diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,3020 +1,1810 @@
 {
-  "best_metric": null,
-  "best_model_checkpoint": null,
-  "epoch": 0.5,
-  "eval_steps": 100,
-  "global_step": 20000,
+  "best_metric": 0.637090802192688,
+  "best_model_checkpoint": "runs/deepseek_CMU-AIR2/math-deepseek_FULL_HardArith_Interm_20240424-065814/checkpoint-2500",
+  "epoch": 0.0625,
+  "eval_steps": 500,
+  "global_step": 2500,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.0,
-      "grad_norm": 11.075758934020996,
-      "learning_rate": 4.000000000000001e-06,
-      "loss": 1.4211,
-      "step": 100
+      "grad_norm": 4.3125,
+      "learning_rate": 4.0000000000000003e-07,
+      "loss": 0.6839,
+      "step": 10
     },
     {
       "epoch": 0.0,
-      "eval_loss": 1.2334846258163452,
-      "eval_runtime": 95.5143,
-      "eval_samples_per_second": 10.47,
-      "eval_steps_per_second": 10.47,
-      "step": 100
-    },
-    {
-      "epoch": 0.01,
-      "grad_norm": 19.510242462158203,
-      "learning_rate": 8.000000000000001e-06,
-      "loss": 1.2244,
-      "step": 200
-    },
-    {
-      "epoch": 0.01,
-      "eval_loss": 1.1949360370635986,
-      "eval_runtime": 95.601,
-      "eval_samples_per_second": 10.46,
-      "eval_steps_per_second": 10.46,
-      "step": 200
-    },
-    {
-      "epoch": 0.01,
-      "grad_norm": 6.514332294464111,
-      "learning_rate": 1.2e-05,
-      "loss": 1.2025,
-      "step": 300
-    },
-    {
-      "epoch": 0.01,
-      "eval_loss": 1.1998025178909302,
-      "eval_runtime": 95.7133,
-      "eval_samples_per_second": 10.448,
-      "eval_steps_per_second": 10.448,
-      "step": 300
-    },
-    {
-      "epoch": 0.01,
-      "grad_norm": 14.404813766479492,
-      "learning_rate": 1.6000000000000003e-05,
-      "loss": 1.2387,
-      "step": 400
-    },
-    {
-      "epoch": 0.01,
-      "eval_loss": 1.2146040201187134,
-      "eval_runtime": 95.6405,
-      "eval_samples_per_second": 10.456,
-      "eval_steps_per_second": 10.456,
-      "step": 400
-    },
-    {
-      "epoch": 0.01,
-      "grad_norm": 12.25019645690918,
-      "learning_rate": 2e-05,
-      "loss": 1.2359,
-      "step": 500
-    },
-    {
-      "epoch": 0.01,
-      "eval_loss": 1.232546329498291,
-      "eval_runtime": 95.701,
-      "eval_samples_per_second": 10.449,
-      "eval_steps_per_second": 10.449,
-      "step": 500
-    },
-    {
-      "epoch": 0.01,
-      "grad_norm": 5.677526473999023,
-      "learning_rate": 1.98974358974359e-05,
-      "loss": 1.2189,
-      "step": 600
-    },
-    {
-      "epoch": 0.01,
-      "eval_loss": 1.2301095724105835,
-      "eval_runtime": 93.7117,
-      "eval_samples_per_second": 10.671,
-      "eval_steps_per_second": 10.671,
-      "step": 600
-    },
-    {
-      "epoch": 0.02,
-      "grad_norm": 13.75782585144043,
-      "learning_rate": 1.9794871794871798e-05,
-      "loss": 1.1982,
-      "step": 700
-    },
-    {
-      "epoch": 0.02,
-      "eval_loss": 1.2097805738449097,
-      "eval_runtime": 95.0989,
-      "eval_samples_per_second": 10.515,
-      "eval_steps_per_second": 10.515,
-      "step": 700
-    },
-    {
-      "epoch": 0.02,
-      "grad_norm": 7.497257232666016,
-      "learning_rate": 1.9692307692307696e-05,
-      "loss": 1.178,
-      "step": 800
-    },
-    {
-      "epoch": 0.02,
-      "eval_loss": 1.1660981178283691,
-      "eval_runtime": 95.0124,
-      "eval_samples_per_second": 10.525,
-      "eval_steps_per_second": 10.525,
-      "step": 800
-    },
-    {
-      "epoch": 0.02,
-      "grad_norm": 6.337695121765137,
-      "learning_rate": 1.958974358974359e-05,
-      "loss": 1.1852,
-      "step": 900
-    },
-    {
-      "epoch": 0.02,
-      "eval_loss": 1.1449871063232422,
-      "eval_runtime": 94.9978,
-      "eval_samples_per_second": 10.527,
-      "eval_steps_per_second": 10.527,
-      "step": 900
-    },
-    {
-      "epoch": 0.03,
-      "grad_norm": 4.1737895011901855,
-      "learning_rate": 1.9487179487179488e-05,
-      "loss": 1.1056,
-      "step": 1000
-    },
-    {
-      "epoch": 0.03,
-      "eval_loss": 1.1298564672470093,
-      "eval_runtime": 94.9792,
-      "eval_samples_per_second": 10.529,
-      "eval_steps_per_second": 10.529,
-      "step": 1000
-    },
-    {
-      "epoch": 0.03,
-      "grad_norm": 6.821949481964111,
-      "learning_rate": 1.9384615384615386e-05,
-      "loss": 1.1681,
-      "step": 1100
-    },
-    {
-      "epoch": 0.03,
-      "eval_loss": 1.1448038816452026,
-      "eval_runtime": 94.8638,
-      "eval_samples_per_second": 10.541,
-      "eval_steps_per_second": 10.541,
-      "step": 1100
-    },
-    {
-      "epoch": 0.03,
-      "grad_norm": 3.908639907836914,
-      "learning_rate": 1.9282051282051284e-05,
-      "loss": 1.1311,
-      "step": 1200
-    },
-    {
-      "epoch": 0.03,
-      "eval_loss": 1.0957874059677124,
-      "eval_runtime": 94.6454,
-      "eval_samples_per_second": 10.566,
-      "eval_steps_per_second": 10.566,
-      "step": 1200
-    },
-    {
-      "epoch": 0.03,
-      "grad_norm": 5.896665096282959,
-      "learning_rate": 1.9179487179487182e-05,
-      "loss": 1.134,
-      "step": 1300
-    },
-    {
-      "epoch": 0.03,
-      "eval_loss": 1.0930886268615723,
-      "eval_runtime": 94.6326,
-      "eval_samples_per_second": 10.567,
-      "eval_steps_per_second": 10.567,
-      "step": 1300
-    },
-    {
-      "epoch": 0.04,
-      "grad_norm": 5.936835765838623,
-      "learning_rate": 1.907692307692308e-05,
-      "loss": 1.0822,
-      "step": 1400
-    },
-    {
-      "epoch": 0.04,
-      "eval_loss": 1.1114310026168823,
-      "eval_runtime": 94.6679,
-      "eval_samples_per_second": 10.563,
-      "eval_steps_per_second": 10.563,
-      "step": 1400
-    },
-    {
-      "epoch": 0.04,
-      "grad_norm": 4.07722806930542,
-      "learning_rate": 1.8974358974358975e-05,
-      "loss": 1.0924,
-      "step": 1500
-    },
-    {
-      "epoch": 0.04,
-      "eval_loss": 1.1060781478881836,
-      "eval_runtime": 94.6419,
-      "eval_samples_per_second": 10.566,
-      "eval_steps_per_second": 10.566,
-      "step": 1500
-    },
-    {
-      "epoch": 0.04,
-      "grad_norm": 8.239760398864746,
-      "learning_rate": 1.8871794871794873e-05,
-      "loss": 1.0801,
-      "step": 1600
-    },
-    {
-      "epoch": 0.04,
-      "eval_loss": 1.0673882961273193,
-      "eval_runtime": 94.6026,
-      "eval_samples_per_second": 10.571,
-      "eval_steps_per_second": 10.571,
-      "step": 1600
-    },
-    {
-      "epoch": 0.04,
-      "grad_norm": 4.63267707824707,
-      "learning_rate": 1.876923076923077e-05,
-      "loss": 1.1051,
-      "step": 1700
-    },
-    {
-      "epoch": 0.04,
-      "eval_loss": 1.0620085000991821,
-      "eval_runtime": 94.6137,
-      "eval_samples_per_second": 10.569,
-      "eval_steps_per_second": 10.569,
-      "step": 1700
-    },
-    {
-      "epoch": 0.04,
-      "grad_norm": 5.941944122314453,
-      "learning_rate": 1.866666666666667e-05,
-      "loss": 1.081,
-      "step": 1800
-    },
-    {
-      "epoch": 0.04,
-      "eval_loss": 1.0232795476913452,
-      "eval_runtime": 94.6201,
-      "eval_samples_per_second": 10.569,
-      "eval_steps_per_second": 10.569,
-      "step": 1800
-    },
-    {
-      "epoch": 0.05,
-      "grad_norm": 4.456423282623291,
-      "learning_rate": 1.8564102564102567e-05,
-      "loss": 1.0303,
-      "step": 1900
-    },
-    {
-      "epoch": 0.05,
-      "eval_loss": 1.052198052406311,
-      "eval_runtime": 94.6135,
-      "eval_samples_per_second": 10.569,
-      "eval_steps_per_second": 10.569,
-      "step": 1900
-    },
-    {
-      "epoch": 0.05,
-      "grad_norm": 8.572375297546387,
-      "learning_rate": 1.8461538461538465e-05,
-      "loss": 1.048,
-      "step": 2000
-    },
-    {
-      "epoch": 0.05,
-      "eval_loss": 1.0538784265518188,
-      "eval_runtime": 94.6091,
-      "eval_samples_per_second": 10.57,
-      "eval_steps_per_second": 10.57,
-      "step": 2000
-    },
-    {
-      "epoch": 0.05,
-      "grad_norm": 4.592229843139648,
-      "learning_rate": 1.835897435897436e-05,
-      "loss": 1.0623,
-      "step": 2100
-    },
-    {
-      "epoch": 0.05,
-      "eval_loss": 1.0412269830703735,
-      "eval_runtime": 94.6145,
-      "eval_samples_per_second": 10.569,
-      "eval_steps_per_second": 10.569,
-      "step": 2100
-    },
-    {
-      "epoch": 0.06,
-      "grad_norm": 3.802849054336548,
-      "learning_rate": 1.8256410256410257e-05,
-      "loss": 1.0699,
-      "step": 2200
-    },
-    {
-      "epoch": 0.06,
-      "eval_loss": 1.0224683284759521,
-      "eval_runtime": 94.6257,
-      "eval_samples_per_second": 10.568,
-      "eval_steps_per_second": 10.568,
-      "step": 2200
-    },
-    {
-      "epoch": 0.06,
-      "grad_norm": 4.145386695861816,
-      "learning_rate": 1.8153846153846155e-05,
-      "loss": 0.9979,
-      "step": 2300
-    },
-    {
-      "epoch": 0.06,
-      "eval_loss": 1.0313537120819092,
-      "eval_runtime": 94.6038,
-      "eval_samples_per_second": 10.57,
-      "eval_steps_per_second": 10.57,
-      "step": 2300
-    },
-    {
-      "epoch": 0.06,
-      "grad_norm": 4.273189067840576,
-      "learning_rate": 1.8051282051282053e-05,
-      "loss": 1.052,
-      "step": 2400
-    },
-    {
-      "epoch": 0.06,
-      "eval_loss": 1.0333129167556763,
-      "eval_runtime": 94.5812,
-      "eval_samples_per_second": 10.573,
-      "eval_steps_per_second": 10.573,
-      "step": 2400
-    },
-    {
-      "epoch": 0.06,
-      "grad_norm": 6.093295574188232,
-      "learning_rate": 1.794871794871795e-05,
-      "loss": 1.0025,
-      "step": 2500
-    },
-    {
-      "epoch": 0.06,
-      "eval_loss": 0.9930676817893982,
-      "eval_runtime": 94.5844,
-      "eval_samples_per_second": 10.573,
-      "eval_steps_per_second": 10.573,
-      "step": 2500
-    },
-    {
-      "epoch": 0.07,
-      "grad_norm": 3.8665239810943604,
-      "learning_rate": 1.784615384615385e-05,
-      "loss": 1.016,
-      "step": 2600
-    },
-    {
-      "epoch": 0.07,
-      "eval_loss": 1.0165327787399292,
-      "eval_runtime": 94.6284,
-      "eval_samples_per_second": 10.568,
-      "eval_steps_per_second": 10.568,
-      "step": 2600
-    },
-    {
-      "epoch": 0.07,
-      "grad_norm": 4.170589923858643,
-      "learning_rate": 1.7743589743589744e-05,
-      "loss": 0.9838,
-      "step": 2700
-    },
-    {
-      "epoch": 0.07,
-      "eval_loss": 1.019612193107605,
-      "eval_runtime": 94.5637,
-      "eval_samples_per_second": 10.575,
-      "eval_steps_per_second": 10.575,
-      "step": 2700
-    },
-    {
-      "epoch": 0.07,
-      "grad_norm": 5.267038345336914,
-      "learning_rate": 1.7641025641025642e-05,
-      "loss": 1.0019,
-      "step": 2800
-    },
-    {
-      "epoch": 0.07,
-      "eval_loss": 0.9611077308654785,
-      "eval_runtime": 94.5638,
-      "eval_samples_per_second": 10.575,
-      "eval_steps_per_second": 10.575,
-      "step": 2800
-    },
-    {
-      "epoch": 0.07,
-      "grad_norm": 6.0030670166015625,
-      "learning_rate": 1.753846153846154e-05,
-      "loss": 0.9659,
-      "step": 2900
-    },
-    {
-      "epoch": 0.07,
-      "eval_loss": 0.9955070614814758,
-      "eval_runtime": 94.5812,
-      "eval_samples_per_second": 10.573,
-      "eval_steps_per_second": 10.573,
-      "step": 2900
-    },
-    {
-      "epoch": 0.07,
-      "grad_norm": 7.910103797912598,
-      "learning_rate": 1.7435897435897438e-05,
-      "loss": 0.9868,
-      "step": 3000
-    },
-    {
-      "epoch": 0.07,
-      "eval_loss": 0.9867352247238159,
-      "eval_runtime": 94.5761,
-      "eval_samples_per_second": 10.573,
-      "eval_steps_per_second": 10.573,
-      "step": 3000
-    },
-    {
-      "epoch": 0.08,
-      "grad_norm": 3.8992385864257812,
-      "learning_rate": 1.7333333333333336e-05,
-      "loss": 0.9562,
-      "step": 3100
-    },
-    {
-      "epoch": 0.08,
-      "eval_loss": 0.9793102741241455,
-      "eval_runtime": 94.5799,
-      "eval_samples_per_second": 10.573,
-      "eval_steps_per_second": 10.573,
-      "step": 3100
-    },
-    {
-      "epoch": 0.08,
-      "grad_norm": 4.803006649017334,
-      "learning_rate": 1.7230769230769234e-05,
-      "loss": 0.9865,
-      "step": 3200
-    },
-    {
-      "epoch": 0.08,
-      "eval_loss": 0.9728935360908508,
-      "eval_runtime": 94.5722,
-      "eval_samples_per_second": 10.574,
-      "eval_steps_per_second": 10.574,
-      "step": 3200
-    },
-    {
-      "epoch": 0.08,
-      "grad_norm": 7.022669792175293,
-      "learning_rate": 1.7128205128205128e-05,
-      "loss": 0.9704,
-      "step": 3300
-    },
-    {
-      "epoch": 0.08,
-      "eval_loss": 0.9876529574394226,
-      "eval_runtime": 94.5739,
-      "eval_samples_per_second": 10.574,
-      "eval_steps_per_second": 10.574,
-      "step": 3300
-    },
-    {
-      "epoch": 0.09,
-      "grad_norm": 4.264179229736328,
-      "learning_rate": 1.7025641025641026e-05,
-      "loss": 0.9809,
-      "step": 3400
-    },
-    {
-      "epoch": 0.09,
-      "eval_loss": 0.9546476006507874,
-      "eval_runtime": 94.5665,
-      "eval_samples_per_second": 10.575,
-      "eval_steps_per_second": 10.575,
-      "step": 3400
-    },
-    {
-      "epoch": 0.09,
-      "grad_norm": 5.092813968658447,
-      "learning_rate": 1.6923076923076924e-05,
-      "loss": 0.9442,
-      "step": 3500
-    },
-    {
-      "epoch": 0.09,
-      "eval_loss": 0.9812406301498413,
-      "eval_runtime": 94.5658,
-      "eval_samples_per_second": 10.575,
-      "eval_steps_per_second": 10.575,
-      "step": 3500
-    },
-    {
-      "epoch": 0.09,
-      "grad_norm": 4.2536234855651855,
-      "learning_rate": 1.6820512820512822e-05,
-      "loss": 0.9708,
-      "step": 3600
-    },
-    {
-      "epoch": 0.09,
-      "eval_loss": 0.9307680726051331,
-      "eval_runtime": 94.5711,
-      "eval_samples_per_second": 10.574,
-      "eval_steps_per_second": 10.574,
-      "step": 3600
-    },
-    {
-      "epoch": 0.09,
-      "grad_norm": 5.6083831787109375,
-      "learning_rate": 1.671794871794872e-05,
-      "loss": 0.9804,
-      "step": 3700
-    },
-    {
-      "epoch": 0.09,
-      "eval_loss": 0.9597137570381165,
-      "eval_runtime": 94.5422,
-      "eval_samples_per_second": 10.577,
-      "eval_steps_per_second": 10.577,
-      "step": 3700
-    },
-    {
-      "epoch": 0.1,
-      "grad_norm": 4.942496299743652,
-      "learning_rate": 1.6615384615384618e-05,
-      "loss": 0.957,
-      "step": 3800
-    },
-    {
-      "epoch": 0.1,
-      "eval_loss": 0.9576207995414734,
-      "eval_runtime": 94.5529,
-      "eval_samples_per_second": 10.576,
-      "eval_steps_per_second": 10.576,
-      "step": 3800
-    },
-    {
-      "epoch": 0.1,
-      "grad_norm": 4.260251522064209,
-      "learning_rate": 1.6512820512820513e-05,
-      "loss": 0.9503,
-      "step": 3900
+      "grad_norm": 6.53125,
+      "learning_rate": 8.000000000000001e-07,
+      "loss": 0.6587,
+      "step": 20
     },
     {
-      "epoch": 0.1,
-      "eval_loss": 0.9324920773506165,
-      "eval_runtime": 94.5562,
-      "eval_samples_per_second": 10.576,
-      "eval_steps_per_second": 10.576,
-      "step": 3900
-    },
-    {
-      "epoch": 0.1,
-      "grad_norm": 2.702169418334961,
-      "learning_rate": 1.641025641025641e-05,
-      "loss": 0.9125,
-      "step": 4000
-    },
-    {
-      "epoch": 0.1,
-      "eval_loss": 0.9403380155563354,
-      "eval_runtime": 95.5173,
-      "eval_samples_per_second": 10.469,
-      "eval_steps_per_second": 10.469,
-      "step": 4000
-    },
-    {
-      "epoch": 0.1,
-      "grad_norm": 6.76617431640625,
-      "learning_rate": 1.630769230769231e-05,
-      "loss": 0.9334,
-      "step": 4100
-    },
-    {
-      "epoch": 0.1,
-      "eval_loss": 0.9736199975013733,
-      "eval_runtime": 95.5516,
-      "eval_samples_per_second": 10.466,
-      "eval_steps_per_second": 10.466,
-      "step": 4100
-    },
-    {
-      "epoch": 0.1,
-      "grad_norm": 10.489439964294434,
-      "learning_rate": 1.6205128205128207e-05,
-      "loss": 0.9424,
-      "step": 4200
-    },
-    {
-      "epoch": 0.1,
-      "eval_loss": 0.9593413472175598,
-      "eval_runtime": 95.4879,
-      "eval_samples_per_second": 10.473,
-      "eval_steps_per_second": 10.473,
-      "step": 4200
-    },
-    {
-      "epoch": 0.11,
-      "grad_norm": 2.2854669094085693,
-      "learning_rate": 1.6102564102564105e-05,
-      "loss": 0.9649,
-      "step": 4300
-    },
-    {
-      "epoch": 0.11,
-      "eval_loss": 0.959586501121521,
-      "eval_runtime": 95.5543,
-      "eval_samples_per_second": 10.465,
-      "eval_steps_per_second": 10.465,
-      "step": 4300
-    },
-    {
-      "epoch": 0.11,
-      "grad_norm": 4.13587760925293,
-      "learning_rate": 1.6000000000000003e-05,
-      "loss": 0.9766,
-      "step": 4400
-    },
-    {
-      "epoch": 0.11,
-      "eval_loss": 0.9532430171966553,
-      "eval_runtime": 95.5431,
-      "eval_samples_per_second": 10.466,
-      "eval_steps_per_second": 10.466,
-      "step": 4400
-    },
-    {
-      "epoch": 0.11,
-      "grad_norm": 4.551805019378662,
-      "learning_rate": 1.5897435897435897e-05,
-      "loss": 0.9286,
-      "step": 4500
-    },
-    {
-      "epoch": 0.11,
-      "eval_loss": 0.9443392157554626,
-      "eval_runtime": 93.5542,
-      "eval_samples_per_second": 10.689,
-      "eval_steps_per_second": 10.689,
-      "step": 4500
-    },
-    {
-      "epoch": 0.12,
-      "grad_norm": 3.9031989574432373,
-      "learning_rate": 1.5794871794871795e-05,
-      "loss": 0.9291,
-      "step": 4600
-    },
-    {
-      "epoch": 0.12,
-      "eval_loss": 0.9149237871170044,
-      "eval_runtime": 93.5444,
-      "eval_samples_per_second": 10.69,
-      "eval_steps_per_second": 10.69,
-      "step": 4600
-    },
-    {
-      "epoch": 0.12,
-      "grad_norm": 8.365277290344238,
-      "learning_rate": 1.5692307692307693e-05,
-      "loss": 0.8988,
-      "step": 4700
-    },
-    {
-      "epoch": 0.12,
-      "eval_loss": 0.905302107334137,
-      "eval_runtime": 93.5589,
-      "eval_samples_per_second": 10.688,
-      "eval_steps_per_second": 10.688,
-      "step": 4700
-    },
-    {
-      "epoch": 0.12,
-      "grad_norm": 4.9096503257751465,
-      "learning_rate": 1.558974358974359e-05,
-      "loss": 0.9506,
-      "step": 4800
-    },
-    {
-      "epoch": 0.12,
-      "eval_loss": 0.9025529623031616,
-      "eval_runtime": 93.5426,
-      "eval_samples_per_second": 10.69,
-      "eval_steps_per_second": 10.69,
-      "step": 4800
-    },
-    {
-      "epoch": 0.12,
-      "grad_norm": 3.766132354736328,
-      "learning_rate": 1.548717948717949e-05,
-      "loss": 0.9171,
-      "step": 4900
-    },
-    {
-      "epoch": 0.12,
-      "eval_loss": 0.9122710227966309,
-      "eval_runtime": 93.5323,
-      "eval_samples_per_second": 10.691,
-      "eval_steps_per_second": 10.691,
-      "step": 4900
-    },
-    {
-      "epoch": 0.12,
-      "grad_norm": 3.239846706390381,
-      "learning_rate": 1.5384615384615387e-05,
-      "loss": 0.9136,
-      "step": 5000
-    },
-    {
-      "epoch": 0.12,
-      "eval_loss": 0.9205535650253296,
-      "eval_runtime": 93.5448,
-      "eval_samples_per_second": 10.69,
-      "eval_steps_per_second": 10.69,
-      "step": 5000
-    },
-    {
-      "epoch": 0.13,
-      "grad_norm": 3.1556451320648193,
-      "learning_rate": 1.5282051282051282e-05,
-      "loss": 0.8966,
-      "step": 5100
-    },
-    {
-      "epoch": 0.13,
-      "eval_loss": 0.900290846824646,
-      "eval_runtime": 93.5687,
-      "eval_samples_per_second": 10.687,
-      "eval_steps_per_second": 10.687,
-      "step": 5100
-    },
-    {
-      "epoch": 0.13,
-      "grad_norm": 5.641241073608398,
-      "learning_rate": 1.517948717948718e-05,
-      "loss": 0.8843,
-      "step": 5200
-    },
-    {
-      "epoch": 0.13,
-      "eval_loss": 0.8797227740287781,
-      "eval_runtime": 93.5479,
-      "eval_samples_per_second": 10.69,
-      "eval_steps_per_second": 10.69,
-      "step": 5200
-    },
-    {
-      "epoch": 0.13,
-      "grad_norm": 18.91106605529785,
-      "learning_rate": 1.5076923076923078e-05,
-      "loss": 0.8859,
-      "step": 5300
-    },
-    {
-      "epoch": 0.13,
-      "eval_loss": 0.9091231822967529,
-      "eval_runtime": 93.5498,
-      "eval_samples_per_second": 10.689,
-      "eval_steps_per_second": 10.689,
-      "step": 5300
-    },
-    {
-      "epoch": 0.14,
-      "grad_norm": 9.203217506408691,
-      "learning_rate": 1.4974358974358976e-05,
-      "loss": 0.9137,
-      "step": 5400
-    },
-    {
-      "epoch": 0.14,
-      "eval_loss": 0.8882608413696289,
-      "eval_runtime": 93.5656,
-      "eval_samples_per_second": 10.688,
-      "eval_steps_per_second": 10.688,
-      "step": 5400
-    },
-    {
-      "epoch": 0.14,
-      "grad_norm": 6.349595069885254,
-      "learning_rate": 1.4871794871794874e-05,
-      "loss": 0.8813,
-      "step": 5500
-    },
-    {
-      "epoch": 0.14,
-      "eval_loss": 0.8685780763626099,
-      "eval_runtime": 93.5448,
-      "eval_samples_per_second": 10.69,
-      "eval_steps_per_second": 10.69,
-      "step": 5500
-    },
-    {
-      "epoch": 0.14,
-      "grad_norm": 4.313300609588623,
-      "learning_rate": 1.4769230769230772e-05,
-      "loss": 0.8643,
-      "step": 5600
-    },
-    {
-      "epoch": 0.14,
-      "eval_loss": 0.8861460089683533,
-      "eval_runtime": 95.5778,
-      "eval_samples_per_second": 10.463,
-      "eval_steps_per_second": 10.463,
-      "step": 5600
-    },
-    {
-      "epoch": 0.14,
-      "grad_norm": 4.9385600090026855,
-      "learning_rate": 1.4666666666666666e-05,
-      "loss": 0.8805,
-      "step": 5700
-    },
-    {
-      "epoch": 0.14,
-      "eval_loss": 0.8668786883354187,
-      "eval_runtime": 95.511,
-      "eval_samples_per_second": 10.47,
-      "eval_steps_per_second": 10.47,
-      "step": 5700
-    },
-    {
-      "epoch": 0.14,
-      "grad_norm": 5.586972713470459,
-      "learning_rate": 1.4564102564102564e-05,
-      "loss": 0.877,
-      "step": 5800
-    },
-    {
-      "epoch": 0.14,
-      "eval_loss": 0.8643443584442139,
-      "eval_runtime": 95.4875,
-      "eval_samples_per_second": 10.473,
-      "eval_steps_per_second": 10.473,
-      "step": 5800
-    },
-    {
-      "epoch": 0.15,
-      "grad_norm": 7.318083763122559,
-      "learning_rate": 1.4461538461538462e-05,
-      "loss": 0.8917,
-      "step": 5900
-    },
-    {
-      "epoch": 0.15,
-      "eval_loss": 0.8765038847923279,
-      "eval_runtime": 95.5428,
-      "eval_samples_per_second": 10.467,
-      "eval_steps_per_second": 10.467,
-      "step": 5900
-    },
-    {
-      "epoch": 0.15,
-      "grad_norm": 4.376430034637451,
-      "learning_rate": 1.435897435897436e-05,
-      "loss": 0.8828,
-      "step": 6000
-    },
-    {
-      "epoch": 0.15,
-      "eval_loss": 0.8941754698753357,
-      "eval_runtime": 94.402,
-      "eval_samples_per_second": 10.593,
-      "eval_steps_per_second": 10.593,
-      "step": 6000
-    },
-    {
-      "epoch": 0.15,
-      "grad_norm": 3.1321024894714355,
-      "learning_rate": 1.4256410256410258e-05,
-      "loss": 0.867,
-      "step": 6100
-    },
-    {
-      "epoch": 0.15,
-      "eval_loss": 0.8598324060440063,
-      "eval_runtime": 93.5419,
-      "eval_samples_per_second": 10.69,
-      "eval_steps_per_second": 10.69,
-      "step": 6100
-    },
-    {
-      "epoch": 0.15,
-      "grad_norm": 3.5920615196228027,
-      "learning_rate": 1.4153846153846156e-05,
-      "loss": 0.869,
-      "step": 6200
-    },
-    {
-      "epoch": 0.15,
-      "eval_loss": 0.8688547611236572,
-      "eval_runtime": 93.5494,
-      "eval_samples_per_second": 10.69,
-      "eval_steps_per_second": 10.69,
-      "step": 6200
-    },
-    {
-      "epoch": 0.16,
-      "grad_norm": 3.6573245525360107,
-      "learning_rate": 1.405128205128205e-05,
-      "loss": 0.8899,
-      "step": 6300
-    },
-    {
-      "epoch": 0.16,
-      "eval_loss": 0.8593968749046326,
-      "eval_runtime": 93.5407,
-      "eval_samples_per_second": 10.691,
-      "eval_steps_per_second": 10.691,
-      "step": 6300
-    },
-    {
-      "epoch": 0.16,
-      "grad_norm": 5.975399971008301,
-      "learning_rate": 1.3948717948717949e-05,
-      "loss": 0.8573,
-      "step": 6400
-    },
-    {
-      "epoch": 0.16,
-      "eval_loss": 0.8987392783164978,
-      "eval_runtime": 93.5376,
-      "eval_samples_per_second": 10.691,
-      "eval_steps_per_second": 10.691,
-      "step": 6400
-    },
-    {
-      "epoch": 0.16,
-      "grad_norm": 6.997458457946777,
-      "learning_rate": 1.3846153846153847e-05,
-      "loss": 0.8447,
-      "step": 6500
-    },
-    {
-      "epoch": 0.16,
-      "eval_loss": 0.8606413006782532,
-      "eval_runtime": 93.5387,
-      "eval_samples_per_second": 10.691,
-      "eval_steps_per_second": 10.691,
-      "step": 6500
-    },
-    {
-      "epoch": 0.17,
-      "grad_norm": 3.1585702896118164,
-      "learning_rate": 1.3743589743589745e-05,
-      "loss": 0.8757,
-      "step": 6600
-    },
-    {
-      "epoch": 0.17,
-      "eval_loss": 0.8433266878128052,
-      "eval_runtime": 93.5234,
-      "eval_samples_per_second": 10.693,
-      "eval_steps_per_second": 10.693,
-      "step": 6600
-    },
-    {
-      "epoch": 0.17,
-      "grad_norm": 3.6385037899017334,
-      "learning_rate": 1.3641025641025643e-05,
-      "loss": 0.8609,
-      "step": 6700
-    },
-    {
-      "epoch": 0.17,
-      "eval_loss": 0.8342506289482117,
-      "eval_runtime": 93.5279,
-      "eval_samples_per_second": 10.692,
-      "eval_steps_per_second": 10.692,
-      "step": 6700
-    },
-    {
-      "epoch": 0.17,
-      "grad_norm": 3.0882773399353027,
-      "learning_rate": 1.353846153846154e-05,
-      "loss": 0.8602,
-      "step": 6800
-    },
-    {
-      "epoch": 0.17,
-      "eval_loss": 0.8522408604621887,
-      "eval_runtime": 93.5252,
-      "eval_samples_per_second": 10.692,
-      "eval_steps_per_second": 10.692,
-      "step": 6800
-    },
-    {
-      "epoch": 0.17,
-      "grad_norm": 7.343814849853516,
-      "learning_rate": 1.3435897435897435e-05,
-      "loss": 0.849,
-      "step": 6900
-    },
-    {
-      "epoch": 0.17,
-      "eval_loss": 0.8569708466529846,
-      "eval_runtime": 97.9814,
-      "eval_samples_per_second": 10.206,
-      "eval_steps_per_second": 10.206,
-      "step": 6900
+      "epoch": 0.0,
+      "grad_norm": 6.34375,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 0.6644,
+      "step": 30
     },
     {
-      "epoch": 0.17,
-      "grad_norm": 4.511021614074707,
-      "learning_rate": 1.3333333333333333e-05,
-      "loss": 0.8304,
-      "step": 7000
+      "epoch": 0.0,
+      "grad_norm": 2.359375,
+      "learning_rate": 1.6000000000000001e-06,
+      "loss": 0.6101,
+      "step": 40
     },
     {
-      "epoch": 0.17,
-      "eval_loss": 0.8397101759910583,
-      "eval_runtime": 104.4026,
-      "eval_samples_per_second": 9.578,
-      "eval_steps_per_second": 9.578,
-      "step": 7000
+      "epoch": 0.0,
+      "grad_norm": 4.90625,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.6188,
+      "step": 50
     },
     {
-      "epoch": 0.18,
-      "grad_norm": 5.263293266296387,
-      "learning_rate": 1.3230769230769231e-05,
-      "loss": 0.8159,
-      "step": 7100
+      "epoch": 0.0,
+      "grad_norm": 3.390625,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 0.6524,
+      "step": 60
     },
     {
-      "epoch": 0.18,
-      "eval_loss": 0.8445966839790344,
-      "eval_runtime": 104.2536,
-      "eval_samples_per_second": 9.592,
-      "eval_steps_per_second": 9.592,
-      "step": 7100
+      "epoch": 0.0,
+      "grad_norm": 3.75,
+      "learning_rate": 2.8000000000000003e-06,
+      "loss": 0.6322,
+      "step": 70
     },
     {
-      "epoch": 0.18,
-      "grad_norm": 3.746361017227173,
-      "learning_rate": 1.312820512820513e-05,
-      "loss": 0.8078,
-      "step": 7200
+      "epoch": 0.0,
+      "grad_norm": 3.484375,
+      "learning_rate": 3.2000000000000003e-06,
+      "loss": 0.676,
+      "step": 80
     },
     {
-      "epoch": 0.18,
-      "eval_loss": 0.8426964282989502,
-      "eval_runtime": 104.3849,
-      "eval_samples_per_second": 9.58,
-      "eval_steps_per_second": 9.58,
-      "step": 7200
+      "epoch": 0.0,
+      "grad_norm": 4.40625,
+      "learning_rate": 3.6000000000000003e-06,
+      "loss": 0.6578,
+      "step": 90
     },
     {
-      "epoch": 0.18,
-      "grad_norm": 5.570109844207764,
-      "learning_rate": 1.3025641025641027e-05,
-      "loss": 0.8289,
-      "step": 7300
+      "epoch": 0.0,
+      "grad_norm": 4.25,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.597,
+      "step": 100
     },
     {
-      "epoch": 0.18,
-      "eval_loss": 0.842907190322876,
-      "eval_runtime": 101.6488,
-      "eval_samples_per_second": 9.838,
-      "eval_steps_per_second": 9.838,
-      "step": 7300
+      "epoch": 0.0,
+      "grad_norm": 4.4375,
+      "learning_rate": 4.4e-06,
+      "loss": 0.6543,
+      "step": 110
     },
     {
-      "epoch": 0.18,
-      "grad_norm": 4.563759803771973,
-      "learning_rate": 1.2923076923076925e-05,
-      "loss": 0.861,
-      "step": 7400
+      "epoch": 0.0,
+      "grad_norm": 7.03125,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 0.6844,
+      "step": 120
     },
     {
-      "epoch": 0.18,
-      "eval_loss": 0.8268536925315857,
-      "eval_runtime": 93.6228,
-      "eval_samples_per_second": 10.681,
-      "eval_steps_per_second": 10.681,
-      "step": 7400
+      "epoch": 0.0,
+      "grad_norm": 6.09375,
+      "learning_rate": 5.2e-06,
+      "loss": 0.6658,
+      "step": 130
     },
     {
-      "epoch": 0.19,
-      "grad_norm": 3.188389301300049,
-      "learning_rate": 1.2820512820512823e-05,
-      "loss": 0.8473,
-      "step": 7500
+      "epoch": 0.0,
+      "grad_norm": 2.96875,
+      "learning_rate": 5.600000000000001e-06,
+      "loss": 0.7388,
+      "step": 140
     },
     {
-      "epoch": 0.19,
-      "eval_loss": 0.8434333205223083,
-      "eval_runtime": 93.9123,
-      "eval_samples_per_second": 10.648,
-      "eval_steps_per_second": 10.648,
-      "step": 7500
+      "epoch": 0.0,
+      "grad_norm": 4.5625,
+      "learning_rate": 6e-06,
+      "loss": 0.6627,
+      "step": 150
     },
     {
-      "epoch": 0.19,
-      "grad_norm": 2.5115840435028076,
-      "learning_rate": 1.2717948717948718e-05,
-      "loss": 0.855,
-      "step": 7600
+      "epoch": 0.0,
+      "grad_norm": 8.125,
+      "learning_rate": 6.4000000000000006e-06,
+      "loss": 0.7036,
+      "step": 160
     },
     {
-      "epoch": 0.19,
-      "eval_loss": 0.8359225988388062,
-      "eval_runtime": 98.358,
-      "eval_samples_per_second": 10.167,
-      "eval_steps_per_second": 10.167,
-      "step": 7600
+      "epoch": 0.0,
+      "grad_norm": 4.9375,
+      "learning_rate": 6.800000000000001e-06,
+      "loss": 0.6136,
+      "step": 170
     },
     {
-      "epoch": 0.19,
-      "grad_norm": 4.867124080657959,
-      "learning_rate": 1.2615384615384616e-05,
-      "loss": 0.8358,
-      "step": 7700
+      "epoch": 0.0,
+      "grad_norm": 2.546875,
+      "learning_rate": 7.2000000000000005e-06,
+      "loss": 0.5902,
+      "step": 180
     },
     {
-      "epoch": 0.19,
-      "eval_loss": 0.8118915557861328,
-      "eval_runtime": 98.3121,
-      "eval_samples_per_second": 10.172,
-      "eval_steps_per_second": 10.172,
-      "step": 7700
+      "epoch": 0.0,
+      "grad_norm": 2.265625,
+      "learning_rate": 7.600000000000001e-06,
+      "loss": 0.6254,
+      "step": 190
     },
     {
-      "epoch": 0.2,
-      "grad_norm": 3.2623794078826904,
-      "learning_rate": 1.2512820512820514e-05,
-      "loss": 0.81,
-      "step": 7800
+      "epoch": 0.01,
+      "grad_norm": 8.9375,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.6322,
+      "step": 200
     },
     {
-      "epoch": 0.2,
-      "eval_loss": 0.8363373279571533,
-      "eval_runtime": 98.3201,
-      "eval_samples_per_second": 10.171,
-      "eval_steps_per_second": 10.171,
-      "step": 7800
+      "epoch": 0.01,
+      "grad_norm": 2.546875,
+      "learning_rate": 8.400000000000001e-06,
+      "loss": 0.6152,
+      "step": 210
     },
     {
-      "epoch": 0.2,
-      "grad_norm": 3.2750914096832275,
-      "learning_rate": 1.2410256410256412e-05,
-      "loss": 0.8187,
-      "step": 7900
+      "epoch": 0.01,
+      "grad_norm": 5.15625,
+      "learning_rate": 8.8e-06,
+      "loss": 0.5843,
+      "step": 220
     },
     {
-      "epoch": 0.2,
-      "eval_loss": 0.8370142579078674,
-      "eval_runtime": 98.2793,
-      "eval_samples_per_second": 10.175,
-      "eval_steps_per_second": 10.175,
-      "step": 7900
+      "epoch": 0.01,
+      "grad_norm": 5.96875,
+      "learning_rate": 9.200000000000002e-06,
+      "loss": 0.7049,
+      "step": 230
     },
     {
-      "epoch": 0.2,
-      "grad_norm": 4.422469139099121,
-      "learning_rate": 1.230769230769231e-05,
-      "loss": 0.8315,
-      "step": 8000
+      "epoch": 0.01,
+      "grad_norm": 4.34375,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 0.4978,
+      "step": 240
     },
     {
-      "epoch": 0.2,
-      "eval_loss": 0.8418492674827576,
-      "eval_runtime": 98.2738,
-      "eval_samples_per_second": 10.176,
-      "eval_steps_per_second": 10.176,
-      "step": 8000
+      "epoch": 0.01,
+      "grad_norm": 3.6875,
+      "learning_rate": 1e-05,
+      "loss": 0.603,
+      "step": 250
     },
     {
-      "epoch": 0.2,
-      "grad_norm": 2.8852996826171875,
-      "learning_rate": 1.2205128205128208e-05,
-      "loss": 0.7983,
-      "step": 8100
+      "epoch": 0.01,
+      "grad_norm": 7.0625,
+      "learning_rate": 1.04e-05,
+      "loss": 0.6845,
+      "step": 260
     },
     {
-      "epoch": 0.2,
-      "eval_loss": 0.8371080160140991,
-      "eval_runtime": 98.3307,
-      "eval_samples_per_second": 10.17,
-      "eval_steps_per_second": 10.17,
-      "step": 8100
+      "epoch": 0.01,
+      "grad_norm": 9.9375,
+      "learning_rate": 1.0800000000000002e-05,
+      "loss": 0.5937,
+      "step": 270
     },
     {
-      "epoch": 0.2,
-      "grad_norm": 7.2119903564453125,
-      "learning_rate": 1.2102564102564102e-05,
-      "loss": 0.8363,
-      "step": 8200
+      "epoch": 0.01,
+      "grad_norm": 5.6875,
+      "learning_rate": 1.1200000000000001e-05,
+      "loss": 0.5744,
+      "step": 280
     },
     {
-      "epoch": 0.2,
-      "eval_loss": 0.8220362663269043,
-      "eval_runtime": 98.3411,
-      "eval_samples_per_second": 10.169,
-      "eval_steps_per_second": 10.169,
-      "step": 8200
+      "epoch": 0.01,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.16e-05,
+      "loss": 0.7565,
+      "step": 290
     },
     {
-      "epoch": 0.21,
-      "grad_norm": 4.324320316314697,
+      "epoch": 0.01,
+      "grad_norm": 8.125,
       "learning_rate": 1.2e-05,
-      "loss": 0.784,
-      "step": 8300
-    },
-    {
-      "epoch": 0.21,
-      "eval_loss": 0.8121130466461182,
-      "eval_runtime": 98.3947,
-      "eval_samples_per_second": 10.163,
-      "eval_steps_per_second": 10.163,
-      "step": 8300
-    },
-    {
-      "epoch": 0.21,
-      "grad_norm": 7.664273262023926,
-      "learning_rate": 1.1897435897435898e-05,
-      "loss": 0.7844,
-      "step": 8400
-    },
-    {
-      "epoch": 0.21,
-      "eval_loss": 0.8250142931938171,
-      "eval_runtime": 98.4654,
-      "eval_samples_per_second": 10.156,
-      "eval_steps_per_second": 10.156,
-      "step": 8400
-    },
-    {
-      "epoch": 0.21,
-      "grad_norm": 4.998061656951904,
-      "learning_rate": 1.1794871794871796e-05,
-      "loss": 0.7711,
-      "step": 8500
-    },
-    {
-      "epoch": 0.21,
-      "eval_loss": 0.8134967684745789,
-      "eval_runtime": 98.4327,
-      "eval_samples_per_second": 10.159,
-      "eval_steps_per_second": 10.159,
-      "step": 8500
-    },
-    {
-      "epoch": 0.21,
-      "grad_norm": 3.266608476638794,
-      "learning_rate": 1.1692307692307694e-05,
-      "loss": 0.8133,
-      "step": 8600
-    },
-    {
-      "epoch": 0.21,
-      "eval_loss": 0.7890066504478455,
-      "eval_runtime": 98.4322,
-      "eval_samples_per_second": 10.159,
-      "eval_steps_per_second": 10.159,
-      "step": 8600
-    },
-    {
-      "epoch": 0.22,
-      "grad_norm": 2.0603463649749756,
-      "learning_rate": 1.1589743589743592e-05,
-      "loss": 0.8355,
-      "step": 8700
-    },
-    {
-      "epoch": 0.22,
-      "eval_loss": 0.8279140591621399,
-      "eval_runtime": 98.4691,
-      "eval_samples_per_second": 10.155,
-      "eval_steps_per_second": 10.155,
-      "step": 8700
-    },
-    {
-      "epoch": 0.22,
-      "grad_norm": 3.618032693862915,
-      "learning_rate": 1.1487179487179487e-05,
-      "loss": 0.8158,
-      "step": 8800
+      "loss": 0.6332,
+      "step": 300
     },
     {
-      "epoch": 0.22,
-      "eval_loss": 0.805887758731842,
-      "eval_runtime": 98.4492,
-      "eval_samples_per_second": 10.158,
-      "eval_steps_per_second": 10.158,
-      "step": 8800
+      "epoch": 0.01,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.2400000000000002e-05,
+      "loss": 0.5884,
+      "step": 310
     },
     {
-      "epoch": 0.22,
-      "grad_norm": 3.650794267654419,
-      "learning_rate": 1.1384615384615385e-05,
-      "loss": 0.7972,
-      "step": 8900
+      "epoch": 0.01,
+      "grad_norm": 5.625,
+      "learning_rate": 1.2800000000000001e-05,
+      "loss": 0.6418,
+      "step": 320
     },
     {
-      "epoch": 0.22,
-      "eval_loss": 0.7881203293800354,
-      "eval_runtime": 98.5045,
-      "eval_samples_per_second": 10.152,
-      "eval_steps_per_second": 10.152,
-      "step": 8900
+      "epoch": 0.01,
+      "grad_norm": 2.421875,
+      "learning_rate": 1.3200000000000002e-05,
+      "loss": 0.6532,
+      "step": 330
     },
     {
-      "epoch": 0.23,
-      "grad_norm": 2.8892910480499268,
-      "learning_rate": 1.1282051282051283e-05,
-      "loss": 0.8165,
-      "step": 9000
+      "epoch": 0.01,
+      "grad_norm": 4.125,
+      "learning_rate": 1.3600000000000002e-05,
+      "loss": 0.5984,
+      "step": 340
     },
     {
-      "epoch": 0.23,
-      "eval_loss": 0.7988776564598083,
-      "eval_runtime": 98.4585,
-      "eval_samples_per_second": 10.157,
-      "eval_steps_per_second": 10.157,
-      "step": 9000
+      "epoch": 0.01,
+      "grad_norm": 5.4375,
+      "learning_rate": 1.4e-05,
+      "loss": 0.7039,
+      "step": 350
     },
     {
-      "epoch": 0.23,
-      "grad_norm": 3.109771728515625,
-      "learning_rate": 1.117948717948718e-05,
-      "loss": 0.8048,
-      "step": 9100
+      "epoch": 0.01,
+      "grad_norm": 4.15625,
+      "learning_rate": 1.4400000000000001e-05,
+      "loss": 0.5297,
+      "step": 360
     },
     {
-      "epoch": 0.23,
-      "eval_loss": 0.8003994226455688,
-      "eval_runtime": 98.471,
-      "eval_samples_per_second": 10.155,
-      "eval_steps_per_second": 10.155,
-      "step": 9100
+      "epoch": 0.01,
+      "grad_norm": 2.71875,
+      "learning_rate": 1.48e-05,
+      "loss": 0.6239,
+      "step": 370
     },
     {
-      "epoch": 0.23,
-      "grad_norm": 4.004971504211426,
-      "learning_rate": 1.1076923076923079e-05,
-      "loss": 0.7855,
-      "step": 9200
+      "epoch": 0.01,
+      "grad_norm": 4.5,
+      "learning_rate": 1.5200000000000002e-05,
+      "loss": 0.6246,
+      "step": 380
     },
     {
-      "epoch": 0.23,
-      "eval_loss": 0.7879841923713684,
-      "eval_runtime": 98.3853,
-      "eval_samples_per_second": 10.164,
-      "eval_steps_per_second": 10.164,
-      "step": 9200
+      "epoch": 0.01,
+      "grad_norm": 3.421875,
+      "learning_rate": 1.5600000000000003e-05,
+      "loss": 0.6023,
+      "step": 390
     },
     {
-      "epoch": 0.23,
-      "grad_norm": 4.2928786277771,
-      "learning_rate": 1.0974358974358977e-05,
-      "loss": 0.8008,
-      "step": 9300
+      "epoch": 0.01,
+      "grad_norm": 2.671875,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.6448,
+      "step": 400
     },
     {
-      "epoch": 0.23,
-      "eval_loss": 0.8041646480560303,
-      "eval_runtime": 98.4192,
-      "eval_samples_per_second": 10.161,
-      "eval_steps_per_second": 10.161,
-      "step": 9300
+      "epoch": 0.01,
+      "grad_norm": 14.75,
+      "learning_rate": 1.64e-05,
+      "loss": 0.7238,
+      "step": 410
     },
     {
-      "epoch": 0.23,
-      "grad_norm": 3.7064008712768555,
-      "learning_rate": 1.0871794871794871e-05,
-      "loss": 0.809,
-      "step": 9400
+      "epoch": 0.01,
+      "grad_norm": 4.1875,
+      "learning_rate": 1.6800000000000002e-05,
+      "loss": 0.6137,
+      "step": 420
     },
     {
-      "epoch": 0.23,
-      "eval_loss": 0.7951182126998901,
-      "eval_runtime": 98.4042,
-      "eval_samples_per_second": 10.162,
-      "eval_steps_per_second": 10.162,
-      "step": 9400
+      "epoch": 0.01,
+      "grad_norm": 2.828125,
+      "learning_rate": 1.72e-05,
+      "loss": 0.6496,
+      "step": 430
     },
     {
-      "epoch": 0.24,
-      "grad_norm": 4.159609317779541,
-      "learning_rate": 1.076923076923077e-05,
-      "loss": 0.7763,
-      "step": 9500
+      "epoch": 0.01,
+      "grad_norm": 8.1875,
+      "learning_rate": 1.76e-05,
+      "loss": 0.5152,
+      "step": 440
     },
     {
-      "epoch": 0.24,
-      "eval_loss": 0.7955417633056641,
-      "eval_runtime": 98.4578,
-      "eval_samples_per_second": 10.157,
-      "eval_steps_per_second": 10.157,
-      "step": 9500
+      "epoch": 0.01,
+      "grad_norm": 7.5625,
+      "learning_rate": 1.8e-05,
+      "loss": 0.5874,
+      "step": 450
     },
     {
-      "epoch": 0.24,
-      "grad_norm": 2.562718152999878,
-      "learning_rate": 1.0666666666666667e-05,
-      "loss": 0.7887,
-      "step": 9600
+      "epoch": 0.01,
+      "grad_norm": 11.875,
+      "learning_rate": 1.8400000000000003e-05,
+      "loss": 0.6662,
+      "step": 460
     },
     {
-      "epoch": 0.24,
-      "eval_loss": 0.8063541054725647,
-      "eval_runtime": 98.4827,
-      "eval_samples_per_second": 10.154,
-      "eval_steps_per_second": 10.154,
-      "step": 9600
+      "epoch": 0.01,
+      "grad_norm": 3.515625,
+      "learning_rate": 1.88e-05,
+      "loss": 0.6845,
+      "step": 470
     },
     {
-      "epoch": 0.24,
-      "grad_norm": 3.551567316055298,
-      "learning_rate": 1.0564102564102565e-05,
-      "loss": 0.8269,
-      "step": 9700
+      "epoch": 0.01,
+      "grad_norm": 10.5625,
+      "learning_rate": 1.9200000000000003e-05,
+      "loss": 0.6917,
+      "step": 480
     },
     {
-      "epoch": 0.24,
-      "eval_loss": 0.7908796668052673,
-      "eval_runtime": 98.4538,
-      "eval_samples_per_second": 10.157,
-      "eval_steps_per_second": 10.157,
-      "step": 9700
+      "epoch": 0.01,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.9600000000000002e-05,
+      "loss": 0.5512,
+      "step": 490
     },
     {
-      "epoch": 0.24,
-      "grad_norm": 4.210842609405518,
-      "learning_rate": 1.0461538461538463e-05,
-      "loss": 0.7872,
-      "step": 9800
+      "epoch": 0.01,
+      "grad_norm": 2.546875,
+      "learning_rate": 2e-05,
+      "loss": 0.6037,
+      "step": 500
     },
     {
-      "epoch": 0.24,
-      "eval_loss": 0.7888299822807312,
-      "eval_runtime": 98.4173,
-      "eval_samples_per_second": 10.161,
-      "eval_steps_per_second": 10.161,
-      "step": 9800
+      "epoch": 0.01,
+      "eval_loss": 0.6478177309036255,
+      "eval_runtime": 38.0753,
+      "eval_samples_per_second": 26.264,
+      "eval_steps_per_second": 26.264,
+      "step": 500
     },
     {
-      "epoch": 0.25,
-      "grad_norm": 3.1311933994293213,
-      "learning_rate": 1.0358974358974361e-05,
-      "loss": 0.7791,
-      "step": 9900
+      "epoch": 0.01,
+      "grad_norm": 5.125,
+      "learning_rate": 1.9978947368421054e-05,
+      "loss": 0.6223,
+      "step": 510
     },
     {
-      "epoch": 0.25,
-      "eval_loss": 0.782675564289093,
-      "eval_runtime": 98.435,
-      "eval_samples_per_second": 10.159,
-      "eval_steps_per_second": 10.159,
-      "step": 9900
+      "epoch": 0.01,
+      "grad_norm": 3.4375,
+      "learning_rate": 1.9957894736842107e-05,
+      "loss": 0.6301,
+      "step": 520
     },
     {
-      "epoch": 0.25,
-      "grad_norm": 3.9567008018493652,
-      "learning_rate": 1.0256410256410256e-05,
-      "loss": 0.7461,
-      "step": 10000
+      "epoch": 0.01,
+      "grad_norm": 7.65625,
+      "learning_rate": 1.993684210526316e-05,
+      "loss": 0.6224,
+      "step": 530
     },
     {
-      "epoch": 0.25,
-      "eval_loss": 0.7776587009429932,
-      "eval_runtime": 98.4484,
-      "eval_samples_per_second": 10.158,
-      "eval_steps_per_second": 10.158,
-      "step": 10000
+      "epoch": 0.01,
+      "grad_norm": 9.125,
+      "learning_rate": 1.9915789473684212e-05,
+      "loss": 0.6255,
+      "step": 540
     },
     {
-      "epoch": 0.25,
-      "grad_norm": 4.541276454925537,
-      "learning_rate": 1.0153846153846154e-05,
-      "loss": 0.7732,
-      "step": 10100
+      "epoch": 0.01,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.9894736842105265e-05,
+      "loss": 0.6737,
+      "step": 550
     },
     {
-      "epoch": 0.25,
-      "eval_loss": 0.7786129117012024,
-      "eval_runtime": 98.4299,
-      "eval_samples_per_second": 10.16,
-      "eval_steps_per_second": 10.16,
-      "step": 10100
+      "epoch": 0.01,
+      "grad_norm": 11.8125,
+      "learning_rate": 1.9873684210526318e-05,
+      "loss": 0.6218,
+      "step": 560
     },
     {
-      "epoch": 0.26,
-      "grad_norm": 3.852530002593994,
-      "learning_rate": 1.0051282051282052e-05,
-      "loss": 0.7708,
-      "step": 10200
+      "epoch": 0.01,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.985263157894737e-05,
+      "loss": 0.5187,
+      "step": 570
     },
     {
-      "epoch": 0.26,
-      "eval_loss": 0.7804587483406067,
-      "eval_runtime": 98.4657,
-      "eval_samples_per_second": 10.156,
-      "eval_steps_per_second": 10.156,
-      "step": 10200
+      "epoch": 0.01,
+      "grad_norm": 6.0625,
+      "learning_rate": 1.9831578947368423e-05,
+      "loss": 0.6786,
+      "step": 580
     },
     {
-      "epoch": 0.26,
-      "grad_norm": 4.2783379554748535,
-      "learning_rate": 9.94871794871795e-06,
-      "loss": 0.7418,
-      "step": 10300
+      "epoch": 0.01,
+      "grad_norm": 5.125,
+      "learning_rate": 1.9810526315789476e-05,
+      "loss": 0.6641,
+      "step": 590
     },
     {
-      "epoch": 0.26,
-      "eval_loss": 0.7932974696159363,
-      "eval_runtime": 98.4438,
-      "eval_samples_per_second": 10.158,
-      "eval_steps_per_second": 10.158,
-      "step": 10300
+      "epoch": 0.01,
+      "grad_norm": 3.75,
+      "learning_rate": 1.9789473684210528e-05,
+      "loss": 0.5541,
+      "step": 600
     },
     {
-      "epoch": 0.26,
-      "grad_norm": 4.555493354797363,
-      "learning_rate": 9.846153846153848e-06,
-      "loss": 0.789,
-      "step": 10400
+      "epoch": 0.02,
+      "grad_norm": 4.4375,
+      "learning_rate": 1.976842105263158e-05,
+      "loss": 0.565,
+      "step": 610
     },
     {
-      "epoch": 0.26,
-      "eval_loss": 0.7759644985198975,
-      "eval_runtime": 98.452,
-      "eval_samples_per_second": 10.157,
-      "eval_steps_per_second": 10.157,
-      "step": 10400
+      "epoch": 0.02,
+      "grad_norm": 10.375,
+      "learning_rate": 1.9747368421052633e-05,
+      "loss": 0.7099,
+      "step": 620
     },
     {
-      "epoch": 0.26,
-      "grad_norm": 3.638544797897339,
-      "learning_rate": 9.743589743589744e-06,
-      "loss": 0.7575,
-      "step": 10500
+      "epoch": 0.02,
+      "grad_norm": 5.09375,
+      "learning_rate": 1.9726315789473686e-05,
+      "loss": 0.5988,
+      "step": 630
     },
     {
-      "epoch": 0.26,
-      "eval_loss": 0.7742583751678467,
-      "eval_runtime": 98.4572,
-      "eval_samples_per_second": 10.157,
-      "eval_steps_per_second": 10.157,
-      "step": 10500
+      "epoch": 0.02,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.970526315789474e-05,
+      "loss": 0.7724,
+      "step": 640
     },
     {
-      "epoch": 0.27,
-      "grad_norm": 4.139896869659424,
-      "learning_rate": 9.641025641025642e-06,
-      "loss": 0.7895,
-      "step": 10600
+      "epoch": 0.02,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.968421052631579e-05,
+      "loss": 0.582,
+      "step": 650
     },
     {
-      "epoch": 0.27,
-      "eval_loss": 0.7806645035743713,
-      "eval_runtime": 98.4502,
-      "eval_samples_per_second": 10.157,
-      "eval_steps_per_second": 10.157,
-      "step": 10600
+      "epoch": 0.02,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.9663157894736844e-05,
+      "loss": 0.6805,
+      "step": 660
     },
     {
-      "epoch": 0.27,
-      "grad_norm": 6.0768866539001465,
-      "learning_rate": 9.53846153846154e-06,
-      "loss": 0.7464,
-      "step": 10700
+      "epoch": 0.02,
+      "grad_norm": 11.0,
+      "learning_rate": 1.9642105263157897e-05,
+      "loss": 0.6248,
+      "step": 670
     },
     {
-      "epoch": 0.27,
-      "eval_loss": 0.7633421421051025,
-      "eval_runtime": 98.4913,
-      "eval_samples_per_second": 10.153,
-      "eval_steps_per_second": 10.153,
-      "step": 10700
+      "epoch": 0.02,
+      "grad_norm": 2.734375,
+      "learning_rate": 1.962105263157895e-05,
+      "loss": 0.5691,
+      "step": 680
     },
     {
-      "epoch": 0.27,
-      "grad_norm": 4.529891014099121,
-      "learning_rate": 9.435897435897436e-06,
-      "loss": 0.7878,
-      "step": 10800
+      "epoch": 0.02,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.9600000000000002e-05,
+      "loss": 0.6657,
+      "step": 690
     },
     {
-      "epoch": 0.27,
-      "eval_loss": 0.7459210157394409,
-      "eval_runtime": 98.4314,
-      "eval_samples_per_second": 10.159,
-      "eval_steps_per_second": 10.159,
-      "step": 10800
+      "epoch": 0.02,
+      "grad_norm": 2.078125,
+      "learning_rate": 1.9578947368421055e-05,
+      "loss": 0.4786,
+      "step": 700
     },
     {
-      "epoch": 0.27,
-      "grad_norm": 4.402657508850098,
-      "learning_rate": 9.333333333333334e-06,
-      "loss": 0.734,
-      "step": 10900
+      "epoch": 0.02,
+      "grad_norm": 7.125,
+      "learning_rate": 1.9557894736842107e-05,
+      "loss": 0.6368,
+      "step": 710
     },
     {
-      "epoch": 0.27,
-      "eval_loss": 0.7673327922821045,
-      "eval_runtime": 98.4432,
-      "eval_samples_per_second": 10.158,
-      "eval_steps_per_second": 10.158,
-      "step": 10900
+      "epoch": 0.02,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.953684210526316e-05,
+      "loss": 0.6108,
+      "step": 720
     },
     {
-      "epoch": 0.28,
-      "grad_norm": 3.691208839416504,
-      "learning_rate": 9.230769230769232e-06,
-      "loss": 0.7504,
-      "step": 11000
+      "epoch": 0.02,
+      "grad_norm": 3.53125,
+      "learning_rate": 1.9515789473684213e-05,
+      "loss": 0.6331,
+      "step": 730
     },
     {
-      "epoch": 0.28,
-      "eval_loss": 0.7593997716903687,
-      "eval_runtime": 98.4682,
-      "eval_samples_per_second": 10.156,
-      "eval_steps_per_second": 10.156,
-      "step": 11000
+      "epoch": 0.02,
+      "grad_norm": 3.96875,
+      "learning_rate": 1.9494736842105265e-05,
+      "loss": 0.5432,
+      "step": 740
     },
     {
-      "epoch": 0.28,
-      "grad_norm": 3.411905527114868,
-      "learning_rate": 9.128205128205129e-06,
-      "loss": 0.7901,
-      "step": 11100
+      "epoch": 0.02,
+      "grad_norm": 4.25,
+      "learning_rate": 1.9473684210526318e-05,
+      "loss": 0.7021,
+      "step": 750
     },
     {
-      "epoch": 0.28,
-      "eval_loss": 0.7573875784873962,
-      "eval_runtime": 98.4374,
-      "eval_samples_per_second": 10.159,
-      "eval_steps_per_second": 10.159,
-      "step": 11100
+      "epoch": 0.02,
+      "grad_norm": 10.375,
+      "learning_rate": 1.945263157894737e-05,
+      "loss": 0.68,
+      "step": 760
     },
     {
-      "epoch": 0.28,
-      "grad_norm": 15.492389678955078,
-      "learning_rate": 9.025641025641027e-06,
-      "loss": 0.7268,
-      "step": 11200
+      "epoch": 0.02,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.9431578947368423e-05,
+      "loss": 0.629,
+      "step": 770
     },
     {
-      "epoch": 0.28,
-      "eval_loss": 0.7712795734405518,
-      "eval_runtime": 98.4057,
-      "eval_samples_per_second": 10.162,
-      "eval_steps_per_second": 10.162,
-      "step": 11200
+      "epoch": 0.02,
+      "grad_norm": 4.5625,
+      "learning_rate": 1.9410526315789476e-05,
+      "loss": 0.6685,
+      "step": 780
     },
     {
-      "epoch": 0.28,
-      "grad_norm": 3.4170589447021484,
-      "learning_rate": 8.923076923076925e-06,
-      "loss": 0.7655,
-      "step": 11300
+      "epoch": 0.02,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.9389473684210525e-05,
+      "loss": 0.5552,
+      "step": 790
     },
     {
-      "epoch": 0.28,
-      "eval_loss": 0.7508448958396912,
-      "eval_runtime": 98.45,
-      "eval_samples_per_second": 10.157,
-      "eval_steps_per_second": 10.157,
-      "step": 11300
+      "epoch": 0.02,
+      "grad_norm": 2.765625,
+      "learning_rate": 1.936842105263158e-05,
+      "loss": 0.657,
+      "step": 800
     },
     {
-      "epoch": 0.28,
-      "grad_norm": 7.331748962402344,
-      "learning_rate": 8.820512820512821e-06,
-      "loss": 0.7459,
-      "step": 11400
+      "epoch": 0.02,
+      "grad_norm": 7.71875,
+      "learning_rate": 1.9347368421052634e-05,
+      "loss": 0.5431,
+      "step": 810
     },
     {
-      "epoch": 0.28,
-      "eval_loss": 0.7602470517158508,
-      "eval_runtime": 98.4042,
-      "eval_samples_per_second": 10.162,
-      "eval_steps_per_second": 10.162,
-      "step": 11400
+      "epoch": 0.02,
+      "grad_norm": 4.5,
+      "learning_rate": 1.9326315789473687e-05,
+      "loss": 0.619,
+      "step": 820
     },
     {
-      "epoch": 0.29,
-      "grad_norm": 4.534581661224365,
-      "learning_rate": 8.717948717948719e-06,
-      "loss": 0.7532,
-      "step": 11500
+      "epoch": 0.02,
+      "grad_norm": 7.53125,
+      "learning_rate": 1.930526315789474e-05,
+      "loss": 0.7058,
+      "step": 830
     },
     {
-      "epoch": 0.29,
-      "eval_loss": 0.7669916152954102,
-      "eval_runtime": 98.3947,
-      "eval_samples_per_second": 10.163,
-      "eval_steps_per_second": 10.163,
-      "step": 11500
+      "epoch": 0.02,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.9284210526315792e-05,
+      "loss": 0.631,
+      "step": 840
     },
     {
-      "epoch": 0.29,
-      "grad_norm": 3.601339101791382,
-      "learning_rate": 8.615384615384617e-06,
-      "loss": 0.7274,
-      "step": 11600
+      "epoch": 0.02,
+      "grad_norm": 6.1875,
+      "learning_rate": 1.9263157894736845e-05,
+      "loss": 0.5991,
+      "step": 850
     },
     {
-      "epoch": 0.29,
-      "eval_loss": 0.7589613199234009,
-      "eval_runtime": 98.411,
-      "eval_samples_per_second": 10.161,
-      "eval_steps_per_second": 10.161,
-      "step": 11600
+      "epoch": 0.02,
+      "grad_norm": 3.890625,
+      "learning_rate": 1.9242105263157894e-05,
+      "loss": 0.6852,
+      "step": 860
     },
     {
-      "epoch": 0.29,
-      "grad_norm": 3.2606701850891113,
-      "learning_rate": 8.512820512820513e-06,
-      "loss": 0.7396,
-      "step": 11700
+      "epoch": 0.02,
+      "grad_norm": 7.46875,
+      "learning_rate": 1.922105263157895e-05,
+      "loss": 0.6474,
+      "step": 870
     },
     {
-      "epoch": 0.29,
-      "eval_loss": 0.7702498435974121,
-      "eval_runtime": 98.3995,
-      "eval_samples_per_second": 10.163,
-      "eval_steps_per_second": 10.163,
-      "step": 11700
+      "epoch": 0.02,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.9200000000000003e-05,
+      "loss": 0.5979,
+      "step": 880
     },
     {
-      "epoch": 0.29,
-      "grad_norm": 4.607340335845947,
-      "learning_rate": 8.410256410256411e-06,
-      "loss": 0.7502,
-      "step": 11800
+      "epoch": 0.02,
+      "grad_norm": 4.625,
+      "learning_rate": 1.9178947368421055e-05,
+      "loss": 0.7035,
+      "step": 890
     },
     {
-      "epoch": 0.29,
-      "eval_loss": 0.7340441346168518,
-      "eval_runtime": 98.3886,
-      "eval_samples_per_second": 10.164,
-      "eval_steps_per_second": 10.164,
-      "step": 11800
+      "epoch": 0.02,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.9157894736842108e-05,
+      "loss": 0.7001,
+      "step": 900
     },
     {
-      "epoch": 0.3,
-      "grad_norm": 6.950789451599121,
-      "learning_rate": 8.307692307692309e-06,
-      "loss": 0.7499,
-      "step": 11900
+      "epoch": 0.02,
+      "grad_norm": 2.578125,
+      "learning_rate": 1.913684210526316e-05,
+      "loss": 0.6453,
+      "step": 910
     },
     {
-      "epoch": 0.3,
-      "eval_loss": 0.7349185943603516,
-      "eval_runtime": 98.4071,
-      "eval_samples_per_second": 10.162,
-      "eval_steps_per_second": 10.162,
-      "step": 11900
+      "epoch": 0.02,
+      "grad_norm": 4.4375,
+      "learning_rate": 1.9115789473684213e-05,
+      "loss": 0.6356,
+      "step": 920
     },
     {
-      "epoch": 0.3,
-      "grad_norm": 6.848160266876221,
-      "learning_rate": 8.205128205128205e-06,
-      "loss": 0.7445,
-      "step": 12000
+      "epoch": 0.02,
+      "grad_norm": 9.1875,
+      "learning_rate": 1.9094736842105262e-05,
+      "loss": 0.5679,
+      "step": 930
     },
     {
-      "epoch": 0.3,
-      "eval_loss": 0.7594106197357178,
-      "eval_runtime": 98.3807,
-      "eval_samples_per_second": 10.165,
-      "eval_steps_per_second": 10.165,
-      "step": 12000
+      "epoch": 0.02,
+      "grad_norm": 3.46875,
+      "learning_rate": 1.907368421052632e-05,
+      "loss": 0.7062,
+      "step": 940
     },
     {
-      "epoch": 0.3,
-      "grad_norm": 5.719851016998291,
-      "learning_rate": 8.102564102564103e-06,
-      "loss": 0.7053,
-      "step": 12100
+      "epoch": 0.02,
+      "grad_norm": 5.5625,
+      "learning_rate": 1.9052631578947368e-05,
+      "loss": 0.665,
+      "step": 950
     },
     {
-      "epoch": 0.3,
-      "eval_loss": 0.7302230000495911,
-      "eval_runtime": 98.3965,
-      "eval_samples_per_second": 10.163,
-      "eval_steps_per_second": 10.163,
-      "step": 12100
+      "epoch": 0.02,
+      "grad_norm": 7.96875,
+      "learning_rate": 1.9031578947368424e-05,
+      "loss": 0.6772,
+      "step": 960
     },
     {
-      "epoch": 0.3,
-      "grad_norm": 2.405026912689209,
-      "learning_rate": 8.000000000000001e-06,
-      "loss": 0.6993,
-      "step": 12200
+      "epoch": 0.02,
+      "grad_norm": 7.09375,
+      "learning_rate": 1.9010526315789476e-05,
+      "loss": 0.5347,
+      "step": 970
     },
     {
-      "epoch": 0.3,
-      "eval_loss": 0.7647203207015991,
-      "eval_runtime": 98.4071,
-      "eval_samples_per_second": 10.162,
-      "eval_steps_per_second": 10.162,
-      "step": 12200
+      "epoch": 0.02,
+      "grad_norm": 4.65625,
+      "learning_rate": 1.898947368421053e-05,
+      "loss": 0.6161,
+      "step": 980
     },
     {
-      "epoch": 0.31,
-      "grad_norm": 3.210000514984131,
-      "learning_rate": 7.897435897435898e-06,
-      "loss": 0.7465,
-      "step": 12300
+      "epoch": 0.02,
+      "grad_norm": 3.796875,
+      "learning_rate": 1.8968421052631582e-05,
+      "loss": 0.595,
+      "step": 990
     },
     {
-      "epoch": 0.31,
-      "eval_loss": 0.7343460321426392,
-      "eval_runtime": 98.3972,
-      "eval_samples_per_second": 10.163,
-      "eval_steps_per_second": 10.163,
-      "step": 12300
+      "epoch": 0.03,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.894736842105263e-05,
+      "loss": 0.6752,
+      "step": 1000
     },
     {
-      "epoch": 0.31,
-      "grad_norm": 5.848453521728516,
-      "learning_rate": 7.794871794871796e-06,
-      "loss": 0.7664,
-      "step": 12400
+      "epoch": 0.03,
+      "eval_loss": 0.6129263639450073,
+      "eval_runtime": 38.0435,
+      "eval_samples_per_second": 26.286,
+      "eval_steps_per_second": 26.286,
+      "step": 1000
     },
     {
-      "epoch": 0.31,
-      "eval_loss": 0.7306160926818848,
-      "eval_runtime": 98.3866,
-      "eval_samples_per_second": 10.164,
-      "eval_steps_per_second": 10.164,
-      "step": 12400
+      "epoch": 0.03,
+      "grad_norm": 5.5,
+      "learning_rate": 1.8926315789473687e-05,
+      "loss": 0.6876,
+      "step": 1010
     },
     {
-      "epoch": 0.31,
-      "grad_norm": 3.4362599849700928,
-      "learning_rate": 7.692307692307694e-06,
-      "loss": 0.7239,
-      "step": 12500
+      "epoch": 0.03,
+      "grad_norm": 3.640625,
+      "learning_rate": 1.8905263157894736e-05,
+      "loss": 0.6364,
+      "step": 1020
     },
     {
-      "epoch": 0.31,
-      "eval_loss": 0.7516223192214966,
-      "eval_runtime": 98.4441,
-      "eval_samples_per_second": 10.158,
-      "eval_steps_per_second": 10.158,
-      "step": 12500
+      "epoch": 0.03,
+      "grad_norm": 6.6875,
+      "learning_rate": 1.8884210526315792e-05,
+      "loss": 0.6598,
+      "step": 1030
     },
     {
-      "epoch": 0.32,
-      "grad_norm": 66.06706237792969,
-      "learning_rate": 7.58974358974359e-06,
-      "loss": 0.7109,
-      "step": 12600
+      "epoch": 0.03,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.886315789473684e-05,
+      "loss": 0.6473,
+      "step": 1040
     },
     {
-      "epoch": 0.32,
-      "eval_loss": 0.7285715341567993,
-      "eval_runtime": 98.5076,
-      "eval_samples_per_second": 10.152,
-      "eval_steps_per_second": 10.152,
-      "step": 12600
+      "epoch": 0.03,
+      "grad_norm": 5.125,
+      "learning_rate": 1.8842105263157898e-05,
+      "loss": 0.5972,
+      "step": 1050
     },
     {
-      "epoch": 0.32,
-      "grad_norm": 4.000461578369141,
-      "learning_rate": 7.487179487179488e-06,
-      "loss": 0.7683,
-      "step": 12700
+      "epoch": 0.03,
+      "grad_norm": 3.765625,
+      "learning_rate": 1.882105263157895e-05,
+      "loss": 0.7538,
+      "step": 1060
     },
     {
-      "epoch": 0.32,
-      "eval_loss": 0.7256894707679749,
-      "eval_runtime": 98.4291,
-      "eval_samples_per_second": 10.16,
-      "eval_steps_per_second": 10.16,
-      "step": 12700
+      "epoch": 0.03,
+      "grad_norm": 4.625,
+      "learning_rate": 1.88e-05,
+      "loss": 0.6282,
+      "step": 1070
     },
     {
-      "epoch": 0.32,
-      "grad_norm": 2.8244261741638184,
-      "learning_rate": 7.384615384615386e-06,
-      "loss": 0.7599,
-      "step": 12800
+      "epoch": 0.03,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.8778947368421056e-05,
+      "loss": 0.7304,
+      "step": 1080
     },
     {
-      "epoch": 0.32,
-      "eval_loss": 0.7099986672401428,
-      "eval_runtime": 98.3772,
-      "eval_samples_per_second": 10.165,
-      "eval_steps_per_second": 10.165,
-      "step": 12800
+      "epoch": 0.03,
+      "grad_norm": 12.625,
+      "learning_rate": 1.8757894736842105e-05,
+      "loss": 0.6744,
+      "step": 1090
     },
     {
-      "epoch": 0.32,
-      "grad_norm": 5.786495685577393,
-      "learning_rate": 7.282051282051282e-06,
-      "loss": 0.7331,
-      "step": 12900
+      "epoch": 0.03,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.873684210526316e-05,
+      "loss": 0.7072,
+      "step": 1100
     },
     {
-      "epoch": 0.32,
-      "eval_loss": 0.7184450626373291,
-      "eval_runtime": 98.386,
-      "eval_samples_per_second": 10.164,
-      "eval_steps_per_second": 10.164,
-      "step": 12900
+      "epoch": 0.03,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.871578947368421e-05,
+      "loss": 0.6935,
+      "step": 1110
     },
     {
-      "epoch": 0.33,
-      "grad_norm": 3.1670029163360596,
-      "learning_rate": 7.17948717948718e-06,
-      "loss": 0.7012,
-      "step": 13000
+      "epoch": 0.03,
+      "grad_norm": 7.125,
+      "learning_rate": 1.8694736842105266e-05,
+      "loss": 0.6507,
+      "step": 1120
     },
     {
-      "epoch": 0.33,
-      "eval_loss": 0.7143362164497375,
-      "eval_runtime": 98.4313,
-      "eval_samples_per_second": 10.159,
-      "eval_steps_per_second": 10.159,
-      "step": 13000
+      "epoch": 0.03,
+      "grad_norm": 0.326171875,
+      "learning_rate": 1.8673684210526316e-05,
+      "loss": 0.5451,
+      "step": 1130
     },
     {
-      "epoch": 0.33,
-      "grad_norm": 3.853381633758545,
-      "learning_rate": 7.076923076923078e-06,
-      "loss": 0.7207,
-      "step": 13100
+      "epoch": 0.03,
+      "grad_norm": 8.0,
+      "learning_rate": 1.8652631578947368e-05,
+      "loss": 0.6145,
+      "step": 1140
     },
     {
-      "epoch": 0.33,
-      "eval_loss": 0.7087370157241821,
-      "eval_runtime": 98.3509,
-      "eval_samples_per_second": 10.168,
-      "eval_steps_per_second": 10.168,
-      "step": 13100
+      "epoch": 0.03,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.8631578947368424e-05,
+      "loss": 0.6846,
+      "step": 1150
     },
     {
-      "epoch": 0.33,
-      "grad_norm": 4.216042518615723,
-      "learning_rate": 6.974358974358974e-06,
-      "loss": 0.7417,
-      "step": 13200
+      "epoch": 0.03,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.8610526315789473e-05,
+      "loss": 0.6139,
+      "step": 1160
     },
     {
-      "epoch": 0.33,
-      "eval_loss": 0.7320001721382141,
-      "eval_runtime": 98.4508,
-      "eval_samples_per_second": 10.157,
-      "eval_steps_per_second": 10.157,
-      "step": 13200
+      "epoch": 0.03,
+      "grad_norm": 6.71875,
+      "learning_rate": 1.858947368421053e-05,
+      "loss": 0.5147,
+      "step": 1170
     },
     {
-      "epoch": 0.33,
-      "grad_norm": 3.6057064533233643,
-      "learning_rate": 6.871794871794872e-06,
-      "loss": 0.7192,
-      "step": 13300
+      "epoch": 0.03,
+      "grad_norm": 4.125,
+      "learning_rate": 1.856842105263158e-05,
+      "loss": 0.5526,
+      "step": 1180
     },
     {
-      "epoch": 0.33,
-      "eval_loss": 0.7073127627372742,
-      "eval_runtime": 98.3946,
-      "eval_samples_per_second": 10.163,
-      "eval_steps_per_second": 10.163,
-      "step": 13300
+      "epoch": 0.03,
+      "grad_norm": 13.6875,
+      "learning_rate": 1.8547368421052635e-05,
+      "loss": 0.6973,
+      "step": 1190
     },
     {
-      "epoch": 0.34,
-      "grad_norm": 6.80657958984375,
-      "learning_rate": 6.76923076923077e-06,
-      "loss": 0.7484,
-      "step": 13400
+      "epoch": 0.03,
+      "grad_norm": 3.8125,
+      "learning_rate": 1.8526315789473684e-05,
+      "loss": 0.6338,
+      "step": 1200
     },
     {
-      "epoch": 0.34,
-      "eval_loss": 0.7168283462524414,
-      "eval_runtime": 98.3837,
-      "eval_samples_per_second": 10.164,
-      "eval_steps_per_second": 10.164,
-      "step": 13400
+      "epoch": 0.03,
+      "grad_norm": 3.234375,
+      "learning_rate": 1.8505263157894737e-05,
+      "loss": 0.5985,
+      "step": 1210
     },
     {
-      "epoch": 0.34,
-      "grad_norm": 3.738083600997925,
-      "learning_rate": 6.666666666666667e-06,
-      "loss": 0.7044,
-      "step": 13500
+      "epoch": 0.03,
+      "grad_norm": 4.625,
+      "learning_rate": 1.8484210526315793e-05,
+      "loss": 0.6832,
+      "step": 1220
     },
     {
-      "epoch": 0.34,
-      "eval_loss": 0.7054859399795532,
-      "eval_runtime": 98.4103,
-      "eval_samples_per_second": 10.162,
-      "eval_steps_per_second": 10.162,
-      "step": 13500
+      "epoch": 0.03,
+      "grad_norm": 4.59375,
+      "learning_rate": 1.8463157894736842e-05,
+      "loss": 0.6766,
+      "step": 1230
     },
     {
-      "epoch": 0.34,
-      "grad_norm": 4.214886665344238,
-      "learning_rate": 6.564102564102565e-06,
-      "loss": 0.7071,
-      "step": 13600
+      "epoch": 0.03,
+      "grad_norm": 5.09375,
+      "learning_rate": 1.8442105263157898e-05,
+      "loss": 0.6342,
+      "step": 1240
     },
     {
-      "epoch": 0.34,
-      "eval_loss": 0.7007602453231812,
-      "eval_runtime": 98.3702,
-      "eval_samples_per_second": 10.166,
-      "eval_steps_per_second": 10.166,
-      "step": 13600
+      "epoch": 0.03,
+      "grad_norm": 5.0625,
+      "learning_rate": 1.8421052631578947e-05,
+      "loss": 0.6043,
+      "step": 1250
     },
     {
-      "epoch": 0.34,
-      "grad_norm": 2.437922716140747,
-      "learning_rate": 6.461538461538463e-06,
-      "loss": 0.6839,
-      "step": 13700
+      "epoch": 0.03,
+      "grad_norm": 3.34375,
+      "learning_rate": 1.8400000000000003e-05,
+      "loss": 0.5879,
+      "step": 1260
     },
     {
-      "epoch": 0.34,
-      "eval_loss": 0.7002619504928589,
-      "eval_runtime": 98.3806,
-      "eval_samples_per_second": 10.165,
-      "eval_steps_per_second": 10.165,
-      "step": 13700
+      "epoch": 0.03,
+      "grad_norm": 4.84375,
+      "learning_rate": 1.8378947368421053e-05,
+      "loss": 0.6553,
+      "step": 1270
     },
     {
-      "epoch": 0.34,
-      "grad_norm": 9.056229591369629,
-      "learning_rate": 6.358974358974359e-06,
-      "loss": 0.7352,
-      "step": 13800
+      "epoch": 0.03,
+      "grad_norm": 2.984375,
+      "learning_rate": 1.8357894736842105e-05,
+      "loss": 0.6256,
+      "step": 1280
     },
     {
-      "epoch": 0.34,
-      "eval_loss": 0.706998884677887,
-      "eval_runtime": 98.3786,
-      "eval_samples_per_second": 10.165,
-      "eval_steps_per_second": 10.165,
-      "step": 13800
+      "epoch": 0.03,
+      "grad_norm": 5.75,
+      "learning_rate": 1.8336842105263158e-05,
+      "loss": 0.7604,
+      "step": 1290
     },
     {
-      "epoch": 0.35,
-      "grad_norm": 7.531219482421875,
-      "learning_rate": 6.256410256410257e-06,
-      "loss": 0.7043,
-      "step": 13900
+      "epoch": 0.03,
+      "grad_norm": 9.0,
+      "learning_rate": 1.831578947368421e-05,
+      "loss": 0.5902,
+      "step": 1300
     },
     {
-      "epoch": 0.35,
-      "eval_loss": 0.7098628878593445,
-      "eval_runtime": 98.4273,
-      "eval_samples_per_second": 10.16,
-      "eval_steps_per_second": 10.16,
-      "step": 13900
+      "epoch": 0.03,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.8294736842105267e-05,
+      "loss": 0.6895,
+      "step": 1310
     },
     {
-      "epoch": 0.35,
-      "grad_norm": 5.21802282333374,
-      "learning_rate": 6.153846153846155e-06,
-      "loss": 0.6914,
-      "step": 14000
+      "epoch": 0.03,
+      "grad_norm": 3.171875,
+      "learning_rate": 1.8273684210526316e-05,
+      "loss": 0.557,
+      "step": 1320
     },
     {
-      "epoch": 0.35,
-      "eval_loss": 0.7066898941993713,
-      "eval_runtime": 98.355,
-      "eval_samples_per_second": 10.167,
-      "eval_steps_per_second": 10.167,
-      "step": 14000
+      "epoch": 0.03,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.8252631578947372e-05,
+      "loss": 0.5967,
+      "step": 1330
     },
     {
-      "epoch": 0.35,
-      "grad_norm": 4.617076873779297,
-      "learning_rate": 6.051282051282051e-06,
-      "loss": 0.6747,
-      "step": 14100
+      "epoch": 0.03,
+      "grad_norm": 2.171875,
+      "learning_rate": 1.823157894736842e-05,
+      "loss": 0.7342,
+      "step": 1340
     },
     {
-      "epoch": 0.35,
-      "eval_loss": 0.7035422921180725,
-      "eval_runtime": 98.4009,
-      "eval_samples_per_second": 10.163,
-      "eval_steps_per_second": 10.163,
-      "step": 14100
+      "epoch": 0.03,
+      "grad_norm": 3.65625,
+      "learning_rate": 1.8210526315789477e-05,
+      "loss": 0.5225,
+      "step": 1350
     },
     {
-      "epoch": 0.35,
-      "grad_norm": 4.323288917541504,
-      "learning_rate": 5.948717948717949e-06,
-      "loss": 0.6973,
-      "step": 14200
+      "epoch": 0.03,
+      "grad_norm": 2.84375,
+      "learning_rate": 1.8189473684210527e-05,
+      "loss": 0.6079,
+      "step": 1360
     },
     {
-      "epoch": 0.35,
-      "eval_loss": 0.6891523599624634,
-      "eval_runtime": 98.3514,
-      "eval_samples_per_second": 10.168,
-      "eval_steps_per_second": 10.168,
-      "step": 14200
+      "epoch": 0.03,
+      "grad_norm": 2.546875,
+      "learning_rate": 1.816842105263158e-05,
+      "loss": 0.5944,
+      "step": 1370
     },
     {
-      "epoch": 0.36,
-      "grad_norm": 3.765428066253662,
-      "learning_rate": 5.846153846153847e-06,
-      "loss": 0.6818,
-      "step": 14300
+      "epoch": 0.03,
+      "grad_norm": 3.0625,
+      "learning_rate": 1.8147368421052632e-05,
+      "loss": 0.5262,
+      "step": 1380
     },
     {
-      "epoch": 0.36,
-      "eval_loss": 0.7228779196739197,
-      "eval_runtime": 98.3527,
-      "eval_samples_per_second": 10.167,
-      "eval_steps_per_second": 10.167,
-      "step": 14300
+      "epoch": 0.03,
+      "grad_norm": 1.7578125,
+      "learning_rate": 1.8126315789473685e-05,
+      "loss": 0.5117,
+      "step": 1390
     },
     {
-      "epoch": 0.36,
-      "grad_norm": 10.397618293762207,
-      "learning_rate": 5.743589743589743e-06,
-      "loss": 0.7123,
-      "step": 14400
+      "epoch": 0.04,
+      "grad_norm": 8.3125,
+      "learning_rate": 1.810526315789474e-05,
+      "loss": 0.7201,
+      "step": 1400
     },
     {
-      "epoch": 0.36,
-      "eval_loss": 0.6805267333984375,
-      "eval_runtime": 98.3834,
-      "eval_samples_per_second": 10.164,
-      "eval_steps_per_second": 10.164,
-      "step": 14400
+      "epoch": 0.04,
+      "grad_norm": 2.171875,
+      "learning_rate": 1.808421052631579e-05,
+      "loss": 0.6592,
+      "step": 1410
     },
     {
-      "epoch": 0.36,
-      "grad_norm": 4.0021867752075195,
-      "learning_rate": 5.641025641025641e-06,
-      "loss": 0.6876,
-      "step": 14500
+      "epoch": 0.04,
+      "grad_norm": 4.34375,
+      "learning_rate": 1.8063157894736846e-05,
+      "loss": 0.6966,
+      "step": 1420
     },
     {
-      "epoch": 0.36,
-      "eval_loss": 0.7112326622009277,
-      "eval_runtime": 98.4081,
-      "eval_samples_per_second": 10.162,
-      "eval_steps_per_second": 10.162,
-      "step": 14500
+      "epoch": 0.04,
+      "grad_norm": 5.9375,
+      "learning_rate": 1.8042105263157895e-05,
+      "loss": 0.5328,
+      "step": 1430
     },
     {
-      "epoch": 0.36,
-      "grad_norm": 4.793341159820557,
-      "learning_rate": 5.538461538461539e-06,
-      "loss": 0.7236,
-      "step": 14600
+      "epoch": 0.04,
+      "grad_norm": 4.03125,
+      "learning_rate": 1.8021052631578948e-05,
+      "loss": 0.6245,
+      "step": 1440
     },
     {
-      "epoch": 0.36,
-      "eval_loss": 0.7055144309997559,
-      "eval_runtime": 98.3771,
-      "eval_samples_per_second": 10.165,
-      "eval_steps_per_second": 10.165,
-      "step": 14600
+      "epoch": 0.04,
+      "grad_norm": 3.03125,
+      "learning_rate": 1.8e-05,
+      "loss": 0.5457,
+      "step": 1450
     },
     {
-      "epoch": 0.37,
-      "grad_norm": 4.1885666847229,
-      "learning_rate": 5.435897435897436e-06,
-      "loss": 0.7125,
-      "step": 14700
+      "epoch": 0.04,
+      "grad_norm": 2.9375,
+      "learning_rate": 1.7978947368421053e-05,
+      "loss": 0.6026,
+      "step": 1460
     },
     {
-      "epoch": 0.37,
-      "eval_loss": 0.6943590641021729,
-      "eval_runtime": 98.3807,
-      "eval_samples_per_second": 10.165,
-      "eval_steps_per_second": 10.165,
-      "step": 14700
+      "epoch": 0.04,
+      "grad_norm": 3.546875,
+      "learning_rate": 1.795789473684211e-05,
+      "loss": 0.6519,
+      "step": 1470
     },
     {
-      "epoch": 0.37,
-      "grad_norm": 4.00455904006958,
-      "learning_rate": 5.333333333333334e-06,
-      "loss": 0.6615,
-      "step": 14800
+      "epoch": 0.04,
+      "grad_norm": 4.8125,
+      "learning_rate": 1.793684210526316e-05,
+      "loss": 0.5364,
+      "step": 1480
     },
     {
-      "epoch": 0.37,
-      "eval_loss": 0.6834198832511902,
-      "eval_runtime": 98.3529,
-      "eval_samples_per_second": 10.167,
-      "eval_steps_per_second": 10.167,
-      "step": 14800
+      "epoch": 0.04,
+      "grad_norm": 4.9375,
+      "learning_rate": 1.7915789473684214e-05,
+      "loss": 0.6421,
+      "step": 1490
     },
     {
-      "epoch": 0.37,
-      "grad_norm": 5.705050945281982,
-      "learning_rate": 5.230769230769232e-06,
-      "loss": 0.6771,
-      "step": 14900
+      "epoch": 0.04,
+      "grad_norm": 5.03125,
+      "learning_rate": 1.7894736842105264e-05,
+      "loss": 0.5376,
+      "step": 1500
     },
     {
-      "epoch": 0.37,
-      "eval_loss": 0.6999417543411255,
-      "eval_runtime": 98.3937,
-      "eval_samples_per_second": 10.163,
-      "eval_steps_per_second": 10.163,
-      "step": 14900
+      "epoch": 0.04,
+      "eval_loss": 0.6211587190628052,
+      "eval_runtime": 38.148,
+      "eval_samples_per_second": 26.214,
+      "eval_steps_per_second": 26.214,
+      "step": 1500
     },
     {
-      "epoch": 0.38,
-      "grad_norm": 3.552046298980713,
-      "learning_rate": 5.128205128205128e-06,
-      "loss": 0.6742,
-      "step": 15000
+      "epoch": 0.04,
+      "grad_norm": 5.125,
+      "learning_rate": 1.7873684210526316e-05,
+      "loss": 0.5633,
+      "step": 1510
     },
     {
-      "epoch": 0.38,
-      "eval_loss": 0.6728054285049438,
-      "eval_runtime": 98.3744,
-      "eval_samples_per_second": 10.165,
-      "eval_steps_per_second": 10.165,
-      "step": 15000
+      "epoch": 0.04,
+      "grad_norm": 14.0,
+      "learning_rate": 1.785263157894737e-05,
+      "loss": 0.6917,
+      "step": 1520
     },
     {
-      "epoch": 0.38,
-      "grad_norm": 6.856844902038574,
-      "learning_rate": 5.025641025641026e-06,
-      "loss": 0.6922,
-      "step": 15100
+      "epoch": 0.04,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.7831578947368422e-05,
+      "loss": 0.5882,
+      "step": 1530
     },
     {
-      "epoch": 0.38,
-      "eval_loss": 0.6971193552017212,
-      "eval_runtime": 98.4333,
-      "eval_samples_per_second": 10.159,
-      "eval_steps_per_second": 10.159,
-      "step": 15100
+      "epoch": 0.04,
+      "grad_norm": 12.0625,
+      "learning_rate": 1.7810526315789474e-05,
+      "loss": 0.5752,
+      "step": 1540
     },
     {
-      "epoch": 0.38,
-      "grad_norm": 4.937108039855957,
-      "learning_rate": 4.923076923076924e-06,
-      "loss": 0.6843,
-      "step": 15200
+      "epoch": 0.04,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.7789473684210527e-05,
+      "loss": 0.6439,
+      "step": 1550
     },
     {
-      "epoch": 0.38,
-      "eval_loss": 0.6769686937332153,
-      "eval_runtime": 98.4292,
-      "eval_samples_per_second": 10.16,
-      "eval_steps_per_second": 10.16,
-      "step": 15200
+      "epoch": 0.04,
+      "grad_norm": 6.0625,
+      "learning_rate": 1.7768421052631583e-05,
+      "loss": 0.6806,
+      "step": 1560
     },
     {
-      "epoch": 0.38,
-      "grad_norm": 7.337194442749023,
-      "learning_rate": 4.820512820512821e-06,
-      "loss": 0.7265,
-      "step": 15300
+      "epoch": 0.04,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.7747368421052632e-05,
+      "loss": 0.5903,
+      "step": 1570
     },
     {
-      "epoch": 0.38,
-      "eval_loss": 0.6937099695205688,
-      "eval_runtime": 98.3534,
-      "eval_samples_per_second": 10.167,
-      "eval_steps_per_second": 10.167,
-      "step": 15300
+      "epoch": 0.04,
+      "grad_norm": 4.46875,
+      "learning_rate": 1.7726315789473685e-05,
+      "loss": 0.6394,
+      "step": 1580
     },
     {
-      "epoch": 0.39,
-      "grad_norm": 6.544970512390137,
-      "learning_rate": 4.717948717948718e-06,
-      "loss": 0.6854,
-      "step": 15400
+      "epoch": 0.04,
+      "grad_norm": 3.5,
+      "learning_rate": 1.7705263157894738e-05,
+      "loss": 0.53,
+      "step": 1590
     },
     {
-      "epoch": 0.39,
-      "eval_loss": 0.6800307631492615,
-      "eval_runtime": 98.3657,
-      "eval_samples_per_second": 10.166,
-      "eval_steps_per_second": 10.166,
-      "step": 15400
+      "epoch": 0.04,
+      "grad_norm": 10.625,
+      "learning_rate": 1.768421052631579e-05,
+      "loss": 0.7113,
+      "step": 1600
     },
     {
-      "epoch": 0.39,
-      "grad_norm": 4.409825801849365,
-      "learning_rate": 4.615384615384616e-06,
-      "loss": 0.7084,
-      "step": 15500
+      "epoch": 0.04,
+      "grad_norm": 10.875,
+      "learning_rate": 1.7663157894736843e-05,
+      "loss": 0.6491,
+      "step": 1610
     },
     {
-      "epoch": 0.39,
-      "eval_loss": 0.6897458434104919,
-      "eval_runtime": 98.3938,
-      "eval_samples_per_second": 10.163,
-      "eval_steps_per_second": 10.163,
-      "step": 15500
+      "epoch": 0.04,
+      "grad_norm": 4.96875,
+      "learning_rate": 1.7642105263157896e-05,
+      "loss": 0.6439,
+      "step": 1620
     },
     {
-      "epoch": 0.39,
-      "grad_norm": 4.764963150024414,
-      "learning_rate": 4.512820512820513e-06,
-      "loss": 0.6835,
-      "step": 15600
+      "epoch": 0.04,
+      "grad_norm": 9.5625,
+      "learning_rate": 1.7621052631578948e-05,
+      "loss": 0.7248,
+      "step": 1630
     },
     {
-      "epoch": 0.39,
-      "eval_loss": 0.6795360445976257,
-      "eval_runtime": 98.3753,
-      "eval_samples_per_second": 10.165,
-      "eval_steps_per_second": 10.165,
-      "step": 15600
+      "epoch": 0.04,
+      "grad_norm": 6.96875,
+      "learning_rate": 1.76e-05,
+      "loss": 0.6754,
+      "step": 1640
     },
     {
-      "epoch": 0.39,
-      "grad_norm": 4.578685283660889,
-      "learning_rate": 4.4102564102564104e-06,
-      "loss": 0.6789,
-      "step": 15700
+      "epoch": 0.04,
+      "grad_norm": 3.828125,
+      "learning_rate": 1.7578947368421054e-05,
+      "loss": 0.6029,
+      "step": 1650
     },
     {
-      "epoch": 0.39,
-      "eval_loss": 0.6842684149742126,
-      "eval_runtime": 98.3424,
-      "eval_samples_per_second": 10.169,
-      "eval_steps_per_second": 10.169,
-      "step": 15700
+      "epoch": 0.04,
+      "grad_norm": 6.0,
+      "learning_rate": 1.7557894736842106e-05,
+      "loss": 0.6225,
+      "step": 1660
     },
     {
-      "epoch": 0.4,
-      "grad_norm": 7.085032939910889,
-      "learning_rate": 4.307692307692308e-06,
-      "loss": 0.6473,
-      "step": 15800
+      "epoch": 0.04,
+      "grad_norm": 3.265625,
+      "learning_rate": 1.753684210526316e-05,
+      "loss": 0.5892,
+      "step": 1670
     },
     {
-      "epoch": 0.4,
-      "eval_loss": 0.6702744364738464,
-      "eval_runtime": 98.3744,
-      "eval_samples_per_second": 10.165,
-      "eval_steps_per_second": 10.165,
-      "step": 15800
+      "epoch": 0.04,
+      "grad_norm": 3.390625,
+      "learning_rate": 1.751578947368421e-05,
+      "loss": 0.7917,
+      "step": 1680
     },
     {
-      "epoch": 0.4,
-      "grad_norm": 9.547301292419434,
-      "learning_rate": 4.2051282051282055e-06,
-      "loss": 0.6666,
-      "step": 15900
+      "epoch": 0.04,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.7494736842105264e-05,
+      "loss": 0.5863,
+      "step": 1690
     },
     {
-      "epoch": 0.4,
-      "eval_loss": 0.6775800585746765,
-      "eval_runtime": 98.3228,
-      "eval_samples_per_second": 10.171,
-      "eval_steps_per_second": 10.171,
-      "step": 15900
+      "epoch": 0.04,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.7473684210526317e-05,
+      "loss": 0.5613,
+      "step": 1700
     },
     {
-      "epoch": 0.4,
-      "grad_norm": 2.161081075668335,
-      "learning_rate": 4.102564102564103e-06,
-      "loss": 0.6968,
-      "step": 16000
+      "epoch": 0.04,
+      "grad_norm": 11.6875,
+      "learning_rate": 1.745263157894737e-05,
+      "loss": 0.6582,
+      "step": 1710
     },
     {
-      "epoch": 0.4,
-      "eval_loss": 0.683419942855835,
-      "eval_runtime": 98.4131,
-      "eval_samples_per_second": 10.161,
-      "eval_steps_per_second": 10.161,
-      "step": 16000
+      "epoch": 0.04,
+      "grad_norm": 5.0,
+      "learning_rate": 1.7431578947368422e-05,
+      "loss": 0.6967,
+      "step": 1720
     },
     {
-      "epoch": 0.4,
-      "grad_norm": 4.315452575683594,
-      "learning_rate": 4.000000000000001e-06,
-      "loss": 0.64,
-      "step": 16100
+      "epoch": 0.04,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.7410526315789475e-05,
+      "loss": 0.6296,
+      "step": 1730
     },
     {
-      "epoch": 0.4,
-      "eval_loss": 0.6916409134864807,
-      "eval_runtime": 98.331,
-      "eval_samples_per_second": 10.17,
-      "eval_steps_per_second": 10.17,
-      "step": 16100
+      "epoch": 0.04,
+      "grad_norm": 26.0,
+      "learning_rate": 1.7389473684210527e-05,
+      "loss": 0.5804,
+      "step": 1740
     },
     {
-      "epoch": 0.41,
-      "grad_norm": 4.9351582527160645,
-      "learning_rate": 3.897435897435898e-06,
-      "loss": 0.6516,
-      "step": 16200
+      "epoch": 0.04,
+      "grad_norm": 3.015625,
+      "learning_rate": 1.736842105263158e-05,
+      "loss": 0.7102,
+      "step": 1750
     },
     {
-      "epoch": 0.41,
-      "eval_loss": 0.6831104755401611,
-      "eval_runtime": 98.342,
-      "eval_samples_per_second": 10.169,
-      "eval_steps_per_second": 10.169,
-      "step": 16200
+      "epoch": 0.04,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.7347368421052633e-05,
+      "loss": 0.6108,
+      "step": 1760
     },
     {
-      "epoch": 0.41,
-      "grad_norm": 2.858851432800293,
-      "learning_rate": 3.794871794871795e-06,
-      "loss": 0.6626,
-      "step": 16300
+      "epoch": 0.04,
+      "grad_norm": 3.9375,
+      "learning_rate": 1.7326315789473685e-05,
+      "loss": 0.6847,
+      "step": 1770
     },
     {
-      "epoch": 0.41,
-      "eval_loss": 0.6660827398300171,
-      "eval_runtime": 98.3507,
-      "eval_samples_per_second": 10.168,
-      "eval_steps_per_second": 10.168,
-      "step": 16300
+      "epoch": 0.04,
+      "grad_norm": 3.890625,
+      "learning_rate": 1.7305263157894738e-05,
+      "loss": 0.7885,
+      "step": 1780
     },
     {
-      "epoch": 0.41,
-      "grad_norm": 10.384642601013184,
-      "learning_rate": 3.692307692307693e-06,
-      "loss": 0.7119,
-      "step": 16400
+      "epoch": 0.04,
+      "grad_norm": 5.53125,
+      "learning_rate": 1.728421052631579e-05,
+      "loss": 0.674,
+      "step": 1790
     },
     {
-      "epoch": 0.41,
-      "eval_loss": 0.6715333461761475,
-      "eval_runtime": 98.3678,
-      "eval_samples_per_second": 10.166,
-      "eval_steps_per_second": 10.166,
-      "step": 16400
+      "epoch": 0.04,
+      "grad_norm": 5.25,
+      "learning_rate": 1.7263157894736843e-05,
+      "loss": 0.6753,
+      "step": 1800
     },
     {
-      "epoch": 0.41,
-      "grad_norm": 2.9744458198547363,
-      "learning_rate": 3.58974358974359e-06,
-      "loss": 0.6827,
-      "step": 16500
+      "epoch": 0.05,
+      "grad_norm": 5.0,
+      "learning_rate": 1.7242105263157896e-05,
+      "loss": 0.5909,
+      "step": 1810
     },
     {
-      "epoch": 0.41,
-      "eval_loss": 0.6574322581291199,
-      "eval_runtime": 98.3789,
-      "eval_samples_per_second": 10.165,
-      "eval_steps_per_second": 10.165,
-      "step": 16500
+      "epoch": 0.05,
+      "grad_norm": 1.8984375,
+      "learning_rate": 1.722105263157895e-05,
+      "loss": 0.6107,
+      "step": 1820
     },
     {
-      "epoch": 0.41,
-      "grad_norm": 3.588651657104492,
-      "learning_rate": 3.487179487179487e-06,
-      "loss": 0.7398,
-      "step": 16600
+      "epoch": 0.05,
+      "grad_norm": 2.875,
+      "learning_rate": 1.72e-05,
+      "loss": 0.623,
+      "step": 1830
     },
     {
-      "epoch": 0.41,
-      "eval_loss": 0.6578336358070374,
-      "eval_runtime": 98.3458,
-      "eval_samples_per_second": 10.168,
-      "eval_steps_per_second": 10.168,
-      "step": 16600
+      "epoch": 0.05,
+      "grad_norm": 4.84375,
+      "learning_rate": 1.7178947368421054e-05,
+      "loss": 0.5906,
+      "step": 1840
     },
     {
-      "epoch": 0.42,
-      "grad_norm": 3.190857410430908,
-      "learning_rate": 3.384615384615385e-06,
-      "loss": 0.6553,
-      "step": 16700
+      "epoch": 0.05,
+      "grad_norm": 3.484375,
+      "learning_rate": 1.7157894736842107e-05,
+      "loss": 0.5833,
+      "step": 1850
     },
     {
-      "epoch": 0.42,
-      "eval_loss": 0.6542092561721802,
-      "eval_runtime": 98.3756,
-      "eval_samples_per_second": 10.165,
-      "eval_steps_per_second": 10.165,
-      "step": 16700
+      "epoch": 0.05,
+      "grad_norm": 9.25,
+      "learning_rate": 1.713684210526316e-05,
+      "loss": 0.5935,
+      "step": 1860
     },
     {
-      "epoch": 0.42,
-      "grad_norm": 3.513176441192627,
-      "learning_rate": 3.2820512820512823e-06,
-      "loss": 0.6531,
-      "step": 16800
+      "epoch": 0.05,
+      "grad_norm": 3.59375,
+      "learning_rate": 1.7115789473684212e-05,
+      "loss": 0.5972,
+      "step": 1870
     },
     {
-      "epoch": 0.42,
-      "eval_loss": 0.6713552474975586,
-      "eval_runtime": 98.3796,
-      "eval_samples_per_second": 10.165,
-      "eval_steps_per_second": 10.165,
-      "step": 16800
+      "epoch": 0.05,
+      "grad_norm": 1.8046875,
+      "learning_rate": 1.7094736842105265e-05,
+      "loss": 0.5128,
+      "step": 1880
     },
     {
-      "epoch": 0.42,
-      "grad_norm": 5.897684574127197,
-      "learning_rate": 3.1794871794871795e-06,
-      "loss": 0.6923,
-      "step": 16900
+      "epoch": 0.05,
+      "grad_norm": 4.9375,
+      "learning_rate": 1.7073684210526317e-05,
+      "loss": 0.6828,
+      "step": 1890
     },
     {
-      "epoch": 0.42,
-      "eval_loss": 0.6415435075759888,
-      "eval_runtime": 98.3575,
-      "eval_samples_per_second": 10.167,
-      "eval_steps_per_second": 10.167,
-      "step": 16900
+      "epoch": 0.05,
+      "grad_norm": 3.671875,
+      "learning_rate": 1.705263157894737e-05,
+      "loss": 0.5969,
+      "step": 1900
     },
     {
-      "epoch": 0.42,
-      "grad_norm": 5.851967811584473,
-      "learning_rate": 3.0769230769230774e-06,
-      "loss": 0.6582,
-      "step": 17000
+      "epoch": 0.05,
+      "grad_norm": 4.65625,
+      "learning_rate": 1.7031578947368423e-05,
+      "loss": 0.5967,
+      "step": 1910
     },
     {
-      "epoch": 0.42,
-      "eval_loss": 0.6755939722061157,
-      "eval_runtime": 98.3485,
-      "eval_samples_per_second": 10.168,
-      "eval_steps_per_second": 10.168,
-      "step": 17000
+      "epoch": 0.05,
+      "grad_norm": 8.5625,
+      "learning_rate": 1.7010526315789475e-05,
+      "loss": 0.6486,
+      "step": 1920
     },
     {
-      "epoch": 0.43,
-      "grad_norm": 3.1814448833465576,
-      "learning_rate": 2.9743589743589746e-06,
-      "loss": 0.6471,
-      "step": 17100
+      "epoch": 0.05,
+      "grad_norm": 2.953125,
+      "learning_rate": 1.6989473684210528e-05,
+      "loss": 0.6343,
+      "step": 1930
     },
     {
-      "epoch": 0.43,
-      "eval_loss": 0.6580842733383179,
-      "eval_runtime": 98.3328,
-      "eval_samples_per_second": 10.17,
-      "eval_steps_per_second": 10.17,
-      "step": 17100
+      "epoch": 0.05,
+      "grad_norm": 4.53125,
+      "learning_rate": 1.696842105263158e-05,
+      "loss": 0.6137,
+      "step": 1940
     },
     {
-      "epoch": 0.43,
-      "grad_norm": 3.3297533988952637,
-      "learning_rate": 2.8717948717948717e-06,
-      "loss": 0.6238,
-      "step": 17200
+      "epoch": 0.05,
+      "grad_norm": 3.1875,
+      "learning_rate": 1.6947368421052633e-05,
+      "loss": 0.5584,
+      "step": 1950
     },
     {
-      "epoch": 0.43,
-      "eval_loss": 0.6540150046348572,
-      "eval_runtime": 98.392,
-      "eval_samples_per_second": 10.163,
-      "eval_steps_per_second": 10.163,
-      "step": 17200
+      "epoch": 0.05,
+      "grad_norm": 4.6875,
+      "learning_rate": 1.6926315789473686e-05,
+      "loss": 0.5924,
+      "step": 1960
     },
     {
-      "epoch": 0.43,
-      "grad_norm": 4.152995586395264,
-      "learning_rate": 2.7692307692307697e-06,
-      "loss": 0.6523,
-      "step": 17300
+      "epoch": 0.05,
+      "grad_norm": 3.359375,
+      "learning_rate": 1.690526315789474e-05,
+      "loss": 0.5658,
+      "step": 1970
     },
     {
-      "epoch": 0.43,
-      "eval_loss": 0.6668370366096497,
-      "eval_runtime": 98.343,
-      "eval_samples_per_second": 10.168,
-      "eval_steps_per_second": 10.168,
-      "step": 17300
+      "epoch": 0.05,
+      "grad_norm": 3.921875,
+      "learning_rate": 1.688421052631579e-05,
+      "loss": 0.6455,
+      "step": 1980
     },
     {
-      "epoch": 0.43,
-      "grad_norm": 8.26444149017334,
-      "learning_rate": 2.666666666666667e-06,
-      "loss": 0.6453,
-      "step": 17400
+      "epoch": 0.05,
+      "grad_norm": 4.0625,
+      "learning_rate": 1.6863157894736844e-05,
+      "loss": 0.7716,
+      "step": 1990
     },
     {
-      "epoch": 0.43,
-      "eval_loss": 0.6706939935684204,
-      "eval_runtime": 98.3728,
-      "eval_samples_per_second": 10.165,
-      "eval_steps_per_second": 10.165,
-      "step": 17400
+      "epoch": 0.05,
+      "grad_norm": 5.34375,
+      "learning_rate": 1.6842105263157896e-05,
+      "loss": 0.6834,
+      "step": 2000
     },
     {
-      "epoch": 0.44,
-      "grad_norm": 4.3345627784729,
-      "learning_rate": 2.564102564102564e-06,
-      "loss": 0.6436,
-      "step": 17500
+      "epoch": 0.05,
+      "eval_loss": 0.6163671612739563,
+      "eval_runtime": 38.1399,
+      "eval_samples_per_second": 26.219,
+      "eval_steps_per_second": 26.219,
+      "step": 2000
     },
     {
-      "epoch": 0.44,
-      "eval_loss": 0.6370129585266113,
-      "eval_runtime": 98.3453,
-      "eval_samples_per_second": 10.168,
-      "eval_steps_per_second": 10.168,
-      "step": 17500
+      "epoch": 0.05,
+      "grad_norm": 5.3125,
+      "learning_rate": 1.682105263157895e-05,
+      "loss": 0.655,
+      "step": 2010
     },
     {
-      "epoch": 0.44,
-      "grad_norm": 4.079812526702881,
-      "learning_rate": 2.461538461538462e-06,
-      "loss": 0.6369,
-      "step": 17600
+      "epoch": 0.05,
+      "grad_norm": 8.625,
+      "learning_rate": 1.6800000000000002e-05,
+      "loss": 0.6942,
+      "step": 2020
     },
     {
-      "epoch": 0.44,
-      "eval_loss": 0.6634325385093689,
-      "eval_runtime": 98.3802,
-      "eval_samples_per_second": 10.165,
-      "eval_steps_per_second": 10.165,
-      "step": 17600
+      "epoch": 0.05,
+      "grad_norm": 4.8125,
+      "learning_rate": 1.6778947368421054e-05,
+      "loss": 0.6266,
+      "step": 2030
     },
     {
-      "epoch": 0.44,
-      "grad_norm": 3.0387985706329346,
-      "learning_rate": 2.358974358974359e-06,
-      "loss": 0.6244,
-      "step": 17700
+      "epoch": 0.05,
+      "grad_norm": 6.46875,
+      "learning_rate": 1.6757894736842107e-05,
+      "loss": 0.6633,
+      "step": 2040
     },
     {
-      "epoch": 0.44,
-      "eval_loss": 0.658197283744812,
-      "eval_runtime": 98.3504,
-      "eval_samples_per_second": 10.168,
-      "eval_steps_per_second": 10.168,
-      "step": 17700
+      "epoch": 0.05,
+      "grad_norm": 6.625,
+      "learning_rate": 1.673684210526316e-05,
+      "loss": 0.6354,
+      "step": 2050
     },
     {
-      "epoch": 0.45,
-      "grad_norm": 4.429220199584961,
-      "learning_rate": 2.2564102564102566e-06,
-      "loss": 0.663,
-      "step": 17800
+      "epoch": 0.05,
+      "grad_norm": 3.203125,
+      "learning_rate": 1.6715789473684212e-05,
+      "loss": 0.6723,
+      "step": 2060
     },
     {
-      "epoch": 0.45,
-      "eval_loss": 0.663162350654602,
-      "eval_runtime": 98.3612,
-      "eval_samples_per_second": 10.167,
-      "eval_steps_per_second": 10.167,
-      "step": 17800
+      "epoch": 0.05,
+      "grad_norm": 3.734375,
+      "learning_rate": 1.6694736842105265e-05,
+      "loss": 0.677,
+      "step": 2070
     },
     {
-      "epoch": 0.45,
-      "grad_norm": 7.286205291748047,
-      "learning_rate": 2.153846153846154e-06,
-      "loss": 0.6692,
-      "step": 17900
+      "epoch": 0.05,
+      "grad_norm": 4.28125,
+      "learning_rate": 1.6673684210526318e-05,
+      "loss": 0.6329,
+      "step": 2080
     },
     {
-      "epoch": 0.45,
-      "eval_loss": 0.6493304371833801,
-      "eval_runtime": 98.3416,
-      "eval_samples_per_second": 10.169,
-      "eval_steps_per_second": 10.169,
-      "step": 17900
+      "epoch": 0.05,
+      "grad_norm": 7.90625,
+      "learning_rate": 1.665263157894737e-05,
+      "loss": 0.6193,
+      "step": 2090
     },
     {
-      "epoch": 0.45,
-      "grad_norm": 3.1025142669677734,
-      "learning_rate": 2.0512820512820513e-06,
-      "loss": 0.6515,
-      "step": 18000
+      "epoch": 0.05,
+      "grad_norm": 6.46875,
+      "learning_rate": 1.6631578947368423e-05,
+      "loss": 0.6225,
+      "step": 2100
     },
     {
-      "epoch": 0.45,
-      "eval_loss": 0.6505727767944336,
-      "eval_runtime": 98.3601,
-      "eval_samples_per_second": 10.167,
-      "eval_steps_per_second": 10.167,
-      "step": 18000
+      "epoch": 0.05,
+      "grad_norm": 2.75,
+      "learning_rate": 1.6610526315789476e-05,
+      "loss": 0.6449,
+      "step": 2110
     },
     {
-      "epoch": 0.45,
-      "grad_norm": 5.170887470245361,
-      "learning_rate": 1.948717948717949e-06,
-      "loss": 0.6554,
-      "step": 18100
+      "epoch": 0.05,
+      "grad_norm": 3.921875,
+      "learning_rate": 1.658947368421053e-05,
+      "loss": 0.5844,
+      "step": 2120
     },
     {
-      "epoch": 0.45,
-      "eval_loss": 0.640792191028595,
-      "eval_runtime": 98.3695,
-      "eval_samples_per_second": 10.166,
-      "eval_steps_per_second": 10.166,
-      "step": 18100
+      "epoch": 0.05,
+      "grad_norm": 6.59375,
+      "learning_rate": 1.656842105263158e-05,
+      "loss": 0.6374,
+      "step": 2130
     },
     {
-      "epoch": 0.46,
-      "grad_norm": 8.582660675048828,
-      "learning_rate": 1.8461538461538465e-06,
-      "loss": 0.6481,
-      "step": 18200
+      "epoch": 0.05,
+      "grad_norm": 6.0,
+      "learning_rate": 1.6547368421052634e-05,
+      "loss": 0.6147,
+      "step": 2140
     },
     {
-      "epoch": 0.46,
-      "eval_loss": 0.6477890014648438,
-      "eval_runtime": 98.3746,
-      "eval_samples_per_second": 10.165,
-      "eval_steps_per_second": 10.165,
-      "step": 18200
+      "epoch": 0.05,
+      "grad_norm": 5.6875,
+      "learning_rate": 1.6526315789473686e-05,
+      "loss": 0.6639,
+      "step": 2150
     },
     {
-      "epoch": 0.46,
-      "grad_norm": 7.4102301597595215,
-      "learning_rate": 1.7435897435897436e-06,
-      "loss": 0.6284,
-      "step": 18300
+      "epoch": 0.05,
+      "grad_norm": 3.109375,
+      "learning_rate": 1.650526315789474e-05,
+      "loss": 0.6715,
+      "step": 2160
     },
     {
-      "epoch": 0.46,
-      "eval_loss": 0.6732329726219177,
-      "eval_runtime": 98.3003,
-      "eval_samples_per_second": 10.173,
-      "eval_steps_per_second": 10.173,
-      "step": 18300
+      "epoch": 0.05,
+      "grad_norm": 2.40625,
+      "learning_rate": 1.648421052631579e-05,
+      "loss": 0.718,
+      "step": 2170
     },
     {
-      "epoch": 0.46,
-      "grad_norm": 3.6458799839019775,
-      "learning_rate": 1.6410256410256412e-06,
-      "loss": 0.6546,
-      "step": 18400
+      "epoch": 0.05,
+      "grad_norm": 5.71875,
+      "learning_rate": 1.6463157894736844e-05,
+      "loss": 0.7205,
+      "step": 2180
     },
     {
-      "epoch": 0.46,
-      "eval_loss": 0.6304293274879456,
-      "eval_runtime": 98.3207,
-      "eval_samples_per_second": 10.171,
-      "eval_steps_per_second": 10.171,
-      "step": 18400
+      "epoch": 0.05,
+      "grad_norm": 4.3125,
+      "learning_rate": 1.6442105263157897e-05,
+      "loss": 0.6613,
+      "step": 2190
     },
     {
-      "epoch": 0.46,
-      "grad_norm": 3.8108537197113037,
-      "learning_rate": 1.5384615384615387e-06,
-      "loss": 0.6665,
-      "step": 18500
+      "epoch": 0.06,
+      "grad_norm": 2.6875,
+      "learning_rate": 1.642105263157895e-05,
+      "loss": 0.7353,
+      "step": 2200
     },
     {
-      "epoch": 0.46,
-      "eval_loss": 0.634360134601593,
-      "eval_runtime": 98.3511,
-      "eval_samples_per_second": 10.168,
-      "eval_steps_per_second": 10.168,
-      "step": 18500
+      "epoch": 0.06,
+      "grad_norm": 2.078125,
+      "learning_rate": 1.64e-05,
+      "loss": 0.5747,
+      "step": 2210
     },
     {
-      "epoch": 0.47,
-      "grad_norm": 9.306204795837402,
-      "learning_rate": 1.4358974358974359e-06,
-      "loss": 0.6559,
-      "step": 18600
+      "epoch": 0.06,
+      "grad_norm": 5.84375,
+      "learning_rate": 1.6378947368421055e-05,
+      "loss": 0.4669,
+      "step": 2220
     },
     {
-      "epoch": 0.47,
-      "eval_loss": 0.6353975534439087,
-      "eval_runtime": 98.3306,
-      "eval_samples_per_second": 10.17,
-      "eval_steps_per_second": 10.17,
-      "step": 18600
+      "epoch": 0.06,
+      "grad_norm": 4.15625,
+      "learning_rate": 1.6357894736842108e-05,
+      "loss": 0.5741,
+      "step": 2230
     },
     {
-      "epoch": 0.47,
-      "grad_norm": 4.001993656158447,
-      "learning_rate": 1.3333333333333334e-06,
-      "loss": 0.6386,
-      "step": 18700
+      "epoch": 0.06,
+      "grad_norm": 2.265625,
+      "learning_rate": 1.633684210526316e-05,
+      "loss": 0.6338,
+      "step": 2240
     },
     {
-      "epoch": 0.47,
-      "eval_loss": 0.6281168460845947,
-      "eval_runtime": 98.2978,
-      "eval_samples_per_second": 10.173,
-      "eval_steps_per_second": 10.173,
-      "step": 18700
+      "epoch": 0.06,
+      "grad_norm": 2.96875,
+      "learning_rate": 1.6315789473684213e-05,
+      "loss": 0.5429,
+      "step": 2250
     },
     {
-      "epoch": 0.47,
-      "grad_norm": 4.9915924072265625,
-      "learning_rate": 1.230769230769231e-06,
-      "loss": 0.6509,
-      "step": 18800
+      "epoch": 0.06,
+      "grad_norm": 2.34375,
+      "learning_rate": 1.6294736842105265e-05,
+      "loss": 0.6671,
+      "step": 2260
     },
     {
-      "epoch": 0.47,
-      "eval_loss": 0.6473774313926697,
-      "eval_runtime": 98.3123,
-      "eval_samples_per_second": 10.172,
-      "eval_steps_per_second": 10.172,
-      "step": 18800
+      "epoch": 0.06,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.6273684210526318e-05,
+      "loss": 0.681,
+      "step": 2270
     },
     {
-      "epoch": 0.47,
-      "grad_norm": 3.8955130577087402,
-      "learning_rate": 1.1282051282051283e-06,
-      "loss": 0.6272,
-      "step": 18900
+      "epoch": 0.06,
+      "grad_norm": 10.125,
+      "learning_rate": 1.6252631578947367e-05,
+      "loss": 0.5757,
+      "step": 2280
     },
     {
-      "epoch": 0.47,
-      "eval_loss": 0.6473217606544495,
-      "eval_runtime": 98.3521,
-      "eval_samples_per_second": 10.168,
-      "eval_steps_per_second": 10.168,
-      "step": 18900
+      "epoch": 0.06,
+      "grad_norm": 4.75,
+      "learning_rate": 1.6231578947368423e-05,
+      "loss": 0.559,
+      "step": 2290
     },
     {
-      "epoch": 0.47,
-      "grad_norm": 9.715432167053223,
-      "learning_rate": 1.0256410256410257e-06,
-      "loss": 0.6086,
-      "step": 19000
+      "epoch": 0.06,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.6210526315789473e-05,
+      "loss": 0.6697,
+      "step": 2300
     },
     {
-      "epoch": 0.47,
-      "eval_loss": 0.620388925075531,
-      "eval_runtime": 98.3358,
-      "eval_samples_per_second": 10.169,
-      "eval_steps_per_second": 10.169,
-      "step": 19000
+      "epoch": 0.06,
+      "grad_norm": 5.15625,
+      "learning_rate": 1.618947368421053e-05,
+      "loss": 0.6065,
+      "step": 2310
     },
     {
-      "epoch": 0.48,
-      "grad_norm": 4.22080659866333,
-      "learning_rate": 9.230769230769232e-07,
-      "loss": 0.6553,
-      "step": 19100
+      "epoch": 0.06,
+      "grad_norm": 5.59375,
+      "learning_rate": 1.616842105263158e-05,
+      "loss": 0.6676,
+      "step": 2320
     },
     {
-      "epoch": 0.48,
-      "eval_loss": 0.6517868041992188,
-      "eval_runtime": 98.326,
-      "eval_samples_per_second": 10.17,
-      "eval_steps_per_second": 10.17,
-      "step": 19100
+      "epoch": 0.06,
+      "grad_norm": 5.1875,
+      "learning_rate": 1.6147368421052634e-05,
+      "loss": 0.5642,
+      "step": 2330
     },
     {
-      "epoch": 0.48,
-      "grad_norm": 3.7308225631713867,
-      "learning_rate": 8.205128205128206e-07,
-      "loss": 0.6119,
-      "step": 19200
+      "epoch": 0.06,
+      "grad_norm": 4.1875,
+      "learning_rate": 1.6126315789473687e-05,
+      "loss": 0.6657,
+      "step": 2340
     },
     {
-      "epoch": 0.48,
-      "eval_loss": 0.639901340007782,
-      "eval_runtime": 98.3,
-      "eval_samples_per_second": 10.173,
-      "eval_steps_per_second": 10.173,
-      "step": 19200
+      "epoch": 0.06,
+      "grad_norm": 2.921875,
+      "learning_rate": 1.6105263157894736e-05,
+      "loss": 0.6927,
+      "step": 2350
     },
     {
-      "epoch": 0.48,
-      "grad_norm": 4.890966892242432,
-      "learning_rate": 7.179487179487179e-07,
-      "loss": 0.6363,
-      "step": 19300
+      "epoch": 0.06,
+      "grad_norm": 5.46875,
+      "learning_rate": 1.6084210526315792e-05,
+      "loss": 0.6291,
+      "step": 2360
     },
     {
-      "epoch": 0.48,
-      "eval_loss": 0.6639401912689209,
-      "eval_runtime": 98.3404,
-      "eval_samples_per_second": 10.169,
-      "eval_steps_per_second": 10.169,
-      "step": 19300
+      "epoch": 0.06,
+      "grad_norm": 2.640625,
+      "learning_rate": 1.606315789473684e-05,
+      "loss": 0.728,
+      "step": 2370
     },
     {
-      "epoch": 0.48,
-      "grad_norm": 6.855196952819824,
-      "learning_rate": 6.153846153846155e-07,
-      "loss": 0.6261,
-      "step": 19400
+      "epoch": 0.06,
+      "grad_norm": 6.4375,
+      "learning_rate": 1.6042105263157897e-05,
+      "loss": 0.621,
+      "step": 2380
     },
     {
-      "epoch": 0.48,
-      "eval_loss": 0.6473622918128967,
-      "eval_runtime": 98.3312,
-      "eval_samples_per_second": 10.17,
-      "eval_steps_per_second": 10.17,
-      "step": 19400
+      "epoch": 0.06,
+      "grad_norm": 3.078125,
+      "learning_rate": 1.6021052631578947e-05,
+      "loss": 0.6176,
+      "step": 2390
     },
     {
-      "epoch": 0.49,
-      "grad_norm": 2.988701581954956,
-      "learning_rate": 5.128205128205128e-07,
-      "loss": 0.6192,
-      "step": 19500
+      "epoch": 0.06,
+      "grad_norm": 4.125,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.6321,
+      "step": 2400
     },
     {
-      "epoch": 0.49,
-      "eval_loss": 0.6319825649261475,
-      "eval_runtime": 98.277,
-      "eval_samples_per_second": 10.175,
-      "eval_steps_per_second": 10.175,
-      "step": 19500
+      "epoch": 0.06,
+      "grad_norm": 3.25,
+      "learning_rate": 1.5978947368421055e-05,
+      "loss": 0.6174,
+      "step": 2410
     },
     {
-      "epoch": 0.49,
-      "grad_norm": 3.922107696533203,
-      "learning_rate": 4.102564102564103e-07,
-      "loss": 0.6123,
-      "step": 19600
+      "epoch": 0.06,
+      "grad_norm": 3.328125,
+      "learning_rate": 1.5957894736842105e-05,
+      "loss": 0.5024,
+      "step": 2420
     },
     {
-      "epoch": 0.49,
-      "eval_loss": 0.6252996325492859,
-      "eval_runtime": 98.3425,
-      "eval_samples_per_second": 10.169,
-      "eval_steps_per_second": 10.169,
-      "step": 19600
+      "epoch": 0.06,
+      "grad_norm": 4.4375,
+      "learning_rate": 1.593684210526316e-05,
+      "loss": 0.5771,
+      "step": 2430
     },
     {
-      "epoch": 0.49,
-      "grad_norm": 13.531473159790039,
-      "learning_rate": 3.0769230769230774e-07,
-      "loss": 0.6128,
-      "step": 19700
+      "epoch": 0.06,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.591578947368421e-05,
+      "loss": 0.6005,
+      "step": 2440
     },
     {
-      "epoch": 0.49,
-      "eval_loss": 0.6362013220787048,
-      "eval_runtime": 98.2906,
-      "eval_samples_per_second": 10.174,
-      "eval_steps_per_second": 10.174,
-      "step": 19700
+      "epoch": 0.06,
+      "grad_norm": 6.875,
+      "learning_rate": 1.5894736842105266e-05,
+      "loss": 0.6532,
+      "step": 2450
     },
     {
-      "epoch": 0.49,
-      "grad_norm": 13.287760734558105,
-      "learning_rate": 2.0512820512820514e-07,
-      "loss": 0.6321,
-      "step": 19800
+      "epoch": 0.06,
+      "grad_norm": 4.21875,
+      "learning_rate": 1.5873684210526315e-05,
+      "loss": 0.7415,
+      "step": 2460
     },
     {
-      "epoch": 0.49,
-      "eval_loss": 0.6311822533607483,
-      "eval_runtime": 98.3206,
-      "eval_samples_per_second": 10.171,
-      "eval_steps_per_second": 10.171,
-      "step": 19800
+      "epoch": 0.06,
+      "grad_norm": 3.84375,
+      "learning_rate": 1.585263157894737e-05,
+      "loss": 0.7431,
+      "step": 2470
     },
     {
-      "epoch": 0.5,
-      "grad_norm": 10.524887084960938,
-      "learning_rate": 1.0256410256410257e-07,
-      "loss": 0.6026,
-      "step": 19900
+      "epoch": 0.06,
+      "grad_norm": 6.40625,
+      "learning_rate": 1.5831578947368424e-05,
+      "loss": 0.4914,
+      "step": 2480
     },
     {
-      "epoch": 0.5,
-      "eval_loss": 0.6232057809829712,
-      "eval_runtime": 98.3191,
-      "eval_samples_per_second": 10.171,
-      "eval_steps_per_second": 10.171,
-      "step": 19900
+      "epoch": 0.06,
+      "grad_norm": 2.59375,
+      "learning_rate": 1.5810526315789473e-05,
+      "loss": 0.5853,
+      "step": 2490
     },
     {
-      "epoch": 0.5,
-      "grad_norm": 4.0852131843566895,
-      "learning_rate": 0.0,
-      "loss": 0.6322,
-      "step": 20000
+      "epoch": 0.06,
+      "grad_norm": 5.375,
+      "learning_rate": 1.578947368421053e-05,
+      "loss": 0.7048,
+      "step": 2500
     },
     {
-      "epoch": 0.5,
-      "eval_loss": 0.632331907749176,
-      "eval_runtime": 98.3118,
-      "eval_samples_per_second": 10.172,
-      "eval_steps_per_second": 10.172,
-      "step": 20000
+      "epoch": 0.06,
+      "eval_loss": 0.637090802192688,
+      "eval_runtime": 38.18,
+      "eval_samples_per_second": 26.192,
+      "eval_steps_per_second": 26.192,
+      "step": 2500
     }
   ],
-  "logging_steps": 100,
-  "max_steps": 20000,
+  "logging_steps": 10,
+  "max_steps": 10000,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 1,
   "save_steps": 2500,
-  "total_flos": 3.1467396661248e+17,
+  "total_flos": 3.933424582656e+16,
   "train_batch_size": 1,
   "trial_name": null,
   "trial_params": null