prateeky2806's picture
Training in progress, step 2600
ded333f
{
"best_metric": 0.598466694355011,
"best_model_checkpoint": "./output_v2/7b_cluster020_Nous-Hermes-llama-2-7b_partitioned_v3_standardized_020/checkpoint-2200",
"epoch": 1.0524185387573366,
"global_step": 2600,
"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": 0.6996,
"step": 10
},
{
"epoch": 0.01,
"learning_rate": 0.0002,
"loss": 0.7986,
"step": 20
},
{
"epoch": 0.01,
"learning_rate": 0.0002,
"loss": 0.5936,
"step": 30
},
{
"epoch": 0.02,
"learning_rate": 0.0002,
"loss": 0.6164,
"step": 40
},
{
"epoch": 0.02,
"learning_rate": 0.0002,
"loss": 0.7464,
"step": 50
},
{
"epoch": 0.02,
"learning_rate": 0.0002,
"loss": 0.8856,
"step": 60
},
{
"epoch": 0.03,
"learning_rate": 0.0002,
"loss": 0.6476,
"step": 70
},
{
"epoch": 0.03,
"learning_rate": 0.0002,
"loss": 0.65,
"step": 80
},
{
"epoch": 0.04,
"learning_rate": 0.0002,
"loss": 0.5282,
"step": 90
},
{
"epoch": 0.04,
"learning_rate": 0.0002,
"loss": 0.5787,
"step": 100
},
{
"epoch": 0.04,
"learning_rate": 0.0002,
"loss": 0.6315,
"step": 110
},
{
"epoch": 0.05,
"learning_rate": 0.0002,
"loss": 0.5419,
"step": 120
},
{
"epoch": 0.05,
"learning_rate": 0.0002,
"loss": 0.593,
"step": 130
},
{
"epoch": 0.06,
"learning_rate": 0.0002,
"loss": 0.6773,
"step": 140
},
{
"epoch": 0.06,
"learning_rate": 0.0002,
"loss": 0.5536,
"step": 150
},
{
"epoch": 0.06,
"learning_rate": 0.0002,
"loss": 0.6384,
"step": 160
},
{
"epoch": 0.07,
"learning_rate": 0.0002,
"loss": 0.5736,
"step": 170
},
{
"epoch": 0.07,
"learning_rate": 0.0002,
"loss": 0.6157,
"step": 180
},
{
"epoch": 0.08,
"learning_rate": 0.0002,
"loss": 0.5551,
"step": 190
},
{
"epoch": 0.08,
"learning_rate": 0.0002,
"loss": 0.6446,
"step": 200
},
{
"epoch": 0.08,
"eval_loss": 0.6395586133003235,
"eval_runtime": 94.1614,
"eval_samples_per_second": 10.62,
"eval_steps_per_second": 5.31,
"step": 200
},
{
"epoch": 0.08,
"mmlu_eval_accuracy": 0.4559132721218583,
"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": 0.3793103448275862,
"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": 0.2727272727272727,
"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": 0.3902439024390244,
"mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
"mmlu_eval_accuracy_global_facts": 0.3,
"mmlu_eval_accuracy_high_school_biology": 0.3125,
"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": 0.7272727272727273,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.32558139534883723,
"mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
"mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
"mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
"mmlu_eval_accuracy_high_school_psychology": 0.7166666666666667,
"mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
"mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
"mmlu_eval_accuracy_high_school_world_history": 0.5,
"mmlu_eval_accuracy_human_aging": 0.6956521739130435,
"mmlu_eval_accuracy_human_sexuality": 0.5,
"mmlu_eval_accuracy_international_law": 0.6923076923076923,
"mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
"mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
"mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
"mmlu_eval_accuracy_management": 0.5454545454545454,
"mmlu_eval_accuracy_marketing": 0.72,
"mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
"mmlu_eval_accuracy_miscellaneous": 0.627906976744186,
"mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
"mmlu_eval_accuracy_moral_scenarios": 0.24,
"mmlu_eval_accuracy_nutrition": 0.5454545454545454,
"mmlu_eval_accuracy_philosophy": 0.47058823529411764,
"mmlu_eval_accuracy_prehistory": 0.4857142857142857,
"mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
"mmlu_eval_accuracy_professional_law": 0.31176470588235294,
"mmlu_eval_accuracy_professional_medicine": 0.3870967741935484,
"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": 0.6363636363636364,
"mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
"mmlu_eval_accuracy_virology": 0.5,
"mmlu_eval_accuracy_world_religions": 0.6842105263157895,
"mmlu_loss": 1.0596903230868493,
"step": 200
},
{
"epoch": 0.09,
"learning_rate": 0.0002,
"loss": 0.7307,
"step": 210
},
{
"epoch": 0.09,
"learning_rate": 0.0002,
"loss": 0.5717,
"step": 220
},
{
"epoch": 0.09,
"learning_rate": 0.0002,
"loss": 0.6836,
"step": 230
},
{
"epoch": 0.1,
"learning_rate": 0.0002,
"loss": 0.5819,
"step": 240
},
{
"epoch": 0.1,
"learning_rate": 0.0002,
"loss": 0.5666,
"step": 250
},
{
"epoch": 0.11,
"learning_rate": 0.0002,
"loss": 0.5266,
"step": 260
},
{
"epoch": 0.11,
"learning_rate": 0.0002,
"loss": 0.5218,
"step": 270
},
{
"epoch": 0.11,
"learning_rate": 0.0002,
"loss": 0.5487,
"step": 280
},
{
"epoch": 0.12,
"learning_rate": 0.0002,
"loss": 0.5345,
"step": 290
},
{
"epoch": 0.12,
"learning_rate": 0.0002,
"loss": 0.6299,
"step": 300
},
{
"epoch": 0.13,
"learning_rate": 0.0002,
"loss": 0.5681,
"step": 310
},
{
"epoch": 0.13,
"learning_rate": 0.0002,
"loss": 0.5553,
"step": 320
},
{
"epoch": 0.13,
"learning_rate": 0.0002,
"loss": 0.575,
"step": 330
},
{
"epoch": 0.14,
"learning_rate": 0.0002,
"loss": 0.5708,
"step": 340
},
{
"epoch": 0.14,
"learning_rate": 0.0002,
"loss": 0.4932,
"step": 350
},
{
"epoch": 0.15,
"learning_rate": 0.0002,
"loss": 0.6957,
"step": 360
},
{
"epoch": 0.15,
"learning_rate": 0.0002,
"loss": 0.6442,
"step": 370
},
{
"epoch": 0.15,
"learning_rate": 0.0002,
"loss": 0.5999,
"step": 380
},
{
"epoch": 0.16,
"learning_rate": 0.0002,
"loss": 0.5086,
"step": 390
},
{
"epoch": 0.16,
"learning_rate": 0.0002,
"loss": 0.7349,
"step": 400
},
{
"epoch": 0.16,
"eval_loss": 0.6260886192321777,
"eval_runtime": 94.1289,
"eval_samples_per_second": 10.624,
"eval_steps_per_second": 5.312,
"step": 400
},
{
"epoch": 0.16,
"mmlu_eval_accuracy": 0.4735216064792921,
"mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
"mmlu_eval_accuracy_anatomy": 0.5714285714285714,
"mmlu_eval_accuracy_astronomy": 0.5,
"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": 0.2727272727272727,
"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": 0.4230769230769231,
"mmlu_eval_accuracy_econometrics": 0.25,
"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": 0.375,
"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": 0.7272727272727273,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
"mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
"mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
"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": 0.6363636363636364,
"mmlu_eval_accuracy_high_school_world_history": 0.5,
"mmlu_eval_accuracy_human_aging": 0.6956521739130435,
"mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
"mmlu_eval_accuracy_international_law": 0.8461538461538461,
"mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
"mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
"mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
"mmlu_eval_accuracy_management": 0.45454545454545453,
"mmlu_eval_accuracy_marketing": 0.8,
"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": 0.6666666666666666,
"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": 0.45161290322580644,
"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": 0.7272727272727273,
"mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
"mmlu_eval_accuracy_virology": 0.4444444444444444,
"mmlu_eval_accuracy_world_religions": 0.7368421052631579,
"mmlu_loss": 0.9645495347814834,
"step": 400
},
{
"epoch": 0.17,
"learning_rate": 0.0002,
"loss": 0.6117,
"step": 410
},
{
"epoch": 0.17,
"learning_rate": 0.0002,
"loss": 0.5963,
"step": 420
},
{
"epoch": 0.17,
"learning_rate": 0.0002,
"loss": 0.5866,
"step": 430
},
{
"epoch": 0.18,
"learning_rate": 0.0002,
"loss": 0.5433,
"step": 440
},
{
"epoch": 0.18,
"learning_rate": 0.0002,
"loss": 0.5432,
"step": 450
},
{
"epoch": 0.19,
"learning_rate": 0.0002,
"loss": 0.5713,
"step": 460
},
{
"epoch": 0.19,
"learning_rate": 0.0002,
"loss": 0.5957,
"step": 470
},
{
"epoch": 0.19,
"learning_rate": 0.0002,
"loss": 0.6526,
"step": 480
},
{
"epoch": 0.2,
"learning_rate": 0.0002,
"loss": 0.57,
"step": 490
},
{
"epoch": 0.2,
"learning_rate": 0.0002,
"loss": 0.5938,
"step": 500
},
{
"epoch": 0.21,
"learning_rate": 0.0002,
"loss": 0.6141,
"step": 510
},
{
"epoch": 0.21,
"learning_rate": 0.0002,
"loss": 0.5262,
"step": 520
},
{
"epoch": 0.21,
"learning_rate": 0.0002,
"loss": 0.7055,
"step": 530
},
{
"epoch": 0.22,
"learning_rate": 0.0002,
"loss": 0.5412,
"step": 540
},
{
"epoch": 0.22,
"learning_rate": 0.0002,
"loss": 0.4956,
"step": 550
},
{
"epoch": 0.23,
"learning_rate": 0.0002,
"loss": 0.6345,
"step": 560
},
{
"epoch": 0.23,
"learning_rate": 0.0002,
"loss": 0.5665,
"step": 570
},
{
"epoch": 0.23,
"learning_rate": 0.0002,
"loss": 0.6687,
"step": 580
},
{
"epoch": 0.24,
"learning_rate": 0.0002,
"loss": 0.5994,
"step": 590
},
{
"epoch": 0.24,
"learning_rate": 0.0002,
"loss": 0.6209,
"step": 600
},
{
"epoch": 0.24,
"eval_loss": 0.6194782853126526,
"eval_runtime": 94.0475,
"eval_samples_per_second": 10.633,
"eval_steps_per_second": 5.316,
"step": 600
},
{
"epoch": 0.24,
"mmlu_eval_accuracy": 0.44690777926110636,
"mmlu_eval_accuracy_abstract_algebra": 0.18181818181818182,
"mmlu_eval_accuracy_anatomy": 0.42857142857142855,
"mmlu_eval_accuracy_astronomy": 0.375,
"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": 0.0,
"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": 0.3076923076923077,
"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": 0.2857142857142857,
"mmlu_eval_accuracy_global_facts": 0.2,
"mmlu_eval_accuracy_high_school_biology": 0.28125,
"mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
"mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_geography": 0.6818181818181818,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.3488372093023256,
"mmlu_eval_accuracy_high_school_mathematics": 0.3103448275862069,
"mmlu_eval_accuracy_high_school_microeconomics": 0.38461538461538464,
"mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
"mmlu_eval_accuracy_high_school_psychology": 0.6833333333333333,
"mmlu_eval_accuracy_high_school_statistics": 0.2608695652173913,
"mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
"mmlu_eval_accuracy_high_school_world_history": 0.5384615384615384,
"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": 0.45454545454545453,
"mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
"mmlu_eval_accuracy_machine_learning": 0.09090909090909091,
"mmlu_eval_accuracy_management": 0.36363636363636365,
"mmlu_eval_accuracy_marketing": 0.72,
"mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
"mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
"mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
"mmlu_eval_accuracy_moral_scenarios": 0.24,
"mmlu_eval_accuracy_nutrition": 0.5454545454545454,
"mmlu_eval_accuracy_philosophy": 0.5588235294117647,
"mmlu_eval_accuracy_prehistory": 0.42857142857142855,
"mmlu_eval_accuracy_professional_accounting": 0.25806451612903225,
"mmlu_eval_accuracy_professional_law": 0.3411764705882353,
"mmlu_eval_accuracy_professional_medicine": 0.3870967741935484,
"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": 0.6818181818181818,
"mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
"mmlu_eval_accuracy_virology": 0.5555555555555556,
"mmlu_eval_accuracy_world_religions": 0.7368421052631579,
"mmlu_loss": 0.9532866988613151,
"step": 600
},
{
"epoch": 0.25,
"learning_rate": 0.0002,
"loss": 0.6145,
"step": 610
},
{
"epoch": 0.25,
"learning_rate": 0.0002,
"loss": 0.6267,
"step": 620
},
{
"epoch": 0.26,
"learning_rate": 0.0002,
"loss": 0.5609,
"step": 630
},
{
"epoch": 0.26,
"learning_rate": 0.0002,
"loss": 0.4955,
"step": 640
},
{
"epoch": 0.26,
"learning_rate": 0.0002,
"loss": 0.551,
"step": 650
},
{
"epoch": 0.27,
"learning_rate": 0.0002,
"loss": 0.5323,
"step": 660
},
{
"epoch": 0.27,
"learning_rate": 0.0002,
"loss": 0.5905,
"step": 670
},
{
"epoch": 0.28,
"learning_rate": 0.0002,
"loss": 0.7368,
"step": 680
},
{
"epoch": 0.28,
"learning_rate": 0.0002,
"loss": 0.573,
"step": 690
},
{
"epoch": 0.28,
"learning_rate": 0.0002,
"loss": 0.5785,
"step": 700
},
{
"epoch": 0.29,
"learning_rate": 0.0002,
"loss": 0.5802,
"step": 710
},
{
"epoch": 0.29,
"learning_rate": 0.0002,
"loss": 0.5823,
"step": 720
},
{
"epoch": 0.3,
"learning_rate": 0.0002,
"loss": 0.5718,
"step": 730
},
{
"epoch": 0.3,
"learning_rate": 0.0002,
"loss": 0.5783,
"step": 740
},
{
"epoch": 0.3,
"learning_rate": 0.0002,
"loss": 0.5367,
"step": 750
},
{
"epoch": 0.31,
"learning_rate": 0.0002,
"loss": 0.6111,
"step": 760
},
{
"epoch": 0.31,
"learning_rate": 0.0002,
"loss": 0.5343,
"step": 770
},
{
"epoch": 0.32,
"learning_rate": 0.0002,
"loss": 0.6399,
"step": 780
},
{
"epoch": 0.32,
"learning_rate": 0.0002,
"loss": 0.6314,
"step": 790
},
{
"epoch": 0.32,
"learning_rate": 0.0002,
"loss": 0.506,
"step": 800
},
{
"epoch": 0.32,
"eval_loss": 0.612710177898407,
"eval_runtime": 94.0353,
"eval_samples_per_second": 10.634,
"eval_steps_per_second": 5.317,
"step": 800
},
{
"epoch": 0.32,
"mmlu_eval_accuracy": 0.451824690933893,
"mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
"mmlu_eval_accuracy_anatomy": 0.5714285714285714,
"mmlu_eval_accuracy_astronomy": 0.4375,
"mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
"mmlu_eval_accuracy_clinical_knowledge": 0.4482758620689655,
"mmlu_eval_accuracy_college_biology": 0.4375,
"mmlu_eval_accuracy_college_chemistry": 0.125,
"mmlu_eval_accuracy_college_computer_science": 0.18181818181818182,
"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": 0.45454545454545453,
"mmlu_eval_accuracy_conceptual_physics": 0.3076923076923077,
"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": 0.2857142857142857,
"mmlu_eval_accuracy_global_facts": 0.3,
"mmlu_eval_accuracy_high_school_biology": 0.40625,
"mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
"mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
"mmlu_eval_accuracy_high_school_european_history": 0.4444444444444444,
"mmlu_eval_accuracy_high_school_geography": 0.6818181818181818,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.37209302325581395,
"mmlu_eval_accuracy_high_school_mathematics": 0.3448275862068966,
"mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
"mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
"mmlu_eval_accuracy_high_school_psychology": 0.75,
"mmlu_eval_accuracy_high_school_statistics": 0.2608695652173913,
"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": 0.3333333333333333,
"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": 0.36363636363636365,
"mmlu_eval_accuracy_marketing": 0.64,
"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": 0.23,
"mmlu_eval_accuracy_nutrition": 0.5454545454545454,
"mmlu_eval_accuracy_philosophy": 0.47058823529411764,
"mmlu_eval_accuracy_prehistory": 0.45714285714285713,
"mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
"mmlu_eval_accuracy_professional_law": 0.3411764705882353,
"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": 0.4444444444444444,
"mmlu_eval_accuracy_sociology": 0.5909090909090909,
"mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
"mmlu_eval_accuracy_virology": 0.5,
"mmlu_eval_accuracy_world_religions": 0.7368421052631579,
"mmlu_loss": 0.9771831466113619,
"step": 800
},
{
"epoch": 0.33,
"learning_rate": 0.0002,
"loss": 0.604,
"step": 810
},
{
"epoch": 0.33,
"learning_rate": 0.0002,
"loss": 0.5633,
"step": 820
},
{
"epoch": 0.34,
"learning_rate": 0.0002,
"loss": 0.5965,
"step": 830
},
{
"epoch": 0.34,
"learning_rate": 0.0002,
"loss": 0.5563,
"step": 840
},
{
"epoch": 0.34,
"learning_rate": 0.0002,
"loss": 0.6227,
"step": 850
},
{
"epoch": 0.35,
"learning_rate": 0.0002,
"loss": 0.6758,
"step": 860
},
{
"epoch": 0.35,
"learning_rate": 0.0002,
"loss": 0.6293,
"step": 870
},
{
"epoch": 0.36,
"learning_rate": 0.0002,
"loss": 0.6711,
"step": 880
},
{
"epoch": 0.36,
"learning_rate": 0.0002,
"loss": 0.5607,
"step": 890
},
{
"epoch": 0.36,
"learning_rate": 0.0002,
"loss": 0.66,
"step": 900
},
{
"epoch": 0.37,
"learning_rate": 0.0002,
"loss": 0.5449,
"step": 910
},
{
"epoch": 0.37,
"learning_rate": 0.0002,
"loss": 0.5715,
"step": 920
},
{
"epoch": 0.38,
"learning_rate": 0.0002,
"loss": 0.5366,
"step": 930
},
{
"epoch": 0.38,
"learning_rate": 0.0002,
"loss": 0.4633,
"step": 940
},
{
"epoch": 0.38,
"learning_rate": 0.0002,
"loss": 0.5635,
"step": 950
},
{
"epoch": 0.39,
"learning_rate": 0.0002,
"loss": 0.5331,
"step": 960
},
{
"epoch": 0.39,
"learning_rate": 0.0002,
"loss": 0.5642,
"step": 970
},
{
"epoch": 0.4,
"learning_rate": 0.0002,
"loss": 0.6002,
"step": 980
},
{
"epoch": 0.4,
"learning_rate": 0.0002,
"loss": 0.5484,
"step": 990
},
{
"epoch": 0.4,
"learning_rate": 0.0002,
"loss": 0.591,
"step": 1000
},
{
"epoch": 0.4,
"eval_loss": 0.6107870936393738,
"eval_runtime": 94.1069,
"eval_samples_per_second": 10.626,
"eval_steps_per_second": 5.313,
"step": 1000
},
{
"epoch": 0.4,
"mmlu_eval_accuracy": 0.44434023866715483,
"mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
"mmlu_eval_accuracy_anatomy": 0.5,
"mmlu_eval_accuracy_astronomy": 0.3125,
"mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
"mmlu_eval_accuracy_clinical_knowledge": 0.41379310344827586,
"mmlu_eval_accuracy_college_biology": 0.375,
"mmlu_eval_accuracy_college_chemistry": 0.0,
"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": 0.34615384615384615,
"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": 0.3,
"mmlu_eval_accuracy_high_school_biology": 0.375,
"mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
"mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
"mmlu_eval_accuracy_high_school_geography": 0.6818181818181818,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.32558139534883723,
"mmlu_eval_accuracy_high_school_mathematics": 0.3103448275862069,
"mmlu_eval_accuracy_high_school_microeconomics": 0.38461538461538464,
"mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
"mmlu_eval_accuracy_high_school_psychology": 0.7,
"mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
"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": 0.36363636363636365,
"mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
"mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
"mmlu_eval_accuracy_management": 0.36363636363636365,
"mmlu_eval_accuracy_marketing": 0.72,
"mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
"mmlu_eval_accuracy_miscellaneous": 0.627906976744186,
"mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
"mmlu_eval_accuracy_moral_scenarios": 0.24,
"mmlu_eval_accuracy_nutrition": 0.5454545454545454,
"mmlu_eval_accuracy_philosophy": 0.47058823529411764,
"mmlu_eval_accuracy_prehistory": 0.4,
"mmlu_eval_accuracy_professional_accounting": 0.22580645161290322,
"mmlu_eval_accuracy_professional_law": 0.3411764705882353,
"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": 0.5185185185185185,
"mmlu_eval_accuracy_sociology": 0.5,
"mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
"mmlu_eval_accuracy_virology": 0.6111111111111112,
"mmlu_eval_accuracy_world_religions": 0.7368421052631579,
"mmlu_loss": 0.9966794518515585,
"step": 1000
},
{
"epoch": 0.41,
"learning_rate": 0.0002,
"loss": 0.5453,
"step": 1010
},
{
"epoch": 0.41,
"learning_rate": 0.0002,
"loss": 0.4825,
"step": 1020
},
{
"epoch": 0.42,
"learning_rate": 0.0002,
"loss": 0.5663,
"step": 1030
},
{
"epoch": 0.42,
"learning_rate": 0.0002,
"loss": 0.4528,
"step": 1040
},
{
"epoch": 0.43,
"learning_rate": 0.0002,
"loss": 0.6646,
"step": 1050
},
{
"epoch": 0.43,
"learning_rate": 0.0002,
"loss": 0.5613,
"step": 1060
},
{
"epoch": 0.43,
"learning_rate": 0.0002,
"loss": 0.6912,
"step": 1070
},
{
"epoch": 0.44,
"learning_rate": 0.0002,
"loss": 0.6891,
"step": 1080
},
{
"epoch": 0.44,
"learning_rate": 0.0002,
"loss": 0.5508,
"step": 1090
},
{
"epoch": 0.45,
"learning_rate": 0.0002,
"loss": 0.6595,
"step": 1100
},
{
"epoch": 0.45,
"learning_rate": 0.0002,
"loss": 0.5936,
"step": 1110
},
{
"epoch": 0.45,
"learning_rate": 0.0002,
"loss": 0.6558,
"step": 1120
},
{
"epoch": 0.46,
"learning_rate": 0.0002,
"loss": 0.6729,
"step": 1130
},
{
"epoch": 0.46,
"learning_rate": 0.0002,
"loss": 0.6205,
"step": 1140
},
{
"epoch": 0.47,
"learning_rate": 0.0002,
"loss": 0.6675,
"step": 1150
},
{
"epoch": 0.47,
"learning_rate": 0.0002,
"loss": 0.5649,
"step": 1160
},
{
"epoch": 0.47,
"learning_rate": 0.0002,
"loss": 0.5922,
"step": 1170
},
{
"epoch": 0.48,
"learning_rate": 0.0002,
"loss": 0.4905,
"step": 1180
},
{
"epoch": 0.48,
"learning_rate": 0.0002,
"loss": 0.6746,
"step": 1190
},
{
"epoch": 0.49,
"learning_rate": 0.0002,
"loss": 0.6171,
"step": 1200
},
{
"epoch": 0.49,
"eval_loss": 0.612968921661377,
"eval_runtime": 94.0813,
"eval_samples_per_second": 10.629,
"eval_steps_per_second": 5.315,
"step": 1200
},
{
"epoch": 0.49,
"mmlu_eval_accuracy": 0.4287111481518256,
"mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
"mmlu_eval_accuracy_anatomy": 0.5,
"mmlu_eval_accuracy_astronomy": 0.375,
"mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
"mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
"mmlu_eval_accuracy_college_biology": 0.375,
"mmlu_eval_accuracy_college_chemistry": 0.0,
"mmlu_eval_accuracy_college_computer_science": 0.09090909090909091,
"mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
"mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
"mmlu_eval_accuracy_college_physics": 0.36363636363636365,
"mmlu_eval_accuracy_computer_security": 0.2727272727272727,
"mmlu_eval_accuracy_conceptual_physics": 0.34615384615384615,
"mmlu_eval_accuracy_econometrics": 0.16666666666666666,
"mmlu_eval_accuracy_electrical_engineering": 0.3125,
"mmlu_eval_accuracy_elementary_mathematics": 0.4146341463414634,
"mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
"mmlu_eval_accuracy_global_facts": 0.3,
"mmlu_eval_accuracy_high_school_biology": 0.25,
"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": 0.5238095238095238,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.3488372093023256,
"mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
"mmlu_eval_accuracy_high_school_microeconomics": 0.3076923076923077,
"mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
"mmlu_eval_accuracy_high_school_psychology": 0.7,
"mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
"mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
"mmlu_eval_accuracy_high_school_world_history": 0.4230769230769231,
"mmlu_eval_accuracy_human_aging": 0.6086956521739131,
"mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
"mmlu_eval_accuracy_international_law": 0.7692307692307693,
"mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
"mmlu_eval_accuracy_logical_fallacies": 0.5,
"mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
"mmlu_eval_accuracy_management": 0.36363636363636365,
"mmlu_eval_accuracy_marketing": 0.76,
"mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
"mmlu_eval_accuracy_miscellaneous": 0.6046511627906976,
"mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
"mmlu_eval_accuracy_moral_scenarios": 0.23,
"mmlu_eval_accuracy_nutrition": 0.5151515151515151,
"mmlu_eval_accuracy_philosophy": 0.5,
"mmlu_eval_accuracy_prehistory": 0.4,
"mmlu_eval_accuracy_professional_accounting": 0.22580645161290322,
"mmlu_eval_accuracy_professional_law": 0.3235294117647059,
"mmlu_eval_accuracy_professional_medicine": 0.3548387096774194,
"mmlu_eval_accuracy_professional_psychology": 0.43478260869565216,
"mmlu_eval_accuracy_public_relations": 0.6666666666666666,
"mmlu_eval_accuracy_security_studies": 0.4444444444444444,
"mmlu_eval_accuracy_sociology": 0.45454545454545453,
"mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
"mmlu_eval_accuracy_virology": 0.6111111111111112,
"mmlu_eval_accuracy_world_religions": 0.7368421052631579,
"mmlu_loss": 1.0120855229213717,
"step": 1200
},
{
"epoch": 0.49,
"learning_rate": 0.0002,
"loss": 0.5031,
"step": 1210
},
{
"epoch": 0.49,
"learning_rate": 0.0002,
"loss": 0.5928,
"step": 1220
},
{
"epoch": 0.5,
"learning_rate": 0.0002,
"loss": 0.5746,
"step": 1230
},
{
"epoch": 0.5,
"learning_rate": 0.0002,
"loss": 0.572,
"step": 1240
},
{
"epoch": 0.51,
"learning_rate": 0.0002,
"loss": 0.5716,
"step": 1250
},
{
"epoch": 0.51,
"learning_rate": 0.0002,
"loss": 0.4872,
"step": 1260
},
{
"epoch": 0.51,
"learning_rate": 0.0002,
"loss": 0.6716,
"step": 1270
},
{
"epoch": 0.52,
"learning_rate": 0.0002,
"loss": 0.6052,
"step": 1280
},
{
"epoch": 0.52,
"learning_rate": 0.0002,
"loss": 0.5711,
"step": 1290
},
{
"epoch": 0.53,
"learning_rate": 0.0002,
"loss": 0.7097,
"step": 1300
},
{
"epoch": 0.53,
"learning_rate": 0.0002,
"loss": 0.5536,
"step": 1310
},
{
"epoch": 0.53,
"learning_rate": 0.0002,
"loss": 0.7815,
"step": 1320
},
{
"epoch": 0.54,
"learning_rate": 0.0002,
"loss": 0.6709,
"step": 1330
},
{
"epoch": 0.54,
"learning_rate": 0.0002,
"loss": 0.5422,
"step": 1340
},
{
"epoch": 0.55,
"learning_rate": 0.0002,
"loss": 0.566,
"step": 1350
},
{
"epoch": 0.55,
"learning_rate": 0.0002,
"loss": 0.4571,
"step": 1360
},
{
"epoch": 0.55,
"learning_rate": 0.0002,
"loss": 0.6572,
"step": 1370
},
{
"epoch": 0.56,
"learning_rate": 0.0002,
"loss": 0.5951,
"step": 1380
},
{
"epoch": 0.56,
"learning_rate": 0.0002,
"loss": 0.6753,
"step": 1390
},
{
"epoch": 0.57,
"learning_rate": 0.0002,
"loss": 0.6247,
"step": 1400
},
{
"epoch": 0.57,
"eval_loss": 0.6076797842979431,
"eval_runtime": 94.144,
"eval_samples_per_second": 10.622,
"eval_steps_per_second": 5.311,
"step": 1400
},
{
"epoch": 0.57,
"mmlu_eval_accuracy": 0.434467608651013,
"mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
"mmlu_eval_accuracy_anatomy": 0.5714285714285714,
"mmlu_eval_accuracy_astronomy": 0.375,
"mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
"mmlu_eval_accuracy_clinical_knowledge": 0.41379310344827586,
"mmlu_eval_accuracy_college_biology": 0.25,
"mmlu_eval_accuracy_college_chemistry": 0.0,
"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": 0.36363636363636365,
"mmlu_eval_accuracy_computer_security": 0.45454545454545453,
"mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
"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": 0.2,
"mmlu_eval_accuracy_high_school_biology": 0.34375,
"mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
"mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_european_history": 0.4444444444444444,
"mmlu_eval_accuracy_high_school_geography": 0.7272727272727273,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.3488372093023256,
"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": 0.6833333333333333,
"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": 0.6086956521739131,
"mmlu_eval_accuracy_human_sexuality": 0.25,
"mmlu_eval_accuracy_international_law": 0.7692307692307693,
"mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
"mmlu_eval_accuracy_logical_fallacies": 0.5,
"mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
"mmlu_eval_accuracy_management": 0.45454545454545453,
"mmlu_eval_accuracy_marketing": 0.68,
"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": 0.23,
"mmlu_eval_accuracy_nutrition": 0.5454545454545454,
"mmlu_eval_accuracy_philosophy": 0.4411764705882353,
"mmlu_eval_accuracy_prehistory": 0.42857142857142855,
"mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
"mmlu_eval_accuracy_professional_law": 0.3235294117647059,
"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": 0.4444444444444444,
"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": 0.9527658471161641,
"step": 1400
},
{
"epoch": 0.57,
"learning_rate": 0.0002,
"loss": 0.6337,
"step": 1410
},
{
"epoch": 0.57,
"learning_rate": 0.0002,
"loss": 0.5077,
"step": 1420
},
{
"epoch": 0.58,
"learning_rate": 0.0002,
"loss": 0.5413,
"step": 1430
},
{
"epoch": 0.58,
"learning_rate": 0.0002,
"loss": 0.6527,
"step": 1440
},
{
"epoch": 0.59,
"learning_rate": 0.0002,
"loss": 0.6435,
"step": 1450
},
{
"epoch": 0.59,
"learning_rate": 0.0002,
"loss": 0.5503,
"step": 1460
},
{
"epoch": 0.6,
"learning_rate": 0.0002,
"loss": 0.5819,
"step": 1470
},
{
"epoch": 0.6,
"learning_rate": 0.0002,
"loss": 0.6342,
"step": 1480
},
{
"epoch": 0.6,
"learning_rate": 0.0002,
"loss": 0.5843,
"step": 1490
},
{
"epoch": 0.61,
"learning_rate": 0.0002,
"loss": 0.5134,
"step": 1500
},
{
"epoch": 0.61,
"learning_rate": 0.0002,
"loss": 0.5694,
"step": 1510
},
{
"epoch": 0.62,
"learning_rate": 0.0002,
"loss": 0.6172,
"step": 1520
},
{
"epoch": 0.62,
"learning_rate": 0.0002,
"loss": 0.5765,
"step": 1530
},
{
"epoch": 0.62,
"learning_rate": 0.0002,
"loss": 0.591,
"step": 1540
},
{
"epoch": 0.63,
"learning_rate": 0.0002,
"loss": 0.5039,
"step": 1550
},
{
"epoch": 0.63,
"learning_rate": 0.0002,
"loss": 0.6288,
"step": 1560
},
{
"epoch": 0.64,
"learning_rate": 0.0002,
"loss": 0.5196,
"step": 1570
},
{
"epoch": 0.64,
"learning_rate": 0.0002,
"loss": 0.5867,
"step": 1580
},
{
"epoch": 0.64,
"learning_rate": 0.0002,
"loss": 0.6002,
"step": 1590
},
{
"epoch": 0.65,
"learning_rate": 0.0002,
"loss": 0.6534,
"step": 1600
},
{
"epoch": 0.65,
"eval_loss": 0.6057603359222412,
"eval_runtime": 94.1839,
"eval_samples_per_second": 10.618,
"eval_steps_per_second": 5.309,
"step": 1600
},
{
"epoch": 0.65,
"mmlu_eval_accuracy": 0.4521968363152983,
"mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
"mmlu_eval_accuracy_anatomy": 0.5714285714285714,
"mmlu_eval_accuracy_astronomy": 0.375,
"mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
"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": 0.36363636363636365,
"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": 0.5454545454545454,
"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": 0.2926829268292683,
"mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
"mmlu_eval_accuracy_global_facts": 0.3,
"mmlu_eval_accuracy_high_school_biology": 0.375,
"mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
"mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
"mmlu_eval_accuracy_high_school_european_history": 0.5,
"mmlu_eval_accuracy_high_school_geography": 0.7272727272727273,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.46511627906976744,
"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": 0.75,
"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": 0.3333333333333333,
"mmlu_eval_accuracy_international_law": 0.7692307692307693,
"mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
"mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
"mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
"mmlu_eval_accuracy_management": 0.5454545454545454,
"mmlu_eval_accuracy_marketing": 0.76,
"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": 0.28,
"mmlu_eval_accuracy_nutrition": 0.5454545454545454,
"mmlu_eval_accuracy_philosophy": 0.4411764705882353,
"mmlu_eval_accuracy_prehistory": 0.5142857142857142,
"mmlu_eval_accuracy_professional_accounting": 0.25806451612903225,
"mmlu_eval_accuracy_professional_law": 0.3411764705882353,
"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": 0.5185185185185185,
"mmlu_eval_accuracy_sociology": 0.5454545454545454,
"mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
"mmlu_eval_accuracy_virology": 0.5,
"mmlu_eval_accuracy_world_religions": 0.7368421052631579,
"mmlu_loss": 0.8609263468582388,
"step": 1600
},
{
"epoch": 0.65,
"learning_rate": 0.0002,
"loss": 0.5552,
"step": 1610
},
{
"epoch": 0.66,
"learning_rate": 0.0002,
"loss": 0.4649,
"step": 1620
},
{
"epoch": 0.66,
"learning_rate": 0.0002,
"loss": 0.5148,
"step": 1630
},
{
"epoch": 0.66,
"learning_rate": 0.0002,
"loss": 0.4968,
"step": 1640
},
{
"epoch": 0.67,
"learning_rate": 0.0002,
"loss": 0.5822,
"step": 1650
},
{
"epoch": 0.67,
"learning_rate": 0.0002,
"loss": 0.4779,
"step": 1660
},
{
"epoch": 0.68,
"learning_rate": 0.0002,
"loss": 0.6367,
"step": 1670
},
{
"epoch": 0.68,
"learning_rate": 0.0002,
"loss": 0.7188,
"step": 1680
},
{
"epoch": 0.68,
"learning_rate": 0.0002,
"loss": 0.5493,
"step": 1690
},
{
"epoch": 0.69,
"learning_rate": 0.0002,
"loss": 0.5365,
"step": 1700
},
{
"epoch": 0.69,
"learning_rate": 0.0002,
"loss": 0.6451,
"step": 1710
},
{
"epoch": 0.7,
"learning_rate": 0.0002,
"loss": 0.5231,
"step": 1720
},
{
"epoch": 0.7,
"learning_rate": 0.0002,
"loss": 0.7517,
"step": 1730
},
{
"epoch": 0.7,
"learning_rate": 0.0002,
"loss": 0.5724,
"step": 1740
},
{
"epoch": 0.71,
"learning_rate": 0.0002,
"loss": 0.4755,
"step": 1750
},
{
"epoch": 0.71,
"learning_rate": 0.0002,
"loss": 0.672,
"step": 1760
},
{
"epoch": 0.72,
"learning_rate": 0.0002,
"loss": 0.6718,
"step": 1770
},
{
"epoch": 0.72,
"learning_rate": 0.0002,
"loss": 0.6726,
"step": 1780
},
{
"epoch": 0.72,
"learning_rate": 0.0002,
"loss": 0.5012,
"step": 1790
},
{
"epoch": 0.73,
"learning_rate": 0.0002,
"loss": 0.4542,
"step": 1800
},
{
"epoch": 0.73,
"eval_loss": 0.6079343557357788,
"eval_runtime": 94.5927,
"eval_samples_per_second": 10.572,
"eval_steps_per_second": 5.286,
"step": 1800
},
{
"epoch": 0.73,
"mmlu_eval_accuracy": 0.45560649806753273,
"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": 0.4482758620689655,
"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": 0.18181818181818182,
"mmlu_eval_accuracy_college_medicine": 0.3181818181818182,
"mmlu_eval_accuracy_college_physics": 0.45454545454545453,
"mmlu_eval_accuracy_computer_security": 0.18181818181818182,
"mmlu_eval_accuracy_conceptual_physics": 0.3076923076923077,
"mmlu_eval_accuracy_econometrics": 0.16666666666666666,
"mmlu_eval_accuracy_electrical_engineering": 0.4375,
"mmlu_eval_accuracy_elementary_mathematics": 0.4634146341463415,
"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": 0.36363636363636365,
"mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
"mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
"mmlu_eval_accuracy_high_school_geography": 0.6818181818181818,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.3023255813953488,
"mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
"mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
"mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
"mmlu_eval_accuracy_high_school_psychology": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
"mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
"mmlu_eval_accuracy_high_school_world_history": 0.5769230769230769,
"mmlu_eval_accuracy_human_aging": 0.6086956521739131,
"mmlu_eval_accuracy_human_sexuality": 0.5,
"mmlu_eval_accuracy_international_law": 0.7692307692307693,
"mmlu_eval_accuracy_jurisprudence": 0.5454545454545454,
"mmlu_eval_accuracy_logical_fallacies": 0.5,
"mmlu_eval_accuracy_machine_learning": 0.09090909090909091,
"mmlu_eval_accuracy_management": 0.5454545454545454,
"mmlu_eval_accuracy_marketing": 0.8,
"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": 0.23,
"mmlu_eval_accuracy_nutrition": 0.5454545454545454,
"mmlu_eval_accuracy_philosophy": 0.5882352941176471,
"mmlu_eval_accuracy_prehistory": 0.37142857142857144,
"mmlu_eval_accuracy_professional_accounting": 0.25806451612903225,
"mmlu_eval_accuracy_professional_law": 0.3411764705882353,
"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": 0.5454545454545454,
"mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
"mmlu_eval_accuracy_virology": 0.4444444444444444,
"mmlu_eval_accuracy_world_religions": 0.6842105263157895,
"mmlu_loss": 0.9425534363398352,
"step": 1800
},
{
"epoch": 0.73,
"learning_rate": 0.0002,
"loss": 0.4603,
"step": 1810
},
{
"epoch": 0.74,
"learning_rate": 0.0002,
"loss": 0.5112,
"step": 1820
},
{
"epoch": 0.74,
"learning_rate": 0.0002,
"loss": 0.6551,
"step": 1830
},
{
"epoch": 0.74,
"learning_rate": 0.0002,
"loss": 0.5428,
"step": 1840
},
{
"epoch": 0.75,
"learning_rate": 0.0002,
"loss": 0.634,
"step": 1850
},
{
"epoch": 0.75,
"learning_rate": 0.0002,
"loss": 0.538,
"step": 1860
},
{
"epoch": 0.76,
"learning_rate": 0.0002,
"loss": 0.5745,
"step": 1870
},
{
"epoch": 0.76,
"learning_rate": 0.0002,
"loss": 0.7127,
"step": 1880
},
{
"epoch": 0.77,
"learning_rate": 0.0002,
"loss": 0.6231,
"step": 1890
},
{
"epoch": 0.77,
"learning_rate": 0.0002,
"loss": 0.5608,
"step": 1900
},
{
"epoch": 0.77,
"learning_rate": 0.0002,
"loss": 0.6482,
"step": 1910
},
{
"epoch": 0.78,
"learning_rate": 0.0002,
"loss": 0.5111,
"step": 1920
},
{
"epoch": 0.78,
"learning_rate": 0.0002,
"loss": 0.6582,
"step": 1930
},
{
"epoch": 0.79,
"learning_rate": 0.0002,
"loss": 0.6121,
"step": 1940
},
{
"epoch": 0.79,
"learning_rate": 0.0002,
"loss": 0.6185,
"step": 1950
},
{
"epoch": 0.79,
"learning_rate": 0.0002,
"loss": 0.5918,
"step": 1960
},
{
"epoch": 0.8,
"learning_rate": 0.0002,
"loss": 0.5883,
"step": 1970
},
{
"epoch": 0.8,
"learning_rate": 0.0002,
"loss": 0.6027,
"step": 1980
},
{
"epoch": 0.81,
"learning_rate": 0.0002,
"loss": 0.4892,
"step": 1990
},
{
"epoch": 0.81,
"learning_rate": 0.0002,
"loss": 0.6467,
"step": 2000
},
{
"epoch": 0.81,
"eval_loss": 0.6004832983016968,
"eval_runtime": 93.9867,
"eval_samples_per_second": 10.64,
"eval_steps_per_second": 5.32,
"step": 2000
},
{
"epoch": 0.81,
"mmlu_eval_accuracy": 0.4582647054610207,
"mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
"mmlu_eval_accuracy_anatomy": 0.5714285714285714,
"mmlu_eval_accuracy_astronomy": 0.375,
"mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
"mmlu_eval_accuracy_clinical_knowledge": 0.41379310344827586,
"mmlu_eval_accuracy_college_biology": 0.375,
"mmlu_eval_accuracy_college_chemistry": 0.375,
"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": 0.45454545454545453,
"mmlu_eval_accuracy_computer_security": 0.09090909090909091,
"mmlu_eval_accuracy_conceptual_physics": 0.3076923076923077,
"mmlu_eval_accuracy_econometrics": 0.16666666666666666,
"mmlu_eval_accuracy_electrical_engineering": 0.375,
"mmlu_eval_accuracy_elementary_mathematics": 0.4146341463414634,
"mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
"mmlu_eval_accuracy_global_facts": 0.3,
"mmlu_eval_accuracy_high_school_biology": 0.34375,
"mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
"mmlu_eval_accuracy_high_school_computer_science": 0.7777777777777778,
"mmlu_eval_accuracy_high_school_european_history": 0.4444444444444444,
"mmlu_eval_accuracy_high_school_geography": 0.7272727272727273,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.37209302325581395,
"mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
"mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
"mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
"mmlu_eval_accuracy_high_school_psychology": 0.6833333333333333,
"mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
"mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
"mmlu_eval_accuracy_high_school_world_history": 0.6153846153846154,
"mmlu_eval_accuracy_human_aging": 0.6956521739130435,
"mmlu_eval_accuracy_human_sexuality": 0.5,
"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": 0.18181818181818182,
"mmlu_eval_accuracy_management": 0.5454545454545454,
"mmlu_eval_accuracy_marketing": 0.8,
"mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
"mmlu_eval_accuracy_miscellaneous": 0.6976744186046512,
"mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
"mmlu_eval_accuracy_moral_scenarios": 0.24,
"mmlu_eval_accuracy_nutrition": 0.5454545454545454,
"mmlu_eval_accuracy_philosophy": 0.5294117647058824,
"mmlu_eval_accuracy_prehistory": 0.37142857142857144,
"mmlu_eval_accuracy_professional_accounting": 0.41935483870967744,
"mmlu_eval_accuracy_professional_law": 0.3235294117647059,
"mmlu_eval_accuracy_professional_medicine": 0.41935483870967744,
"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": 0.5454545454545454,
"mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
"mmlu_eval_accuracy_virology": 0.5555555555555556,
"mmlu_eval_accuracy_world_religions": 0.6842105263157895,
"mmlu_loss": 0.8762637240612787,
"step": 2000
},
{
"epoch": 0.81,
"learning_rate": 0.0002,
"loss": 0.6325,
"step": 2010
},
{
"epoch": 0.82,
"learning_rate": 0.0002,
"loss": 0.5258,
"step": 2020
},
{
"epoch": 0.82,
"learning_rate": 0.0002,
"loss": 0.5538,
"step": 2030
},
{
"epoch": 0.83,
"learning_rate": 0.0002,
"loss": 0.598,
"step": 2040
},
{
"epoch": 0.83,
"learning_rate": 0.0002,
"loss": 0.5337,
"step": 2050
},
{
"epoch": 0.83,
"learning_rate": 0.0002,
"loss": 0.5749,
"step": 2060
},
{
"epoch": 0.84,
"learning_rate": 0.0002,
"loss": 0.664,
"step": 2070
},
{
"epoch": 0.84,
"learning_rate": 0.0002,
"loss": 0.6095,
"step": 2080
},
{
"epoch": 0.85,
"learning_rate": 0.0002,
"loss": 0.5729,
"step": 2090
},
{
"epoch": 0.85,
"learning_rate": 0.0002,
"loss": 0.6395,
"step": 2100
},
{
"epoch": 0.85,
"learning_rate": 0.0002,
"loss": 0.5581,
"step": 2110
},
{
"epoch": 0.86,
"learning_rate": 0.0002,
"loss": 0.6305,
"step": 2120
},
{
"epoch": 0.86,
"learning_rate": 0.0002,
"loss": 0.6186,
"step": 2130
},
{
"epoch": 0.87,
"learning_rate": 0.0002,
"loss": 0.4686,
"step": 2140
},
{
"epoch": 0.87,
"learning_rate": 0.0002,
"loss": 0.6395,
"step": 2150
},
{
"epoch": 0.87,
"learning_rate": 0.0002,
"loss": 0.5673,
"step": 2160
},
{
"epoch": 0.88,
"learning_rate": 0.0002,
"loss": 0.5648,
"step": 2170
},
{
"epoch": 0.88,
"learning_rate": 0.0002,
"loss": 0.5265,
"step": 2180
},
{
"epoch": 0.89,
"learning_rate": 0.0002,
"loss": 0.542,
"step": 2190
},
{
"epoch": 0.89,
"learning_rate": 0.0002,
"loss": 0.488,
"step": 2200
},
{
"epoch": 0.89,
"eval_loss": 0.598466694355011,
"eval_runtime": 93.9641,
"eval_samples_per_second": 10.642,
"eval_steps_per_second": 5.321,
"step": 2200
},
{
"epoch": 0.89,
"mmlu_eval_accuracy": 0.453684883010787,
"mmlu_eval_accuracy_abstract_algebra": 0.09090909090909091,
"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": 0.0,
"mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
"mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
"mmlu_eval_accuracy_college_medicine": 0.2727272727272727,
"mmlu_eval_accuracy_college_physics": 0.5454545454545454,
"mmlu_eval_accuracy_computer_security": 0.2727272727272727,
"mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
"mmlu_eval_accuracy_econometrics": 0.16666666666666666,
"mmlu_eval_accuracy_electrical_engineering": 0.4375,
"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": 0.5555555555555556,
"mmlu_eval_accuracy_high_school_geography": 0.7272727272727273,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.32558139534883723,
"mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
"mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
"mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
"mmlu_eval_accuracy_high_school_psychology": 0.6833333333333333,
"mmlu_eval_accuracy_high_school_statistics": 0.21739130434782608,
"mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
"mmlu_eval_accuracy_high_school_world_history": 0.5384615384615384,
"mmlu_eval_accuracy_human_aging": 0.7391304347826086,
"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": 0.5555555555555556,
"mmlu_eval_accuracy_machine_learning": 0.09090909090909091,
"mmlu_eval_accuracy_management": 0.6363636363636364,
"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": 0.23,
"mmlu_eval_accuracy_nutrition": 0.5151515151515151,
"mmlu_eval_accuracy_philosophy": 0.47058823529411764,
"mmlu_eval_accuracy_prehistory": 0.34285714285714286,
"mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
"mmlu_eval_accuracy_professional_law": 0.3588235294117647,
"mmlu_eval_accuracy_professional_medicine": 0.45161290322580644,
"mmlu_eval_accuracy_professional_psychology": 0.34782608695652173,
"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": 0.5454545454545454,
"mmlu_eval_accuracy_virology": 0.6111111111111112,
"mmlu_eval_accuracy_world_religions": 0.6842105263157895,
"mmlu_loss": 0.9464586163158516,
"step": 2200
},
{
"epoch": 0.89,
"learning_rate": 0.0002,
"loss": 0.5428,
"step": 2210
},
{
"epoch": 0.9,
"learning_rate": 0.0002,
"loss": 0.6717,
"step": 2220
},
{
"epoch": 0.9,
"learning_rate": 0.0002,
"loss": 0.6128,
"step": 2230
},
{
"epoch": 0.91,
"learning_rate": 0.0002,
"loss": 0.5053,
"step": 2240
},
{
"epoch": 0.91,
"learning_rate": 0.0002,
"loss": 0.5135,
"step": 2250
},
{
"epoch": 0.91,
"learning_rate": 0.0002,
"loss": 0.5352,
"step": 2260
},
{
"epoch": 0.92,
"learning_rate": 0.0002,
"loss": 0.5411,
"step": 2270
},
{
"epoch": 0.92,
"learning_rate": 0.0002,
"loss": 0.7386,
"step": 2280
},
{
"epoch": 0.93,
"learning_rate": 0.0002,
"loss": 0.5334,
"step": 2290
},
{
"epoch": 0.93,
"learning_rate": 0.0002,
"loss": 0.5402,
"step": 2300
},
{
"epoch": 0.94,
"learning_rate": 0.0002,
"loss": 0.7309,
"step": 2310
},
{
"epoch": 0.94,
"learning_rate": 0.0002,
"loss": 0.7377,
"step": 2320
},
{
"epoch": 0.94,
"learning_rate": 0.0002,
"loss": 0.4948,
"step": 2330
},
{
"epoch": 0.95,
"learning_rate": 0.0002,
"loss": 0.5601,
"step": 2340
},
{
"epoch": 0.95,
"learning_rate": 0.0002,
"loss": 0.5611,
"step": 2350
},
{
"epoch": 0.96,
"learning_rate": 0.0002,
"loss": 0.5769,
"step": 2360
},
{
"epoch": 0.96,
"learning_rate": 0.0002,
"loss": 0.4425,
"step": 2370
},
{
"epoch": 0.96,
"learning_rate": 0.0002,
"loss": 0.5148,
"step": 2380
},
{
"epoch": 0.97,
"learning_rate": 0.0002,
"loss": 0.5422,
"step": 2390
},
{
"epoch": 0.97,
"learning_rate": 0.0002,
"loss": 0.5161,
"step": 2400
},
{
"epoch": 0.97,
"eval_loss": 0.6037020683288574,
"eval_runtime": 94.2295,
"eval_samples_per_second": 10.612,
"eval_steps_per_second": 5.306,
"step": 2400
},
{
"epoch": 0.97,
"mmlu_eval_accuracy": 0.4454679892995579,
"mmlu_eval_accuracy_abstract_algebra": 0.18181818181818182,
"mmlu_eval_accuracy_anatomy": 0.5714285714285714,
"mmlu_eval_accuracy_astronomy": 0.5625,
"mmlu_eval_accuracy_business_ethics": 0.7272727272727273,
"mmlu_eval_accuracy_clinical_knowledge": 0.3103448275862069,
"mmlu_eval_accuracy_college_biology": 0.25,
"mmlu_eval_accuracy_college_chemistry": 0.25,
"mmlu_eval_accuracy_college_computer_science": 0.5454545454545454,
"mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
"mmlu_eval_accuracy_college_medicine": 0.2727272727272727,
"mmlu_eval_accuracy_college_physics": 0.36363636363636365,
"mmlu_eval_accuracy_computer_security": 0.18181818181818182,
"mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
"mmlu_eval_accuracy_econometrics": 0.16666666666666666,
"mmlu_eval_accuracy_electrical_engineering": 0.5,
"mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
"mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
"mmlu_eval_accuracy_global_facts": 0.2,
"mmlu_eval_accuracy_high_school_biology": 0.28125,
"mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
"mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
"mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_geography": 0.6363636363636364,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.32558139534883723,
"mmlu_eval_accuracy_high_school_mathematics": 0.3448275862068966,
"mmlu_eval_accuracy_high_school_microeconomics": 0.34615384615384615,
"mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
"mmlu_eval_accuracy_high_school_psychology": 0.65,
"mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
"mmlu_eval_accuracy_high_school_us_history": 0.5454545454545454,
"mmlu_eval_accuracy_high_school_world_history": 0.6153846153846154,
"mmlu_eval_accuracy_human_aging": 0.6086956521739131,
"mmlu_eval_accuracy_human_sexuality": 0.5,
"mmlu_eval_accuracy_international_law": 0.6923076923076923,
"mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
"mmlu_eval_accuracy_logical_fallacies": 0.5,
"mmlu_eval_accuracy_machine_learning": 0.09090909090909091,
"mmlu_eval_accuracy_management": 0.45454545454545453,
"mmlu_eval_accuracy_marketing": 0.72,
"mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
"mmlu_eval_accuracy_miscellaneous": 0.6395348837209303,
"mmlu_eval_accuracy_moral_disputes": 0.5,
"mmlu_eval_accuracy_moral_scenarios": 0.24,
"mmlu_eval_accuracy_nutrition": 0.5151515151515151,
"mmlu_eval_accuracy_philosophy": 0.5588235294117647,
"mmlu_eval_accuracy_prehistory": 0.2857142857142857,
"mmlu_eval_accuracy_professional_accounting": 0.3870967741935484,
"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": 0.37037037037037035,
"mmlu_eval_accuracy_sociology": 0.5454545454545454,
"mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
"mmlu_eval_accuracy_virology": 0.4444444444444444,
"mmlu_eval_accuracy_world_religions": 0.8421052631578947,
"mmlu_loss": 0.8906538285723554,
"step": 2400
},
{
"epoch": 0.98,
"learning_rate": 0.0002,
"loss": 0.5139,
"step": 2410
},
{
"epoch": 0.98,
"learning_rate": 0.0002,
"loss": 0.5291,
"step": 2420
},
{
"epoch": 0.98,
"learning_rate": 0.0002,
"loss": 0.6079,
"step": 2430
},
{
"epoch": 0.99,
"learning_rate": 0.0002,
"loss": 0.5692,
"step": 2440
},
{
"epoch": 0.99,
"learning_rate": 0.0002,
"loss": 0.6136,
"step": 2450
},
{
"epoch": 1.0,
"learning_rate": 0.0002,
"loss": 0.5858,
"step": 2460
},
{
"epoch": 1.0,
"learning_rate": 0.0002,
"loss": 0.4679,
"step": 2470
},
{
"epoch": 1.0,
"learning_rate": 0.0002,
"loss": 0.5018,
"step": 2480
},
{
"epoch": 1.01,
"learning_rate": 0.0002,
"loss": 0.551,
"step": 2490
},
{
"epoch": 1.01,
"learning_rate": 0.0002,
"loss": 0.528,
"step": 2500
},
{
"epoch": 1.02,
"learning_rate": 0.0002,
"loss": 0.4489,
"step": 2510
},
{
"epoch": 1.02,
"learning_rate": 0.0002,
"loss": 0.4718,
"step": 2520
},
{
"epoch": 1.02,
"learning_rate": 0.0002,
"loss": 0.4079,
"step": 2530
},
{
"epoch": 1.03,
"learning_rate": 0.0002,
"loss": 0.4827,
"step": 2540
},
{
"epoch": 1.03,
"learning_rate": 0.0002,
"loss": 0.5017,
"step": 2550
},
{
"epoch": 1.04,
"learning_rate": 0.0002,
"loss": 0.4425,
"step": 2560
},
{
"epoch": 1.04,
"learning_rate": 0.0002,
"loss": 0.4271,
"step": 2570
},
{
"epoch": 1.04,
"learning_rate": 0.0002,
"loss": 0.5164,
"step": 2580
},
{
"epoch": 1.05,
"learning_rate": 0.0002,
"loss": 0.3981,
"step": 2590
},
{
"epoch": 1.05,
"learning_rate": 0.0002,
"loss": 0.645,
"step": 2600
},
{
"epoch": 1.05,
"eval_loss": 0.6178489327430725,
"eval_runtime": 94.1423,
"eval_samples_per_second": 10.622,
"eval_steps_per_second": 5.311,
"step": 2600
},
{
"epoch": 1.05,
"mmlu_eval_accuracy": 0.461544966521083,
"mmlu_eval_accuracy_abstract_algebra": 0.18181818181818182,
"mmlu_eval_accuracy_anatomy": 0.5714285714285714,
"mmlu_eval_accuracy_astronomy": 0.4375,
"mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
"mmlu_eval_accuracy_clinical_knowledge": 0.3793103448275862,
"mmlu_eval_accuracy_college_biology": 0.3125,
"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": 0.45454545454545453,
"mmlu_eval_accuracy_conceptual_physics": 0.34615384615384615,
"mmlu_eval_accuracy_econometrics": 0.16666666666666666,
"mmlu_eval_accuracy_electrical_engineering": 0.4375,
"mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
"mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
"mmlu_eval_accuracy_global_facts": 0.3,
"mmlu_eval_accuracy_high_school_biology": 0.3125,
"mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
"mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_european_history": 0.5,
"mmlu_eval_accuracy_high_school_geography": 0.6363636363636364,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.37209302325581395,
"mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
"mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
"mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
"mmlu_eval_accuracy_high_school_psychology": 0.8,
"mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
"mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
"mmlu_eval_accuracy_high_school_world_history": 0.6153846153846154,
"mmlu_eval_accuracy_human_aging": 0.6521739130434783,
"mmlu_eval_accuracy_human_sexuality": 0.5,
"mmlu_eval_accuracy_international_law": 0.6923076923076923,
"mmlu_eval_accuracy_jurisprudence": 0.5454545454545454,
"mmlu_eval_accuracy_logical_fallacies": 0.5,
"mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
"mmlu_eval_accuracy_management": 0.45454545454545453,
"mmlu_eval_accuracy_marketing": 0.76,
"mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
"mmlu_eval_accuracy_miscellaneous": 0.6976744186046512,
"mmlu_eval_accuracy_moral_disputes": 0.39473684210526316,
"mmlu_eval_accuracy_moral_scenarios": 0.23,
"mmlu_eval_accuracy_nutrition": 0.5454545454545454,
"mmlu_eval_accuracy_philosophy": 0.5882352941176471,
"mmlu_eval_accuracy_prehistory": 0.37142857142857144,
"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": 0.391304347826087,
"mmlu_eval_accuracy_public_relations": 0.5,
"mmlu_eval_accuracy_security_studies": 0.48148148148148145,
"mmlu_eval_accuracy_sociology": 0.6818181818181818,
"mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
"mmlu_eval_accuracy_virology": 0.5555555555555556,
"mmlu_eval_accuracy_world_religions": 0.7368421052631579,
"mmlu_loss": 0.9348930491732711,
"step": 2600
}
],
"max_steps": 5000,
"num_train_epochs": 3,
"total_flos": 2.1956158906232832e+17,
"trial_name": null,
"trial_params": null
}