247927900942305e-05,
+      "loss": 3.0923,
+      "step": 28585
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.246807263913849e-05,
+      "loss": 2.9137,
+      "step": 28590
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.245686560097742e-05,
+      "loss": 3.0841,
+      "step": 28595
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.244565789554016e-05,
+      "loss": 3.0691,
+      "step": 28600
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.243444952342707e-05,
+      "loss": 3.0029,
+      "step": 28605
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.242324048523854e-05,
+      "loss": 2.9777,
+      "step": 28610
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.241203078157503e-05,
+      "loss": 3.1058,
+      "step": 28615
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.240082041303697e-05,
+      "loss": 3.1003,
+      "step": 28620
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.238960938022489e-05,
+      "loss": 3.0842,
+      "step": 28625
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.237839768373932e-05,
+      "loss": 3.0623,
+      "step": 28630
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.236718532418085e-05,
+      "loss": 2.9281,
+      "step": 28635
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.235597230215008e-05,
+      "loss": 2.9577,
+      "step": 28640
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.234475861824769e-05,
+      "loss": 3.0664,
+      "step": 28645
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.233354427307431e-05,
+      "loss": 2.9434,
+      "step": 28650
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.232232926723069e-05,
+      "loss": 3.1107,
+      "step": 28655
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.231111360131757e-05,
+      "loss": 3.0046,
+      "step": 28660
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.229989727593574e-05,
+      "loss": 2.8551,
+      "step": 28665
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.228868029168606e-05,
+      "loss": 2.9794,
+      "step": 28670
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.227746264916934e-05,
+      "loss": 3.0407,
+      "step": 28675
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.22662443489865e-05,
+      "loss": 2.9904,
+      "step": 28680
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.225502539173845e-05,
+      "loss": 2.9354,
+      "step": 28685
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.224380577802618e-05,
+      "loss": 3.1048,
+      "step": 28690
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.22325855084507e-05,
+      "loss": 3.0118,
+      "step": 28695
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.222136458361299e-05,
+      "loss": 3.0694,
+      "step": 28700
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.221014300411417e-05,
+      "loss": 3.1428,
+      "step": 28705
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.219892077055533e-05,
+      "loss": 3.1058,
+      "step": 28710
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.218769788353761e-05,
+      "loss": 3.1137,
+      "step": 28715
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.21764743436622e-05,
+      "loss": 3.0034,
+      "step": 28720
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.21652501515303e-05,
+      "loss": 2.997,
+      "step": 28725
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.215402530774315e-05,
+      "loss": 2.9709,
+      "step": 28730
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.214279981290201e-05,
+      "loss": 3.1166,
+      "step": 28735
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.213157366760825e-05,
+      "loss": 3.0002,
+      "step": 28740
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.21203468724632e-05,
+      "loss": 2.9363,
+      "step": 28745
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.210911942806822e-05,
+      "loss": 3.2578,
+      "step": 28750
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.209789133502474e-05,
+      "loss": 2.9895,
+      "step": 28755
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.208666259393423e-05,
+      "loss": 2.9223,
+      "step": 28760
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.207543320539815e-05,
+      "loss": 3.148,
+      "step": 28765
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.206420317001806e-05,
+      "loss": 2.9902,
+      "step": 28770
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.205297248839553e-05,
+      "loss": 3.1232,
+      "step": 28775
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.204174116113209e-05,
+      "loss": 3.2072,
+      "step": 28780
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.203050918882941e-05,
+      "loss": 3.0867,
+      "step": 28785
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.201927657208914e-05,
+      "loss": 3.0456,
+      "step": 28790
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.200804331151299e-05,
+      "loss": 2.9925,
+      "step": 28795
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.19968094077027e-05,
+      "loss": 3.0859,
+      "step": 28800
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.198557486126e-05,
+      "loss": 3.0953,
+      "step": 28805
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.197433967278672e-05,
+      "loss": 3.0479,
+      "step": 28810
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.196310384288468e-05,
+      "loss": 3.0414,
+      "step": 28815
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.195186737215575e-05,
+      "loss": 2.8716,
+      "step": 28820
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.194063026120185e-05,
+      "loss": 3.0113,
+      "step": 28825
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.19293925106249e-05,
+      "loss": 2.8594,
+      "step": 28830
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.191815412102687e-05,
+      "loss": 3.0377,
+      "step": 28835
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.190691509300978e-05,
+      "loss": 3.0124,
+      "step": 28840
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.189567542717566e-05,
+      "loss": 2.971,
+      "step": 28845
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.188443512412661e-05,
+      "loss": 3.008,
+      "step": 28850
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.187319418446469e-05,
+      "loss": 2.9325,
+      "step": 28855
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.186195260879207e-05,
+      "loss": 3.0228,
+      "step": 28860
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.185071039771092e-05,
+      "loss": 2.9465,
+      "step": 28865
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.183946755182349e-05,
+      "loss": 2.9793,
+      "step": 28870
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.182822407173197e-05,
+      "loss": 2.99,
+      "step": 28875
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.181697995803867e-05,
+      "loss": 2.9504,
+      "step": 28880
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.180573521134589e-05,
+      "loss": 2.9587,
+      "step": 28885
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.179448983225597e-05,
+      "loss": 3.0528,
+      "step": 28890
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.178324382137131e-05,
+      "loss": 2.9375,
+      "step": 28895
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.177199717929431e-05,
+      "loss": 2.9785,
+      "step": 28900
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.176074990662744e-05,
+      "loss": 3.0618,
+      "step": 28905
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.174950200397316e-05,
+      "loss": 3.067,
+      "step": 28910
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.173825347193398e-05,
+      "loss": 3.0343,
+      "step": 28915
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.172700431111246e-05,
+      "loss": 3.1328,
+      "step": 28920
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.171575452211119e-05,
+      "loss": 3.052,
+      "step": 28925
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.170450410553278e-05,
+      "loss": 2.9882,
+      "step": 28930
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.169325306197988e-05,
+      "loss": 2.907,
+      "step": 28935
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.168200139205518e-05,
+      "loss": 3.0498,
+      "step": 28940
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.16707490963614e-05,
+      "loss": 3.0009,
+      "step": 28945
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.165949617550128e-05,
+      "loss": 2.9083,
+      "step": 28950
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.16482426300776e-05,
+      "loss": 3.036,
+      "step": 28955
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.163698846069318e-05,
+      "loss": 2.9795,
+      "step": 28960
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.162573366795089e-05,
+      "loss": 3.0286,
+      "step": 28965
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.16144782524536e-05,
+      "loss": 2.9536,
+      "step": 28970
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.160322221480424e-05,
+      "loss": 2.9552,
+      "step": 28975
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.159196555560575e-05,
+      "loss": 3.1266,
+      "step": 28980
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.158070827546111e-05,
+      "loss": 3.0008,
+      "step": 28985
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.156945037497334e-05,
+      "loss": 3.0002,
+      "step": 28990
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.155819185474549e-05,
+      "loss": 3.0589,
+      "step": 28995
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.154693271538068e-05,
+      "loss": 2.9602,
+      "step": 29000
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.153567295748199e-05,
+      "loss": 2.8897,
+      "step": 29005
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.152441258165257e-05,
+      "loss": 3.0128,
+      "step": 29010
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.151315158849561e-05,
+      "loss": 3.0807,
+      "step": 29015
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.150188997861433e-05,
+      "loss": 2.9474,
+      "step": 29020
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.1490627752612e-05,
+      "loss": 3.048,
+      "step": 29025
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.147936491109186e-05,
+      "loss": 2.9677,
+      "step": 29030
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.146810145465727e-05,
+      "loss": 3.0188,
+      "step": 29035
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.145683738391154e-05,
+      "loss": 3.0603,
+      "step": 29040
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.144557269945808e-05,
+      "loss": 3.0583,
+      "step": 29045
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.143430740190029e-05,
+      "loss": 3.0262,
+      "step": 29050
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.142304149184163e-05,
+      "loss": 2.8707,
+      "step": 29055
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.141177496988557e-05,
+      "loss": 2.9983,
+      "step": 29060
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.140050783663562e-05,
+      "loss": 3.0971,
+      "step": 29065
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.138924009269534e-05,
+      "loss": 3.0247,
+      "step": 29070
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.13779717386683e-05,
+      "loss": 3.0534,
+      "step": 29075
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.13667027751581e-05,
+      "loss": 3.2237,
+      "step": 29080
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.13554332027684e-05,
+      "loss": 2.8973,
+      "step": 29085
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.134416302210287e-05,
+      "loss": 3.0261,
+      "step": 29090
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.133289223376523e-05,
+      "loss": 3.0594,
+      "step": 29095
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.13216208383592e-05,
+      "loss": 2.9996,
+      "step": 29100
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.131034883648858e-05,
+      "loss": 2.9181,
+      "step": 29105
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.129907622875716e-05,
+      "loss": 3.0987,
+      "step": 29110
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.128780301576879e-05,
+      "loss": 3.0393,
+      "step": 29115
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.127652919812732e-05,
+      "loss": 3.1046,
+      "step": 29120
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.126525477643666e-05,
+      "loss": 2.8816,
+      "step": 29125
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.125397975130078e-05,
+      "loss": 3.0162,
+      "step": 29130
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.124270412332362e-05,
+      "loss": 2.9294,
+      "step": 29135
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.123142789310917e-05,
+      "loss": 3.0204,
+      "step": 29140
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.122015106126148e-05,
+      "loss": 3.0092,
+      "step": 29145
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.120887362838461e-05,
+      "loss": 3.0292,
+      "step": 29150
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.119759559508268e-05,
+      "loss": 3.0173,
+      "step": 29155
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.118631696195978e-05,
+      "loss": 2.9599,
+      "step": 29160
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.117503772962008e-05,
+      "loss": 3.0545,
+      "step": 29165
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.116375789866781e-05,
+      "loss": 2.8344,
+      "step": 29170
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.115247746970716e-05,
+      "loss": 3.0289,
+      "step": 29175
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.114119644334242e-05,
+      "loss": 2.9514,
+      "step": 29180
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.112991482017784e-05,
+      "loss": 3.087,
+      "step": 29185
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.111863260081776e-05,
+      "loss": 3.0111,
+      "step": 29190
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.110734978586654e-05,
+      "loss": 2.9474,
+      "step": 29195
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.109606637592856e-05,
+      "loss": 3.0433,
+      "step": 29200
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.108478237160825e-05,
+      "loss": 3.086,
+      "step": 29205
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.107349777351005e-05,
+      "loss": 3.0437,
+      "step": 29210
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.106221258223843e-05,
+      "loss": 3.0455,
+      "step": 29215
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.105092679839791e-05,
+      "loss": 3.0018,
+      "step": 29220
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.103964042259305e-05,
+      "loss": 2.9386,
+      "step": 29225
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.102835345542841e-05,
+      "loss": 3.001,
+      "step": 29230
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.101706589750861e-05,
+      "loss": 3.0721,
+      "step": 29235
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.1005777749438274e-05,
+      "loss": 2.9659,
+      "step": 29240
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0994489011822076e-05,
+      "loss": 2.9924,
+      "step": 29245
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.098319968526476e-05,
+      "loss": 3.0663,
+      "step": 29250
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.097190977037099e-05,
+      "loss": 2.9326,
+      "step": 29255
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.096061926774559e-05,
+      "loss": 3.1044,
+      "step": 29260
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.094932817799333e-05,
+      "loss": 3.1136,
+      "step": 29265
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0938036501719045e-05,
+      "loss": 3.0711,
+      "step": 29270
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.09267442395276e-05,
+      "loss": 3.1616,
+      "step": 29275
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.091545139202389e-05,
+      "loss": 2.9911,
+      "step": 29280
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0904157959812824e-05,
+      "loss": 2.9272,
+      "step": 29285
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0892863943499354e-05,
+      "loss": 2.9555,
+      "step": 29290
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0881569343688496e-05,
+      "loss": 3.0822,
+      "step": 29295
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.087027416098523e-05,
+      "loss": 2.9706,
+      "step": 29300
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0858978395994646e-05,
+      "loss": 3.0351,
+      "step": 29305
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.084768204932179e-05,
+      "loss": 3.1402,
+      "step": 29310
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.083638512157177e-05,
+      "loss": 3.0638,
+      "step": 29315
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.082508761334975e-05,
+      "loss": 2.8763,
+      "step": 29320
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.081378952526089e-05,
+      "loss": 3.0142,
+      "step": 29325
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.08024908579104e-05,
+      "loss": 2.9796,
+      "step": 29330
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.079119161190351e-05,
+      "loss": 2.8704,
+      "step": 29335
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.077989178784549e-05,
+      "loss": 3.0416,
+      "step": 29340
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0768591386341636e-05,
+      "loss": 3.1375,
+      "step": 29345
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.075729040799727e-05,
+      "loss": 3.1011,
+      "step": 29350
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.074598885341777e-05,
+      "loss": 3.0041,
+      "step": 29355
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.07346867232085e-05,
+      "loss": 3.0722,
+      "step": 29360
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.072338401797489e-05,
+      "loss": 2.8835,
+      "step": 29365
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.07120807383224e-05,
+      "loss": 2.9879,
+      "step": 29370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.070077688485649e-05,
+      "loss": 3.1136,
+      "step": 29375
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.068947245818272e-05,
+      "loss": 2.9646,
+      "step": 29380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0678167458906575e-05,
+      "loss": 3.0331,
+      "step": 29385
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0666861887633676e-05,
+      "loss": 3.1039,
+      "step": 29390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0655555744969596e-05,
+      "loss": 3.1802,
+      "step": 29395
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0644249031519985e-05,
+      "loss": 3.009,
+      "step": 29400
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.063294174789052e-05,
+      "loss": 2.9857,
+      "step": 29405
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.062163389468687e-05,
+      "loss": 3.0603,
+      "step": 29410
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0610325472514775e-05,
+      "loss": 2.9659,
+      "step": 29415
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0599016481979986e-05,
+      "loss": 2.9834,
+      "step": 29420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.058770692368832e-05,
+      "loss": 2.9762,
+      "step": 29425
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.057639679824556e-05,
+      "loss": 2.9527,
+      "step": 29430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.056508610625758e-05,
+      "loss": 3.011,
+      "step": 29435
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.055377484833025e-05,
+      "loss": 3.0107,
+      "step": 29440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0542463025069475e-05,
+      "loss": 3.0594,
+      "step": 29445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.053115063708119e-05,
+      "loss": 3.1415,
+      "step": 29450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0519837684971384e-05,
+      "loss": 3.1448,
+      "step": 29455
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.050852416934605e-05,
+      "loss": 2.9815,
+      "step": 29460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0497210090811197e-05,
+      "loss": 2.8913,
+      "step": 29465
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.048589544997291e-05,
+      "loss": 2.9906,
+      "step": 29470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0474580247437274e-05,
+      "loss": 2.935,
+      "step": 29475
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.046326448381041e-05,
+      "loss": 2.9148,
+      "step": 29480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.045194815969848e-05,
+      "loss": 3.0662,
+      "step": 29485
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0440631275707635e-05,
+      "loss": 3.0625,
+      "step": 29490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.042931383244411e-05,
+      "loss": 3.1033,
+      "step": 29495
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.041799583051413e-05,
+      "loss": 3.013,
+      "step": 29500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.040667727052398e-05,
+      "loss": 3.1619,
+      "step": 29505
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0395358153079975e-05,
+      "loss": 3.0157,
+      "step": 29510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.038403847878841e-05,
+      "loss": 3.0018,
+      "step": 29515
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0372718248255655e-05,
+      "loss": 3.0781,
+      "step": 29520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.036139746208811e-05,
+      "loss": 3.1222,
+      "step": 29525
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.03500761208922e-05,
+      "loss": 2.9852,
+      "step": 29530
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0338754225274355e-05,
+      "loss": 2.9446,
+      "step": 29535
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.032743177584107e-05,
+      "loss": 3.0131,
+      "step": 29540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0316108773198846e-05,
+      "loss": 2.965,
+      "step": 29545
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.0304785217954227e-05,
+      "loss": 3.009,
+      "step": 29550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.029346111071379e-05,
+      "loss": 3.0284,
+      "step": 29555
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.028213645208411e-05,
+      "loss": 2.9751,
+      "step": 29560
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.027081124267183e-05,
+      "loss": 2.8563,
+      "step": 29565
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.025948548308361e-05,
+      "loss": 3.0101,
+      "step": 29570
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.0248159173926125e-05,
+      "loss": 3.0035,
+      "step": 29575
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.023683231580608e-05,
+      "loss": 3.0044,
+      "step": 29580
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.022550490933027e-05,
+      "loss": 2.9962,
+      "step": 29585
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.0214176955105405e-05,
+      "loss": 3.119,
+      "step": 29590
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.020284845373833e-05,
+      "loss": 3.0059,
+      "step": 29595
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.019151940583586e-05,
+      "loss": 3.09,
+      "step": 29600
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.018018981200486e-05,
+      "loss": 3.1404,
+      "step": 29605
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.016885967285225e-05,
+      "loss": 2.9325,
+      "step": 29610
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.015752898898491e-05,
+      "loss": 2.9374,
+      "step": 29615
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.01461977610098e-05,
+      "loss": 3.0584,
+      "step": 29620
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.013486598953393e-05,
+      "loss": 2.8898,
+      "step": 29625
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.0123533675164255e-05,
+      "loss": 3.0208,
+      "step": 29630
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.011220081850787e-05,
+      "loss": 3.2667,
+      "step": 29635
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.01008674201718e-05,
+      "loss": 3.0851,
+      "step": 29640
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.008953348076316e-05,
+      "loss": 3.0528,
+      "step": 29645
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.007819900088906e-05,
+      "loss": 3.0161,
+      "step": 29650
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.006686398115666e-05,
+      "loss": 2.9313,
+      "step": 29655
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.005552842217316e-05,
+      "loss": 3.0052,
+      "step": 29660
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.0044192324545744e-05,
+      "loss": 2.9532,
+      "step": 29665
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.003285568888166e-05,
+      "loss": 3.1209,
+      "step": 29670
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.002151851578819e-05,
+      "loss": 2.9819,
+      "step": 29675
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.001018080587262e-05,
+      "loss": 2.96,
+      "step": 29680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9998842559742276e-05,
+      "loss": 2.8837,
+      "step": 29685
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9987503778004515e-05,
+      "loss": 3.0237,
+      "step": 29690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9976164461266725e-05,
+      "loss": 3.0665,
+      "step": 29695
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.996482461013631e-05,
+      "loss": 2.8544,
+      "step": 29700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.995348422522071e-05,
+      "loss": 2.967,
+      "step": 29705
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.994214330712743e-05,
+      "loss": 3.1255,
+      "step": 29710
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9930801856463915e-05,
+      "loss": 2.956,
+      "step": 29715
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9919459873837714e-05,
+      "loss": 3.07,
+      "step": 29720
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9908117359856394e-05,
+      "loss": 2.9923,
+      "step": 29725
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.989677431512752e-05,
+      "loss": 3.0924,
+      "step": 29730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.988543074025873e-05,
+      "loss": 2.9338,
+      "step": 29735
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9874086635857644e-05,
+      "loss": 2.9208,
+      "step": 29740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9862742002531924e-05,
+      "loss": 3.15,
+      "step": 29745
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9851396840889285e-05,
+      "loss": 2.9809,
+      "step": 29750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9840051151537436e-05,
+      "loss": 3.078,
+      "step": 29755
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9828704935084146e-05,
+      "loss": 2.9205,
+      "step": 29760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9817358192137206e-05,
+      "loss": 2.9324,
+      "step": 29765
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.98060109233044e-05,
+      "loss": 3.0692,
+      "step": 29770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.979466312919358e-05,
+      "loss": 3.1072,
+      "step": 29775
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9783314810412615e-05,
+      "loss": 2.8993,
+      "step": 29780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.977196596756939e-05,
+      "loss": 3.129,
+      "step": 29785
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.976061660127185e-05,
+      "loss": 2.9539,
+      "step": 29790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9749266712127915e-05,
+      "loss": 2.9264,
+      "step": 29795
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.973791630074558e-05,
+      "loss": 2.9359,
+      "step": 29800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9726565367732845e-05,
+      "loss": 3.0185,
+      "step": 29805
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9715213913697757e-05,
+      "loss": 3.0232,
+      "step": 29810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9703861939248383e-05,
+      "loss": 2.9956,
+      "step": 29815
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.969250944499279e-05,
+      "loss": 2.9561,
+      "step": 29820
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.968115643153911e-05,
+      "loss": 3.0261,
+      "step": 29825
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9669802899495486e-05,
+      "loss": 3.0652,
+      "step": 29830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9658448849470094e-05,
+      "loss": 3.0626,
+      "step": 29835
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.964709428207114e-05,
+      "loss": 3.0632,
+      "step": 29840
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.963573919790684e-05,
+      "loss": 2.9889,
+      "step": 29845
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.962438359758546e-05,
+      "loss": 2.8978,
+      "step": 29850
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.961302748171531e-05,
+      "loss": 3.1549,
+      "step": 29855
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9601670850904654e-05,
+      "loss": 2.9877,
+      "step": 29860
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.959031370576186e-05,
+      "loss": 3.0025,
+      "step": 29865
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9578956046895294e-05,
+      "loss": 3.1018,
+      "step": 29870
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.956759787491335e-05,
+      "loss": 3.0571,
+      "step": 29875
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9556239190424436e-05,
+      "loss": 3.0888,
+      "step": 29880
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.954487999403703e-05,
+      "loss": 3.1283,
+      "step": 29885
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.953352028635959e-05,
+      "loss": 2.9917,
+      "step": 29890
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.952216006800062e-05,
+      "loss": 2.8479,
+      "step": 29895
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.951079933956867e-05,
+      "loss": 3.1385,
+      "step": 29900
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9499438101672264e-05,
+      "loss": 3.1843,
+      "step": 29905
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.948807635492003e-05,
+      "loss": 3.0592,
+      "step": 29910
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.947671409992057e-05,
+      "loss": 2.9151,
+      "step": 29915
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.94653513372825e-05,
+      "loss": 3.0708,
+      "step": 29920
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.945398806761451e-05,
+      "loss": 2.958,
+      "step": 29925
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.944262429152529e-05,
+      "loss": 2.9117,
+      "step": 29930
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.943126000962358e-05,
+      "loss": 3.031,
+      "step": 29935
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.941989522251812e-05,
+      "loss": 2.937,
+      "step": 29940
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9408529930817657e-05,
+      "loss": 3.0351,
+      "step": 29945
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.939716413513103e-05,
+      "loss": 2.9787,
+      "step": 29950
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.938579783606706e-05,
+      "loss": 2.993,
+      "step": 29955
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9374431034234615e-05,
+      "loss": 2.9715,
+      "step": 29960
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.936306373024256e-05,
+      "loss": 2.9966,
+      "step": 29965
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.935169592469981e-05,
+      "loss": 3.0191,
+      "step": 29970
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.934032761821532e-05,
+      "loss": 3.1671,
+      "step": 29975
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.932895881139804e-05,
+      "loss": 3.0668,
+      "step": 29980
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.931758950485697e-05,
+      "loss": 3.1031,
+      "step": 29985
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.930621969920114e-05,
+      "loss": 2.9812,
+      "step": 29990
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9294849395039566e-05,
+      "loss": 2.9953,
+      "step": 29995
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9283478592981335e-05,
+      "loss": 3.0271,
+      "step": 30000
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.927210729363555e-05,
+      "loss": 3.1123,
+      "step": 30005
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9260735497611334e-05,
+      "loss": 3.166,
+      "step": 30010
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.924936320551785e-05,
+      "loss": 3.1239,
+      "step": 30015
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.923799041796425e-05,
+      "loss": 3.0866,
+      "step": 30020
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.922661713555977e-05,
+      "loss": 3.0064,
+      "step": 30025
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9215243358913616e-05,
+      "loss": 2.8963,
+      "step": 30030
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9203869088635065e-05,
+      "loss": 3.0246,
+      "step": 30035
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9192494325333405e-05,
+      "loss": 3.0501,
+      "step": 30040
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.918111906961793e-05,
+      "loss": 3.053,
+      "step": 30045
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.916974332209798e-05,
+      "loss": 3.1609,
+      "step": 30050
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.915836708338293e-05,
+      "loss": 2.9026,
+      "step": 30055
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9146990354082155e-05,
+      "loss": 2.9792,
+      "step": 30060
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.91356131348051e-05,
+      "loss": 3.0112,
+      "step": 30065
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.912423542616117e-05,
+      "loss": 2.9676,
+      "step": 30070
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9112857228759865e-05,
+      "loss": 2.979,
+      "step": 30075
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.910147854321066e-05,
+      "loss": 3.161,
+      "step": 30080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9090099370123084e-05,
+      "loss": 3.0539,
+      "step": 30085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.90787197101067e-05,
+      "loss": 3.0183,
+      "step": 30090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9067339563771065e-05,
+      "loss": 3.0218,
+      "step": 30095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.905595893172576e-05,
+      "loss": 3.0895,
+      "step": 30100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.904457781458045e-05,
+      "loss": 2.9884,
+      "step": 30105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.903319621294475e-05,
+      "loss": 2.9994,
+      "step": 30110
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9021814127428365e-05,
+      "loss": 3.123,
+      "step": 30115
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.9010431558641e-05,
+      "loss": 3.0046,
+      "step": 30120
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.8999048507192365e-05,
+      "loss": 3.0052,
+      "step": 30125
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.8987664973692216e-05,
+      "loss": 2.9847,
+      "step": 30130
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.897628095875035e-05,
+      "loss": 3.1643,
+      "step": 30135
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.896489646297656e-05,
+      "loss": 3.1843,
+      "step": 30140
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.895351148698068e-05,
+      "loss": 2.8822,
+      "step": 30145
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.894212603137258e-05,
+      "loss": 3.0233,
+      "step": 30150
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.893074009676215e-05,
+      "loss": 2.9222,
+      "step": 30155
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.891935368375927e-05,
+      "loss": 3.0278,
+      "step": 30160
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.890796679297389e-05,
+      "loss": 2.9708,
+      "step": 30165
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.8896579425015976e-05,
+      "loss": 2.9799,
+      "step": 30170
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.88851915804955e-05,
+      "loss": 2.9482,
+      "step": 30175
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.88738032600225e-05,
+      "loss": 2.9377,
+      "step": 30180
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.8862414464206994e-05,
+      "loss": 3.1085,
+      "step": 30185
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.885102519365905e-05,
+      "loss": 3.0685,
+      "step": 30190
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.883963544898876e-05,
+      "loss": 3.0152,
+      "step": 30195
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.8828245230806236e-05,
+      "loss": 2.9439,
+      "step": 30200
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.88168545397216e-05,
+      "loss": 3.0213,
+      "step": 30205
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.8805463376345036e-05,
+      "loss": 3.1261,
+      "step": 30210
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.8794071741286726e-05,
+      "loss": 2.808,
+      "step": 30215
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.87826796351569e-05,
+      "loss": 3.1083,
+      "step": 30220
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.8771287058565784e-05,
+      "loss": 2.9697,
+      "step": 30225
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.875989401212364e-05,
+      "loss": 2.9953,
+      "step": 30230
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.874850049644076e-05,
+      "loss": 2.8994,
+      "step": 30235
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.873710651212747e-05,
+      "loss": 2.8671,
+      "step": 30240
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 5.8725712059794114e-05,
+      "loss": 3.1031,
+      "step": 30245
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.871431714005103e-05,
+      "loss": 2.88,
+      "step": 30250
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8702921753508645e-05,
+      "loss": 3.0833,
+      "step": 30255
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8691525900777344e-05,
+      "loss": 2.9014,
+      "step": 30260
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8680129582467593e-05,
+      "loss": 2.9771,
+      "step": 30265
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8668732799189854e-05,
+      "loss": 3.0075,
+      "step": 30270
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.865733555155459e-05,
+      "loss": 2.8986,
+      "step": 30275
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.864593784017235e-05,
+      "loss": 3.0695,
+      "step": 30280
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.863453966565365e-05,
+      "loss": 2.9052,
+      "step": 30285
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.862314102860908e-05,
+      "loss": 2.9154,
+      "step": 30290
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.861174192964922e-05,
+      "loss": 2.9125,
+      "step": 30295
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8600342369384675e-05,
+      "loss": 2.9375,
+      "step": 30300
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.858894234842609e-05,
+      "loss": 2.9337,
+      "step": 30305
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.857754186738413e-05,
+      "loss": 3.0034,
+      "step": 30310
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.856614092686949e-05,
+      "loss": 3.1396,
+      "step": 30315
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8554739527492895e-05,
+      "loss": 2.9741,
+      "step": 30320
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.854333766986505e-05,
+      "loss": 3.0809,
+      "step": 30325
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8531935354596733e-05,
+      "loss": 2.9735,
+      "step": 30330
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8520532582298726e-05,
+      "loss": 2.9977,
+      "step": 30335
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.850912935358186e-05,
+      "loss": 2.9679,
+      "step": 30340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8497725669056955e-05,
+      "loss": 3.0556,
+      "step": 30345
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.848632152933488e-05,
+      "loss": 2.9104,
+      "step": 30350
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.84749169350265e-05,
+      "loss": 3.027,
+      "step": 30355
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.846351188674275e-05,
+      "loss": 3.0751,
+      "step": 30360
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.845210638509454e-05,
+      "loss": 3.0003,
+      "step": 30365
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.844070043069286e-05,
+      "loss": 3.0017,
+      "step": 30370
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.842929402414866e-05,
+      "loss": 2.9124,
+      "step": 30375
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.841788716607296e-05,
+      "loss": 3.1825,
+      "step": 30380
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.840647985707679e-05,
+      "loss": 3.0773,
+      "step": 30385
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.83950720977712e-05,
+      "loss": 2.9788,
+      "step": 30390
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8383663888767284e-05,
+      "loss": 2.9592,
+      "step": 30395
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.837225523067612e-05,
+      "loss": 2.9518,
+      "step": 30400
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.836084612410885e-05,
+      "loss": 3.0003,
+      "step": 30405
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.834943656967663e-05,
+      "loss": 3.0852,
+      "step": 30410
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.833802656799062e-05,
+      "loss": 3.111,
+      "step": 30415
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8326616119662034e-05,
+      "loss": 3.0353,
+      "step": 30420
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.831520522530209e-05,
+      "loss": 2.8794,
+      "step": 30425
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8303793885522026e-05,
+      "loss": 2.9915,
+      "step": 30430
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.829238210093312e-05,
+      "loss": 3.0147,
+      "step": 30435
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.828096987214666e-05,
+      "loss": 3.0747,
+      "step": 30440
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8269557199773985e-05,
+      "loss": 3.028,
+      "step": 30445
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8258144084426414e-05,
+      "loss": 2.9851,
+      "step": 30450
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8246730526715324e-05,
+      "loss": 3.0459,
+      "step": 30455
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.82353165272521e-05,
+      "loss": 3.0435,
+      "step": 30460
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.822390208664815e-05,
+      "loss": 3.0227,
+      "step": 30465
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.821248720551492e-05,
+      "loss": 3.0836,
+      "step": 30470
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.820107188446386e-05,
+      "loss": 2.9707,
+      "step": 30475
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.818965612410647e-05,
+      "loss": 3.0436,
+      "step": 30480
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.817823992505425e-05,
+      "loss": 3.1509,
+      "step": 30485
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.816682328791871e-05,
+      "loss": 3.0019,
+      "step": 30490
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.815540621331145e-05,
+      "loss": 3.0271,
+      "step": 30495
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.814398870184401e-05,
+      "loss": 3.0377,
+      "step": 30500
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8132570754128e-05,
+      "loss": 3.0343,
+      "step": 30505
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.812115237077504e-05,
+      "loss": 3.0043,
+      "step": 30510
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8109733552396806e-05,
+      "loss": 3.018,
+      "step": 30515
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.809831429960493e-05,
+      "loss": 3.0575,
+      "step": 30520
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.808689461301113e-05,
+      "loss": 2.886,
+      "step": 30525
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.807547449322712e-05,
+      "loss": 3.0709,
+      "step": 30530
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.806405394086464e-05,
+      "loss": 2.9876,
+      "step": 30535
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8052632956535444e-05,
+      "loss": 2.9709,
+      "step": 30540
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8041211540851326e-05,
+      "loss": 2.9938,
+      "step": 30545
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.802978969442412e-05,
+      "loss": 2.9956,
+      "step": 30550
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.8018367417865615e-05,
+      "loss": 2.9789,
+      "step": 30555
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.80069447117877e-05,
+      "loss": 3.123,
+      "step": 30560
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.799552157680224e-05,
+      "loss": 2.9467,
+      "step": 30565
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.798409801352115e-05,
+      "loss": 3.1211,
+      "step": 30570
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.797267402255635e-05,
+      "loss": 2.9533,
+      "step": 30575
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7961249604519775e-05,
+      "loss": 3.0007,
+      "step": 30580
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.794982476002341e-05,
+      "loss": 3.0238,
+      "step": 30585
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.793839948967924e-05,
+      "loss": 3.0205,
+      "step": 30590
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.792697379409929e-05,
+      "loss": 3.0254,
+      "step": 30595
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.791554767389561e-05,
+      "loss": 2.9566,
+      "step": 30600
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7904121129680234e-05,
+      "loss": 3.1262,
+      "step": 30605
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.789269416206527e-05,
+      "loss": 3.1002,
+      "step": 30610
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7881266771662815e-05,
+      "loss": 2.9471,
+      "step": 30615
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7869838959085e-05,
+      "loss": 3.06,
+      "step": 30620
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7858410724944e-05,
+      "loss": 3.048,
+      "step": 30625
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.784698206985195e-05,
+      "loss": 3.0082,
+      "step": 30630
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.783555299442107e-05,
+      "loss": 3.118,
+      "step": 30635
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7824123499263585e-05,
+      "loss": 2.9972,
+      "step": 30640
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.781269358499173e-05,
+      "loss": 3.1355,
+      "step": 30645
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7801263252217786e-05,
+      "loss": 3.078,
+      "step": 30650
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7789832501554006e-05,
+      "loss": 2.9505,
+      "step": 30655
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.777840133361274e-05,
+      "loss": 3.1011,
+      "step": 30660
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.776696974900629e-05,
+      "loss": 2.9695,
+      "step": 30665
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.775553774834703e-05,
+      "loss": 3.1342,
+      "step": 30670
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7744105332247354e-05,
+      "loss": 3.0361,
+      "step": 30675
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7732672501319615e-05,
+      "loss": 2.8466,
+      "step": 30680
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7721239256176264e-05,
+      "loss": 3.0794,
+      "step": 30685
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7709805597429734e-05,
+      "loss": 3.1661,
+      "step": 30690
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.769837152569251e-05,
+      "loss": 2.9265,
+      "step": 30695
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.768693704157706e-05,
+      "loss": 2.9183,
+      "step": 30700
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.76755021456959e-05,
+      "loss": 3.0098,
+      "step": 30705
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.766406683866157e-05,
+      "loss": 3.0912,
+      "step": 30710
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.765263112108662e-05,
+      "loss": 2.9605,
+      "step": 30715
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7641194993583616e-05,
+      "loss": 3.1159,
+      "step": 30720
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.762975845676518e-05,
+      "loss": 3.0554,
+      "step": 30725
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.76183215112439e-05,
+      "loss": 2.9717,
+      "step": 30730
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.760688415763245e-05,
+      "loss": 3.0524,
+      "step": 30735
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.759544639654346e-05,
+      "loss": 3.0669,
+      "step": 30740
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.758400822858966e-05,
+      "loss": 3.0062,
+      "step": 30745
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.757256965438371e-05,
+      "loss": 3.1091,
+      "step": 30750
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.756113067453839e-05,
+      "loss": 3.0749,
+      "step": 30755
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.754969128966641e-05,
+      "loss": 2.9532,
+      "step": 30760
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.753825150038056e-05,
+      "loss": 2.8749,
+      "step": 30765
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.752681130729363e-05,
+      "loss": 2.9929,
+      "step": 30770
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7515370711018436e-05,
+      "loss": 3.1188,
+      "step": 30775
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7503929712167815e-05,
+      "loss": 3.0154,
+      "step": 30780
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.749248831135463e-05,
+      "loss": 2.9641,
+      "step": 30785
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.748104650919176e-05,
+      "loss": 3.0659,
+      "step": 30790
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.74696043062921e-05,
+      "loss": 2.9751,
+      "step": 30795
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.74581617032686e-05,
+      "loss": 2.9293,
+      "step": 30800
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.744671870073418e-05,
+      "loss": 2.889,
+      "step": 30805
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7435275299301816e-05,
+      "loss": 2.8779,
+      "step": 30810
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.742383149958449e-05,
+      "loss": 3.0492,
+      "step": 30815
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.741238730219522e-05,
+      "loss": 3.0065,
+      "step": 30820
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.740094270774702e-05,
+      "loss": 3.1227,
+      "step": 30825
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.738949771685297e-05,
+      "loss": 2.953,
+      "step": 30830
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.737805233012612e-05,
+      "loss": 3.0282,
+      "step": 30835
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.736660654817957e-05,
+      "loss": 2.992,
+      "step": 30840
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.735516037162643e-05,
+      "loss": 3.0333,
+      "step": 30845
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7343713801079845e-05,
+      "loss": 3.0309,
+      "step": 30850
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.733226683715299e-05,
+      "loss": 3.0879,
+      "step": 30855
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.732081948045901e-05,
+      "loss": 3.0278,
+      "step": 30860
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.730937173161112e-05,
+      "loss": 3.0425,
+      "step": 30865
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7297923591222535e-05,
+      "loss": 2.8369,
+      "step": 30870
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7286475059906506e-05,
+      "loss": 3.0228,
+      "step": 30875
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7275026138276313e-05,
+      "loss": 3.0346,
+      "step": 30880
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7263576826945194e-05,
+      "loss": 3.1244,
+      "step": 30885
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7252127126526476e-05,
+      "loss": 3.1653,
+      "step": 30890
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7240677037633504e-05,
+      "loss": 3.1198,
+      "step": 30895
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.72292265608796e-05,
+      "loss": 2.9685,
+      "step": 30900
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7217775696878154e-05,
+      "loss": 3.1061,
+      "step": 30905
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7206324446242524e-05,
+      "loss": 2.9328,
+      "step": 30910
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.719487280958612e-05,
+      "loss": 2.8975,
+      "step": 30915
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7183420787522405e-05,
+      "loss": 3.0955,
+      "step": 30920
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.7171968380664785e-05,
+      "loss": 2.9027,
+      "step": 30925
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.716051558962679e-05,
+      "loss": 2.9134,
+      "step": 30930
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.714906241502184e-05,
+      "loss": 3.1016,
+      "step": 30935
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.713760885746349e-05,
+      "loss": 2.7805,
+      "step": 30940
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.712615491756527e-05,
+      "loss": 2.861,
+      "step": 30945
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.7114700595940707e-05,
+      "loss": 2.9436,
+      "step": 30950
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.7103245893203414e-05,
+      "loss": 2.9348,
+      "step": 30955
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.709179080996696e-05,
+      "loss": 2.9529,
+      "step": 30960
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.708033534684495e-05,
+      "loss": 3.0112,
+      "step": 30965
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.7068879504451035e-05,
+      "loss": 2.9599,
+      "step": 30970
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.705742328339886e-05,
+      "loss": 3.051,
+      "step": 30975
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.704596668430211e-05,
+      "loss": 3.0207,
+      "step": 30980
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.703450970777447e-05,
+      "loss": 3.1242,
+      "step": 30985
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.702305235442966e-05,
+      "loss": 3.0119,
+      "step": 30990
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.701159462488142e-05,
+      "loss": 3.0088,
+      "step": 30995
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.700013651974348e-05,
+      "loss": 3.1143,
+      "step": 31000
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.698867803962966e-05,
+      "loss": 3.0633,
+      "step": 31005
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.697721918515372e-05,
+      "loss": 3.0412,
+      "step": 31010
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.696575995692948e-05,
+      "loss": 3.0258,
+      "step": 31015
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.695430035557079e-05,
+      "loss": 3.038,
+      "step": 31020
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.694284038169149e-05,
+      "loss": 2.8331,
+      "step": 31025
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6931380035905476e-05,
+      "loss": 3.1269,
+      "step": 31030
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.691991931882663e-05,
+      "loss": 2.8357,
+      "step": 31035
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.690845823106887e-05,
+      "loss": 3.0624,
+      "step": 31040
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.689699677324613e-05,
+      "loss": 3.0153,
+      "step": 31045
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6885534945972354e-05,
+      "loss": 2.9669,
+      "step": 31050
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.687407274986154e-05,
+      "loss": 3.0459,
+      "step": 31055
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.686261018552767e-05,
+      "loss": 2.9633,
+      "step": 31060
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.685114725358477e-05,
+      "loss": 2.9803,
+      "step": 31065
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.683968395464685e-05,
+      "loss": 3.0949,
+      "step": 31070
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.682822028932798e-05,
+      "loss": 3.0932,
+      "step": 31075
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.681675625824224e-05,
+      "loss": 2.9885,
+      "step": 31080
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6805291862003704e-05,
+      "loss": 2.9264,
+      "step": 31085
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.679382710122651e-05,
+      "loss": 2.9662,
+      "step": 31090
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.678236197652477e-05,
+      "loss": 3.057,
+      "step": 31095
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.677089648851264e-05,
+      "loss": 3.0911,
+      "step": 31100
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.67594306378043e-05,
+      "loss": 3.0909,
+      "step": 31105
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.674796442501392e-05,
+      "loss": 3.014,
+      "step": 31110
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.673649785075573e-05,
+      "loss": 3.0673,
+      "step": 31115
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.672503091564395e-05,
+      "loss": 2.9631,
+      "step": 31120
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.671356362029283e-05,
+      "loss": 2.9145,
+      "step": 31125
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6702095965316646e-05,
+      "loss": 3.0977,
+      "step": 31130
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.669062795132969e-05,
+      "loss": 2.9602,
+      "step": 31135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.667915957894624e-05,
+      "loss": 2.9579,
+      "step": 31140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.666769084878063e-05,
+      "loss": 3.0913,
+      "step": 31145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6656221761447226e-05,
+      "loss": 3.0526,
+      "step": 31150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6644752317560376e-05,
+      "loss": 3.0526,
+      "step": 31155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.663328251773448e-05,
+      "loss": 3.0093,
+      "step": 31160
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6621812362583905e-05,
+      "loss": 2.8871,
+      "step": 31165
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6610341852723106e-05,
+      "loss": 3.0501,
+      "step": 31170
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6598870988766505e-05,
+      "loss": 3.0652,
+      "step": 31175
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.658739977132856e-05,
+      "loss": 3.1035,
+      "step": 31180
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.657592820102377e-05,
+      "loss": 3.0068,
+      "step": 31185
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.656445627846662e-05,
+      "loss": 3.0069,
+      "step": 31190
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6552984004271616e-05,
+      "loss": 3.1242,
+      "step": 31195
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.654151137905329e-05,
+      "loss": 3.0396,
+      "step": 31200
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6530038403426214e-05,
+      "loss": 3.0447,
+      "step": 31205
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.651856507800496e-05,
+      "loss": 2.9342,
+      "step": 31210
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.65070914034041e-05,
+      "loss": 3.0205,
+      "step": 31215
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6495617380238255e-05,
+      "loss": 3.0814,
+      "step": 31220
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.648414300912206e-05,
+      "loss": 3.0355,
+      "step": 31225
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.647266829067014e-05,
+      "loss": 2.9431,
+      "step": 31230
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.64611932254972e-05,
+      "loss": 2.938,
+      "step": 31235
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6449717814217884e-05,
+      "loss": 3.0076,
+      "step": 31240
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.64382420574469e-05,
+      "loss": 3.0223,
+      "step": 31245
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.642676595579899e-05,
+      "loss": 3.0002,
+      "step": 31250
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.641528950988888e-05,
+      "loss": 3.1078,
+      "step": 31255
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6403812720331336e-05,
+      "loss": 2.7951,
+      "step": 31260
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.639233558774113e-05,
+      "loss": 2.8531,
+      "step": 31265
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.638085811273305e-05,
+      "loss": 3.0438,
+      "step": 31270
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.636938029592191e-05,
+      "loss": 2.9513,
+      "step": 31275
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.635790213792256e-05,
+      "loss": 2.9948,
+      "step": 31280
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.634642363934984e-05,
+      "loss": 3.042,
+      "step": 31285
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6334944800818615e-05,
+      "loss": 2.9686,
+      "step": 31290
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.632346562294376e-05,
+      "loss": 3.0409,
+      "step": 31295
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6311986106340196e-05,
+      "loss": 3.0481,
+      "step": 31300
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.630050625162283e-05,
+      "loss": 3.0375,
+      "step": 31305
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6289026059406636e-05,
+      "loss": 2.9139,
+      "step": 31310
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.627754553030653e-05,
+      "loss": 3.1551,
+      "step": 31315
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.626606466493752e-05,
+      "loss": 3.0785,
+      "step": 31320
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.625458346391458e-05,
+      "loss": 3.1638,
+      "step": 31325
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.624310192785274e-05,
+      "loss": 3.0,
+      "step": 31330
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.623162005736703e-05,
+      "loss": 3.0173,
+      "step": 31335
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.622013785307248e-05,
+      "loss": 3.1036,
+      "step": 31340
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.620865531558417e-05,
+      "loss": 3.0988,
+      "step": 31345
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.619717244551718e-05,
+      "loss": 3.096,
+      "step": 31350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.618568924348662e-05,
+      "loss": 3.1712,
+      "step": 31355
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.61742057101076e-05,
+      "loss": 2.9554,
+      "step": 31360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.616272184599527e-05,
+      "loss": 3.0272,
+      "step": 31365
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6151237651764775e-05,
+      "loss": 2.8984,
+      "step": 31370
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.613975312803129e-05,
+      "loss": 3.1488,
+      "step": 31375
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6128268275410013e-05,
+      "loss": 2.8665,
+      "step": 31380
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6116783094516135e-05,
+      "loss": 2.8901,
+      "step": 31385
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.61052975859649e-05,
+      "loss": 3.0384,
+      "step": 31390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.609381175037154e-05,
+      "loss": 3.019,
+      "step": 31395
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6082325588351327e-05,
+      "loss": 3.0898,
+      "step": 31400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.607083910051953e-05,
+      "loss": 2.9671,
+      "step": 31405
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6059352287491476e-05,
+      "loss": 2.9329,
+      "step": 31410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.604786514988242e-05,
+      "loss": 2.9917,
+      "step": 31415
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.603637768830773e-05,
+      "loss": 3.0032,
+      "step": 31420
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.602488990338275e-05,
+      "loss": 3.1196,
+      "step": 31425
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.601340179572285e-05,
+      "loss": 3.0066,
+      "step": 31430
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.6001913365943406e-05,
+      "loss": 3.1774,
+      "step": 31435
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.599042461465982e-05,
+      "loss": 2.9039,
+      "step": 31440
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.59789355424875e-05,
+      "loss": 2.9507,
+      "step": 31445
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.596744615004189e-05,
+      "loss": 3.0253,
+      "step": 31450
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.595595643793844e-05,
+      "loss": 2.8871,
+      "step": 31455
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5944466406792614e-05,
+      "loss": 3.0303,
+      "step": 31460
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5932976057219924e-05,
+      "loss": 2.9503,
+      "step": 31465
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.592148538983584e-05,
+      "loss": 3.0262,
+      "step": 31470
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.590999440525588e-05,
+      "loss": 3.006,
+      "step": 31475
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5898503104095614e-05,
+      "loss": 2.9659,
+      "step": 31480
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.588701148697056e-05,
+      "loss": 3.067,
+      "step": 31485
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5875519554496325e-05,
+      "loss": 3.0286,
+      "step": 31490
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5864027307288455e-05,
+      "loss": 3.0263,
+      "step": 31495
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.585253474596259e-05,
+      "loss": 3.1,
+      "step": 31500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.584104187113432e-05,
+      "loss": 3.1201,
+      "step": 31505
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.582954868341932e-05,
+      "loss": 2.8905,
+      "step": 31510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.581805518343323e-05,
+      "loss": 3.0559,
+      "step": 31515
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.580656137179171e-05,
+      "loss": 3.0275,
+      "step": 31520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5795067249110445e-05,
+      "loss": 2.9867,
+      "step": 31525
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.578357281600516e-05,
+      "loss": 3.0466,
+      "step": 31530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.577207807309156e-05,
+      "loss": 3.06,
+      "step": 31535
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5760583020985414e-05,
+      "loss": 2.962,
+      "step": 31540
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.574908766030242e-05,
+      "loss": 2.9574,
+      "step": 31545
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.57375919916584e-05,
+      "loss": 3.0962,
+      "step": 31550
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5726096015669124e-05,
+      "loss": 3.0438,
+      "step": 31555
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5714599732950404e-05,
+      "loss": 3.0013,
+      "step": 31560
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.570310314411804e-05,
+      "loss": 2.991,
+      "step": 31565
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5691606249787885e-05,
+      "loss": 3.0608,
+      "step": 31570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.56801090505758e-05,
+      "loss": 3.0217,
+      "step": 31575
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.566861154709764e-05,
+      "loss": 3.0529,
+      "step": 31580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.565711373996928e-05,
+      "loss": 3.0717,
+      "step": 31585
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.564561562980667e-05,
+      "loss": 3.0465,
+      "step": 31590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.563411721722567e-05,
+      "loss": 2.9554,
+      "step": 31595
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.562261850284225e-05,
+      "loss": 3.1117,
+      "step": 31600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.5611119487272344e-05,
+      "loss": 3.0113,
+      "step": 31605
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.559962017113194e-05,
+      "loss": 2.877,
+      "step": 31610
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.558812055503701e-05,
+      "loss": 2.8802,
+      "step": 31615
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.557662063960355e-05,
+      "loss": 3.0109,
+      "step": 31620
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.556512042544757e-05,
+      "loss": 3.0417,
+      "step": 31625
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.555361991318512e-05,
+      "loss": 2.8732,
+      "step": 31630
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.554211910343222e-05,
+      "loss": 3.124,
+      "step": 31635
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.553061799680497e-05,
+      "loss": 2.9667,
+      "step": 31640
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.5519116593919417e-05,
+      "loss": 2.9371,
+      "step": 31645
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.550761489539166e-05,
+      "loss": 3.085,
+      "step": 31650
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.549611290183783e-05,
+      "loss": 3.0246,
+      "step": 31655
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.548461061387403e-05,
+      "loss": 2.9711,
+      "step": 31660
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.547310803211642e-05,
+      "loss": 3.0176,
+      "step": 31665
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.546160515718115e-05,
+      "loss": 3.0065,
+      "step": 31670
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.54501019896844e-05,
+      "loss": 2.9486,
+      "step": 31675
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.543859853024235e-05,
+      "loss": 2.9755,
+      "step": 31680
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.54270947794712e-05,
+      "loss": 2.9502,
+      "step": 31685
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.541559073798719e-05,
+      "loss": 3.056,
+      "step": 31690
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.540408640640654e-05,
+      "loss": 2.8505,
+      "step": 31695
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.539258178534552e-05,
+      "loss": 2.9939,
+      "step": 31700
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.538107687542038e-05,
+      "loss": 2.9701,
+      "step": 31705
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.536957167724741e-05,
+      "loss": 2.9755,
+      "step": 31710
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.535806619144291e-05,
+      "loss": 3.0849,
+      "step": 31715
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.5346560418623185e-05,
+      "loss": 3.0601,
+      "step": 31720
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.533505435940457e-05,
+      "loss": 3.0142,
+      "step": 31725
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.5323548014403414e-05,
+      "loss": 2.9771,
+      "step": 31730
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.5312041384236067e-05,
+      "loss": 2.9346,
+      "step": 31735
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.5300534469518924e-05,
+      "loss": 2.8691,
+      "step": 31740
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.528902727086834e-05,
+      "loss": 3.0136,
+      "step": 31745
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.527751978890076e-05,
+      "loss": 3.0993,
+      "step": 31750
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.526601202423256e-05,
+      "loss": 2.9917,
+      "step": 31755
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.525450397748022e-05,
+      "loss": 3.0483,
+      "step": 31760
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.5242995649260165e-05,
+      "loss": 3.0675,
+      "step": 31765
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.523148704018888e-05,
+      "loss": 2.953,
+      "step": 31770
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.521997815088282e-05,
+      "loss": 3.0823,
+      "step": 31775
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.52084689819585e-05,
+      "loss": 3.0306,
+      "step": 31780
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.519695953403242e-05,
+      "loss": 2.9374,
+      "step": 31785
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.5185449807721114e-05,
+      "loss": 3.029,
+      "step": 31790
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.517393980364114e-05,
+      "loss": 3.0819,
+      "step": 31795
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.516242952240901e-05,
+      "loss": 2.976,
+      "step": 31800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.515091896464133e-05,
+      "loss": 2.8674,
+      "step": 31805
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.513940813095466e-05,
+      "loss": 3.045,
+      "step": 31810
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.512789702196563e-05,
+      "loss": 3.0184,
+      "step": 31815
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.511638563829084e-05,
+      "loss": 3.003,
+      "step": 31820
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.510487398054691e-05,
+      "loss": 3.1121,
+      "step": 31825
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.50933620493505e-05,
+      "loss": 3.0293,
+      "step": 31830
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.5081849845318253e-05,
+      "loss": 2.8717,
+      "step": 31835
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.507033736906686e-05,
+      "loss": 2.7364,
+      "step": 31840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.5058824621213e-05,
+      "loss": 3.0522,
+      "step": 31845
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.5047311602373377e-05,
+      "loss": 2.9712,
+      "step": 31850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.5035798313164696e-05,
+      "loss": 2.9567,
+      "step": 31855
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.502428475420372e-05,
+      "loss": 2.879,
+      "step": 31860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.501277092610716e-05,
+      "loss": 3.0641,
+      "step": 31865
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.500125682949181e-05,
+      "loss": 3.1044,
+      "step": 31870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4989742464974414e-05,
+      "loss": 2.9331,
+      "step": 31875
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.497822783317177e-05,
+      "loss": 3.0987,
+      "step": 31880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.496671293470069e-05,
+      "loss": 2.9496,
+      "step": 31885
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.495519777017799e-05,
+      "loss": 3.0437,
+      "step": 31890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4943682340220505e-05,
+      "loss": 2.8915,
+      "step": 31895
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.493216664544507e-05,
+      "loss": 3.0,
+      "step": 31900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4920650686468556e-05,
+      "loss": 2.8724,
+      "step": 31905
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.490913446390782e-05,
+      "loss": 3.0476,
+      "step": 31910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.489761797837978e-05,
+      "loss": 3.0051,
+      "step": 31915
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.488610123050133e-05,
+      "loss": 2.9452,
+      "step": 31920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4874584220889355e-05,
+      "loss": 2.9747,
+      "step": 31925
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4863066950160835e-05,
+      "loss": 3.0802,
+      "step": 31930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.485154941893268e-05,
+      "loss": 2.9534,
+      "step": 31935
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.484003162782185e-05,
+      "loss": 3.0431,
+      "step": 31940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.482851357744535e-05,
+      "loss": 2.9535,
+      "step": 31945
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.481699526842013e-05,
+      "loss": 3.0232,
+      "step": 31950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.480547670136321e-05,
+      "loss": 3.0147,
+      "step": 31955
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.47939578768916e-05,
+      "loss": 2.9708,
+      "step": 31960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4782438795622326e-05,
+      "loss": 2.9925,
+      "step": 31965
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4770919458172454e-05,
+      "loss": 3.0593,
+      "step": 31970
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.475939986515899e-05,
+      "loss": 2.9006,
+      "step": 31975
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.474788001719904e-05,
+      "loss": 2.9484,
+      "step": 31980
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4736359914909676e-05,
+      "loss": 3.0968,
+      "step": 31985
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.472483955890799e-05,
+      "loss": 3.0384,
+      "step": 31990
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.471331894981111e-05,
+      "loss": 3.0577,
+      "step": 31995
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.470179808823615e-05,
+      "loss": 3.0745,
+      "step": 32000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4690276974800225e-05,
+      "loss": 2.9737,
+      "step": 32005
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.467875561012052e-05,
+      "loss": 2.9111,
+      "step": 32010
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4667233994814204e-05,
+      "loss": 3.099,
+      "step": 32015
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4655712129498414e-05,
+      "loss": 3.0974,
+      "step": 32020
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4644190014790375e-05,
+      "loss": 2.9596,
+      "step": 32025
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.463266765130728e-05,
+      "loss": 2.9985,
+      "step": 32030
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4621145039666344e-05,
+      "loss": 3.0659,
+      "step": 32035
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.46096221804848e-05,
+      "loss": 3.0613,
+      "step": 32040
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.459809907437989e-05,
+      "loss": 3.0072,
+      "step": 32045
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4586575721968893e-05,
+      "loss": 2.9643,
+      "step": 32050
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.457505212386904e-05,
+      "loss": 2.8228,
+      "step": 32055
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.456352828069764e-05,
+      "loss": 2.8728,
+      "step": 32060
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.455200419307199e-05,
+      "loss": 3.025,
+      "step": 32065
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4540479861609415e-05,
+      "loss": 2.9367,
+      "step": 32070
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.452895528692721e-05,
+      "loss": 3.0401,
+      "step": 32075
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4517430469642706e-05,
+      "loss": 3.0807,
+      "step": 32080
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.450590541037328e-05,
+      "loss": 2.9506,
+      "step": 32085
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.449438010973628e-05,
+      "loss": 3.0206,
+      "step": 32090
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.448285456834907e-05,
+      "loss": 2.9961,
+      "step": 32095
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4471328786829076e-05,
+      "loss": 3.1578,
+      "step": 32100
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4459802765793653e-05,
+      "loss": 3.0504,
+      "step": 32105
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.444827650586024e-05,
+      "loss": 3.0557,
+      "step": 32110
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4436750007646265e-05,
+      "loss": 2.928,
+      "step": 32115
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.442522327176916e-05,
+      "loss": 3.1425,
+      "step": 32120
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.441369629884638e-05,
+      "loss": 2.9633,
+      "step": 32125
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.440216908949538e-05,
+      "loss": 2.9366,
+      "step": 32130
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.439064164433365e-05,
+      "loss": 3.0473,
+      "step": 32135
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4379113963978664e-05,
+      "loss": 3.0296,
+      "step": 32140
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.436758604904795e-05,
+      "loss": 2.9075,
+      "step": 32145
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.435605790015902e-05,
+      "loss": 3.0214,
+      "step": 32150
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.434452951792937e-05,
+      "loss": 2.9918,
+      "step": 32155
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.433300090297656e-05,
+      "loss": 3.0894,
+      "step": 32160
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4321472055918156e-05,
+      "loss": 2.941,
+      "step": 32165
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4309942977371695e-05,
+      "loss": 2.8445,
+      "step": 32170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4298413667954784e-05,
+      "loss": 2.8094,
+      "step": 32175
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4286884128284996e-05,
+      "loss": 2.9933,
+      "step": 32180
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4275354358979935e-05,
+      "loss": 2.9889,
+      "step": 32185
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.426382436065721e-05,
+      "loss": 3.1609,
+      "step": 32190
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4252294133934455e-05,
+      "loss": 2.8609,
+      "step": 32195
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.424076367942932e-05,
+      "loss": 2.9628,
+      "step": 32200
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.422923299775944e-05,
+      "loss": 3.0578,
+      "step": 32205
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4217702089542475e-05,
+      "loss": 3.1014,
+      "step": 32210
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.420617095539611e-05,
+      "loss": 3.0741,
+      "step": 32215
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.4194639595938024e-05,
+      "loss": 3.0051,
+      "step": 32220
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.418310801178594e-05,
+      "loss": 2.9494,
+      "step": 32225
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.417157620355755e-05,
+      "loss": 2.8897,
+      "step": 32230
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.416004417187057e-05,
+      "loss": 3.1294,
+      "step": 32235
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.414851191734276e-05,
+      "loss": 3.0197,
+      "step": 32240
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.413697944059184e-05,
+      "loss": 2.9891,
+      "step": 32245
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.41254467422356e-05,
+      "loss": 2.9774,
+      "step": 32250
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.411391382289178e-05,
+      "loss": 3.0103,
+      "step": 32255
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.410238068317819e-05,
+      "loss": 2.8553,
+      "step": 32260
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.40908473237126e-05,
+      "loss": 3.0603,
+      "step": 32265
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.407931374511284e-05,
+      "loss": 2.9287,
+      "step": 32270
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.406777994799672e-05,
+      "loss": 3.0018,
+      "step": 32275
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.405624593298206e-05,
+      "loss": 3.0614,
+      "step": 32280
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 5.404471170068671e-05,
+      "loss": 2.9488,
+      "step": 32285
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.403317725172853e-05,
+      "loss": 3.0491,
+      "step": 32290
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.402164258672538e-05,
+      "loss": 2.9501,
+      "step": 32295
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.4010107706295136e-05,
+      "loss": 3.0173,
+      "step": 32300
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.399857261105568e-05,
+      "loss": 3.0805,
+      "step": 32305
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.398703730162492e-05,
+      "loss": 3.0433,
+      "step": 32310
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3975501778620774e-05,
+      "loss": 2.9943,
+      "step": 32315
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3963966042661154e-05,
+      "loss": 2.9796,
+      "step": 32320
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3952430094363994e-05,
+      "loss": 2.9634,
+      "step": 32325
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3940893934347236e-05,
+      "loss": 2.9282,
+      "step": 32330
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.392935756322884e-05,
+      "loss": 2.9908,
+      "step": 32335
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.391782098162679e-05,
+      "loss": 3.0331,
+      "step": 32340
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.390628419015904e-05,
+      "loss": 3.1187,
+      "step": 32345
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.389474718944361e-05,
+      "loss": 2.8929,
+      "step": 32350
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.388320998009847e-05,
+      "loss": 3.0454,
+      "step": 32355
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.387167256274165e-05,
+      "loss": 3.0062,
+      "step": 32360
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.386013493799117e-05,
+      "loss": 3.0121,
+      "step": 32365
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.384859710646506e-05,
+      "loss": 2.9146,
+      "step": 32370
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3837059068781394e-05,
+      "loss": 3.0819,
+      "step": 32375
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.38255208255582e-05,
+      "loss": 2.9318,
+      "step": 32380
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.381398237741354e-05,
+      "loss": 3.0616,
+      "step": 32385
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.380244372496552e-05,
+      "loss": 2.876,
+      "step": 32390
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3790904868832226e-05,
+      "loss": 3.0899,
+      "step": 32395
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3779365809631746e-05,
+      "loss": 2.942,
+      "step": 32400
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.37678265479822e-05,
+      "loss": 3.0648,
+      "step": 32405
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.37562870845017e-05,
+      "loss": 2.8548,
+      "step": 32410
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3744747419808396e-05,
+      "loss": 3.0783,
+      "step": 32415
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.373320755452043e-05,
+      "loss": 2.9189,
+      "step": 32420
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3721667489255935e-05,
+      "loss": 3.021,
+      "step": 32425
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3710127224633124e-05,
+      "loss": 2.9596,
+      "step": 32430
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.369858676127012e-05,
+      "loss": 3.0133,
+      "step": 32435
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3687046099785145e-05,
+      "loss": 3.0525,
+      "step": 32440
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3675505240796374e-05,
+      "loss": 2.9808,
+      "step": 32445
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.366396418492203e-05,
+      "loss": 3.0066,
+      "step": 32450
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.365242293278036e-05,
+      "loss": 2.9312,
+      "step": 32455
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3640881484989535e-05,
+      "loss": 3.0322,
+      "step": 32460
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3629339842167825e-05,
+      "loss": 3.0972,
+      "step": 32465
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.361779800493349e-05,
+      "loss": 2.9307,
+      "step": 32470
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.360625597390477e-05,
+      "loss": 3.0936,
+      "step": 32475
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3594713749699964e-05,
+      "loss": 3.096,
+      "step": 32480
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.358317133293732e-05,
+      "loss": 3.0121,
+      "step": 32485
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.357162872423514e-05,
+      "loss": 3.0205,
+      "step": 32490
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3560085924211745e-05,
+      "loss": 2.9009,
+      "step": 32495
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.354854293348543e-05,
+      "loss": 2.951,
+      "step": 32500
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.353699975267453e-05,
+      "loss": 2.9597,
+      "step": 32505
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3525456382397375e-05,
+      "loss": 2.9969,
+      "step": 32510
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.351391282327229e-05,
+      "loss": 3.0029,
+      "step": 32515
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3502369075917645e-05,
+      "loss": 3.0133,
+      "step": 32520
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.349082514095179e-05,
+      "loss": 3.0997,
+      "step": 32525
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.347928101899313e-05,
+      "loss": 3.0533,
+      "step": 32530
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.346773671066002e-05,
+      "loss": 2.9797,
+      "step": 32535
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.345619221657086e-05,
+      "loss": 2.9481,
+      "step": 32540
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3444647537344036e-05,
+      "loss": 2.856,
+      "step": 32545
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.343310267359799e-05,
+      "loss": 2.9808,
+      "step": 32550
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.342155762595115e-05,
+      "loss": 3.0499,
+      "step": 32555
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.341001239502191e-05,
+      "loss": 3.0752,
+      "step": 32560
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.339846698142874e-05,
+      "loss": 3.1774,
+      "step": 32565
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3386921385790086e-05,
+      "loss": 3.0294,
+      "step": 32570
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.337537560872442e-05,
+      "loss": 2.9655,
+      "step": 32575
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.336382965085021e-05,
+      "loss": 3.0976,
+      "step": 32580
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3352283512785914e-05,
+      "loss": 3.0617,
+      "step": 32585
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.334073719515006e-05,
+      "loss": 3.0205,
+      "step": 32590
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.332919069856111e-05,
+      "loss": 2.9772,
+      "step": 32595
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.331764402363761e-05,
+      "loss": 2.9985,
+      "step": 32600
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.330609717099807e-05,
+      "loss": 2.9498,
+      "step": 32605
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.329455014126101e-05,
+      "loss": 3.0724,
+      "step": 32610
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.328300293504499e-05,
+      "loss": 3.0887,
+      "step": 32615
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3271455552968516e-05,
+      "loss": 2.9932,
+      "step": 32620
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.325990799565019e-05,
+      "loss": 2.8937,
+      "step": 32625
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3248360263708564e-05,
+      "loss": 2.8832,
+      "step": 32630
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.32368123577622e-05,
+      "loss": 2.9018,
+      "step": 32635
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.322526427842971e-05,
+      "loss": 2.8558,
+      "step": 32640
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3213716026329666e-05,
+      "loss": 3.067,
+      "step": 32645
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.320216760208069e-05,
+      "loss": 3.1805,
+      "step": 32650
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.31906190063014e-05,
+      "loss": 2.9635,
+      "step": 32655
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.317907023961039e-05,
+      "loss": 3.0123,
+      "step": 32660
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.316752130262631e-05,
+      "loss": 3.0694,
+      "step": 32665
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.315597219596781e-05,
+      "loss": 2.8429,
+      "step": 32670
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3144422920253524e-05,
+      "loss": 3.046,
+      "step": 32675
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.313287347610214e-05,
+      "loss": 3.0158,
+      "step": 32680
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.312132386413228e-05,
+      "loss": 2.8685,
+      "step": 32685
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.310977408496266e-05,
+      "loss": 2.9512,
+      "step": 32690
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.309822413921194e-05,
+      "loss": 2.9968,
+      "step": 32695
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.308667402749884e-05,
+      "loss": 3.0138,
+      "step": 32700
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.307512375044206e-05,
+      "loss": 2.8488,
+      "step": 32705
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.306357330866029e-05,
+      "loss": 2.9712,
+      "step": 32710
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3052022702772265e-05,
+      "loss": 2.9863,
+      "step": 32715
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.304047193339673e-05,
+      "loss": 3.0667,
+      "step": 32720
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.30289210011524e-05,
+      "loss": 3.0767,
+      "step": 32725
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.3017369906658056e-05,
+      "loss": 3.016,
+      "step": 32730
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.300581865053241e-05,
+      "loss": 2.8908,
+      "step": 32735
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.299426723339427e-05,
+      "loss": 2.9525,
+      "step": 32740
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.298271565586238e-05,
+      "loss": 3.1107,
+      "step": 32745
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.297116391855553e-05,
+      "loss": 2.9515,
+      "step": 32750
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2959612022092516e-05,
+      "loss": 3.1814,
+      "step": 32755
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2948059967092166e-05,
+      "loss": 3.015,
+      "step": 32760
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2936507754173236e-05,
+      "loss": 2.9157,
+      "step": 32765
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.292495538395457e-05,
+      "loss": 3.0093,
+      "step": 32770
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2913402857054986e-05,
+      "loss": 2.9118,
+      "step": 32775
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.290185017409332e-05,
+      "loss": 3.0192,
+      "step": 32780
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.289029733568843e-05,
+      "loss": 2.9431,
+      "step": 32785
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.287874434245914e-05,
+      "loss": 2.8161,
+      "step": 32790
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.286719119502431e-05,
+      "loss": 3.0858,
+      "step": 32795
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.285563789400283e-05,
+      "loss": 3.0733,
+      "step": 32800
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.284408444001355e-05,
+      "loss": 2.9882,
+      "step": 32805
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.283253083367539e-05,
+      "loss": 3.0377,
+      "step": 32810
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.282097707560719e-05,
+      "loss": 3.0186,
+      "step": 32815
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2809423166427885e-05,
+      "loss": 3.0114,
+      "step": 32820
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2797869106756373e-05,
+      "loss": 2.9852,
+      "step": 32825
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.278631489721156e-05,
+      "loss": 3.0914,
+      "step": 32830
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2774760538412406e-05,
+      "loss": 2.8411,
+      "step": 32835
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.27632060309778e-05,
+      "loss": 2.9457,
+      "step": 32840
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.275165137552669e-05,
+      "loss": 2.9077,
+      "step": 32845
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.274009657267803e-05,
+      "loss": 3.0645,
+      "step": 32850
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2728541623050786e-05,
+      "loss": 2.9941,
+      "step": 32855
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.271698652726392e-05,
+      "loss": 2.9745,
+      "step": 32860
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.270543128593638e-05,
+      "loss": 2.9616,
+      "step": 32865
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2693875899687163e-05,
+      "loss": 3.0993,
+      "step": 32870
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2682320369135254e-05,
+      "loss": 3.0026,
+      "step": 32875
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.267076469489964e-05,
+      "loss": 3.0625,
+      "step": 32880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.265920887759935e-05,
+      "loss": 2.9781,
+      "step": 32885
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.264765291785336e-05,
+      "loss": 2.9882,
+      "step": 32890
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.26360968162807e-05,
+      "loss": 3.1267,
+      "step": 32895
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2624540573500394e-05,
+      "loss": 3.087,
+      "step": 32900
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2612984190131475e-05,
+      "loss": 2.9336,
+      "step": 32905
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.260142766679299e-05,
+      "loss": 2.9957,
+      "step": 32910
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.258987100410397e-05,
+      "loss": 2.8912,
+      "step": 32915
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.257831420268351e-05,
+      "loss": 2.9219,
+      "step": 32920
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2566757263150614e-05,
+      "loss": 3.0869,
+      "step": 32925
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2555200186124385e-05,
+      "loss": 3.0365,
+      "step": 32930
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2543642972223906e-05,
+      "loss": 2.8853,
+      "step": 32935
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.253208562206825e-05,
+      "loss": 2.9871,
+      "step": 32940
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.2520528136276505e-05,
+      "loss": 2.9711,
+      "step": 32945
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.250897051546778e-05,
+      "loss": 2.9786,
+      "step": 32950
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.249741276026118e-05,
+      "loss": 2.8213,
+      "step": 32955
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.248585487127582e-05,
+      "loss": 2.9784,
+      "step": 32960
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.247429684913081e-05,
+      "loss": 2.8993,
+      "step": 32965
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2462738694445294e-05,
+      "loss": 3.0302,
+      "step": 32970
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2451180407838384e-05,
+      "loss": 2.9416,
+      "step": 32975
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2439621989929244e-05,
+      "loss": 2.9383,
+      "step": 32980
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2428063441337016e-05,
+      "loss": 3.0359,
+      "step": 32985
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.241650476268085e-05,
+      "loss": 2.9347,
+      "step": 32990
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.240494595457992e-05,
+      "loss": 3.0191,
+      "step": 32995
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.239338701765338e-05,
+      "loss": 2.9101,
+      "step": 33000
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.238182795252043e-05,
+      "loss": 3.0076,
+      "step": 33005
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2370268759800244e-05,
+      "loss": 2.9961,
+      "step": 33010
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2358709440112006e-05,
+      "loss": 3.0532,
+      "step": 33015
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2347149994074905e-05,
+      "loss": 2.8427,
+      "step": 33020
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2335590422308164e-05,
+      "loss": 2.8755,
+      "step": 33025
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2324030725431e-05,
+      "loss": 2.975,
+      "step": 33030
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.231247090406261e-05,
+      "loss": 2.8868,
+      "step": 33035
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2300910958822214e-05,
+      "loss": 2.9136,
+      "step": 33040
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.228935089032906e-05,
+      "loss": 2.9353,
+      "step": 33045
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.227779069920239e-05,
+      "loss": 3.0932,
+      "step": 33050
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.226623038606142e-05,
+      "loss": 2.976,
+      "step": 33055
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.225466995152545e-05,
+      "loss": 2.9526,
+      "step": 33060
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2243109396213674e-05,
+      "loss": 2.9136,
+      "step": 33065
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2231548720745404e-05,
+      "loss": 2.9712,
+      "step": 33070
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.221998792573989e-05,
+      "loss": 2.893,
+      "step": 33075
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.220842701181641e-05,
+      "loss": 2.8996,
+      "step": 33080
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2196865979594246e-05,
+      "loss": 3.068,
+      "step": 33085
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2185304829692706e-05,
+      "loss": 2.9455,
+      "step": 33090
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.217374356273106e-05,
+      "loss": 2.9119,
+      "step": 33095
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.216218217932862e-05,
+      "loss": 3.0194,
+      "step": 33100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.215062068010469e-05,
+      "loss": 2.9261,
+      "step": 33105
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.213905906567859e-05,
+      "loss": 2.9277,
+      "step": 33110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.212749733666964e-05,
+      "loss": 2.8953,
+      "step": 33115
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.211593549369717e-05,
+      "loss": 3.0912,
+      "step": 33120
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.210437353738049e-05,
+      "loss": 2.9552,
+      "step": 33125
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2092811468338964e-05,
+      "loss": 2.9482,
+      "step": 33130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.208124928719193e-05,
+      "loss": 2.9194,
+      "step": 33135
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.206968699455874e-05,
+      "loss": 2.8623,
+      "step": 33140
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2058124591058745e-05,
+      "loss": 3.0444,
+      "step": 33145
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.20465620773113e-05,
+      "loss": 2.9366,
+      "step": 33150
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.2034999453935795e-05,
+      "loss": 3.0308,
+      "step": 33155
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.202343672155159e-05,
+      "loss": 2.966,
+      "step": 33160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.201187388077807e-05,
+      "loss": 3.0494,
+      "step": 33165
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.200031093223462e-05,
+      "loss": 2.9881,
+      "step": 33170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.198874787654062e-05,
+      "loss": 3.0188,
+      "step": 33175
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1977184714315485e-05,
+      "loss": 2.9741,
+      "step": 33180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1965621446178617e-05,
+      "loss": 3.043,
+      "step": 33185
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.195405807274942e-05,
+      "loss": 2.9598,
+      "step": 33190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.194249459464731e-05,
+      "loss": 2.9793,
+      "step": 33195
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.193093101249169e-05,
+      "loss": 2.9696,
+      "step": 33200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.191936732690201e-05,
+      "loss": 2.8765,
+      "step": 33205
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.190780353849769e-05,
+      "loss": 2.8353,
+      "step": 33210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.189623964789817e-05,
+      "loss": 2.7881,
+      "step": 33215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1884675655722905e-05,
+      "loss": 2.9739,
+      "step": 33220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1873111562591315e-05,
+      "loss": 2.8819,
+      "step": 33225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1861547369122864e-05,
+      "loss": 3.0596,
+      "step": 33230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.184998307593701e-05,
+      "loss": 3.0399,
+      "step": 33235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.183841868365323e-05,
+      "loss": 2.9881,
+      "step": 33240
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.182685419289098e-05,
+      "loss": 2.8348,
+      "step": 33245
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.181528960426975e-05,
+      "loss": 3.0606,
+      "step": 33250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.180372491840899e-05,
+      "loss": 2.8389,
+      "step": 33255
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1792160135928214e-05,
+      "loss": 2.9807,
+      "step": 33260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1780595257446906e-05,
+      "loss": 2.9662,
+      "step": 33265
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1769030283584544e-05,
+      "loss": 2.8788,
+      "step": 33270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.175746521496064e-05,
+      "loss": 2.8924,
+      "step": 33275
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.174590005219471e-05,
+      "loss": 3.0525,
+      "step": 33280
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.173433479590624e-05,
+      "loss": 2.8383,
+      "step": 33285
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.172276944671477e-05,
+      "loss": 3.1002,
+      "step": 33290
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.171120400523981e-05,
+      "loss": 2.8905,
+      "step": 33295
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1699638472100873e-05,
+      "loss": 3.0961,
+      "step": 33300
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1688072847917515e-05,
+      "loss": 3.0241,
+      "step": 33305
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1676507133309235e-05,
+      "loss": 2.9986,
+      "step": 33310
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.166494132889562e-05,
+      "loss": 3.1069,
+      "step": 33315
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.165337543529618e-05,
+      "loss": 3.0197,
+      "step": 33320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1641809453130476e-05,
+      "loss": 3.1279,
+      "step": 33325
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.163024338301806e-05,
+      "loss": 2.9609,
+      "step": 33330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.161867722557849e-05,
+      "loss": 2.9726,
+      "step": 33335
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1607110981431336e-05,
+      "loss": 3.1746,
+      "step": 33340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1595544651196156e-05,
+      "loss": 3.0607,
+      "step": 33345
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.158397823549253e-05,
+      "loss": 2.8961,
+      "step": 33350
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.157241173494004e-05,
+      "loss": 3.0715,
+      "step": 33355
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.156084515015825e-05,
+      "loss": 2.8892,
+      "step": 33360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1549278481766785e-05,
+      "loss": 2.9241,
+      "step": 33365
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1537711730385184e-05,
+      "loss": 2.997,
+      "step": 33370
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.152614489663308e-05,
+      "loss": 3.0843,
+      "step": 33375
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1514577981130054e-05,
+      "loss": 2.906,
+      "step": 33380
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1503010984495714e-05,
+      "loss": 3.082,
+      "step": 33385
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.149144390734969e-05,
+      "loss": 3.0013,
+      "step": 33390
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.147987675031157e-05,
+      "loss": 3.112,
+      "step": 33395
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1468309514000966e-05,
+      "loss": 2.8747,
+      "step": 33400
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.145674219903752e-05,
+      "loss": 2.8739,
+      "step": 33405
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.144517480604085e-05,
+      "loss": 2.9701,
+      "step": 33410
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1433607335630586e-05,
+      "loss": 3.01,
+      "step": 33415
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.142203978842638e-05,
+      "loss": 3.0337,
+      "step": 33420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.141047216504783e-05,
+      "loss": 2.9446,
+      "step": 33425
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.139890446611459e-05,
+      "loss": 2.9228,
+      "step": 33430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1387336692246336e-05,
+      "loss": 3.0284,
+      "step": 33435
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.137576884406269e-05,
+      "loss": 3.0245,
+      "step": 33440
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.136420092218333e-05,
+      "loss": 2.9728,
+      "step": 33445
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.135263292722787e-05,
+      "loss": 3.161,
+      "step": 33450
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1341064859816027e-05,
+      "loss": 2.8936,
+      "step": 33455
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.132949672056742e-05,
+      "loss": 3.0183,
+      "step": 33460
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.131792851010174e-05,
+      "loss": 2.9387,
+      "step": 33465
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.130636022903868e-05,
+      "loss": 2.9527,
+      "step": 33470
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1294791877997894e-05,
+      "loss": 2.9834,
+      "step": 33475
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.128322345759906e-05,
+      "loss": 2.9725,
+      "step": 33480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.127165496846187e-05,
+      "loss": 3.0898,
+      "step": 33485
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.126008641120601e-05,
+      "loss": 2.9349,
+      "step": 33490
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.124851778645118e-05,
+      "loss": 3.0005,
+      "step": 33495
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.123694909481707e-05,
+      "loss": 2.9499,
+      "step": 33500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.122538033692338e-05,
+      "loss": 3.0695,
+      "step": 33505
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.121381151338981e-05,
+      "loss": 3.0951,
+      "step": 33510
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1202242624836063e-05,
+      "loss": 2.9794,
+      "step": 33515
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.119067367188186e-05,
+      "loss": 2.9946,
+      "step": 33520
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.117910465514693e-05,
+      "loss": 2.9,
+      "step": 33525
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1167535575250944e-05,
+      "loss": 2.9249,
+      "step": 33530
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.115596643281365e-05,
+      "loss": 3.0423,
+      "step": 33535
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.114439722845478e-05,
+      "loss": 3.0769,
+      "step": 33540
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.113282796279404e-05,
+      "loss": 3.0176,
+      "step": 33545
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.112125863645118e-05,
+      "loss": 2.9561,
+      "step": 33550
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1109689250045914e-05,
+      "loss": 2.8831,
+      "step": 33555
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1098119804198e-05,
+      "loss": 2.9111,
+      "step": 33560
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.108655029952717e-05,
+      "loss": 3.1591,
+      "step": 33565
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1074980736653157e-05,
+      "loss": 2.8573,
+      "step": 33570
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.10634111161957e-05,
+      "loss": 3.0056,
+      "step": 33575
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.105184143877457e-05,
+      "loss": 3.0059,
+      "step": 33580
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1040271705009524e-05,
+      "loss": 2.8463,
+      "step": 33585
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1028701915520295e-05,
+      "loss": 3.1161,
+      "step": 33590
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1017132070926654e-05,
+      "loss": 2.8296,
+      "step": 33595
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.1005562171848356e-05,
+      "loss": 2.9965,
+      "step": 33600
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.099399221890516e-05,
+      "loss": 3.0376,
+      "step": 33605
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.098242221271685e-05,
+      "loss": 2.9699,
+      "step": 33610
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.097085215390318e-05,
+      "loss": 3.0099,
+      "step": 33615
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.095928204308395e-05,
+      "loss": 3.0458,
+      "step": 33620
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.094771188087889e-05,
+      "loss": 3.0438,
+      "step": 33625
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.0936141667907814e-05,
+      "loss": 3.0046,
+      "step": 33630
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.092457140479049e-05,
+      "loss": 2.8687,
+      "step": 33635
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 5.0913001092146696e-05,
+      "loss": 2.9985,
+      "step": 33640
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.090143073059624e-05,
+      "loss": 2.9542,
+      "step": 33645
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.088986032075889e-05,
+      "loss": 3.1604,
+      "step": 33650
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0878289863254435e-05,
+      "loss": 3.0828,
+      "step": 33655
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.086671935870269e-05,
+      "loss": 2.9031,
+      "step": 33660
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.085514880772343e-05,
+      "loss": 2.9958,
+      "step": 33665
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.084357821093648e-05,
+      "loss": 3.16,
+      "step": 33670
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0832007568961615e-05,
+      "loss": 3.102,
+      "step": 33675
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0820436882418654e-05,
+      "loss": 3.0133,
+      "step": 33680
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.080886615192738e-05,
+      "loss": 2.9994,
+      "step": 33685
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0797295378107636e-05,
+      "loss": 2.9364,
+      "step": 33690
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.078572456157923e-05,
+      "loss": 2.8627,
+      "step": 33695
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0774153702961944e-05,
+      "loss": 3.0316,
+      "step": 33700
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0762582802875614e-05,
+      "loss": 2.8847,
+      "step": 33705
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.075101186194005e-05,
+      "loss": 3.1633,
+      "step": 33710
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.073944088077508e-05,
+      "loss": 3.0275,
+      "step": 33715
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0727869860000535e-05,
+      "loss": 3.112,
+      "step": 33720
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0716298800236215e-05,
+      "loss": 2.8896,
+      "step": 33725
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.070472770210196e-05,
+      "loss": 3.0138,
+      "step": 33730
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.06931565662176e-05,
+      "loss": 3.007,
+      "step": 33735
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0681585393202965e-05,
+      "loss": 2.9015,
+      "step": 33740
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.067001418367788e-05,
+      "loss": 2.9657,
+      "step": 33745
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0658442938262195e-05,
+      "loss": 3.0209,
+      "step": 33750
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.064687165757573e-05,
+      "loss": 3.0682,
+      "step": 33755
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.063530034223832e-05,
+      "loss": 3.0168,
+      "step": 33760
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0623728992869826e-05,
+      "loss": 2.983,
+      "step": 33765
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0612157610090075e-05,
+      "loss": 2.9961,
+      "step": 33770
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.060058619451892e-05,
+      "loss": 2.8739,
+      "step": 33775
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.058901474677619e-05,
+      "loss": 3.0294,
+      "step": 33780
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.057744326748173e-05,
+      "loss": 3.0213,
+      "step": 33785
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0565871757255414e-05,
+      "loss": 2.906,
+      "step": 33790
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0554300216717064e-05,
+      "loss": 3.0705,
+      "step": 33795
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.054272864648657e-05,
+      "loss": 3.0694,
+      "step": 33800
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.053115704718375e-05,
+      "loss": 3.0573,
+      "step": 33805
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.051958541942847e-05,
+      "loss": 3.0743,
+      "step": 33810
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.050801376384059e-05,
+      "loss": 2.9657,
+      "step": 33815
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.049644208103996e-05,
+      "loss": 2.9684,
+      "step": 33820
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0484870371646466e-05,
+      "loss": 3.0862,
+      "step": 33825
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.047329863627993e-05,
+      "loss": 2.9801,
+      "step": 33830
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0461726875560234e-05,
+      "loss": 2.873,
+      "step": 33835
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0450155090107254e-05,
+      "loss": 2.8626,
+      "step": 33840
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.043858328054083e-05,
+      "loss": 3.0364,
+      "step": 33845
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0427011447480845e-05,
+      "loss": 3.0324,
+      "step": 33850
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0415439591547155e-05,
+      "loss": 3.0227,
+      "step": 33855
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0403867713359646e-05,
+      "loss": 3.0487,
+      "step": 33860
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.039229581353818e-05,
+      "loss": 2.9886,
+      "step": 33865
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.038072389270263e-05,
+      "loss": 2.9735,
+      "step": 33870
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.036915195147286e-05,
+      "loss": 2.9944,
+      "step": 33875
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.035757999046875e-05,
+      "loss": 2.9707,
+      "step": 33880
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0346008010310164e-05,
+      "loss": 3.0259,
+      "step": 33885
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0334436011617005e-05,
+      "loss": 3.0108,
+      "step": 33890
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.032286399500912e-05,
+      "loss": 3.0371,
+      "step": 33895
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.03112919611064e-05,
+      "loss": 2.9402,
+      "step": 33900
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.029971991052872e-05,
+      "loss": 3.03,
+      "step": 33905
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.028814784389596e-05,
+      "loss": 2.9488,
+      "step": 33910
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0276575761828e-05,
+      "loss": 3.0295,
+      "step": 33915
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.026500366494473e-05,
+      "loss": 3.0655,
+      "step": 33920
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0253431553866036e-05,
+      "loss": 2.9054,
+      "step": 33925
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0241859429211766e-05,
+      "loss": 3.0194,
+      "step": 33930
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.023028729160184e-05,
+      "loss": 2.9336,
+      "step": 33935
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.021871514165611e-05,
+      "loss": 3.0011,
+      "step": 33940
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.020714297999449e-05,
+      "loss": 2.9939,
+      "step": 33945
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.019557080723687e-05,
+      "loss": 3.0667,
+      "step": 33950
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.01839986240031e-05,
+      "loss": 2.9986,
+      "step": 33955
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.017242643091309e-05,
+      "loss": 2.9849,
+      "step": 33960
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.016085422858673e-05,
+      "loss": 2.9398,
+      "step": 33965
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0149282017643895e-05,
+      "loss": 3.0282,
+      "step": 33970
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.013770979870448e-05,
+      "loss": 2.9746,
+      "step": 33975
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.012613757238838e-05,
+      "loss": 3.0484,
+      "step": 33980
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0114565339315456e-05,
+      "loss": 3.0021,
+      "step": 33985
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.010299310010563e-05,
+      "loss": 2.9748,
+      "step": 33990
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.009142085537878e-05,
+      "loss": 3.0614,
+      "step": 33995
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.00798486057548e-05,
+      "loss": 2.8609,
+      "step": 34000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.006827635185356e-05,
+      "loss": 2.9655,
+      "step": 34005
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.005670409429497e-05,
+      "loss": 3.0475,
+      "step": 34010
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.004513183369891e-05,
+      "loss": 2.8466,
+      "step": 34015
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0033559570685286e-05,
+      "loss": 2.948,
+      "step": 34020
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.0021987305873986e-05,
+      "loss": 3.0955,
+      "step": 34025
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.001041503988487e-05,
+      "loss": 2.9556,
+      "step": 34030
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.999884277333787e-05,
+      "loss": 2.868,
+      "step": 34035
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.998727050685286e-05,
+      "loss": 2.8862,
+      "step": 34040
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.997569824104971e-05,
+      "loss": 2.8835,
+      "step": 34045
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.996412597654835e-05,
+      "loss": 2.9936,
+      "step": 34050
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.995255371396864e-05,
+      "loss": 2.948,
+      "step": 34055
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.994098145393049e-05,
+      "loss": 2.9759,
+      "step": 34060
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9929409197053794e-05,
+      "loss": 2.9953,
+      "step": 34065
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.99178369439584e-05,
+      "loss": 3.0139,
+      "step": 34070
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.990626469526426e-05,
+      "loss": 3.0991,
+      "step": 34075
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9894692451591204e-05,
+      "loss": 3.0694,
+      "step": 34080
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.988312021355918e-05,
+      "loss": 2.965,
+      "step": 34085
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.987154798178805e-05,
+      "loss": 3.0214,
+      "step": 34090
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9859975756897676e-05,
+      "loss": 3.0226,
+      "step": 34095
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.984840353950799e-05,
+      "loss": 3.0562,
+      "step": 34100
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9836831330238856e-05,
+      "loss": 3.115,
+      "step": 34105
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.982525912971018e-05,
+      "loss": 2.9385,
+      "step": 34110
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.981368693854182e-05,
+      "loss": 3.0534,
+      "step": 34115
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9802114757353694e-05,
+      "loss": 3.0083,
+      "step": 34120
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.979054258676567e-05,
+      "loss": 3.1207,
+      "step": 34125
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9778970427397636e-05,
+      "loss": 3.0145,
+      "step": 34130
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.976739827986949e-05,
+      "loss": 3.0357,
+      "step": 34135
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.97558261448011e-05,
+      "loss": 3.0486,
+      "step": 34140
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.974425402281236e-05,
+      "loss": 3.0416,
+      "step": 34145
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9732681914523144e-05,
+      "loss": 2.7456,
+      "step": 34150
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.972110982055334e-05,
+      "loss": 3.043,
+      "step": 34155
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9709537741522834e-05,
+      "loss": 2.9802,
+      "step": 34160
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.969796567805151e-05,
+      "loss": 3.1373,
+      "step": 34165
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.968639363075924e-05,
+      "loss": 2.9808,
+      "step": 34170
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9674821600265895e-05,
+      "loss": 2.8455,
+      "step": 34175
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.966324958719136e-05,
+      "loss": 3.0021,
+      "step": 34180
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.965167759215552e-05,
+      "loss": 3.0102,
+      "step": 34185
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9640105615778255e-05,
+      "loss": 2.7999,
+      "step": 34190
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.962853365867941e-05,
+      "loss": 3.0523,
+      "step": 34195
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9616961721478896e-05,
+      "loss": 3.1072,
+      "step": 34200
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.960538980479656e-05,
+      "loss": 2.9771,
+      "step": 34205
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.959381790925231e-05,
+      "loss": 2.9556,
+      "step": 34210
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.958224603546598e-05,
+      "loss": 2.9337,
+      "step": 34215
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.957067418405743e-05,
+      "loss": 2.8494,
+      "step": 34220
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.955910235564658e-05,
+      "loss": 3.0305,
+      "step": 34225
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.954753055085325e-05,
+      "loss": 2.9483,
+      "step": 34230
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.953595877029734e-05,
+      "loss": 3.0709,
+      "step": 34235
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9524387014598675e-05,
+      "loss": 2.9722,
+      "step": 34240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9512815284377176e-05,
+      "loss": 2.9046,
+      "step": 34245
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.950124358025266e-05,
+      "loss": 2.9849,
+      "step": 34250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.948967190284499e-05,
+      "loss": 2.9615,
+      "step": 34255
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.947810025277405e-05,
+      "loss": 3.0658,
+      "step": 34260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9466528630659656e-05,
+      "loss": 3.0755,
+      "step": 34265
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9454957037121716e-05,
+      "loss": 3.0027,
+      "step": 34270
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9443385472780066e-05,
+      "loss": 2.9824,
+      "step": 34275
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.943181393825453e-05,
+      "loss": 3.0525,
+      "step": 34280
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.942024243416501e-05,
+      "loss": 3.0664,
+      "step": 34285
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9408670961131305e-05,
+      "loss": 2.9189,
+      "step": 34290
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.939709951977331e-05,
+      "loss": 3.022,
+      "step": 34295
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.938552811071085e-05,
+      "loss": 2.9657,
+      "step": 34300
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.937395673456374e-05,
+      "loss": 2.963,
+      "step": 34305
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.9362385391951885e-05,
+      "loss": 2.8681,
+      "step": 34310
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.935081408349507e-05,
+      "loss": 2.8932,
+      "step": 34315
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.933924280981318e-05,
+      "loss": 3.0207,
+      "step": 34320
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.932767157152603e-05,
+      "loss": 2.9621,
+      "step": 34325
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.9316100369253436e-05,
+      "loss": 3.0975,
+      "step": 34330
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.930452920361528e-05,
+      "loss": 3.0414,
+      "step": 34335
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.929295807523134e-05,
+      "loss": 2.9124,
+      "step": 34340
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.92813869847215e-05,
+      "loss": 2.9238,
+      "step": 34345
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.926981593270557e-05,
+      "loss": 2.9853,
+      "step": 34350
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.9258244919803326e-05,
+      "loss": 2.9271,
+      "step": 34355
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.924667394663467e-05,
+      "loss": 2.9355,
+      "step": 34360
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.9235103013819357e-05,
+      "loss": 3.1488,
+      "step": 34365
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.922353212197726e-05,
+      "loss": 2.9873,
+      "step": 34370
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.921196127172817e-05,
+      "loss": 3.1687,
+      "step": 34375
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.920039046369189e-05,
+      "loss": 2.9754,
+      "step": 34380
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.918881969848826e-05,
+      "loss": 2.9259,
+      "step": 34385
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.917724897673707e-05,
+      "loss": 2.8648,
+      "step": 34390
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.9165678299058144e-05,
+      "loss": 3.0557,
+      "step": 34395
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.915410766607129e-05,
+      "loss": 3.0133,
+      "step": 34400
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.914253707839627e-05,
+      "loss": 2.9496,
+      "step": 34405
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.913096653665294e-05,
+      "loss": 2.9706,
+      "step": 34410
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.9119396041461055e-05,
+      "loss": 2.9724,
+      "step": 34415
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.910782559344046e-05,
+      "loss": 2.8023,
+      "step": 34420
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.90962551932109e-05,
+      "loss": 2.9659,
+      "step": 34425
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.908468484139218e-05,
+      "loss": 2.995,
+      "step": 34430
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.9073114538604115e-05,
+      "loss": 2.984,
+      "step": 34435
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.9061544285466446e-05,
+      "loss": 2.8514,
+      "step": 34440
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.9049974082598996e-05,
+      "loss": 2.9169,
+      "step": 34445
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.903840393062152e-05,
+      "loss": 3.0138,
+      "step": 34450
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.902683383015381e-05,
+      "loss": 2.9247,
+      "step": 34455
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.9015263781815636e-05,
+      "loss": 3.0935,
+      "step": 34460
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.900369378622676e-05,
+      "loss": 3.0429,
+      "step": 34465
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8992123844006964e-05,
+      "loss": 2.988,
+      "step": 34470
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.898055395577603e-05,
+      "loss": 2.8001,
+      "step": 34475
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8968984122153685e-05,
+      "loss": 3.1062,
+      "step": 34480
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.895741434375971e-05,
+      "loss": 2.9474,
+      "step": 34485
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.894584462121386e-05,
+      "loss": 3.0307,
+      "step": 34490
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8934274955135885e-05,
+      "loss": 2.7888,
+      "step": 34495
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8922705346145556e-05,
+      "loss": 3.0879,
+      "step": 34500
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8911135794862585e-05,
+      "loss": 3.0596,
+      "step": 34505
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8899566301906756e-05,
+      "loss": 2.8885,
+      "step": 34510
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.888799686789779e-05,
+      "loss": 2.8574,
+      "step": 34515
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.887642749345543e-05,
+      "loss": 3.0146,
+      "step": 34520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.886485817919943e-05,
+      "loss": 2.8654,
+      "step": 34525
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.885328892574948e-05,
+      "loss": 3.0242,
+      "step": 34530
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8841719733725345e-05,
+      "loss": 2.961,
+      "step": 34535
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.883015060374673e-05,
+      "loss": 3.0277,
+      "step": 34540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.881858153643339e-05,
+      "loss": 2.9979,
+      "step": 34545
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.880701253240502e-05,
+      "loss": 3.1769,
+      "step": 34550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8795443592281325e-05,
+      "loss": 2.8994,
+      "step": 34555
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.878387471668205e-05,
+      "loss": 3.0811,
+      "step": 34560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.877230590622687e-05,
+      "loss": 2.9667,
+      "step": 34565
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.876073716153553e-05,
+      "loss": 2.9375,
+      "step": 34570
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8749168483227705e-05,
+      "loss": 2.8703,
+      "step": 34575
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.873759987192308e-05,
+      "loss": 2.9678,
+      "step": 34580
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8726031328241387e-05,
+      "loss": 2.8922,
+      "step": 34585
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8714462852802284e-05,
+      "loss": 2.8294,
+      "step": 34590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8702894446225496e-05,
+      "loss": 3.0127,
+      "step": 34595
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.869132610913068e-05,
+      "loss": 3.016,
+      "step": 34600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.86797578421375e-05,
+      "loss": 2.9195,
+      "step": 34605
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.866818964586568e-05,
+      "loss": 2.9338,
+      "step": 34610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8656621520934846e-05,
+      "loss": 3.0451,
+      "step": 34615
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.864505346796471e-05,
+      "loss": 3.0145,
+      "step": 34620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.863348548757491e-05,
+      "loss": 2.9396,
+      "step": 34625
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.86219175803851e-05,
+      "loss": 3.0926,
+      "step": 34630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.861034974701497e-05,
+      "loss": 3.0172,
+      "step": 34635
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.859878198808413e-05,
+      "loss": 3.0335,
+      "step": 34640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8587214304212276e-05,
+      "loss": 3.0216,
+      "step": 34645
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.857564669601904e-05,
+      "loss": 2.972,
+      "step": 34650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.856407916412402e-05,
+      "loss": 3.0456,
+      "step": 34655
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.855251170914691e-05,
+      "loss": 2.9406,
+      "step": 34660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8540944331707306e-05,
+      "loss": 2.9961,
+      "step": 34665
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.852937703242487e-05,
+      "loss": 2.8427,
+      "step": 34670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8517809811919206e-05,
+      "loss": 2.9375,
+      "step": 34675
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.850624267080992e-05,
+      "loss": 3.0694,
+      "step": 34680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8494675609716664e-05,
+      "loss": 2.9191,
+      "step": 34685
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.848310862925901e-05,
+      "loss": 2.8858,
+      "step": 34690
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.847154173005661e-05,
+      "loss": 3.0332,
+      "step": 34695
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.845997491272905e-05,
+      "loss": 2.9965,
+      "step": 34700
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8448408177895896e-05,
+      "loss": 3.0162,
+      "step": 34705
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.84368415261768e-05,
+      "loss": 2.9551,
+      "step": 34710
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.84252749581913e-05,
+      "loss": 2.9988,
+      "step": 34715
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.841370847455901e-05,
+      "loss": 2.8589,
+      "step": 34720
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8402142075899516e-05,
+      "loss": 2.9311,
+      "step": 34725
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.839057576283237e-05,
+      "loss": 3.0441,
+      "step": 34730
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.837900953597718e-05,
+      "loss": 2.9431,
+      "step": 34735
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.836744339595346e-05,
+      "loss": 3.0412,
+      "step": 34740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8355877343380834e-05,
+      "loss": 3.0065,
+      "step": 34745
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8344311378878814e-05,
+      "loss": 2.9657,
+      "step": 34750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.833274550306697e-05,
+      "loss": 2.9694,
+      "step": 34755
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.832117971656486e-05,
+      "loss": 2.8957,
+      "step": 34760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8309614019992004e-05,
+      "loss": 2.9847,
+      "step": 34765
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.829804841396796e-05,
+      "loss": 2.9634,
+      "step": 34770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.828648289911227e-05,
+      "loss": 2.9425,
+      "step": 34775
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.827491747604444e-05,
+      "loss": 3.0066,
+      "step": 34780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.826335214538401e-05,
+      "loss": 2.9313,
+      "step": 34785
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8251786907750484e-05,
+      "loss": 3.0004,
+      "step": 34790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.824022176376339e-05,
+      "loss": 2.9554,
+      "step": 34795
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.822865671404225e-05,
+      "loss": 3.0391,
+      "step": 34800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.821709175920653e-05,
+      "loss": 3.0751,
+      "step": 34805
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.820552689987577e-05,
+      "loss": 2.975,
+      "step": 34810
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.819396213666943e-05,
+      "loss": 3.1709,
+      "step": 34815
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.818239747020702e-05,
+      "loss": 2.9406,
+      "step": 34820
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8170832901108036e-05,
+      "loss": 3.0116,
+      "step": 34825
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.815926842999191e-05,
+      "loss": 3.0126,
+      "step": 34830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.814770405747817e-05,
+      "loss": 2.9813,
+      "step": 34835
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8136139784186235e-05,
+      "loss": 3.0403,
+      "step": 34840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.812457561073561e-05,
+      "loss": 3.1132,
+      "step": 34845
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.811301153774574e-05,
+      "loss": 2.9579,
+      "step": 34850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.810144756583604e-05,
+      "loss": 2.9657,
+      "step": 34855
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.808988369562603e-05,
+      "loss": 2.9569,
+      "step": 34860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.8078319927735064e-05,
+      "loss": 2.9865,
+      "step": 34865
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.806675626278265e-05,
+      "loss": 2.9283,
+      "step": 34870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.80551927013882e-05,
+      "loss": 3.0053,
+      "step": 34875
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.804362924417109e-05,
+      "loss": 2.9992,
+      "step": 34880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.803206589175082e-05,
+      "loss": 2.9686,
+      "step": 34885
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.802050264474673e-05,
+      "loss": 3.1036,
+      "step": 34890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.800893950377828e-05,
+      "loss": 2.9693,
+      "step": 34895
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.7997376469464836e-05,
+      "loss": 2.8689,
+      "step": 34900
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.7985813542425825e-05,
+      "loss": 3.0182,
+      "step": 34905
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.797425072328062e-05,
+      "loss": 3.0009,
+      "step": 34910
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.7962688012648595e-05,
+      "loss": 2.9366,
+      "step": 34915
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.795112541114915e-05,
+      "loss": 3.1229,
+      "step": 34920
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.793956291940165e-05,
+      "loss": 2.9687,
+      "step": 34925
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.792800053802547e-05,
+      "loss": 2.9836,
+      "step": 34930
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.7916438267639966e-05,
+      "loss": 2.8592,
+      "step": 34935
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.790487610886447e-05,
+      "loss": 3.0205,
+      "step": 34940
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.7893314062318385e-05,
+      "loss": 3.0622,
+      "step": 34945
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.7881752128620986e-05,
+      "loss": 3.114,
+      "step": 34950
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.787019030839168e-05,
+      "loss": 2.9586,
+      "step": 34955
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.785862860224976e-05,
+      "loss": 3.0067,
+      "step": 34960
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.784706701081454e-05,
+      "loss": 3.0157,
+      "step": 34965
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.783550553470537e-05,
+      "loss": 3.1036,
+      "step": 34970
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.782394417454153e-05,
+      "loss": 2.9442,
+      "step": 34975
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.781238293094237e-05,
+      "loss": 3.0324,
+      "step": 34980
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.780082180452716e-05,
+      "loss": 2.9982,
+      "step": 34985
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.7789260795915184e-05,
+      "loss": 3.0726,
+      "step": 34990
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.777769990572577e-05,
+      "loss": 3.0485,
+      "step": 34995
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.776613913457815e-05,
+      "loss": 3.12,
+      "step": 35000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7754578483091647e-05,
+      "loss": 2.9659,
+      "step": 35005
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.77430179518855e-05,
+      "loss": 2.8094,
+      "step": 35010
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.773145754157897e-05,
+      "loss": 2.8579,
+      "step": 35015
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7719897252791334e-05,
+      "loss": 2.8951,
+      "step": 35020
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7708337086141814e-05,
+      "loss": 2.9659,
+      "step": 35025
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.769677704224968e-05,
+      "loss": 2.9768,
+      "step": 35030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.768521712173417e-05,
+      "loss": 3.169,
+      "step": 35035
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7673657325214464e-05,
+      "loss": 2.9686,
+      "step": 35040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.766209765330985e-05,
+      "loss": 3.052,
+      "step": 35045
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7650538106639494e-05,
+      "loss": 3.0077,
+      "step": 35050
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.763897868582264e-05,
+      "loss": 3.077,
+      "step": 35055
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.762741939147847e-05,
+      "loss": 3.0368,
+      "step": 35060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7615860224226186e-05,
+      "loss": 2.9179,
+      "step": 35065
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.760430118468498e-05,
+      "loss": 3.0297,
+      "step": 35070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.759274227347403e-05,
+      "loss": 3.0991,
+      "step": 35075
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.758118349121253e-05,
+      "loss": 2.9691,
+      "step": 35080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.756962483851962e-05,
+      "loss": 3.1355,
+      "step": 35085
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7558066316014476e-05,
+      "loss": 2.9515,
+      "step": 35090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.754650792431625e-05,
+      "loss": 2.9245,
+      "step": 35095
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.753494966404409e-05,
+      "loss": 3.1213,
+      "step": 35100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7523391535817134e-05,
+      "loss": 3.0271,
+      "step": 35105
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.751183354025453e-05,
+      "loss": 3.0098,
+      "step": 35110
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7500275677975384e-05,
+      "loss": 2.9928,
+      "step": 35115
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.748871794959884e-05,
+      "loss": 2.9685,
+      "step": 35120
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7477160355743976e-05,
+      "loss": 2.9256,
+      "step": 35125
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.746560289702993e-05,
+      "loss": 2.9528,
+      "step": 35130
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.745404557407579e-05,
+      "loss": 2.9908,
+      "step": 35135
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.744248838750062e-05,
+      "loss": 2.9997,
+      "step": 35140
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.743093133792355e-05,
+      "loss": 3.1306,
+      "step": 35145
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.74193744259636e-05,
+      "loss": 3.0069,
+      "step": 35150
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7407817652239896e-05,
+      "loss": 3.0683,
+      "step": 35155
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.739626101737147e-05,
+      "loss": 2.8367,
+      "step": 35160
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7384704521977364e-05,
+      "loss": 3.0019,
+      "step": 35165
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.737314816667666e-05,
+      "loss": 2.9243,
+      "step": 35170
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.736159195208835e-05,
+      "loss": 3.0458,
+      "step": 35175
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7350035878831514e-05,
+      "loss": 2.9701,
+      "step": 35180
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.733847994752515e-05,
+      "loss": 2.9925,
+      "step": 35185
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.732692415878825e-05,
+      "loss": 3.0722,
+      "step": 35190
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.731536851323987e-05,
+      "loss": 2.9267,
+      "step": 35195
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.730381301149897e-05,
+      "loss": 3.0272,
+      "step": 35200
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.729225765418458e-05,
+      "loss": 2.9272,
+      "step": 35205
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.728070244191567e-05,
+      "loss": 3.0195,
+      "step": 35210
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.726914737531119e-05,
+      "loss": 2.9429,
+      "step": 35215
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7257592454990156e-05,
+      "loss": 2.9204,
+      "step": 35220
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.724603768157149e-05,
+      "loss": 2.9225,
+      "step": 35225
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.723448305567417e-05,
+      "loss": 3.0154,
+      "step": 35230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7222928577917144e-05,
+      "loss": 2.8696,
+      "step": 35235
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.721137424891932e-05,
+      "loss": 2.9865,
+      "step": 35240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.719982006929967e-05,
+      "loss": 3.0717,
+      "step": 35245
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7188266039677074e-05,
+      "loss": 2.9117,
+      "step": 35250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7176712160670486e-05,
+      "loss": 2.911,
+      "step": 35255
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.71651584328988e-05,
+      "loss": 2.9836,
+      "step": 35260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7153604856980874e-05,
+      "loss": 3.0346,
+      "step": 35265
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7142051433535664e-05,
+      "loss": 3.1592,
+      "step": 35270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.713049816318199e-05,
+      "loss": 3.0554,
+      "step": 35275
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.711894504653878e-05,
+      "loss": 3.0334,
+      "step": 35280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7107392084224864e-05,
+      "loss": 3.1323,
+      "step": 35285
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7095839276859086e-05,
+      "loss": 3.0084,
+      "step": 35290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.708428662506035e-05,
+      "loss": 2.7726,
+      "step": 35295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7072734129447424e-05,
+      "loss": 2.9219,
+      "step": 35300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.706118179063921e-05,
+      "loss": 2.9985,
+      "step": 35305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7049629609254506e-05,
+      "loss": 3.0123,
+      "step": 35310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.7038077585912087e-05,
+      "loss": 2.9881,
+      "step": 35315
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.702652572123082e-05,
+      "loss": 2.9571,
+      "step": 35320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.701497401582945e-05,
+      "loss": 2.9554,
+      "step": 35325
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.700342247032682e-05,
+      "loss": 3.0354,
+      "step": 35330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6991871085341673e-05,
+      "loss": 3.0357,
+      "step": 35335
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.698031986149279e-05,
+      "loss": 3.0637,
+      "step": 35340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.696876879939895e-05,
+      "loss": 2.9981,
+      "step": 35345
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.695721789967887e-05,
+      "loss": 3.0268,
+      "step": 35350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.694566716295135e-05,
+      "loss": 2.965,
+      "step": 35355
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.693411658983508e-05,
+      "loss": 3.0064,
+      "step": 35360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6922566180948807e-05,
+      "loss": 2.8921,
+      "step": 35365
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6911015936911254e-05,
+      "loss": 2.9014,
+      "step": 35370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6899465858341127e-05,
+      "loss": 2.9077,
+      "step": 35375
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6887915945857145e-05,
+      "loss": 3.0157,
+      "step": 35380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.687636620007798e-05,
+      "loss": 2.9178,
+      "step": 35385
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6864816621622324e-05,
+      "loss": 3.0288,
+      "step": 35390
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.685326721110885e-05,
+      "loss": 2.9516,
+      "step": 35395
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6841717969156235e-05,
+      "loss": 2.8864,
+      "step": 35400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.683016889638312e-05,
+      "loss": 3.0191,
+      "step": 35405
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.681861999340818e-05,
+      "loss": 3.0469,
+      "step": 35410
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.680707126085002e-05,
+      "loss": 3.1145,
+      "step": 35415
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.67955226993273e-05,
+      "loss": 2.9755,
+      "step": 35420
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6783974309458615e-05,
+      "loss": 2.9418,
+      "step": 35425
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.67724260918626e-05,
+      "loss": 3.0584,
+      "step": 35430
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.676087804715785e-05,
+      "loss": 3.0545,
+      "step": 35435
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.674933017596295e-05,
+      "loss": 2.9287,
+      "step": 35440
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.67377824788965e-05,
+      "loss": 3.1767,
+      "step": 35445
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.672623495657705e-05,
+      "loss": 2.9736,
+      "step": 35450
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.67146876096232e-05,
+      "loss": 3.0176,
+      "step": 35455
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.670314043865348e-05,
+      "loss": 2.9612,
+      "step": 35460
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.669159344428643e-05,
+      "loss": 2.9852,
+      "step": 35465
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6680046627140625e-05,
+      "loss": 2.9348,
+      "step": 35470
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.666849998783454e-05,
+      "loss": 2.9594,
+      "step": 35475
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.665695352698675e-05,
+      "loss": 2.8817,
+      "step": 35480
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.664540724521573e-05,
+      "loss": 2.9364,
+      "step": 35485
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.663386114313997e-05,
+      "loss": 2.9338,
+      "step": 35490
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.662231522137799e-05,
+      "loss": 3.1053,
+      "step": 35495
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.661076948054823e-05,
+      "loss": 2.9421,
+      "step": 35500
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.659922392126921e-05,
+      "loss": 2.8797,
+      "step": 35505
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6587678544159364e-05,
+      "loss": 3.207,
+      "step": 35510
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6576133349837116e-05,
+      "loss": 3.0465,
+      "step": 35515
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6564588338920954e-05,
+      "loss": 3.0928,
+      "step": 35520
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.655304351202927e-05,
+      "loss": 2.9738,
+      "step": 35525
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.654149886978053e-05,
+      "loss": 3.1027,
+      "step": 35530
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.652995441279311e-05,
+      "loss": 2.8918,
+      "step": 35535
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6518410141685396e-05,
+      "loss": 2.9183,
+      "step": 35540
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.650686605707584e-05,
+      "loss": 2.9355,
+      "step": 35545
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.649532215958275e-05,
+      "loss": 2.8643,
+      "step": 35550
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6483778449824555e-05,
+      "loss": 3.0426,
+      "step": 35555
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.647223492841958e-05,
+      "loss": 3.0615,
+      "step": 35560
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.646069159598621e-05,
+      "loss": 3.0759,
+      "step": 35565
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6449148453142766e-05,
+      "loss": 3.0134,
+      "step": 35570
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.643760550050756e-05,
+      "loss": 3.028,
+      "step": 35575
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.642606273869895e-05,
+      "loss": 3.0084,
+      "step": 35580
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.641452016833521e-05,
+      "loss": 3.0358,
+      "step": 35585
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.640297779003468e-05,
+      "loss": 2.8724,
+      "step": 35590
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6391435604415627e-05,
+      "loss": 3.0989,
+      "step": 35595
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.637989361209631e-05,
+      "loss": 2.9671,
+      "step": 35600
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6368351813695046e-05,
+      "loss": 2.9187,
+      "step": 35605
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.635681020983005e-05,
+      "loss": 3.1479,
+      "step": 35610
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6345268801119605e-05,
+      "loss": 3.0186,
+      "step": 35615
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.633372758818194e-05,
+      "loss": 2.8918,
+      "step": 35620
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6322186571635245e-05,
+      "loss": 2.8913,
+      "step": 35625
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.631064575209779e-05,
+      "loss": 3.0035,
+      "step": 35630
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.629910513018774e-05,
+      "loss": 3.0094,
+      "step": 35635
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6287564706523336e-05,
+      "loss": 3.0521,
+      "step": 35640
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6276024481722736e-05,
+      "loss": 2.8411,
+      "step": 35645
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6264484456404087e-05,
+      "loss": 3.1406,
+      "step": 35650
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.62529446311856e-05,
+      "loss": 2.9934,
+      "step": 35655
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6241405006685395e-05,
+      "loss": 3.078,
+      "step": 35660
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6229865583521655e-05,
+      "loss": 2.9137,
+      "step": 35665
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6218326362312464e-05,
+      "loss": 2.9043,
+      "step": 35670
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.620678734367597e-05,
+      "loss": 2.9672,
+      "step": 35675
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.6195248528230284e-05,
+      "loss": 2.8721,
+      "step": 35680
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.618370991659348e-05,
+      "loss": 3.0472,
+      "step": 35685
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6172171509383685e-05,
+      "loss": 3.0157,
+      "step": 35690
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.616063330721895e-05,
+      "loss": 3.1525,
+      "step": 35695
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.614909531071734e-05,
+      "loss": 2.9701,
+      "step": 35700
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6137557520496914e-05,
+      "loss": 2.9342,
+      "step": 35705
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.612601993717572e-05,
+      "loss": 2.8354,
+      "step": 35710
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6114482561371795e-05,
+      "loss": 2.9817,
+      "step": 35715
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.610294539370317e-05,
+      "loss": 2.8399,
+      "step": 35720
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6091408434787825e-05,
+      "loss": 2.9818,
+      "step": 35725
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.607987168524377e-05,
+      "loss": 2.8971,
+      "step": 35730
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.606833514568901e-05,
+      "loss": 2.9304,
+      "step": 35735
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6056798816741515e-05,
+      "loss": 2.9228,
+      "step": 35740
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.604526269901925e-05,
+      "loss": 3.0004,
+      "step": 35745
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6033726793140155e-05,
+      "loss": 2.8914,
+      "step": 35750
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6022191099722195e-05,
+      "loss": 2.9922,
+      "step": 35755
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.6010655619383275e-05,
+      "loss": 2.9152,
+      "step": 35760
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.599912035274136e-05,
+      "loss": 2.9154,
+      "step": 35765
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.598758530041433e-05,
+      "loss": 3.0781,
+      "step": 35770
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5976050463020056e-05,
+      "loss": 2.9135,
+      "step": 35775
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.596451584117648e-05,
+      "loss": 3.0924,
+      "step": 35780
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.595298143550143e-05,
+      "loss": 3.1263,
+      "step": 35785
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.59414472466128e-05,
+      "loss": 2.8981,
+      "step": 35790
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.592991327512843e-05,
+      "loss": 2.8348,
+      "step": 35795
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.591837952166614e-05,
+      "loss": 2.9695,
+      "step": 35800
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5906845986843794e-05,
+      "loss": 3.05,
+      "step": 35805
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5895312671279174e-05,
+      "loss": 2.9498,
+      "step": 35810
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.588377957559012e-05,
+      "loss": 2.9102,
+      "step": 35815
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.58722467003944e-05,
+      "loss": 2.9897,
+      "step": 35820
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.586071404630978e-05,
+      "loss": 3.01,
+      "step": 35825
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.584918161395408e-05,
+      "loss": 2.9937,
+      "step": 35830
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5837649403945e-05,
+      "loss": 2.9473,
+      "step": 35835
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.582611741690033e-05,
+      "loss": 2.9873,
+      "step": 35840
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.581458565343779e-05,
+      "loss": 2.9679,
+      "step": 35845
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.580305411417507e-05,
+      "loss": 2.9172,
+      "step": 35850
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.579152279972993e-05,
+      "loss": 2.9995,
+      "step": 35855
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5779991710720024e-05,
+      "loss": 3.0768,
+      "step": 35860
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5768460847763074e-05,
+      "loss": 3.014,
+      "step": 35865
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5756930211476735e-05,
+      "loss": 3.0598,
+      "step": 35870
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.574539980247864e-05,
+      "loss": 2.8036,
+      "step": 35875
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5733869621386495e-05,
+      "loss": 2.9495,
+      "step": 35880
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.572233966881788e-05,
+      "loss": 2.9393,
+      "step": 35885
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5710809945390475e-05,
+      "loss": 2.9191,
+      "step": 35890
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.569928045172186e-05,
+      "loss": 2.9308,
+      "step": 35895
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.568775118842962e-05,
+      "loss": 2.9843,
+      "step": 35900
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5676222156131376e-05,
+      "loss": 3.0094,
+      "step": 35905
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.566469335544467e-05,
+      "loss": 2.9443,
+      "step": 35910
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.56531647869871e-05,
+      "loss": 2.916,
+      "step": 35915
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.56416364513762e-05,
+      "loss": 2.9202,
+      "step": 35920
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5630108349229486e-05,
+      "loss": 2.9906,
+      "step": 35925
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.561858048116453e-05,
+      "loss": 2.8824,
+      "step": 35930
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.560705284779879e-05,
+      "loss": 2.8081,
+      "step": 35935
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.559552544974982e-05,
+      "loss": 2.8501,
+      "step": 35940
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.558399828763507e-05,
+      "loss": 2.9514,
+      "step": 35945
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.557247136207201e-05,
+      "loss": 2.9754,
+      "step": 35950
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.556094467367815e-05,
+      "loss": 2.9492,
+      "step": 35955
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5549418223070874e-05,
+      "loss": 2.9907,
+      "step": 35960
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.553789201086768e-05,
+      "loss": 3.0609,
+      "step": 35965
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5526366037685945e-05,
+      "loss": 2.9747,
+      "step": 35970
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.551484030414311e-05,
+      "loss": 3.0065,
+      "step": 35975
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5503314810856556e-05,
+      "loss": 2.852,
+      "step": 35980
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.549178955844368e-05,
+      "loss": 2.9254,
+      "step": 35985
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.548026454752186e-05,
+      "loss": 3.0529,
+      "step": 35990
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.546873977870843e-05,
+      "loss": 3.0621,
+      "step": 35995
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5457215252620756e-05,
+      "loss": 2.8787,
+      "step": 36000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.544569096987617e-05,
+      "loss": 2.8426,
+      "step": 36005
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.543416693109199e-05,
+      "loss": 2.9491,
+      "step": 36010
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.542264313688552e-05,
+      "loss": 3.0156,
+      "step": 36015
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.541111958787407e-05,
+      "loss": 2.9489,
+      "step": 36020
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.539959628467489e-05,
+      "loss": 2.8423,
+      "step": 36025
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.538807322790528e-05,
+      "loss": 3.0058,
+      "step": 36030
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5376550418182475e-05,
+      "loss": 3.0819,
+      "step": 36035
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.536502785612373e-05,
+      "loss": 2.9693,
+      "step": 36040
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.535350554234628e-05,
+      "loss": 3.0829,
+      "step": 36045
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5341983477467296e-05,
+      "loss": 2.8807,
+      "step": 36050
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.533046166210404e-05,
+      "loss": 3.0004,
+      "step": 36055
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.531894009687365e-05,
+      "loss": 3.0284,
+      "step": 36060
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.530741878239334e-05,
+      "loss": 3.0619,
+      "step": 36065
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.529589771928025e-05,
+      "loss": 2.906,
+      "step": 36070
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.528437690815153e-05,
+      "loss": 3.0336,
+      "step": 36075
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5272856349624326e-05,
+      "loss": 2.8775,
+      "step": 36080
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5261336044315725e-05,
+      "loss": 2.9578,
+      "step": 36085
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.524981599284289e-05,
+      "loss": 3.0527,
+      "step": 36090
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5238296195822876e-05,
+      "loss": 2.7472,
+      "step": 36095
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.522677665387276e-05,
+      "loss": 2.9166,
+      "step": 36100
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.521525736760964e-05,
+      "loss": 3.0755,
+      "step": 36105
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.520373833765052e-05,
+      "loss": 2.9672,
+      "step": 36110
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5192219564612494e-05,
+      "loss": 2.879,
+      "step": 36115
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5180701049112565e-05,
+      "loss": 2.9826,
+      "step": 36120
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.516918279176772e-05,
+      "loss": 2.9292,
+      "step": 36125
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5157664793194996e-05,
+      "loss": 2.9383,
+      "step": 36130
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.514614705401134e-05,
+      "loss": 2.9582,
+      "step": 36135
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5134629574833766e-05,
+      "loss": 3.0576,
+      "step": 36140
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.512311235627921e-05,
+      "loss": 2.989,
+      "step": 36145
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5111595398964576e-05,
+      "loss": 2.8808,
+      "step": 36150
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.510007870350685e-05,
+      "loss": 2.8502,
+      "step": 36155
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.508856227052291e-05,
+      "loss": 3.0222,
+      "step": 36160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.5077046100629684e-05,
+      "loss": 3.0009,
+      "step": 36165
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.506553019444404e-05,
+      "loss": 2.9594,
+      "step": 36170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.505401455258283e-05,
+      "loss": 3.0506,
+      "step": 36175
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.504249917566297e-05,
+      "loss": 2.9289,
+      "step": 36180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.503098406430123e-05,
+      "loss": 2.8252,
+      "step": 36185
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.501946921911451e-05,
+      "loss": 3.0045,
+      "step": 36190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.500795464071958e-05,
+      "loss": 2.8378,
+      "step": 36195
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.4996440329733234e-05,
+      "loss": 2.8967,
+      "step": 36200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.49849262867723e-05,
+      "loss": 3.0202,
+      "step": 36205
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.497341251245351e-05,
+      "loss": 3.0109,
+      "step": 36210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.496189900739366e-05,
+      "loss": 3.0367,
+      "step": 36215
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.4950385772209467e-05,
+      "loss": 3.1611,
+      "step": 36220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.493887280751765e-05,
+      "loss": 2.8921,
+      "step": 36225
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.492736011393495e-05,
+      "loss": 3.0802,
+      "step": 36230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.491584769207804e-05,
+      "loss": 3.0383,
+      "step": 36235
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.490433554256363e-05,
+      "loss": 2.9686,
+      "step": 36240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.4892823666008374e-05,
+      "loss": 2.965,
+      "step": 36245
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.4881312063028954e-05,
+      "loss": 3.0607,
+      "step": 36250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.486980073424199e-05,
+      "loss": 2.9663,
+      "step": 36255
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.4858289680264087e-05,
+      "loss": 2.9609,
+      "step": 36260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.4846778901711904e-05,
+      "loss": 2.8704,
+      "step": 36265
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.483526839920199e-05,
+      "loss": 3.2059,
+      "step": 36270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.4823758173350986e-05,
+      "loss": 2.9978,
+      "step": 36275
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.481224822477541e-05,
+      "loss": 2.9685,
+      "step": 36280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.480073855409182e-05,
+      "loss": 2.9333,
+      "step": 36285
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.4789229161916786e-05,
+      "loss": 2.9295,
+      "step": 36290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.4777720048866786e-05,
+      "loss": 3.0037,
+      "step": 36295
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.476621121555837e-05,
+      "loss": 2.8659,
+      "step": 36300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.475470266260801e-05,
+      "loss": 2.9657,
+      "step": 36305
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.4743194390632174e-05,
+      "loss": 2.955,
+      "step": 36310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.473168640024734e-05,
+      "loss": 3.09,
+      "step": 36315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.4720178692069956e-05,
+      "loss": 2.9714,
+      "step": 36320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.4708671266716453e-05,
+      "loss": 3.0104,
+      "step": 36325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.469716412480326e-05,
+      "loss": 3.1678,
+      "step": 36330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.468565726694676e-05,
+      "loss": 2.9849,
+      "step": 36335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.467415069376334e-05,
+      "loss": 2.9878,
+      "step": 36340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.466264440586938e-05,
+      "loss": 2.9194,
+      "step": 36345
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.4651138403881234e-05,
+      "loss": 3.0277,
+      "step": 36350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.463963268841526e-05,
+      "loss": 2.9463,
+      "step": 36355
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.462812726008775e-05,
+      "loss": 3.0782,
+      "step": 36360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.461662211951505e-05,
+      "loss": 2.9607,
+      "step": 36365
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.460511726731343e-05,
+      "loss": 2.7744,
+      "step": 36370
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.459361270409918e-05,
+      "loss": 2.9219,
+      "step": 36375
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4582108430488574e-05,
+      "loss": 2.966,
+      "step": 36380
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4570604447097824e-05,
+      "loss": 3.0584,
+      "step": 36385
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4559100754543216e-05,
+      "loss": 3.0074,
+      "step": 36390
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4547597353440915e-05,
+      "loss": 2.8579,
+      "step": 36395
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4536094244407175e-05,
+      "loss": 2.9323,
+      "step": 36400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.452459142805815e-05,
+      "loss": 3.0132,
+      "step": 36405
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4513088905009994e-05,
+      "loss": 2.9227,
+      "step": 36410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.450158667587891e-05,
+      "loss": 2.8004,
+      "step": 36415
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.449008474128099e-05,
+      "loss": 3.0775,
+      "step": 36420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.44785831018324e-05,
+      "loss": 3.0501,
+      "step": 36425
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.446708175814923e-05,
+      "loss": 2.8207,
+      "step": 36430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4455580710847546e-05,
+      "loss": 2.921,
+      "step": 36435
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.444407996054346e-05,
+      "loss": 3.0085,
+      "step": 36440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4432579507853004e-05,
+      "loss": 3.0251,
+      "step": 36445
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.442107935339226e-05,
+      "loss": 2.9324,
+      "step": 36450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.440957949777723e-05,
+      "loss": 2.992,
+      "step": 36455
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.439807994162391e-05,
+      "loss": 3.0534,
+      "step": 36460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4386580685548337e-05,
+      "loss": 3.0982,
+      "step": 36465
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4375081730166446e-05,
+      "loss": 2.9387,
+      "step": 36470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.436358307609424e-05,
+      "loss": 3.0122,
+      "step": 36475
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.435208472394766e-05,
+      "loss": 2.9863,
+      "step": 36480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4340586674342596e-05,
+      "loss": 2.8817,
+      "step": 36485
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4329088927895025e-05,
+      "loss": 3.0291,
+      "step": 36490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.43175914852208e-05,
+      "loss": 2.916,
+      "step": 36495
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.430609434693583e-05,
+      "loss": 2.9611,
+      "step": 36500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.429459751365598e-05,
+      "loss": 3.0233,
+      "step": 36505
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.428310098599707e-05,
+      "loss": 3.0012,
+      "step": 36510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.427160476457497e-05,
+      "loss": 3.0746,
+      "step": 36515
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4260108850005466e-05,
+      "loss": 2.8767,
+      "step": 36520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.42486132429044e-05,
+      "loss": 2.9635,
+      "step": 36525
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4237117943887536e-05,
+      "loss": 3.0132,
+      "step": 36530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.422562295357061e-05,
+      "loss": 2.9484,
+      "step": 36535
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.421412827256943e-05,
+      "loss": 3.0833,
+      "step": 36540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.420263390149968e-05,
+      "loss": 3.0992,
+      "step": 36545
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.419113984097713e-05,
+      "loss": 3.1152,
+      "step": 36550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.417964609161744e-05,
+      "loss": 2.8615,
+      "step": 36555
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.416815265403629e-05,
+      "loss": 2.9233,
+      "step": 36560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.415665952884939e-05,
+      "loss": 2.9414,
+      "step": 36565
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4145166716672346e-05,
+      "loss": 2.9968,
+      "step": 36570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.413367421812084e-05,
+      "loss": 2.9699,
+      "step": 36575
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.412218203381046e-05,
+      "loss": 3.0179,
+      "step": 36580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.41106901643568e-05,
+      "loss": 2.8829,
+      "step": 36585
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4099198610375466e-05,
+      "loss": 2.8897,
+      "step": 36590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.408770737248201e-05,
+      "loss": 2.9547,
+      "step": 36595
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4076216451292e-05,
+      "loss": 2.9685,
+      "step": 36600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.406472584742094e-05,
+      "loss": 2.924,
+      "step": 36605
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.4053235561484374e-05,
+      "loss": 2.9596,
+      "step": 36610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.404174559409779e-05,
+      "loss": 2.9925,
+      "step": 36615
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.403025594587667e-05,
+      "loss": 3.0715,
+      "step": 36620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.401876661743649e-05,
+      "loss": 2.9246,
+      "step": 36625
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.400727760939269e-05,
+      "loss": 3.005,
+      "step": 36630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3995788922360695e-05,
+      "loss": 2.8628,
+      "step": 36635
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.398430055695592e-05,
+      "loss": 2.9471,
+      "step": 36640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.397281251379377e-05,
+      "loss": 3.1572,
+      "step": 36645
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.396132479348962e-05,
+      "loss": 2.9607,
+      "step": 36650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3949837396658846e-05,
+      "loss": 2.8587,
+      "step": 36655
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.393835032391675e-05,
+      "loss": 2.9431,
+      "step": 36660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.392686357587872e-05,
+      "loss": 3.0676,
+      "step": 36665
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3915377153160017e-05,
+      "loss": 3.1119,
+      "step": 36670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.390389105637595e-05,
+      "loss": 2.9247,
+      "step": 36675
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.389240528614181e-05,
+      "loss": 2.9023,
+      "step": 36680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.388091984307282e-05,
+      "loss": 3.0999,
+      "step": 36685
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.386943472778426e-05,
+      "loss": 2.8554,
+      "step": 36690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3857949940891306e-05,
+      "loss": 3.0252,
+      "step": 36695
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3846465483009214e-05,
+      "loss": 3.099,
+      "step": 36700
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3834981354753144e-05,
+      "loss": 3.0216,
+      "step": 36705
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.382349755673825e-05,
+      "loss": 2.9352,
+      "step": 36710
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.381201408957973e-05,
+      "loss": 3.026,
+      "step": 36715
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.380053095389266e-05,
+      "loss": 3.1341,
+      "step": 36720
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.378904815029221e-05,
+      "loss": 2.9437,
+      "step": 36725
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3777565679393455e-05,
+      "loss": 3.0599,
+      "step": 36730
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3766083541811456e-05,
+      "loss": 3.101,
+      "step": 36735
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3754601738161315e-05,
+      "loss": 2.8491,
+      "step": 36740
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.374312026905804e-05,
+      "loss": 3.051,
+      "step": 36745
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.37316391351167e-05,
+      "loss": 2.9666,
+      "step": 36750
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3720158336952275e-05,
+      "loss": 2.833,
+      "step": 36755
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.370867787517975e-05,
+      "loss": 2.9236,
+      "step": 36760
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.369719775041412e-05,
+      "loss": 2.803,
+      "step": 36765
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.368571796327032e-05,
+      "loss": 2.9941,
+      "step": 36770
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3674238514363305e-05,
+      "loss": 2.9651,
+      "step": 36775
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3662759404308e-05,
+      "loss": 2.9992,
+      "step": 36780
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3651280633719265e-05,
+      "loss": 2.9962,
+      "step": 36785
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3639802203212034e-05,
+      "loss": 2.8468,
+      "step": 36790
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3628324113401126e-05,
+      "loss": 3.0151,
+      "step": 36795
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.361684636490143e-05,
+      "loss": 3.0065,
+      "step": 36800
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.360536895832775e-05,
+      "loss": 2.8303,
+      "step": 36805
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.359389189429488e-05,
+      "loss": 3.0192,
+      "step": 36810
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.358241517341765e-05,
+      "loss": 2.9026,
+      "step": 36815
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.357093879631078e-05,
+      "loss": 2.9267,
+      "step": 36820
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3559462763589094e-05,
+      "loss": 2.8382,
+      "step": 36825
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.354798707586728e-05,
+      "loss": 2.9241,
+      "step": 36830
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.353651173376005e-05,
+      "loss": 2.9543,
+      "step": 36835
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.352503673788213e-05,
+      "loss": 2.9201,
+      "step": 36840
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.351356208884818e-05,
+      "loss": 2.9064,
+      "step": 36845
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.350208778727289e-05,
+      "loss": 3.0353,
+      "step": 36850
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.349061383377089e-05,
+      "loss": 2.889,
+      "step": 36855
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3479140228956764e-05,
+      "loss": 3.0072,
+      "step": 36860
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3467666973445185e-05,
+      "loss": 3.0501,
+      "step": 36865
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.345619406785069e-05,
+      "loss": 3.0077,
+      "step": 36870
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.344472151278787e-05,
+      "loss": 2.9884,
+      "step": 36875
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.343324930887127e-05,
+      "loss": 2.9976,
+      "step": 36880
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3421777456715424e-05,
+      "loss": 2.9588,
+      "step": 36885
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.341030595693485e-05,
+      "loss": 3.0755,
+      "step": 36890
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3398834810144014e-05,
+      "loss": 2.9458,
+      "step": 36895
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.338736401695742e-05,
+      "loss": 2.98,
+      "step": 36900
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.33758935779895e-05,
+      "loss": 2.8721,
+      "step": 36905
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3364423493854724e-05,
+      "loss": 2.8992,
+      "step": 36910
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.335295376516747e-05,
+      "loss": 3.026,
+      "step": 36915
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.334148439254216e-05,
+      "loss": 2.9828,
+      "step": 36920
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.333001537659317e-05,
+      "loss": 2.9578,
+      "step": 36925
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.331854671793485e-05,
+      "loss": 3.0067,
+      "step": 36930
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.330707841718156e-05,
+      "loss": 3.039,
+      "step": 36935
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.32956104749476e-05,
+      "loss": 2.9301,
+      "step": 36940
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.328414289184728e-05,
+      "loss": 3.0665,
+      "step": 36945
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.327267566849489e-05,
+      "loss": 3.04,
+      "step": 36950
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.326120880550469e-05,
+      "loss": 3.0104,
+      "step": 36955
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3249742303490924e-05,
+      "loss": 3.0627,
+      "step": 36960
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.323827616306783e-05,
+      "loss": 3.0801,
+      "step": 36965
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.322681038484958e-05,
+      "loss": 2.9331,
+      "step": 36970
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3215344969450404e-05,
+      "loss": 2.8684,
+      "step": 36975
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.320387991748444e-05,
+      "loss": 3.0059,
+      "step": 36980
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.319241522956584e-05,
+      "loss": 3.0147,
+      "step": 36985
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3180950906308755e-05,
+      "loss": 3.0255,
+      "step": 36990
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.316948694832724e-05,
+      "loss": 2.9839,
+      "step": 36995
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.315802335623545e-05,
+      "loss": 3.0941,
+      "step": 37000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.314656013064741e-05,
+      "loss": 3.0683,
+      "step": 37005
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.31350972721772e-05,
+      "loss": 2.9182,
+      "step": 37010
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.312363478143882e-05,
+      "loss": 2.9637,
+      "step": 37015
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.3112172659046284e-05,
+      "loss": 2.9811,
+      "step": 37020
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.310071090561362e-05,
+      "loss": 2.8981,
+      "step": 37025
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.308924952175474e-05,
+      "loss": 2.9311,
+      "step": 37030
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.307778850808366e-05,
+      "loss": 3.0326,
+      "step": 37035
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.306632786521427e-05,
+      "loss": 2.9358,
+      "step": 37040
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.305486759376047e-05,
+      "loss": 3.0051,
+      "step": 37045
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3043407694336205e-05,
+      "loss": 3.0807,
+      "step": 37050
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.303194816755528e-05,
+      "loss": 2.8706,
+      "step": 37055
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3020489014031605e-05,
+      "loss": 2.8418,
+      "step": 37060
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.300903023437899e-05,
+      "loss": 3.0153,
+      "step": 37065
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.299757182921122e-05,
+      "loss": 2.8726,
+      "step": 37070
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.298611379914214e-05,
+      "loss": 2.9114,
+      "step": 37075
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.297465614478547e-05,
+      "loss": 3.0299,
+      "step": 37080
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.296319886675501e-05,
+      "loss": 2.9996,
+      "step": 37085
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2951741965664455e-05,
+      "loss": 2.8394,
+      "step": 37090
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.294028544212751e-05,
+      "loss": 3.0815,
+      "step": 37095
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2928829296757906e-05,
+      "loss": 2.9185,
+      "step": 37100
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.291737353016926e-05,
+      "loss": 2.9086,
+      "step": 37105
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.290591814297529e-05,
+      "loss": 2.9348,
+      "step": 37110
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.289446313578958e-05,
+      "loss": 2.9835,
+      "step": 37115
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.288300850922573e-05,
+      "loss": 3.0054,
+      "step": 37120
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.287155426389737e-05,
+      "loss": 3.1204,
+      "step": 37125
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.286010040041803e-05,
+      "loss": 2.9795,
+      "step": 37130
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.284864691940128e-05,
+      "loss": 2.8655,
+      "step": 37135
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2837193821460655e-05,
+      "loss": 2.8155,
+      "step": 37140
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.282574110720963e-05,
+      "loss": 2.8796,
+      "step": 37145
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2814288777261734e-05,
+      "loss": 2.9777,
+      "step": 37150
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2802836832230384e-05,
+      "loss": 2.9243,
+      "step": 37155
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.279138527272908e-05,
+      "loss": 3.0608,
+      "step": 37160
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.277993409937122e-05,
+      "loss": 3.018,
+      "step": 37165
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.276848331277019e-05,
+      "loss": 3.1141,
+      "step": 37170
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.275703291353941e-05,
+      "loss": 2.999,
+      "step": 37175
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.274558290229221e-05,
+      "loss": 3.0041,
+      "step": 37180
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.273413327964196e-05,
+      "loss": 2.9546,
+      "step": 37185
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.272268404620197e-05,
+      "loss": 3.0,
+      "step": 37190
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.271123520258553e-05,
+      "loss": 2.9624,
+      "step": 37195
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2699786749405945e-05,
+      "loss": 2.9247,
+      "step": 37200
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.268833868727643e-05,
+      "loss": 2.9275,
+      "step": 37205
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.267689101681028e-05,
+      "loss": 3.0437,
+      "step": 37210
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.266544373862066e-05,
+      "loss": 3.0708,
+      "step": 37215
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2653996853320796e-05,
+      "loss": 3.0488,
+      "step": 37220
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.264255036152385e-05,
+      "loss": 3.1312,
+      "step": 37225
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.263110426384298e-05,
+      "loss": 2.9479,
+      "step": 37230
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2619658560891326e-05,
+      "loss": 2.9251,
+      "step": 37235
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.260821325328199e-05,
+      "loss": 3.0727,
+      "step": 37240
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.259676834162806e-05,
+      "loss": 3.0313,
+      "step": 37245
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.25853238265426e-05,
+      "loss": 2.9775,
+      "step": 37250
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.257387970863868e-05,
+      "loss": 3.1006,
+      "step": 37255
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2562435988529306e-05,
+      "loss": 2.9085,
+      "step": 37260
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.255099266682749e-05,
+      "loss": 3.0303,
+      "step": 37265
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.253954974414621e-05,
+      "loss": 2.8988,
+      "step": 37270
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2528107221098445e-05,
+      "loss": 3.0273,
+      "step": 37275
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.251666509829711e-05,
+      "loss": 2.8535,
+      "step": 37280
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2505223376355153e-05,
+      "loss": 2.8089,
+      "step": 37285
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.249378205588546e-05,
+      "loss": 3.0752,
+      "step": 37290
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2482341137500895e-05,
+      "loss": 3.0481,
+      "step": 37295
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.247090062181433e-05,
+      "loss": 3.0411,
+      "step": 37300
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.245946050943858e-05,
+      "loss": 2.9466,
+      "step": 37305
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.244802080098649e-05,
+      "loss": 3.0523,
+      "step": 37310
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2436581497070823e-05,
+      "loss": 2.995,
+      "step": 37315
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.242514259830434e-05,
+      "loss": 2.9801,
+      "step": 37320
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2413704105299824e-05,
+      "loss": 2.9218,
+      "step": 37325
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.240226601866995e-05,
+      "loss": 2.9019,
+      "step": 37330
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.239082833902747e-05,
+      "loss": 2.9785,
+      "step": 37335
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2379391066985044e-05,
+      "loss": 2.9273,
+      "step": 37340
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.236795420315532e-05,
+      "loss": 2.8176,
+      "step": 37345
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.235651774815096e-05,
+      "loss": 2.9001,
+      "step": 37350
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2345081702584546e-05,
+      "loss": 2.9897,
+      "step": 37355
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.233364606706872e-05,
+      "loss": 2.9964,
+      "step": 37360
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2322210842216034e-05,
+      "loss": 2.9866,
+      "step": 37365
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2310776028639e-05,
+      "loss": 2.942,
+      "step": 37370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.229934162695021e-05,
+      "loss": 3.0139,
+      "step": 37375
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2287907637762105e-05,
+      "loss": 2.9708,
+      "step": 37380
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2276474061687224e-05,
+      "loss": 3.0456,
+      "step": 37385
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2265040899338017e-05,
+      "loss": 3.0272,
+      "step": 37390
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.225360815132688e-05,
+      "loss": 2.9905,
+      "step": 37395
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.22421758182663e-05,
+      "loss": 2.9872,
+      "step": 37400
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.22307439007686e-05,
+      "loss": 2.9587,
+      "step": 37405
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2219312399446207e-05,
+      "loss": 3.0268,
+      "step": 37410
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.220788131491146e-05,
+      "loss": 3.2118,
+      "step": 37415
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2196450647776664e-05,
+      "loss": 2.8796,
+      "step": 37420
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.218502039865415e-05,
+      "loss": 3.0503,
+      "step": 37425
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.217359056815618e-05,
+      "loss": 3.0221,
+      "step": 37430
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2162161156895046e-05,
+      "loss": 3.0044,
+      "step": 37435
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.215073216548297e-05,
+      "loss": 2.983,
+      "step": 37440
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.213930359453214e-05,
+      "loss": 3.0043,
+      "step": 37445
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.212787544465481e-05,
+      "loss": 3.0575,
+      "step": 37450
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.211644771646308e-05,
+      "loss": 2.8779,
+      "step": 37455
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.210502041056916e-05,
+      "loss": 2.7936,
+      "step": 37460
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2093593527585165e-05,
+      "loss": 3.0551,
+      "step": 37465
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2082167068123145e-05,
+      "loss": 3.0822,
+      "step": 37470
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.207074103279525e-05,
+      "loss": 3.0447,
+      "step": 37475
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2059315422213486e-05,
+      "loss": 2.9733,
+      "step": 37480
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2047890236989926e-05,
+      "loss": 3.1556,
+      "step": 37485
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.2036465477736564e-05,
+      "loss": 2.9142,
+      "step": 37490
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.202504114506538e-05,
+      "loss": 2.967,
+      "step": 37495
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.201361723958837e-05,
+      "loss": 2.9194,
+      "step": 37500
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.200219376191744e-05,
+      "loss": 2.924,
+      "step": 37505
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.199077071266454e-05,
+      "loss": 2.9352,
+      "step": 37510
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.197934809244156e-05,
+      "loss": 3.0546,
+      "step": 37515
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.196792590186037e-05,
+      "loss": 2.9808,
+      "step": 37520
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1956504141532824e-05,
+      "loss": 2.9391,
+      "step": 37525
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.194508281207075e-05,
+      "loss": 2.9563,
+      "step": 37530
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.193366191408597e-05,
+      "loss": 2.862,
+      "step": 37535
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1922241448190234e-05,
+      "loss": 3.0899,
+      "step": 37540
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1910821414995325e-05,
+      "loss": 3.0588,
+      "step": 37545
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.189940181511297e-05,
+      "loss": 3.0303,
+      "step": 37550
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.188798264915489e-05,
+      "loss": 2.9793,
+      "step": 37555
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.187656391773277e-05,
+      "loss": 3.0004,
+      "step": 37560
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.186514562145827e-05,
+      "loss": 3.0443,
+      "step": 37565
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.185372776094305e-05,
+      "loss": 3.0085,
+      "step": 37570
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1842310336798716e-05,
+      "loss": 2.8023,
+      "step": 37575
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.183089334963687e-05,
+      "loss": 2.9557,
+      "step": 37580
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.181947680006908e-05,
+      "loss": 2.9681,
+      "step": 37585
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.18080606887069e-05,
+      "loss": 2.9575,
+      "step": 37590
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.179664501616185e-05,
+      "loss": 2.9638,
+      "step": 37595
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1785229783045456e-05,
+      "loss": 3.0295,
+      "step": 37600
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.177381498996915e-05,
+      "loss": 2.9547,
+      "step": 37605
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1762400637544445e-05,
+      "loss": 2.7942,
+      "step": 37610
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.175098672638272e-05,
+      "loss": 2.9131,
+      "step": 37615
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1739573257095434e-05,
+      "loss": 3.1731,
+      "step": 37620
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.172816023029394e-05,
+      "loss": 2.8972,
+      "step": 37625
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.171674764658958e-05,
+      "loss": 3.0159,
+      "step": 37630
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.170533550659375e-05,
+      "loss": 2.93,
+      "step": 37635
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.16939238109177e-05,
+      "loss": 2.8698,
+      "step": 37640
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1682512560172763e-05,
+      "loss": 2.9792,
+      "step": 37645
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.16711017549702e-05,
+      "loss": 2.9385,
+      "step": 37650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1659691395921217e-05,
+      "loss": 2.9357,
+      "step": 37655
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1648281483637084e-05,
+      "loss": 3.0582,
+      "step": 37660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1636872018728947e-05,
+      "loss": 3.0751,
+      "step": 37665
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.162546300180802e-05,
+      "loss": 2.8956,
+      "step": 37670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.161405443348543e-05,
+      "loss": 3.1188,
+      "step": 37675
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1602646314372266e-05,
+      "loss": 2.9352,
+      "step": 37680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1591238645079686e-05,
+      "loss": 3.0537,
+      "step": 37685
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.157983142621872e-05,
+      "loss": 3.038,
+      "step": 37690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1568424658400444e-05,
+      "loss": 2.9734,
+      "step": 37695
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1557018342235874e-05,
+      "loss": 3.056,
+      "step": 37700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1545612478335996e-05,
+      "loss": 2.9295,
+      "step": 37705
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1534207067311816e-05,
+      "loss": 2.9292,
+      "step": 37710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.152280210977425e-05,
+      "loss": 2.94,
+      "step": 37715
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.1511397606334273e-05,
+      "loss": 2.9401,
+      "step": 37720
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1499993557602765e-05,
+      "loss": 2.9602,
+      "step": 37725
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1488589964190585e-05,
+      "loss": 2.9679,
+      "step": 37730
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1477186826708634e-05,
+      "loss": 2.867,
+      "step": 37735
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.14657841457677e-05,
+      "loss": 2.9537,
+      "step": 37740
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1454381921978614e-05,
+      "loss": 2.9842,
+      "step": 37745
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1442980155952164e-05,
+      "loss": 3.1356,
+      "step": 37750
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1431578848299077e-05,
+      "loss": 2.9271,
+      "step": 37755
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1420177999630125e-05,
+      "loss": 2.8589,
+      "step": 37760
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1408777610555985e-05,
+      "loss": 3.0385,
+      "step": 37765
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.139737768168737e-05,
+      "loss": 2.8512,
+      "step": 37770
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1385978213634916e-05,
+      "loss": 3.0768,
+      "step": 37775
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1374579207009246e-05,
+      "loss": 2.9924,
+      "step": 37780
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.136318066242101e-05,
+      "loss": 3.0893,
+      "step": 37785
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1351782580480756e-05,
+      "loss": 2.9908,
+      "step": 37790
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.134038496179907e-05,
+      "loss": 2.9107,
+      "step": 37795
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1328987806986484e-05,
+      "loss": 2.9875,
+      "step": 37800
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.131759111665349e-05,
+      "loss": 2.9687,
+      "step": 37805
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.13061948914106e-05,
+      "loss": 3.0201,
+      "step": 37810
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.129479913186823e-05,
+      "loss": 3.0765,
+      "step": 37815
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.128340383863688e-05,
+      "loss": 2.9885,
+      "step": 37820
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1272009012326916e-05,
+      "loss": 2.9125,
+      "step": 37825
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.126061465354874e-05,
+      "loss": 2.9665,
+      "step": 37830
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.12492207629127e-05,
+      "loss": 2.8859,
+      "step": 37835
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.123782734102915e-05,
+      "loss": 2.9216,
+      "step": 37840
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1226434388508394e-05,
+      "loss": 3.0268,
+      "step": 37845
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1215041905960714e-05,
+      "loss": 2.824,
+      "step": 37850
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1203649893996364e-05,
+      "loss": 2.9588,
+      "step": 37855
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.119225835322559e-05,
+      "loss": 2.9831,
+      "step": 37860
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.11808672842586e-05,
+      "loss": 2.8898,
+      "step": 37865
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1169476687705576e-05,
+      "loss": 3.111,
+      "step": 37870
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.115808656417669e-05,
+      "loss": 2.9985,
+      "step": 37875
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.114669691428206e-05,
+      "loss": 2.9269,
+      "step": 37880
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.113530773863179e-05,
+      "loss": 3.1019,
+      "step": 37885
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.112391903783598e-05,
+      "loss": 3.0282,
+      "step": 37890
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.111253081250468e-05,
+      "loss": 2.9751,
+      "step": 37895
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.110114306324793e-05,
+      "loss": 2.9681,
+      "step": 37900
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1089755790675716e-05,
+      "loss": 2.9558,
+      "step": 37905
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.1078368995398055e-05,
+      "loss": 2.8659,
+      "step": 37910
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.106698267802486e-05,
+      "loss": 2.9332,
+      "step": 37915
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.10555968391661e-05,
+      "loss": 3.0192,
+      "step": 37920
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.104421147943166e-05,
+      "loss": 3.0766,
+      "step": 37925
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.103282659943141e-05,
+      "loss": 2.9033,
+      "step": 37930
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.102144219977522e-05,
+      "loss": 3.0883,
+      "step": 37935
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.10100582810729e-05,
+      "loss": 3.1615,
+      "step": 37940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.099867484393429e-05,
+      "loss": 3.0271,
+      "step": 37945
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.098729188896914e-05,
+      "loss": 2.8922,
+      "step": 37950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.097590941678717e-05,
+      "loss": 3.0411,
+      "step": 37955
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0964527427998157e-05,
+      "loss": 2.9667,
+      "step": 37960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.095314592321175e-05,
+      "loss": 3.0018,
+      "step": 37965
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0941764903037685e-05,
+      "loss": 2.9681,
+      "step": 37970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.093038436808556e-05,
+      "loss": 2.9965,
+      "step": 37975
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0919004318964985e-05,
+      "loss": 2.9576,
+      "step": 37980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.09076247562856e-05,
+      "loss": 2.9318,
+      "step": 37985
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.089624568065693e-05,
+      "loss": 3.0233,
+      "step": 37990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0884867092688554e-05,
+      "loss": 2.8693,
+      "step": 37995
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.087348899298997e-05,
+      "loss": 2.8774,
+      "step": 38000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0862111382170654e-05,
+      "loss": 2.8697,
+      "step": 38005
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.08507342608401e-05,
+      "loss": 3.0321,
+      "step": 38010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.083935762960771e-05,
+      "loss": 2.9525,
+      "step": 38015
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.082798148908294e-05,
+      "loss": 2.9518,
+      "step": 38020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0816605839875145e-05,
+      "loss": 2.8457,
+      "step": 38025
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.080523068259367e-05,
+      "loss": 3.0223,
+      "step": 38030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0793856017847894e-05,
+      "loss": 2.9305,
+      "step": 38035
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.078248184624707e-05,
+      "loss": 2.9528,
+      "step": 38040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.077110816840052e-05,
+      "loss": 2.9951,
+      "step": 38045
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0759734984917485e-05,
+      "loss": 2.8378,
+      "step": 38050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0748362296407166e-05,
+      "loss": 2.9943,
+      "step": 38055
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.073699010347881e-05,
+      "loss": 2.9641,
+      "step": 38060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.072561840674153e-05,
+      "loss": 3.0395,
+      "step": 38065
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.071424720680454e-05,
+      "loss": 3.0196,
+      "step": 38070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0702876504276925e-05,
+      "loss": 2.9689,
+      "step": 38075
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0691506299767763e-05,
+      "loss": 2.9596,
+      "step": 38080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.068013659388616e-05,
+      "loss": 2.9903,
+      "step": 38085
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.066876738724111e-05,
+      "loss": 3.0002,
+      "step": 38090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.065739868044168e-05,
+      "loss": 3.0223,
+      "step": 38095
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0646030474096835e-05,
+      "loss": 2.9377,
+      "step": 38100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.06346627688155e-05,
+      "loss": 2.9398,
+      "step": 38105
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.062329556520666e-05,
+      "loss": 2.9685,
+      "step": 38110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0611928863879186e-05,
+      "loss": 3.0016,
+      "step": 38115
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.060056266544199e-05,
+      "loss": 3.0985,
+      "step": 38120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.058919697050389e-05,
+      "loss": 2.8351,
+      "step": 38125
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0577831779673724e-05,
+      "loss": 2.9402,
+      "step": 38130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.05664670935603e-05,
+      "loss": 2.9013,
+      "step": 38135
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.055510291277238e-05,
+      "loss": 2.9542,
+      "step": 38140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0543739237918723e-05,
+      "loss": 3.0051,
+      "step": 38145
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.053237606960803e-05,
+      "loss": 2.903,
+      "step": 38150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.052101340844899e-05,
+      "loss": 2.9133,
+      "step": 38155
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0509651255050276e-05,
+      "loss": 3.005,
+      "step": 38160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.049828961002051e-05,
+      "loss": 3.0332,
+      "step": 38165
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0486928473968314e-05,
+      "loss": 2.9659,
+      "step": 38170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0475567847502284e-05,
+      "loss": 3.082,
+      "step": 38175
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.046420773123093e-05,
+      "loss": 3.008,
+      "step": 38180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.045284812576282e-05,
+      "loss": 2.9982,
+      "step": 38185
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0441489031706434e-05,
+      "loss": 2.8817,
+      "step": 38190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.043013044967024e-05,
+      "loss": 3.0,
+      "step": 38195
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.041877238026271e-05,
+      "loss": 2.9422,
+      "step": 38200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.040741482409222e-05,
+      "loss": 3.1553,
+      "step": 38205
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0396057781767195e-05,
+      "loss": 2.8981,
+      "step": 38210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.038470125389596e-05,
+      "loss": 2.9659,
+      "step": 38215
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.037334524108689e-05,
+      "loss": 2.8563,
+      "step": 38220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.036198974394827e-05,
+      "loss": 2.997,
+      "step": 38225
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.035063476308839e-05,
+      "loss": 3.0811,
+      "step": 38230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.03392802991155e-05,
+      "loss": 3.0588,
+      "step": 38235
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0327926352637795e-05,
+      "loss": 2.9639,
+      "step": 38240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.031657292426352e-05,
+      "loss": 3.0502,
+      "step": 38245
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.030522001460079e-05,
+      "loss": 2.9164,
+      "step": 38250
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.02938676242578e-05,
+      "loss": 3.0998,
+      "step": 38255
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0282515753842633e-05,
+      "loss": 2.9624,
+      "step": 38260
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.027116440396337e-05,
+      "loss": 3.0254,
+      "step": 38265
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.025981357522808e-05,
+      "loss": 3.0851,
+      "step": 38270
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0248463268244775e-05,
+      "loss": 3.0595,
+      "step": 38275
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.023711348362149e-05,
+      "loss": 3.0064,
+      "step": 38280
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0225764221966175e-05,
+      "loss": 3.0544,
+      "step": 38285
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.021441548388676e-05,
+      "loss": 3.0567,
+      "step": 38290
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.02030672699912e-05,
+      "loss": 2.9139,
+      "step": 38295
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0191719580887344e-05,
+      "loss": 2.8745,
+      "step": 38300
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0180372417183096e-05,
+      "loss": 2.9638,
+      "step": 38305
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.016902577948626e-05,
+      "loss": 2.9386,
+      "step": 38310
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0157679668404635e-05,
+      "loss": 3.0123,
+      "step": 38315
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.014633408454602e-05,
+      "loss": 2.9684,
+      "step": 38320
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.013498902851814e-05,
+      "loss": 2.8421,
+      "step": 38325
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.012364450092874e-05,
+      "loss": 3.0649,
+      "step": 38330
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0112300502385506e-05,
+      "loss": 3.0495,
+      "step": 38335
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.010095703349607e-05,
+      "loss": 2.9143,
+      "step": 38340
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.008961409486811e-05,
+      "loss": 2.9275,
+      "step": 38345
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.007827168710919e-05,
+      "loss": 2.9698,
+      "step": 38350
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0066929810826944e-05,
+      "loss": 3.0272,
+      "step": 38355
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.005558846662888e-05,
+      "loss": 2.9883,
+      "step": 38360
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.0044247655122516e-05,
+      "loss": 3.1335,
+      "step": 38365
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.003290737691538e-05,
+      "loss": 2.927,
+      "step": 38370
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.002156763261488e-05,
+      "loss": 2.7921,
+      "step": 38375
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.001022842282851e-05,
+      "loss": 2.9998,
+      "step": 38380
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.999888974816365e-05,
+      "loss": 2.9543,
+      "step": 38385
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.998755160922766e-05,
+      "loss": 2.912,
+      "step": 38390
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.9976214006627935e-05,
+      "loss": 2.8075,
+      "step": 38395
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 3.996487694097174e-05,
+      "loss": 2.874,
+      "step": 38400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.995354041286642e-05,
+      "loss": 3.0157,
+      "step": 38405
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.994220442291922e-05,
+      "loss": 2.9535,
+      "step": 38410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9930868971737346e-05,
+      "loss": 3.0721,
+      "step": 38415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9919534059928044e-05,
+      "loss": 2.931,
+      "step": 38420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.990819968809846e-05,
+      "loss": 2.9469,
+      "step": 38425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.989686585685576e-05,
+      "loss": 2.9824,
+      "step": 38430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.988553256680706e-05,
+      "loss": 3.0182,
+      "step": 38435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.987419981855944e-05,
+      "loss": 2.931,
+      "step": 38440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.986286761271999e-05,
+      "loss": 2.9769,
+      "step": 38445
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.985153594989569e-05,
+      "loss": 2.9355,
+      "step": 38450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9840204830693596e-05,
+      "loss": 2.8734,
+      "step": 38455
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.982887425572064e-05,
+      "loss": 2.9654,
+      "step": 38460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.981754422558378e-05,
+      "loss": 2.9196,
+      "step": 38465
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9806214740889944e-05,
+      "loss": 3.0354,
+      "step": 38470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.979488580224601e-05,
+      "loss": 3.0272,
+      "step": 38475
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9783557410258835e-05,
+      "loss": 2.8908,
+      "step": 38480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9772229565535247e-05,
+      "loss": 2.9861,
+      "step": 38485
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9760902268682035e-05,
+      "loss": 2.9366,
+      "step": 38490
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.974957552030597e-05,
+      "loss": 2.8854,
+      "step": 38495
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.97382493210138e-05,
+      "loss": 3.0246,
+      "step": 38500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9726923671412216e-05,
+      "loss": 2.9965,
+      "step": 38505
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9715598572107944e-05,
+      "loss": 2.9599,
+      "step": 38510
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.970427402370757e-05,
+      "loss": 3.0191,
+      "step": 38515
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.969295002681777e-05,
+      "loss": 2.8932,
+      "step": 38520
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9681626582045114e-05,
+      "loss": 2.9919,
+      "step": 38525
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.967030368999615e-05,
+      "loss": 2.9337,
+      "step": 38530
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.965898135127745e-05,
+      "loss": 2.9714,
+      "step": 38535
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9647659566495463e-05,
+      "loss": 2.9616,
+      "step": 38540
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.963633833625672e-05,
+      "loss": 2.9624,
+      "step": 38545
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.962501766116761e-05,
+      "loss": 2.9667,
+      "step": 38550
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.961369754183459e-05,
+      "loss": 2.8818,
+      "step": 38555
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.960237797886404e-05,
+      "loss": 2.9781,
+      "step": 38560
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.959105897286228e-05,
+      "loss": 2.9587,
+      "step": 38565
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.957974052443567e-05,
+      "loss": 3.0045,
+      "step": 38570
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.956842263419047e-05,
+      "loss": 2.9618,
+      "step": 38575
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9557105302733e-05,
+      "loss": 2.9678,
+      "step": 38580
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9545788530669445e-05,
+      "loss": 2.8913,
+      "step": 38585
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.953447231860602e-05,
+      "loss": 2.9442,
+      "step": 38590
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.952315666714892e-05,
+      "loss": 2.8641,
+      "step": 38595
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.951184157690425e-05,
+      "loss": 2.9398,
+      "step": 38600
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.950052704847818e-05,
+      "loss": 2.8648,
+      "step": 38605
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.948921308247676e-05,
+      "loss": 2.9323,
+      "step": 38610
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.947789967950603e-05,
+      "loss": 3.0207,
+      "step": 38615
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9466586840172056e-05,
+      "loss": 2.922,
+      "step": 38620
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.945527456508079e-05,
+      "loss": 2.9252,
+      "step": 38625
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9443962854838234e-05,
+      "loss": 2.9451,
+      "step": 38630
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.94326517100503e-05,
+      "loss": 3.0092,
+      "step": 38635
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9421341131322884e-05,
+      "loss": 2.9609,
+      "step": 38640
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9410031119261895e-05,
+      "loss": 2.9044,
+      "step": 38645
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9398721674473115e-05,
+      "loss": 3.0553,
+      "step": 38650
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.938741279756243e-05,
+      "loss": 3.0296,
+      "step": 38655
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9376104489135576e-05,
+      "loss": 2.9904,
+      "step": 38660
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.936479674979829e-05,
+      "loss": 2.8328,
+      "step": 38665
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.935348958015634e-05,
+      "loss": 2.94,
+      "step": 38670
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.934218298081537e-05,
+      "loss": 3.1859,
+      "step": 38675
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.933087695238109e-05,
+      "loss": 2.8939,
+      "step": 38680
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.93195714954591e-05,
+      "loss": 2.9188,
+      "step": 38685
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.930826661065498e-05,
+      "loss": 2.9365,
+      "step": 38690
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.929696229857433e-05,
+      "loss": 2.9042,
+      "step": 38695
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.928565855982266e-05,
+      "loss": 3.0399,
+      "step": 38700
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9274355395005516e-05,
+      "loss": 2.892,
+      "step": 38705
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.926305280472835e-05,
+      "loss": 2.834,
+      "step": 38710
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.925175078959659e-05,
+      "loss": 2.9197,
+      "step": 38715
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.924044935021569e-05,
+      "loss": 3.044,
+      "step": 38720
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9229148487190995e-05,
+      "loss": 2.9551,
+      "step": 38725
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9217848201127886e-05,
+      "loss": 2.9687,
+      "step": 38730
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.920654849263167e-05,
+      "loss": 2.7822,
+      "step": 38735
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.919524936230764e-05,
+      "loss": 2.8786,
+      "step": 38740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.918395081076106e-05,
+      "loss": 2.9737,
+      "step": 38745
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9172652838597143e-05,
+      "loss": 3.0395,
+      "step": 38750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9161355446421116e-05,
+      "loss": 2.913,
+      "step": 38755
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9150058634838124e-05,
+      "loss": 2.9548,
+      "step": 38760
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.913876240445331e-05,
+      "loss": 3.0427,
+      "step": 38765
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9127466755871766e-05,
+      "loss": 2.9529,
+      "step": 38770
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.911617168969858e-05,
+      "loss": 2.8982,
+      "step": 38775
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9104877206538795e-05,
+      "loss": 3.0728,
+      "step": 38780
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.909358330699743e-05,
+      "loss": 2.9541,
+      "step": 38785
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9082289991679436e-05,
+      "loss": 2.9555,
+      "step": 38790
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9070997261189776e-05,
+      "loss": 3.1048,
+      "step": 38795
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.905970511613337e-05,
+      "loss": 3.0077,
+      "step": 38800
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9048413557115104e-05,
+      "loss": 3.17,
+      "step": 38805
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.9037122584739843e-05,
+      "loss": 2.7852,
+      "step": 38810
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.902583219961237e-05,
+      "loss": 3.0486,
+      "step": 38815
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.901454240233753e-05,
+      "loss": 3.0344,
+      "step": 38820
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.900325319352003e-05,
+      "loss": 3.0318,
+      "step": 38825
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.899196457376465e-05,
+      "loss": 2.9668,
+      "step": 38830
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.898067654367606e-05,
+      "loss": 2.9074,
+      "step": 38835
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.896938910385891e-05,
+      "loss": 2.9849,
+      "step": 38840
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.895810225491786e-05,
+      "loss": 3.0546,
+      "step": 38845
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.894681599745749e-05,
+      "loss": 2.8676,
+      "step": 38850
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.893553033208241e-05,
+      "loss": 2.9191,
+      "step": 38855
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.892424525939712e-05,
+      "loss": 2.9829,
+      "step": 38860
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.891296078000612e-05,
+      "loss": 3.0994,
+      "step": 38865
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.890167689451393e-05,
+      "loss": 2.9607,
+      "step": 38870
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.889039360352493e-05,
+      "loss": 2.8815,
+      "step": 38875
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.88791109076436e-05,
+      "loss": 2.888,
+      "step": 38880
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.886782880747427e-05,
+      "loss": 2.9454,
+      "step": 38885
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.885654730362132e-05,
+      "loss": 2.8092,
+      "step": 38890
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8845266396689056e-05,
+      "loss": 2.9824,
+      "step": 38895
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8833986087281734e-05,
+      "loss": 2.7925,
+      "step": 38900
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8822706376003644e-05,
+      "loss": 3.1195,
+      "step": 38905
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.881142726345897e-05,
+      "loss": 2.9307,
+      "step": 38910
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.880014875025194e-05,
+      "loss": 3.049,
+      "step": 38915
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.878887083698669e-05,
+      "loss": 3.0165,
+      "step": 38920
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.877759352426732e-05,
+      "loss": 2.961,
+      "step": 38925
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8766316812697964e-05,
+      "loss": 2.8956,
+      "step": 38930
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8755040702882644e-05,
+      "loss": 2.868,
+      "step": 38935
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.874376519542542e-05,
+      "loss": 2.9647,
+      "step": 38940
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8732490290930266e-05,
+      "loss": 3.028,
+      "step": 38945
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.872121599000114e-05,
+      "loss": 2.9915,
+      "step": 38950
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.870994229324199e-05,
+      "loss": 3.0321,
+      "step": 38955
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.869866920125668e-05,
+      "loss": 2.8578,
+      "step": 38960
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.868739671464912e-05,
+      "loss": 2.939,
+      "step": 38965
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.867612483402312e-05,
+      "loss": 3.045,
+      "step": 38970
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8664853559982454e-05,
+      "loss": 3.0295,
+      "step": 38975
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.865358289313095e-05,
+      "loss": 2.9634,
+      "step": 38980
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.864231283407228e-05,
+      "loss": 3.0382,
+      "step": 38985
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8631043383410196e-05,
+      "loss": 3.1798,
+      "step": 38990
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8619774541748356e-05,
+      "loss": 2.8437,
+      "step": 38995
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.860850630969036e-05,
+      "loss": 2.9053,
+      "step": 39000
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.859723868783986e-05,
+      "loss": 2.9343,
+      "step": 39005
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8585971676800406e-05,
+      "loss": 3.181,
+      "step": 39010
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.857470527717555e-05,
+      "loss": 2.9737,
+      "step": 39015
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8563439489568785e-05,
+      "loss": 2.918,
+      "step": 39020
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.855217431458358e-05,
+      "loss": 2.9494,
+      "step": 39025
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.85409097528234e-05,
+      "loss": 2.9109,
+      "step": 39030
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.852964580489162e-05,
+      "loss": 2.8879,
+      "step": 39035
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8518382471391654e-05,
+      "loss": 2.9821,
+      "step": 39040
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8507119752926816e-05,
+      "loss": 3.0314,
+      "step": 39045
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8495857650100424e-05,
+      "loss": 2.9817,
+      "step": 39050
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.848459616351577e-05,
+      "loss": 3.0828,
+      "step": 39055
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.847333529377605e-05,
+      "loss": 2.8586,
+      "step": 39060
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8462075041484534e-05,
+      "loss": 3.0567,
+      "step": 39065
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8450815407244356e-05,
+      "loss": 2.9236,
+      "step": 39070
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.843955639165868e-05,
+      "loss": 2.8307,
+      "step": 39075
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 3.8428297995330606e-05,
+      "loss": 3.052,
+      "step": 39080
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.841704021886323e-05,
+      "loss": 3.0774,
+      "step": 39085
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8405783062859587e-05,
+      "loss": 2.9384,
+      "step": 39090
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8394526527922674e-05,
+      "loss": 2.8763,
+      "step": 39095
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.838327061465549e-05,
+      "loss": 2.944,
+      "step": 39100
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.837201532366096e-05,
+      "loss": 2.9971,
+      "step": 39105
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.836076065554202e-05,
+      "loss": 2.9423,
+      "step": 39110
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.834950661090152e-05,
+      "loss": 2.8414,
+      "step": 39115
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.833825319034235e-05,
+      "loss": 2.9172,
+      "step": 39120
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.832700039446725e-05,
+      "loss": 2.9486,
+      "step": 39125
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.831574822387907e-05,
+      "loss": 3.043,
+      "step": 39130
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.830449667918051e-05,
+      "loss": 3.0098,
+      "step": 39135
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8293245760974296e-05,
+      "loss": 2.9361,
+      "step": 39140
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.828199546986311e-05,
+      "loss": 2.9384,
+      "step": 39145
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8270745806449575e-05,
+      "loss": 2.9988,
+      "step": 39150
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8259496771336334e-05,
+      "loss": 3.0796,
+      "step": 39155
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.824824836512592e-05,
+      "loss": 3.037,
+      "step": 39160
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.823700058842092e-05,
+      "loss": 2.9821,
+      "step": 39165
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.822575344182383e-05,
+      "loss": 2.8218,
+      "step": 39170
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8214506925937097e-05,
+      "loss": 2.9638,
+      "step": 39175
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.82032610413632e-05,
+      "loss": 2.9238,
+      "step": 39180
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8192015788704514e-05,
+      "loss": 3.0421,
+      "step": 39185
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.818077116856346e-05,
+      "loss": 2.8674,
+      "step": 39190
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.816952718154233e-05,
+      "loss": 3.1414,
+      "step": 39195
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8158283828243436e-05,
+      "loss": 2.9262,
+      "step": 39200
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.814704110926908e-05,
+      "loss": 2.9817,
+      "step": 39205
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8135799025221455e-05,
+      "loss": 3.0598,
+      "step": 39210
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.812455757670282e-05,
+      "loss": 2.9283,
+      "step": 39215
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.811331676431532e-05,
+      "loss": 3.0519,
+      "step": 39220
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8102076588661054e-05,
+      "loss": 3.1502,
+      "step": 39225
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.809083705034219e-05,
+      "loss": 3.0234,
+      "step": 39230
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.807959814996074e-05,
+      "loss": 2.8932,
+      "step": 39235
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.806835988811878e-05,
+      "loss": 2.8524,
+      "step": 39240
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.805712226541829e-05,
+      "loss": 2.9409,
+      "step": 39245
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.804588528246121e-05,
+      "loss": 2.8261,
+      "step": 39250
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.803464893984952e-05,
+      "loss": 2.7855,
+      "step": 39255
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.802341323818507e-05,
+      "loss": 2.7807,
+      "step": 39260
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.8012178178069766e-05,
+      "loss": 2.943,
+      "step": 39265
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.800094376010542e-05,
+      "loss": 3.0593,
+      "step": 39270
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.798970998489379e-05,
+      "loss": 2.8516,
+      "step": 39275
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.797847685303669e-05,
+      "loss": 2.9488,
+      "step": 39280
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7967244365135803e-05,
+      "loss": 3.0248,
+      "step": 39285
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.795601252179285e-05,
+      "loss": 2.9507,
+      "step": 39290
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.794478132360947e-05,
+      "loss": 3.0642,
+      "step": 39295
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.793355077118727e-05,
+      "loss": 3.0651,
+      "step": 39300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.792232086512787e-05,
+      "loss": 2.8988,
+      "step": 39305
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.791109160603278e-05,
+      "loss": 2.8343,
+      "step": 39310
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.789986299450357e-05,
+      "loss": 2.9075,
+      "step": 39315
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7888635031141686e-05,
+      "loss": 2.9659,
+      "step": 39320
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.787740771654856e-05,
+      "loss": 3.0277,
+      "step": 39325
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.786618105132566e-05,
+      "loss": 2.9554,
+      "step": 39330
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.78549550360743e-05,
+      "loss": 2.9827,
+      "step": 39335
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7843729671395886e-05,
+      "loss": 2.9903,
+      "step": 39340
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7832504957891676e-05,
+      "loss": 3.0326,
+      "step": 39345
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.782128089616297e-05,
+      "loss": 3.04,
+      "step": 39350
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7810057486811e-05,
+      "loss": 2.9254,
+      "step": 39355
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.779883473043696e-05,
+      "loss": 2.8637,
+      "step": 39360
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7787612627642046e-05,
+      "loss": 2.9022,
+      "step": 39365
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.777639117902736e-05,
+      "loss": 3.094,
+      "step": 39370
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.776517038519401e-05,
+      "loss": 2.9181,
+      "step": 39375
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.775395024674308e-05,
+      "loss": 2.9621,
+      "step": 39380
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7742730764275574e-05,
+      "loss": 2.9837,
+      "step": 39385
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7731511938392504e-05,
+      "loss": 2.8916,
+      "step": 39390
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7720293769694814e-05,
+      "loss": 2.8425,
+      "step": 39395
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.770907625878343e-05,
+      "loss": 2.931,
+      "step": 39400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7697859406259226e-05,
+      "loss": 2.8977,
+      "step": 39405
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7686643212723086e-05,
+      "loss": 3.0674,
+      "step": 39410
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.767542767877581e-05,
+      "loss": 3.0822,
+      "step": 39415
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7664212805018186e-05,
+      "loss": 2.8757,
+      "step": 39420
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.765299859205094e-05,
+      "loss": 2.8326,
+      "step": 39425
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.764178504047481e-05,
+      "loss": 2.9111,
+      "step": 39430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.763057215089045e-05,
+      "loss": 3.1038,
+      "step": 39435
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.761935992389852e-05,
+      "loss": 3.0248,
+      "step": 39440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.760814836009961e-05,
+      "loss": 2.9063,
+      "step": 39445
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.759693746009429e-05,
+      "loss": 2.9407,
+      "step": 39450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.758572722448311e-05,
+      "loss": 3.0131,
+      "step": 39455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.757451765386654e-05,
+      "loss": 2.9725,
+      "step": 39460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.756330874884507e-05,
+      "loss": 2.884,
+      "step": 39465
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.755210051001912e-05,
+      "loss": 3.062,
+      "step": 39470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7540892937989056e-05,
+      "loss": 2.8976,
+      "step": 39475
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.752968603335528e-05,
+      "loss": 3.0506,
+      "step": 39480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.751847979671805e-05,
+      "loss": 2.9601,
+      "step": 39485
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7507274228677715e-05,
+      "loss": 2.9694,
+      "step": 39490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.749606932983449e-05,
+      "loss": 3.0055,
+      "step": 39495
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.748486510078857e-05,
+      "loss": 3.1086,
+      "step": 39500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.747366154214017e-05,
+      "loss": 2.9743,
+      "step": 39505
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7462458654489396e-05,
+      "loss": 2.9292,
+      "step": 39510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.745125643843638e-05,
+      "loss": 2.9044,
+      "step": 39515
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.744005489458119e-05,
+      "loss": 3.0591,
+      "step": 39520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.742885402352382e-05,
+      "loss": 2.9992,
+      "step": 39525
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.741765382586431e-05,
+      "loss": 2.862,
+      "step": 39530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.740645430220259e-05,
+      "loss": 2.8674,
+      "step": 39535
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.739525545313861e-05,
+      "loss": 2.9254,
+      "step": 39540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7384057279272234e-05,
+      "loss": 3.0429,
+      "step": 39545
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.737285978120334e-05,
+      "loss": 2.9237,
+      "step": 39550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.736166295953174e-05,
+      "loss": 3.0201,
+      "step": 39555
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.735046681485717e-05,
+      "loss": 3.015,
+      "step": 39560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7339271347779433e-05,
+      "loss": 2.8719,
+      "step": 39565
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7328076558898184e-05,
+      "loss": 2.9428,
+      "step": 39570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.731688244881314e-05,
+      "loss": 3.0147,
+      "step": 39575
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7305689018123924e-05,
+      "loss": 2.8985,
+      "step": 39580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.729449626743009e-05,
+      "loss": 2.9847,
+      "step": 39585
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.728330419733127e-05,
+      "loss": 2.9979,
+      "step": 39590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.727211280842692e-05,
+      "loss": 2.8655,
+      "step": 39595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.726092210131657e-05,
+      "loss": 3.0306,
+      "step": 39600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.724973207659969e-05,
+      "loss": 3.0621,
+      "step": 39605
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.723854273487563e-05,
+      "loss": 3.0085,
+      "step": 39610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.722735407674383e-05,
+      "loss": 2.8726,
+      "step": 39615
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.721616610280359e-05,
+      "loss": 3.0381,
+      "step": 39620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.720497881365426e-05,
+      "loss": 2.9748,
+      "step": 39625
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.719379220989507e-05,
+      "loss": 3.0352,
+      "step": 39630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.718260629212526e-05,
+      "loss": 3.0379,
+      "step": 39635
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7171421060944046e-05,
+      "loss": 3.0373,
+      "step": 39640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.716023651695055e-05,
+      "loss": 2.8666,
+      "step": 39645
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7149052660743936e-05,
+      "loss": 2.9957,
+      "step": 39650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.713786949292327e-05,
+      "loss": 2.9257,
+      "step": 39655
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.712668701408758e-05,
+      "loss": 3.0206,
+      "step": 39660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.711550522483591e-05,
+      "loss": 2.8739,
+      "step": 39665
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.710432412576721e-05,
+      "loss": 3.0092,
+      "step": 39670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.709314371748044e-05,
+      "loss": 2.8715,
+      "step": 39675
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7081964000574475e-05,
+      "loss": 2.8851,
+      "step": 39680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.70707849756482e-05,
+      "loss": 2.9207,
+      "step": 39685
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7059606643300425e-05,
+      "loss": 3.0862,
+      "step": 39690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7048429004129945e-05,
+      "loss": 3.0221,
+      "step": 39695
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.703725205873553e-05,
+      "loss": 2.9626,
+      "step": 39700
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.7026075807715866e-05,
+      "loss": 2.9418,
+      "step": 39705
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.701490025166964e-05,
+      "loss": 2.9758,
+      "step": 39710
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.70037253911955e-05,
+      "loss": 2.8814,
+      "step": 39715
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.6992551226892034e-05,
+      "loss": 3.011,
+      "step": 39720
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.698137775935783e-05,
+      "loss": 2.8936,
+      "step": 39725
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.6970204989191406e-05,
+      "loss": 2.9997,
+      "step": 39730
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.6959032916991245e-05,
+      "loss": 2.9734,
+      "step": 39735
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.69478615433558e-05,
+      "loss": 2.9746,
+      "step": 39740
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.69366908688835e-05,
+      "loss": 2.9766,
+      "step": 39745
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.6925520894172715e-05,
+      "loss": 3.0203,
+      "step": 39750
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.691435161982181e-05,
+      "loss": 2.9073,
+      "step": 39755
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 3.6903183046429036e-05,
+      "loss": 2.843,
+      "step": 39760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.689201517459272e-05,
+      "loss": 3.031,
+      "step": 39765
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.688084800491104e-05,
+      "loss": 2.9047,
+      "step": 39770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.686968153798223e-05,
+      "loss": 2.9307,
+      "step": 39775
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.685851577440442e-05,
+      "loss": 3.0502,
+      "step": 39780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.684735071477571e-05,
+      "loss": 2.9081,
+      "step": 39785
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.683618635969422e-05,
+      "loss": 2.9453,
+      "step": 39790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6825022709757944e-05,
+      "loss": 2.8952,
+      "step": 39795
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.681385976556493e-05,
+      "loss": 2.9645,
+      "step": 39800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.680269752771312e-05,
+      "loss": 2.8968,
+      "step": 39805
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.679153599680042e-05,
+      "loss": 2.9995,
+      "step": 39810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.678037517342476e-05,
+      "loss": 2.9726,
+      "step": 39815
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.676921505818396e-05,
+      "loss": 2.988,
+      "step": 39820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.675805565167586e-05,
+      "loss": 2.9549,
+      "step": 39825
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.674689695449822e-05,
+      "loss": 2.9973,
+      "step": 39830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.673573896724877e-05,
+      "loss": 2.9602,
+      "step": 39835
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.672458169052522e-05,
+      "loss": 2.8937,
+      "step": 39840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.671342512492522e-05,
+      "loss": 2.9736,
+      "step": 39845
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.670226927104641e-05,
+      "loss": 3.0338,
+      "step": 39850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.669111412948637e-05,
+      "loss": 2.8421,
+      "step": 39855
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6679959700842625e-05,
+      "loss": 2.9895,
+      "step": 39860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.666880598571273e-05,
+      "loss": 2.9737,
+      "step": 39865
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.665765298469409e-05,
+      "loss": 2.8572,
+      "step": 39870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.664650069838421e-05,
+      "loss": 2.891,
+      "step": 39875
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.663534912738044e-05,
+      "loss": 2.919,
+      "step": 39880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6624198272280125e-05,
+      "loss": 2.8201,
+      "step": 39885
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6613048133680625e-05,
+      "loss": 2.9505,
+      "step": 39890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.660189871217917e-05,
+      "loss": 3.077,
+      "step": 39895
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.659075000837304e-05,
+      "loss": 2.9548,
+      "step": 39900
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.657960202285944e-05,
+      "loss": 2.9605,
+      "step": 39905
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.656845475623548e-05,
+      "loss": 2.8533,
+      "step": 39910
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.655730820909835e-05,
+      "loss": 2.9198,
+      "step": 39915
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.654616238204508e-05,
+      "loss": 2.9935,
+      "step": 39920
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.653501727567277e-05,
+      "loss": 2.8939,
+      "step": 39925
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.652387289057839e-05,
+      "loss": 2.9581,
+      "step": 39930
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.651272922735892e-05,
+      "loss": 2.9178,
+      "step": 39935
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6501586286611316e-05,
+      "loss": 2.7552,
+      "step": 39940
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.649044406893243e-05,
+      "loss": 2.8875,
+      "step": 39945
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6479302574919155e-05,
+      "loss": 3.0578,
+      "step": 39950
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6468161805168294e-05,
+      "loss": 2.9736,
+      "step": 39955
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.64570217602766e-05,
+      "loss": 2.8913,
+      "step": 39960
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.644588244084085e-05,
+      "loss": 3.0077,
+      "step": 39965
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.643474384745771e-05,
+      "loss": 2.9018,
+      "step": 39970
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6423605980723866e-05,
+      "loss": 2.9008,
+      "step": 39975
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6412468841235925e-05,
+      "loss": 3.1142,
+      "step": 39980
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.640133242959047e-05,
+      "loss": 2.9219,
+      "step": 39985
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.639019674638405e-05,
+      "loss": 2.9624,
+      "step": 39990
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.637906179221316e-05,
+      "loss": 3.0158,
+      "step": 39995
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.63679275676743e-05,
+      "loss": 2.9015,
+      "step": 40000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.635679407336384e-05,
+      "loss": 3.0985,
+      "step": 40005
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.63456613098782e-05,
+      "loss": 3.1643,
+      "step": 40010
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.633452927781373e-05,
+      "loss": 2.9738,
+      "step": 40015
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.632339797776673e-05,
+      "loss": 2.9331,
+      "step": 40020
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.631226741033348e-05,
+      "loss": 3.1133,
+      "step": 40025
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.63011375761102e-05,
+      "loss": 2.9643,
+      "step": 40030
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6290008475693075e-05,
+      "loss": 3.0236,
+      "step": 40035
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.627888010967828e-05,
+      "loss": 2.9504,
+      "step": 40040
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.626775247866191e-05,
+      "loss": 3.0719,
+      "step": 40045
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6256625583240036e-05,
+      "loss": 3.0025,
+      "step": 40050
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.624549942400871e-05,
+      "loss": 2.9268,
+      "step": 40055
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6234374001563896e-05,
+      "loss": 3.0504,
+      "step": 40060
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.622324931650158e-05,
+      "loss": 3.0163,
+      "step": 40065
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.621212536941765e-05,
+      "loss": 2.9479,
+      "step": 40070
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.620100216090802e-05,
+      "loss": 3.0322,
+      "step": 40075
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.618987969156849e-05,
+      "loss": 3.026,
+      "step": 40080
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.617875796199487e-05,
+      "loss": 2.9618,
+      "step": 40085
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.616763697278292e-05,
+      "loss": 2.8604,
+      "step": 40090
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6156516724528336e-05,
+      "loss": 3.0715,
+      "step": 40095
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6145397217826835e-05,
+      "loss": 2.9641,
+      "step": 40100
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6134278453274026e-05,
+      "loss": 2.8912,
+      "step": 40105
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.612316043146551e-05,
+      "loss": 3.0557,
+      "step": 40110
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.611204315299686e-05,
+      "loss": 2.8304,
+      "step": 40115
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.610092661846356e-05,
+      "loss": 2.95,
+      "step": 40120
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6089810828461135e-05,
+      "loss": 3.0475,
+      "step": 40125
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6078695783585006e-05,
+      "loss": 2.9042,
+      "step": 40130
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6067581484430545e-05,
+      "loss": 2.8679,
+      "step": 40135
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.605646793159316e-05,
+      "loss": 2.9927,
+      "step": 40140
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.604535512566812e-05,
+      "loss": 2.9154,
+      "step": 40145
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.603424306725075e-05,
+      "loss": 2.9391,
+      "step": 40150
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6023131756936264e-05,
+      "loss": 3.0243,
+      "step": 40155
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.6012021195319846e-05,
+      "loss": 2.9868,
+      "step": 40160
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.600091138299669e-05,
+      "loss": 2.986,
+      "step": 40165
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.598980232056188e-05,
+      "loss": 2.945,
+      "step": 40170
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.597869400861052e-05,
+      "loss": 2.8602,
+      "step": 40175
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.596758644773766e-05,
+      "loss": 2.963,
+      "step": 40180
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.595647963853825e-05,
+      "loss": 2.835,
+      "step": 40185
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5945373581607295e-05,
+      "loss": 2.9351,
+      "step": 40190
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5934268277539674e-05,
+      "loss": 2.8329,
+      "step": 40195
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.59231637269303e-05,
+      "loss": 3.089,
+      "step": 40200
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.591205993037398e-05,
+      "loss": 2.896,
+      "step": 40205
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5900956888465544e-05,
+      "loss": 2.8863,
+      "step": 40210
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.588985460179972e-05,
+      "loss": 3.1403,
+      "step": 40215
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.587875307097122e-05,
+      "loss": 2.9906,
+      "step": 40220
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.586765229657475e-05,
+      "loss": 2.8584,
+      "step": 40225
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.58565522792049e-05,
+      "loss": 2.9516,
+      "step": 40230
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.584545301945631e-05,
+      "loss": 2.9896,
+      "step": 40235
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.583435451792352e-05,
+      "loss": 2.8805,
+      "step": 40240
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.582325677520101e-05,
+      "loss": 2.9811,
+      "step": 40245
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.581215979188329e-05,
+      "loss": 2.8963,
+      "step": 40250
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.580106356856477e-05,
+      "loss": 3.0605,
+      "step": 40255
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5789968105839864e-05,
+      "loss": 2.9371,
+      "step": 40260
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.577887340430292e-05,
+      "loss": 2.8988,
+      "step": 40265
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5767779464548196e-05,
+      "loss": 3.0208,
+      "step": 40270
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5756686287170025e-05,
+      "loss": 3.0397,
+      "step": 40275
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5745593872762596e-05,
+      "loss": 2.959,
+      "step": 40280
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.573450222192012e-05,
+      "loss": 2.967,
+      "step": 40285
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.572341133523672e-05,
+      "loss": 2.9601,
+      "step": 40290
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.571232121330652e-05,
+      "loss": 2.9177,
+      "step": 40295
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.570123185672359e-05,
+      "loss": 2.9667,
+      "step": 40300
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.569014326608191e-05,
+      "loss": 2.9977,
+      "step": 40305
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5679055441975516e-05,
+      "loss": 2.9977,
+      "step": 40310
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5667968384998305e-05,
+      "loss": 2.9279,
+      "step": 40315
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5656882095744205e-05,
+      "loss": 2.8931,
+      "step": 40320
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.564579657480707e-05,
+      "loss": 3.0564,
+      "step": 40325
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5634711822780696e-05,
+      "loss": 3.0057,
+      "step": 40330
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.562362784025889e-05,
+      "loss": 2.9207,
+      "step": 40335
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5612544627835375e-05,
+      "loss": 2.9767,
+      "step": 40340
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.560146218610384e-05,
+      "loss": 2.9566,
+      "step": 40345
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.559038051565793e-05,
+      "loss": 2.9031,
+      "step": 40350
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5579299617091275e-05,
+      "loss": 2.8421,
+      "step": 40355
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5568219490997435e-05,
+      "loss": 3.019,
+      "step": 40360
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5557140137969946e-05,
+      "loss": 2.9723,
+      "step": 40365
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.554606155860227e-05,
+      "loss": 3.122,
+      "step": 40370
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5534983753487894e-05,
+      "loss": 2.9697,
+      "step": 40375
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5523906723220193e-05,
+      "loss": 3.0824,
+      "step": 40380
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5512830468392535e-05,
+      "loss": 3.0247,
+      "step": 40385
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.550175498959825e-05,
+      "loss": 3.0819,
+      "step": 40390
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.549068028743058e-05,
+      "loss": 3.1175,
+      "step": 40395
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5479606362482825e-05,
+      "loss": 2.8535,
+      "step": 40400
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.546853321534813e-05,
+      "loss": 2.948,
+      "step": 40405
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5457460846619684e-05,
+      "loss": 3.1108,
+      "step": 40410
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5446389256890575e-05,
+      "loss": 2.9509,
+      "step": 40415
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.543531844675388e-05,
+      "loss": 3.116,
+      "step": 40420
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.542424841680265e-05,
+      "loss": 2.969,
+      "step": 40425
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.541317916762983e-05,
+      "loss": 3.0621,
+      "step": 40430
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5402110699828404e-05,
+      "loss": 2.8535,
+      "step": 40435
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 3.5391043013991275e-05,
+      "loss": 2.9964,
+      "step": 40440
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.537997611071126e-05,
+      "loss": 2.7772,
+      "step": 40445
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5368909990581244e-05,
+      "loss": 3.0666,
+      "step": 40450
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.535784465419395e-05,
+      "loss": 3.0181,
+      "step": 40455
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.534678010214216e-05,
+      "loss": 3.0533,
+      "step": 40460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5335716335018544e-05,
+      "loss": 2.8915,
+      "step": 40465
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.532465335341573e-05,
+      "loss": 2.9067,
+      "step": 40470
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.531359115792638e-05,
+      "loss": 2.9326,
+      "step": 40475
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.530252974914301e-05,
+      "loss": 2.8743,
+      "step": 40480
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.529146912765819e-05,
+      "loss": 2.9876,
+      "step": 40485
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.52804092940644e-05,
+      "loss": 2.9246,
+      "step": 40490
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.526935024895404e-05,
+      "loss": 2.9516,
+      "step": 40495
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.525829199291954e-05,
+      "loss": 2.9798,
+      "step": 40500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.524723452655325e-05,
+      "loss": 3.0661,
+      "step": 40505
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.523617785044751e-05,
+      "loss": 2.8919,
+      "step": 40510
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5225121965194565e-05,
+      "loss": 3.0387,
+      "step": 40515
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.521406687138663e-05,
+      "loss": 3.1082,
+      "step": 40520
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.520301256961593e-05,
+      "loss": 3.087,
+      "step": 40525
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5191959060474574e-05,
+      "loss": 2.9034,
+      "step": 40530
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.51809063445547e-05,
+      "loss": 2.7998,
+      "step": 40535
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5169854422448356e-05,
+      "loss": 2.8408,
+      "step": 40540
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.515880329474753e-05,
+      "loss": 3.0029,
+      "step": 40545
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.514775296204425e-05,
+      "loss": 3.0977,
+      "step": 40550
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5136703424930395e-05,
+      "loss": 2.8729,
+      "step": 40555
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5125654683997905e-05,
+      "loss": 3.0924,
+      "step": 40560
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5114606739838604e-05,
+      "loss": 2.9119,
+      "step": 40565
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5103559593044265e-05,
+      "loss": 2.9925,
+      "step": 40570
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.509251324420671e-05,
+      "loss": 2.8705,
+      "step": 40575
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.508146769391761e-05,
+      "loss": 3.0466,
+      "step": 40580
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.507042294276868e-05,
+      "loss": 2.9222,
+      "step": 40585
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.505937899135151e-05,
+      "loss": 2.8872,
+      "step": 40590
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.504833584025773e-05,
+      "loss": 2.8405,
+      "step": 40595
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.503729349007887e-05,
+      "loss": 3.0498,
+      "step": 40600
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.5026251941406424e-05,
+      "loss": 2.9819,
+      "step": 40605
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.501521119483189e-05,
+      "loss": 2.8992,
+      "step": 40610
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.500417125094665e-05,
+      "loss": 3.0427,
+      "step": 40615
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4993132110342084e-05,
+      "loss": 2.8499,
+      "step": 40620
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.498209377360955e-05,
+      "loss": 3.1157,
+      "step": 40625
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4971056241340315e-05,
+      "loss": 3.0267,
+      "step": 40630
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.496001951412564e-05,
+      "loss": 2.9878,
+      "step": 40635
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4948983592556725e-05,
+      "loss": 2.9535,
+      "step": 40640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4937948477224726e-05,
+      "loss": 2.8135,
+      "step": 40645
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.492691416872075e-05,
+      "loss": 2.8456,
+      "step": 40650
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.491588066763589e-05,
+      "loss": 3.0193,
+      "step": 40655
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.490484797456117e-05,
+      "loss": 2.9039,
+      "step": 40660
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.48938160900876e-05,
+      "loss": 2.89,
+      "step": 40665
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4882785014806065e-05,
+      "loss": 2.8689,
+      "step": 40670
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4871754749307525e-05,
+      "loss": 2.9944,
+      "step": 40675
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.486072529418281e-05,
+      "loss": 2.975,
+      "step": 40680
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.484969665002275e-05,
+      "loss": 3.0194,
+      "step": 40685
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.483866881741811e-05,
+      "loss": 2.7786,
+      "step": 40690
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.482764179695959e-05,
+      "loss": 2.978,
+      "step": 40695
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4816615589237917e-05,
+      "loss": 3.1301,
+      "step": 40700
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.480559019484369e-05,
+      "loss": 2.9184,
+      "step": 40705
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.479456561436755e-05,
+      "loss": 2.9113,
+      "step": 40710
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.478354184840003e-05,
+      "loss": 3.0147,
+      "step": 40715
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4772518897531604e-05,
+      "loss": 3.0357,
+      "step": 40720
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.47614967623528e-05,
+      "loss": 2.9229,
+      "step": 40725
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4750475443453986e-05,
+      "loss": 2.8925,
+      "step": 40730
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.473945494142558e-05,
+      "loss": 2.9603,
+      "step": 40735
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.472843525685792e-05,
+      "loss": 2.8823,
+      "step": 40740
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.471741639034124e-05,
+      "loss": 2.8478,
+      "step": 40745
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4706398342465854e-05,
+      "loss": 2.9813,
+      "step": 40750
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4695381113821903e-05,
+      "loss": 3.1117,
+      "step": 40755
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4684364704999614e-05,
+      "loss": 2.9822,
+      "step": 40760
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.467334911658906e-05,
+      "loss": 3.0261,
+      "step": 40765
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.46623343491803e-05,
+      "loss": 2.8558,
+      "step": 40770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4651320403363404e-05,
+      "loss": 2.8483,
+      "step": 40775
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4640307279728315e-05,
+      "loss": 2.9825,
+      "step": 40780
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4629294978865e-05,
+      "loss": 2.9434,
+      "step": 40785
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4618283501363356e-05,
+      "loss": 3.0763,
+      "step": 40790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.46072728478132e-05,
+      "loss": 2.8805,
+      "step": 40795
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4596263018804374e-05,
+      "loss": 3.0625,
+      "step": 40800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4585254014926614e-05,
+      "loss": 2.977,
+      "step": 40805
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.457424583676967e-05,
+      "loss": 2.9389,
+      "step": 40810
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.45632384849232e-05,
+      "loss": 2.9759,
+      "step": 40815
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4552231959976824e-05,
+      "loss": 2.7213,
+      "step": 40820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.454122626252014e-05,
+      "loss": 2.857,
+      "step": 40825
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.453022139314268e-05,
+      "loss": 2.9844,
+      "step": 40830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.451921735243396e-05,
+      "loss": 3.1563,
+      "step": 40835
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.450821414098343e-05,
+      "loss": 2.9761,
+      "step": 40840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.449721175938046e-05,
+      "loss": 2.8986,
+      "step": 40845
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.448621020821448e-05,
+      "loss": 2.9988,
+      "step": 40850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.447520948807474e-05,
+      "loss": 2.9167,
+      "step": 40855
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4464209599550566e-05,
+      "loss": 2.9328,
+      "step": 40860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4453210543231155e-05,
+      "loss": 3.0462,
+      "step": 40865
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.444221231970573e-05,
+      "loss": 3.0842,
+      "step": 40870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.443121492956341e-05,
+      "loss": 3.0198,
+      "step": 40875
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.442021837339327e-05,
+      "loss": 2.8731,
+      "step": 40880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.44092226517844e-05,
+      "loss": 2.9647,
+      "step": 40885
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.439822776532578e-05,
+      "loss": 2.94,
+      "step": 40890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4387233714606385e-05,
+      "loss": 2.9637,
+      "step": 40895
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.437624050021513e-05,
+      "loss": 2.9728,
+      "step": 40900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.43652481227409e-05,
+      "loss": 3.001,
+      "step": 40905
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4354256582772515e-05,
+      "loss": 3.0801,
+      "step": 40910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4343265880898725e-05,
+      "loss": 2.9514,
+      "step": 40915
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.433227601770833e-05,
+      "loss": 2.9493,
+      "step": 40920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.432128699378998e-05,
+      "loss": 2.888,
+      "step": 40925
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.431029880973233e-05,
+      "loss": 2.9666,
+      "step": 40930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4299311466123995e-05,
+      "loss": 2.9086,
+      "step": 40935
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4288324963553524e-05,
+      "loss": 2.9577,
+      "step": 40940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4277339302609444e-05,
+      "loss": 2.9289,
+      "step": 40945
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4266354483880204e-05,
+      "loss": 2.8677,
+      "step": 40950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.425537050795423e-05,
+      "loss": 3.0554,
+      "step": 40955
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4244387375419915e-05,
+      "loss": 3.0265,
+      "step": 40960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.423340508686557e-05,
+      "loss": 2.8517,
+      "step": 40965
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4222423642879503e-05,
+      "loss": 2.8517,
+      "step": 40970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.421144304404996e-05,
+      "loss": 2.8862,
+      "step": 40975
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.42004632909651e-05,
+      "loss": 2.9942,
+      "step": 40980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4189484384213127e-05,
+      "loss": 2.9268,
+      "step": 40985
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.417850632438211e-05,
+      "loss": 2.8778,
+      "step": 40990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.416752911206013e-05,
+      "loss": 2.9002,
+      "step": 40995
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.415655274783519e-05,
+      "loss": 2.8629,
+      "step": 41000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.414557723229525e-05,
+      "loss": 2.9278,
+      "step": 41005
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.413460256602827e-05,
+      "loss": 3.0173,
+      "step": 41010
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4123628749622095e-05,
+      "loss": 2.86,
+      "step": 41015
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.411265578366458e-05,
+      "loss": 3.0111,
+      "step": 41020
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4101683668743516e-05,
+      "loss": 2.9053,
+      "step": 41025
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.409071240544661e-05,
+      "loss": 2.9712,
+      "step": 41030
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.407974199436161e-05,
+      "loss": 3.093,
+      "step": 41035
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.406877243607613e-05,
+      "loss": 2.9431,
+      "step": 41040
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.405780373117779e-05,
+      "loss": 3.0157,
+      "step": 41045
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.404683588025416e-05,
+      "loss": 2.9053,
+      "step": 41050
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4035868883892714e-05,
+      "loss": 2.8628,
+      "step": 41055
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.4024902742680974e-05,
+      "loss": 2.8793,
+      "step": 41060
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.401393745720631e-05,
+      "loss": 2.9733,
+      "step": 41065
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.400297302805615e-05,
+      "loss": 2.9114,
+      "step": 41070
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.3992009455817804e-05,
+      "loss": 2.9718,
+      "step": 41075
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.398104674107853e-05,
+      "loss": 2.9467,
+      "step": 41080
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.3970084884425615e-05,
+      "loss": 2.8526,
+      "step": 41085
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.39591238864462e-05,
+      "loss": 2.8505,
+      "step": 41090
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.394816374772749e-05,
+      "loss": 2.8938,
+      "step": 41095
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.393720446885655e-05,
+      "loss": 3.0148,
+      "step": 41100
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.3926246050420415e-05,
+      "loss": 2.9828,
+      "step": 41105
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.391528849300615e-05,
+      "loss": 3.0564,
+      "step": 41110
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.3904331797200676e-05,
+      "loss": 2.967,
+      "step": 41115
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 3.389337596359092e-05,
+      "loss": 2.9806,
+      "step": 41120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3882420992763774e-05,
+      "loss": 3.0317,
+      "step": 41125
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.387146688530601e-05,
+      "loss": 3.0788,
+      "step": 41130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.386051364180447e-05,
+      "loss": 2.8481,
+      "step": 41135
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3849561262845825e-05,
+      "loss": 2.9225,
+      "step": 41140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.383860974901681e-05,
+      "loss": 2.8916,
+      "step": 41145
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.382765910090405e-05,
+      "loss": 3.0664,
+      "step": 41150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.381670931909411e-05,
+      "loss": 3.0145,
+      "step": 41155
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.380576040417357e-05,
+      "loss": 3.0594,
+      "step": 41160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3794812356728905e-05,
+      "loss": 2.9514,
+      "step": 41165
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.37838651773466e-05,
+      "loss": 2.9486,
+      "step": 41170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.377291886661305e-05,
+      "loss": 2.787,
+      "step": 41175
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.376197342511458e-05,
+      "loss": 3.1277,
+      "step": 41180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.375102885343755e-05,
+      "loss": 2.9889,
+      "step": 41185
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.37400851521682e-05,
+      "loss": 2.9503,
+      "step": 41190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.372914232189277e-05,
+      "loss": 2.9455,
+      "step": 41195
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.371820036319742e-05,
+      "loss": 2.8914,
+      "step": 41200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.370725927666828e-05,
+      "loss": 2.9656,
+      "step": 41205
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.369631906289144e-05,
+      "loss": 2.9435,
+      "step": 41210
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.368537972245291e-05,
+      "loss": 3.0259,
+      "step": 41215
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.367444125593871e-05,
+      "loss": 2.8261,
+      "step": 41220
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.366350366393476e-05,
+      "loss": 3.0638,
+      "step": 41225
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.365256694702694e-05,
+      "loss": 3.0116,
+      "step": 41230
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.364163110580113e-05,
+      "loss": 3.0449,
+      "step": 41235
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.36306961408431e-05,
+      "loss": 2.8873,
+      "step": 41240
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3619762052738625e-05,
+      "loss": 2.9898,
+      "step": 41245
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3608828842073406e-05,
+      "loss": 3.0051,
+      "step": 41250
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.359789650943309e-05,
+      "loss": 2.9159,
+      "step": 41255
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3586965055403286e-05,
+      "loss": 2.9594,
+      "step": 41260
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3576034480569575e-05,
+      "loss": 2.9953,
+      "step": 41265
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.356510478551746e-05,
+      "loss": 2.8981,
+      "step": 41270
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.355417597083243e-05,
+      "loss": 2.9502,
+      "step": 41275
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3543248037099886e-05,
+      "loss": 2.918,
+      "step": 41280
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.353232098490523e-05,
+      "loss": 2.9803,
+      "step": 41285
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.352139481483376e-05,
+      "loss": 2.9307,
+      "step": 41290
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.351046952747078e-05,
+      "loss": 3.0026,
+      "step": 41295
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.349954512340152e-05,
+      "loss": 2.8763,
+      "step": 41300
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.348862160321115e-05,
+      "loss": 2.8736,
+      "step": 41305
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.347769896748484e-05,
+      "loss": 2.8495,
+      "step": 41310
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.346677721680766e-05,
+      "loss": 2.8825,
+      "step": 41315
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.345585635176466e-05,
+      "loss": 3.0119,
+      "step": 41320
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.344493637294084e-05,
+      "loss": 2.9589,
+      "step": 41325
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.343401728092114e-05,
+      "loss": 3.0527,
+      "step": 41330
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3423099076290475e-05,
+      "loss": 2.8936,
+      "step": 41335
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.341218175963369e-05,
+      "loss": 2.9981,
+      "step": 41340
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3401265331535606e-05,
+      "loss": 2.9247,
+      "step": 41345
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.339034979258098e-05,
+      "loss": 2.9354,
+      "step": 41350
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3379435143354495e-05,
+      "loss": 2.9087,
+      "step": 41355
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.336852138444086e-05,
+      "loss": 2.9943,
+      "step": 41360
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.335760851642465e-05,
+      "loss": 3.0562,
+      "step": 41365
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.334669653989047e-05,
+      "loss": 2.9944,
+      "step": 41370
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.333578545542283e-05,
+      "loss": 2.9272,
+      "step": 41375
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3324875263606176e-05,
+      "loss": 2.9138,
+      "step": 41380
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.331396596502497e-05,
+      "loss": 3.032,
+      "step": 41385
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.330305756026357e-05,
+      "loss": 2.9038,
+      "step": 41390
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3292150049906323e-05,
+      "loss": 2.8915,
+      "step": 41395
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3281243434537504e-05,
+      "loss": 3.0032,
+      "step": 41400
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3270337714741316e-05,
+      "loss": 2.8443,
+      "step": 41405
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3259432891102e-05,
+      "loss": 2.8773,
+      "step": 41410
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.324852896420364e-05,
+      "loss": 3.1521,
+      "step": 41415
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3237625934630374e-05,
+      "loss": 2.8712,
+      "step": 41420
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.322672380296622e-05,
+      "loss": 2.8002,
+      "step": 41425
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3215822569795155e-05,
+      "loss": 2.9905,
+      "step": 41430
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3204922235701166e-05,
+      "loss": 2.9475,
+      "step": 41435
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3194022801268104e-05,
+      "loss": 2.7485,
+      "step": 41440
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3183124267079866e-05,
+      "loss": 3.0928,
+      "step": 41445
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.317222663372023e-05,
+      "loss": 2.8818,
+      "step": 41450
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.316132990177292e-05,
+      "loss": 2.942,
+      "step": 41455
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.315043407182169e-05,
+      "loss": 2.956,
+      "step": 41460
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.313953914445016e-05,
+      "loss": 2.9438,
+      "step": 41465
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.312864512024198e-05,
+      "loss": 2.9194,
+      "step": 41470
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.311775199978068e-05,
+      "loss": 3.0187,
+      "step": 41475
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3106859783649744e-05,
+      "loss": 2.8454,
+      "step": 41480
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.309596847243269e-05,
+      "loss": 2.9784,
+      "step": 41485
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.308507806671289e-05,
+      "loss": 2.8894,
+      "step": 41490
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.307418856707374e-05,
+      "loss": 2.9085,
+      "step": 41495
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3063299974098545e-05,
+      "loss": 2.9069,
+      "step": 41500
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.305241228837056e-05,
+      "loss": 3.0009,
+      "step": 41505
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.304152551047304e-05,
+      "loss": 3.0891,
+      "step": 41510
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.30306396409891e-05,
+      "loss": 2.8614,
+      "step": 41515
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.3019754680501926e-05,
+      "loss": 2.9197,
+      "step": 41520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.300887062959454e-05,
+      "loss": 3.0404,
+      "step": 41525
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.299798748885001e-05,
+      "loss": 2.8765,
+      "step": 41530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.298710525885128e-05,
+      "loss": 2.9952,
+      "step": 41535
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.29762239401813e-05,
+      "loss": 2.9219,
+      "step": 41540
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.296534353342293e-05,
+      "loss": 3.0118,
+      "step": 41545
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2954464039159e-05,
+      "loss": 2.9336,
+      "step": 41550
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.294358545797233e-05,
+      "loss": 2.8741,
+      "step": 41555
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2932707790445607e-05,
+      "loss": 3.0306,
+      "step": 41560
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.292183103716153e-05,
+      "loss": 3.0636,
+      "step": 41565
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.291095519870273e-05,
+      "loss": 3.0279,
+      "step": 41570
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2900080275651794e-05,
+      "loss": 2.9295,
+      "step": 41575
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.288920626859127e-05,
+      "loss": 2.9434,
+      "step": 41580
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2878333178103636e-05,
+      "loss": 2.9621,
+      "step": 41585
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2867461004771326e-05,
+      "loss": 2.8972,
+      "step": 41590
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.285658974917673e-05,
+      "loss": 2.9548,
+      "step": 41595
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.284571941190219e-05,
+      "loss": 2.9743,
+      "step": 41600
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2834849993530006e-05,
+      "loss": 2.8646,
+      "step": 41605
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.282398149464242e-05,
+      "loss": 2.9303,
+      "step": 41610
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.281311391582158e-05,
+      "loss": 2.9846,
+      "step": 41615
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.28022472576497e-05,
+      "loss": 2.9818,
+      "step": 41620
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.279138152070881e-05,
+      "loss": 2.9718,
+      "step": 41625
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.278051670558101e-05,
+      "loss": 3.0338,
+      "step": 41630
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2769652812848264e-05,
+      "loss": 3.0074,
+      "step": 41635
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2758789843092494e-05,
+      "loss": 3.1038,
+      "step": 41640
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.274792779689565e-05,
+      "loss": 3.0002,
+      "step": 41645
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.273706667483953e-05,
+      "loss": 3.074,
+      "step": 41650
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.272620647750597e-05,
+      "loss": 2.9278,
+      "step": 41655
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2715347205476705e-05,
+      "loss": 2.9195,
+      "step": 41660
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.27044888593334e-05,
+      "loss": 2.968,
+      "step": 41665
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2693631439657763e-05,
+      "loss": 2.8418,
+      "step": 41670
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.268277494703133e-05,
+      "loss": 3.0442,
+      "step": 41675
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.267191938203571e-05,
+      "loss": 2.8754,
+      "step": 41680
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.266106474525237e-05,
+      "loss": 2.9294,
+      "step": 41685
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2650211037262746e-05,
+      "loss": 3.1024,
+      "step": 41690
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.263935825864827e-05,
+      "loss": 2.8648,
+      "step": 41695
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.262850640999027e-05,
+      "loss": 2.8464,
+      "step": 41700
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2617655491870055e-05,
+      "loss": 2.9255,
+      "step": 41705
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.260680550486889e-05,
+      "loss": 2.8858,
+      "step": 41710
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.259595644956793e-05,
+      "loss": 2.9911,
+      "step": 41715
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.258510832654839e-05,
+      "loss": 2.96,
+      "step": 41720
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2574261136391296e-05,
+      "loss": 2.9788,
+      "step": 41725
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.256341487967777e-05,
+      "loss": 2.9248,
+      "step": 41730
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.255256955698879e-05,
+      "loss": 2.9899,
+      "step": 41735
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.254172516890527e-05,
+      "loss": 3.0226,
+      "step": 41740
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.253088171600816e-05,
+      "loss": 3.033,
+      "step": 41745
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.252003919887827e-05,
+      "loss": 3.0071,
+      "step": 41750
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.250919761809644e-05,
+      "loss": 3.0036,
+      "step": 41755
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.24983569742434e-05,
+      "loss": 3.0235,
+      "step": 41760
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2487517267899835e-05,
+      "loss": 2.9666,
+      "step": 41765
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2476678499646426e-05,
+      "loss": 2.9842,
+      "step": 41770
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.246584067006374e-05,
+      "loss": 2.9424,
+      "step": 41775
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2455003779732366e-05,
+      "loss": 2.8375,
+      "step": 41780
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.244416782923279e-05,
+      "loss": 2.9484,
+      "step": 41785
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.243333281914541e-05,
+      "loss": 2.9979,
+      "step": 41790
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.24224987500507e-05,
+      "loss": 2.9951,
+      "step": 41795
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 3.2411665622528956e-05,
+      "loss": 2.7823,
+      "step": 41800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.240083343716051e-05,
+      "loss": 2.998,
+      "step": 41805
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2390002194525594e-05,
+      "loss": 3.0192,
+      "step": 41810
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2379171895204386e-05,
+      "loss": 3.0766,
+      "step": 41815
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2368342539777066e-05,
+      "loss": 2.965,
+      "step": 41820
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.235751412882369e-05,
+      "loss": 3.0677,
+      "step": 41825
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2346686662924354e-05,
+      "loss": 2.9203,
+      "step": 41830
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2335860142659004e-05,
+      "loss": 2.8834,
+      "step": 41835
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2325034568607606e-05,
+      "loss": 2.9657,
+      "step": 41840
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.231420994135005e-05,
+      "loss": 2.9802,
+      "step": 41845
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2303386261466184e-05,
+      "loss": 3.1997,
+      "step": 41850
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.22925635295358e-05,
+      "loss": 2.8268,
+      "step": 41855
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.228174174613863e-05,
+      "loss": 2.9539,
+      "step": 41860
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.227092091185435e-05,
+      "loss": 2.9498,
+      "step": 41865
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2260101027262625e-05,
+      "loss": 3.0011,
+      "step": 41870
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2249282092943035e-05,
+      "loss": 2.9761,
+      "step": 41875
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.223846410947512e-05,
+      "loss": 2.9541,
+      "step": 41880
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.222764707743836e-05,
+      "loss": 3.0393,
+      "step": 41885
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.221683099741218e-05,
+      "loss": 2.8657,
+      "step": 41890
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2206015869976e-05,
+      "loss": 2.8399,
+      "step": 41895
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.219520169570911e-05,
+      "loss": 3.0021,
+      "step": 41900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.218438847519082e-05,
+      "loss": 2.9596,
+      "step": 41905
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.217357620900036e-05,
+      "loss": 2.9192,
+      "step": 41910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.216276489771688e-05,
+      "loss": 2.9232,
+      "step": 41915
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.215195454191956e-05,
+      "loss": 2.8719,
+      "step": 41920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.214114514218743e-05,
+      "loss": 2.8137,
+      "step": 41925
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.213033669909955e-05,
+      "loss": 3.0701,
+      "step": 41930
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.211952921323489e-05,
+      "loss": 3.0573,
+      "step": 41935
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2108722685172344e-05,
+      "loss": 3.0066,
+      "step": 41940
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.209791711549082e-05,
+      "loss": 2.963,
+      "step": 41945
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2087112504769104e-05,
+      "loss": 2.9816,
+      "step": 41950
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.207630885358602e-05,
+      "loss": 3.0099,
+      "step": 41955
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.206550616252024e-05,
+      "loss": 2.9764,
+      "step": 41960
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.205470443215044e-05,
+      "loss": 2.9386,
+      "step": 41965
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.204390366305525e-05,
+      "loss": 2.9256,
+      "step": 41970
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.203310385581321e-05,
+      "loss": 2.8621,
+      "step": 41975
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2022305011002863e-05,
+      "loss": 2.9255,
+      "step": 41980
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.2011507129202654e-05,
+      "loss": 2.9423,
+      "step": 41985
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.200071021099098e-05,
+      "loss": 3.0226,
+      "step": 41990
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1989914256946227e-05,
+      "loss": 2.9831,
+      "step": 41995
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1979119267646676e-05,
+      "loss": 2.9848,
+      "step": 42000
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.196832524367061e-05,
+      "loss": 3.0359,
+      "step": 42005
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.195753218559621e-05,
+      "loss": 2.9602,
+      "step": 42010
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.19467400940016e-05,
+      "loss": 2.8627,
+      "step": 42015
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1935948969464944e-05,
+      "loss": 2.941,
+      "step": 42020
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.192515881256424e-05,
+      "loss": 2.7718,
+      "step": 42025
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1914369623877505e-05,
+      "loss": 2.9747,
+      "step": 42030
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1903581403982686e-05,
+      "loss": 2.8911,
+      "step": 42035
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.189279415345765e-05,
+      "loss": 2.9188,
+      "step": 42040
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.188200787288026e-05,
+      "loss": 2.9672,
+      "step": 42045
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.187122256282829e-05,
+      "loss": 2.9206,
+      "step": 42050
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.18604382238795e-05,
+      "loss": 2.9157,
+      "step": 42055
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.184965485661156e-05,
+      "loss": 2.8917,
+      "step": 42060
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.183887246160208e-05,
+      "loss": 2.7863,
+      "step": 42065
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1828091039428675e-05,
+      "loss": 2.9096,
+      "step": 42070
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1817310590668836e-05,
+      "loss": 2.8519,
+      "step": 42075
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.180653111590009e-05,
+      "loss": 2.8696,
+      "step": 42080
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.179575261569981e-05,
+      "loss": 2.8224,
+      "step": 42085
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1784975090645386e-05,
+      "loss": 3.016,
+      "step": 42090
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.177419854131416e-05,
+      "loss": 2.9066,
+      "step": 42095
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.176342296828335e-05,
+      "loss": 2.9862,
+      "step": 42100
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.175264837213022e-05,
+      "loss": 2.8718,
+      "step": 42105
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.17418747534319e-05,
+      "loss": 2.8251,
+      "step": 42110
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.173110211276551e-05,
+      "loss": 2.8668,
+      "step": 42115
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.172033045070812e-05,
+      "loss": 3.0527,
+      "step": 42120
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.17095597678367e-05,
+      "loss": 3.0113,
+      "step": 42125
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.169879006472825e-05,
+      "loss": 2.9485,
+      "step": 42130
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.168802134195963e-05,
+      "loss": 2.9364,
+      "step": 42135
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1677253600107705e-05,
+      "loss": 2.9218,
+      "step": 42140
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1666486839749264e-05,
+      "loss": 2.9079,
+      "step": 42145
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.165572106146105e-05,
+      "loss": 2.8699,
+      "step": 42150
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1644956265819775e-05,
+      "loss": 2.8912,
+      "step": 42155
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.163419245340205e-05,
+      "loss": 2.7708,
+      "step": 42160
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.162342962478445e-05,
+      "loss": 2.8686,
+      "step": 42165
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.161266778054354e-05,
+      "loss": 3.1158,
+      "step": 42170
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.160190692125578e-05,
+      "loss": 2.9213,
+      "step": 42175
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.159114704749759e-05,
+      "loss": 2.7865,
+      "step": 42180
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1580388159845363e-05,
+      "loss": 2.9567,
+      "step": 42185
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.156963025887539e-05,
+      "loss": 2.9835,
+      "step": 42190
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.155887334516399e-05,
+      "loss": 3.0571,
+      "step": 42195
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.154811741928732e-05,
+      "loss": 3.1194,
+      "step": 42200
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.153736248182158e-05,
+      "loss": 2.9832,
+      "step": 42205
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.152660853334286e-05,
+      "loss": 2.9451,
+      "step": 42210
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.151585557442723e-05,
+      "loss": 2.8274,
+      "step": 42215
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.150510360565069e-05,
+      "loss": 2.9836,
+      "step": 42220
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.149435262758917e-05,
+      "loss": 2.9029,
+      "step": 42225
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.14836026408186e-05,
+      "loss": 2.9197,
+      "step": 42230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.14728536459148e-05,
+      "loss": 2.9538,
+      "step": 42235
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1462105643453564e-05,
+      "loss": 2.8857,
+      "step": 42240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.145135863401065e-05,
+      "loss": 2.9304,
+      "step": 42245
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.14406126181617e-05,
+      "loss": 2.8489,
+      "step": 42250
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1429867596482396e-05,
+      "loss": 2.8862,
+      "step": 42255
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.141912356954827e-05,
+      "loss": 3.1061,
+      "step": 42260
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.140838053793489e-05,
+      "loss": 2.8808,
+      "step": 42265
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.13976385022177e-05,
+      "loss": 2.8775,
+      "step": 42270
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.138689746297211e-05,
+      "loss": 2.9715,
+      "step": 42275
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.137615742077351e-05,
+      "loss": 2.9699,
+      "step": 42280
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.136541837619719e-05,
+      "loss": 2.8401,
+      "step": 42285
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.135468032981843e-05,
+      "loss": 2.9733,
+      "step": 42290
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.134394328221242e-05,
+      "loss": 3.0795,
+      "step": 42295
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1333207233954294e-05,
+      "loss": 2.9636,
+      "step": 42300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.132247218561919e-05,
+      "loss": 2.8767,
+      "step": 42305
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.13117381377821e-05,
+      "loss": 2.9734,
+      "step": 42310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.130100509101807e-05,
+      "loss": 2.9351,
+      "step": 42315
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.129027304590201e-05,
+      "loss": 2.9444,
+      "step": 42320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.127954200300878e-05,
+      "loss": 2.9184,
+      "step": 42325
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.126881196291326e-05,
+      "loss": 2.9529,
+      "step": 42330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.125808292619017e-05,
+      "loss": 3.0176,
+      "step": 42335
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.124735489341427e-05,
+      "loss": 3.0031,
+      "step": 42340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.123662786516023e-05,
+      "loss": 3.081,
+      "step": 42345
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.122590184200263e-05,
+      "loss": 3.0028,
+      "step": 42350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.121517682451606e-05,
+      "loss": 2.938,
+      "step": 42355
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.120445281327501e-05,
+      "loss": 2.9327,
+      "step": 42360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.119372980885395e-05,
+      "loss": 3.0551,
+      "step": 42365
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.118300781182728e-05,
+      "loss": 2.9277,
+      "step": 42370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1172286822769304e-05,
+      "loss": 2.9853,
+      "step": 42375
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1161566842254364e-05,
+      "loss": 2.8835,
+      "step": 42380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1150847870856645e-05,
+      "loss": 2.906,
+      "step": 42385
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1140129909150396e-05,
+      "loss": 2.9297,
+      "step": 42390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1129412957709694e-05,
+      "loss": 2.9702,
+      "step": 42395
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.111869701710861e-05,
+      "loss": 3.0236,
+      "step": 42400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1107982087921205e-05,
+      "loss": 3.0173,
+      "step": 42405
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1097268170721405e-05,
+      "loss": 2.8776,
+      "step": 42410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1086555266083154e-05,
+      "loss": 3.0052,
+      "step": 42415
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.107584337458029e-05,
+      "loss": 2.9357,
+      "step": 42420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.106513249678661e-05,
+      "loss": 3.0238,
+      "step": 42425
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.105442263327589e-05,
+      "loss": 2.9503,
+      "step": 42430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.104371378462179e-05,
+      "loss": 2.9889,
+      "step": 42435
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.103300595139798e-05,
+      "loss": 2.919,
+      "step": 42440
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.102229913417802e-05,
+      "loss": 3.0734,
+      "step": 42445
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.1011593333535466e-05,
+      "loss": 2.8694,
+      "step": 42450
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.100088855004379e-05,
+      "loss": 2.7919,
+      "step": 42455
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.099018478427639e-05,
+      "loss": 2.9624,
+      "step": 42460
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.097948203680667e-05,
+      "loss": 2.8346,
+      "step": 42465
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.0968780308207926e-05,
+      "loss": 2.8873,
+      "step": 42470
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.0958079599053407e-05,
+      "loss": 3.0379,
+      "step": 42475
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 3.094737990991633e-05,
+      "loss": 2.841,
+      "step": 42480
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0936681241369844e-05,
+      "loss": 2.9015,
+      "step": 42485
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.092598359398704e-05,
+      "loss": 2.9872,
+      "step": 42490
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0915286968340974e-05,
+      "loss": 2.916,
+      "step": 42495
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.09045913650046e-05,
+      "loss": 3.0365,
+      "step": 42500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.089389678455087e-05,
+      "loss": 2.9102,
+      "step": 42505
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.088320322755267e-05,
+      "loss": 2.9648,
+      "step": 42510
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.087251069458279e-05,
+      "loss": 2.8473,
+      "step": 42515
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.086181918621403e-05,
+      "loss": 2.9502,
+      "step": 42520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.085112870301906e-05,
+      "loss": 2.9119,
+      "step": 42525
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.084043924557059e-05,
+      "loss": 2.8808,
+      "step": 42530
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.082975081444117e-05,
+      "loss": 3.024,
+      "step": 42535
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.081906341020339e-05,
+      "loss": 2.8144,
+      "step": 42540
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0808377033429726e-05,
+      "loss": 2.9341,
+      "step": 42545
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0797691684692585e-05,
+      "loss": 3.0346,
+      "step": 42550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.078700736456439e-05,
+      "loss": 3.0512,
+      "step": 42555
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.077632407361744e-05,
+      "loss": 2.9248,
+      "step": 42560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.076564181242403e-05,
+      "loss": 3.0629,
+      "step": 42565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.075496058155637e-05,
+      "loss": 2.814,
+      "step": 42570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.074428038158659e-05,
+      "loss": 2.9676,
+      "step": 42575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0733601213086834e-05,
+      "loss": 2.9121,
+      "step": 42580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0722923076629126e-05,
+      "loss": 2.9555,
+      "step": 42585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.071224597278549e-05,
+      "loss": 2.9212,
+      "step": 42590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.070156990212785e-05,
+      "loss": 2.9317,
+      "step": 42595
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0690894865228074e-05,
+      "loss": 2.9612,
+      "step": 42600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0680220862658014e-05,
+      "loss": 3.022,
+      "step": 42605
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.066954789498943e-05,
+      "loss": 2.7825,
+      "step": 42610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.065887596279406e-05,
+      "loss": 2.9226,
+      "step": 42615
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.064820506664355e-05,
+      "loss": 2.9539,
+      "step": 42620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.063753520710949e-05,
+      "loss": 2.9759,
+      "step": 42625
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.062686638476348e-05,
+      "loss": 3.1369,
+      "step": 42630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0616198600176966e-05,
+      "loss": 2.9909,
+      "step": 42635
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0605531853921434e-05,
+      "loss": 3.1617,
+      "step": 42640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.059486614656824e-05,
+      "loss": 2.8942,
+      "step": 42645
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0584201478688704e-05,
+      "loss": 2.9621,
+      "step": 42650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0573537850854125e-05,
+      "loss": 2.8761,
+      "step": 42655
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.05628752636357e-05,
+      "loss": 2.9603,
+      "step": 42660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0552213717604616e-05,
+      "loss": 2.9286,
+      "step": 42665
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.054155321333197e-05,
+      "loss": 2.9813,
+      "step": 42670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.053089375138879e-05,
+      "loss": 2.9908,
+      "step": 42675
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0520235332346105e-05,
+      "loss": 3.0237,
+      "step": 42680
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0509577956774814e-05,
+      "loss": 3.0236,
+      "step": 42685
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.049892162524585e-05,
+      "loss": 3.06,
+      "step": 42690
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0488266338330012e-05,
+      "loss": 3.0592,
+      "step": 42695
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.047761209659805e-05,
+      "loss": 2.9138,
+      "step": 42700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.046695890062073e-05,
+      "loss": 2.9009,
+      "step": 42705
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0456306750968655e-05,
+      "loss": 2.9777,
+      "step": 42710
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0445655648212484e-05,
+      "loss": 2.6402,
+      "step": 42715
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0435005592922734e-05,
+      "loss": 2.9792,
+      "step": 42720
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.042435658566988e-05,
+      "loss": 2.8956,
+      "step": 42725
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0413708627024384e-05,
+      "loss": 3.0362,
+      "step": 42730
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0403061717556603e-05,
+      "loss": 2.9498,
+      "step": 42735
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0392415857836898e-05,
+      "loss": 2.9545,
+      "step": 42740
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0381771048435494e-05,
+      "loss": 3.0273,
+      "step": 42745
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.037112728992261e-05,
+      "loss": 2.8694,
+      "step": 42750
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.036048458286841e-05,
+      "loss": 2.8943,
+      "step": 42755
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0349842927842987e-05,
+      "loss": 2.9313,
+      "step": 42760
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0339202325416395e-05,
+      "loss": 3.0743,
+      "step": 42765
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0328562776158593e-05,
+      "loss": 2.8943,
+      "step": 42770
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.031792428063952e-05,
+      "loss": 3.0015,
+      "step": 42775
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.030728683942905e-05,
+      "loss": 2.9613,
+      "step": 42780
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0296650453096996e-05,
+      "loss": 2.9985,
+      "step": 42785
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0286015122213118e-05,
+      "loss": 2.9103,
+      "step": 42790
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0275380847347136e-05,
+      "loss": 2.8115,
+      "step": 42795
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0264747629068656e-05,
+      "loss": 3.1285,
+      "step": 42800
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0254115467947295e-05,
+      "loss": 2.7853,
+      "step": 42805
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0243484364552578e-05,
+      "loss": 2.8528,
+      "step": 42810
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0232854319453984e-05,
+      "loss": 2.924,
+      "step": 42815
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0222225333220934e-05,
+      "loss": 3.1143,
+      "step": 42820
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0211597406422772e-05,
+      "loss": 2.9356,
+      "step": 42825
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0200970539628837e-05,
+      "loss": 2.9557,
+      "step": 42830
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.019034473340833e-05,
+      "loss": 2.7682,
+      "step": 42835
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.01797199883305e-05,
+      "loss": 3.0342,
+      "step": 42840
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.016909630496445e-05,
+      "loss": 2.8604,
+      "step": 42845
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0158473683879242e-05,
+      "loss": 2.8676,
+      "step": 42850
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.014785212564394e-05,
+      "loss": 2.9944,
+      "step": 42855
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0137231630827457e-05,
+      "loss": 3.0223,
+      "step": 42860
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0126612199998756e-05,
+      "loss": 2.9017,
+      "step": 42865
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.011599383372664e-05,
+      "loss": 2.8569,
+      "step": 42870
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0105376532579937e-05,
+      "loss": 2.961,
+      "step": 42875
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0094760297127377e-05,
+      "loss": 2.9201,
+      "step": 42880
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0084145127937603e-05,
+      "loss": 2.9054,
+      "step": 42885
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0073531025579295e-05,
+      "loss": 2.8629,
+      "step": 42890
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0062917990620952e-05,
+      "loss": 3.0718,
+      "step": 42895
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.005230602363115e-05,
+      "loss": 2.9504,
+      "step": 42900
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.004169512517831e-05,
+      "loss": 3.0117,
+      "step": 42905
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0031085295830797e-05,
+      "loss": 3.0061,
+      "step": 42910
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0020476536156995e-05,
+      "loss": 2.8891,
+      "step": 42915
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 3.0009868846725144e-05,
+      "loss": 3.1129,
+      "step": 42920
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9999262228103498e-05,
+      "loss": 2.9691,
+      "step": 42925
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.998865668086021e-05,
+      "loss": 3.0019,
+      "step": 42930
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9978052205563355e-05,
+      "loss": 2.8987,
+      "step": 42935
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.996744880278104e-05,
+      "loss": 2.9436,
+      "step": 42940
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9956846473081203e-05,
+      "loss": 2.9612,
+      "step": 42945
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.994624521703182e-05,
+      "loss": 2.8954,
+      "step": 42950
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.993564503520075e-05,
+      "loss": 2.9035,
+      "step": 42955
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9925045928155792e-05,
+      "loss": 2.8612,
+      "step": 42960
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9914447896464748e-05,
+      "loss": 2.9434,
+      "step": 42965
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9903850940695276e-05,
+      "loss": 3.0241,
+      "step": 42970
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9893255061415084e-05,
+      "loss": 2.9616,
+      "step": 42975
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9882660259191707e-05,
+      "loss": 2.9776,
+      "step": 42980
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.987206653459269e-05,
+      "loss": 2.9647,
+      "step": 42985
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.986147388818552e-05,
+      "loss": 2.9985,
+      "step": 42990
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.985088232053759e-05,
+      "loss": 2.8455,
+      "step": 42995
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.984029183221629e-05,
+      "loss": 2.99,
+      "step": 43000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9829702423788895e-05,
+      "loss": 2.8723,
+      "step": 43005
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.981911409582264e-05,
+      "loss": 3.0741,
+      "step": 43010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9808526848884743e-05,
+      "loss": 2.8328,
+      "step": 43015
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.979794068354229e-05,
+      "loss": 3.054,
+      "step": 43020
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.97873556003624e-05,
+      "loss": 2.959,
+      "step": 43025
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.977677159991204e-05,
+      "loss": 3.0548,
+      "step": 43030
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9766188682758155e-05,
+      "loss": 3.0055,
+      "step": 43035
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.975560684946769e-05,
+      "loss": 2.8161,
+      "step": 43040
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.974502610060743e-05,
+      "loss": 2.9116,
+      "step": 43045
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.973444643674419e-05,
+      "loss": 2.9387,
+      "step": 43050
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9723867858444675e-05,
+      "loss": 3.0449,
+      "step": 43055
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9713290366275537e-05,
+      "loss": 2.9276,
+      "step": 43060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.970271396080341e-05,
+      "loss": 2.9806,
+      "step": 43065
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.969213864259479e-05,
+      "loss": 2.9744,
+      "step": 43070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9681564412216223e-05,
+      "loss": 2.9198,
+      "step": 43075
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.96709912702341e-05,
+      "loss": 2.8892,
+      "step": 43080
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9660419217214796e-05,
+      "loss": 3.0155,
+      "step": 43085
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9649848253724637e-05,
+      "loss": 3.0538,
+      "step": 43090
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9639278380329864e-05,
+      "loss": 2.8657,
+      "step": 43095
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9628709597596684e-05,
+      "loss": 2.954,
+      "step": 43100
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9618141906091234e-05,
+      "loss": 3.035,
+      "step": 43105
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9607575306379575e-05,
+      "loss": 2.767,
+      "step": 43110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9597009799027747e-05,
+      "loss": 2.8431,
+      "step": 43115
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.95864453846017e-05,
+      "loss": 3.1194,
+      "step": 43120
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.957588206366735e-05,
+      "loss": 2.9708,
+      "step": 43125
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.956531983679054e-05,
+      "loss": 2.971,
+      "step": 43130
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9554758704537024e-05,
+      "loss": 3.0972,
+      "step": 43135
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.954419866747258e-05,
+      "loss": 2.8772,
+      "step": 43140
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.953363972616284e-05,
+      "loss": 3.0543,
+      "step": 43145
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9523081881173437e-05,
+      "loss": 3.0872,
+      "step": 43150
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9512525133069914e-05,
+      "loss": 2.9541,
+      "step": 43155
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2.9501969482417747e-05,
+      "loss": 2.8503,
+      "step": 43160
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9491414929782403e-05,
+      "loss": 3.0089,
+      "step": 43165
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9480861475729215e-05,
+      "loss": 2.9031,
+      "step": 43170
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9470309120823554e-05,
+      "loss": 2.8601,
+      "step": 43175
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9459757865630638e-05,
+      "loss": 3.0308,
+      "step": 43180
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9449207710715654e-05,
+      "loss": 3.0009,
+      "step": 43185
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.943865865664379e-05,
+      "loss": 2.9479,
+      "step": 43190
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.942811070398007e-05,
+      "loss": 2.8524,
+      "step": 43195
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9417563853289566e-05,
+      "loss": 2.9316,
+      "step": 43200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9407018105137223e-05,
+      "loss": 3.0403,
+      "step": 43205
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9396473460087916e-05,
+      "loss": 2.9572,
+      "step": 43210
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9385929918706533e-05,
+      "loss": 3.0026,
+      "step": 43215
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9375387481557825e-05,
+      "loss": 2.9425,
+      "step": 43220
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9364846149206547e-05,
+      "loss": 2.9568,
+      "step": 43225
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.935430592221735e-05,
+      "loss": 2.9569,
+      "step": 43230
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9343766801154826e-05,
+      "loss": 3.012,
+      "step": 43235
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9333228786583566e-05,
+      "loss": 2.9628,
+      "step": 43240
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.932269187906801e-05,
+      "loss": 2.8035,
+      "step": 43245
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.931215607917263e-05,
+      "loss": 2.9663,
+      "step": 43250
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9301621387461786e-05,
+      "loss": 2.908,
+      "step": 43255
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.929108780449975e-05,
+      "loss": 2.9135,
+      "step": 43260
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9280555330850834e-05,
+      "loss": 2.959,
+      "step": 43265
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9270023967079185e-05,
+      "loss": 2.9184,
+      "step": 43270
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9259493713748963e-05,
+      "loss": 2.7767,
+      "step": 43275
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.924896457142424e-05,
+      "loss": 2.9698,
+      "step": 43280
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9238436540669e-05,
+      "loss": 2.8781,
+      "step": 43285
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9227909622047234e-05,
+      "loss": 2.8663,
+      "step": 43290
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9217383816122807e-05,
+      "loss": 2.9752,
+      "step": 43295
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9206859123459583e-05,
+      "loss": 3.0463,
+      "step": 43300
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9196335544621334e-05,
+      "loss": 3.0627,
+      "step": 43305
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9185813080171736e-05,
+      "loss": 3.0292,
+      "step": 43310
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9175291730674493e-05,
+      "loss": 3.077,
+      "step": 43315
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9164771496693178e-05,
+      "loss": 2.9574,
+      "step": 43320
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9154252378791323e-05,
+      "loss": 3.0238,
+      "step": 43325
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.914373437753245e-05,
+      "loss": 2.9126,
+      "step": 43330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9133217493479898e-05,
+      "loss": 3.0309,
+      "step": 43335
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9122701727197105e-05,
+      "loss": 3.0414,
+      "step": 43340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9112187079247288e-05,
+      "loss": 2.9721,
+      "step": 43345
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9101673550193777e-05,
+      "loss": 2.7587,
+      "step": 43350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9091161140599667e-05,
+      "loss": 2.9217,
+      "step": 43355
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.908064985102812e-05,
+      "loss": 2.9469,
+      "step": 43360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9070139682042176e-05,
+      "loss": 3.0227,
+      "step": 43365
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9059630634204837e-05,
+      "loss": 2.8322,
+      "step": 43370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9049122708079045e-05,
+      "loss": 2.8016,
+      "step": 43375
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.90386159042277e-05,
+      "loss": 2.8137,
+      "step": 43380
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.902811022321355e-05,
+      "loss": 3.0307,
+      "step": 43385
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9017605665599434e-05,
+      "loss": 3.006,
+      "step": 43390
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.9007102231947976e-05,
+      "loss": 2.8718,
+      "step": 43395
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8996599922821888e-05,
+      "loss": 3.0877,
+      "step": 43400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8986098738783683e-05,
+      "loss": 3.1647,
+      "step": 43405
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8975598680395904e-05,
+      "loss": 2.8384,
+      "step": 43410
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8965099748221003e-05,
+      "loss": 2.924,
+      "step": 43415
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.895460194282138e-05,
+      "loss": 2.8724,
+      "step": 43420
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8944105264759358e-05,
+      "loss": 2.9227,
+      "step": 43425
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8933609714597254e-05,
+      "loss": 2.8917,
+      "step": 43430
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8923115292897202e-05,
+      "loss": 3.1292,
+      "step": 43435
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8912622000221447e-05,
+      "loss": 2.9427,
+      "step": 43440
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8902129837131998e-05,
+      "loss": 2.8301,
+      "step": 43445
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8891638804190966e-05,
+      "loss": 3.0445,
+      "step": 43450
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.888114890196027e-05,
+      "loss": 2.8526,
+      "step": 43455
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8870660131001836e-05,
+      "loss": 3.0287,
+      "step": 43460
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8860172491877523e-05,
+      "loss": 2.9524,
+      "step": 43465
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8849685985149118e-05,
+      "loss": 2.8764,
+      "step": 43470
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.883920061137835e-05,
+      "loss": 2.9451,
+      "step": 43475
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.88287163711269e-05,
+      "loss": 2.9384,
+      "step": 43480
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8818233264956336e-05,
+      "loss": 2.992,
+      "step": 43485
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.880775129342827e-05,
+      "loss": 2.8794,
+      "step": 43490
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8797270457104112e-05,
+      "loss": 2.9856,
+      "step": 43495
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.878679075654538e-05,
+      "loss": 2.9581,
+      "step": 43500
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.877631219231336e-05,
+      "loss": 2.9039,
+      "step": 43505
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.876583476496939e-05,
+      "loss": 2.9663,
+      "step": 43510
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8755358475074713e-05,
+      "loss": 3.1169,
+      "step": 43515
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8744883323190507e-05,
+      "loss": 2.9582,
+      "step": 43520
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8734409309877895e-05,
+      "loss": 2.9245,
+      "step": 43525
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8723936435697946e-05,
+      "loss": 2.9028,
+      "step": 43530
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8713464701211645e-05,
+      "loss": 2.9386,
+      "step": 43535
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8702994106979942e-05,
+      "loss": 2.955,
+      "step": 43540
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8692524653563713e-05,
+      "loss": 3.0673,
+      "step": 43545
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8682056341523773e-05,
+      "loss": 3.0346,
+      "step": 43550
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8671589171420876e-05,
+      "loss": 2.8598,
+      "step": 43555
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8661123143815737e-05,
+      "loss": 2.9533,
+      "step": 43560
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8650658259268954e-05,
+      "loss": 3.1074,
+      "step": 43565
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.864019451834111e-05,
+      "loss": 3.1522,
+      "step": 43570
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8629731921592724e-05,
+      "loss": 2.831,
+      "step": 43575
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8619270469584238e-05,
+      "loss": 2.964,
+      "step": 43580
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8608810162876044e-05,
+      "loss": 3.0993,
+      "step": 43585
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8598351002028474e-05,
+      "loss": 2.9274,
+      "step": 43590
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8587892987601784e-05,
+      "loss": 2.9997,
+      "step": 43595
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8577436120156174e-05,
+      "loss": 2.8754,
+      "step": 43600
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8566980400251807e-05,
+      "loss": 2.8765,
+      "step": 43605
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8556525828448745e-05,
+      "loss": 3.0574,
+      "step": 43610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8546072405307035e-05,
+      "loss": 3.0834,
+      "step": 43615
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8535620131386563e-05,
+      "loss": 2.8876,
+      "step": 43620
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8525169007247327e-05,
+      "loss": 2.8413,
+      "step": 43625
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8514719033449065e-05,
+      "loss": 3.0049,
+      "step": 43630
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8504270210551644e-05,
+      "loss": 2.8709,
+      "step": 43635
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8493822539114694e-05,
+      "loss": 3.0257,
+      "step": 43640
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8483376019697904e-05,
+      "loss": 3.0304,
+      "step": 43645
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.847293065286085e-05,
+      "loss": 2.952,
+      "step": 43650
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8462486439163072e-05,
+      "loss": 2.9075,
+      "step": 43655
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.845204337916403e-05,
+      "loss": 2.9408,
+      "step": 43660
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8441601473423135e-05,
+      "loss": 2.8734,
+      "step": 43665
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8431160722499677e-05,
+      "loss": 2.9363,
+      "step": 43670
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.842072112695302e-05,
+      "loss": 2.9552,
+      "step": 43675
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8410282687342292e-05,
+      "loss": 2.873,
+      "step": 43680
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8399845404226733e-05,
+      "loss": 2.8963,
+      "step": 43685
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.838940927816537e-05,
+      "loss": 2.9703,
+      "step": 43690
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8378974309717265e-05,
+      "loss": 2.966,
+      "step": 43695
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.836854049944138e-05,
+      "loss": 2.8919,
+      "step": 43700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.835810784789663e-05,
+      "loss": 2.9892,
+      "step": 43705
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.834767635564185e-05,
+      "loss": 3.0243,
+      "step": 43710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8337246023235848e-05,
+      "loss": 2.9891,
+      "step": 43715
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8326816851237282e-05,
+      "loss": 3.0247,
+      "step": 43720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.83163888402049e-05,
+      "loss": 2.8624,
+      "step": 43725
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8305961990697206e-05,
+      "loss": 2.9531,
+      "step": 43730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8295536303272824e-05,
+      "loss": 3.0038,
+      "step": 43735
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8285111778490165e-05,
+      "loss": 3.028,
+      "step": 43740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.827468841690765e-05,
+      "loss": 2.8753,
+      "step": 43745
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8264266219083642e-05,
+      "loss": 2.7801,
+      "step": 43750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.825384518557641e-05,
+      "loss": 2.9975,
+      "step": 43755
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8243425316944187e-05,
+      "loss": 2.988,
+      "step": 43760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.823300661374515e-05,
+      "loss": 2.9157,
+      "step": 43765
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.822258907653734e-05,
+      "loss": 2.9009,
+      "step": 43770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8212172705878875e-05,
+      "loss": 3.0711,
+      "step": 43775
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8201757502327642e-05,
+      "loss": 2.9422,
+      "step": 43780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8191343466441627e-05,
+      "loss": 2.9946,
+      "step": 43785
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.818093059877863e-05,
+      "loss": 3.0514,
+      "step": 43790
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8170518899896446e-05,
+      "loss": 2.9401,
+      "step": 43795
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.816010837035281e-05,
+      "loss": 2.9578,
+      "step": 43800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.814969901070537e-05,
+      "loss": 2.9558,
+      "step": 43805
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8139290821511732e-05,
+      "loss": 2.9165,
+      "step": 43810
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.812888380332942e-05,
+      "loss": 2.853,
+      "step": 43815
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.811847795671592e-05,
+      "loss": 3.0297,
+      "step": 43820
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8108073282228653e-05,
+      "loss": 3.085,
+      "step": 43825
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.809766978042491e-05,
+      "loss": 2.8922,
+      "step": 43830
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.8087267451862053e-05,
+      "loss": 2.8058,
+      "step": 43835
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2.807686629709724e-05,
+      "loss": 2.9518,
+      "step": 43840
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8066466316687655e-05,
+      "loss": 2.9723,
+      "step": 43845
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8056067511190388e-05,
+      "loss": 2.8853,
+      "step": 43850
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8045669881162466e-05,
+      "loss": 2.9196,
+      "step": 43855
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.803527342716087e-05,
+      "loss": 2.9204,
+      "step": 43860
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8024878149742506e-05,
+      "loss": 2.9377,
+      "step": 43865
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8014484049464207e-05,
+      "loss": 3.0082,
+      "step": 43870
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.8004091126882758e-05,
+      "loss": 3.0083,
+      "step": 43875
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.799369938255487e-05,
+      "loss": 2.9633,
+      "step": 43880
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.798330881703722e-05,
+      "loss": 3.0758,
+      "step": 43885
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7972919430886392e-05,
+      "loss": 3.0375,
+      "step": 43890
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.796253122465886e-05,
+      "loss": 2.9257,
+      "step": 43895
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7952144198911183e-05,
+      "loss": 2.9193,
+      "step": 43900
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7941758354199687e-05,
+      "loss": 2.8282,
+      "step": 43905
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7931373691080732e-05,
+      "loss": 2.9534,
+      "step": 43910
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7920990210110597e-05,
+      "loss": 2.7923,
+      "step": 43915
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.79106079118455e-05,
+      "loss": 2.8708,
+      "step": 43920
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7900226796841565e-05,
+      "loss": 2.9342,
+      "step": 43925
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7889846865654907e-05,
+      "loss": 2.9152,
+      "step": 43930
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7879468118841523e-05,
+      "loss": 3.07,
+      "step": 43935
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.78690905569574e-05,
+      "loss": 2.8054,
+      "step": 43940
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7858714180558378e-05,
+      "loss": 2.8491,
+      "step": 43945
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7848338990200363e-05,
+      "loss": 2.9878,
+      "step": 43950
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7837964986439034e-05,
+      "loss": 2.94,
+      "step": 43955
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7827592169830186e-05,
+      "loss": 2.9717,
+      "step": 43960
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7817220540929396e-05,
+      "loss": 2.9253,
+      "step": 43965
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7806850100292258e-05,
+      "loss": 2.8545,
+      "step": 43970
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.779648084847428e-05,
+      "loss": 2.9184,
+      "step": 43975
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7786112786030926e-05,
+      "loss": 2.9732,
+      "step": 43980
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7775745913517572e-05,
+      "loss": 2.8554,
+      "step": 43985
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7765380231489556e-05,
+      "loss": 2.7291,
+      "step": 43990
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.775501574050209e-05,
+      "loss": 2.9791,
+      "step": 43995
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7744652441110426e-05,
+      "loss": 2.8488,
+      "step": 44000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.773429033386963e-05,
+      "loss": 2.9187,
+      "step": 44005
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7723929419334848e-05,
+      "loss": 2.9789,
+      "step": 44010
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7713569698061014e-05,
+      "loss": 2.9202,
+      "step": 44015
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.77032111706031e-05,
+      "loss": 2.9703,
+      "step": 44020
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7692853837515968e-05,
+      "loss": 2.9665,
+      "step": 44025
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7682497699354437e-05,
+      "loss": 3.0641,
+      "step": 44030
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7672142756673248e-05,
+      "loss": 2.8781,
+      "step": 44035
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7661789010027105e-05,
+      "loss": 3.0262,
+      "step": 44040
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7651436459970565e-05,
+      "loss": 2.9634,
+      "step": 44045
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7641085107058267e-05,
+      "loss": 2.9487,
+      "step": 44050
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7630734951844617e-05,
+      "loss": 2.904,
+      "step": 44055
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7620385994884123e-05,
+      "loss": 2.9327,
+      "step": 44060
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7610038236731095e-05,
+      "loss": 2.995,
+      "step": 44065
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7599691677939833e-05,
+      "loss": 3.0187,
+      "step": 44070
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7589346319064585e-05,
+      "loss": 2.9023,
+      "step": 44075
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7579002160659518e-05,
+      "loss": 2.9726,
+      "step": 44080
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.756865920327873e-05,
+      "loss": 2.8312,
+      "step": 44085
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7558317447476288e-05,
+      "loss": 3.0087,
+      "step": 44090
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7547976893806105e-05,
+      "loss": 2.8772,
+      "step": 44095
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.753763754282217e-05,
+      "loss": 2.8851,
+      "step": 44100
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7527299395078264e-05,
+      "loss": 2.9522,
+      "step": 44105
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7516962451128237e-05,
+      "loss": 2.8713,
+      "step": 44110
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.750662671152575e-05,
+      "loss": 3.0472,
+      "step": 44115
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7496292176824478e-05,
+      "loss": 2.946,
+      "step": 44120
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.748595884757801e-05,
+      "loss": 2.8748,
+      "step": 44125
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.747562672433987e-05,
+      "loss": 2.8317,
+      "step": 44130
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7465295807663522e-05,
+      "loss": 2.9326,
+      "step": 44135
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7454966098102354e-05,
+      "loss": 2.9052,
+      "step": 44140
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.744463759620971e-05,
+      "loss": 2.9539,
+      "step": 44145
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7434310302538847e-05,
+      "loss": 2.8946,
+      "step": 44150
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.742398421764296e-05,
+      "loss": 2.8979,
+      "step": 44155
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.741365934207522e-05,
+      "loss": 2.8811,
+      "step": 44160
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7403335676388654e-05,
+      "loss": 2.8996,
+      "step": 44165
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7393013221136283e-05,
+      "loss": 2.9742,
+      "step": 44170
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7382691976871057e-05,
+      "loss": 2.8907,
+      "step": 44175
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7372371944145847e-05,
+      "loss": 2.9667,
+      "step": 44180
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7362053123513475e-05,
+      "loss": 2.9228,
+      "step": 44185
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7351735515526678e-05,
+      "loss": 3.0407,
+      "step": 44190
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7341419120738147e-05,
+      "loss": 2.8432,
+      "step": 44195
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7331103939700498e-05,
+      "loss": 2.8145,
+      "step": 44200
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7320789972966276e-05,
+      "loss": 2.9013,
+      "step": 44205
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7310477221087982e-05,
+      "loss": 2.9237,
+      "step": 44210
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.730016568461803e-05,
+      "loss": 2.9135,
+      "step": 44215
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7289855364108776e-05,
+      "loss": 2.8684,
+      "step": 44220
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7279546260112533e-05,
+      "loss": 2.9344,
+      "step": 44225
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.726923837318147e-05,
+      "loss": 3.0438,
+      "step": 44230
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7258931703867834e-05,
+      "loss": 3.0438,
+      "step": 44235
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7248626252723637e-05,
+      "loss": 3.0138,
+      "step": 44240
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7238322020300988e-05,
+      "loss": 2.9003,
+      "step": 44245
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7228019007151793e-05,
+      "loss": 2.9471,
+      "step": 44250
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.721771721382798e-05,
+      "loss": 3.0185,
+      "step": 44255
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7207416640881378e-05,
+      "loss": 2.9553,
+      "step": 44260
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7197117288863762e-05,
+      "loss": 2.8703,
+      "step": 44265
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7186819158326833e-05,
+      "loss": 2.8526,
+      "step": 44270
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.717652224982225e-05,
+      "loss": 3.0246,
+      "step": 44275
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.716622656390153e-05,
+      "loss": 3.0749,
+      "step": 44280
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7155932101116255e-05,
+      "loss": 3.0604,
+      "step": 44285
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7145638862017797e-05,
+      "loss": 2.9665,
+      "step": 44290
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7135346847157617e-05,
+      "loss": 2.9963,
+      "step": 44295
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.712505605708695e-05,
+      "loss": 2.9591,
+      "step": 44300
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7114766492357068e-05,
+      "loss": 2.953,
+      "step": 44305
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.710447815351916e-05,
+      "loss": 2.7435,
+      "step": 44310
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7094191041124335e-05,
+      "loss": 2.9187,
+      "step": 44315
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7083905155723637e-05,
+      "loss": 2.8022,
+      "step": 44320
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.707362049786808e-05,
+      "loss": 3.019,
+      "step": 44325
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.706333706810852e-05,
+      "loss": 3.0669,
+      "step": 44330
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7053054866995875e-05,
+      "loss": 2.9757,
+      "step": 44335
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7042773895080868e-05,
+      "loss": 2.8133,
+      "step": 44340
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7032494152914295e-05,
+      "loss": 2.8831,
+      "step": 44345
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7022215641046743e-05,
+      "loss": 3.2001,
+      "step": 44350
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.7011938360028832e-05,
+      "loss": 2.8393,
+      "step": 44355
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.700166231041107e-05,
+      "loss": 3.0617,
+      "step": 44360
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6991387492743926e-05,
+      "loss": 2.9107,
+      "step": 44365
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6981113907577782e-05,
+      "loss": 2.8812,
+      "step": 44370
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6970841555462995e-05,
+      "loss": 2.8953,
+      "step": 44375
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.696057043694975e-05,
+      "loss": 3.0513,
+      "step": 44380
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6950300552588325e-05,
+      "loss": 2.8861,
+      "step": 44385
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6940031902928764e-05,
+      "loss": 2.8208,
+      "step": 44390
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6929764488521213e-05,
+      "loss": 2.8506,
+      "step": 44395
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6919498309915603e-05,
+      "loss": 3.0213,
+      "step": 44400
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6909233367661874e-05,
+      "loss": 2.9754,
+      "step": 44405
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6898969662309905e-05,
+      "loss": 2.8974,
+      "step": 44410
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.688870719440947e-05,
+      "loss": 3.0052,
+      "step": 44415
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.687844596451031e-05,
+      "loss": 3.0597,
+      "step": 44420
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6868185973162106e-05,
+      "loss": 2.8318,
+      "step": 44425
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6857927220914402e-05,
+      "loss": 2.9585,
+      "step": 44430
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6847669708316797e-05,
+      "loss": 2.9747,
+      "step": 44435
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.683741343591868e-05,
+      "loss": 2.9545,
+      "step": 44440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.682715840426952e-05,
+      "loss": 2.9372,
+      "step": 44445
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6816904613918603e-05,
+      "loss": 2.7272,
+      "step": 44450
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6806652065415193e-05,
+      "loss": 2.8869,
+      "step": 44455
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.679640075930851e-05,
+      "loss": 2.9068,
+      "step": 44460
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6786150696147673e-05,
+      "loss": 2.9977,
+      "step": 44465
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.677590187648174e-05,
+      "loss": 2.9946,
+      "step": 44470
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.676565430085972e-05,
+      "loss": 2.865,
+      "step": 44475
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.675540796983055e-05,
+      "loss": 2.8909,
+      "step": 44480
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6745162883943074e-05,
+      "loss": 2.8933,
+      "step": 44485
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6734919043746103e-05,
+      "loss": 2.9013,
+      "step": 44490
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6724676449788365e-05,
+      "loss": 2.8277,
+      "step": 44495
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6714435102618552e-05,
+      "loss": 2.8733,
+      "step": 44500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6704195002785214e-05,
+      "loss": 3.038,
+      "step": 44505
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.669395615083691e-05,
+      "loss": 2.8294,
+      "step": 44510
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6683718547322085e-05,
+      "loss": 3.0219,
+      "step": 44515
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2.6673482192789156e-05,
+      "loss": 2.9892,
+      "step": 44520
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.666324708778644e-05,
+      "loss": 2.9179,
+      "step": 44525
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.665301323286221e-05,
+      "loss": 2.8241,
+      "step": 44530
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6642780628564655e-05,
+      "loss": 2.912,
+      "step": 44535
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.663254927544192e-05,
+      "loss": 2.8901,
+      "step": 44540
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6622319174042045e-05,
+      "loss": 2.9045,
+      "step": 44545
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.661209032491305e-05,
+      "loss": 2.8622,
+      "step": 44550
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6601862728602823e-05,
+      "loss": 2.9907,
+      "step": 44555
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.659163638565928e-05,
+      "loss": 2.9562,
+      "step": 44560
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6581411296630145e-05,
+      "loss": 2.9089,
+      "step": 44565
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.657118746206323e-05,
+      "loss": 2.9398,
+      "step": 44570
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.656096488250612e-05,
+      "loss": 2.7626,
+      "step": 44575
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.655074355850645e-05,
+      "loss": 3.0325,
+      "step": 44580
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6540523490611723e-05,
+      "loss": 2.8106,
+      "step": 44585
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.653030467936941e-05,
+      "loss": 2.7754,
+      "step": 44590
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6520087125326898e-05,
+      "loss": 2.8143,
+      "step": 44595
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.650987082903153e-05,
+      "loss": 2.8726,
+      "step": 44600
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6499655791030497e-05,
+      "loss": 2.9391,
+      "step": 44605
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.648944201187108e-05,
+      "loss": 3.1005,
+      "step": 44610
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.64792294921003e-05,
+      "loss": 2.8543,
+      "step": 44615
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6469018232265307e-05,
+      "loss": 2.8279,
+      "step": 44620
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6458808232913024e-05,
+      "loss": 2.96,
+      "step": 44625
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6448599494590388e-05,
+      "loss": 2.9316,
+      "step": 44630
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6438392017844242e-05,
+      "loss": 3.0586,
+      "step": 44635
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6428185803221382e-05,
+      "loss": 2.908,
+      "step": 44640
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.641798085126852e-05,
+      "loss": 2.9444,
+      "step": 44645
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6407777162532323e-05,
+      "loss": 2.9993,
+      "step": 44650
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6397574737559304e-05,
+      "loss": 2.9035,
+      "step": 44655
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.638737357689606e-05,
+      "loss": 2.9287,
+      "step": 44660
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.637717368108897e-05,
+      "loss": 2.7672,
+      "step": 44665
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.636697505068447e-05,
+      "loss": 2.9418,
+      "step": 44670
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6356777686228823e-05,
+      "loss": 2.9282,
+      "step": 44675
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.634658158826829e-05,
+      "loss": 2.9165,
+      "step": 44680
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6336386757349037e-05,
+      "loss": 2.996,
+      "step": 44685
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6326193194017172e-05,
+      "loss": 2.8937,
+      "step": 44690
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.631600089881874e-05,
+      "loss": 2.9894,
+      "step": 44695
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6305809872299726e-05,
+      "loss": 2.9223,
+      "step": 44700
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.629562011500597e-05,
+      "loss": 2.9607,
+      "step": 44705
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6285431627483392e-05,
+      "loss": 2.9697,
+      "step": 44710
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6275244410277666e-05,
+      "loss": 3.0276,
+      "step": 44715
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6265058463934587e-05,
+      "loss": 2.9164,
+      "step": 44720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.625487378899972e-05,
+      "loss": 3.0474,
+      "step": 44725
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.624469038601864e-05,
+      "loss": 2.8495,
+      "step": 44730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.623450825553684e-05,
+      "loss": 2.9649,
+      "step": 44735
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.622432739809976e-05,
+      "loss": 2.8958,
+      "step": 44740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6214147814252737e-05,
+      "loss": 2.9788,
+      "step": 44745
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6203969504541078e-05,
+      "loss": 2.8577,
+      "step": 44750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.619379246950999e-05,
+      "loss": 3.0304,
+      "step": 44755
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.618361670970464e-05,
+      "loss": 2.8527,
+      "step": 44760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6173442225670096e-05,
+      "loss": 3.0163,
+      "step": 44765
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.616326901795141e-05,
+      "loss": 3.0269,
+      "step": 44770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.615309708709348e-05,
+      "loss": 2.9791,
+      "step": 44775
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6142926433641202e-05,
+      "loss": 2.8751,
+      "step": 44780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.61327570581394e-05,
+      "loss": 2.9806,
+      "step": 44785
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6122588961132805e-05,
+      "loss": 2.9285,
+      "step": 44790
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6112422143166093e-05,
+      "loss": 3.1147,
+      "step": 44795
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6102256604783874e-05,
+      "loss": 2.9565,
+      "step": 44800
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6092092346530684e-05,
+      "loss": 2.9354,
+      "step": 44805
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6081929368950985e-05,
+      "loss": 2.8885,
+      "step": 44810
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6071767672589175e-05,
+      "loss": 2.9562,
+      "step": 44815
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.60616072579896e-05,
+      "loss": 2.9467,
+      "step": 44820
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6051448125696532e-05,
+      "loss": 2.9689,
+      "step": 44825
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.60412902762541e-05,
+      "loss": 2.8634,
+      "step": 44830
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6031133710206514e-05,
+      "loss": 2.8801,
+      "step": 44835
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6020978428097753e-05,
+      "loss": 2.9294,
+      "step": 44840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.6010824430471882e-05,
+      "loss": 3.0684,
+      "step": 44845
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.600067171787276e-05,
+      "loss": 2.9325,
+      "step": 44850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5990520290844256e-05,
+      "loss": 3.0767,
+      "step": 44855
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5980370149930155e-05,
+      "loss": 2.9235,
+      "step": 44860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.597022129567416e-05,
+      "loss": 2.7738,
+      "step": 44865
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5960073728619915e-05,
+      "loss": 3.0461,
+      "step": 44870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5949927449311006e-05,
+      "loss": 2.8977,
+      "step": 44875
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.593978245829093e-05,
+      "loss": 2.8707,
+      "step": 44880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.592963875610314e-05,
+      "loss": 2.9995,
+      "step": 44885
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.591949634329095e-05,
+      "loss": 2.866,
+      "step": 44890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5909355220397736e-05,
+      "loss": 3.035,
+      "step": 44895
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5899215387966642e-05,
+      "loss": 3.0145,
+      "step": 44900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.588907684654091e-05,
+      "loss": 3.0995,
+      "step": 44905
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5878939596663575e-05,
+      "loss": 2.9678,
+      "step": 44910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.586880363887768e-05,
+      "loss": 2.86,
+      "step": 44915
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5858668973726165e-05,
+      "loss": 2.9636,
+      "step": 44920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.584853560175193e-05,
+      "loss": 2.9269,
+      "step": 44925
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5838403523497774e-05,
+      "loss": 2.855,
+      "step": 44930
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5828272739506466e-05,
+      "loss": 2.9161,
+      "step": 44935
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5818143250320625e-05,
+      "loss": 2.9512,
+      "step": 44940
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.580801505648293e-05,
+      "loss": 2.8636,
+      "step": 44945
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5797888158535843e-05,
+      "loss": 2.9628,
+      "step": 44950
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5787762557021917e-05,
+      "loss": 2.9851,
+      "step": 44955
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5777638252483472e-05,
+      "loss": 2.9419,
+      "step": 44960
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5767515245462868e-05,
+      "loss": 2.8446,
+      "step": 44965
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5757393536502367e-05,
+      "loss": 3.1126,
+      "step": 44970
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.574727312614415e-05,
+      "loss": 2.9922,
+      "step": 44975
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5737154014930337e-05,
+      "loss": 2.8938,
+      "step": 44980
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5727036203403e-05,
+      "loss": 2.9739,
+      "step": 44985
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5716919692104058e-05,
+      "loss": 2.907,
+      "step": 44990
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5706804481575507e-05,
+      "loss": 2.953,
+      "step": 44995
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5696690572359095e-05,
+      "loss": 2.9684,
+      "step": 45000
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5686577964996688e-05,
+      "loss": 2.8517,
+      "step": 45005
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.567646666002992e-05,
+      "loss": 2.9605,
+      "step": 45010
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5666356658000447e-05,
+      "loss": 2.9302,
+      "step": 45015
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.565624795944982e-05,
+      "loss": 2.8766,
+      "step": 45020
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.564614056491954e-05,
+      "loss": 2.8973,
+      "step": 45025
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.563603447495103e-05,
+      "loss": 3.0139,
+      "step": 45030
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5625929690085664e-05,
+      "loss": 3.0386,
+      "step": 45035
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.561582621086466e-05,
+      "loss": 2.9252,
+      "step": 45040
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5605724037829305e-05,
+      "loss": 3.0152,
+      "step": 45045
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5595623171520678e-05,
+      "loss": 2.9135,
+      "step": 45050
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5585523612479913e-05,
+      "loss": 3.0134,
+      "step": 45055
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5575425361247962e-05,
+      "loss": 3.0233,
+      "step": 45060
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5565328418365775e-05,
+      "loss": 2.8856,
+      "step": 45065
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5555232784374217e-05,
+      "loss": 3.0272,
+      "step": 45070
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5545138459814076e-05,
+      "loss": 2.8249,
+      "step": 45075
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.553504544522608e-05,
+      "loss": 3.0144,
+      "step": 45080
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5524953741150876e-05,
+      "loss": 2.9665,
+      "step": 45085
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5514863348129048e-05,
+      "loss": 2.8856,
+      "step": 45090
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5504774266701102e-05,
+      "loss": 2.9863,
+      "step": 45095
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5494686497407494e-05,
+      "loss": 2.9061,
+      "step": 45100
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5484600040788577e-05,
+      "loss": 2.967,
+      "step": 45105
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.547451489738468e-05,
+      "loss": 2.9967,
+      "step": 45110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5464431067736e-05,
+      "loss": 2.8284,
+      "step": 45115
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.54543485523827e-05,
+      "loss": 3.0317,
+      "step": 45120
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5444267351864892e-05,
+      "loss": 3.0037,
+      "step": 45125
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.543418746672257e-05,
+      "loss": 2.893,
+      "step": 45130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5424108897495702e-05,
+      "loss": 2.9482,
+      "step": 45135
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5414031644724158e-05,
+      "loss": 2.9663,
+      "step": 45140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5403955708947745e-05,
+      "loss": 2.938,
+      "step": 45145
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5393881090706206e-05,
+      "loss": 2.9428,
+      "step": 45150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5383807790539206e-05,
+      "loss": 2.8533,
+      "step": 45155
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5373735808986348e-05,
+      "loss": 3.0074,
+      "step": 45160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5363665146587112e-05,
+      "loss": 2.9561,
+      "step": 45165
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.535359580388103e-05,
+      "loss": 2.9906,
+      "step": 45170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5343527781407405e-05,
+      "loss": 2.9087,
+      "step": 45175
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.533346107970562e-05,
+      "loss": 3.0915,
+      "step": 45180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.532339569931488e-05,
+      "loss": 2.7615,
+      "step": 45185
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.531333164077436e-05,
+      "loss": 2.9239,
+      "step": 45190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2.5303268904623156e-05,
+      "loss": 2.9721,
+      "step": 45195
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5293207491400316e-05,
+      "loss": 2.9428,
+      "step": 45200
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.528314740164478e-05,
+      "loss": 2.8388,
+      "step": 45205
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.527308863589547e-05,
+      "loss": 3.0345,
+      "step": 45210
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5263031194691146e-05,
+      "loss": 3.0658,
+      "step": 45215
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.525297507857062e-05,
+      "loss": 2.9428,
+      "step": 45220
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.524292028807249e-05,
+      "loss": 2.8181,
+      "step": 45225
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5232866823735446e-05,
+      "loss": 3.122,
+      "step": 45230
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5222814686097968e-05,
+      "loss": 2.8844,
+      "step": 45235
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5212763875698527e-05,
+      "loss": 2.7614,
+      "step": 45240
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5202714393075516e-05,
+      "loss": 2.9493,
+      "step": 45245
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.519266623876726e-05,
+      "loss": 3.0123,
+      "step": 45250
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5182619413312004e-05,
+      "loss": 2.9385,
+      "step": 45255
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.517257391724794e-05,
+      "loss": 2.9224,
+      "step": 45260
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5162529751113128e-05,
+      "loss": 2.8656,
+      "step": 45265
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5152486915445673e-05,
+      "loss": 2.9647,
+      "step": 45270
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.514244541078346e-05,
+      "loss": 2.8911,
+      "step": 45275
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.513240523766447e-05,
+      "loss": 3.022,
+      "step": 45280
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5122366396626453e-05,
+      "loss": 2.8984,
+      "step": 45285
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.511232888820718e-05,
+      "loss": 2.9819,
+      "step": 45290
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5102292712944342e-05,
+      "loss": 2.8763,
+      "step": 45295
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5092257871375542e-05,
+      "loss": 2.9028,
+      "step": 45300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5082224364038302e-05,
+      "loss": 2.967,
+      "step": 45305
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5072192191470122e-05,
+      "loss": 2.9496,
+      "step": 45310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5062161354208323e-05,
+      "loss": 2.8937,
+      "step": 45315
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5052131852790317e-05,
+      "loss": 2.8063,
+      "step": 45320
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.504210368775327e-05,
+      "loss": 2.7055,
+      "step": 45325
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5032076859634434e-05,
+      "loss": 2.9654,
+      "step": 45330
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5022051368970867e-05,
+      "loss": 2.9436,
+      "step": 45335
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5012027216299615e-05,
+      "loss": 2.9682,
+      "step": 45340
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.5002004402157643e-05,
+      "loss": 2.9007,
+      "step": 45345
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.499198292708184e-05,
+      "loss": 2.9522,
+      "step": 45350
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4981962791609032e-05,
+      "loss": 2.9356,
+      "step": 45355
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4971943996275966e-05,
+      "loss": 2.8677,
+      "step": 45360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4961926541619314e-05,
+      "loss": 2.9639,
+      "step": 45365
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4951910428175702e-05,
+      "loss": 2.8587,
+      "step": 45370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4941895656481602e-05,
+      "loss": 2.8932,
+      "step": 45375
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4931882227073556e-05,
+      "loss": 2.9541,
+      "step": 45380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.492187014048789e-05,
+      "loss": 2.905,
+      "step": 45385
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.491185939726095e-05,
+      "loss": 2.9052,
+      "step": 45390
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4901849997928973e-05,
+      "loss": 2.9091,
+      "step": 45395
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4891841943028133e-05,
+      "loss": 2.8335,
+      "step": 45400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4881835233094524e-05,
+      "loss": 2.9639,
+      "step": 45405
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4871829868664194e-05,
+      "loss": 2.9609,
+      "step": 45410
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.486182585027308e-05,
+      "loss": 2.9733,
+      "step": 45415
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4851823178457073e-05,
+      "loss": 2.9861,
+      "step": 45420
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4841821853751996e-05,
+      "loss": 2.859,
+      "step": 45425
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4831821876693574e-05,
+      "loss": 2.9158,
+      "step": 45430
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.48218232478175e-05,
+      "loss": 2.901,
+      "step": 45435
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4811825967659313e-05,
+      "loss": 2.8882,
+      "step": 45440
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4801830036754615e-05,
+      "loss": 2.832,
+      "step": 45445
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.479183545563878e-05,
+      "loss": 2.9193,
+      "step": 45450
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4781842224847263e-05,
+      "loss": 3.0091,
+      "step": 45455
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4771850344915315e-05,
+      "loss": 2.8877,
+      "step": 45460
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4761859816378187e-05,
+      "loss": 2.9684,
+      "step": 45465
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4751870639771036e-05,
+      "loss": 2.989,
+      "step": 45470
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4741882815628965e-05,
+      "loss": 2.9434,
+      "step": 45475
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4731896344486984e-05,
+      "loss": 2.8991,
+      "step": 45480
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.472191122688005e-05,
+      "loss": 3.0393,
+      "step": 45485
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4711927463342994e-05,
+      "loss": 2.8963,
+      "step": 45490
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4701945054410674e-05,
+      "loss": 2.7682,
+      "step": 45495
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4691964000617756e-05,
+      "loss": 2.8205,
+      "step": 45500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.468198430249896e-05,
+      "loss": 2.9242,
+      "step": 45505
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.46720059605888e-05,
+      "loss": 2.9026,
+      "step": 45510
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4662028975421856e-05,
+      "loss": 2.8176,
+      "step": 45515
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.465205334753251e-05,
+      "loss": 2.965,
+      "step": 45520
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4642079077455144e-05,
+      "loss": 2.8884,
+      "step": 45525
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4632106165724043e-05,
+      "loss": 2.9582,
+      "step": 45530
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.462213461287344e-05,
+      "loss": 2.9523,
+      "step": 45535
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4612164419437467e-05,
+      "loss": 2.9248,
+      "step": 45540
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4602195585950223e-05,
+      "loss": 2.9548,
+      "step": 45545
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4592228112945648e-05,
+      "loss": 2.8818,
+      "step": 45550
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4582262000957746e-05,
+      "loss": 2.9626,
+      "step": 45555
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4572297250520294e-05,
+      "loss": 2.8466,
+      "step": 45560
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4562333862167146e-05,
+      "loss": 3.0152,
+      "step": 45565
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4552371836431964e-05,
+      "loss": 2.9932,
+      "step": 45570
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4542411173848394e-05,
+      "loss": 2.9865,
+      "step": 45575
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4532451874949996e-05,
+      "loss": 2.9259,
+      "step": 45580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.452249394027027e-05,
+      "loss": 2.8372,
+      "step": 45585
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.451253737034262e-05,
+      "loss": 2.8849,
+      "step": 45590
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.450258216570041e-05,
+      "loss": 2.9591,
+      "step": 45595
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4492628326876858e-05,
+      "loss": 2.7391,
+      "step": 45600
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.448267585440524e-05,
+      "loss": 2.921,
+      "step": 45605
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4472724748818592e-05,
+      "loss": 2.9462,
+      "step": 45610
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.446277501065005e-05,
+      "loss": 3.0103,
+      "step": 45615
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.445282664043253e-05,
+      "loss": 2.9973,
+      "step": 45620
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4442879638698958e-05,
+      "loss": 2.7972,
+      "step": 45625
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4432934005982154e-05,
+      "loss": 2.921,
+      "step": 45630
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.442298974281489e-05,
+      "loss": 2.9269,
+      "step": 45635
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4413046849729843e-05,
+      "loss": 2.953,
+      "step": 45640
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4403105327259633e-05,
+      "loss": 2.9188,
+      "step": 45645
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.439316517593676e-05,
+      "loss": 2.9156,
+      "step": 45650
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4383226396293744e-05,
+      "loss": 2.9134,
+      "step": 45655
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4373288988862912e-05,
+      "loss": 2.9711,
+      "step": 45660
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.436335295417665e-05,
+      "loss": 2.9763,
+      "step": 45665
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.435341829276715e-05,
+      "loss": 2.9874,
+      "step": 45670
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4343485005166593e-05,
+      "loss": 2.9751,
+      "step": 45675
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.433355309190708e-05,
+      "loss": 2.8617,
+      "step": 45680
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4323622553520624e-05,
+      "loss": 2.9776,
+      "step": 45685
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.431369339053919e-05,
+      "loss": 3.053,
+      "step": 45690
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4303765603494644e-05,
+      "loss": 2.9768,
+      "step": 45695
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4293839192918782e-05,
+      "loss": 2.8868,
+      "step": 45700
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.428391415934334e-05,
+      "loss": 2.9514,
+      "step": 45705
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4273990503299965e-05,
+      "loss": 2.9281,
+      "step": 45710
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4264068225320264e-05,
+      "loss": 2.9034,
+      "step": 45715
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4254147325935694e-05,
+      "loss": 2.8786,
+      "step": 45720
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4244227805677717e-05,
+      "loss": 2.8997,
+      "step": 45725
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4234309665077687e-05,
+      "loss": 3.0886,
+      "step": 45730
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4224392904666888e-05,
+      "loss": 2.9818,
+      "step": 45735
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.421447752497653e-05,
+      "loss": 3.0948,
+      "step": 45740
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4204563526537748e-05,
+      "loss": 2.9179,
+      "step": 45745
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.419465090988161e-05,
+      "loss": 2.805,
+      "step": 45750
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.41847396755391e-05,
+      "loss": 2.9374,
+      "step": 45755
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.417482982404114e-05,
+      "loss": 2.9736,
+      "step": 45760
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4164921355918556e-05,
+      "loss": 2.8954,
+      "step": 45765
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.415501427170215e-05,
+      "loss": 3.0484,
+      "step": 45770
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4145108571922544e-05,
+      "loss": 2.9367,
+      "step": 45775
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4135204257110437e-05,
+      "loss": 3.0441,
+      "step": 45780
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4125301327796296e-05,
+      "loss": 2.8968,
+      "step": 45785
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4115399784510674e-05,
+      "loss": 2.8471,
+      "step": 45790
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4105499627783895e-05,
+      "loss": 2.9381,
+      "step": 45795
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.409560085814631e-05,
+      "loss": 2.8816,
+      "step": 45800
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4085703476128164e-05,
+      "loss": 3.0351,
+      "step": 45805
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4075807482259628e-05,
+      "loss": 2.964,
+      "step": 45810
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.40659128770708e-05,
+      "loss": 2.8203,
+      "step": 45815
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4056019661091722e-05,
+      "loss": 2.9116,
+      "step": 45820
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.404612783485229e-05,
+      "loss": 2.9068,
+      "step": 45825
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4036237398882448e-05,
+      "loss": 2.7929,
+      "step": 45830
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.402634835371193e-05,
+      "loss": 2.9752,
+      "step": 45835
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.401646069987053e-05,
+      "loss": 2.9864,
+      "step": 45840
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.4006574437887845e-05,
+      "loss": 2.9331,
+      "step": 45845
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.3996689568293473e-05,
+      "loss": 2.8054,
+      "step": 45850
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.3986806091616915e-05,
+      "loss": 3.0362,
+      "step": 45855
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.3976924008387598e-05,
+      "loss": 3.1212,
+      "step": 45860
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.3967043319134873e-05,
+      "loss": 2.9131,
+      "step": 45865
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.395716402438804e-05,
+      "loss": 3.0268,
+      "step": 45870
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2.3947286124676245e-05,
+      "loss": 2.9024,
+      "step": 45875
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.393740962052869e-05,
+      "loss": 2.9554,
+      "step": 45880
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3927534512474358e-05,
+      "loss": 2.7938,
+      "step": 45885
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.39176608010423e-05,
+      "loss": 2.8916,
+      "step": 45890
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3907788486761367e-05,
+      "loss": 2.8255,
+      "step": 45895
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3897917570160405e-05,
+      "loss": 2.9237,
+      "step": 45900
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.388804805176817e-05,
+      "loss": 2.9771,
+      "step": 45905
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.387817993211334e-05,
+      "loss": 3.0201,
+      "step": 45910
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.386831321172453e-05,
+      "loss": 2.8774,
+      "step": 45915
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3858447891130276e-05,
+      "loss": 2.8467,
+      "step": 45920
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.384858397085898e-05,
+      "loss": 2.7946,
+      "step": 45925
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.38387214514391e-05,
+      "loss": 2.9928,
+      "step": 45930
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.382886033339887e-05,
+      "loss": 2.9837,
+      "step": 45935
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.381900061726658e-05,
+      "loss": 2.9319,
+      "step": 45940
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.380914230357034e-05,
+      "loss": 2.8824,
+      "step": 45945
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.379928539283825e-05,
+      "loss": 2.8991,
+      "step": 45950
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.378942988559831e-05,
+      "loss": 2.9394,
+      "step": 45955
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3779575782378448e-05,
+      "loss": 2.8971,
+      "step": 45960
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3769723083706523e-05,
+      "loss": 2.8457,
+      "step": 45965
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3759871790110317e-05,
+      "loss": 2.9644,
+      "step": 45970
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.375002190211752e-05,
+      "loss": 3.0087,
+      "step": 45975
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.374017342025579e-05,
+      "loss": 2.9132,
+      "step": 45980
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3730326345052623e-05,
+      "loss": 2.7695,
+      "step": 45985
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3720480677035565e-05,
+      "loss": 2.8997,
+      "step": 45990
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3710636416731975e-05,
+      "loss": 2.9812,
+      "step": 45995
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3700793564669187e-05,
+      "loss": 2.6713,
+      "step": 46000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3690952121374454e-05,
+      "loss": 2.8978,
+      "step": 46005
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3681112087374957e-05,
+      "loss": 2.9232,
+      "step": 46010
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.367127346319779e-05,
+      "loss": 3.0184,
+      "step": 46015
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3661436249369988e-05,
+      "loss": 2.962,
+      "step": 46020
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3651600446418492e-05,
+      "loss": 2.9639,
+      "step": 46025
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3641766054870172e-05,
+      "loss": 2.9431,
+      "step": 46030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3631933075251837e-05,
+      "loss": 2.9671,
+      "step": 46035
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3622101508090204e-05,
+      "loss": 2.8991,
+      "step": 46040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3612271353911946e-05,
+      "loss": 2.8552,
+      "step": 46045
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3602442613243564e-05,
+      "loss": 2.8324,
+      "step": 46050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3592615286611642e-05,
+      "loss": 2.9115,
+      "step": 46055
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.358278937454253e-05,
+      "loss": 2.9579,
+      "step": 46060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3572964877562604e-05,
+      "loss": 2.9847,
+      "step": 46065
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.356314179619813e-05,
+      "loss": 2.864,
+      "step": 46070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3553320130975293e-05,
+      "loss": 2.8735,
+      "step": 46075
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.354349988242021e-05,
+      "loss": 2.8975,
+      "step": 46080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3533681051058935e-05,
+      "loss": 2.9673,
+      "step": 46085
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.352386363741742e-05,
+      "loss": 2.9208,
+      "step": 46090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3514047642021576e-05,
+      "loss": 2.7645,
+      "step": 46095
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.350423306539716e-05,
+      "loss": 2.8844,
+      "step": 46100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3494419908069974e-05,
+      "loss": 3.0154,
+      "step": 46105
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3484608170565626e-05,
+      "loss": 2.9821,
+      "step": 46110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3474797853409756e-05,
+      "loss": 2.7378,
+      "step": 46115
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3464988957127827e-05,
+      "loss": 2.9182,
+      "step": 46120
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3455181482245286e-05,
+      "loss": 2.8571,
+      "step": 46125
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.344537542928749e-05,
+      "loss": 2.784,
+      "step": 46130
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.343557079877972e-05,
+      "loss": 3.0097,
+      "step": 46135
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3425767591247176e-05,
+      "loss": 3.0162,
+      "step": 46140
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.341596580721501e-05,
+      "loss": 2.961,
+      "step": 46145
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3406165447208223e-05,
+      "loss": 2.9042,
+      "step": 46150
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.339636651175185e-05,
+      "loss": 2.8734,
+      "step": 46155
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3386569001370722e-05,
+      "loss": 2.8837,
+      "step": 46160
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3376772916589744e-05,
+      "loss": 2.9773,
+      "step": 46165
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3366978257933575e-05,
+      "loss": 3.0046,
+      "step": 46170
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.335718502592697e-05,
+      "loss": 2.9993,
+      "step": 46175
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.334739322109446e-05,
+      "loss": 2.8366,
+      "step": 46180
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3337602843960583e-05,
+      "loss": 2.8932,
+      "step": 46185
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3327813895049777e-05,
+      "loss": 2.9753,
+      "step": 46190
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.331802637488641e-05,
+      "loss": 3.0064,
+      "step": 46195
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3308240283994766e-05,
+      "loss": 3.0339,
+      "step": 46200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.329845562289908e-05,
+      "loss": 2.931,
+      "step": 46205
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3288672392123423e-05,
+      "loss": 2.9278,
+      "step": 46210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.327889059219194e-05,
+      "loss": 2.8768,
+      "step": 46215
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.326911022362852e-05,
+      "loss": 2.8625,
+      "step": 46220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3259331286957152e-05,
+      "loss": 2.9893,
+      "step": 46225
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3249553782701605e-05,
+      "loss": 3.0139,
+      "step": 46230
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3239777711385657e-05,
+      "loss": 2.9813,
+      "step": 46235
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3230003073532967e-05,
+      "loss": 2.9832,
+      "step": 46240
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.322022986966715e-05,
+      "loss": 2.9904,
+      "step": 46245
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.321045810031171e-05,
+      "loss": 2.9413,
+      "step": 46250
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3200687765990127e-05,
+      "loss": 2.881,
+      "step": 46255
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3190918867225692e-05,
+      "loss": 2.8896,
+      "step": 46260
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3181151404541784e-05,
+      "loss": 2.963,
+      "step": 46265
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3171385378461536e-05,
+      "loss": 2.8419,
+      "step": 46270
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3161620789508158e-05,
+      "loss": 2.8763,
+      "step": 46275
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3151857638204655e-05,
+      "loss": 2.9904,
+      "step": 46280
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3142095925074033e-05,
+      "loss": 3.0757,
+      "step": 46285
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3132335650639188e-05,
+      "loss": 2.9503,
+      "step": 46290
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3122576815422954e-05,
+      "loss": 2.9679,
+      "step": 46295
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.311281941994808e-05,
+      "loss": 3.0203,
+      "step": 46300
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3103063464737242e-05,
+      "loss": 3.0116,
+      "step": 46305
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3093308950313026e-05,
+      "loss": 2.91,
+      "step": 46310
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3083555877197966e-05,
+      "loss": 2.9319,
+      "step": 46315
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.30738042459145e-05,
+      "loss": 2.9205,
+      "step": 46320
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3064054056985008e-05,
+      "loss": 3.0209,
+      "step": 46325
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3054305310931745e-05,
+      "loss": 3.0183,
+      "step": 46330
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3044558008276933e-05,
+      "loss": 3.0113,
+      "step": 46335
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3034812149542713e-05,
+      "loss": 2.977,
+      "step": 46340
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3025067735251138e-05,
+      "loss": 2.8432,
+      "step": 46345
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3015324765924185e-05,
+      "loss": 2.8461,
+      "step": 46350
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.3005583242083757e-05,
+      "loss": 3.1529,
+      "step": 46355
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2995843164251674e-05,
+      "loss": 2.9105,
+      "step": 46360
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2986104532949692e-05,
+      "loss": 2.8807,
+      "step": 46365
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.297636734869947e-05,
+      "loss": 2.8854,
+      "step": 46370
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.29666316120226e-05,
+      "loss": 2.8936,
+      "step": 46375
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2956897323440614e-05,
+      "loss": 2.9486,
+      "step": 46380
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2947164483474898e-05,
+      "loss": 2.899,
+      "step": 46385
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.293743309264688e-05,
+      "loss": 3.0396,
+      "step": 46390
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.292770315147777e-05,
+      "loss": 2.9123,
+      "step": 46395
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2917974660488844e-05,
+      "loss": 2.9658,
+      "step": 46400
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2908247620201167e-05,
+      "loss": 2.9134,
+      "step": 46405
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2898522031135818e-05,
+      "loss": 2.8595,
+      "step": 46410
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2888797893813745e-05,
+      "loss": 3.018,
+      "step": 46415
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.287907520875586e-05,
+      "loss": 2.8246,
+      "step": 46420
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2869353976482976e-05,
+      "loss": 2.9896,
+      "step": 46425
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2859634197515846e-05,
+      "loss": 2.8394,
+      "step": 46430
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.284991587237506e-05,
+      "loss": 2.9889,
+      "step": 46435
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.284019900158129e-05,
+      "loss": 2.9897,
+      "step": 46440
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.283048358565496e-05,
+      "loss": 2.8407,
+      "step": 46445
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2820769625116568e-05,
+      "loss": 2.848,
+      "step": 46450
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.28110571204864e-05,
+      "loss": 2.8259,
+      "step": 46455
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2801346072284755e-05,
+      "loss": 3.0364,
+      "step": 46460
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.279163648103182e-05,
+      "loss": 2.9178,
+      "step": 46465
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.27819283472477e-05,
+      "loss": 2.8991,
+      "step": 46470
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2772221671452442e-05,
+      "loss": 2.9618,
+      "step": 46475
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2762516454166017e-05,
+      "loss": 2.8959,
+      "step": 46480
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2752812695908242e-05,
+      "loss": 3.0032,
+      "step": 46485
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2743110397199004e-05,
+      "loss": 2.8613,
+      "step": 46490
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.273340955855794e-05,
+      "loss": 2.8591,
+      "step": 46495
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.272371018050477e-05,
+      "loss": 2.9981,
+      "step": 46500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.271401226355901e-05,
+      "loss": 2.9136,
+      "step": 46505
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2704315808240166e-05,
+      "loss": 2.958,
+      "step": 46510
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.269462081506764e-05,
+      "loss": 2.8476,
+      "step": 46515
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2684927284560774e-05,
+      "loss": 2.8889,
+      "step": 46520
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.267523521723881e-05,
+      "loss": 2.7934,
+      "step": 46525
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.266554461362095e-05,
+      "loss": 3.0127,
+      "step": 46530
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.265585547422623e-05,
+      "loss": 2.9611,
+      "step": 46535
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2646167799573742e-05,
+      "loss": 3.0045,
+      "step": 46540
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2636481590182346e-05,
+      "loss": 2.8447,
+      "step": 46545
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2626796846570986e-05,
+      "loss": 2.9475,
+      "step": 46550
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2.2617113569258385e-05,
+      "loss": 2.8956,
+      "step": 46555
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.260743175876326e-05,
+      "loss": 2.9473,
+      "step": 46560
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2597751415604246e-05,
+      "loss": 3.0981,
+      "step": 46565
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2588072540299876e-05,
+      "loss": 2.8815,
+      "step": 46570
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.257839513336863e-05,
+      "loss": 3.0288,
+      "step": 46575
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2568719195328913e-05,
+      "loss": 2.8963,
+      "step": 46580
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2559044726698973e-05,
+      "loss": 2.99,
+      "step": 46585
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2549371727997126e-05,
+      "loss": 2.8588,
+      "step": 46590
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2539700199741442e-05,
+      "loss": 2.9524,
+      "step": 46595
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2530030142450076e-05,
+      "loss": 2.9621,
+      "step": 46600
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.252036155664096e-05,
+      "loss": 2.8875,
+      "step": 46605
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2510694442832038e-05,
+      "loss": 3.0716,
+      "step": 46610
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2501028801541142e-05,
+      "loss": 2.8259,
+      "step": 46615
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2491364633286027e-05,
+      "loss": 3.0928,
+      "step": 46620
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2481701938584386e-05,
+      "loss": 2.8966,
+      "step": 46625
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2472040717953807e-05,
+      "loss": 3.1257,
+      "step": 46630
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2462380971911823e-05,
+      "loss": 2.8691,
+      "step": 46635
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2452722700975864e-05,
+      "loss": 2.9745,
+      "step": 46640
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2443065905663307e-05,
+      "loss": 3.0149,
+      "step": 46645
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.243341058649143e-05,
+      "loss": 2.9935,
+      "step": 46650
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2423756743977454e-05,
+      "loss": 2.9152,
+      "step": 46655
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.241410437863848e-05,
+      "loss": 2.8612,
+      "step": 46660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2404453490991562e-05,
+      "loss": 2.9532,
+      "step": 46665
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2394804081553676e-05,
+      "loss": 2.8764,
+      "step": 46670
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2385156150841714e-05,
+      "loss": 2.8787,
+      "step": 46675
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2375509699372476e-05,
+      "loss": 2.8717,
+      "step": 46680
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2365864727662705e-05,
+      "loss": 3.0409,
+      "step": 46685
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.235622123622904e-05,
+      "loss": 2.9597,
+      "step": 46690
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2346579225588065e-05,
+      "loss": 2.9329,
+      "step": 46695
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2336938696256265e-05,
+      "loss": 2.8892,
+      "step": 46700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2327299648750076e-05,
+      "loss": 3.0573,
+      "step": 46705
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.231766208358577e-05,
+      "loss": 2.8825,
+      "step": 46710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2308026001279692e-05,
+      "loss": 2.9818,
+      "step": 46715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2298391402347923e-05,
+      "loss": 2.8979,
+      "step": 46720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2288758287306648e-05,
+      "loss": 3.021,
+      "step": 46725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2279126656671822e-05,
+      "loss": 2.9834,
+      "step": 46730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2269496510959405e-05,
+      "loss": 2.878,
+      "step": 46735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2259867850685245e-05,
+      "loss": 2.878,
+      "step": 46740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.225024067636513e-05,
+      "loss": 2.8632,
+      "step": 46745
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.224061498851475e-05,
+      "loss": 2.972,
+      "step": 46750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.223099078764975e-05,
+      "loss": 3.0478,
+      "step": 46755
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.222136807428561e-05,
+      "loss": 2.942,
+      "step": 46760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2211746848937865e-05,
+      "loss": 2.8382,
+      "step": 46765
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.220212711212181e-05,
+      "loss": 2.8812,
+      "step": 46770
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2192508864352833e-05,
+      "loss": 2.9965,
+      "step": 46775
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.21828921061461e-05,
+      "loss": 2.9662,
+      "step": 46780
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2173276838016753e-05,
+      "loss": 2.9346,
+      "step": 46785
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2163663060479868e-05,
+      "loss": 2.976,
+      "step": 46790
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2154050774050415e-05,
+      "loss": 2.9476,
+      "step": 46795
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.214443997924331e-05,
+      "loss": 2.9106,
+      "step": 46800
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.213483067657337e-05,
+      "loss": 2.8094,
+      "step": 46805
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.21252228665553e-05,
+      "loss": 3.0247,
+      "step": 46810
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2115616549703826e-05,
+      "loss": 2.8257,
+      "step": 46815
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.210601172653346e-05,
+      "loss": 3.0218,
+      "step": 46820
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2096408397558772e-05,
+      "loss": 2.9475,
+      "step": 46825
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.208680656329411e-05,
+      "loss": 2.8813,
+      "step": 46830
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.207720622425389e-05,
+      "loss": 2.8866,
+      "step": 46835
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2067607380952317e-05,
+      "loss": 2.7846,
+      "step": 46840
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2058010033903587e-05,
+      "loss": 2.8669,
+      "step": 46845
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2048414183621807e-05,
+      "loss": 2.9673,
+      "step": 46850
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2038819830620995e-05,
+      "loss": 2.9263,
+      "step": 46855
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2029226975415092e-05,
+      "loss": 2.9348,
+      "step": 46860
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.201963561851797e-05,
+      "loss": 2.8672,
+      "step": 46865
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.201004576044336e-05,
+      "loss": 2.8604,
+      "step": 46870
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.2000457401705033e-05,
+      "loss": 2.8157,
+      "step": 46875
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1990870542816532e-05,
+      "loss": 2.8633,
+      "step": 46880
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1981285184291474e-05,
+      "loss": 2.8942,
+      "step": 46885
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.197170132664326e-05,
+      "loss": 3.0384,
+      "step": 46890
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.196211897038529e-05,
+      "loss": 2.9456,
+      "step": 46895
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1952538116030853e-05,
+      "loss": 2.9533,
+      "step": 46900
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.194295876409317e-05,
+      "loss": 2.8806,
+      "step": 46905
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1933380915085378e-05,
+      "loss": 2.9391,
+      "step": 46910
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1923804569520535e-05,
+      "loss": 2.9194,
+      "step": 46915
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1914229727911618e-05,
+      "loss": 2.9265,
+      "step": 46920
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1904656390771532e-05,
+      "loss": 2.9502,
+      "step": 46925
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1895084558613048e-05,
+      "loss": 2.8311,
+      "step": 46930
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1885514231948962e-05,
+      "loss": 2.8976,
+      "step": 46935
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1875945411291882e-05,
+      "loss": 2.7422,
+      "step": 46940
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.186637809715439e-05,
+      "loss": 2.7723,
+      "step": 46945
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1856812290048977e-05,
+      "loss": 3.0057,
+      "step": 46950
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.184724799048807e-05,
+      "loss": 3.0701,
+      "step": 46955
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1837685198983975e-05,
+      "loss": 2.9757,
+      "step": 46960
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.182812391604896e-05,
+      "loss": 2.8627,
+      "step": 46965
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.181856414219519e-05,
+      "loss": 2.9713,
+      "step": 46970
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1809005877934753e-05,
+      "loss": 2.9412,
+      "step": 46975
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1799449123779646e-05,
+      "loss": 2.9891,
+      "step": 46980
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1789893880241806e-05,
+      "loss": 2.9096,
+      "step": 46985
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1780340147833096e-05,
+      "loss": 2.8479,
+      "step": 46990
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1770787927065224e-05,
+      "loss": 2.8825,
+      "step": 46995
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.176123721844994e-05,
+      "loss": 2.9945,
+      "step": 47000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1751688022498802e-05,
+      "loss": 2.8956,
+      "step": 47005
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.174214033972334e-05,
+      "loss": 2.9239,
+      "step": 47010
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1732594170634997e-05,
+      "loss": 2.9253,
+      "step": 47015
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1723049515745137e-05,
+      "loss": 2.8896,
+      "step": 47020
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1713506375565028e-05,
+      "loss": 2.9412,
+      "step": 47025
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1703964750605882e-05,
+      "loss": 2.9021,
+      "step": 47030
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1694424641378802e-05,
+      "loss": 2.9894,
+      "step": 47035
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1684886048394843e-05,
+      "loss": 2.9087,
+      "step": 47040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1675348972164905e-05,
+      "loss": 3.0123,
+      "step": 47045
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.166581341319994e-05,
+      "loss": 2.9023,
+      "step": 47050
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.165627937201065e-05,
+      "loss": 3.0434,
+      "step": 47055
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.164674684910783e-05,
+      "loss": 3.0459,
+      "step": 47060
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1637215845002045e-05,
+      "loss": 2.9235,
+      "step": 47065
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.162768636020387e-05,
+      "loss": 3.0074,
+      "step": 47070
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1618158395223765e-05,
+      "loss": 2.9203,
+      "step": 47075
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.160863195057211e-05,
+      "loss": 2.9588,
+      "step": 47080
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1599107026759215e-05,
+      "loss": 2.9285,
+      "step": 47085
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.158958362429531e-05,
+      "loss": 2.9955,
+      "step": 47090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1580061743690493e-05,
+      "loss": 2.8754,
+      "step": 47095
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1570541385454884e-05,
+      "loss": 2.9736,
+      "step": 47100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1561022550098388e-05,
+      "loss": 3.0172,
+      "step": 47105
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1551505238130976e-05,
+      "loss": 2.8174,
+      "step": 47110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1541989450062406e-05,
+      "loss": 3.006,
+      "step": 47115
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.153247518640243e-05,
+      "loss": 2.9453,
+      "step": 47120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1522962447660687e-05,
+      "loss": 2.86,
+      "step": 47125
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1513451234346758e-05,
+      "loss": 2.9426,
+      "step": 47130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.150394154697013e-05,
+      "loss": 2.9716,
+      "step": 47135
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1494433386040218e-05,
+      "loss": 2.8914,
+      "step": 47140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1484926752066293e-05,
+      "loss": 2.8302,
+      "step": 47145
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.147542164555767e-05,
+      "loss": 2.8771,
+      "step": 47150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1465918067023433e-05,
+      "loss": 2.941,
+      "step": 47155
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1456416016972742e-05,
+      "loss": 2.9464,
+      "step": 47160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.144691549591453e-05,
+      "loss": 2.9528,
+      "step": 47165
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.143741650435772e-05,
+      "loss": 2.9821,
+      "step": 47170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1427919042811163e-05,
+      "loss": 2.8434,
+      "step": 47175
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1418423111783596e-05,
+      "loss": 2.8866,
+      "step": 47180
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1408928711783694e-05,
+      "loss": 3.0008,
+      "step": 47185
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.139943584332006e-05,
+      "loss": 2.9226,
+      "step": 47190
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1389944506901138e-05,
+      "loss": 2.9267,
+      "step": 47195
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.138045470303543e-05,
+      "loss": 2.9623,
+      "step": 47200
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1370966432231192e-05,
+      "loss": 2.899,
+      "step": 47205
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1361479694996765e-05,
+      "loss": 2.7968,
+      "step": 47210
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.135199449184027e-05,
+      "loss": 2.9611,
+      "step": 47215
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1342510823269812e-05,
+      "loss": 2.963,
+      "step": 47220
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.133302868979341e-05,
+      "loss": 2.9091,
+      "step": 47225
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.1323548091918983e-05,
+      "loss": 3.0256,
+      "step": 47230
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2.131406903015438e-05,
+      "loss": 3.0406,
+      "step": 47235
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1304591505007377e-05,
+      "loss": 2.9563,
+      "step": 47240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.129511551698564e-05,
+      "loss": 3.007,
+      "step": 47245
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.128564106659678e-05,
+      "loss": 2.9707,
+      "step": 47250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.127616815434831e-05,
+      "loss": 2.9959,
+      "step": 47255
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1266696780747662e-05,
+      "loss": 2.9043,
+      "step": 47260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.125722694630221e-05,
+      "loss": 2.9955,
+      "step": 47265
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1247758651519188e-05,
+      "loss": 2.909,
+      "step": 47270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1238291896905805e-05,
+      "loss": 3.0095,
+      "step": 47275
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.122882668296916e-05,
+      "loss": 3.075,
+      "step": 47280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1219363010216275e-05,
+      "loss": 2.928,
+      "step": 47285
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1209900879154092e-05,
+      "loss": 2.8265,
+      "step": 47290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1200440290289474e-05,
+      "loss": 2.9445,
+      "step": 47295
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1190981244129192e-05,
+      "loss": 2.9223,
+      "step": 47300
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1181523741179938e-05,
+      "loss": 2.8158,
+      "step": 47305
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1172067781948318e-05,
+      "loss": 2.9792,
+      "step": 47310
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1162613366940875e-05,
+      "loss": 2.8381,
+      "step": 47315
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.115316049666401e-05,
+      "loss": 3.0021,
+      "step": 47320
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.114370917162415e-05,
+      "loss": 2.9079,
+      "step": 47325
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.11342593923275e-05,
+      "loss": 2.8838,
+      "step": 47330
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.112481115928033e-05,
+      "loss": 2.9752,
+      "step": 47335
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1115364472988696e-05,
+      "loss": 2.8889,
+      "step": 47340
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1105919333958653e-05,
+      "loss": 2.7792,
+      "step": 47345
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1096475742696142e-05,
+      "loss": 2.7864,
+      "step": 47350
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1087033699707026e-05,
+      "loss": 2.9106,
+      "step": 47355
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.107759320549709e-05,
+      "loss": 2.9291,
+      "step": 47360
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1068154260572055e-05,
+      "loss": 2.9276,
+      "step": 47365
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1058716865437478e-05,
+      "loss": 2.9906,
+      "step": 47370
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1049281020598953e-05,
+      "loss": 3.0658,
+      "step": 47375
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1039846726561874e-05,
+      "loss": 2.8741,
+      "step": 47380
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1030413983831672e-05,
+      "loss": 3.0647,
+      "step": 47385
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1020982792913574e-05,
+      "loss": 2.9772,
+      "step": 47390
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1011553154312803e-05,
+      "loss": 2.8974,
+      "step": 47395
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.1002125068534474e-05,
+      "loss": 3.0588,
+      "step": 47400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0992698536083616e-05,
+      "loss": 2.8152,
+      "step": 47405
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.098327355746519e-05,
+      "loss": 2.8831,
+      "step": 47410
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0973850133184068e-05,
+      "loss": 3.0322,
+      "step": 47415
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.096442826374499e-05,
+      "loss": 2.8851,
+      "step": 47420
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.095500794965272e-05,
+      "loss": 2.8875,
+      "step": 47425
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.094558919141182e-05,
+      "loss": 2.7853,
+      "step": 47430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0936171989526876e-05,
+      "loss": 2.7491,
+      "step": 47435
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0926756344502303e-05,
+      "loss": 2.8305,
+      "step": 47440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.091734225684247e-05,
+      "loss": 2.8684,
+      "step": 47445
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0907929727051668e-05,
+      "loss": 2.8402,
+      "step": 47450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.08985187556341e-05,
+      "loss": 2.9077,
+      "step": 47455
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0889109343093878e-05,
+      "loss": 2.9449,
+      "step": 47460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0879701489935057e-05,
+      "loss": 2.9722,
+      "step": 47465
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0870295196661533e-05,
+      "loss": 2.9049,
+      "step": 47470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.086089046377724e-05,
+      "loss": 2.8603,
+      "step": 47475
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0851487291785892e-05,
+      "loss": 2.9276,
+      "step": 47480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0842085681191266e-05,
+      "loss": 2.9309,
+      "step": 47485
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0832685632496897e-05,
+      "loss": 2.9183,
+      "step": 47490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.082328714620639e-05,
+      "loss": 2.806,
+      "step": 47495
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.081389022282314e-05,
+      "loss": 2.9533,
+      "step": 47500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0804494862850533e-05,
+      "loss": 2.8876,
+      "step": 47505
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0795101066791844e-05,
+      "loss": 2.8842,
+      "step": 47510
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.078570883515027e-05,
+      "loss": 2.8962,
+      "step": 47515
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0776318168428933e-05,
+      "loss": 2.7721,
+      "step": 47520
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0766929067130853e-05,
+      "loss": 2.9992,
+      "step": 47525
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0757541531758978e-05,
+      "loss": 2.9158,
+      "step": 47530
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.074815556281619e-05,
+      "loss": 2.8435,
+      "step": 47535
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.073877116080521e-05,
+      "loss": 2.8504,
+      "step": 47540
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0729388326228806e-05,
+      "loss": 3.0569,
+      "step": 47545
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.072000705958953e-05,
+      "loss": 2.9219,
+      "step": 47550
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.071062736138993e-05,
+      "loss": 2.9127,
+      "step": 47555
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.070124923213244e-05,
+      "loss": 2.9478,
+      "step": 47560
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.069187267231943e-05,
+      "loss": 2.9825,
+      "step": 47565
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0682497682453167e-05,
+      "loss": 2.9288,
+      "step": 47570
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0673124263035835e-05,
+      "loss": 2.9994,
+      "step": 47575
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0663752414569555e-05,
+      "loss": 2.9292,
+      "step": 47580
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.065438213755633e-05,
+      "loss": 3.0654,
+      "step": 47585
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0645013432498105e-05,
+      "loss": 2.9681,
+      "step": 47590
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0635646299896737e-05,
+      "loss": 3.0584,
+      "step": 47595
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.062628074025401e-05,
+      "loss": 2.8583,
+      "step": 47600
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0616916754071552e-05,
+      "loss": 2.8572,
+      "step": 47605
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0607554341851032e-05,
+      "loss": 2.8661,
+      "step": 47610
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.059819350409392e-05,
+      "loss": 2.9398,
+      "step": 47615
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0588834241301663e-05,
+      "loss": 3.0179,
+      "step": 47620
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0579476553975607e-05,
+      "loss": 2.9772,
+      "step": 47625
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0570120442617012e-05,
+      "loss": 2.8692,
+      "step": 47630
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0560765907727058e-05,
+      "loss": 2.9361,
+      "step": 47635
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0551412949806843e-05,
+      "loss": 2.9832,
+      "step": 47640
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0542061569357367e-05,
+      "loss": 2.8816,
+      "step": 47645
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0532711766879586e-05,
+      "loss": 2.8455,
+      "step": 47650
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0523363542874274e-05,
+      "loss": 2.8714,
+      "step": 47655
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0514016897842264e-05,
+      "loss": 3.0016,
+      "step": 47660
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0504671832284157e-05,
+      "loss": 3.0956,
+      "step": 47665
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.049532834670061e-05,
+      "loss": 3.1058,
+      "step": 47670
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.048598644159207e-05,
+      "loss": 2.9698,
+      "step": 47675
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.047664611745897e-05,
+      "loss": 2.8694,
+      "step": 47680
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0467307374801646e-05,
+      "loss": 2.8548,
+      "step": 47685
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.045797021412035e-05,
+      "loss": 2.9067,
+      "step": 47690
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0448634635915236e-05,
+      "loss": 2.8785,
+      "step": 47695
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.043930064068641e-05,
+      "loss": 3.0029,
+      "step": 47700
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.04299682289338e-05,
+      "loss": 2.8577,
+      "step": 47705
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0420637401157396e-05,
+      "loss": 2.9807,
+      "step": 47710
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0411308157856947e-05,
+      "loss": 2.9173,
+      "step": 47715
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0401980499532264e-05,
+      "loss": 2.8574,
+      "step": 47720
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0392654426682943e-05,
+      "loss": 2.9493,
+      "step": 47725
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0383329939808582e-05,
+      "loss": 3.0199,
+      "step": 47730
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.037400703940866e-05,
+      "loss": 2.8355,
+      "step": 47735
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0364685725982568e-05,
+      "loss": 2.8476,
+      "step": 47740
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.035536600002963e-05,
+      "loss": 2.9746,
+      "step": 47745
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.034604786204909e-05,
+      "loss": 2.9784,
+      "step": 47750
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0336731312540042e-05,
+      "loss": 2.9246,
+      "step": 47755
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.032741635200161e-05,
+      "loss": 2.9504,
+      "step": 47760
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.03181029809327e-05,
+      "loss": 2.8059,
+      "step": 47765
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0308791199832274e-05,
+      "loss": 3.0051,
+      "step": 47770
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.029948100919908e-05,
+      "loss": 2.9487,
+      "step": 47775
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.029017240953185e-05,
+      "loss": 3.0035,
+      "step": 47780
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0280865401329223e-05,
+      "loss": 3.0111,
+      "step": 47785
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0271559985089748e-05,
+      "loss": 2.9022,
+      "step": 47790
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.026225616131189e-05,
+      "loss": 2.9898,
+      "step": 47795
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0252953930494033e-05,
+      "loss": 2.8859,
+      "step": 47800
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0243653293134422e-05,
+      "loss": 2.9084,
+      "step": 47805
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0234354249731335e-05,
+      "loss": 2.8953,
+      "step": 47810
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0225056800782828e-05,
+      "loss": 2.8558,
+      "step": 47815
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0215760946786992e-05,
+      "loss": 2.8478,
+      "step": 47820
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.020646668824174e-05,
+      "loss": 3.0759,
+      "step": 47825
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0197174025644955e-05,
+      "loss": 2.9038,
+      "step": 47830
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0187882959494404e-05,
+      "loss": 2.9938,
+      "step": 47835
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.017859349028779e-05,
+      "loss": 3.116,
+      "step": 47840
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0169305618522717e-05,
+      "loss": 2.8271,
+      "step": 47845
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0160019344696713e-05,
+      "loss": 2.9015,
+      "step": 47850
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.015073466930721e-05,
+      "loss": 2.9565,
+      "step": 47855
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0141451592851563e-05,
+      "loss": 2.8566,
+      "step": 47860
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.013217011582703e-05,
+      "loss": 2.8379,
+      "step": 47865
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0122890238730825e-05,
+      "loss": 2.975,
+      "step": 47870
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0113611962059998e-05,
+      "loss": 2.9739,
+      "step": 47875
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0104335286311572e-05,
+      "loss": 2.99,
+      "step": 47880
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0095060211982474e-05,
+      "loss": 2.9876,
+      "step": 47885
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0085786739569547e-05,
+      "loss": 2.9359,
+      "step": 47890
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.007651486956953e-05,
+      "loss": 2.8245,
+      "step": 47895
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0067244602479096e-05,
+      "loss": 3.0202,
+      "step": 47900
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0057975938794827e-05,
+      "loss": 3.0155,
+      "step": 47905
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0048708879013217e-05,
+      "loss": 2.8095,
+      "step": 47910
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2.0039443423630665e-05,
+      "loss": 2.7826,
+      "step": 47915
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0030179573143505e-05,
+      "loss": 2.8121,
+      "step": 47920
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0020917328047983e-05,
+      "loss": 2.8859,
+      "step": 47925
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0011656688840196e-05,
+      "loss": 2.9373,
+      "step": 47930
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2.0002397656016286e-05,
+      "loss": 2.911,
+      "step": 47935
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9993140230072156e-05,
+      "loss": 2.8903,
+      "step": 47940
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9983884411503767e-05,
+      "loss": 2.8908,
+      "step": 47945
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9974630200806872e-05,
+      "loss": 2.976,
+      "step": 47950
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9965377598477215e-05,
+      "loss": 2.7045,
+      "step": 47955
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9956126605010418e-05,
+      "loss": 2.9997,
+      "step": 47960
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.994687722090204e-05,
+      "loss": 2.9859,
+      "step": 47965
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9937629446647534e-05,
+      "loss": 2.9882,
+      "step": 47970
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9928383282742292e-05,
+      "loss": 2.8185,
+      "step": 47975
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9919138729681558e-05,
+      "loss": 2.9535,
+      "step": 47980
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.99098957879606e-05,
+      "loss": 2.8936,
+      "step": 47985
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9900654458074464e-05,
+      "loss": 2.8405,
+      "step": 47990
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9891414740518248e-05,
+      "loss": 2.9453,
+      "step": 47995
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.988217663578685e-05,
+      "loss": 3.0427,
+      "step": 48000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.987294014437513e-05,
+      "loss": 2.9186,
+      "step": 48005
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9863705266777876e-05,
+      "loss": 2.7554,
+      "step": 48010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.985447200348976e-05,
+      "loss": 2.8271,
+      "step": 48015
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.984524035500539e-05,
+      "loss": 3.0575,
+      "step": 48020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.983601032181928e-05,
+      "loss": 2.8348,
+      "step": 48025
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.982678190442582e-05,
+      "loss": 2.863,
+      "step": 48030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.98175551033194e-05,
+      "loss": 2.9257,
+      "step": 48035
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9808329918994218e-05,
+      "loss": 2.7255,
+      "step": 48040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9799106351944495e-05,
+      "loss": 2.8523,
+      "step": 48045
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9789884402664272e-05,
+      "loss": 3.0285,
+      "step": 48050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9780664071647542e-05,
+      "loss": 3.017,
+      "step": 48055
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9771445359388217e-05,
+      "loss": 2.8512,
+      "step": 48060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9762228266380117e-05,
+      "loss": 2.9245,
+      "step": 48065
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.975301279311697e-05,
+      "loss": 2.7864,
+      "step": 48070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.974379894009244e-05,
+      "loss": 2.8985,
+      "step": 48075
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9734586707800034e-05,
+      "loss": 2.8758,
+      "step": 48080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9725376096733294e-05,
+      "loss": 2.8382,
+      "step": 48085
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9716167107385525e-05,
+      "loss": 2.8959,
+      "step": 48090
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.970695974025011e-05,
+      "loss": 2.823,
+      "step": 48095
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.969775399582019e-05,
+      "loss": 2.9621,
+      "step": 48100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9688549874588918e-05,
+      "loss": 2.824,
+      "step": 48105
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.967934737704933e-05,
+      "loss": 2.9016,
+      "step": 48110
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9670146503694365e-05,
+      "loss": 2.879,
+      "step": 48115
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.96609472550169e-05,
+      "loss": 2.9779,
+      "step": 48120
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9651749631509697e-05,
+      "loss": 2.9642,
+      "step": 48125
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9642553633665455e-05,
+      "loss": 3.0461,
+      "step": 48130
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9633359261976793e-05,
+      "loss": 3.1288,
+      "step": 48135
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9624166516936166e-05,
+      "loss": 2.8898,
+      "step": 48140
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9614975399036072e-05,
+      "loss": 2.9305,
+      "step": 48145
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9605785908768805e-05,
+      "loss": 2.9663,
+      "step": 48150
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.959659804662663e-05,
+      "loss": 3.0805,
+      "step": 48155
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9587411813101714e-05,
+      "loss": 3.04,
+      "step": 48160
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9578227208686135e-05,
+      "loss": 2.9609,
+      "step": 48165
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9569044233871887e-05,
+      "loss": 2.8563,
+      "step": 48170
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.955986288915087e-05,
+      "loss": 2.7489,
+      "step": 48175
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.95506831750149e-05,
+      "loss": 2.9712,
+      "step": 48180
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9541505091955716e-05,
+      "loss": 2.9636,
+      "step": 48185
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.953232864046494e-05,
+      "loss": 2.8856,
+      "step": 48190
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9523153821034146e-05,
+      "loss": 2.8635,
+      "step": 48195
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9513980634154792e-05,
+      "loss": 2.8791,
+      "step": 48200
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9504809080318264e-05,
+      "loss": 2.9496,
+      "step": 48205
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.949563916001586e-05,
+      "loss": 2.7587,
+      "step": 48210
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.948647087373876e-05,
+      "loss": 2.9018,
+      "step": 48215
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9477304221978094e-05,
+      "loss": 2.8504,
+      "step": 48220
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9468139205224893e-05,
+      "loss": 2.807,
+      "step": 48225
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.94589758239701e-05,
+      "loss": 2.896,
+      "step": 48230
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9449814078704563e-05,
+      "loss": 2.9337,
+      "step": 48235
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.944065396991906e-05,
+      "loss": 2.7936,
+      "step": 48240
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.943149549810426e-05,
+      "loss": 2.8866,
+      "step": 48245
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9422338663750755e-05,
+      "loss": 2.9921,
+      "step": 48250
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.941318346734905e-05,
+      "loss": 2.8704,
+      "step": 48255
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9404029909389587e-05,
+      "loss": 3.0133,
+      "step": 48260
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.939487799036263e-05,
+      "loss": 2.9259,
+      "step": 48265
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9385727710758496e-05,
+      "loss": 2.8922,
+      "step": 48270
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9376579071067256e-05,
+      "loss": 2.9055,
+      "step": 48275
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9367432071779064e-05,
+      "loss": 3.0077,
+      "step": 48280
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9358286713383832e-05,
+      "loss": 2.9172,
+      "step": 48285
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9349142996371465e-05,
+      "loss": 2.9758,
+      "step": 48290
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.934000092123177e-05,
+      "loss": 3.0686,
+      "step": 48295
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9330860488454457e-05,
+      "loss": 2.8091,
+      "step": 48300
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9321721698529148e-05,
+      "loss": 2.9492,
+      "step": 48305
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9312584551945406e-05,
+      "loss": 2.9238,
+      "step": 48310
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9303449049192624e-05,
+      "loss": 3.0078,
+      "step": 48315
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9294315190760226e-05,
+      "loss": 2.983,
+      "step": 48320
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9285182977137422e-05,
+      "loss": 2.8432,
+      "step": 48325
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9276052408813468e-05,
+      "loss": 3.0846,
+      "step": 48330
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.92669234862774e-05,
+      "loss": 3.0138,
+      "step": 48335
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9257796210018248e-05,
+      "loss": 2.9098,
+      "step": 48340
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9248670580524942e-05,
+      "loss": 2.8696,
+      "step": 48345
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9239546598286296e-05,
+      "loss": 2.8944,
+      "step": 48350
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9230424263791068e-05,
+      "loss": 3.0069,
+      "step": 48355
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9221303577527928e-05,
+      "loss": 2.7918,
+      "step": 48360
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9212184539985383e-05,
+      "loss": 3.0623,
+      "step": 48365
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9203067151651992e-05,
+      "loss": 2.951,
+      "step": 48370
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.919395141301607e-05,
+      "loss": 2.9668,
+      "step": 48375
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.918483732456599e-05,
+      "loss": 3.0674,
+      "step": 48380
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9175724886789918e-05,
+      "loss": 2.9086,
+      "step": 48385
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.916661410017599e-05,
+      "loss": 2.8281,
+      "step": 48390
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9157504965212246e-05,
+      "loss": 2.8652,
+      "step": 48395
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.914839748238664e-05,
+      "loss": 2.868,
+      "step": 48400
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9139291652187018e-05,
+      "loss": 2.9823,
+      "step": 48405
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9130187475101186e-05,
+      "loss": 2.9482,
+      "step": 48410
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9121084951616758e-05,
+      "loss": 3.027,
+      "step": 48415
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9111984082221413e-05,
+      "loss": 2.8042,
+      "step": 48420
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9102884867402578e-05,
+      "loss": 3.0567,
+      "step": 48425
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.909378730764774e-05,
+      "loss": 3.0583,
+      "step": 48430
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9084691403444177e-05,
+      "loss": 2.8779,
+      "step": 48435
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.907559715527915e-05,
+      "loss": 2.9757,
+      "step": 48440
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9066504563639805e-05,
+      "loss": 2.8203,
+      "step": 48445
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9057413629013203e-05,
+      "loss": 3.0438,
+      "step": 48450
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.904832435188632e-05,
+      "loss": 2.9956,
+      "step": 48455
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9039236732746045e-05,
+      "loss": 2.9207,
+      "step": 48460
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.903015077207917e-05,
+      "loss": 2.8756,
+      "step": 48465
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9021066470372402e-05,
+      "loss": 2.8115,
+      "step": 48470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9011983828112356e-05,
+      "loss": 2.8564,
+      "step": 48475
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9002902845785586e-05,
+      "loss": 2.9875,
+      "step": 48480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8993823523878494e-05,
+      "loss": 2.8004,
+      "step": 48485
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8984745862877445e-05,
+      "loss": 3.1233,
+      "step": 48490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.897566986326871e-05,
+      "loss": 2.9696,
+      "step": 48495
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8966595525538457e-05,
+      "loss": 2.9247,
+      "step": 48500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8957522850172773e-05,
+      "loss": 2.9989,
+      "step": 48505
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8948451837657644e-05,
+      "loss": 2.8838,
+      "step": 48510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.893938248847899e-05,
+      "loss": 2.8698,
+      "step": 48515
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8930314803122623e-05,
+      "loss": 2.9673,
+      "step": 48520
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8921248782074267e-05,
+      "loss": 3.0656,
+      "step": 48525
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8912184425819568e-05,
+      "loss": 2.9565,
+      "step": 48530
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8903121734844087e-05,
+      "loss": 3.0604,
+      "step": 48535
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.889406070963324e-05,
+      "loss": 2.9235,
+      "step": 48540
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.888500135067246e-05,
+      "loss": 2.9342,
+      "step": 48545
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8875943658446965e-05,
+      "loss": 2.949,
+      "step": 48550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8866887633442014e-05,
+      "loss": 2.8828,
+      "step": 48555
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8857833276142662e-05,
+      "loss": 2.8679,
+      "step": 48560
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8848780587033938e-05,
+      "loss": 2.9562,
+      "step": 48565
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8839729566600767e-05,
+      "loss": 3.1561,
+      "step": 48570
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8830680215327984e-05,
+      "loss": 2.8816,
+      "step": 48575
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8821632533700332e-05,
+      "loss": 3.0273,
+      "step": 48580
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8812586522202495e-05,
+      "loss": 2.9333,
+      "step": 48585
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8803542181318978e-05,
+      "loss": 2.9399,
+      "step": 48590
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.8794499511534335e-05,
+      "loss": 2.9347,
+      "step": 48595
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8785458513332883e-05,
+      "loss": 2.9487,
+      "step": 48600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8776419187198985e-05,
+      "loss": 2.9354,
+      "step": 48605
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.87673815336168e-05,
+      "loss": 2.9352,
+      "step": 48610
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.875834555307047e-05,
+      "loss": 2.8359,
+      "step": 48615
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8749311246044027e-05,
+      "loss": 2.8866,
+      "step": 48620
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.87402786130214e-05,
+      "loss": 3.0533,
+      "step": 48625
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8731247654486444e-05,
+      "loss": 2.8466,
+      "step": 48630
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.872221837092294e-05,
+      "loss": 2.8321,
+      "step": 48635
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.871319076281451e-05,
+      "loss": 2.7919,
+      "step": 48640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8704164830644792e-05,
+      "loss": 2.9739,
+      "step": 48645
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.869514057489722e-05,
+      "loss": 2.903,
+      "step": 48650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8686117996055262e-05,
+      "loss": 2.8197,
+      "step": 48655
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8677097094602175e-05,
+      "loss": 2.8962,
+      "step": 48660
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8668077871021207e-05,
+      "loss": 2.9394,
+      "step": 48665
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.865906032579548e-05,
+      "loss": 3.0274,
+      "step": 48670
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8650044459408038e-05,
+      "loss": 2.985,
+      "step": 48675
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.864103027234184e-05,
+      "loss": 2.9377,
+      "step": 48680
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8632017765079762e-05,
+      "loss": 2.8222,
+      "step": 48685
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8623006938104525e-05,
+      "loss": 2.8923,
+      "step": 48690
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8613997791898875e-05,
+      "loss": 2.9394,
+      "step": 48695
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8604990326945344e-05,
+      "loss": 2.9452,
+      "step": 48700
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8595984543726497e-05,
+      "loss": 2.834,
+      "step": 48705
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.85869804427247e-05,
+      "loss": 3.0529,
+      "step": 48710
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8577978024422288e-05,
+      "loss": 2.9277,
+      "step": 48715
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8568977289301493e-05,
+      "loss": 2.9487,
+      "step": 48720
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.855997823784446e-05,
+      "loss": 3.0044,
+      "step": 48725
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8550980870533237e-05,
+      "loss": 2.9167,
+      "step": 48730
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.854198518784981e-05,
+      "loss": 2.8341,
+      "step": 48735
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8532991190275983e-05,
+      "loss": 2.9605,
+      "step": 48740
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8523998878293625e-05,
+      "loss": 2.9849,
+      "step": 48745
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8515008252384346e-05,
+      "loss": 2.9962,
+      "step": 48750
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8506019313029815e-05,
+      "loss": 2.9639,
+      "step": 48755
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.84970320607115e-05,
+      "loss": 2.8616,
+      "step": 48760
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8488046495910827e-05,
+      "loss": 2.9497,
+      "step": 48765
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8479062619109127e-05,
+      "loss": 2.981,
+      "step": 48770
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8470080430787645e-05,
+      "loss": 2.7402,
+      "step": 48775
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.846109993142752e-05,
+      "loss": 2.909,
+      "step": 48780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8452121121509824e-05,
+      "loss": 2.8682,
+      "step": 48785
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.844314400151551e-05,
+      "loss": 2.9183,
+      "step": 48790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8434168571925465e-05,
+      "loss": 3.0646,
+      "step": 48795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8425194833220467e-05,
+      "loss": 2.9989,
+      "step": 48800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8416222785881216e-05,
+      "loss": 2.84,
+      "step": 48805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8407252430388334e-05,
+      "loss": 2.8291,
+      "step": 48810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8398283767222284e-05,
+      "loss": 2.8312,
+      "step": 48815
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.838931679686356e-05,
+      "loss": 2.8434,
+      "step": 48820
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8380351519792438e-05,
+      "loss": 2.9326,
+      "step": 48825
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8371387936489187e-05,
+      "loss": 2.8733,
+      "step": 48830
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.836242604743395e-05,
+      "loss": 2.983,
+      "step": 48835
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8353465853106794e-05,
+      "loss": 2.972,
+      "step": 48840
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.834450735398769e-05,
+      "loss": 2.9776,
+      "step": 48845
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.833555055055651e-05,
+      "loss": 2.8281,
+      "step": 48850
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8326595443293048e-05,
+      "loss": 2.8836,
+      "step": 48855
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8317642032676997e-05,
+      "loss": 3.0397,
+      "step": 48860
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.830869031918797e-05,
+      "loss": 2.7814,
+      "step": 48865
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.82997403033055e-05,
+      "loss": 3.0585,
+      "step": 48870
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8290791985508953e-05,
+      "loss": 2.8844,
+      "step": 48875
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.828184536627774e-05,
+      "loss": 2.8655,
+      "step": 48880
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.827290044609103e-05,
+      "loss": 2.9774,
+      "step": 48885
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8263957225428047e-05,
+      "loss": 2.8912,
+      "step": 48890
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.82550157047678e-05,
+      "loss": 2.9689,
+      "step": 48895
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.824607588458927e-05,
+      "loss": 2.9169,
+      "step": 48900
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8237137765371348e-05,
+      "loss": 3.0176,
+      "step": 48905
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.822820134759281e-05,
+      "loss": 2.7617,
+      "step": 48910
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8219266631732358e-05,
+      "loss": 2.9493,
+      "step": 48915
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8210333618268614e-05,
+      "loss": 2.9862,
+      "step": 48920
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8201402307680043e-05,
+      "loss": 2.899,
+      "step": 48925
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.819247270044513e-05,
+      "loss": 3.0283,
+      "step": 48930
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8183544797042147e-05,
+      "loss": 2.9357,
+      "step": 48935
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8174618597949394e-05,
+      "loss": 3.0243,
+      "step": 48940
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8165694103644976e-05,
+      "loss": 2.9039,
+      "step": 48945
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8156771314606962e-05,
+      "loss": 2.8699,
+      "step": 48950
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8147850231313323e-05,
+      "loss": 2.8951,
+      "step": 48955
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8138930854241935e-05,
+      "loss": 3.0636,
+      "step": 48960
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.813001318387058e-05,
+      "loss": 2.8437,
+      "step": 48965
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8121097220676962e-05,
+      "loss": 2.9086,
+      "step": 48970
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.811218296513864e-05,
+      "loss": 2.8818,
+      "step": 48975
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8103270417733188e-05,
+      "loss": 2.9542,
+      "step": 48980
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.809435957893795e-05,
+      "loss": 2.8569,
+      "step": 48985
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8085450449230335e-05,
+      "loss": 2.8906,
+      "step": 48990
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8076543029087508e-05,
+      "loss": 2.9201,
+      "step": 48995
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8067637318986642e-05,
+      "loss": 2.864,
+      "step": 49000
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8058733319404785e-05,
+      "loss": 2.9479,
+      "step": 49005
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8049831030818897e-05,
+      "loss": 2.9574,
+      "step": 49010
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8040930453705847e-05,
+      "loss": 2.9574,
+      "step": 49015
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8032031588542426e-05,
+      "loss": 2.9254,
+      "step": 49020
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.802313443580527e-05,
+      "loss": 2.8345,
+      "step": 49025
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.8014238995971045e-05,
+      "loss": 3.0113,
+      "step": 49030
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.800534526951618e-05,
+      "loss": 2.9962,
+      "step": 49035
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7996453256917152e-05,
+      "loss": 2.8555,
+      "step": 49040
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7987562958650222e-05,
+      "loss": 2.9009,
+      "step": 49045
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7978674375191645e-05,
+      "loss": 2.8574,
+      "step": 49050
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7969787507017554e-05,
+      "loss": 2.9226,
+      "step": 49055
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7960902354603982e-05,
+      "loss": 2.9663,
+      "step": 49060
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7952018918426887e-05,
+      "loss": 3.0126,
+      "step": 49065
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7943137198962123e-05,
+      "loss": 2.8957,
+      "step": 49070
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7934257196685463e-05,
+      "loss": 2.9316,
+      "step": 49075
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7925378912072595e-05,
+      "loss": 2.9433,
+      "step": 49080
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7916502345599047e-05,
+      "loss": 2.9748,
+      "step": 49085
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7907627497740377e-05,
+      "loss": 2.865,
+      "step": 49090
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7898754368971942e-05,
+      "loss": 3.0206,
+      "step": 49095
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7889882959769055e-05,
+      "loss": 2.8835,
+      "step": 49100
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7881013270606938e-05,
+      "loss": 2.7782,
+      "step": 49105
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7872145301960708e-05,
+      "loss": 2.9614,
+      "step": 49110
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7863279054305398e-05,
+      "loss": 2.89,
+      "step": 49115
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7854414528115943e-05,
+      "loss": 3.0055,
+      "step": 49120
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7845551723867187e-05,
+      "loss": 2.816,
+      "step": 49125
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7836690642033892e-05,
+      "loss": 2.8562,
+      "step": 49130
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7827831283090714e-05,
+      "loss": 2.8059,
+      "step": 49135
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7818973647512215e-05,
+      "loss": 3.032,
+      "step": 49140
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7810117735772897e-05,
+      "loss": 2.8307,
+      "step": 49145
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.780126354834709e-05,
+      "loss": 2.9125,
+      "step": 49150
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.779241108570916e-05,
+      "loss": 2.9589,
+      "step": 49155
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7783560348333227e-05,
+      "loss": 2.919,
+      "step": 49160
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7774711336693473e-05,
+      "loss": 3.0232,
+      "step": 49165
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7765864051263864e-05,
+      "loss": 2.8452,
+      "step": 49170
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7757018492518328e-05,
+      "loss": 3.0076,
+      "step": 49175
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.774817466093071e-05,
+      "loss": 3.0179,
+      "step": 49180
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7739332556974736e-05,
+      "loss": 2.9048,
+      "step": 49185
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7730492181124047e-05,
+      "loss": 2.9711,
+      "step": 49190
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7721653533852227e-05,
+      "loss": 2.8524,
+      "step": 49195
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.771281661563267e-05,
+      "loss": 2.917,
+      "step": 49200
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7703981426938822e-05,
+      "loss": 2.9441,
+      "step": 49205
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7695147968243875e-05,
+      "loss": 2.8685,
+      "step": 49210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7686316240021088e-05,
+      "loss": 2.9656,
+      "step": 49215
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.76774862427435e-05,
+      "loss": 2.8701,
+      "step": 49220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7668657976884123e-05,
+      "loss": 2.9202,
+      "step": 49225
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7659831442915857e-05,
+      "loss": 3.0037,
+      "step": 49230
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.765100664131151e-05,
+      "loss": 2.973,
+      "step": 49235
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7642183572543808e-05,
+      "loss": 3.0641,
+      "step": 49240
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7633362237085388e-05,
+      "loss": 2.9256,
+      "step": 49245
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.762454263540873e-05,
+      "loss": 2.889,
+      "step": 49250
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7615724767986342e-05,
+      "loss": 2.9517,
+      "step": 49255
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7606908635290504e-05,
+      "loss": 3.0646,
+      "step": 49260
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7598094237793533e-05,
+      "loss": 2.946,
+      "step": 49265
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7589281575967543e-05,
+      "loss": 2.8581,
+      "step": 49270
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.7580470650284615e-05,
+      "loss": 2.979,
+      "step": 49275
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7571661461216727e-05,
+      "loss": 2.9811,
+      "step": 49280
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7562854009235758e-05,
+      "loss": 3.0127,
+      "step": 49285
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.755404829481349e-05,
+      "loss": 2.9622,
+      "step": 49290
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.754524431842165e-05,
+      "loss": 2.9655,
+      "step": 49295
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7536442080531775e-05,
+      "loss": 2.9007,
+      "step": 49300
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.752764158161545e-05,
+      "loss": 2.8875,
+      "step": 49305
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7518842822144023e-05,
+      "loss": 3.0861,
+      "step": 49310
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7510045802588874e-05,
+      "loss": 2.8167,
+      "step": 49315
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.75012505234212e-05,
+      "loss": 2.99,
+      "step": 49320
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7492456985112137e-05,
+      "loss": 2.9005,
+      "step": 49325
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7483665188132737e-05,
+      "loss": 2.8882,
+      "step": 49330
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7474875132953943e-05,
+      "loss": 2.8782,
+      "step": 49335
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7466086820046617e-05,
+      "loss": 2.8197,
+      "step": 49340
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7457300249881543e-05,
+      "loss": 2.9277,
+      "step": 49345
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.744851542292933e-05,
+      "loss": 3.0083,
+      "step": 49350
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7439732339660624e-05,
+      "loss": 2.8776,
+      "step": 49355
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7430951000545847e-05,
+      "loss": 2.8901,
+      "step": 49360
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7422171406055453e-05,
+      "loss": 2.8344,
+      "step": 49365
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7413393556659684e-05,
+      "loss": 2.8864,
+      "step": 49370
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7404617452828763e-05,
+      "loss": 2.9086,
+      "step": 49375
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.73958430950328e-05,
+      "loss": 3.0156,
+      "step": 49380
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7387070483741803e-05,
+      "loss": 2.9155,
+      "step": 49385
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7378299619425697e-05,
+      "loss": 2.9529,
+      "step": 49390
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.736953050255432e-05,
+      "loss": 2.9461,
+      "step": 49395
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7360763133597398e-05,
+      "loss": 2.8897,
+      "step": 49400
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.735199751302457e-05,
+      "loss": 2.866,
+      "step": 49405
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.734323364130539e-05,
+      "loss": 3.0845,
+      "step": 49410
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7334471518909306e-05,
+      "loss": 2.9137,
+      "step": 49415
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.73257111463057e-05,
+      "loss": 2.8281,
+      "step": 49420
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.73169525239638e-05,
+      "loss": 2.8969,
+      "step": 49425
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.73081956523528e-05,
+      "loss": 2.9742,
+      "step": 49430
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7299440531941775e-05,
+      "loss": 3.0714,
+      "step": 49435
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7290687163199713e-05,
+      "loss": 2.9642,
+      "step": 49440
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.72819355465955e-05,
+      "loss": 2.9916,
+      "step": 49445
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.727318568259794e-05,
+      "loss": 2.9689,
+      "step": 49450
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.726443757167574e-05,
+      "loss": 2.9686,
+      "step": 49455
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.72556912142975e-05,
+      "loss": 2.9467,
+      "step": 49460
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7246946610931734e-05,
+      "loss": 3.035,
+      "step": 49465
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7238203762046888e-05,
+      "loss": 2.9146,
+      "step": 49470
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7229462668111235e-05,
+      "loss": 2.9117,
+      "step": 49475
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7220723329593082e-05,
+      "loss": 2.8787,
+      "step": 49480
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.721198574696049e-05,
+      "loss": 3.0099,
+      "step": 49485
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7203249920681587e-05,
+      "loss": 2.8945,
+      "step": 49490
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.719451585122424e-05,
+      "loss": 2.9396,
+      "step": 49495
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7185783539056387e-05,
+      "loss": 3.0392,
+      "step": 49500
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7177052984645736e-05,
+      "loss": 3.0484,
+      "step": 49505
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7168324188459978e-05,
+      "loss": 2.9874,
+      "step": 49510
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7159597150966678e-05,
+      "loss": 2.9442,
+      "step": 49515
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.715087187263332e-05,
+      "loss": 2.9634,
+      "step": 49520
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7142148353927294e-05,
+      "loss": 3.0258,
+      "step": 49525
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.713342659531591e-05,
+      "loss": 2.8776,
+      "step": 49530
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7124706597266315e-05,
+      "loss": 2.818,
+      "step": 49535
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.711598836024567e-05,
+      "loss": 2.9112,
+      "step": 49540
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.710727188472093e-05,
+      "loss": 2.9229,
+      "step": 49545
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7098557171159075e-05,
+      "loss": 3.0427,
+      "step": 49550
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7089844220026868e-05,
+      "loss": 2.9916,
+      "step": 49555
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.708113303179106e-05,
+      "loss": 2.8268,
+      "step": 49560
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.707242360691827e-05,
+      "loss": 2.8672,
+      "step": 49565
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7063715945875053e-05,
+      "loss": 2.8747,
+      "step": 49570
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.705501004912784e-05,
+      "loss": 2.862,
+      "step": 49575
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7046305917142996e-05,
+      "loss": 3.0489,
+      "step": 49580
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7037603550386728e-05,
+      "loss": 3.0463,
+      "step": 49585
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7028902949325265e-05,
+      "loss": 2.8597,
+      "step": 49590
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.7020204114424598e-05,
+      "loss": 2.9608,
+      "step": 49595
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.701150704615077e-05,
+      "loss": 3.0513,
+      "step": 49600
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.70028117449696e-05,
+      "loss": 2.9095,
+      "step": 49605
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6994118211346894e-05,
+      "loss": 3.0271,
+      "step": 49610
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6985426445748327e-05,
+      "loss": 2.9497,
+      "step": 49615
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6976736448639496e-05,
+      "loss": 2.9373,
+      "step": 49620
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.69680482204859e-05,
+      "loss": 2.9344,
+      "step": 49625
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6959361761752958e-05,
+      "loss": 3.0011,
+      "step": 49630
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.695067707290592e-05,
+      "loss": 3.047,
+      "step": 49635
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.694199415441007e-05,
+      "loss": 2.8488,
+      "step": 49640
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.693331300673045e-05,
+      "loss": 2.8814,
+      "step": 49645
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6924633630332166e-05,
+      "loss": 2.8288,
+      "step": 49650
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6915956025680086e-05,
+      "loss": 3.0356,
+      "step": 49655
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6907280193239056e-05,
+      "loss": 2.9373,
+      "step": 49660
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.689860613347382e-05,
+      "loss": 2.8517,
+      "step": 49665
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6889933846849022e-05,
+      "loss": 2.9737,
+      "step": 49670
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.68812633338292e-05,
+      "loss": 2.911,
+      "step": 49675
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6872594594878817e-05,
+      "loss": 2.9204,
+      "step": 49680
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6863927630462222e-05,
+      "loss": 2.9317,
+      "step": 49685
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6855262441043707e-05,
+      "loss": 2.9243,
+      "step": 49690
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6846599027087378e-05,
+      "loss": 3.0026,
+      "step": 49695
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.683793738905738e-05,
+      "loss": 3.065,
+      "step": 49700
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6829277527417637e-05,
+      "loss": 2.8681,
+      "step": 49705
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.682061944263205e-05,
+      "loss": 2.8412,
+      "step": 49710
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.681196313516441e-05,
+      "loss": 2.8986,
+      "step": 49715
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6803308605478403e-05,
+      "loss": 3.0037,
+      "step": 49720
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.679465585403763e-05,
+      "loss": 2.97,
+      "step": 49725
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6786004881305583e-05,
+      "loss": 3.0311,
+      "step": 49730
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6777355687745677e-05,
+      "loss": 2.8295,
+      "step": 49735
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6768708273821228e-05,
+      "loss": 2.8288,
+      "step": 49740
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.676006263999544e-05,
+      "loss": 2.9135,
+      "step": 49745
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6751418786731438e-05,
+      "loss": 2.9575,
+      "step": 49750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6742776714492263e-05,
+      "loss": 2.9904,
+      "step": 49755
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6734136423740793e-05,
+      "loss": 2.8491,
+      "step": 49760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.672549791493993e-05,
+      "loss": 2.924,
+      "step": 49765
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.671686118855237e-05,
+      "loss": 2.8723,
+      "step": 49770
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6708226245040758e-05,
+      "loss": 2.9203,
+      "step": 49775
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.669959308486765e-05,
+      "loss": 2.9873,
+      "step": 49780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.66909617084955e-05,
+      "loss": 2.7997,
+      "step": 49785
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6682332116386656e-05,
+      "loss": 2.8824,
+      "step": 49790
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.667370430900339e-05,
+      "loss": 2.8607,
+      "step": 49795
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6665078286807857e-05,
+      "loss": 2.9191,
+      "step": 49800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.665645405026215e-05,
+      "loss": 3.0232,
+      "step": 49805
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.664783159982819e-05,
+      "loss": 2.9262,
+      "step": 49810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6639210935967926e-05,
+      "loss": 3.0606,
+      "step": 49815
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6630592059143067e-05,
+      "loss": 2.9075,
+      "step": 49820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6621974969815367e-05,
+      "loss": 2.9985,
+      "step": 49825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6613359668446378e-05,
+      "loss": 2.8824,
+      "step": 49830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6604746155497603e-05,
+      "loss": 2.9125,
+      "step": 49835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.659613443143044e-05,
+      "loss": 3.1521,
+      "step": 49840
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.658752449670619e-05,
+      "loss": 2.9985,
+      "step": 49845
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6578916351786077e-05,
+      "loss": 2.9491,
+      "step": 49850
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.657030999713121e-05,
+      "loss": 2.9276,
+      "step": 49855
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6561705433202575e-05,
+      "loss": 2.9826,
+      "step": 49860
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6553102660461138e-05,
+      "loss": 3.011,
+      "step": 49865
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6544501679367668e-05,
+      "loss": 2.9894,
+      "step": 49870
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6535902490382966e-05,
+      "loss": 2.9288,
+      "step": 49875
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.65273050939676e-05,
+      "loss": 3.045,
+      "step": 49880
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.651870949058213e-05,
+      "loss": 2.9061,
+      "step": 49885
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6510115680687e-05,
+      "loss": 2.9534,
+      "step": 49890
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6501523664742547e-05,
+      "loss": 3.0019,
+      "step": 49895
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.649293344320902e-05,
+      "loss": 2.9682,
+      "step": 49900
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6484345016546593e-05,
+      "loss": 3.0305,
+      "step": 49905
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6475758385215267e-05,
+      "loss": 2.8317,
+      "step": 49910
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.646717354967507e-05,
+      "loss": 2.9885,
+      "step": 49915
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6458590510385795e-05,
+      "loss": 2.9639,
+      "step": 49920
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6450009267807282e-05,
+      "loss": 2.9405,
+      "step": 49925
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6441429822399153e-05,
+      "loss": 2.9193,
+      "step": 49930
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6432852174620987e-05,
+      "loss": 2.8023,
+      "step": 49935
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6424276324932276e-05,
+      "loss": 2.7204,
+      "step": 49940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6415702273792394e-05,
+      "loss": 2.8812,
+      "step": 49945
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6407130021660634e-05,
+      "loss": 2.9133,
+      "step": 49950
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.6398559568996192e-05,
+      "loss": 2.9414,
+      "step": 49955
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6389990916258117e-05,
+      "loss": 2.9071,
+      "step": 49960
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6381424063905476e-05,
+      "loss": 2.8184,
+      "step": 49965
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6372859012397094e-05,
+      "loss": 2.8426,
+      "step": 49970
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.636429576219185e-05,
+      "loss": 3.025,
+      "step": 49975
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6355734313748393e-05,
+      "loss": 2.8727,
+      "step": 49980
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.634717466752535e-05,
+      "loss": 2.872,
+      "step": 49985
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6338616823981244e-05,
+      "loss": 2.9813,
+      "step": 49990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.6330060783574484e-05,
+      "loss": 2.8158,
+      "step": 49995
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.63215065467634e-05,
+      "loss": 2.9495,
+      "step": 50000
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 67969,
+  "num_train_epochs": 1,
+  "save_steps": 1000,
+  "total_flos": 7.6970578673664e+18,
+  "trial_name": null,
+  "trial_params": null
+}