prateeky2806's picture
Training in progress, step 1400
e23ffbf
{
"best_metric": 0.5195611119270325,
"best_model_checkpoint": "./output_v2/7b_cluster030_Nous-Hermes-llama-2-7b_partitioned_v3_standardized_030/checkpoint-1200",
"epoch": 0.548213411649535,
"global_step": 1400,
"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": 0.9902,
"step": 10
},
{
"epoch": 0.01,
"learning_rate": 0.0002,
"loss": 0.7549,
"step": 20
},
{
"epoch": 0.01,
"learning_rate": 0.0002,
"loss": 0.8421,
"step": 30
},
{
"epoch": 0.02,
"learning_rate": 0.0002,
"loss": 0.6897,
"step": 40
},
{
"epoch": 0.02,
"learning_rate": 0.0002,
"loss": 0.8507,
"step": 50
},
{
"epoch": 0.02,
"learning_rate": 0.0002,
"loss": 0.6511,
"step": 60
},
{
"epoch": 0.03,
"learning_rate": 0.0002,
"loss": 0.6798,
"step": 70
},
{
"epoch": 0.03,
"learning_rate": 0.0002,
"loss": 0.7609,
"step": 80
},
{
"epoch": 0.04,
"learning_rate": 0.0002,
"loss": 0.7702,
"step": 90
},
{
"epoch": 0.04,
"learning_rate": 0.0002,
"loss": 0.6088,
"step": 100
},
{
"epoch": 0.04,
"learning_rate": 0.0002,
"loss": 0.694,
"step": 110
},
{
"epoch": 0.05,
"learning_rate": 0.0002,
"loss": 0.6922,
"step": 120
},
{
"epoch": 0.05,
"learning_rate": 0.0002,
"loss": 0.6326,
"step": 130
},
{
"epoch": 0.05,
"learning_rate": 0.0002,
"loss": 0.4704,
"step": 140
},
{
"epoch": 0.06,
"learning_rate": 0.0002,
"loss": 0.6479,
"step": 150
},
{
"epoch": 0.06,
"learning_rate": 0.0002,
"loss": 0.6442,
"step": 160
},
{
"epoch": 0.07,
"learning_rate": 0.0002,
"loss": 0.5064,
"step": 170
},
{
"epoch": 0.07,
"learning_rate": 0.0002,
"loss": 0.5357,
"step": 180
},
{
"epoch": 0.07,
"learning_rate": 0.0002,
"loss": 0.671,
"step": 190
},
{
"epoch": 0.08,
"learning_rate": 0.0002,
"loss": 0.7203,
"step": 200
},
{
"epoch": 0.08,
"eval_loss": 0.5725088715553284,
"eval_runtime": 110.5239,
"eval_samples_per_second": 9.048,
"eval_steps_per_second": 4.524,
"step": 200
},
{
"epoch": 0.08,
"mmlu_eval_accuracy": 0.4726934353480768,
"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": 0.4827586206896552,
"mmlu_eval_accuracy_college_biology": 0.5,
"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": 0.36363636363636365,
"mmlu_eval_accuracy_college_physics": 0.45454545454545453,
"mmlu_eval_accuracy_computer_security": 0.6363636363636364,
"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": 0.2857142857142857,
"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": 0.5,
"mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.7142857142857143,
"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": 0.29411764705882354,
"mmlu_eval_accuracy_high_school_psychology": 0.7666666666666667,
"mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
"mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
"mmlu_eval_accuracy_high_school_world_history": 0.4230769230769231,
"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": 0.45454545454545453,
"mmlu_eval_accuracy_marketing": 0.64,
"mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
"mmlu_eval_accuracy_miscellaneous": 0.6046511627906976,
"mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
"mmlu_eval_accuracy_moral_scenarios": 0.31,
"mmlu_eval_accuracy_nutrition": 0.5454545454545454,
"mmlu_eval_accuracy_philosophy": 0.4411764705882353,
"mmlu_eval_accuracy_prehistory": 0.5142857142857142,
"mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
"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": 0.5185185185185185,
"mmlu_eval_accuracy_sociology": 0.6818181818181818,
"mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
"mmlu_eval_accuracy_virology": 0.3888888888888889,
"mmlu_eval_accuracy_world_religions": 0.7368421052631579,
"mmlu_loss": 0.9052433924598732,
"step": 200
},
{
"epoch": 0.08,
"learning_rate": 0.0002,
"loss": 0.6101,
"step": 210
},
{
"epoch": 0.09,
"learning_rate": 0.0002,
"loss": 0.481,
"step": 220
},
{
"epoch": 0.09,
"learning_rate": 0.0002,
"loss": 0.5222,
"step": 230
},
{
"epoch": 0.09,
"learning_rate": 0.0002,
"loss": 0.5102,
"step": 240
},
{
"epoch": 0.1,
"learning_rate": 0.0002,
"loss": 0.6642,
"step": 250
},
{
"epoch": 0.1,
"learning_rate": 0.0002,
"loss": 0.5723,
"step": 260
},
{
"epoch": 0.11,
"learning_rate": 0.0002,
"loss": 0.4273,
"step": 270
},
{
"epoch": 0.11,
"learning_rate": 0.0002,
"loss": 0.6994,
"step": 280
},
{
"epoch": 0.11,
"learning_rate": 0.0002,
"loss": 0.7166,
"step": 290
},
{
"epoch": 0.12,
"learning_rate": 0.0002,
"loss": 0.6693,
"step": 300
},
{
"epoch": 0.12,
"learning_rate": 0.0002,
"loss": 0.5636,
"step": 310
},
{
"epoch": 0.13,
"learning_rate": 0.0002,
"loss": 0.6241,
"step": 320
},
{
"epoch": 0.13,
"learning_rate": 0.0002,
"loss": 0.5453,
"step": 330
},
{
"epoch": 0.13,
"learning_rate": 0.0002,
"loss": 0.6589,
"step": 340
},
{
"epoch": 0.14,
"learning_rate": 0.0002,
"loss": 0.6073,
"step": 350
},
{
"epoch": 0.14,
"learning_rate": 0.0002,
"loss": 0.5931,
"step": 360
},
{
"epoch": 0.14,
"learning_rate": 0.0002,
"loss": 0.5405,
"step": 370
},
{
"epoch": 0.15,
"learning_rate": 0.0002,
"loss": 0.6522,
"step": 380
},
{
"epoch": 0.15,
"learning_rate": 0.0002,
"loss": 0.672,
"step": 390
},
{
"epoch": 0.16,
"learning_rate": 0.0002,
"loss": 0.5791,
"step": 400
},
{
"epoch": 0.16,
"eval_loss": 0.5623835921287537,
"eval_runtime": 111.0199,
"eval_samples_per_second": 9.007,
"eval_steps_per_second": 4.504,
"step": 400
},
{
"epoch": 0.16,
"mmlu_eval_accuracy": 0.4759225748253283,
"mmlu_eval_accuracy_abstract_algebra": 0.18181818181818182,
"mmlu_eval_accuracy_anatomy": 0.6428571428571429,
"mmlu_eval_accuracy_astronomy": 0.4375,
"mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
"mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
"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": 0.3181818181818182,
"mmlu_eval_accuracy_college_physics": 0.45454545454545453,
"mmlu_eval_accuracy_computer_security": 0.2727272727272727,
"mmlu_eval_accuracy_conceptual_physics": 0.5384615384615384,
"mmlu_eval_accuracy_econometrics": 0.16666666666666666,
"mmlu_eval_accuracy_electrical_engineering": 0.5,
"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": 0.34375,
"mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
"mmlu_eval_accuracy_high_school_computer_science": 0.7777777777777778,
"mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
"mmlu_eval_accuracy_high_school_geography": 0.7272727272727273,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.3488372093023256,
"mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
"mmlu_eval_accuracy_high_school_microeconomics": 0.38461538461538464,
"mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
"mmlu_eval_accuracy_high_school_psychology": 0.7,
"mmlu_eval_accuracy_high_school_statistics": 0.21739130434782608,
"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": 0.7692307692307693,
"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": 0.72,
"mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
"mmlu_eval_accuracy_miscellaneous": 0.686046511627907,
"mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
"mmlu_eval_accuracy_moral_scenarios": 0.23,
"mmlu_eval_accuracy_nutrition": 0.6666666666666666,
"mmlu_eval_accuracy_philosophy": 0.47058823529411764,
"mmlu_eval_accuracy_prehistory": 0.37142857142857144,
"mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
"mmlu_eval_accuracy_professional_law": 0.3352941176470588,
"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": 0.48148148148148145,
"mmlu_eval_accuracy_sociology": 0.7727272727272727,
"mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
"mmlu_eval_accuracy_virology": 0.3888888888888889,
"mmlu_eval_accuracy_world_religions": 0.7368421052631579,
"mmlu_loss": 0.7658495643363376,
"step": 400
},
{
"epoch": 0.16,
"learning_rate": 0.0002,
"loss": 0.6453,
"step": 410
},
{
"epoch": 0.16,
"learning_rate": 0.0002,
"loss": 0.6714,
"step": 420
},
{
"epoch": 0.17,
"learning_rate": 0.0002,
"loss": 0.6399,
"step": 430
},
{
"epoch": 0.17,
"learning_rate": 0.0002,
"loss": 0.481,
"step": 440
},
{
"epoch": 0.18,
"learning_rate": 0.0002,
"loss": 0.5834,
"step": 450
},
{
"epoch": 0.18,
"learning_rate": 0.0002,
"loss": 0.6174,
"step": 460
},
{
"epoch": 0.18,
"learning_rate": 0.0002,
"loss": 0.6331,
"step": 470
},
{
"epoch": 0.19,
"learning_rate": 0.0002,
"loss": 0.5989,
"step": 480
},
{
"epoch": 0.19,
"learning_rate": 0.0002,
"loss": 0.6293,
"step": 490
},
{
"epoch": 0.2,
"learning_rate": 0.0002,
"loss": 0.4968,
"step": 500
},
{
"epoch": 0.2,
"learning_rate": 0.0002,
"loss": 0.609,
"step": 510
},
{
"epoch": 0.2,
"learning_rate": 0.0002,
"loss": 0.52,
"step": 520
},
{
"epoch": 0.21,
"learning_rate": 0.0002,
"loss": 0.5791,
"step": 530
},
{
"epoch": 0.21,
"learning_rate": 0.0002,
"loss": 0.5389,
"step": 540
},
{
"epoch": 0.22,
"learning_rate": 0.0002,
"loss": 0.5664,
"step": 550
},
{
"epoch": 0.22,
"learning_rate": 0.0002,
"loss": 0.5939,
"step": 560
},
{
"epoch": 0.22,
"learning_rate": 0.0002,
"loss": 0.5111,
"step": 570
},
{
"epoch": 0.23,
"learning_rate": 0.0002,
"loss": 0.5947,
"step": 580
},
{
"epoch": 0.23,
"learning_rate": 0.0002,
"loss": 0.6401,
"step": 590
},
{
"epoch": 0.23,
"learning_rate": 0.0002,
"loss": 0.5482,
"step": 600
},
{
"epoch": 0.23,
"eval_loss": 0.5499725341796875,
"eval_runtime": 111.3864,
"eval_samples_per_second": 8.978,
"eval_steps_per_second": 4.489,
"step": 600
},
{
"epoch": 0.23,
"mmlu_eval_accuracy": 0.4639735323699154,
"mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
"mmlu_eval_accuracy_anatomy": 0.5714285714285714,
"mmlu_eval_accuracy_astronomy": 0.4375,
"mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
"mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
"mmlu_eval_accuracy_college_biology": 0.5,
"mmlu_eval_accuracy_college_chemistry": 0.125,
"mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
"mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
"mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
"mmlu_eval_accuracy_college_physics": 0.36363636363636365,
"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": 0.375,
"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": 0.22727272727272727,
"mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_european_history": 0.4444444444444444,
"mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
"mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
"mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
"mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
"mmlu_eval_accuracy_high_school_psychology": 0.7333333333333333,
"mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
"mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
"mmlu_eval_accuracy_high_school_world_history": 0.46153846153846156,
"mmlu_eval_accuracy_human_aging": 0.7391304347826086,
"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": 0.2727272727272727,
"mmlu_eval_accuracy_management": 0.5454545454545454,
"mmlu_eval_accuracy_marketing": 0.68,
"mmlu_eval_accuracy_medical_genetics": 0.6363636363636364,
"mmlu_eval_accuracy_miscellaneous": 0.627906976744186,
"mmlu_eval_accuracy_moral_disputes": 0.5789473684210527,
"mmlu_eval_accuracy_moral_scenarios": 0.32,
"mmlu_eval_accuracy_nutrition": 0.6060606060606061,
"mmlu_eval_accuracy_philosophy": 0.38235294117647056,
"mmlu_eval_accuracy_prehistory": 0.5142857142857142,
"mmlu_eval_accuracy_professional_accounting": 0.22580645161290322,
"mmlu_eval_accuracy_professional_law": 0.36470588235294116,
"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": 0.5185185185185185,
"mmlu_eval_accuracy_sociology": 0.6363636363636364,
"mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
"mmlu_eval_accuracy_virology": 0.4444444444444444,
"mmlu_eval_accuracy_world_religions": 0.7368421052631579,
"mmlu_loss": 0.8004742157074085,
"step": 600
},
{
"epoch": 0.24,
"learning_rate": 0.0002,
"loss": 0.5413,
"step": 610
},
{
"epoch": 0.24,
"learning_rate": 0.0002,
"loss": 0.5972,
"step": 620
},
{
"epoch": 0.25,
"learning_rate": 0.0002,
"loss": 0.601,
"step": 630
},
{
"epoch": 0.25,
"learning_rate": 0.0002,
"loss": 0.6284,
"step": 640
},
{
"epoch": 0.25,
"learning_rate": 0.0002,
"loss": 0.6391,
"step": 650
},
{
"epoch": 0.26,
"learning_rate": 0.0002,
"loss": 0.7461,
"step": 660
},
{
"epoch": 0.26,
"learning_rate": 0.0002,
"loss": 0.5584,
"step": 670
},
{
"epoch": 0.27,
"learning_rate": 0.0002,
"loss": 0.5523,
"step": 680
},
{
"epoch": 0.27,
"learning_rate": 0.0002,
"loss": 0.6266,
"step": 690
},
{
"epoch": 0.27,
"learning_rate": 0.0002,
"loss": 0.6217,
"step": 700
},
{
"epoch": 0.28,
"learning_rate": 0.0002,
"loss": 0.5673,
"step": 710
},
{
"epoch": 0.28,
"learning_rate": 0.0002,
"loss": 0.608,
"step": 720
},
{
"epoch": 0.29,
"learning_rate": 0.0002,
"loss": 0.6208,
"step": 730
},
{
"epoch": 0.29,
"learning_rate": 0.0002,
"loss": 0.5609,
"step": 740
},
{
"epoch": 0.29,
"learning_rate": 0.0002,
"loss": 0.4951,
"step": 750
},
{
"epoch": 0.3,
"learning_rate": 0.0002,
"loss": 0.5513,
"step": 760
},
{
"epoch": 0.3,
"learning_rate": 0.0002,
"loss": 0.6386,
"step": 770
},
{
"epoch": 0.31,
"learning_rate": 0.0002,
"loss": 0.4915,
"step": 780
},
{
"epoch": 0.31,
"learning_rate": 0.0002,
"loss": 0.6105,
"step": 790
},
{
"epoch": 0.31,
"learning_rate": 0.0002,
"loss": 0.5794,
"step": 800
},
{
"epoch": 0.31,
"eval_loss": 0.5391483306884766,
"eval_runtime": 111.9656,
"eval_samples_per_second": 8.931,
"eval_steps_per_second": 4.466,
"step": 800
},
{
"epoch": 0.31,
"mmlu_eval_accuracy": 0.4865307699515832,
"mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
"mmlu_eval_accuracy_anatomy": 0.6428571428571429,
"mmlu_eval_accuracy_astronomy": 0.375,
"mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
"mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
"mmlu_eval_accuracy_college_biology": 0.5,
"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": 0.5,
"mmlu_eval_accuracy_college_physics": 0.5454545454545454,
"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": 0.2926829268292683,
"mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
"mmlu_eval_accuracy_global_facts": 0.5,
"mmlu_eval_accuracy_high_school_biology": 0.4375,
"mmlu_eval_accuracy_high_school_chemistry": 0.22727272727272727,
"mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_european_history": 0.5,
"mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
"mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
"mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
"mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
"mmlu_eval_accuracy_high_school_psychology": 0.7833333333333333,
"mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
"mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
"mmlu_eval_accuracy_high_school_world_history": 0.5384615384615384,
"mmlu_eval_accuracy_human_aging": 0.6086956521739131,
"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": 0.5555555555555556,
"mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
"mmlu_eval_accuracy_management": 0.7272727272727273,
"mmlu_eval_accuracy_marketing": 0.68,
"mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
"mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
"mmlu_eval_accuracy_moral_disputes": 0.6052631578947368,
"mmlu_eval_accuracy_moral_scenarios": 0.26,
"mmlu_eval_accuracy_nutrition": 0.6060606060606061,
"mmlu_eval_accuracy_philosophy": 0.4411764705882353,
"mmlu_eval_accuracy_prehistory": 0.4,
"mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
"mmlu_eval_accuracy_professional_law": 0.34705882352941175,
"mmlu_eval_accuracy_professional_medicine": 0.41935483870967744,
"mmlu_eval_accuracy_professional_psychology": 0.43478260869565216,
"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": 0.7272727272727273,
"mmlu_eval_accuracy_virology": 0.3333333333333333,
"mmlu_eval_accuracy_world_religions": 0.7368421052631579,
"mmlu_loss": 0.734274251542149,
"step": 800
},
{
"epoch": 0.32,
"learning_rate": 0.0002,
"loss": 0.5351,
"step": 810
},
{
"epoch": 0.32,
"learning_rate": 0.0002,
"loss": 0.7598,
"step": 820
},
{
"epoch": 0.33,
"learning_rate": 0.0002,
"loss": 0.5416,
"step": 830
},
{
"epoch": 0.33,
"learning_rate": 0.0002,
"loss": 0.5895,
"step": 840
},
{
"epoch": 0.33,
"learning_rate": 0.0002,
"loss": 0.6436,
"step": 850
},
{
"epoch": 0.34,
"learning_rate": 0.0002,
"loss": 0.714,
"step": 860
},
{
"epoch": 0.34,
"learning_rate": 0.0002,
"loss": 0.7195,
"step": 870
},
{
"epoch": 0.34,
"learning_rate": 0.0002,
"loss": 0.5647,
"step": 880
},
{
"epoch": 0.35,
"learning_rate": 0.0002,
"loss": 0.5616,
"step": 890
},
{
"epoch": 0.35,
"learning_rate": 0.0002,
"loss": 0.7585,
"step": 900
},
{
"epoch": 0.36,
"learning_rate": 0.0002,
"loss": 0.6748,
"step": 910
},
{
"epoch": 0.36,
"learning_rate": 0.0002,
"loss": 0.622,
"step": 920
},
{
"epoch": 0.36,
"learning_rate": 0.0002,
"loss": 0.5912,
"step": 930
},
{
"epoch": 0.37,
"learning_rate": 0.0002,
"loss": 0.5217,
"step": 940
},
{
"epoch": 0.37,
"learning_rate": 0.0002,
"loss": 0.5993,
"step": 950
},
{
"epoch": 0.38,
"learning_rate": 0.0002,
"loss": 0.5852,
"step": 960
},
{
"epoch": 0.38,
"learning_rate": 0.0002,
"loss": 0.586,
"step": 970
},
{
"epoch": 0.38,
"learning_rate": 0.0002,
"loss": 0.5978,
"step": 980
},
{
"epoch": 0.39,
"learning_rate": 0.0002,
"loss": 0.6072,
"step": 990
},
{
"epoch": 0.39,
"learning_rate": 0.0002,
"loss": 0.7357,
"step": 1000
},
{
"epoch": 0.39,
"eval_loss": 0.540454089641571,
"eval_runtime": 111.6241,
"eval_samples_per_second": 8.959,
"eval_steps_per_second": 4.479,
"step": 1000
},
{
"epoch": 0.39,
"mmlu_eval_accuracy": 0.46609677068605504,
"mmlu_eval_accuracy_abstract_algebra": 0.18181818181818182,
"mmlu_eval_accuracy_anatomy": 0.5,
"mmlu_eval_accuracy_astronomy": 0.4375,
"mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
"mmlu_eval_accuracy_clinical_knowledge": 0.3793103448275862,
"mmlu_eval_accuracy_college_biology": 0.25,
"mmlu_eval_accuracy_college_chemistry": 0.375,
"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": 0.36363636363636365,
"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": 0.375,
"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": 0.375,
"mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
"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": 0.6190476190476191,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.32558139534883723,
"mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
"mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
"mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
"mmlu_eval_accuracy_high_school_psychology": 0.7,
"mmlu_eval_accuracy_high_school_statistics": 0.2608695652173913,
"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": 0.5,
"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": 0.2727272727272727,
"mmlu_eval_accuracy_management": 0.7272727272727273,
"mmlu_eval_accuracy_marketing": 0.72,
"mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
"mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
"mmlu_eval_accuracy_moral_disputes": 0.39473684210526316,
"mmlu_eval_accuracy_moral_scenarios": 0.23,
"mmlu_eval_accuracy_nutrition": 0.6363636363636364,
"mmlu_eval_accuracy_philosophy": 0.47058823529411764,
"mmlu_eval_accuracy_prehistory": 0.34285714285714286,
"mmlu_eval_accuracy_professional_accounting": 0.41935483870967744,
"mmlu_eval_accuracy_professional_law": 0.31176470588235294,
"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": 0.5555555555555556,
"mmlu_eval_accuracy_sociology": 0.7272727272727273,
"mmlu_eval_accuracy_us_foreign_policy": 0.45454545454545453,
"mmlu_eval_accuracy_virology": 0.5,
"mmlu_eval_accuracy_world_religions": 0.7368421052631579,
"mmlu_loss": 0.7961046382707031,
"step": 1000
},
{
"epoch": 0.4,
"learning_rate": 0.0002,
"loss": 0.6952,
"step": 1010
},
{
"epoch": 0.4,
"learning_rate": 0.0002,
"loss": 0.5948,
"step": 1020
},
{
"epoch": 0.4,
"learning_rate": 0.0002,
"loss": 0.6301,
"step": 1030
},
{
"epoch": 0.41,
"learning_rate": 0.0002,
"loss": 0.6787,
"step": 1040
},
{
"epoch": 0.41,
"learning_rate": 0.0002,
"loss": 0.5183,
"step": 1050
},
{
"epoch": 0.42,
"learning_rate": 0.0002,
"loss": 0.7002,
"step": 1060
},
{
"epoch": 0.42,
"learning_rate": 0.0002,
"loss": 0.6534,
"step": 1070
},
{
"epoch": 0.42,
"learning_rate": 0.0002,
"loss": 0.5553,
"step": 1080
},
{
"epoch": 0.43,
"learning_rate": 0.0002,
"loss": 0.499,
"step": 1090
},
{
"epoch": 0.43,
"learning_rate": 0.0002,
"loss": 0.6952,
"step": 1100
},
{
"epoch": 0.43,
"learning_rate": 0.0002,
"loss": 0.5279,
"step": 1110
},
{
"epoch": 0.44,
"learning_rate": 0.0002,
"loss": 0.6835,
"step": 1120
},
{
"epoch": 0.44,
"learning_rate": 0.0002,
"loss": 0.5202,
"step": 1130
},
{
"epoch": 0.45,
"learning_rate": 0.0002,
"loss": 0.5252,
"step": 1140
},
{
"epoch": 0.45,
"learning_rate": 0.0002,
"loss": 0.5192,
"step": 1150
},
{
"epoch": 0.45,
"learning_rate": 0.0002,
"loss": 0.5952,
"step": 1160
},
{
"epoch": 0.46,
"learning_rate": 0.0002,
"loss": 0.5739,
"step": 1170
},
{
"epoch": 0.46,
"learning_rate": 0.0002,
"loss": 0.6092,
"step": 1180
},
{
"epoch": 0.47,
"learning_rate": 0.0002,
"loss": 0.7525,
"step": 1190
},
{
"epoch": 0.47,
"learning_rate": 0.0002,
"loss": 0.6786,
"step": 1200
},
{
"epoch": 0.47,
"eval_loss": 0.5195611119270325,
"eval_runtime": 111.24,
"eval_samples_per_second": 8.99,
"eval_steps_per_second": 4.495,
"step": 1200
},
{
"epoch": 0.47,
"mmlu_eval_accuracy": 0.4533161015344259,
"mmlu_eval_accuracy_abstract_algebra": 0.18181818181818182,
"mmlu_eval_accuracy_anatomy": 0.5714285714285714,
"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": 0.4375,
"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": 0.5454545454545454,
"mmlu_eval_accuracy_computer_security": 0.2727272727272727,
"mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
"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": 0.4,
"mmlu_eval_accuracy_high_school_biology": 0.375,
"mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
"mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
"mmlu_eval_accuracy_high_school_geography": 0.7272727272727273,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.38095238095238093,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.3023255813953488,
"mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
"mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
"mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
"mmlu_eval_accuracy_high_school_psychology": 0.7166666666666667,
"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": 0.7692307692307693,
"mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
"mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
"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": 0.6395348837209303,
"mmlu_eval_accuracy_moral_disputes": 0.5,
"mmlu_eval_accuracy_moral_scenarios": 0.26,
"mmlu_eval_accuracy_nutrition": 0.5454545454545454,
"mmlu_eval_accuracy_philosophy": 0.4411764705882353,
"mmlu_eval_accuracy_prehistory": 0.42857142857142855,
"mmlu_eval_accuracy_professional_accounting": 0.22580645161290322,
"mmlu_eval_accuracy_professional_law": 0.3588235294117647,
"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": 0.45454545454545453,
"mmlu_eval_accuracy_virology": 0.5555555555555556,
"mmlu_eval_accuracy_world_religions": 0.7368421052631579,
"mmlu_loss": 0.7577510231390946,
"step": 1200
},
{
"epoch": 0.47,
"learning_rate": 0.0002,
"loss": 0.676,
"step": 1210
},
{
"epoch": 0.48,
"learning_rate": 0.0002,
"loss": 0.5373,
"step": 1220
},
{
"epoch": 0.48,
"learning_rate": 0.0002,
"loss": 0.6367,
"step": 1230
},
{
"epoch": 0.49,
"learning_rate": 0.0002,
"loss": 0.5926,
"step": 1240
},
{
"epoch": 0.49,
"learning_rate": 0.0002,
"loss": 0.5637,
"step": 1250
},
{
"epoch": 0.49,
"learning_rate": 0.0002,
"loss": 0.6318,
"step": 1260
},
{
"epoch": 0.5,
"learning_rate": 0.0002,
"loss": 0.6401,
"step": 1270
},
{
"epoch": 0.5,
"learning_rate": 0.0002,
"loss": 0.56,
"step": 1280
},
{
"epoch": 0.51,
"learning_rate": 0.0002,
"loss": 0.5247,
"step": 1290
},
{
"epoch": 0.51,
"learning_rate": 0.0002,
"loss": 0.6248,
"step": 1300
},
{
"epoch": 0.51,
"learning_rate": 0.0002,
"loss": 0.551,
"step": 1310
},
{
"epoch": 0.52,
"learning_rate": 0.0002,
"loss": 0.5318,
"step": 1320
},
{
"epoch": 0.52,
"learning_rate": 0.0002,
"loss": 0.4965,
"step": 1330
},
{
"epoch": 0.52,
"learning_rate": 0.0002,
"loss": 0.6144,
"step": 1340
},
{
"epoch": 0.53,
"learning_rate": 0.0002,
"loss": 0.5746,
"step": 1350
},
{
"epoch": 0.53,
"learning_rate": 0.0002,
"loss": 0.5896,
"step": 1360
},
{
"epoch": 0.54,
"learning_rate": 0.0002,
"loss": 0.5302,
"step": 1370
},
{
"epoch": 0.54,
"learning_rate": 0.0002,
"loss": 0.5363,
"step": 1380
},
{
"epoch": 0.54,
"learning_rate": 0.0002,
"loss": 0.4933,
"step": 1390
},
{
"epoch": 0.55,
"learning_rate": 0.0002,
"loss": 0.4435,
"step": 1400
},
{
"epoch": 0.55,
"eval_loss": 0.5208213925361633,
"eval_runtime": 112.8123,
"eval_samples_per_second": 8.864,
"eval_steps_per_second": 4.432,
"step": 1400
},
{
"epoch": 0.55,
"mmlu_eval_accuracy": 0.47006780054926284,
"mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
"mmlu_eval_accuracy_anatomy": 0.5,
"mmlu_eval_accuracy_astronomy": 0.5625,
"mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
"mmlu_eval_accuracy_clinical_knowledge": 0.4482758620689655,
"mmlu_eval_accuracy_college_biology": 0.5,
"mmlu_eval_accuracy_college_chemistry": 0.375,
"mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
"mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
"mmlu_eval_accuracy_college_medicine": 0.5454545454545454,
"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": 0.25,
"mmlu_eval_accuracy_elementary_mathematics": 0.2926829268292683,
"mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
"mmlu_eval_accuracy_global_facts": 0.2,
"mmlu_eval_accuracy_high_school_biology": 0.40625,
"mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
"mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
"mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
"mmlu_eval_accuracy_high_school_government_and_politics": 0.47619047619047616,
"mmlu_eval_accuracy_high_school_macroeconomics": 0.32558139534883723,
"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": 0.7,
"mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
"mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
"mmlu_eval_accuracy_high_school_world_history": 0.46153846153846156,
"mmlu_eval_accuracy_human_aging": 0.782608695652174,
"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": 0.5555555555555556,
"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": 0.6395348837209303,
"mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
"mmlu_eval_accuracy_moral_scenarios": 0.28,
"mmlu_eval_accuracy_nutrition": 0.5757575757575758,
"mmlu_eval_accuracy_philosophy": 0.5,
"mmlu_eval_accuracy_prehistory": 0.37142857142857144,
"mmlu_eval_accuracy_professional_accounting": 0.45161290322580644,
"mmlu_eval_accuracy_professional_law": 0.34705882352941175,
"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": 0.48148148148148145,
"mmlu_eval_accuracy_sociology": 0.5454545454545454,
"mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
"mmlu_eval_accuracy_virology": 0.3888888888888889,
"mmlu_eval_accuracy_world_religions": 0.6842105263157895,
"mmlu_loss": 0.7375706565364534,
"step": 1400
}
],
"max_steps": 5000,
"num_train_epochs": 2,
"total_flos": 1.1029560251970355e+17,
"trial_name": null,
"trial_params": null
}