diff --git "a/checkpoint-4800/trainer_state.json" "b/checkpoint-4800/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-4800/trainer_state.json"
@@ -0,0 +1,4600 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.5513897866839044,
+  "global_step": 4800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 0.0002,
+      "loss": 0.9529,
+      "step": 10
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002,
+      "loss": 0.7908,
+      "step": 20
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002,
+      "loss": 0.7173,
+      "step": 30
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0.0002,
+      "loss": 0.7776,
+      "step": 40
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002,
+      "loss": 0.6779,
+      "step": 50
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002,
+      "loss": 0.7011,
+      "step": 60
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0002,
+      "loss": 0.7156,
+      "step": 70
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002,
+      "loss": 0.7418,
+      "step": 80
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002,
+      "loss": 0.7465,
+      "step": 90
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002,
+      "loss": 0.7595,
+      "step": 100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002,
+      "loss": 0.6816,
+      "step": 110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002,
+      "loss": 0.7981,
+      "step": 120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0002,
+      "loss": 0.6494,
+      "step": 130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002,
+      "loss": 0.7423,
+      "step": 140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002,
+      "loss": 0.7555,
+      "step": 150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002,
+      "loss": 0.5799,
+      "step": 160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0002,
+      "loss": 0.699,
+      "step": 170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002,
+      "loss": 0.7154,
+      "step": 180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002,
+      "loss": 0.6096,
+      "step": 190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0002,
+      "loss": 0.5713,
+      "step": 200
+    },
+    {
+      "epoch": 0.06,
+      "eval_loss": 0.6557502150535583,
+      "eval_runtime": 92.7109,
+      "eval_samples_per_second": 10.786,
+      "eval_steps_per_second": 5.393,
+      "step": 200
+    },
+    {
+      "epoch": 0.06,
+      "mmlu_eval_accuracy": 0.47294392264572244,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.4375,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.41379310344827586,
+      "mmlu_eval_accuracy_college_biology": 0.5,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_medicine": 0.3181818181818182,
+      "mmlu_eval_accuracy_college_physics": 0.5454545454545454,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+      "mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.40625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.3333333333333333,
+      "mmlu_eval_accuracy_high_school_european_history": 0.4444444444444444,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.3488372093023256,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+      "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7166666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
+      "mmlu_eval_accuracy_high_school_us_history": 0.5,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5769230769230769,
+      "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.18181818181818182,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.5454545454545454,
+      "mmlu_eval_accuracy_marketing": 0.72,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.6395348837209303,
+      "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
+      "mmlu_eval_accuracy_moral_scenarios": 0.25,
+      "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
+      "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
+      "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
+      "mmlu_eval_accuracy_professional_accounting": 0.25806451612903225,
+      "mmlu_eval_accuracy_professional_law": 0.3176470588235294,
+      "mmlu_eval_accuracy_professional_medicine": 0.4838709677419355,
+      "mmlu_eval_accuracy_professional_psychology": 0.391304347826087,
+      "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
+      "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
+      "mmlu_eval_accuracy_sociology": 0.7727272727272727,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 0.8432865626915472,
+      "step": 200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002,
+      "loss": 0.6339,
+      "step": 210
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002,
+      "loss": 0.701,
+      "step": 220
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0002,
+      "loss": 0.689,
+      "step": 230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002,
+      "loss": 0.6955,
+      "step": 240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002,
+      "loss": 0.6124,
+      "step": 250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0002,
+      "loss": 0.6648,
+      "step": 260
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002,
+      "loss": 0.7364,
+      "step": 270
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002,
+      "loss": 0.558,
+      "step": 280
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0002,
+      "loss": 0.7089,
+      "step": 290
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002,
+      "loss": 0.5691,
+      "step": 300
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002,
+      "loss": 0.7079,
+      "step": 310
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0002,
+      "loss": 0.6558,
+      "step": 320
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002,
+      "loss": 0.5887,
+      "step": 330
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002,
+      "loss": 0.704,
+      "step": 340
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002,
+      "loss": 0.638,
+      "step": 350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002,
+      "loss": 0.708,
+      "step": 360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002,
+      "loss": 0.6203,
+      "step": 370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0002,
+      "loss": 0.6288,
+      "step": 380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002,
+      "loss": 0.6725,
+      "step": 390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002,
+      "loss": 0.6849,
+      "step": 400
+    },
+    {
+      "epoch": 0.13,
+      "eval_loss": 0.6309903860092163,
+      "eval_runtime": 92.7675,
+      "eval_samples_per_second": 10.78,
+      "eval_steps_per_second": 5.39,
+      "step": 400
+    },
+    {
+      "epoch": 0.13,
+      "mmlu_eval_accuracy": 0.46759845160702646,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+      "mmlu_eval_accuracy_astronomy": 0.4375,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4482758620689655,
+      "mmlu_eval_accuracy_college_biology": 0.375,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.5454545454545454,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.5,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.4375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.2926829268292683,
+      "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
+      "mmlu_eval_accuracy_global_facts": 0.4,
+      "mmlu_eval_accuracy_high_school_biology": 0.375,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.32558139534883723,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.23076923076923078,
+      "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+      "mmlu_eval_accuracy_high_school_psychology": 0.75,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5,
+      "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+      "mmlu_eval_accuracy_human_sexuality": 0.5,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.72,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
+      "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
+      "mmlu_eval_accuracy_moral_scenarios": 0.25,
+      "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
+      "mmlu_eval_accuracy_philosophy": 0.5,
+      "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
+      "mmlu_eval_accuracy_professional_accounting": 0.25806451612903225,
+      "mmlu_eval_accuracy_professional_law": 0.3176470588235294,
+      "mmlu_eval_accuracy_professional_medicine": 0.41935483870967744,
+      "mmlu_eval_accuracy_professional_psychology": 0.391304347826087,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
+      "mmlu_eval_accuracy_sociology": 0.7272727272727273,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 0.8323084239996289,
+      "step": 400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0002,
+      "loss": 0.5261,
+      "step": 410
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002,
+      "loss": 0.6614,
+      "step": 420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002,
+      "loss": 0.6213,
+      "step": 430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0002,
+      "loss": 0.7206,
+      "step": 440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002,
+      "loss": 0.6944,
+      "step": 450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002,
+      "loss": 0.675,
+      "step": 460
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0002,
+      "loss": 0.5907,
+      "step": 470
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002,
+      "loss": 0.5778,
+      "step": 480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002,
+      "loss": 0.5583,
+      "step": 490
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002,
+      "loss": 0.7171,
+      "step": 500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0002,
+      "loss": 0.728,
+      "step": 510
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002,
+      "loss": 0.6037,
+      "step": 520
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002,
+      "loss": 0.6598,
+      "step": 530
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0002,
+      "loss": 0.6913,
+      "step": 540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002,
+      "loss": 0.6092,
+      "step": 550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002,
+      "loss": 0.589,
+      "step": 560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0002,
+      "loss": 0.5608,
+      "step": 570
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002,
+      "loss": 0.6251,
+      "step": 580
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002,
+      "loss": 0.6527,
+      "step": 590
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0002,
+      "loss": 0.6518,
+      "step": 600
+    },
+    {
+      "epoch": 0.19,
+      "eval_loss": 0.618711531162262,
+      "eval_runtime": 92.7537,
+      "eval_samples_per_second": 10.781,
+      "eval_steps_per_second": 5.391,
+      "step": 600
+    },
+    {
+      "epoch": 0.19,
+      "mmlu_eval_accuracy": 0.4660645174643213,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+      "mmlu_eval_accuracy_astronomy": 0.4375,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.4,
+      "mmlu_eval_accuracy_high_school_biology": 0.46875,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.37209302325581395,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.3103448275862069,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
+      "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7333333333333333,
+      "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5769230769230769,
+      "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.7692307692307693,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.68,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
+      "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
+      "mmlu_eval_accuracy_moral_scenarios": 0.26,
+      "mmlu_eval_accuracy_nutrition": 0.5454545454545454,
+      "mmlu_eval_accuracy_philosophy": 0.5,
+      "mmlu_eval_accuracy_prehistory": 0.4,
+      "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
+      "mmlu_eval_accuracy_professional_law": 0.32941176470588235,
+      "mmlu_eval_accuracy_professional_medicine": 0.45161290322580644,
+      "mmlu_eval_accuracy_professional_psychology": 0.42028985507246375,
+      "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
+      "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 0.8410176051775718,
+      "step": 600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002,
+      "loss": 0.6876,
+      "step": 610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002,
+      "loss": 0.7174,
+      "step": 620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0002,
+      "loss": 0.6002,
+      "step": 630
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002,
+      "loss": 0.7426,
+      "step": 640
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002,
+      "loss": 0.5213,
+      "step": 650
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002,
+      "loss": 0.574,
+      "step": 660
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002,
+      "loss": 0.6669,
+      "step": 670
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002,
+      "loss": 0.5847,
+      "step": 680
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0002,
+      "loss": 0.7233,
+      "step": 690
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002,
+      "loss": 0.6199,
+      "step": 700
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002,
+      "loss": 0.5977,
+      "step": 710
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0002,
+      "loss": 0.6853,
+      "step": 720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002,
+      "loss": 0.5777,
+      "step": 730
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002,
+      "loss": 0.5943,
+      "step": 740
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0002,
+      "loss": 0.6323,
+      "step": 750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002,
+      "loss": 0.6367,
+      "step": 760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002,
+      "loss": 0.6367,
+      "step": 770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0002,
+      "loss": 0.6197,
+      "step": 780
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002,
+      "loss": 0.5639,
+      "step": 790
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002,
+      "loss": 0.6281,
+      "step": 800
+    },
+    {
+      "epoch": 0.26,
+      "eval_loss": 0.6189825534820557,
+      "eval_runtime": 92.99,
+      "eval_samples_per_second": 10.754,
+      "eval_steps_per_second": 5.377,
+      "step": 800
+    },
+    {
+      "epoch": 0.26,
+      "mmlu_eval_accuracy": 0.48167538061853127,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.4375,
+      "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.6363636363636364,
+      "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
+      "mmlu_eval_accuracy_econometrics": 0.08333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
+      "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.40625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.4444444444444444,
+      "mmlu_eval_accuracy_high_school_european_history": 0.3888888888888889,
+      "mmlu_eval_accuracy_high_school_geography": 0.7272727272727273,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
+      "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7666666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5769230769230769,
+      "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+      "mmlu_eval_accuracy_human_sexuality": 0.5,
+      "mmlu_eval_accuracy_international_law": 0.7692307692307693,
+      "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.72,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.686046511627907,
+      "mmlu_eval_accuracy_moral_disputes": 0.5,
+      "mmlu_eval_accuracy_moral_scenarios": 0.27,
+      "mmlu_eval_accuracy_nutrition": 0.5454545454545454,
+      "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
+      "mmlu_eval_accuracy_prehistory": 0.5142857142857142,
+      "mmlu_eval_accuracy_professional_accounting": 0.25806451612903225,
+      "mmlu_eval_accuracy_professional_law": 0.3352941176470588,
+      "mmlu_eval_accuracy_professional_medicine": 0.4838709677419355,
+      "mmlu_eval_accuracy_professional_psychology": 0.4057971014492754,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 0.9606302572954105,
+      "step": 800
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0002,
+      "loss": 0.637,
+      "step": 810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002,
+      "loss": 0.6902,
+      "step": 820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002,
+      "loss": 0.5872,
+      "step": 830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002,
+      "loss": 0.6115,
+      "step": 840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0002,
+      "loss": 0.6499,
+      "step": 850
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002,
+      "loss": 0.5977,
+      "step": 860
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002,
+      "loss": 0.5873,
+      "step": 870
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0002,
+      "loss": 0.6776,
+      "step": 880
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002,
+      "loss": 0.6804,
+      "step": 890
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002,
+      "loss": 0.5914,
+      "step": 900
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0002,
+      "loss": 0.6313,
+      "step": 910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002,
+      "loss": 0.5869,
+      "step": 920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002,
+      "loss": 0.519,
+      "step": 930
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0002,
+      "loss": 0.4877,
+      "step": 940
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002,
+      "loss": 0.5931,
+      "step": 950
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002,
+      "loss": 0.6614,
+      "step": 960
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0002,
+      "loss": 0.6287,
+      "step": 970
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002,
+      "loss": 0.5139,
+      "step": 980
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002,
+      "loss": 0.6392,
+      "step": 990
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002,
+      "loss": 0.7001,
+      "step": 1000
+    },
+    {
+      "epoch": 0.32,
+      "eval_loss": 0.5796898007392883,
+      "eval_runtime": 92.7581,
+      "eval_samples_per_second": 10.781,
+      "eval_steps_per_second": 5.39,
+      "step": 1000
+    },
+    {
+      "epoch": 0.32,
+      "mmlu_eval_accuracy": 0.47460361123181055,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
+      "mmlu_eval_accuracy_astronomy": 0.375,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4482758620689655,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_physics": 0.5454545454545454,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
+      "mmlu_eval_accuracy_econometrics": 0.25,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.53125,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.4444444444444444,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5,
+      "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.37209302325581395,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.34615384615384615,
+      "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+      "mmlu_eval_accuracy_high_school_psychology": 0.75,
+      "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
+      "mmlu_eval_accuracy_high_school_us_history": 0.5454545454545454,
+      "mmlu_eval_accuracy_high_school_world_history": 0.46153846153846156,
+      "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+      "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+      "mmlu_eval_accuracy_international_law": 0.7692307692307693,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.72,
+      "mmlu_eval_accuracy_medical_genetics": 0.6363636363636364,
+      "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
+      "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
+      "mmlu_eval_accuracy_moral_scenarios": 0.29,
+      "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
+      "mmlu_eval_accuracy_philosophy": 0.5588235294117647,
+      "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
+      "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
+      "mmlu_eval_accuracy_professional_law": 0.3176470588235294,
+      "mmlu_eval_accuracy_professional_medicine": 0.41935483870967744,
+      "mmlu_eval_accuracy_professional_psychology": 0.4927536231884058,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
+      "mmlu_eval_accuracy_virology": 0.6111111111111112,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 0.8773848874586836,
+      "step": 1000
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002,
+      "loss": 0.6671,
+      "step": 1010
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002,
+      "loss": 0.5942,
+      "step": 1020
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0002,
+      "loss": 0.6236,
+      "step": 1030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002,
+      "loss": 0.6162,
+      "step": 1040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002,
+      "loss": 0.734,
+      "step": 1050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0002,
+      "loss": 0.6108,
+      "step": 1060
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002,
+      "loss": 0.6669,
+      "step": 1070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002,
+      "loss": 0.6991,
+      "step": 1080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0002,
+      "loss": 0.6696,
+      "step": 1090
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002,
+      "loss": 0.6511,
+      "step": 1100
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002,
+      "loss": 0.6007,
+      "step": 1110
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0002,
+      "loss": 0.5543,
+      "step": 1120
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002,
+      "loss": 0.6399,
+      "step": 1130
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002,
+      "loss": 0.569,
+      "step": 1140
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002,
+      "loss": 0.6436,
+      "step": 1150
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002,
+      "loss": 0.4886,
+      "step": 1160
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002,
+      "loss": 0.5431,
+      "step": 1170
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002,
+      "loss": 0.5516,
+      "step": 1180
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0002,
+      "loss": 0.5041,
+      "step": 1190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002,
+      "loss": 0.6737,
+      "step": 1200
+    },
+    {
+      "epoch": 0.39,
+      "eval_loss": 0.5647590160369873,
+      "eval_runtime": 92.7185,
+      "eval_samples_per_second": 10.785,
+      "eval_steps_per_second": 5.393,
+      "step": 1200
+    },
+    {
+      "epoch": 0.39,
+      "mmlu_eval_accuracy": 0.47729744414466774,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.5,
+      "mmlu_eval_accuracy_business_ethics": 0.36363636363636365,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
+      "mmlu_eval_accuracy_conceptual_physics": 0.5384615384615384,
+      "mmlu_eval_accuracy_econometrics": 0.08333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.4375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.53125,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.4444444444444444,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.46511627906976744,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8,
+      "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
+      "mmlu_eval_accuracy_high_school_us_history": 0.5454545454545454,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5384615384615384,
+      "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.7692307692307693,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.8181818181818182,
+      "mmlu_eval_accuracy_marketing": 0.76,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
+      "mmlu_eval_accuracy_moral_disputes": 0.5,
+      "mmlu_eval_accuracy_moral_scenarios": 0.27,
+      "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
+      "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
+      "mmlu_eval_accuracy_prehistory": 0.5428571428571428,
+      "mmlu_eval_accuracy_professional_accounting": 0.25806451612903225,
+      "mmlu_eval_accuracy_professional_law": 0.3352941176470588,
+      "mmlu_eval_accuracy_professional_medicine": 0.45161290322580644,
+      "mmlu_eval_accuracy_professional_psychology": 0.34782608695652173,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 0.9123951537490825,
+      "step": 1200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002,
+      "loss": 0.4926,
+      "step": 1210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0002,
+      "loss": 0.5814,
+      "step": 1220
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002,
+      "loss": 0.4985,
+      "step": 1230
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002,
+      "loss": 0.5966,
+      "step": 1240
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0002,
+      "loss": 0.6493,
+      "step": 1250
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002,
+      "loss": 0.5811,
+      "step": 1260
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002,
+      "loss": 0.563,
+      "step": 1270
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0002,
+      "loss": 0.6206,
+      "step": 1280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002,
+      "loss": 0.5657,
+      "step": 1290
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002,
+      "loss": 0.6061,
+      "step": 1300
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0002,
+      "loss": 0.5776,
+      "step": 1310
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002,
+      "loss": 0.6702,
+      "step": 1320
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002,
+      "loss": 0.5688,
+      "step": 1330
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0002,
+      "loss": 0.5237,
+      "step": 1340
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002,
+      "loss": 0.5883,
+      "step": 1350
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002,
+      "loss": 0.5206,
+      "step": 1360
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0002,
+      "loss": 0.6948,
+      "step": 1370
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002,
+      "loss": 0.5656,
+      "step": 1380
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002,
+      "loss": 0.6026,
+      "step": 1390
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0002,
+      "loss": 0.6319,
+      "step": 1400
+    },
+    {
+      "epoch": 0.45,
+      "eval_loss": 0.5552607774734497,
+      "eval_runtime": 92.7391,
+      "eval_samples_per_second": 10.783,
+      "eval_steps_per_second": 5.391,
+      "step": 1400
+    },
+    {
+      "epoch": 0.45,
+      "mmlu_eval_accuracy": 0.4812787579159337,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.375,
+      "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.41379310344827586,
+      "mmlu_eval_accuracy_college_biology": 0.5,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.7272727272727273,
+      "mmlu_eval_accuracy_conceptual_physics": 0.5,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+      "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.5,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.3333333333333333,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.37209302325581395,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5384615384615384,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7833333333333333,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.5909090909090909,
+      "mmlu_eval_accuracy_high_school_world_history": 0.6153846153846154,
+      "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.7692307692307693,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.8181818181818182,
+      "mmlu_eval_accuracy_marketing": 0.76,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
+      "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
+      "mmlu_eval_accuracy_moral_scenarios": 0.25,
+      "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
+      "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
+      "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+      "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
+      "mmlu_eval_accuracy_professional_law": 0.3,
+      "mmlu_eval_accuracy_professional_medicine": 0.4838709677419355,
+      "mmlu_eval_accuracy_professional_psychology": 0.391304347826087,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.4444444444444444,
+      "mmlu_eval_accuracy_sociology": 0.7272727272727273,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.4444444444444444,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 0.8453598066862896,
+      "step": 1400
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002,
+      "loss": 0.4931,
+      "step": 1410
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002,
+      "loss": 0.4544,
+      "step": 1420
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0002,
+      "loss": 0.563,
+      "step": 1430
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002,
+      "loss": 0.4629,
+      "step": 1440
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002,
+      "loss": 0.4903,
+      "step": 1450
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0002,
+      "loss": 0.5581,
+      "step": 1460
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002,
+      "loss": 0.6571,
+      "step": 1470
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002,
+      "loss": 0.5562,
+      "step": 1480
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002,
+      "loss": 0.4745,
+      "step": 1490
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002,
+      "loss": 0.6383,
+      "step": 1500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002,
+      "loss": 0.6477,
+      "step": 1510
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002,
+      "loss": 0.5758,
+      "step": 1520
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002,
+      "loss": 0.5195,
+      "step": 1530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002,
+      "loss": 0.5691,
+      "step": 1540
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002,
+      "loss": 0.5451,
+      "step": 1550
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0002,
+      "loss": 0.5753,
+      "step": 1560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0002,
+      "loss": 0.5859,
+      "step": 1570
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0002,
+      "loss": 0.6325,
+      "step": 1580
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0002,
+      "loss": 0.7023,
+      "step": 1590
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002,
+      "loss": 0.5009,
+      "step": 1600
+    },
+    {
+      "epoch": 0.52,
+      "eval_loss": 0.5365247130393982,
+      "eval_runtime": 92.8248,
+      "eval_samples_per_second": 10.773,
+      "eval_steps_per_second": 5.386,
+      "step": 1600
+    },
+    {
+      "epoch": 0.52,
+      "mmlu_eval_accuracy": 0.4625997452457202,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.4375,
+      "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.0,
+      "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
+      "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
+      "mmlu_eval_accuracy_econometrics": 0.08333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+      "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.5625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.38461538461538464,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7333333333333333,
+      "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
+      "mmlu_eval_accuracy_high_school_us_history": 0.5454545454545454,
+      "mmlu_eval_accuracy_high_school_world_history": 0.4230769230769231,
+      "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+      "mmlu_eval_accuracy_human_sexuality": 0.25,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.76,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.6162790697674418,
+      "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
+      "mmlu_eval_accuracy_moral_scenarios": 0.24,
+      "mmlu_eval_accuracy_nutrition": 0.5454545454545454,
+      "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
+      "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+      "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
+      "mmlu_eval_accuracy_professional_law": 0.34705882352941175,
+      "mmlu_eval_accuracy_professional_medicine": 0.45161290322580644,
+      "mmlu_eval_accuracy_professional_psychology": 0.463768115942029,
+      "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
+      "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
+      "mmlu_eval_accuracy_sociology": 0.5454545454545454,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 0.8016339474243377,
+      "step": 1600
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002,
+      "loss": 0.5698,
+      "step": 1610
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002,
+      "loss": 0.5115,
+      "step": 1620
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002,
+      "loss": 0.5908,
+      "step": 1630
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002,
+      "loss": 0.6114,
+      "step": 1640
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002,
+      "loss": 0.5335,
+      "step": 1650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0002,
+      "loss": 0.6275,
+      "step": 1660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0002,
+      "loss": 0.4862,
+      "step": 1670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0002,
+      "loss": 0.6334,
+      "step": 1680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002,
+      "loss": 0.6494,
+      "step": 1690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002,
+      "loss": 0.5298,
+      "step": 1700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002,
+      "loss": 0.5359,
+      "step": 1710
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002,
+      "loss": 0.5049,
+      "step": 1720
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002,
+      "loss": 0.5015,
+      "step": 1730
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002,
+      "loss": 0.6523,
+      "step": 1740
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002,
+      "loss": 0.5047,
+      "step": 1750
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002,
+      "loss": 0.5321,
+      "step": 1760
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002,
+      "loss": 0.5595,
+      "step": 1770
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002,
+      "loss": 0.7094,
+      "step": 1780
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002,
+      "loss": 0.6828,
+      "step": 1790
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002,
+      "loss": 0.5355,
+      "step": 1800
+    },
+    {
+      "epoch": 0.58,
+      "eval_loss": 0.533126711845398,
+      "eval_runtime": 92.8481,
+      "eval_samples_per_second": 10.77,
+      "eval_steps_per_second": 5.385,
+      "step": 1800
+    },
+    {
+      "epoch": 0.58,
+      "mmlu_eval_accuracy": 0.4894584764122198,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.5,
+      "mmlu_eval_accuracy_astronomy": 0.4375,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.3793103448275862,
+      "mmlu_eval_accuracy_college_biology": 0.375,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.5384615384615384,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.6,
+      "mmlu_eval_accuracy_high_school_biology": 0.40625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
+      "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.3488372093023256,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
+      "mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7333333333333333,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.5909090909090909,
+      "mmlu_eval_accuracy_high_school_world_history": 0.46153846153846156,
+      "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+      "mmlu_eval_accuracy_human_sexuality": 0.5833333333333334,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.8181818181818182,
+      "mmlu_eval_accuracy_marketing": 0.72,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
+      "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
+      "mmlu_eval_accuracy_moral_scenarios": 0.25,
+      "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
+      "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
+      "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+      "mmlu_eval_accuracy_professional_accounting": 0.3870967741935484,
+      "mmlu_eval_accuracy_professional_law": 0.3411764705882353,
+      "mmlu_eval_accuracy_professional_medicine": 0.41935483870967744,
+      "mmlu_eval_accuracy_professional_psychology": 0.37681159420289856,
+      "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
+      "mmlu_eval_accuracy_security_studies": 0.5925925925925926,
+      "mmlu_eval_accuracy_sociology": 0.7272727272727273,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 0.9346268427978919,
+      "step": 1800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002,
+      "loss": 0.4742,
+      "step": 1810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002,
+      "loss": 0.5088,
+      "step": 1820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002,
+      "loss": 0.459,
+      "step": 1830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002,
+      "loss": 0.5195,
+      "step": 1840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002,
+      "loss": 0.617,
+      "step": 1850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002,
+      "loss": 0.4523,
+      "step": 1860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002,
+      "loss": 0.646,
+      "step": 1870
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002,
+      "loss": 0.4423,
+      "step": 1880
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002,
+      "loss": 0.5904,
+      "step": 1890
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002,
+      "loss": 0.5508,
+      "step": 1900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002,
+      "loss": 0.4987,
+      "step": 1910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002,
+      "loss": 0.5466,
+      "step": 1920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002,
+      "loss": 0.5342,
+      "step": 1930
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002,
+      "loss": 0.6875,
+      "step": 1940
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002,
+      "loss": 0.5443,
+      "step": 1950
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002,
+      "loss": 0.5367,
+      "step": 1960
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002,
+      "loss": 0.5529,
+      "step": 1970
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002,
+      "loss": 0.438,
+      "step": 1980
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002,
+      "loss": 0.4422,
+      "step": 1990
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002,
+      "loss": 0.5624,
+      "step": 2000
+    },
+    {
+      "epoch": 0.65,
+      "eval_loss": 0.5018913745880127,
+      "eval_runtime": 92.7994,
+      "eval_samples_per_second": 10.776,
+      "eval_steps_per_second": 5.388,
+      "step": 2000
+    },
+    {
+      "epoch": 0.65,
+      "mmlu_eval_accuracy": 0.4873839805836957,
+      "mmlu_eval_accuracy_abstract_algebra": 0.18181818181818182,
+      "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+      "mmlu_eval_accuracy_astronomy": 0.5625,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.41379310344827586,
+      "mmlu_eval_accuracy_college_biology": 0.375,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
+      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_econometrics": 0.08333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.4375,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.45454545454545453,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7666666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5384615384615384,
+      "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.8181818181818182,
+      "mmlu_eval_accuracy_marketing": 0.72,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
+      "mmlu_eval_accuracy_moral_disputes": 0.5789473684210527,
+      "mmlu_eval_accuracy_moral_scenarios": 0.27,
+      "mmlu_eval_accuracy_nutrition": 0.5454545454545454,
+      "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
+      "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+      "mmlu_eval_accuracy_professional_law": 0.3588235294117647,
+      "mmlu_eval_accuracy_professional_medicine": 0.4838709677419355,
+      "mmlu_eval_accuracy_professional_psychology": 0.4492753623188406,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
+      "mmlu_eval_accuracy_sociology": 0.5909090909090909,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 0.8898517729972268,
+      "step": 2000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002,
+      "loss": 0.4835,
+      "step": 2010
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002,
+      "loss": 0.5138,
+      "step": 2020
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002,
+      "loss": 0.4579,
+      "step": 2030
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002,
+      "loss": 0.6192,
+      "step": 2040
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002,
+      "loss": 0.5606,
+      "step": 2050
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002,
+      "loss": 0.5427,
+      "step": 2060
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002,
+      "loss": 0.5302,
+      "step": 2070
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002,
+      "loss": 0.5107,
+      "step": 2080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002,
+      "loss": 0.5849,
+      "step": 2090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002,
+      "loss": 0.4736,
+      "step": 2100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002,
+      "loss": 0.5274,
+      "step": 2110
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002,
+      "loss": 0.5004,
+      "step": 2120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002,
+      "loss": 0.4559,
+      "step": 2130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002,
+      "loss": 0.4954,
+      "step": 2140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002,
+      "loss": 0.5255,
+      "step": 2150
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002,
+      "loss": 0.3956,
+      "step": 2160
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002,
+      "loss": 0.446,
+      "step": 2170
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002,
+      "loss": 0.4923,
+      "step": 2180
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002,
+      "loss": 0.5505,
+      "step": 2190
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002,
+      "loss": 0.6295,
+      "step": 2200
+    },
+    {
+      "epoch": 0.71,
+      "eval_loss": 0.49923792481422424,
+      "eval_runtime": 92.7241,
+      "eval_samples_per_second": 10.785,
+      "eval_steps_per_second": 5.392,
+      "step": 2200
+    },
+    {
+      "epoch": 0.71,
+      "mmlu_eval_accuracy": 0.46882014178817705,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.375,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
+      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_econometrics": 0.08333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.25,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.40625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
+      "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.37209302325581395,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+      "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
+      "mmlu_eval_accuracy_high_school_psychology": 0.75,
+      "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5,
+      "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+      "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.68,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.627906976744186,
+      "mmlu_eval_accuracy_moral_disputes": 0.5,
+      "mmlu_eval_accuracy_moral_scenarios": 0.23,
+      "mmlu_eval_accuracy_nutrition": 0.5454545454545454,
+      "mmlu_eval_accuracy_philosophy": 0.38235294117647056,
+      "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
+      "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
+      "mmlu_eval_accuracy_professional_law": 0.36470588235294116,
+      "mmlu_eval_accuracy_professional_medicine": 0.41935483870967744,
+      "mmlu_eval_accuracy_professional_psychology": 0.42028985507246375,
+      "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
+      "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
+      "mmlu_eval_accuracy_sociology": 0.6363636363636364,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 0.8042322283090328,
+      "step": 2200
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002,
+      "loss": 0.5447,
+      "step": 2210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002,
+      "loss": 0.5842,
+      "step": 2220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002,
+      "loss": 0.4748,
+      "step": 2230
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002,
+      "loss": 0.5358,
+      "step": 2240
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002,
+      "loss": 0.4141,
+      "step": 2250
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002,
+      "loss": 0.4624,
+      "step": 2260
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002,
+      "loss": 0.5596,
+      "step": 2270
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002,
+      "loss": 0.5622,
+      "step": 2280
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002,
+      "loss": 0.5275,
+      "step": 2290
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002,
+      "loss": 0.4613,
+      "step": 2300
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002,
+      "loss": 0.5599,
+      "step": 2310
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002,
+      "loss": 0.5232,
+      "step": 2320
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002,
+      "loss": 0.485,
+      "step": 2330
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0002,
+      "loss": 0.5091,
+      "step": 2340
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0002,
+      "loss": 0.5513,
+      "step": 2350
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0002,
+      "loss": 0.5351,
+      "step": 2360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0002,
+      "loss": 0.5042,
+      "step": 2370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0002,
+      "loss": 0.5317,
+      "step": 2380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.0002,
+      "loss": 0.5015,
+      "step": 2390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0002,
+      "loss": 0.4993,
+      "step": 2400
+    },
+    {
+      "epoch": 0.78,
+      "eval_loss": 0.4968625605106354,
+      "eval_runtime": 92.831,
+      "eval_samples_per_second": 10.772,
+      "eval_steps_per_second": 5.386,
+      "step": 2400
+    },
+    {
+      "epoch": 0.78,
+      "mmlu_eval_accuracy": 0.4950372743506906,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+      "mmlu_eval_accuracy_astronomy": 0.5625,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4482758620689655,
+      "mmlu_eval_accuracy_college_biology": 0.5625,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
+      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_econometrics": 0.08333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
+      "mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.5,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.45454545454545453,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7833333333333333,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.46153846153846156,
+      "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
+      "mmlu_eval_accuracy_human_sexuality": 0.5,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.8181818181818182,
+      "mmlu_eval_accuracy_marketing": 0.76,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.627906976744186,
+      "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
+      "mmlu_eval_accuracy_moral_scenarios": 0.25,
+      "mmlu_eval_accuracy_nutrition": 0.5454545454545454,
+      "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
+      "mmlu_eval_accuracy_prehistory": 0.5428571428571428,
+      "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
+      "mmlu_eval_accuracy_professional_law": 0.3,
+      "mmlu_eval_accuracy_professional_medicine": 0.41935483870967744,
+      "mmlu_eval_accuracy_professional_psychology": 0.391304347826087,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.6111111111111112,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 0.9267964519307446,
+      "step": 2400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0002,
+      "loss": 0.4947,
+      "step": 2410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0002,
+      "loss": 0.4746,
+      "step": 2420
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0002,
+      "loss": 0.5745,
+      "step": 2430
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0002,
+      "loss": 0.5345,
+      "step": 2440
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0002,
+      "loss": 0.5506,
+      "step": 2450
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0002,
+      "loss": 0.5117,
+      "step": 2460
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0002,
+      "loss": 0.5503,
+      "step": 2470
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0002,
+      "loss": 0.6092,
+      "step": 2480
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0002,
+      "loss": 0.4803,
+      "step": 2490
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0002,
+      "loss": 0.5619,
+      "step": 2500
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0002,
+      "loss": 0.5649,
+      "step": 2510
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0002,
+      "loss": 0.4675,
+      "step": 2520
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0002,
+      "loss": 0.3825,
+      "step": 2530
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0002,
+      "loss": 0.3774,
+      "step": 2540
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.0002,
+      "loss": 0.6572,
+      "step": 2550
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0002,
+      "loss": 0.5841,
+      "step": 2560
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0002,
+      "loss": 0.4091,
+      "step": 2570
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0002,
+      "loss": 0.4902,
+      "step": 2580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0002,
+      "loss": 0.3788,
+      "step": 2590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0002,
+      "loss": 0.5437,
+      "step": 2600
+    },
+    {
+      "epoch": 0.84,
+      "eval_loss": 0.49346765875816345,
+      "eval_runtime": 92.8693,
+      "eval_samples_per_second": 10.768,
+      "eval_steps_per_second": 5.384,
+      "step": 2600
+    },
+    {
+      "epoch": 0.84,
+      "mmlu_eval_accuracy": 0.48036715840748917,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+      "mmlu_eval_accuracy_astronomy": 0.5625,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.3793103448275862,
+      "mmlu_eval_accuracy_college_biology": 0.3125,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
+      "mmlu_eval_accuracy_econometrics": 0.08333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
+      "mmlu_eval_accuracy_formal_logic": 0.42857142857142855,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.5,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.4444444444444444,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.37209302325581395,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7333333333333333,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
+      "mmlu_eval_accuracy_high_school_world_history": 0.46153846153846156,
+      "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+      "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+      "mmlu_eval_accuracy_international_law": 0.7692307692307693,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.8,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.6395348837209303,
+      "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
+      "mmlu_eval_accuracy_moral_scenarios": 0.27,
+      "mmlu_eval_accuracy_nutrition": 0.5454545454545454,
+      "mmlu_eval_accuracy_philosophy": 0.5,
+      "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
+      "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
+      "mmlu_eval_accuracy_professional_law": 0.3235294117647059,
+      "mmlu_eval_accuracy_professional_medicine": 0.45161290322580644,
+      "mmlu_eval_accuracy_professional_psychology": 0.4057971014492754,
+      "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
+      "mmlu_eval_accuracy_security_studies": 0.4444444444444444,
+      "mmlu_eval_accuracy_sociology": 0.5454545454545454,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.6111111111111112,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 0.9441872774270745,
+      "step": 2600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0002,
+      "loss": 0.3932,
+      "step": 2610
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0002,
+      "loss": 0.4533,
+      "step": 2620
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0002,
+      "loss": 0.55,
+      "step": 2630
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0002,
+      "loss": 0.5148,
+      "step": 2640
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0002,
+      "loss": 0.4371,
+      "step": 2650
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0002,
+      "loss": 0.4694,
+      "step": 2660
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.0002,
+      "loss": 0.6459,
+      "step": 2670
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0002,
+      "loss": 0.4738,
+      "step": 2680
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0002,
+      "loss": 0.532,
+      "step": 2690
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0002,
+      "loss": 0.516,
+      "step": 2700
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0002,
+      "loss": 0.505,
+      "step": 2710
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0002,
+      "loss": 0.5387,
+      "step": 2720
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0002,
+      "loss": 0.5307,
+      "step": 2730
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0002,
+      "loss": 0.4942,
+      "step": 2740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0002,
+      "loss": 0.5934,
+      "step": 2750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.0002,
+      "loss": 0.4751,
+      "step": 2760
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0002,
+      "loss": 0.6256,
+      "step": 2770
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0002,
+      "loss": 0.4482,
+      "step": 2780
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0002,
+      "loss": 0.525,
+      "step": 2790
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0002,
+      "loss": 0.4123,
+      "step": 2800
+    },
+    {
+      "epoch": 0.9,
+      "eval_loss": 0.4873571991920471,
+      "eval_runtime": 92.9027,
+      "eval_samples_per_second": 10.764,
+      "eval_steps_per_second": 5.382,
+      "step": 2800
+    },
+    {
+      "epoch": 0.9,
+      "mmlu_eval_accuracy": 0.4788798955105805,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+      "mmlu_eval_accuracy_astronomy": 0.5625,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.41379310344827586,
+      "mmlu_eval_accuracy_college_biology": 0.5,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
+      "mmlu_eval_accuracy_econometrics": 0.25,
+      "mmlu_eval_accuracy_electrical_engineering": 0.375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
+      "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.46875,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.4444444444444444,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.37209302325581395,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+      "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7833333333333333,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5,
+      "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+      "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.72,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
+      "mmlu_eval_accuracy_moral_disputes": 0.5,
+      "mmlu_eval_accuracy_moral_scenarios": 0.26,
+      "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
+      "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
+      "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
+      "mmlu_eval_accuracy_professional_accounting": 0.3870967741935484,
+      "mmlu_eval_accuracy_professional_law": 0.3411764705882353,
+      "mmlu_eval_accuracy_professional_medicine": 0.45161290322580644,
+      "mmlu_eval_accuracy_professional_psychology": 0.391304347826087,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.4444444444444444,
+      "mmlu_eval_accuracy_sociology": 0.6363636363636364,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 0.9054499416181686,
+      "step": 2800
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0002,
+      "loss": 0.4438,
+      "step": 2810
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0002,
+      "loss": 0.4154,
+      "step": 2820
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0002,
+      "loss": 0.4255,
+      "step": 2830
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0002,
+      "loss": 0.4833,
+      "step": 2840
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0002,
+      "loss": 0.5216,
+      "step": 2850
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0002,
+      "loss": 0.5155,
+      "step": 2860
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0002,
+      "loss": 0.3956,
+      "step": 2870
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0002,
+      "loss": 0.4986,
+      "step": 2880
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0002,
+      "loss": 0.4809,
+      "step": 2890
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0002,
+      "loss": 0.4222,
+      "step": 2900
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0002,
+      "loss": 0.4685,
+      "step": 2910
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0002,
+      "loss": 0.4816,
+      "step": 2920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0002,
+      "loss": 0.559,
+      "step": 2930
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0002,
+      "loss": 0.4976,
+      "step": 2940
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0002,
+      "loss": 0.4644,
+      "step": 2950
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0002,
+      "loss": 0.4328,
+      "step": 2960
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0002,
+      "loss": 0.3755,
+      "step": 2970
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.0002,
+      "loss": 0.4862,
+      "step": 2980
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0002,
+      "loss": 0.3602,
+      "step": 2990
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0002,
+      "loss": 0.4969,
+      "step": 3000
+    },
+    {
+      "epoch": 0.97,
+      "eval_loss": 0.4759189486503601,
+      "eval_runtime": 92.7764,
+      "eval_samples_per_second": 10.779,
+      "eval_steps_per_second": 5.389,
+      "step": 3000
+    },
+    {
+      "epoch": 0.97,
+      "mmlu_eval_accuracy": 0.47644674735610876,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+      "mmlu_eval_accuracy_astronomy": 0.625,
+      "mmlu_eval_accuracy_business_ethics": 0.36363636363636365,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.41379310344827586,
+      "mmlu_eval_accuracy_college_biology": 0.375,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_physics": 0.5454545454545454,
+      "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
+      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_econometrics": 0.25,
+      "mmlu_eval_accuracy_electrical_engineering": 0.375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
+      "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.375,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.3333333333333333,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5384615384615384,
+      "mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8166666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.4230769230769231,
+      "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.7692307692307693,
+      "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.8181818181818182,
+      "mmlu_eval_accuracy_marketing": 0.72,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
+      "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
+      "mmlu_eval_accuracy_moral_scenarios": 0.27,
+      "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
+      "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
+      "mmlu_eval_accuracy_prehistory": 0.6,
+      "mmlu_eval_accuracy_professional_accounting": 0.25806451612903225,
+      "mmlu_eval_accuracy_professional_law": 0.3235294117647059,
+      "mmlu_eval_accuracy_professional_medicine": 0.4838709677419355,
+      "mmlu_eval_accuracy_professional_psychology": 0.2898550724637681,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 0.8384745381577986,
+      "step": 3000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0002,
+      "loss": 0.3811,
+      "step": 3010
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0002,
+      "loss": 0.4609,
+      "step": 3020
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0002,
+      "loss": 0.4465,
+      "step": 3030
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0002,
+      "loss": 0.4319,
+      "step": 3040
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0002,
+      "loss": 0.3987,
+      "step": 3050
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0002,
+      "loss": 0.4333,
+      "step": 3060
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0002,
+      "loss": 0.4409,
+      "step": 3070
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0002,
+      "loss": 0.5345,
+      "step": 3080
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0002,
+      "loss": 0.4889,
+      "step": 3090
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0002,
+      "loss": 0.5426,
+      "step": 3100
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0002,
+      "loss": 0.3533,
+      "step": 3110
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0002,
+      "loss": 0.2978,
+      "step": 3120
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0002,
+      "loss": 0.3878,
+      "step": 3130
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0002,
+      "loss": 0.2189,
+      "step": 3140
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0002,
+      "loss": 0.4635,
+      "step": 3150
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0002,
+      "loss": 0.3747,
+      "step": 3160
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0002,
+      "loss": 0.3886,
+      "step": 3170
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0002,
+      "loss": 0.405,
+      "step": 3180
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0002,
+      "loss": 0.3307,
+      "step": 3190
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0002,
+      "loss": 0.3314,
+      "step": 3200
+    },
+    {
+      "epoch": 1.03,
+      "eval_loss": 0.4710354208946228,
+      "eval_runtime": 92.8548,
+      "eval_samples_per_second": 10.769,
+      "eval_steps_per_second": 5.385,
+      "step": 3200
+    },
+    {
+      "epoch": 1.03,
+      "mmlu_eval_accuracy": 0.48093919104536653,
+      "mmlu_eval_accuracy_abstract_algebra": 0.18181818181818182,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.5625,
+      "mmlu_eval_accuracy_business_ethics": 0.36363636363636365,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.41379310344827586,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_mathematics": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.6363636363636364,
+      "mmlu_eval_accuracy_conceptual_physics": 0.5,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.2926829268292683,
+      "mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.5625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.4444444444444444,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
+      "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
+      "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7666666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5,
+      "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+      "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+      "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.72,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.6395348837209303,
+      "mmlu_eval_accuracy_moral_disputes": 0.5526315789473685,
+      "mmlu_eval_accuracy_moral_scenarios": 0.22,
+      "mmlu_eval_accuracy_nutrition": 0.5454545454545454,
+      "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
+      "mmlu_eval_accuracy_prehistory": 0.5142857142857142,
+      "mmlu_eval_accuracy_professional_accounting": 0.25806451612903225,
+      "mmlu_eval_accuracy_professional_law": 0.3176470588235294,
+      "mmlu_eval_accuracy_professional_medicine": 0.41935483870967744,
+      "mmlu_eval_accuracy_professional_psychology": 0.42028985507246375,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 0.9746437882104835,
+      "step": 3200
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0002,
+      "loss": 0.3397,
+      "step": 3210
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0002,
+      "loss": 0.3717,
+      "step": 3220
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0002,
+      "loss": 0.3644,
+      "step": 3230
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0002,
+      "loss": 0.408,
+      "step": 3240
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0002,
+      "loss": 0.3467,
+      "step": 3250
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0002,
+      "loss": 0.3149,
+      "step": 3260
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0002,
+      "loss": 0.3619,
+      "step": 3270
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0002,
+      "loss": 0.387,
+      "step": 3280
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0002,
+      "loss": 0.365,
+      "step": 3290
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0002,
+      "loss": 0.4566,
+      "step": 3300
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0002,
+      "loss": 0.2898,
+      "step": 3310
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0002,
+      "loss": 0.3508,
+      "step": 3320
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0002,
+      "loss": 0.3188,
+      "step": 3330
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0002,
+      "loss": 0.3325,
+      "step": 3340
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0002,
+      "loss": 0.3382,
+      "step": 3350
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0002,
+      "loss": 0.2875,
+      "step": 3360
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0002,
+      "loss": 0.3696,
+      "step": 3370
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0002,
+      "loss": 0.3008,
+      "step": 3380
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0002,
+      "loss": 0.4654,
+      "step": 3390
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0002,
+      "loss": 0.4212,
+      "step": 3400
+    },
+    {
+      "epoch": 1.1,
+      "eval_loss": 0.4838115870952606,
+      "eval_runtime": 92.8008,
+      "eval_samples_per_second": 10.776,
+      "eval_steps_per_second": 5.388,
+      "step": 3400
+    },
+    {
+      "epoch": 1.1,
+      "mmlu_eval_accuracy": 0.4748766842351904,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.5,
+      "mmlu_eval_accuracy_astronomy": 0.4375,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.3448275862068966,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_physics": 0.5454545454545454,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
+      "mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.46875,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
+      "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.32558139534883723,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.38461538461538464,
+      "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7333333333333333,
+      "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5384615384615384,
+      "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+      "mmlu_eval_accuracy_human_sexuality": 0.5833333333333334,
+      "mmlu_eval_accuracy_international_law": 0.7692307692307693,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5,
+      "mmlu_eval_accuracy_machine_learning": 0.09090909090909091,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.76,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.627906976744186,
+      "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
+      "mmlu_eval_accuracy_moral_scenarios": 0.29,
+      "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
+      "mmlu_eval_accuracy_philosophy": 0.5,
+      "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+      "mmlu_eval_accuracy_professional_law": 0.3058823529411765,
+      "mmlu_eval_accuracy_professional_medicine": 0.3870967741935484,
+      "mmlu_eval_accuracy_professional_psychology": 0.37681159420289856,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.4444444444444444,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 0.852231516500366,
+      "step": 3400
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0002,
+      "loss": 0.3756,
+      "step": 3410
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0002,
+      "loss": 0.448,
+      "step": 3420
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0002,
+      "loss": 0.3674,
+      "step": 3430
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0002,
+      "loss": 0.3219,
+      "step": 3440
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0002,
+      "loss": 0.3858,
+      "step": 3450
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0002,
+      "loss": 0.3591,
+      "step": 3460
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0002,
+      "loss": 0.4104,
+      "step": 3470
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0002,
+      "loss": 0.3449,
+      "step": 3480
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0002,
+      "loss": 0.3806,
+      "step": 3490
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0002,
+      "loss": 0.3801,
+      "step": 3500
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0002,
+      "loss": 0.3832,
+      "step": 3510
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0002,
+      "loss": 0.2705,
+      "step": 3520
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0002,
+      "loss": 0.2491,
+      "step": 3530
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0002,
+      "loss": 0.306,
+      "step": 3540
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0002,
+      "loss": 0.3281,
+      "step": 3550
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0002,
+      "loss": 0.3485,
+      "step": 3560
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0002,
+      "loss": 0.4544,
+      "step": 3570
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0002,
+      "loss": 0.327,
+      "step": 3580
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0002,
+      "loss": 0.3357,
+      "step": 3590
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0002,
+      "loss": 0.498,
+      "step": 3600
+    },
+    {
+      "epoch": 1.16,
+      "eval_loss": 0.4635956287384033,
+      "eval_runtime": 92.9064,
+      "eval_samples_per_second": 10.764,
+      "eval_steps_per_second": 5.382,
+      "step": 3600
+    },
+    {
+      "epoch": 1.16,
+      "mmlu_eval_accuracy": 0.4806370203028222,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.4375,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4482758620689655,
+      "mmlu_eval_accuracy_college_biology": 0.375,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_medicine": 0.3181818181818182,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.6363636363636364,
+      "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
+      "mmlu_eval_accuracy_econometrics": 0.08333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
+      "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.5,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.4444444444444444,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
+      "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.37209302325581395,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7666666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.46153846153846156,
+      "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+      "mmlu_eval_accuracy_human_sexuality": 0.5,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+      "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
+      "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
+      "mmlu_eval_accuracy_moral_scenarios": 0.24,
+      "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
+      "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
+      "mmlu_eval_accuracy_prehistory": 0.5428571428571428,
+      "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
+      "mmlu_eval_accuracy_professional_law": 0.32941176470588235,
+      "mmlu_eval_accuracy_professional_medicine": 0.3870967741935484,
+      "mmlu_eval_accuracy_professional_psychology": 0.391304347826087,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
+      "mmlu_eval_accuracy_sociology": 0.6363636363636364,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.4444444444444444,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 0.9088846339549937,
+      "step": 3600
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0002,
+      "loss": 0.3504,
+      "step": 3610
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0002,
+      "loss": 0.3137,
+      "step": 3620
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0002,
+      "loss": 0.3356,
+      "step": 3630
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0002,
+      "loss": 0.389,
+      "step": 3640
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0002,
+      "loss": 0.3724,
+      "step": 3650
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0002,
+      "loss": 0.3269,
+      "step": 3660
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0002,
+      "loss": 0.2583,
+      "step": 3670
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0002,
+      "loss": 0.4136,
+      "step": 3680
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0002,
+      "loss": 0.3767,
+      "step": 3690
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0002,
+      "loss": 0.4162,
+      "step": 3700
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0002,
+      "loss": 0.3691,
+      "step": 3710
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0002,
+      "loss": 0.3066,
+      "step": 3720
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0002,
+      "loss": 0.4164,
+      "step": 3730
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0002,
+      "loss": 0.3298,
+      "step": 3740
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0002,
+      "loss": 0.4793,
+      "step": 3750
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002,
+      "loss": 0.3191,
+      "step": 3760
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002,
+      "loss": 0.3991,
+      "step": 3770
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002,
+      "loss": 0.3472,
+      "step": 3780
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002,
+      "loss": 0.398,
+      "step": 3790
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002,
+      "loss": 0.3852,
+      "step": 3800
+    },
+    {
+      "epoch": 1.23,
+      "eval_loss": 0.45864441990852356,
+      "eval_runtime": 92.8133,
+      "eval_samples_per_second": 10.774,
+      "eval_steps_per_second": 5.387,
+      "step": 3800
+    },
+    {
+      "epoch": 1.23,
+      "mmlu_eval_accuracy": 0.4741328353924988,
+      "mmlu_eval_accuracy_abstract_algebra": 0.18181818181818182,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.5625,
+      "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_physics": 0.5454545454545454,
+      "mmlu_eval_accuracy_computer_security": 0.6363636363636364,
+      "mmlu_eval_accuracy_conceptual_physics": 0.3076923076923077,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.2682926829268293,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.4,
+      "mmlu_eval_accuracy_high_school_biology": 0.5,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.22727272727272727,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.4444444444444444,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5384615384615384,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8,
+      "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5,
+      "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
+      "mmlu_eval_accuracy_human_sexuality": 0.5,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.76,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
+      "mmlu_eval_accuracy_moral_disputes": 0.5,
+      "mmlu_eval_accuracy_moral_scenarios": 0.28,
+      "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
+      "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
+      "mmlu_eval_accuracy_prehistory": 0.5714285714285714,
+      "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
+      "mmlu_eval_accuracy_professional_law": 0.3176470588235294,
+      "mmlu_eval_accuracy_professional_medicine": 0.41935483870967744,
+      "mmlu_eval_accuracy_professional_psychology": 0.4057971014492754,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.4444444444444444,
+      "mmlu_eval_accuracy_sociology": 0.7727272727272727,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 0.8557860668330373,
+      "step": 3800
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002,
+      "loss": 0.3525,
+      "step": 3810
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002,
+      "loss": 0.3821,
+      "step": 3820
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0002,
+      "loss": 0.3034,
+      "step": 3830
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0002,
+      "loss": 0.4402,
+      "step": 3840
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0002,
+      "loss": 0.4403,
+      "step": 3850
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002,
+      "loss": 0.3717,
+      "step": 3860
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002,
+      "loss": 0.2817,
+      "step": 3870
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002,
+      "loss": 0.3473,
+      "step": 3880
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0002,
+      "loss": 0.3621,
+      "step": 3890
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0002,
+      "loss": 0.3355,
+      "step": 3900
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0002,
+      "loss": 0.3012,
+      "step": 3910
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0002,
+      "loss": 0.3306,
+      "step": 3920
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0002,
+      "loss": 0.419,
+      "step": 3930
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0002,
+      "loss": 0.3715,
+      "step": 3940
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0002,
+      "loss": 0.403,
+      "step": 3950
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0002,
+      "loss": 0.4401,
+      "step": 3960
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0002,
+      "loss": 0.3677,
+      "step": 3970
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002,
+      "loss": 0.3942,
+      "step": 3980
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002,
+      "loss": 0.3138,
+      "step": 3990
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002,
+      "loss": 0.4405,
+      "step": 4000
+    },
+    {
+      "epoch": 1.29,
+      "eval_loss": 0.4552951157093048,
+      "eval_runtime": 92.8831,
+      "eval_samples_per_second": 10.766,
+      "eval_steps_per_second": 5.383,
+      "step": 4000
+    },
+    {
+      "epoch": 1.29,
+      "mmlu_eval_accuracy": 0.49984985070715504,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.375,
+      "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
+      "mmlu_eval_accuracy_college_biology": 0.5625,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.3181818181818182,
+      "mmlu_eval_accuracy_college_physics": 0.6363636363636364,
+      "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
+      "mmlu_eval_accuracy_conceptual_physics": 0.5,
+      "mmlu_eval_accuracy_econometrics": 0.4166666666666667,
+      "mmlu_eval_accuracy_electrical_engineering": 0.25,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.24390243902439024,
+      "mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
+      "mmlu_eval_accuracy_global_facts": 0.4,
+      "mmlu_eval_accuracy_high_school_biology": 0.5,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.4444444444444444,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
+      "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5769230769230769,
+      "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+      "mmlu_eval_accuracy_human_sexuality": 0.5,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.8181818181818182,
+      "mmlu_eval_accuracy_marketing": 0.8,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.686046511627907,
+      "mmlu_eval_accuracy_moral_disputes": 0.5,
+      "mmlu_eval_accuracy_moral_scenarios": 0.24,
+      "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
+      "mmlu_eval_accuracy_philosophy": 0.4117647058823529,
+      "mmlu_eval_accuracy_prehistory": 0.5714285714285714,
+      "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
+      "mmlu_eval_accuracy_professional_law": 0.35294117647058826,
+      "mmlu_eval_accuracy_professional_medicine": 0.45161290322580644,
+      "mmlu_eval_accuracy_professional_psychology": 0.37681159420289856,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
+      "mmlu_eval_accuracy_sociology": 0.7727272727272727,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 0.8634209213413628,
+      "step": 4000
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0002,
+      "loss": 0.3778,
+      "step": 4010
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0002,
+      "loss": 0.3565,
+      "step": 4020
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0002,
+      "loss": 0.2519,
+      "step": 4030
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0002,
+      "loss": 0.379,
+      "step": 4040
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0002,
+      "loss": 0.33,
+      "step": 4050
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0002,
+      "loss": 0.3186,
+      "step": 4060
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002,
+      "loss": 0.3546,
+      "step": 4070
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002,
+      "loss": 0.3288,
+      "step": 4080
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002,
+      "loss": 0.2546,
+      "step": 4090
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0002,
+      "loss": 0.3622,
+      "step": 4100
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0002,
+      "loss": 0.3814,
+      "step": 4110
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0002,
+      "loss": 0.3508,
+      "step": 4120
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0002,
+      "loss": 0.311,
+      "step": 4130
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0002,
+      "loss": 0.3012,
+      "step": 4140
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0002,
+      "loss": 0.4021,
+      "step": 4150
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0002,
+      "loss": 0.3811,
+      "step": 4160
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0002,
+      "loss": 0.4127,
+      "step": 4170
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0002,
+      "loss": 0.3644,
+      "step": 4180
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0002,
+      "loss": 0.3852,
+      "step": 4190
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0002,
+      "loss": 0.2721,
+      "step": 4200
+    },
+    {
+      "epoch": 1.36,
+      "eval_loss": 0.44986042380332947,
+      "eval_runtime": 92.8765,
+      "eval_samples_per_second": 10.767,
+      "eval_steps_per_second": 5.383,
+      "step": 4200
+    },
+    {
+      "epoch": 1.36,
+      "mmlu_eval_accuracy": 0.4833839114547751,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+      "mmlu_eval_accuracy_astronomy": 0.4375,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.41379310344827586,
+      "mmlu_eval_accuracy_college_biology": 0.375,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_physics": 0.5454545454545454,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.5,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.38461538461538464,
+      "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7166666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5769230769230769,
+      "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+      "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.8181818181818182,
+      "mmlu_eval_accuracy_marketing": 0.76,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.686046511627907,
+      "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
+      "mmlu_eval_accuracy_moral_scenarios": 0.23,
+      "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
+      "mmlu_eval_accuracy_philosophy": 0.5,
+      "mmlu_eval_accuracy_prehistory": 0.5142857142857142,
+      "mmlu_eval_accuracy_professional_accounting": 0.41935483870967744,
+      "mmlu_eval_accuracy_professional_law": 0.3411764705882353,
+      "mmlu_eval_accuracy_professional_medicine": 0.41935483870967744,
+      "mmlu_eval_accuracy_professional_psychology": 0.43478260869565216,
+      "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
+      "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.8181818181818182,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
+      "mmlu_loss": 0.9393174315148639,
+      "step": 4200
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0002,
+      "loss": 0.3953,
+      "step": 4210
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0002,
+      "loss": 0.2784,
+      "step": 4220
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002,
+      "loss": 0.3217,
+      "step": 4230
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002,
+      "loss": 0.3874,
+      "step": 4240
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002,
+      "loss": 0.3192,
+      "step": 4250
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002,
+      "loss": 0.3759,
+      "step": 4260
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002,
+      "loss": 0.3781,
+      "step": 4270
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002,
+      "loss": 0.256,
+      "step": 4280
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0002,
+      "loss": 0.3781,
+      "step": 4290
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0002,
+      "loss": 0.4478,
+      "step": 4300
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0002,
+      "loss": 0.4456,
+      "step": 4310
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0002,
+      "loss": 0.405,
+      "step": 4320
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0002,
+      "loss": 0.4139,
+      "step": 4330
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0002,
+      "loss": 0.2726,
+      "step": 4340
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0002,
+      "loss": 0.3414,
+      "step": 4350
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0002,
+      "loss": 0.4043,
+      "step": 4360
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0002,
+      "loss": 0.3363,
+      "step": 4370
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002,
+      "loss": 0.4315,
+      "step": 4380
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002,
+      "loss": 0.3872,
+      "step": 4390
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002,
+      "loss": 0.358,
+      "step": 4400
+    },
+    {
+      "epoch": 1.42,
+      "eval_loss": 0.4541943669319153,
+      "eval_runtime": 92.816,
+      "eval_samples_per_second": 10.774,
+      "eval_steps_per_second": 5.387,
+      "step": 4400
+    },
+    {
+      "epoch": 1.42,
+      "mmlu_eval_accuracy": 0.46572975756143253,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.5,
+      "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
+      "mmlu_eval_accuracy_college_biology": 0.5,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.1875,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+      "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
+      "mmlu_eval_accuracy_global_facts": 0.4,
+      "mmlu_eval_accuracy_high_school_biology": 0.5,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.22727272727272727,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.4444444444444444,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.37209302325581395,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
+      "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7666666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
+      "mmlu_eval_accuracy_high_school_us_history": 0.5909090909090909,
+      "mmlu_eval_accuracy_high_school_world_history": 0.46153846153846156,
+      "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
+      "mmlu_eval_accuracy_human_sexuality": 0.5,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.76,
+      "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+      "mmlu_eval_accuracy_miscellaneous": 0.686046511627907,
+      "mmlu_eval_accuracy_moral_disputes": 0.5,
+      "mmlu_eval_accuracy_moral_scenarios": 0.24,
+      "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
+      "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
+      "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+      "mmlu_eval_accuracy_professional_law": 0.3235294117647059,
+      "mmlu_eval_accuracy_professional_medicine": 0.45161290322580644,
+      "mmlu_eval_accuracy_professional_psychology": 0.4057971014492754,
+      "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
+      "mmlu_eval_accuracy_security_studies": 0.4074074074074074,
+      "mmlu_eval_accuracy_sociology": 0.7727272727272727,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
+      "mmlu_loss": 1.1100451363051218,
+      "step": 4400
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0002,
+      "loss": 0.405,
+      "step": 4410
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0002,
+      "loss": 0.3002,
+      "step": 4420
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0002,
+      "loss": 0.3862,
+      "step": 4430
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002,
+      "loss": 0.4145,
+      "step": 4440
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002,
+      "loss": 0.3624,
+      "step": 4450
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002,
+      "loss": 0.4279,
+      "step": 4460
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002,
+      "loss": 0.315,
+      "step": 4470
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0002,
+      "loss": 0.4288,
+      "step": 4480
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0002,
+      "loss": 0.2975,
+      "step": 4490
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0002,
+      "loss": 0.3429,
+      "step": 4500
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002,
+      "loss": 0.3685,
+      "step": 4510
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002,
+      "loss": 0.3391,
+      "step": 4520
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002,
+      "loss": 0.4046,
+      "step": 4530
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0002,
+      "loss": 0.361,
+      "step": 4540
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0002,
+      "loss": 0.3229,
+      "step": 4550
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0002,
+      "loss": 0.38,
+      "step": 4560
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002,
+      "loss": 0.2996,
+      "step": 4570
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002,
+      "loss": 0.333,
+      "step": 4580
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002,
+      "loss": 0.4243,
+      "step": 4590
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002,
+      "loss": 0.4146,
+      "step": 4600
+    },
+    {
+      "epoch": 1.49,
+      "eval_loss": 0.454068660736084,
+      "eval_runtime": 92.7711,
+      "eval_samples_per_second": 10.779,
+      "eval_steps_per_second": 5.39,
+      "step": 4600
+    },
+    {
+      "epoch": 1.49,
+      "mmlu_eval_accuracy": 0.47302137804215916,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
+      "mmlu_eval_accuracy_astronomy": 0.5,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.3448275862068966,
+      "mmlu_eval_accuracy_college_biology": 0.375,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+      "mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
+      "mmlu_eval_accuracy_global_facts": 0.4,
+      "mmlu_eval_accuracy_high_school_biology": 0.4375,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.4444444444444444,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5,
+      "mmlu_eval_accuracy_high_school_geography": 0.7272727272727273,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.3488372093023256,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.38461538461538464,
+      "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7,
+      "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5384615384615384,
+      "mmlu_eval_accuracy_human_aging": 0.782608695652174,
+      "mmlu_eval_accuracy_human_sexuality": 0.5,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.8,
+      "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
+      "mmlu_eval_accuracy_miscellaneous": 0.686046511627907,
+      "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
+      "mmlu_eval_accuracy_moral_scenarios": 0.3,
+      "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
+      "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
+      "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+      "mmlu_eval_accuracy_professional_law": 0.32941176470588235,
+      "mmlu_eval_accuracy_professional_medicine": 0.41935483870967744,
+      "mmlu_eval_accuracy_professional_psychology": 0.43478260869565216,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.4444444444444444,
+      "mmlu_eval_accuracy_sociology": 0.7727272727272727,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
+      "mmlu_loss": 0.9443787644585648,
+      "step": 4600
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002,
+      "loss": 0.249,
+      "step": 4610
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002,
+      "loss": 0.3483,
+      "step": 4620
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0002,
+      "loss": 0.3234,
+      "step": 4630
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0002,
+      "loss": 0.3997,
+      "step": 4640
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0002,
+      "loss": 0.409,
+      "step": 4650
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0002,
+      "loss": 0.4366,
+      "step": 4660
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0002,
+      "loss": 0.3959,
+      "step": 4670
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0002,
+      "loss": 0.3452,
+      "step": 4680
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002,
+      "loss": 0.3565,
+      "step": 4690
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002,
+      "loss": 0.2879,
+      "step": 4700
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002,
+      "loss": 0.4298,
+      "step": 4710
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0002,
+      "loss": 0.3548,
+      "step": 4720
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0002,
+      "loss": 0.4156,
+      "step": 4730
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0002,
+      "loss": 0.3201,
+      "step": 4740
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002,
+      "loss": 0.325,
+      "step": 4750
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002,
+      "loss": 0.2887,
+      "step": 4760
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002,
+      "loss": 0.3626,
+      "step": 4770
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002,
+      "loss": 0.3627,
+      "step": 4780
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0002,
+      "loss": 0.3928,
+      "step": 4790
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0002,
+      "loss": 0.2977,
+      "step": 4800
+    },
+    {
+      "epoch": 1.55,
+      "eval_loss": 0.46080464124679565,
+      "eval_runtime": 93.0123,
+      "eval_samples_per_second": 10.751,
+      "eval_steps_per_second": 5.376,
+      "step": 4800
+    },
+    {
+      "epoch": 1.55,
+      "mmlu_eval_accuracy": 0.4760097213526257,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
+      "mmlu_eval_accuracy_astronomy": 0.5,
+      "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
+      "mmlu_eval_accuracy_college_biology": 0.5,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.5454545454545454,
+      "mmlu_eval_accuracy_computer_security": 0.2727272727272727,
+      "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.4,
+      "mmlu_eval_accuracy_high_school_biology": 0.40625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.4444444444444444,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5,
+      "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.32558139534883723,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.34615384615384615,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7666666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.4230769230769231,
+      "mmlu_eval_accuracy_human_aging": 0.8260869565217391,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
+      "mmlu_eval_accuracy_management": 0.5454545454545454,
+      "mmlu_eval_accuracy_marketing": 0.76,
+      "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
+      "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
+      "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
+      "mmlu_eval_accuracy_moral_scenarios": 0.28,
+      "mmlu_eval_accuracy_nutrition": 0.696969696969697,
+      "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
+      "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+      "mmlu_eval_accuracy_professional_law": 0.29411764705882354,
+      "mmlu_eval_accuracy_professional_medicine": 0.41935483870967744,
+      "mmlu_eval_accuracy_professional_psychology": 0.4492753623188406,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
+      "mmlu_eval_accuracy_sociology": 0.7727272727272727,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
+      "mmlu_loss": 0.9548846610843669,
+      "step": 4800
+    }
+  ],
+  "max_steps": 5000,
+  "num_train_epochs": 2,
+  "total_flos": 3.8905982439491174e+17,
+  "trial_name": null,
+  "trial_params": null
+}