diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,4696 @@
+{
+  "best_metric": 0.2831316888332367,
+  "best_model_checkpoint": "./results/checkpoint-6000",
+  "epoch": 1.9601437438745508,
+  "global_step": 6000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.6936,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.6923,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3e-06,
+      "loss": 0.6935,
+      "step": 30
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.6924,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5e-06,
+      "loss": 0.6939,
+      "step": 50
+    },
+    {
+      "epoch": 0.02,
+      "eval_accuracy": 0.5167850852012634,
+      "eval_loss": 0.6928304433822632,
+      "eval_runtime": 3067.7165,
+      "eval_samples_per_second": 3.991,
+      "eval_steps_per_second": 0.063,
+      "step": 50
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6e-06,
+      "loss": 0.6924,
+      "step": 60
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.000000000000001e-06,
+      "loss": 0.694,
+      "step": 70
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.693,
+      "step": 80
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9e-06,
+      "loss": 0.6924,
+      "step": 90
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1e-05,
+      "loss": 0.6928,
+      "step": 100
+    },
+    {
+      "epoch": 0.03,
+      "eval_accuracy": 0.5167850852012634,
+      "eval_loss": 0.6926709413528442,
+      "eval_runtime": 3341.1644,
+      "eval_samples_per_second": 3.664,
+      "eval_steps_per_second": 0.057,
+      "step": 100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 0.6934,
+      "step": 110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2e-05,
+      "loss": 0.693,
+      "step": 120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 0.693,
+      "step": 130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.4000000000000001e-05,
+      "loss": 0.6888,
+      "step": 140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5e-05,
+      "loss": 0.68,
+      "step": 150
+    },
+    {
+      "epoch": 0.05,
+      "eval_accuracy": 0.6862697005271912,
+      "eval_loss": 0.6718223690986633,
+      "eval_runtime": 2945.3733,
+      "eval_samples_per_second": 4.157,
+      "eval_steps_per_second": 0.065,
+      "step": 150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.691,
+      "step": 160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.7000000000000003e-05,
+      "loss": 0.6657,
+      "step": 170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8e-05,
+      "loss": 0.6501,
+      "step": 180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9e-05,
+      "loss": 0.6251,
+      "step": 190
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2e-05,
+      "loss": 0.6026,
+      "step": 200
+    },
+    {
+      "epoch": 0.07,
+      "eval_accuracy": 0.7803642749786377,
+      "eval_loss": 0.5917710065841675,
+      "eval_runtime": 2925.3749,
+      "eval_samples_per_second": 4.185,
+      "eval_steps_per_second": 0.066,
+      "step": 200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.1e-05,
+      "loss": 0.6115,
+      "step": 210
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.2000000000000003e-05,
+      "loss": 0.6038,
+      "step": 220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.3000000000000003e-05,
+      "loss": 0.5834,
+      "step": 230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.4e-05,
+      "loss": 0.5658,
+      "step": 240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.5e-05,
+      "loss": 0.5625,
+      "step": 250
+    },
+    {
+      "epoch": 0.08,
+      "eval_accuracy": 0.7969452142715454,
+      "eval_loss": 0.5428493618965149,
+      "eval_runtime": 2992.1851,
+      "eval_samples_per_second": 4.092,
+      "eval_steps_per_second": 0.064,
+      "step": 250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.6000000000000002e-05,
+      "loss": 0.5542,
+      "step": 260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.7000000000000002e-05,
+      "loss": 0.5558,
+      "step": 270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.8000000000000003e-05,
+      "loss": 0.5051,
+      "step": 280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.9e-05,
+      "loss": 0.5449,
+      "step": 290
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3e-05,
+      "loss": 0.5634,
+      "step": 300
+    },
+    {
+      "epoch": 0.1,
+      "eval_accuracy": 0.799967348575592,
+      "eval_loss": 0.5014462471008301,
+      "eval_runtime": 2928.187,
+      "eval_samples_per_second": 4.181,
+      "eval_steps_per_second": 0.066,
+      "step": 300
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.1e-05,
+      "loss": 0.5279,
+      "step": 310
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 3.2000000000000005e-05,
+      "loss": 0.5098,
+      "step": 320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.3e-05,
+      "loss": 0.5059,
+      "step": 330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.4000000000000007e-05,
+      "loss": 0.494,
+      "step": 340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3.5e-05,
+      "loss": 0.4813,
+      "step": 350
+    },
+    {
+      "epoch": 0.11,
+      "eval_accuracy": 0.826839804649353,
+      "eval_loss": 0.4505186676979065,
+      "eval_runtime": 2897.6636,
+      "eval_samples_per_second": 4.225,
+      "eval_steps_per_second": 0.066,
+      "step": 350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.6e-05,
+      "loss": 0.4341,
+      "step": 360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.7e-05,
+      "loss": 0.4513,
+      "step": 370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.8e-05,
+      "loss": 0.4994,
+      "step": 380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.9000000000000006e-05,
+      "loss": 0.4634,
+      "step": 390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4e-05,
+      "loss": 0.505,
+      "step": 400
+    },
+    {
+      "epoch": 0.13,
+      "eval_accuracy": 0.7340521216392517,
+      "eval_loss": 0.5640795230865479,
+      "eval_runtime": 2900.3415,
+      "eval_samples_per_second": 4.221,
+      "eval_steps_per_second": 0.066,
+      "step": 400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 4.1e-05,
+      "loss": 0.4905,
+      "step": 410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.2e-05,
+      "loss": 0.5808,
+      "step": 420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.3e-05,
+      "loss": 0.5112,
+      "step": 430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.4000000000000006e-05,
+      "loss": 0.5161,
+      "step": 440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.5e-05,
+      "loss": 0.58,
+      "step": 450
+    },
+    {
+      "epoch": 0.15,
+      "eval_accuracy": 0.8172833323478699,
+      "eval_loss": 0.45767492055892944,
+      "eval_runtime": 2972.4836,
+      "eval_samples_per_second": 4.119,
+      "eval_steps_per_second": 0.065,
+      "step": 450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.600000000000001e-05,
+      "loss": 0.5511,
+      "step": 460
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.7e-05,
+      "loss": 0.4798,
+      "step": 470
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.8e-05,
+      "loss": 0.4224,
+      "step": 480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.9e-05,
+      "loss": 0.373,
+      "step": 490
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 5e-05,
+      "loss": 0.5414,
+      "step": 500
+    },
+    {
+      "epoch": 0.16,
+      "eval_accuracy": 0.8167932629585266,
+      "eval_loss": 0.4663669466972351,
+      "eval_runtime": 2920.5709,
+      "eval_samples_per_second": 4.192,
+      "eval_steps_per_second": 0.066,
+      "step": 500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.994241621559369e-05,
+      "loss": 0.5669,
+      "step": 510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.988483243118738e-05,
+      "loss": 0.5316,
+      "step": 520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.9827248646781064e-05,
+      "loss": 0.4944,
+      "step": 530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.976966486237476e-05,
+      "loss": 0.5185,
+      "step": 540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.971208107796845e-05,
+      "loss": 0.461,
+      "step": 550
+    },
+    {
+      "epoch": 0.18,
+      "eval_accuracy": 0.8294535875320435,
+      "eval_loss": 0.4393218457698822,
+      "eval_runtime": 2952.3148,
+      "eval_samples_per_second": 4.147,
+      "eval_steps_per_second": 0.065,
+      "step": 550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 4.965449729356213e-05,
+      "loss": 0.4398,
+      "step": 560
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.959691350915582e-05,
+      "loss": 0.3987,
+      "step": 570
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.953932972474951e-05,
+      "loss": 0.5216,
+      "step": 580
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 4.94817459403432e-05,
+      "loss": 0.4074,
+      "step": 590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.942416215593689e-05,
+      "loss": 0.5235,
+      "step": 600
+    },
+    {
+      "epoch": 0.2,
+      "eval_accuracy": 0.8164665699005127,
+      "eval_loss": 0.4412307143211365,
+      "eval_runtime": 3037.7485,
+      "eval_samples_per_second": 4.03,
+      "eval_steps_per_second": 0.063,
+      "step": 600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.936657837153058e-05,
+      "loss": 0.5466,
+      "step": 610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 4.930899458712427e-05,
+      "loss": 0.5202,
+      "step": 620
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.9251410802717954e-05,
+      "loss": 0.4504,
+      "step": 630
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.919382701831164e-05,
+      "loss": 0.558,
+      "step": 640
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 4.913624323390533e-05,
+      "loss": 0.5234,
+      "step": 650
+    },
+    {
+      "epoch": 0.21,
+      "eval_accuracy": 0.8162215352058411,
+      "eval_loss": 0.48874276876449585,
+      "eval_runtime": 2990.9698,
+      "eval_samples_per_second": 4.093,
+      "eval_steps_per_second": 0.064,
+      "step": 650
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.907865944949903e-05,
+      "loss": 0.4322,
+      "step": 660
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.902107566509271e-05,
+      "loss": 0.6654,
+      "step": 670
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 4.89634918806864e-05,
+      "loss": 0.3895,
+      "step": 680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.890590809628009e-05,
+      "loss": 0.4972,
+      "step": 690
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.884832431187378e-05,
+      "loss": 0.4382,
+      "step": 700
+    },
+    {
+      "epoch": 0.23,
+      "eval_accuracy": 0.7774238586425781,
+      "eval_loss": 0.5239776968955994,
+      "eval_runtime": 3050.3978,
+      "eval_samples_per_second": 4.014,
+      "eval_steps_per_second": 0.063,
+      "step": 700
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 4.8790740527467464e-05,
+      "loss": 0.4849,
+      "step": 710
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.873315674306116e-05,
+      "loss": 0.6191,
+      "step": 720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.867557295865485e-05,
+      "loss": 0.4692,
+      "step": 730
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 4.861798917424853e-05,
+      "loss": 0.4188,
+      "step": 740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.856040538984222e-05,
+      "loss": 0.4231,
+      "step": 750
+    },
+    {
+      "epoch": 0.25,
+      "eval_accuracy": 0.8423588871955872,
+      "eval_loss": 0.395524799823761,
+      "eval_runtime": 3016.9494,
+      "eval_samples_per_second": 4.058,
+      "eval_steps_per_second": 0.064,
+      "step": 750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.850282160543591e-05,
+      "loss": 0.4108,
+      "step": 760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.84452378210296e-05,
+      "loss": 0.424,
+      "step": 770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 4.8387654036623284e-05,
+      "loss": 0.3364,
+      "step": 780
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.833007025221698e-05,
+      "loss": 0.3836,
+      "step": 790
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.827248646781067e-05,
+      "loss": 0.4643,
+      "step": 800
+    },
+    {
+      "epoch": 0.26,
+      "eval_accuracy": 0.755533754825592,
+      "eval_loss": 0.5963751673698425,
+      "eval_runtime": 3041.0395,
+      "eval_samples_per_second": 4.026,
+      "eval_steps_per_second": 0.063,
+      "step": 800
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 4.821490268340435e-05,
+      "loss": 0.4687,
+      "step": 810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.815731889899804e-05,
+      "loss": 0.34,
+      "step": 820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.809973511459173e-05,
+      "loss": 0.3561,
+      "step": 830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 4.804215133018542e-05,
+      "loss": 0.4232,
+      "step": 840
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.798456754577911e-05,
+      "loss": 0.4381,
+      "step": 850
+    },
+    {
+      "epoch": 0.28,
+      "eval_accuracy": 0.848648190498352,
+      "eval_loss": 0.40665099024772644,
+      "eval_runtime": 2945.2579,
+      "eval_samples_per_second": 4.157,
+      "eval_steps_per_second": 0.065,
+      "step": 850
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.79269837613728e-05,
+      "loss": 0.4271,
+      "step": 860
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 4.786939997696649e-05,
+      "loss": 0.4871,
+      "step": 870
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.781181619256018e-05,
+      "loss": 0.5369,
+      "step": 880
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.775423240815386e-05,
+      "loss": 0.548,
+      "step": 890
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 4.769664862374755e-05,
+      "loss": 0.4326,
+      "step": 900
+    },
+    {
+      "epoch": 0.29,
+      "eval_accuracy": 0.8188352584838867,
+      "eval_loss": 0.40551692247390747,
+      "eval_runtime": 2910.8132,
+      "eval_samples_per_second": 4.206,
+      "eval_steps_per_second": 0.066,
+      "step": 900
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.763906483934125e-05,
+      "loss": 0.3592,
+      "step": 910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.758148105493493e-05,
+      "loss": 0.29,
+      "step": 920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.752389727052862e-05,
+      "loss": 0.5536,
+      "step": 930
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.746631348612231e-05,
+      "loss": 0.3507,
+      "step": 940
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.7408729701716e-05,
+      "loss": 0.4433,
+      "step": 950
+    },
+    {
+      "epoch": 0.31,
+      "eval_accuracy": 0.8152413368225098,
+      "eval_loss": 0.4874822497367859,
+      "eval_runtime": 3097.8896,
+      "eval_samples_per_second": 3.952,
+      "eval_steps_per_second": 0.062,
+      "step": 950
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 4.735114591730968e-05,
+      "loss": 0.5522,
+      "step": 960
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.729356213290337e-05,
+      "loss": 0.478,
+      "step": 970
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.723597834849707e-05,
+      "loss": 0.5597,
+      "step": 980
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 4.717839456409076e-05,
+      "loss": 0.5501,
+      "step": 990
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.712081077968444e-05,
+      "loss": 0.6363,
+      "step": 1000
+    },
+    {
+      "epoch": 0.33,
+      "eval_accuracy": 0.5706117749214172,
+      "eval_loss": 0.6323691010475159,
+      "eval_runtime": 3016.3709,
+      "eval_samples_per_second": 4.059,
+      "eval_steps_per_second": 0.064,
+      "step": 1000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.706322699527813e-05,
+      "loss": 0.5962,
+      "step": 1010
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 4.700564321087182e-05,
+      "loss": 0.5009,
+      "step": 1020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.694805942646551e-05,
+      "loss": 0.4466,
+      "step": 1030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.68904756420592e-05,
+      "loss": 0.3369,
+      "step": 1040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 4.683289185765289e-05,
+      "loss": 0.4967,
+      "step": 1050
+    },
+    {
+      "epoch": 0.34,
+      "eval_accuracy": 0.8430123329162598,
+      "eval_loss": 0.39603397250175476,
+      "eval_runtime": 3109.9986,
+      "eval_samples_per_second": 3.937,
+      "eval_steps_per_second": 0.062,
+      "step": 1050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.677530807324658e-05,
+      "loss": 0.3918,
+      "step": 1060
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.671772428884026e-05,
+      "loss": 0.3638,
+      "step": 1070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 4.666014050443395e-05,
+      "loss": 0.4857,
+      "step": 1080
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.660255672002764e-05,
+      "loss": 0.4537,
+      "step": 1090
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.654497293562133e-05,
+      "loss": 0.4589,
+      "step": 1100
+    },
+    {
+      "epoch": 0.36,
+      "eval_accuracy": 0.7901657819747925,
+      "eval_loss": 0.4916388690471649,
+      "eval_runtime": 3035.7809,
+      "eval_samples_per_second": 4.033,
+      "eval_steps_per_second": 0.063,
+      "step": 1100
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 4.648738915121502e-05,
+      "loss": 0.4679,
+      "step": 1110
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.642980536680871e-05,
+      "loss": 0.371,
+      "step": 1120
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.63722215824024e-05,
+      "loss": 0.4058,
+      "step": 1130
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 4.631463779799608e-05,
+      "loss": 0.5627,
+      "step": 1140
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.625705401358977e-05,
+      "loss": 0.5172,
+      "step": 1150
+    },
+    {
+      "epoch": 0.38,
+      "eval_accuracy": 0.8287184238433838,
+      "eval_loss": 0.4040413200855255,
+      "eval_runtime": 3074.7633,
+      "eval_samples_per_second": 3.982,
+      "eval_steps_per_second": 0.062,
+      "step": 1150
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.619947022918347e-05,
+      "loss": 0.3419,
+      "step": 1160
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 4.614188644477716e-05,
+      "loss": 0.372,
+      "step": 1170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.608430266037084e-05,
+      "loss": 0.3666,
+      "step": 1180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.602671887596453e-05,
+      "loss": 0.428,
+      "step": 1190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 4.596913509155822e-05,
+      "loss": 0.4445,
+      "step": 1200
+    },
+    {
+      "epoch": 0.39,
+      "eval_accuracy": 0.8600016236305237,
+      "eval_loss": 0.34572649002075195,
+      "eval_runtime": 2914.5824,
+      "eval_samples_per_second": 4.201,
+      "eval_steps_per_second": 0.066,
+      "step": 1200
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.591155130715191e-05,
+      "loss": 0.4485,
+      "step": 1210
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.585396752274559e-05,
+      "loss": 0.3881,
+      "step": 1220
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 4.579638373833929e-05,
+      "loss": 0.4353,
+      "step": 1230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.573879995393298e-05,
+      "loss": 0.3187,
+      "step": 1240
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.568121616952666e-05,
+      "loss": 0.4123,
+      "step": 1250
+    },
+    {
+      "epoch": 0.41,
+      "eval_accuracy": 0.8562443852424622,
+      "eval_loss": 0.3715479373931885,
+      "eval_runtime": 2913.7922,
+      "eval_samples_per_second": 4.202,
+      "eval_steps_per_second": 0.066,
+      "step": 1250
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.562363238512035e-05,
+      "loss": 0.3874,
+      "step": 1260
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 4.556604860071404e-05,
+      "loss": 0.3507,
+      "step": 1270
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.550846481630773e-05,
+      "loss": 0.3694,
+      "step": 1280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.545088103190142e-05,
+      "loss": 0.3705,
+      "step": 1290
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 4.539329724749511e-05,
+      "loss": 0.3237,
+      "step": 1300
+    },
+    {
+      "epoch": 0.42,
+      "eval_accuracy": 0.8576329350471497,
+      "eval_loss": 0.36021336913108826,
+      "eval_runtime": 2909.6601,
+      "eval_samples_per_second": 4.208,
+      "eval_steps_per_second": 0.066,
+      "step": 1300
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.53357134630888e-05,
+      "loss": 0.3813,
+      "step": 1310
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.527812967868249e-05,
+      "loss": 0.3765,
+      "step": 1320
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 4.522054589427617e-05,
+      "loss": 0.5093,
+      "step": 1330
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.516296210986986e-05,
+      "loss": 0.4042,
+      "step": 1340
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.510537832546356e-05,
+      "loss": 0.3936,
+      "step": 1350
+    },
+    {
+      "epoch": 0.44,
+      "eval_accuracy": 0.8592665195465088,
+      "eval_loss": 0.34257298707962036,
+      "eval_runtime": 2947.6678,
+      "eval_samples_per_second": 4.153,
+      "eval_steps_per_second": 0.065,
+      "step": 1350
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 4.504779454105724e-05,
+      "loss": 0.3917,
+      "step": 1360
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.499021075665093e-05,
+      "loss": 0.3531,
+      "step": 1370
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.493262697224462e-05,
+      "loss": 0.3605,
+      "step": 1380
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 4.487504318783831e-05,
+      "loss": 0.3927,
+      "step": 1390
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.481745940343199e-05,
+      "loss": 0.3594,
+      "step": 1400
+    },
+    {
+      "epoch": 0.46,
+      "eval_accuracy": 0.8171200156211853,
+      "eval_loss": 0.45752185583114624,
+      "eval_runtime": 2897.6038,
+      "eval_samples_per_second": 4.225,
+      "eval_steps_per_second": 0.066,
+      "step": 1400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.475987561902568e-05,
+      "loss": 0.3721,
+      "step": 1410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 4.470229183461938e-05,
+      "loss": 0.7249,
+      "step": 1420
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.464470805021306e-05,
+      "loss": 0.4163,
+      "step": 1430
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.458712426580675e-05,
+      "loss": 0.3804,
+      "step": 1440
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 4.452954048140044e-05,
+      "loss": 0.3449,
+      "step": 1450
+    },
+    {
+      "epoch": 0.47,
+      "eval_accuracy": 0.8485665321350098,
+      "eval_loss": 0.4127117097377777,
+      "eval_runtime": 2920.5854,
+      "eval_samples_per_second": 4.192,
+      "eval_steps_per_second": 0.066,
+      "step": 1450
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.447195669699413e-05,
+      "loss": 0.3974,
+      "step": 1460
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.441437291258782e-05,
+      "loss": 0.4296,
+      "step": 1470
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 4.435678912818151e-05,
+      "loss": 0.4072,
+      "step": 1480
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.42992053437752e-05,
+      "loss": 0.5211,
+      "step": 1490
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.424162155936889e-05,
+      "loss": 0.3895,
+      "step": 1500
+    },
+    {
+      "epoch": 0.49,
+      "eval_accuracy": 0.8578779697418213,
+      "eval_loss": 0.3651716709136963,
+      "eval_runtime": 2933.5659,
+      "eval_samples_per_second": 4.173,
+      "eval_steps_per_second": 0.065,
+      "step": 1500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 4.418403777496257e-05,
+      "loss": 0.3782,
+      "step": 1510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.412645399055626e-05,
+      "loss": 0.3356,
+      "step": 1520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.406887020614995e-05,
+      "loss": 0.4293,
+      "step": 1530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 4.401128642174364e-05,
+      "loss": 0.338,
+      "step": 1540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.395370263733733e-05,
+      "loss": 0.344,
+      "step": 1550
+    },
+    {
+      "epoch": 0.51,
+      "eval_accuracy": 0.8640856146812439,
+      "eval_loss": 0.33252376317977905,
+      "eval_runtime": 2994.1013,
+      "eval_samples_per_second": 4.089,
+      "eval_steps_per_second": 0.064,
+      "step": 1550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.389611885293102e-05,
+      "loss": 0.2727,
+      "step": 1560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 4.383853506852471e-05,
+      "loss": 0.4033,
+      "step": 1570
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.378095128411839e-05,
+      "loss": 0.2831,
+      "step": 1580
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.372336749971208e-05,
+      "loss": 0.3471,
+      "step": 1590
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 4.3665783715305776e-05,
+      "loss": 0.364,
+      "step": 1600
+    },
+    {
+      "epoch": 0.52,
+      "eval_accuracy": 0.8733970522880554,
+      "eval_loss": 0.3105648458003998,
+      "eval_runtime": 3022.9008,
+      "eval_samples_per_second": 4.05,
+      "eval_steps_per_second": 0.064,
+      "step": 1600
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.3608199930899466e-05,
+      "loss": 0.3323,
+      "step": 1610
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.355061614649315e-05,
+      "loss": 0.272,
+      "step": 1620
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 4.349303236208684e-05,
+      "loss": 0.4414,
+      "step": 1630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.343544857768053e-05,
+      "loss": 0.3218,
+      "step": 1640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.337786479327422e-05,
+      "loss": 0.6309,
+      "step": 1650
+    },
+    {
+      "epoch": 0.54,
+      "eval_accuracy": 0.843910813331604,
+      "eval_loss": 0.364272803068161,
+      "eval_runtime": 2915.6475,
+      "eval_samples_per_second": 4.199,
+      "eval_steps_per_second": 0.066,
+      "step": 1650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 4.33202810088679e-05,
+      "loss": 0.2782,
+      "step": 1660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.32626972244616e-05,
+      "loss": 0.4917,
+      "step": 1670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.3205113440055286e-05,
+      "loss": 0.4247,
+      "step": 1680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 4.314752965564897e-05,
+      "loss": 0.4354,
+      "step": 1690
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.308994587124266e-05,
+      "loss": 0.4705,
+      "step": 1700
+    },
+    {
+      "epoch": 0.56,
+      "eval_accuracy": 0.8681695461273193,
+      "eval_loss": 0.3355846703052521,
+      "eval_runtime": 2930.2464,
+      "eval_samples_per_second": 4.178,
+      "eval_steps_per_second": 0.066,
+      "step": 1700
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.303236208683635e-05,
+      "loss": 0.3883,
+      "step": 1710
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 4.297477830243004e-05,
+      "loss": 0.3621,
+      "step": 1720
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.291719451802373e-05,
+      "loss": 0.3832,
+      "step": 1730
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.285961073361742e-05,
+      "loss": 0.3807,
+      "step": 1740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.2802026949211107e-05,
+      "loss": 0.4715,
+      "step": 1750
+    },
+    {
+      "epoch": 0.57,
+      "eval_accuracy": 0.8703749179840088,
+      "eval_loss": 0.3262869417667389,
+      "eval_runtime": 2984.533,
+      "eval_samples_per_second": 4.102,
+      "eval_steps_per_second": 0.064,
+      "step": 1750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 4.274444316480479e-05,
+      "loss": 0.4,
+      "step": 1760
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.268685938039848e-05,
+      "loss": 0.3577,
+      "step": 1770
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.262927559599217e-05,
+      "loss": 0.3346,
+      "step": 1780
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 4.2571691811585865e-05,
+      "loss": 0.4483,
+      "step": 1790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.251410802717955e-05,
+      "loss": 0.3575,
+      "step": 1800
+    },
+    {
+      "epoch": 0.59,
+      "eval_accuracy": 0.8698031306266785,
+      "eval_loss": 0.35289597511291504,
+      "eval_runtime": 2934.3355,
+      "eval_samples_per_second": 4.172,
+      "eval_steps_per_second": 0.065,
+      "step": 1800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.245652424277324e-05,
+      "loss": 0.3892,
+      "step": 1810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 4.239894045836693e-05,
+      "loss": 0.3594,
+      "step": 1820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.2341356673960617e-05,
+      "loss": 0.3949,
+      "step": 1830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.22837728895543e-05,
+      "loss": 0.4108,
+      "step": 1840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 4.222618910514799e-05,
+      "loss": 0.3551,
+      "step": 1850
+    },
+    {
+      "epoch": 0.6,
+      "eval_accuracy": 0.8766642212867737,
+      "eval_loss": 0.31740882992744446,
+      "eval_runtime": 2928.4953,
+      "eval_samples_per_second": 4.181,
+      "eval_steps_per_second": 0.066,
+      "step": 1850
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.2168605320741685e-05,
+      "loss": 0.3809,
+      "step": 1860
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.211102153633537e-05,
+      "loss": 0.4427,
+      "step": 1870
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 4.205343775192906e-05,
+      "loss": 0.2968,
+      "step": 1880
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.199585396752275e-05,
+      "loss": 0.5083,
+      "step": 1890
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.193827018311644e-05,
+      "loss": 0.3751,
+      "step": 1900
+    },
+    {
+      "epoch": 0.62,
+      "eval_accuracy": 0.8611451387405396,
+      "eval_loss": 0.39074617624282837,
+      "eval_runtime": 2974.5471,
+      "eval_samples_per_second": 4.116,
+      "eval_steps_per_second": 0.065,
+      "step": 1900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 4.188068639871012e-05,
+      "loss": 0.3789,
+      "step": 1910
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.1823102614303816e-05,
+      "loss": 0.3044,
+      "step": 1920
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.1765518829897506e-05,
+      "loss": 0.4,
+      "step": 1930
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 4.1707935045491195e-05,
+      "loss": 0.5459,
+      "step": 1940
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.165035126108488e-05,
+      "loss": 0.3304,
+      "step": 1950
+    },
+    {
+      "epoch": 0.64,
+      "eval_accuracy": 0.871273398399353,
+      "eval_loss": 0.3362810015678406,
+      "eval_runtime": 3072.5062,
+      "eval_samples_per_second": 3.985,
+      "eval_steps_per_second": 0.062,
+      "step": 1950
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.159276747667857e-05,
+      "loss": 0.2947,
+      "step": 1960
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 4.153518369227226e-05,
+      "loss": 0.3824,
+      "step": 1970
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.147759990786595e-05,
+      "loss": 0.4153,
+      "step": 1980
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.1420016123459636e-05,
+      "loss": 0.473,
+      "step": 1990
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 4.1362432339053326e-05,
+      "loss": 0.4142,
+      "step": 2000
+    },
+    {
+      "epoch": 0.65,
+      "eval_accuracy": 0.8678428530693054,
+      "eval_loss": 0.3490336239337921,
+      "eval_runtime": 3028.5373,
+      "eval_samples_per_second": 4.043,
+      "eval_steps_per_second": 0.063,
+      "step": 2000
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.1304848554647016e-05,
+      "loss": 0.2809,
+      "step": 2010
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.12472647702407e-05,
+      "loss": 0.4147,
+      "step": 2020
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 4.118968098583439e-05,
+      "loss": 0.3299,
+      "step": 2030
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.1132097201428084e-05,
+      "loss": 0.3565,
+      "step": 2040
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.107451341702177e-05,
+      "loss": 0.3607,
+      "step": 2050
+    },
+    {
+      "epoch": 0.67,
+      "eval_accuracy": 0.8337008953094482,
+      "eval_loss": 0.440717488527298,
+      "eval_runtime": 2955.0057,
+      "eval_samples_per_second": 4.143,
+      "eval_steps_per_second": 0.065,
+      "step": 2050
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 4.101692963261546e-05,
+      "loss": 0.3997,
+      "step": 2060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.0959345848209146e-05,
+      "loss": 0.4368,
+      "step": 2070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.0901762063802836e-05,
+      "loss": 0.3253,
+      "step": 2080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.084417827939652e-05,
+      "loss": 0.3693,
+      "step": 2090
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.078659449499021e-05,
+      "loss": 0.3482,
+      "step": 2100
+    },
+    {
+      "epoch": 0.69,
+      "eval_accuracy": 0.8706199526786804,
+      "eval_loss": 0.31695759296417236,
+      "eval_runtime": 2927.7031,
+      "eval_samples_per_second": 4.182,
+      "eval_steps_per_second": 0.066,
+      "step": 2100
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.0729010710583905e-05,
+      "loss": 0.3172,
+      "step": 2110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.0671426926177594e-05,
+      "loss": 0.3125,
+      "step": 2120
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.061384314177128e-05,
+      "loss": 0.5427,
+      "step": 2130
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.055625935736497e-05,
+      "loss": 0.2956,
+      "step": 2140
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.0498675572958656e-05,
+      "loss": 0.4052,
+      "step": 2150
+    },
+    {
+      "epoch": 0.7,
+      "eval_accuracy": 0.8591031432151794,
+      "eval_loss": 0.40359142422676086,
+      "eval_runtime": 3099.542,
+      "eval_samples_per_second": 3.95,
+      "eval_steps_per_second": 0.062,
+      "step": 2150
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0441091788552346e-05,
+      "loss": 0.4393,
+      "step": 2160
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0383508004146036e-05,
+      "loss": 0.3817,
+      "step": 2170
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0325924219739725e-05,
+      "loss": 0.3882,
+      "step": 2180
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.0268340435333415e-05,
+      "loss": 0.321,
+      "step": 2190
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.02107566509271e-05,
+      "loss": 0.3144,
+      "step": 2200
+    },
+    {
+      "epoch": 0.72,
+      "eval_accuracy": 0.8765008449554443,
+      "eval_loss": 0.33840280771255493,
+      "eval_runtime": 3038.9712,
+      "eval_samples_per_second": 4.029,
+      "eval_steps_per_second": 0.063,
+      "step": 2200
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 4.015317286652079e-05,
+      "loss": 0.4218,
+      "step": 2210
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.009558908211448e-05,
+      "loss": 0.3699,
+      "step": 2220
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 4.003800529770817e-05,
+      "loss": 0.4581,
+      "step": 2230
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.9980421513301856e-05,
+      "loss": 0.3644,
+      "step": 2240
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.9922837728895546e-05,
+      "loss": 0.3325,
+      "step": 2250
+    },
+    {
+      "epoch": 0.74,
+      "eval_accuracy": 0.8622886538505554,
+      "eval_loss": 0.3353598713874817,
+      "eval_runtime": 3031.1476,
+      "eval_samples_per_second": 4.039,
+      "eval_steps_per_second": 0.063,
+      "step": 2250
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.9865253944489235e-05,
+      "loss": 0.3336,
+      "step": 2260
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.9807670160082925e-05,
+      "loss": 0.3224,
+      "step": 2270
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.975008637567661e-05,
+      "loss": 0.2383,
+      "step": 2280
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.96925025912703e-05,
+      "loss": 0.4136,
+      "step": 2290
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.9634918806863994e-05,
+      "loss": 0.3907,
+      "step": 2300
+    },
+    {
+      "epoch": 0.75,
+      "eval_accuracy": 0.859756588935852,
+      "eval_loss": 0.3931679129600525,
+      "eval_runtime": 3070.5805,
+      "eval_samples_per_second": 3.987,
+      "eval_steps_per_second": 0.063,
+      "step": 2300
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.9577335022457676e-05,
+      "loss": 0.3771,
+      "step": 2310
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9519751238051366e-05,
+      "loss": 0.2548,
+      "step": 2320
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9462167453645056e-05,
+      "loss": 0.4178,
+      "step": 2330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 3.9404583669238745e-05,
+      "loss": 0.248,
+      "step": 2340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.934699988483243e-05,
+      "loss": 0.3951,
+      "step": 2350
+    },
+    {
+      "epoch": 0.77,
+      "eval_accuracy": 0.8809115290641785,
+      "eval_loss": 0.3367299735546112,
+      "eval_runtime": 3007.758,
+      "eval_samples_per_second": 4.07,
+      "eval_steps_per_second": 0.064,
+      "step": 2350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9289416100426124e-05,
+      "loss": 0.5165,
+      "step": 2360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 3.9231832316019814e-05,
+      "loss": 0.3466,
+      "step": 2370
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.91742485316135e-05,
+      "loss": 0.2617,
+      "step": 2380
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.9116664747207186e-05,
+      "loss": 0.2396,
+      "step": 2390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 3.9059080962800876e-05,
+      "loss": 0.4248,
+      "step": 2400
+    },
+    {
+      "epoch": 0.78,
+      "eval_accuracy": 0.8034795522689819,
+      "eval_loss": 0.5644595623016357,
+      "eval_runtime": 2905.2597,
+      "eval_samples_per_second": 4.214,
+      "eval_steps_per_second": 0.066,
+      "step": 2400
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.9001497178394565e-05,
+      "loss": 0.5326,
+      "step": 2410
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.8943913393988255e-05,
+      "loss": 0.3532,
+      "step": 2420
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 3.8886329609581945e-05,
+      "loss": 0.2889,
+      "step": 2430
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.8828745825175634e-05,
+      "loss": 0.3974,
+      "step": 2440
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.8771162040769324e-05,
+      "loss": 0.3762,
+      "step": 2450
+    },
+    {
+      "epoch": 0.8,
+      "eval_accuracy": 0.8773993253707886,
+      "eval_loss": 0.34061020612716675,
+      "eval_runtime": 2912.8057,
+      "eval_samples_per_second": 4.203,
+      "eval_steps_per_second": 0.066,
+      "step": 2450
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 3.871357825636301e-05,
+      "loss": 0.3972,
+      "step": 2460
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.8655994471956696e-05,
+      "loss": 0.3076,
+      "step": 2470
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.859841068755039e-05,
+      "loss": 0.4305,
+      "step": 2480
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 3.8540826903144075e-05,
+      "loss": 0.3924,
+      "step": 2490
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8483243118737765e-05,
+      "loss": 0.3243,
+      "step": 2500
+    },
+    {
+      "epoch": 0.82,
+      "eval_accuracy": 0.8704565763473511,
+      "eval_loss": 0.34552812576293945,
+      "eval_runtime": 2909.9177,
+      "eval_samples_per_second": 4.207,
+      "eval_steps_per_second": 0.066,
+      "step": 2500
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8425659334331455e-05,
+      "loss": 0.3736,
+      "step": 2510
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 3.8368075549925144e-05,
+      "loss": 0.3722,
+      "step": 2520
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.831049176551883e-05,
+      "loss": 0.3887,
+      "step": 2530
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.825290798111252e-05,
+      "loss": 0.2656,
+      "step": 2540
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 3.819532419670621e-05,
+      "loss": 0.3665,
+      "step": 2550
+    },
+    {
+      "epoch": 0.83,
+      "eval_accuracy": 0.8765008449554443,
+      "eval_loss": 0.3324890732765198,
+      "eval_runtime": 2898.9921,
+      "eval_samples_per_second": 4.223,
+      "eval_steps_per_second": 0.066,
+      "step": 2550
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.81377404122999e-05,
+      "loss": 0.3224,
+      "step": 2560
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.8080156627893585e-05,
+      "loss": 0.3492,
+      "step": 2570
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 3.8022572843487275e-05,
+      "loss": 0.2999,
+      "step": 2580
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.7964989059080965e-05,
+      "loss": 0.315,
+      "step": 2590
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.7907405274674654e-05,
+      "loss": 0.3133,
+      "step": 2600
+    },
+    {
+      "epoch": 0.85,
+      "eval_accuracy": 0.8835252523422241,
+      "eval_loss": 0.3392607569694519,
+      "eval_runtime": 2888.7867,
+      "eval_samples_per_second": 4.238,
+      "eval_steps_per_second": 0.066,
+      "step": 2600
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 3.7849821490268344e-05,
+      "loss": 0.4609,
+      "step": 2610
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.779223770586203e-05,
+      "loss": 0.3426,
+      "step": 2620
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.773465392145572e-05,
+      "loss": 0.4421,
+      "step": 2630
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 3.7677070137049406e-05,
+      "loss": 0.3832,
+      "step": 2640
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 3.7619486352643095e-05,
+      "loss": 0.3089,
+      "step": 2650
+    },
+    {
+      "epoch": 0.87,
+      "eval_accuracy": 0.8714367151260376,
+      "eval_loss": 0.35768425464630127,
+      "eval_runtime": 2888.0542,
+      "eval_samples_per_second": 4.239,
+      "eval_steps_per_second": 0.066,
+      "step": 2650
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 3.7561902568236785e-05,
+      "loss": 0.3473,
+      "step": 2660
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 3.7504318783830475e-05,
+      "loss": 0.3031,
+      "step": 2670
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7446734999424164e-05,
+      "loss": 0.3098,
+      "step": 2680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.7389151215017854e-05,
+      "loss": 0.3391,
+      "step": 2690
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 3.733156743061154e-05,
+      "loss": 0.4607,
+      "step": 2700
+    },
+    {
+      "epoch": 0.88,
+      "eval_accuracy": 0.8618802428245544,
+      "eval_loss": 0.382795125246048,
+      "eval_runtime": 2882.6679,
+      "eval_samples_per_second": 4.247,
+      "eval_steps_per_second": 0.067,
+      "step": 2700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.7273983646205226e-05,
+      "loss": 0.4121,
+      "step": 2710
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.7216399861798916e-05,
+      "loss": 0.3283,
+      "step": 2720
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 3.7158816077392605e-05,
+      "loss": 0.3795,
+      "step": 2730
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.71012322929863e-05,
+      "loss": 0.3245,
+      "step": 2740
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.7043648508579985e-05,
+      "loss": 0.3135,
+      "step": 2750
+    },
+    {
+      "epoch": 0.9,
+      "eval_accuracy": 0.8852405548095703,
+      "eval_loss": 0.2938101291656494,
+      "eval_runtime": 2869.0497,
+      "eval_samples_per_second": 4.267,
+      "eval_steps_per_second": 0.067,
+      "step": 2750
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.6986064724173674e-05,
+      "loss": 0.3218,
+      "step": 2760
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 3.6928480939767364e-05,
+      "loss": 0.2949,
+      "step": 2770
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.687089715536105e-05,
+      "loss": 0.2688,
+      "step": 2780
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.6813313370954736e-05,
+      "loss": 0.3828,
+      "step": 2790
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 3.675572958654843e-05,
+      "loss": 0.3256,
+      "step": 2800
+    },
+    {
+      "epoch": 0.91,
+      "eval_accuracy": 0.882381796836853,
+      "eval_loss": 0.2991408407688141,
+      "eval_runtime": 2911.5784,
+      "eval_samples_per_second": 4.205,
+      "eval_steps_per_second": 0.066,
+      "step": 2800
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.669814580214212e-05,
+      "loss": 0.4311,
+      "step": 2810
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6640562017735805e-05,
+      "loss": 0.3457,
+      "step": 2820
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 3.6582978233329494e-05,
+      "loss": 0.3137,
+      "step": 2830
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6525394448923184e-05,
+      "loss": 0.2957,
+      "step": 2840
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.6467810664516874e-05,
+      "loss": 0.322,
+      "step": 2850
+    },
+    {
+      "epoch": 0.93,
+      "eval_accuracy": 0.8836886286735535,
+      "eval_loss": 0.2957231402397156,
+      "eval_runtime": 3000.5532,
+      "eval_samples_per_second": 4.08,
+      "eval_steps_per_second": 0.064,
+      "step": 2850
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 3.641022688011056e-05,
+      "loss": 0.3722,
+      "step": 2860
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.635264309570425e-05,
+      "loss": 0.3227,
+      "step": 2870
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.629505931129794e-05,
+      "loss": 0.3077,
+      "step": 2880
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 3.623747552689163e-05,
+      "loss": 0.3331,
+      "step": 2890
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.6179891742485315e-05,
+      "loss": 0.2935,
+      "step": 2900
+    },
+    {
+      "epoch": 0.95,
+      "eval_accuracy": 0.8792779445648193,
+      "eval_loss": 0.32762154936790466,
+      "eval_runtime": 2999.2249,
+      "eval_samples_per_second": 4.082,
+      "eval_steps_per_second": 0.064,
+      "step": 2900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.6122307958079004e-05,
+      "loss": 0.4533,
+      "step": 2910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 3.60647241736727e-05,
+      "loss": 0.3215,
+      "step": 2920
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.6007140389266384e-05,
+      "loss": 0.2866,
+      "step": 2930
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.594955660486007e-05,
+      "loss": 0.3395,
+      "step": 2940
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 3.589197282045376e-05,
+      "loss": 0.3111,
+      "step": 2950
+    },
+    {
+      "epoch": 0.96,
+      "eval_accuracy": 0.8871191740036011,
+      "eval_loss": 0.2872740924358368,
+      "eval_runtime": 3014.0668,
+      "eval_samples_per_second": 4.062,
+      "eval_steps_per_second": 0.064,
+      "step": 2950
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.583438903604745e-05,
+      "loss": 0.2482,
+      "step": 2960
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5776805251641135e-05,
+      "loss": 0.3019,
+      "step": 2970
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 3.5719221467234825e-05,
+      "loss": 0.3043,
+      "step": 2980
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.566163768282852e-05,
+      "loss": 0.3144,
+      "step": 2990
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.5604053898422204e-05,
+      "loss": 0.2705,
+      "step": 3000
+    },
+    {
+      "epoch": 0.98,
+      "eval_accuracy": 0.8801764249801636,
+      "eval_loss": 0.31194931268692017,
+      "eval_runtime": 2990.6485,
+      "eval_samples_per_second": 4.094,
+      "eval_steps_per_second": 0.064,
+      "step": 3000
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 3.5546470114015894e-05,
+      "loss": 0.4068,
+      "step": 3010
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.548888632960958e-05,
+      "loss": 0.2519,
+      "step": 3020
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.543130254520327e-05,
+      "loss": 0.2913,
+      "step": 3030
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 3.5373718760796956e-05,
+      "loss": 0.3195,
+      "step": 3040
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.531613497639065e-05,
+      "loss": 0.1961,
+      "step": 3050
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.8861390352249146,
+      "eval_loss": 0.31677114963531494,
+      "eval_runtime": 3046.4306,
+      "eval_samples_per_second": 4.019,
+      "eval_steps_per_second": 0.063,
+      "step": 3050
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.525855119198434e-05,
+      "loss": 0.3411,
+      "step": 3060
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 3.520096740757803e-05,
+      "loss": 0.2577,
+      "step": 3070
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 3.5143383623171714e-05,
+      "loss": 0.2633,
+      "step": 3080
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 3.5085799838765404e-05,
+      "loss": 0.3464,
+      "step": 3090
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 3.502821605435909e-05,
+      "loss": 0.2509,
+      "step": 3100
+    },
+    {
+      "epoch": 1.01,
+      "eval_accuracy": 0.8817283511161804,
+      "eval_loss": 0.3493081033229828,
+      "eval_runtime": 2940.1695,
+      "eval_samples_per_second": 4.164,
+      "eval_steps_per_second": 0.065,
+      "step": 3100
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 3.497063226995278e-05,
+      "loss": 0.2918,
+      "step": 3110
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 3.491304848554647e-05,
+      "loss": 0.2767,
+      "step": 3120
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 3.485546470114016e-05,
+      "loss": 0.3438,
+      "step": 3130
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 3.479788091673385e-05,
+      "loss": 0.2656,
+      "step": 3140
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 3.4740297132327534e-05,
+      "loss": 0.353,
+      "step": 3150
+    },
+    {
+      "epoch": 1.03,
+      "eval_accuracy": 0.876745879650116,
+      "eval_loss": 0.35529011487960815,
+      "eval_runtime": 2957.9472,
+      "eval_samples_per_second": 4.139,
+      "eval_steps_per_second": 0.065,
+      "step": 3150
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 3.4682713347921224e-05,
+      "loss": 0.3958,
+      "step": 3160
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 3.4625129563514914e-05,
+      "loss": 0.2825,
+      "step": 3170
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 3.456754577910861e-05,
+      "loss": 0.3076,
+      "step": 3180
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 3.450996199470229e-05,
+      "loss": 0.3207,
+      "step": 3190
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 3.445237821029598e-05,
+      "loss": 0.3193,
+      "step": 3200
+    },
+    {
+      "epoch": 1.05,
+      "eval_accuracy": 0.8861390352249146,
+      "eval_loss": 0.3049357235431671,
+      "eval_runtime": 2946.4042,
+      "eval_samples_per_second": 4.155,
+      "eval_steps_per_second": 0.065,
+      "step": 3200
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 3.439479442588967e-05,
+      "loss": 0.3157,
+      "step": 3210
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 3.433721064148336e-05,
+      "loss": 0.2552,
+      "step": 3220
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 3.4279626857077044e-05,
+      "loss": 0.1984,
+      "step": 3230
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 3.422204307267074e-05,
+      "loss": 0.2895,
+      "step": 3240
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 3.416445928826443e-05,
+      "loss": 0.3835,
+      "step": 3250
+    },
+    {
+      "epoch": 1.06,
+      "eval_accuracy": 0.8836069703102112,
+      "eval_loss": 0.31986963748931885,
+      "eval_runtime": 2977.2171,
+      "eval_samples_per_second": 4.112,
+      "eval_steps_per_second": 0.064,
+      "step": 3250
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 3.410687550385811e-05,
+      "loss": 0.1779,
+      "step": 3260
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 3.40492917194518e-05,
+      "loss": 0.201,
+      "step": 3270
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 3.399170793504549e-05,
+      "loss": 0.2527,
+      "step": 3280
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 3.393412415063918e-05,
+      "loss": 0.4082,
+      "step": 3290
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 3.387654036623287e-05,
+      "loss": 0.1875,
+      "step": 3300
+    },
+    {
+      "epoch": 1.08,
+      "eval_accuracy": 0.8863023519515991,
+      "eval_loss": 0.3436829149723053,
+      "eval_runtime": 3037.4605,
+      "eval_samples_per_second": 4.031,
+      "eval_steps_per_second": 0.063,
+      "step": 3300
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 3.381895658182656e-05,
+      "loss": 0.2471,
+      "step": 3310
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 3.376137279742025e-05,
+      "loss": 0.2886,
+      "step": 3320
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 3.3703789013013933e-05,
+      "loss": 0.2543,
+      "step": 3330
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 3.364620522860762e-05,
+      "loss": 0.3081,
+      "step": 3340
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 3.358862144420131e-05,
+      "loss": 0.3401,
+      "step": 3350
+    },
+    {
+      "epoch": 1.09,
+      "eval_accuracy": 0.8604916930198669,
+      "eval_loss": 0.33765077590942383,
+      "eval_runtime": 3045.8041,
+      "eval_samples_per_second": 4.02,
+      "eval_steps_per_second": 0.063,
+      "step": 3350
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 3.353103765979501e-05,
+      "loss": 0.2812,
+      "step": 3360
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 3.347345387538869e-05,
+      "loss": 0.3059,
+      "step": 3370
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 3.341587009098238e-05,
+      "loss": 0.1746,
+      "step": 3380
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 3.335828630657607e-05,
+      "loss": 0.3549,
+      "step": 3390
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 3.330070252216976e-05,
+      "loss": 0.2638,
+      "step": 3400
+    },
+    {
+      "epoch": 1.11,
+      "eval_accuracy": 0.8876092433929443,
+      "eval_loss": 0.3165830075740814,
+      "eval_runtime": 3077.5964,
+      "eval_samples_per_second": 3.978,
+      "eval_steps_per_second": 0.062,
+      "step": 3400
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 3.3243118737763443e-05,
+      "loss": 0.2538,
+      "step": 3410
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 3.318553495335713e-05,
+      "loss": 0.2884,
+      "step": 3420
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 3.312795116895083e-05,
+      "loss": 0.2238,
+      "step": 3430
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 3.307036738454451e-05,
+      "loss": 0.2464,
+      "step": 3440
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 3.30127836001382e-05,
+      "loss": 0.2292,
+      "step": 3450
+    },
+    {
+      "epoch": 1.13,
+      "eval_accuracy": 0.8859756588935852,
+      "eval_loss": 0.3086280822753906,
+      "eval_runtime": 3051.2098,
+      "eval_samples_per_second": 4.013,
+      "eval_steps_per_second": 0.063,
+      "step": 3450
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 3.295519981573189e-05,
+      "loss": 0.2235,
+      "step": 3460
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 3.289761603132558e-05,
+      "loss": 0.2191,
+      "step": 3470
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 3.2840032246919264e-05,
+      "loss": 0.2011,
+      "step": 3480
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 3.278244846251296e-05,
+      "loss": 0.2728,
+      "step": 3490
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 3.272486467810665e-05,
+      "loss": 0.3802,
+      "step": 3500
+    },
+    {
+      "epoch": 1.14,
+      "eval_accuracy": 0.8824634552001953,
+      "eval_loss": 0.34480205178260803,
+      "eval_runtime": 2914.7217,
+      "eval_samples_per_second": 4.2,
+      "eval_steps_per_second": 0.066,
+      "step": 3500
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 3.266728089370034e-05,
+      "loss": 0.2859,
+      "step": 3510
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 3.260969710929402e-05,
+      "loss": 0.3239,
+      "step": 3520
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 3.255211332488771e-05,
+      "loss": 0.285,
+      "step": 3530
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 3.24945295404814e-05,
+      "loss": 0.2368,
+      "step": 3540
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 3.243694575607509e-05,
+      "loss": 0.3109,
+      "step": 3550
+    },
+    {
+      "epoch": 1.16,
+      "eval_accuracy": 0.8840153813362122,
+      "eval_loss": 0.3186437487602234,
+      "eval_runtime": 2887.6289,
+      "eval_samples_per_second": 4.24,
+      "eval_steps_per_second": 0.066,
+      "step": 3550
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 3.237936197166878e-05,
+      "loss": 0.2621,
+      "step": 3560
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 3.232177818726247e-05,
+      "loss": 0.3249,
+      "step": 3570
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 3.226419440285616e-05,
+      "loss": 0.2032,
+      "step": 3580
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 3.220661061844984e-05,
+      "loss": 0.3035,
+      "step": 3590
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 3.214902683404353e-05,
+      "loss": 0.1687,
+      "step": 3600
+    },
+    {
+      "epoch": 1.18,
+      "eval_accuracy": 0.879196286201477,
+      "eval_loss": 0.3766389787197113,
+      "eval_runtime": 2934.3858,
+      "eval_samples_per_second": 4.172,
+      "eval_steps_per_second": 0.065,
+      "step": 3600
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 3.209144304963722e-05,
+      "loss": 0.2218,
+      "step": 3610
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 3.203385926523091e-05,
+      "loss": 0.3318,
+      "step": 3620
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 3.19762754808246e-05,
+      "loss": 0.2759,
+      "step": 3630
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 3.191869169641829e-05,
+      "loss": 0.2841,
+      "step": 3640
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 3.186110791201198e-05,
+      "loss": 0.4197,
+      "step": 3650
+    },
+    {
+      "epoch": 1.19,
+      "eval_accuracy": 0.8806664943695068,
+      "eval_loss": 0.35255807638168335,
+      "eval_runtime": 2885.4305,
+      "eval_samples_per_second": 4.243,
+      "eval_steps_per_second": 0.067,
+      "step": 3650
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.180352412760566e-05,
+      "loss": 0.3488,
+      "step": 3660
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.174594034319935e-05,
+      "loss": 0.2732,
+      "step": 3670
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 3.168835655879305e-05,
+      "loss": 0.2353,
+      "step": 3680
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.163077277438674e-05,
+      "loss": 0.284,
+      "step": 3690
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.157318898998042e-05,
+      "loss": 0.4672,
+      "step": 3700
+    },
+    {
+      "epoch": 1.21,
+      "eval_accuracy": 0.8781344294548035,
+      "eval_loss": 0.3553983271121979,
+      "eval_runtime": 2899.2733,
+      "eval_samples_per_second": 4.223,
+      "eval_steps_per_second": 0.066,
+      "step": 3700
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 3.151560520557411e-05,
+      "loss": 0.3399,
+      "step": 3710
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.14580214211678e-05,
+      "loss": 0.1954,
+      "step": 3720
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.140043763676149e-05,
+      "loss": 0.2319,
+      "step": 3730
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 3.134285385235518e-05,
+      "loss": 0.2601,
+      "step": 3740
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.128527006794887e-05,
+      "loss": 0.2089,
+      "step": 3750
+    },
+    {
+      "epoch": 1.23,
+      "eval_accuracy": 0.8892428278923035,
+      "eval_loss": 0.34623658657073975,
+      "eval_runtime": 2894.5359,
+      "eval_samples_per_second": 4.23,
+      "eval_steps_per_second": 0.066,
+      "step": 3750
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.122768628354256e-05,
+      "loss": 0.3792,
+      "step": 3760
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.117010249913624e-05,
+      "loss": 0.2881,
+      "step": 3770
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 3.111251871472993e-05,
+      "loss": 0.2984,
+      "step": 3780
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.105493493032362e-05,
+      "loss": 0.2787,
+      "step": 3790
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.099735114591731e-05,
+      "loss": 0.2619,
+      "step": 3800
+    },
+    {
+      "epoch": 1.24,
+      "eval_accuracy": 0.8899779319763184,
+      "eval_loss": 0.27432918548583984,
+      "eval_runtime": 2916.9308,
+      "eval_samples_per_second": 4.197,
+      "eval_steps_per_second": 0.066,
+      "step": 3800
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 3.0939767361511e-05,
+      "loss": 0.2945,
+      "step": 3810
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.088218357710469e-05,
+      "loss": 0.2507,
+      "step": 3820
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.082459979269838e-05,
+      "loss": 0.2165,
+      "step": 3830
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 3.076701600829207e-05,
+      "loss": 0.3021,
+      "step": 3840
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.070943222388575e-05,
+      "loss": 0.2475,
+      "step": 3850
+    },
+    {
+      "epoch": 1.26,
+      "eval_accuracy": 0.8839336633682251,
+      "eval_loss": 0.33079230785369873,
+      "eval_runtime": 2924.5803,
+      "eval_samples_per_second": 4.186,
+      "eval_steps_per_second": 0.066,
+      "step": 3850
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.065184843947944e-05,
+      "loss": 0.4076,
+      "step": 3860
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 3.059426465507314e-05,
+      "loss": 0.3001,
+      "step": 3870
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.053668087066682e-05,
+      "loss": 0.3737,
+      "step": 3880
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.047909708626051e-05,
+      "loss": 0.3211,
+      "step": 3890
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 3.04215133018542e-05,
+      "loss": 0.2555,
+      "step": 3900
+    },
+    {
+      "epoch": 1.27,
+      "eval_accuracy": 0.8938985466957092,
+      "eval_loss": 0.2678125202655792,
+      "eval_runtime": 3020.8478,
+      "eval_samples_per_second": 4.053,
+      "eval_steps_per_second": 0.064,
+      "step": 3900
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.0363929517447886e-05,
+      "loss": 0.2218,
+      "step": 3910
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.0306345733041575e-05,
+      "loss": 0.2281,
+      "step": 3920
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 3.024876194863527e-05,
+      "loss": 0.2133,
+      "step": 3930
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.0191178164228955e-05,
+      "loss": 0.3718,
+      "step": 3940
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.0133594379822644e-05,
+      "loss": 0.3422,
+      "step": 3950
+    },
+    {
+      "epoch": 1.29,
+      "eval_accuracy": 0.8835252523422241,
+      "eval_loss": 0.3514479696750641,
+      "eval_runtime": 3121.751,
+      "eval_samples_per_second": 3.922,
+      "eval_steps_per_second": 0.062,
+      "step": 3950
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 3.007601059541633e-05,
+      "loss": 0.2236,
+      "step": 3960
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 3.001842681101002e-05,
+      "loss": 0.3787,
+      "step": 3970
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.9960843026603706e-05,
+      "loss": 0.3707,
+      "step": 3980
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2.9903259242197403e-05,
+      "loss": 0.2872,
+      "step": 3990
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.984567545779109e-05,
+      "loss": 0.4241,
+      "step": 4000
+    },
+    {
+      "epoch": 1.31,
+      "eval_accuracy": 0.8759291172027588,
+      "eval_loss": 0.3576194643974304,
+      "eval_runtime": 3025.1279,
+      "eval_samples_per_second": 4.047,
+      "eval_steps_per_second": 0.063,
+      "step": 4000
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.978809167338478e-05,
+      "loss": 0.2719,
+      "step": 4010
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2.9730507888978465e-05,
+      "loss": 0.3052,
+      "step": 4020
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.9672924104572154e-05,
+      "loss": 0.3074,
+      "step": 4030
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.961534032016584e-05,
+      "loss": 0.2776,
+      "step": 4040
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2.9557756535759533e-05,
+      "loss": 0.2679,
+      "step": 4050
+    },
+    {
+      "epoch": 1.32,
+      "eval_accuracy": 0.892591655254364,
+      "eval_loss": 0.28911352157592773,
+      "eval_runtime": 3026.0398,
+      "eval_samples_per_second": 4.046,
+      "eval_steps_per_second": 0.063,
+      "step": 4050
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.9500172751353223e-05,
+      "loss": 0.2663,
+      "step": 4060
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.944258896694691e-05,
+      "loss": 0.2623,
+      "step": 4070
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2.93850051825406e-05,
+      "loss": 0.2172,
+      "step": 4080
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.9327421398134285e-05,
+      "loss": 0.2142,
+      "step": 4090
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.9269837613727974e-05,
+      "loss": 0.2927,
+      "step": 4100
+    },
+    {
+      "epoch": 1.34,
+      "eval_accuracy": 0.8939802050590515,
+      "eval_loss": 0.33391907811164856,
+      "eval_runtime": 3047.4975,
+      "eval_samples_per_second": 4.017,
+      "eval_steps_per_second": 0.063,
+      "step": 4100
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2.921225382932166e-05,
+      "loss": 0.1962,
+      "step": 4110
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.9154670044915354e-05,
+      "loss": 0.3582,
+      "step": 4120
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.9097086260509043e-05,
+      "loss": 0.5313,
+      "step": 4130
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2.9039502476102733e-05,
+      "loss": 0.2979,
+      "step": 4140
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.898191869169642e-05,
+      "loss": 0.1658,
+      "step": 4150
+    },
+    {
+      "epoch": 1.36,
+      "eval_accuracy": 0.8950420618057251,
+      "eval_loss": 0.3208805024623871,
+      "eval_runtime": 3049.3637,
+      "eval_samples_per_second": 4.015,
+      "eval_steps_per_second": 0.063,
+      "step": 4150
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.892433490729011e-05,
+      "loss": 0.3613,
+      "step": 4160
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2.8866751122883795e-05,
+      "loss": 0.229,
+      "step": 4170
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.8809167338477488e-05,
+      "loss": 0.2697,
+      "step": 4180
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.8751583554071177e-05,
+      "loss": 0.2549,
+      "step": 4190
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2.8693999769664864e-05,
+      "loss": 0.2813,
+      "step": 4200
+    },
+    {
+      "epoch": 1.37,
+      "eval_accuracy": 0.8891611695289612,
+      "eval_loss": 0.3334508538246155,
+      "eval_runtime": 2959.4164,
+      "eval_samples_per_second": 4.137,
+      "eval_steps_per_second": 0.065,
+      "step": 4200
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.8636415985258553e-05,
+      "loss": 0.2688,
+      "step": 4210
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.857883220085224e-05,
+      "loss": 0.2659,
+      "step": 4220
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2.852124841644593e-05,
+      "loss": 0.2095,
+      "step": 4230
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.8463664632039622e-05,
+      "loss": 0.2683,
+      "step": 4240
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.8406080847633308e-05,
+      "loss": 0.2146,
+      "step": 4250
+    },
+    {
+      "epoch": 1.39,
+      "eval_accuracy": 0.8797680139541626,
+      "eval_loss": 0.3393837809562683,
+      "eval_runtime": 2895.1322,
+      "eval_samples_per_second": 4.229,
+      "eval_steps_per_second": 0.066,
+      "step": 4250
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.8348497063226998e-05,
+      "loss": 0.1797,
+      "step": 4260
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2.8290913278820684e-05,
+      "loss": 0.2529,
+      "step": 4270
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.8233329494414374e-05,
+      "loss": 0.237,
+      "step": 4280
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.817574571000806e-05,
+      "loss": 0.1763,
+      "step": 4290
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2.811816192560175e-05,
+      "loss": 0.3035,
+      "step": 4300
+    },
+    {
+      "epoch": 1.4,
+      "eval_accuracy": 0.8946336507797241,
+      "eval_loss": 0.300825297832489,
+      "eval_runtime": 2943.6237,
+      "eval_samples_per_second": 4.159,
+      "eval_steps_per_second": 0.065,
+      "step": 4300
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.8060578141195442e-05,
+      "loss": 0.2886,
+      "step": 4310
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.8002994356789132e-05,
+      "loss": 0.2885,
+      "step": 4320
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2.7945410572382818e-05,
+      "loss": 0.1489,
+      "step": 4330
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7887826787976508e-05,
+      "loss": 0.3775,
+      "step": 4340
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7830243003570194e-05,
+      "loss": 0.2093,
+      "step": 4350
+    },
+    {
+      "epoch": 1.42,
+      "eval_accuracy": 0.8903863430023193,
+      "eval_loss": 0.3052256405353546,
+      "eval_runtime": 2954.2408,
+      "eval_samples_per_second": 4.144,
+      "eval_steps_per_second": 0.065,
+      "step": 4350
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2.7772659219163884e-05,
+      "loss": 0.3467,
+      "step": 4360
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.7715075434757577e-05,
+      "loss": 0.1922,
+      "step": 4370
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.7657491650351263e-05,
+      "loss": 0.1993,
+      "step": 4380
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2.7599907865944952e-05,
+      "loss": 0.2085,
+      "step": 4390
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.754232408153864e-05,
+      "loss": 0.2988,
+      "step": 4400
+    },
+    {
+      "epoch": 1.44,
+      "eval_accuracy": 0.8940619230270386,
+      "eval_loss": 0.3122096061706543,
+      "eval_runtime": 3120.6357,
+      "eval_samples_per_second": 3.923,
+      "eval_steps_per_second": 0.062,
+      "step": 4400
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.7484740297132328e-05,
+      "loss": 0.3208,
+      "step": 4410
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2.7427156512726014e-05,
+      "loss": 0.1466,
+      "step": 4420
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7369572728319707e-05,
+      "loss": 0.4314,
+      "step": 4430
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7311988943913397e-05,
+      "loss": 0.3066,
+      "step": 4440
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2.7254405159507087e-05,
+      "loss": 0.2778,
+      "step": 4450
+    },
+    {
+      "epoch": 1.45,
+      "eval_accuracy": 0.8876092433929443,
+      "eval_loss": 0.3001978099346161,
+      "eval_runtime": 3122.4892,
+      "eval_samples_per_second": 3.921,
+      "eval_steps_per_second": 0.061,
+      "step": 4450
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.7196821375100773e-05,
+      "loss": 0.3345,
+      "step": 4460
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.7139237590694462e-05,
+      "loss": 0.2502,
+      "step": 4470
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2.708165380628815e-05,
+      "loss": 0.2756,
+      "step": 4480
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.702407002188184e-05,
+      "loss": 0.2278,
+      "step": 4490
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.696648623747553e-05,
+      "loss": 0.2852,
+      "step": 4500
+    },
+    {
+      "epoch": 1.47,
+      "eval_accuracy": 0.8934901356697083,
+      "eval_loss": 0.28812864422798157,
+      "eval_runtime": 3109.2546,
+      "eval_samples_per_second": 3.938,
+      "eval_steps_per_second": 0.062,
+      "step": 4500
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2.6908902453069217e-05,
+      "loss": 0.2484,
+      "step": 4510
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6851318668662907e-05,
+      "loss": 0.2932,
+      "step": 4520
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6793734884256593e-05,
+      "loss": 0.3673,
+      "step": 4530
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2.6736151099850283e-05,
+      "loss": 0.2133,
+      "step": 4540
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.667856731544397e-05,
+      "loss": 0.2544,
+      "step": 4550
+    },
+    {
+      "epoch": 1.49,
+      "eval_accuracy": 0.8926733732223511,
+      "eval_loss": 0.28385019302368164,
+      "eval_runtime": 2905.8678,
+      "eval_samples_per_second": 4.213,
+      "eval_steps_per_second": 0.066,
+      "step": 4550
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.6620983531037662e-05,
+      "loss": 0.2157,
+      "step": 4560
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2.656339974663135e-05,
+      "loss": 0.3033,
+      "step": 4570
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.6505815962225038e-05,
+      "loss": 0.2199,
+      "step": 4580
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.6448232177818727e-05,
+      "loss": 0.2694,
+      "step": 4590
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2.6390648393412413e-05,
+      "loss": 0.2696,
+      "step": 4600
+    },
+    {
+      "epoch": 1.5,
+      "eval_accuracy": 0.8880176544189453,
+      "eval_loss": 0.285013347864151,
+      "eval_runtime": 2911.0552,
+      "eval_samples_per_second": 4.206,
+      "eval_steps_per_second": 0.066,
+      "step": 4600
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.6333064609006103e-05,
+      "loss": 0.3933,
+      "step": 4610
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.6275480824599796e-05,
+      "loss": 0.2221,
+      "step": 4620
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2.6217897040193486e-05,
+      "loss": 0.3228,
+      "step": 4630
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.6160313255787172e-05,
+      "loss": 0.2925,
+      "step": 4640
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.610272947138086e-05,
+      "loss": 0.343,
+      "step": 4650
+    },
+    {
+      "epoch": 1.52,
+      "eval_accuracy": 0.8952054381370544,
+      "eval_loss": 0.289019376039505,
+      "eval_runtime": 2918.1534,
+      "eval_samples_per_second": 4.195,
+      "eval_steps_per_second": 0.066,
+      "step": 4650
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2.6045145686974548e-05,
+      "loss": 0.315,
+      "step": 4660
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5987561902568237e-05,
+      "loss": 0.2391,
+      "step": 4670
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.592997811816193e-05,
+      "loss": 0.1955,
+      "step": 4680
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2.5872394333755616e-05,
+      "loss": 0.2877,
+      "step": 4690
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5814810549349306e-05,
+      "loss": 0.1999,
+      "step": 4700
+    },
+    {
+      "epoch": 1.54,
+      "eval_accuracy": 0.8983908891677856,
+      "eval_loss": 0.2899695336818695,
+      "eval_runtime": 2966.3751,
+      "eval_samples_per_second": 4.127,
+      "eval_steps_per_second": 0.065,
+      "step": 4700
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5757226764942992e-05,
+      "loss": 0.2439,
+      "step": 4710
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2.5699642980536682e-05,
+      "loss": 0.29,
+      "step": 4720
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5642059196130368e-05,
+      "loss": 0.227,
+      "step": 4730
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.5584475411724058e-05,
+      "loss": 0.3447,
+      "step": 4740
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2.552689162731775e-05,
+      "loss": 0.1606,
+      "step": 4750
+    },
+    {
+      "epoch": 1.55,
+      "eval_accuracy": 0.8899779319763184,
+      "eval_loss": 0.31522202491760254,
+      "eval_runtime": 2914.7065,
+      "eval_samples_per_second": 4.2,
+      "eval_steps_per_second": 0.066,
+      "step": 4750
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5469307842911437e-05,
+      "loss": 0.2824,
+      "step": 4760
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5411724058505126e-05,
+      "loss": 0.2872,
+      "step": 4770
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5354140274098816e-05,
+      "loss": 0.1249,
+      "step": 4780
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2.5296556489692502e-05,
+      "loss": 0.2302,
+      "step": 4790
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5238972705286192e-05,
+      "loss": 0.2931,
+      "step": 4800
+    },
+    {
+      "epoch": 1.57,
+      "eval_accuracy": 0.8784611821174622,
+      "eval_loss": 0.3652476668357849,
+      "eval_runtime": 2910.9256,
+      "eval_samples_per_second": 4.206,
+      "eval_steps_per_second": 0.066,
+      "step": 4800
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.5181388920879885e-05,
+      "loss": 0.2541,
+      "step": 4810
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2.512380513647357e-05,
+      "loss": 0.3191,
+      "step": 4820
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.506622135206726e-05,
+      "loss": 0.2633,
+      "step": 4830
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.5008637567660947e-05,
+      "loss": 0.2819,
+      "step": 4840
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2.4951053783254636e-05,
+      "loss": 0.2449,
+      "step": 4850
+    },
+    {
+      "epoch": 1.58,
+      "eval_accuracy": 0.8983908891677856,
+      "eval_loss": 0.2624811828136444,
+      "eval_runtime": 2898.1844,
+      "eval_samples_per_second": 4.224,
+      "eval_steps_per_second": 0.066,
+      "step": 4850
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.4893469998848326e-05,
+      "loss": 0.1688,
+      "step": 4860
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.4835886214442016e-05,
+      "loss": 0.2743,
+      "step": 4870
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2.4778302430035702e-05,
+      "loss": 0.2757,
+      "step": 4880
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.472071864562939e-05,
+      "loss": 0.2341,
+      "step": 4890
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.466313486122308e-05,
+      "loss": 0.1984,
+      "step": 4900
+    },
+    {
+      "epoch": 1.6,
+      "eval_accuracy": 0.8992077112197876,
+      "eval_loss": 0.2779299020767212,
+      "eval_runtime": 2931.606,
+      "eval_samples_per_second": 4.176,
+      "eval_steps_per_second": 0.065,
+      "step": 4900
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2.4605551076816767e-05,
+      "loss": 0.2568,
+      "step": 4910
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.454796729241046e-05,
+      "loss": 0.2987,
+      "step": 4920
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.4490383508004146e-05,
+      "loss": 0.2404,
+      "step": 4930
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2.4432799723597836e-05,
+      "loss": 0.326,
+      "step": 4940
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.4375215939191525e-05,
+      "loss": 0.2954,
+      "step": 4950
+    },
+    {
+      "epoch": 1.62,
+      "eval_accuracy": 0.8992893695831299,
+      "eval_loss": 0.2730477452278137,
+      "eval_runtime": 2995.6287,
+      "eval_samples_per_second": 4.087,
+      "eval_steps_per_second": 0.064,
+      "step": 4950
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.4317632154785215e-05,
+      "loss": 0.2491,
+      "step": 4960
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2.42600483703789e-05,
+      "loss": 0.2666,
+      "step": 4970
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.420246458597259e-05,
+      "loss": 0.2994,
+      "step": 4980
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.414488080156628e-05,
+      "loss": 0.2328,
+      "step": 4990
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2.4087297017159967e-05,
+      "loss": 0.213,
+      "step": 5000
+    },
+    {
+      "epoch": 1.63,
+      "eval_accuracy": 0.8965939879417419,
+      "eval_loss": 0.29905205965042114,
+      "eval_runtime": 2889.7678,
+      "eval_samples_per_second": 4.237,
+      "eval_steps_per_second": 0.066,
+      "step": 5000
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.402971323275366e-05,
+      "loss": 0.3081,
+      "step": 5010
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.3972129448347346e-05,
+      "loss": 0.3024,
+      "step": 5020
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2.3914545663941035e-05,
+      "loss": 0.1831,
+      "step": 5030
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.3856961879534725e-05,
+      "loss": 0.3269,
+      "step": 5040
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.3799378095128415e-05,
+      "loss": 0.2451,
+      "step": 5050
+    },
+    {
+      "epoch": 1.65,
+      "eval_accuracy": 0.898145854473114,
+      "eval_loss": 0.3344215154647827,
+      "eval_runtime": 2870.5816,
+      "eval_samples_per_second": 4.265,
+      "eval_steps_per_second": 0.067,
+      "step": 5050
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2.37417943107221e-05,
+      "loss": 0.2023,
+      "step": 5060
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.368421052631579e-05,
+      "loss": 0.3063,
+      "step": 5070
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.362662674190948e-05,
+      "loss": 0.3241,
+      "step": 5080
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2.356904295750317e-05,
+      "loss": 0.2943,
+      "step": 5090
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.3511459173096856e-05,
+      "loss": 0.3022,
+      "step": 5100
+    },
+    {
+      "epoch": 1.67,
+      "eval_accuracy": 0.8949604034423828,
+      "eval_loss": 0.2846279740333557,
+      "eval_runtime": 2901.6913,
+      "eval_samples_per_second": 4.219,
+      "eval_steps_per_second": 0.066,
+      "step": 5100
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.3453875388690545e-05,
+      "loss": 0.2632,
+      "step": 5110
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2.3396291604284235e-05,
+      "loss": 0.2458,
+      "step": 5120
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.333870781987792e-05,
+      "loss": 0.2831,
+      "step": 5130
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.3281124035471614e-05,
+      "loss": 0.2299,
+      "step": 5140
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2.32235402510653e-05,
+      "loss": 0.1725,
+      "step": 5150
+    },
+    {
+      "epoch": 1.68,
+      "eval_accuracy": 0.8927550315856934,
+      "eval_loss": 0.339284747838974,
+      "eval_runtime": 2881.5176,
+      "eval_samples_per_second": 4.249,
+      "eval_steps_per_second": 0.067,
+      "step": 5150
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.316595646665899e-05,
+      "loss": 0.165,
+      "step": 5160
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.310837268225268e-05,
+      "loss": 0.3811,
+      "step": 5170
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2.305078889784637e-05,
+      "loss": 0.2255,
+      "step": 5180
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.2993205113440055e-05,
+      "loss": 0.2664,
+      "step": 5190
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.2935621329033745e-05,
+      "loss": 0.2801,
+      "step": 5200
+    },
+    {
+      "epoch": 1.7,
+      "eval_accuracy": 0.9001878499984741,
+      "eval_loss": 0.261701375246048,
+      "eval_runtime": 2890.3977,
+      "eval_samples_per_second": 4.236,
+      "eval_steps_per_second": 0.066,
+      "step": 5200
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2.2878037544627435e-05,
+      "loss": 0.2242,
+      "step": 5210
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.282045376022112e-05,
+      "loss": 0.2252,
+      "step": 5220
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.2762869975814814e-05,
+      "loss": 0.2854,
+      "step": 5230
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.27052861914085e-05,
+      "loss": 0.2086,
+      "step": 5240
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.264770240700219e-05,
+      "loss": 0.2925,
+      "step": 5250
+    },
+    {
+      "epoch": 1.72,
+      "eval_accuracy": 0.8987993001937866,
+      "eval_loss": 0.30346158146858215,
+      "eval_runtime": 2875.143,
+      "eval_samples_per_second": 4.258,
+      "eval_steps_per_second": 0.067,
+      "step": 5250
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.259011862259588e-05,
+      "loss": 0.2909,
+      "step": 5260
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.253253483818957e-05,
+      "loss": 0.2984,
+      "step": 5270
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.2474951053783255e-05,
+      "loss": 0.2474,
+      "step": 5280
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.2417367269376945e-05,
+      "loss": 0.1422,
+      "step": 5290
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.2359783484970634e-05,
+      "loss": 0.3178,
+      "step": 5300
+    },
+    {
+      "epoch": 1.73,
+      "eval_accuracy": 0.898554265499115,
+      "eval_loss": 0.3339698910713196,
+      "eval_runtime": 2886.0568,
+      "eval_samples_per_second": 4.242,
+      "eval_steps_per_second": 0.067,
+      "step": 5300
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.230219970056432e-05,
+      "loss": 0.2817,
+      "step": 5310
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.224461591615801e-05,
+      "loss": 0.1896,
+      "step": 5320
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.21870321317517e-05,
+      "loss": 0.1787,
+      "step": 5330
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.212944834734539e-05,
+      "loss": 0.344,
+      "step": 5340
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.2071864562939075e-05,
+      "loss": 0.2909,
+      "step": 5350
+    },
+    {
+      "epoch": 1.75,
+      "eval_accuracy": 0.898554265499115,
+      "eval_loss": 0.3080793023109436,
+      "eval_runtime": 2867.0175,
+      "eval_samples_per_second": 4.27,
+      "eval_steps_per_second": 0.067,
+      "step": 5350
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.2014280778532768e-05,
+      "loss": 0.3911,
+      "step": 5360
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.1956696994126454e-05,
+      "loss": 0.2377,
+      "step": 5370
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.1899113209720144e-05,
+      "loss": 0.3122,
+      "step": 5380
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.1841529425313834e-05,
+      "loss": 0.2893,
+      "step": 5390
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.178394564090752e-05,
+      "loss": 0.2247,
+      "step": 5400
+    },
+    {
+      "epoch": 1.76,
+      "eval_accuracy": 0.900922954082489,
+      "eval_loss": 0.2840961217880249,
+      "eval_runtime": 2863.4685,
+      "eval_samples_per_second": 4.276,
+      "eval_steps_per_second": 0.067,
+      "step": 5400
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.172636185650121e-05,
+      "loss": 0.1947,
+      "step": 5410
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.16687780720949e-05,
+      "loss": 0.24,
+      "step": 5420
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.161119428768859e-05,
+      "loss": 0.4273,
+      "step": 5430
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.1553610503282275e-05,
+      "loss": 0.148,
+      "step": 5440
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.1496026718875968e-05,
+      "loss": 0.2023,
+      "step": 5450
+    },
+    {
+      "epoch": 1.78,
+      "eval_accuracy": 0.9002695679664612,
+      "eval_loss": 0.2983114421367645,
+      "eval_runtime": 2872.2543,
+      "eval_samples_per_second": 4.263,
+      "eval_steps_per_second": 0.067,
+      "step": 5450
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.1438442934469654e-05,
+      "loss": 0.2521,
+      "step": 5460
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1380859150063344e-05,
+      "loss": 0.2531,
+      "step": 5470
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1323275365657033e-05,
+      "loss": 0.3155,
+      "step": 5480
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1265691581250723e-05,
+      "loss": 0.2677,
+      "step": 5490
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.120810779684441e-05,
+      "loss": 0.2107,
+      "step": 5500
+    },
+    {
+      "epoch": 1.8,
+      "eval_accuracy": 0.89577716588974,
+      "eval_loss": 0.31875815987586975,
+      "eval_runtime": 2853.814,
+      "eval_samples_per_second": 4.29,
+      "eval_steps_per_second": 0.067,
+      "step": 5500
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.11505240124381e-05,
+      "loss": 0.2495,
+      "step": 5510
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.1092940228031788e-05,
+      "loss": 0.1656,
+      "step": 5520
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.1035356443625474e-05,
+      "loss": 0.2836,
+      "step": 5530
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.0977772659219164e-05,
+      "loss": 0.3072,
+      "step": 5540
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.0920188874812854e-05,
+      "loss": 0.2074,
+      "step": 5550
+    },
+    {
+      "epoch": 1.81,
+      "eval_accuracy": 0.9019031524658203,
+      "eval_loss": 0.26809632778167725,
+      "eval_runtime": 2846.6005,
+      "eval_samples_per_second": 4.301,
+      "eval_steps_per_second": 0.067,
+      "step": 5550
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.0862605090406543e-05,
+      "loss": 0.2032,
+      "step": 5560
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.080502130600023e-05,
+      "loss": 0.205,
+      "step": 5570
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2.0747437521593922e-05,
+      "loss": 0.2309,
+      "step": 5580
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.068985373718761e-05,
+      "loss": 0.3076,
+      "step": 5590
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.0632269952781298e-05,
+      "loss": 0.2089,
+      "step": 5600
+    },
+    {
+      "epoch": 1.83,
+      "eval_accuracy": 0.9031283259391785,
+      "eval_loss": 0.27018752694129944,
+      "eval_runtime": 2917.6978,
+      "eval_samples_per_second": 4.196,
+      "eval_steps_per_second": 0.066,
+      "step": 5600
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2.0574686168374988e-05,
+      "loss": 0.2181,
+      "step": 5610
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.0517102383968674e-05,
+      "loss": 0.2765,
+      "step": 5620
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.0459518599562364e-05,
+      "loss": 0.1267,
+      "step": 5630
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2.0401934815156053e-05,
+      "loss": 0.3208,
+      "step": 5640
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.0344351030749743e-05,
+      "loss": 0.2785,
+      "step": 5650
+    },
+    {
+      "epoch": 1.85,
+      "eval_accuracy": 0.8988810181617737,
+      "eval_loss": 0.31067556142807007,
+      "eval_runtime": 2879.816,
+      "eval_samples_per_second": 4.251,
+      "eval_steps_per_second": 0.067,
+      "step": 5650
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.028676724634343e-05,
+      "loss": 0.1782,
+      "step": 5660
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2.0229183461937122e-05,
+      "loss": 0.2338,
+      "step": 5670
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.0171599677530808e-05,
+      "loss": 0.2991,
+      "step": 5680
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.0114015893124498e-05,
+      "loss": 0.3224,
+      "step": 5690
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2.0056432108718187e-05,
+      "loss": 0.2416,
+      "step": 5700
+    },
+    {
+      "epoch": 1.86,
+      "eval_accuracy": 0.8996977806091309,
+      "eval_loss": 0.28742799162864685,
+      "eval_runtime": 2868.7203,
+      "eval_samples_per_second": 4.268,
+      "eval_steps_per_second": 0.067,
+      "step": 5700
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9998848324311874e-05,
+      "loss": 0.2502,
+      "step": 5710
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9941264539905563e-05,
+      "loss": 0.2489,
+      "step": 5720
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.9883680755499253e-05,
+      "loss": 0.2386,
+      "step": 5730
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9826096971092942e-05,
+      "loss": 0.2654,
+      "step": 5740
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.976851318668663e-05,
+      "loss": 0.1684,
+      "step": 5750
+    },
+    {
+      "epoch": 1.88,
+      "eval_accuracy": 0.9012497067451477,
+      "eval_loss": 0.30475884675979614,
+      "eval_runtime": 2876.0701,
+      "eval_samples_per_second": 4.257,
+      "eval_steps_per_second": 0.067,
+      "step": 5750
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.9710929402280318e-05,
+      "loss": 0.3422,
+      "step": 5760
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.9653345617874008e-05,
+      "loss": 0.2296,
+      "step": 5770
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.9595761833467697e-05,
+      "loss": 0.3785,
+      "step": 5780
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.9538178049061383e-05,
+      "loss": 0.2553,
+      "step": 5790
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.9480594264655076e-05,
+      "loss": 0.2557,
+      "step": 5800
+    },
+    {
+      "epoch": 1.89,
+      "eval_accuracy": 0.9028832912445068,
+      "eval_loss": 0.26270949840545654,
+      "eval_runtime": 2881.7997,
+      "eval_samples_per_second": 4.248,
+      "eval_steps_per_second": 0.067,
+      "step": 5800
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.9423010480248763e-05,
+      "loss": 0.1637,
+      "step": 5810
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.9365426695842452e-05,
+      "loss": 0.3039,
+      "step": 5820
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.9307842911436142e-05,
+      "loss": 0.2283,
+      "step": 5830
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.9250259127029828e-05,
+      "loss": 0.2509,
+      "step": 5840
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.9192675342623518e-05,
+      "loss": 0.2598,
+      "step": 5850
+    },
+    {
+      "epoch": 1.91,
+      "eval_accuracy": 0.8999428153038025,
+      "eval_loss": 0.26313936710357666,
+      "eval_runtime": 2904.1486,
+      "eval_samples_per_second": 4.216,
+      "eval_steps_per_second": 0.066,
+      "step": 5850
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 1.9135091558217207e-05,
+      "loss": 0.2911,
+      "step": 5860
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.9077507773810897e-05,
+      "loss": 0.1678,
+      "step": 5870
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.9019923989404583e-05,
+      "loss": 0.2472,
+      "step": 5880
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 1.8962340204998276e-05,
+      "loss": 0.2314,
+      "step": 5890
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.8904756420591962e-05,
+      "loss": 0.2912,
+      "step": 5900
+    },
+    {
+      "epoch": 1.93,
+      "eval_accuracy": 0.8992893695831299,
+      "eval_loss": 0.26787182688713074,
+      "eval_runtime": 2904.0893,
+      "eval_samples_per_second": 4.216,
+      "eval_steps_per_second": 0.066,
+      "step": 5900
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.8847172636185652e-05,
+      "loss": 0.2041,
+      "step": 5910
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 1.878958885177934e-05,
+      "loss": 0.3211,
+      "step": 5920
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.8732005067373028e-05,
+      "loss": 0.1922,
+      "step": 5930
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.8674421282966717e-05,
+      "loss": 0.2857,
+      "step": 5940
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 1.8616837498560403e-05,
+      "loss": 0.2408,
+      "step": 5950
+    },
+    {
+      "epoch": 1.94,
+      "eval_accuracy": 0.9022298455238342,
+      "eval_loss": 0.25954896211624146,
+      "eval_runtime": 2892.6055,
+      "eval_samples_per_second": 4.233,
+      "eval_steps_per_second": 0.066,
+      "step": 5950
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.8559253714154096e-05,
+      "loss": 0.2656,
+      "step": 5960
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.8501669929747783e-05,
+      "loss": 0.2583,
+      "step": 5970
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 1.8444086145341472e-05,
+      "loss": 0.3119,
+      "step": 5980
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.8386502360935162e-05,
+      "loss": 0.1984,
+      "step": 5990
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 1.832891857652885e-05,
+      "loss": 0.1835,
+      "step": 6000
+    },
+    {
+      "epoch": 1.96,
+      "eval_accuracy": 0.9010863304138184,
+      "eval_loss": 0.2831316888332367,
+      "eval_runtime": 2891.4361,
+      "eval_samples_per_second": 4.234,
+      "eval_steps_per_second": 0.066,
+      "step": 6000
+    }
+  ],
+  "max_steps": 9183,
+  "num_train_epochs": 3,
+  "total_flos": 2.525681953717248e+16,
+  "trial_name": null,
+  "trial_params": null
+}