Farouk commited on
Commit
c1aa80f
Β·
1 Parent(s): 19da3f9

Training in progress, step 10000

Browse files
adapter_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:832fe1bf4c1262a05e4e8abdb9b9fcf332b3d23a12789e0e35934df54e3ebf69
3
  size 319977229
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c1d5bb11c428194d7d9b329c96d41bbc574bca23917af98883315f13db8729c7
3
  size 319977229
{checkpoint-8000 β†’ checkpoint-10000}/README.md RENAMED
File without changes
{checkpoint-8000 β†’ checkpoint-10000}/adapter_config.json RENAMED
File without changes
{checkpoint-8000 β†’ checkpoint-10000}/adapter_model.bin RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:68aa3c70d4e9d5c3f0e1669e4f68215d39e00c2e640196f32ab6e13e5df7aee7
3
  size 319977229
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c1d5bb11c428194d7d9b329c96d41bbc574bca23917af98883315f13db8729c7
3
  size 319977229
{checkpoint-8000 β†’ checkpoint-10000}/added_tokens.json RENAMED
File without changes
{checkpoint-8000 β†’ checkpoint-10000}/optimizer.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8d5e4aab9297e4d1e4cd3558a3127810791170a8e8cfb0ce2478a505a26d656e
3
  size 1279539973
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1cd5e9ad5a14efd7ad5ae67cf54902b95218af89a9f8f83f26e2b2357880a4ad
3
  size 1279539973
{checkpoint-8000 β†’ checkpoint-10000}/rng_state.pth RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c4d1eb8e66f0770ad1ddb569a9718c0f0e77bca5e6d6472818be1e55337941b5
3
  size 14511
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e35d9ca636abe6dc43cfabc438e43d8d32c31b381c811b50067ca58252b7386b
3
  size 14511
{checkpoint-8000 β†’ checkpoint-10000}/scheduler.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ec0a535d2c9c4c62a74336a7f93b6d947a1152f53a6066eccd4123d6b477c15c
3
  size 627
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a4d1f3cad6e54148c289328646571ee0d43f9671c93cc919adafe5bd27a33d05
3
  size 627
{checkpoint-8000 β†’ checkpoint-10000}/special_tokens_map.json RENAMED
File without changes
{checkpoint-8000 β†’ checkpoint-10000}/tokenizer.model RENAMED
File without changes
{checkpoint-8000 β†’ checkpoint-10000}/tokenizer_config.json RENAMED
File without changes
{checkpoint-8000 β†’ checkpoint-10000}/trainer_state.json RENAMED
@@ -1,8 +1,8 @@
1
  {
2
  "best_metric": 0.9177722930908203,
3
  "best_model_checkpoint": "experts/expert-14/checkpoint-5800",
4
- "epoch": 2.4605920799692425,
5
- "global_step": 8000,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
@@ -7646,11 +7646,1921 @@
7646
  "mmlu_eval_accuracy_world_religions": 0.631578947368421,
7647
  "mmlu_loss": 1.2873948419140773,
7648
  "step": 8000
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7649
  }
7650
  ],
7651
  "max_steps": 10000,
7652
  "num_train_epochs": 4,
7653
- "total_flos": 1.265562365336912e+18,
7654
  "trial_name": null,
7655
  "trial_params": null
7656
  }
 
1
  {
2
  "best_metric": 0.9177722930908203,
3
  "best_model_checkpoint": "experts/expert-14/checkpoint-5800",
4
+ "epoch": 3.0757400999615534,
5
+ "global_step": 10000,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
 
7646
  "mmlu_eval_accuracy_world_religions": 0.631578947368421,
7647
  "mmlu_loss": 1.2873948419140773,
7648
  "step": 8000
7649
+ },
7650
+ {
7651
+ "epoch": 2.46,
7652
+ "learning_rate": 0.0002,
7653
+ "loss": 0.7958,
7654
+ "step": 8010
7655
+ },
7656
+ {
7657
+ "epoch": 2.47,
7658
+ "learning_rate": 0.0002,
7659
+ "loss": 0.7313,
7660
+ "step": 8020
7661
+ },
7662
+ {
7663
+ "epoch": 2.47,
7664
+ "learning_rate": 0.0002,
7665
+ "loss": 0.7099,
7666
+ "step": 8030
7667
+ },
7668
+ {
7669
+ "epoch": 2.47,
7670
+ "learning_rate": 0.0002,
7671
+ "loss": 0.7862,
7672
+ "step": 8040
7673
+ },
7674
+ {
7675
+ "epoch": 2.48,
7676
+ "learning_rate": 0.0002,
7677
+ "loss": 0.7337,
7678
+ "step": 8050
7679
+ },
7680
+ {
7681
+ "epoch": 2.48,
7682
+ "learning_rate": 0.0002,
7683
+ "loss": 0.7595,
7684
+ "step": 8060
7685
+ },
7686
+ {
7687
+ "epoch": 2.48,
7688
+ "learning_rate": 0.0002,
7689
+ "loss": 0.754,
7690
+ "step": 8070
7691
+ },
7692
+ {
7693
+ "epoch": 2.49,
7694
+ "learning_rate": 0.0002,
7695
+ "loss": 0.7609,
7696
+ "step": 8080
7697
+ },
7698
+ {
7699
+ "epoch": 2.49,
7700
+ "learning_rate": 0.0002,
7701
+ "loss": 0.7941,
7702
+ "step": 8090
7703
+ },
7704
+ {
7705
+ "epoch": 2.49,
7706
+ "learning_rate": 0.0002,
7707
+ "loss": 0.8151,
7708
+ "step": 8100
7709
+ },
7710
+ {
7711
+ "epoch": 2.49,
7712
+ "learning_rate": 0.0002,
7713
+ "loss": 0.7898,
7714
+ "step": 8110
7715
+ },
7716
+ {
7717
+ "epoch": 2.5,
7718
+ "learning_rate": 0.0002,
7719
+ "loss": 0.7658,
7720
+ "step": 8120
7721
+ },
7722
+ {
7723
+ "epoch": 2.5,
7724
+ "learning_rate": 0.0002,
7725
+ "loss": 0.7499,
7726
+ "step": 8130
7727
+ },
7728
+ {
7729
+ "epoch": 2.5,
7730
+ "learning_rate": 0.0002,
7731
+ "loss": 0.7247,
7732
+ "step": 8140
7733
+ },
7734
+ {
7735
+ "epoch": 2.51,
7736
+ "learning_rate": 0.0002,
7737
+ "loss": 0.7225,
7738
+ "step": 8150
7739
+ },
7740
+ {
7741
+ "epoch": 2.51,
7742
+ "learning_rate": 0.0002,
7743
+ "loss": 0.8,
7744
+ "step": 8160
7745
+ },
7746
+ {
7747
+ "epoch": 2.51,
7748
+ "learning_rate": 0.0002,
7749
+ "loss": 0.7601,
7750
+ "step": 8170
7751
+ },
7752
+ {
7753
+ "epoch": 2.52,
7754
+ "learning_rate": 0.0002,
7755
+ "loss": 0.7922,
7756
+ "step": 8180
7757
+ },
7758
+ {
7759
+ "epoch": 2.52,
7760
+ "learning_rate": 0.0002,
7761
+ "loss": 0.7752,
7762
+ "step": 8190
7763
+ },
7764
+ {
7765
+ "epoch": 2.52,
7766
+ "learning_rate": 0.0002,
7767
+ "loss": 0.7663,
7768
+ "step": 8200
7769
+ },
7770
+ {
7771
+ "epoch": 2.52,
7772
+ "eval_loss": 0.9532663226127625,
7773
+ "eval_runtime": 67.4923,
7774
+ "eval_samples_per_second": 14.817,
7775
+ "eval_steps_per_second": 7.408,
7776
+ "step": 8200
7777
+ },
7778
+ {
7779
+ "epoch": 2.52,
7780
+ "mmlu_eval_accuracy": 0.496708696985176,
7781
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
7782
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
7783
+ "mmlu_eval_accuracy_astronomy": 0.375,
7784
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
7785
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
7786
+ "mmlu_eval_accuracy_college_biology": 0.5,
7787
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
7788
+ "mmlu_eval_accuracy_college_computer_science": 0.5454545454545454,
7789
+ "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
7790
+ "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
7791
+ "mmlu_eval_accuracy_college_physics": 0.2727272727272727,
7792
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
7793
+ "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
7794
+ "mmlu_eval_accuracy_econometrics": 0.25,
7795
+ "mmlu_eval_accuracy_electrical_engineering": 0.1875,
7796
+ "mmlu_eval_accuracy_elementary_mathematics": 0.2682926829268293,
7797
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
7798
+ "mmlu_eval_accuracy_global_facts": 0.6,
7799
+ "mmlu_eval_accuracy_high_school_biology": 0.46875,
7800
+ "mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
7801
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
7802
+ "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
7803
+ "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
7804
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
7805
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
7806
+ "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
7807
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
7808
+ "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
7809
+ "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
7810
+ "mmlu_eval_accuracy_high_school_statistics": 0.4782608695652174,
7811
+ "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
7812
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
7813
+ "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
7814
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
7815
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
7816
+ "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
7817
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
7818
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
7819
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
7820
+ "mmlu_eval_accuracy_marketing": 0.84,
7821
+ "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
7822
+ "mmlu_eval_accuracy_miscellaneous": 0.686046511627907,
7823
+ "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
7824
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
7825
+ "mmlu_eval_accuracy_nutrition": 0.5151515151515151,
7826
+ "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
7827
+ "mmlu_eval_accuracy_prehistory": 0.5714285714285714,
7828
+ "mmlu_eval_accuracy_professional_accounting": 0.3870967741935484,
7829
+ "mmlu_eval_accuracy_professional_law": 0.35294117647058826,
7830
+ "mmlu_eval_accuracy_professional_medicine": 0.6129032258064516,
7831
+ "mmlu_eval_accuracy_professional_psychology": 0.43478260869565216,
7832
+ "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
7833
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
7834
+ "mmlu_eval_accuracy_sociology": 0.6818181818181818,
7835
+ "mmlu_eval_accuracy_us_foreign_policy": 0.9090909090909091,
7836
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
7837
+ "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
7838
+ "mmlu_loss": 1.3628051146359108,
7839
+ "step": 8200
7840
+ },
7841
+ {
7842
+ "epoch": 2.53,
7843
+ "learning_rate": 0.0002,
7844
+ "loss": 0.6874,
7845
+ "step": 8210
7846
+ },
7847
+ {
7848
+ "epoch": 2.53,
7849
+ "learning_rate": 0.0002,
7850
+ "loss": 0.7562,
7851
+ "step": 8220
7852
+ },
7853
+ {
7854
+ "epoch": 2.53,
7855
+ "learning_rate": 0.0002,
7856
+ "loss": 0.774,
7857
+ "step": 8230
7858
+ },
7859
+ {
7860
+ "epoch": 2.53,
7861
+ "learning_rate": 0.0002,
7862
+ "loss": 0.7271,
7863
+ "step": 8240
7864
+ },
7865
+ {
7866
+ "epoch": 2.54,
7867
+ "learning_rate": 0.0002,
7868
+ "loss": 0.7389,
7869
+ "step": 8250
7870
+ },
7871
+ {
7872
+ "epoch": 2.54,
7873
+ "learning_rate": 0.0002,
7874
+ "loss": 0.7315,
7875
+ "step": 8260
7876
+ },
7877
+ {
7878
+ "epoch": 2.54,
7879
+ "learning_rate": 0.0002,
7880
+ "loss": 0.7251,
7881
+ "step": 8270
7882
+ },
7883
+ {
7884
+ "epoch": 2.55,
7885
+ "learning_rate": 0.0002,
7886
+ "loss": 0.7357,
7887
+ "step": 8280
7888
+ },
7889
+ {
7890
+ "epoch": 2.55,
7891
+ "learning_rate": 0.0002,
7892
+ "loss": 0.7861,
7893
+ "step": 8290
7894
+ },
7895
+ {
7896
+ "epoch": 2.55,
7897
+ "learning_rate": 0.0002,
7898
+ "loss": 0.7303,
7899
+ "step": 8300
7900
+ },
7901
+ {
7902
+ "epoch": 2.56,
7903
+ "learning_rate": 0.0002,
7904
+ "loss": 0.7912,
7905
+ "step": 8310
7906
+ },
7907
+ {
7908
+ "epoch": 2.56,
7909
+ "learning_rate": 0.0002,
7910
+ "loss": 0.7137,
7911
+ "step": 8320
7912
+ },
7913
+ {
7914
+ "epoch": 2.56,
7915
+ "learning_rate": 0.0002,
7916
+ "loss": 0.7976,
7917
+ "step": 8330
7918
+ },
7919
+ {
7920
+ "epoch": 2.57,
7921
+ "learning_rate": 0.0002,
7922
+ "loss": 0.7737,
7923
+ "step": 8340
7924
+ },
7925
+ {
7926
+ "epoch": 2.57,
7927
+ "learning_rate": 0.0002,
7928
+ "loss": 0.7494,
7929
+ "step": 8350
7930
+ },
7931
+ {
7932
+ "epoch": 2.57,
7933
+ "learning_rate": 0.0002,
7934
+ "loss": 0.7697,
7935
+ "step": 8360
7936
+ },
7937
+ {
7938
+ "epoch": 2.57,
7939
+ "learning_rate": 0.0002,
7940
+ "loss": 0.7982,
7941
+ "step": 8370
7942
+ },
7943
+ {
7944
+ "epoch": 2.58,
7945
+ "learning_rate": 0.0002,
7946
+ "loss": 0.729,
7947
+ "step": 8380
7948
+ },
7949
+ {
7950
+ "epoch": 2.58,
7951
+ "learning_rate": 0.0002,
7952
+ "loss": 0.8068,
7953
+ "step": 8390
7954
+ },
7955
+ {
7956
+ "epoch": 2.58,
7957
+ "learning_rate": 0.0002,
7958
+ "loss": 0.7601,
7959
+ "step": 8400
7960
+ },
7961
+ {
7962
+ "epoch": 2.58,
7963
+ "eval_loss": 0.9522873759269714,
7964
+ "eval_runtime": 67.418,
7965
+ "eval_samples_per_second": 14.833,
7966
+ "eval_steps_per_second": 7.416,
7967
+ "step": 8400
7968
+ },
7969
+ {
7970
+ "epoch": 2.58,
7971
+ "mmlu_eval_accuracy": 0.4995362526191875,
7972
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
7973
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
7974
+ "mmlu_eval_accuracy_astronomy": 0.5,
7975
+ "mmlu_eval_accuracy_business_ethics": 0.36363636363636365,
7976
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
7977
+ "mmlu_eval_accuracy_college_biology": 0.5,
7978
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
7979
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
7980
+ "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
7981
+ "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
7982
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
7983
+ "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
7984
+ "mmlu_eval_accuracy_conceptual_physics": 0.5,
7985
+ "mmlu_eval_accuracy_econometrics": 0.25,
7986
+ "mmlu_eval_accuracy_electrical_engineering": 0.1875,
7987
+ "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
7988
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
7989
+ "mmlu_eval_accuracy_global_facts": 0.6,
7990
+ "mmlu_eval_accuracy_high_school_biology": 0.4375,
7991
+ "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
7992
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
7993
+ "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
7994
+ "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
7995
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
7996
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
7997
+ "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
7998
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
7999
+ "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
8000
+ "mmlu_eval_accuracy_high_school_psychology": 0.8166666666666667,
8001
+ "mmlu_eval_accuracy_high_school_statistics": 0.4782608695652174,
8002
+ "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
8003
+ "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
8004
+ "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
8005
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
8006
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
8007
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
8008
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
8009
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
8010
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
8011
+ "mmlu_eval_accuracy_marketing": 0.84,
8012
+ "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
8013
+ "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
8014
+ "mmlu_eval_accuracy_moral_disputes": 0.5526315789473685,
8015
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
8016
+ "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
8017
+ "mmlu_eval_accuracy_philosophy": 0.5,
8018
+ "mmlu_eval_accuracy_prehistory": 0.6285714285714286,
8019
+ "mmlu_eval_accuracy_professional_accounting": 0.41935483870967744,
8020
+ "mmlu_eval_accuracy_professional_law": 0.36470588235294116,
8021
+ "mmlu_eval_accuracy_professional_medicine": 0.5806451612903226,
8022
+ "mmlu_eval_accuracy_professional_psychology": 0.5072463768115942,
8023
+ "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
8024
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
8025
+ "mmlu_eval_accuracy_sociology": 0.7272727272727273,
8026
+ "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
8027
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
8028
+ "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
8029
+ "mmlu_loss": 1.3607582125044366,
8030
+ "step": 8400
8031
+ },
8032
+ {
8033
+ "epoch": 2.59,
8034
+ "learning_rate": 0.0002,
8035
+ "loss": 0.7996,
8036
+ "step": 8410
8037
+ },
8038
+ {
8039
+ "epoch": 2.59,
8040
+ "learning_rate": 0.0002,
8041
+ "loss": 0.7926,
8042
+ "step": 8420
8043
+ },
8044
+ {
8045
+ "epoch": 2.59,
8046
+ "learning_rate": 0.0002,
8047
+ "loss": 0.7554,
8048
+ "step": 8430
8049
+ },
8050
+ {
8051
+ "epoch": 2.6,
8052
+ "learning_rate": 0.0002,
8053
+ "loss": 0.7062,
8054
+ "step": 8440
8055
+ },
8056
+ {
8057
+ "epoch": 2.6,
8058
+ "learning_rate": 0.0002,
8059
+ "loss": 0.7646,
8060
+ "step": 8450
8061
+ },
8062
+ {
8063
+ "epoch": 2.6,
8064
+ "learning_rate": 0.0002,
8065
+ "loss": 0.7543,
8066
+ "step": 8460
8067
+ },
8068
+ {
8069
+ "epoch": 2.61,
8070
+ "learning_rate": 0.0002,
8071
+ "loss": 0.7408,
8072
+ "step": 8470
8073
+ },
8074
+ {
8075
+ "epoch": 2.61,
8076
+ "learning_rate": 0.0002,
8077
+ "loss": 0.7622,
8078
+ "step": 8480
8079
+ },
8080
+ {
8081
+ "epoch": 2.61,
8082
+ "learning_rate": 0.0002,
8083
+ "loss": 0.7204,
8084
+ "step": 8490
8085
+ },
8086
+ {
8087
+ "epoch": 2.61,
8088
+ "learning_rate": 0.0002,
8089
+ "loss": 0.7642,
8090
+ "step": 8500
8091
+ },
8092
+ {
8093
+ "epoch": 2.62,
8094
+ "learning_rate": 0.0002,
8095
+ "loss": 0.7625,
8096
+ "step": 8510
8097
+ },
8098
+ {
8099
+ "epoch": 2.62,
8100
+ "learning_rate": 0.0002,
8101
+ "loss": 0.7782,
8102
+ "step": 8520
8103
+ },
8104
+ {
8105
+ "epoch": 2.62,
8106
+ "learning_rate": 0.0002,
8107
+ "loss": 0.7365,
8108
+ "step": 8530
8109
+ },
8110
+ {
8111
+ "epoch": 2.63,
8112
+ "learning_rate": 0.0002,
8113
+ "loss": 0.8038,
8114
+ "step": 8540
8115
+ },
8116
+ {
8117
+ "epoch": 2.63,
8118
+ "learning_rate": 0.0002,
8119
+ "loss": 0.7693,
8120
+ "step": 8550
8121
+ },
8122
+ {
8123
+ "epoch": 2.63,
8124
+ "learning_rate": 0.0002,
8125
+ "loss": 0.7791,
8126
+ "step": 8560
8127
+ },
8128
+ {
8129
+ "epoch": 2.64,
8130
+ "learning_rate": 0.0002,
8131
+ "loss": 0.7742,
8132
+ "step": 8570
8133
+ },
8134
+ {
8135
+ "epoch": 2.64,
8136
+ "learning_rate": 0.0002,
8137
+ "loss": 0.7836,
8138
+ "step": 8580
8139
+ },
8140
+ {
8141
+ "epoch": 2.64,
8142
+ "learning_rate": 0.0002,
8143
+ "loss": 0.7413,
8144
+ "step": 8590
8145
+ },
8146
+ {
8147
+ "epoch": 2.65,
8148
+ "learning_rate": 0.0002,
8149
+ "loss": 0.7824,
8150
+ "step": 8600
8151
+ },
8152
+ {
8153
+ "epoch": 2.65,
8154
+ "eval_loss": 0.9485287666320801,
8155
+ "eval_runtime": 67.4906,
8156
+ "eval_samples_per_second": 14.817,
8157
+ "eval_steps_per_second": 7.408,
8158
+ "step": 8600
8159
+ },
8160
+ {
8161
+ "epoch": 2.65,
8162
+ "mmlu_eval_accuracy": 0.49526726033631285,
8163
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
8164
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
8165
+ "mmlu_eval_accuracy_astronomy": 0.4375,
8166
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
8167
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
8168
+ "mmlu_eval_accuracy_college_biology": 0.5,
8169
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
8170
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
8171
+ "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
8172
+ "mmlu_eval_accuracy_college_medicine": 0.5,
8173
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
8174
+ "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
8175
+ "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
8176
+ "mmlu_eval_accuracy_econometrics": 0.3333333333333333,
8177
+ "mmlu_eval_accuracy_electrical_engineering": 0.1875,
8178
+ "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
8179
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
8180
+ "mmlu_eval_accuracy_global_facts": 0.6,
8181
+ "mmlu_eval_accuracy_high_school_biology": 0.46875,
8182
+ "mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
8183
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
8184
+ "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
8185
+ "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
8186
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
8187
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
8188
+ "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
8189
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
8190
+ "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
8191
+ "mmlu_eval_accuracy_high_school_psychology": 0.8166666666666667,
8192
+ "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
8193
+ "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
8194
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
8195
+ "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
8196
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
8197
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
8198
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
8199
+ "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
8200
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
8201
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
8202
+ "mmlu_eval_accuracy_marketing": 0.76,
8203
+ "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
8204
+ "mmlu_eval_accuracy_miscellaneous": 0.6976744186046512,
8205
+ "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
8206
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
8207
+ "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
8208
+ "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
8209
+ "mmlu_eval_accuracy_prehistory": 0.6,
8210
+ "mmlu_eval_accuracy_professional_accounting": 0.45161290322580644,
8211
+ "mmlu_eval_accuracy_professional_law": 0.34705882352941175,
8212
+ "mmlu_eval_accuracy_professional_medicine": 0.5806451612903226,
8213
+ "mmlu_eval_accuracy_professional_psychology": 0.463768115942029,
8214
+ "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
8215
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
8216
+ "mmlu_eval_accuracy_sociology": 0.7727272727272727,
8217
+ "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
8218
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
8219
+ "mmlu_eval_accuracy_world_religions": 0.5263157894736842,
8220
+ "mmlu_loss": 1.4527632203488063,
8221
+ "step": 8600
8222
+ },
8223
+ {
8224
+ "epoch": 2.65,
8225
+ "learning_rate": 0.0002,
8226
+ "loss": 0.7723,
8227
+ "step": 8610
8228
+ },
8229
+ {
8230
+ "epoch": 2.65,
8231
+ "learning_rate": 0.0002,
8232
+ "loss": 0.7681,
8233
+ "step": 8620
8234
+ },
8235
+ {
8236
+ "epoch": 2.65,
8237
+ "learning_rate": 0.0002,
8238
+ "loss": 0.8204,
8239
+ "step": 8630
8240
+ },
8241
+ {
8242
+ "epoch": 2.66,
8243
+ "learning_rate": 0.0002,
8244
+ "loss": 0.7274,
8245
+ "step": 8640
8246
+ },
8247
+ {
8248
+ "epoch": 2.66,
8249
+ "learning_rate": 0.0002,
8250
+ "loss": 0.7819,
8251
+ "step": 8650
8252
+ },
8253
+ {
8254
+ "epoch": 2.66,
8255
+ "learning_rate": 0.0002,
8256
+ "loss": 0.7369,
8257
+ "step": 8660
8258
+ },
8259
+ {
8260
+ "epoch": 2.67,
8261
+ "learning_rate": 0.0002,
8262
+ "loss": 0.7847,
8263
+ "step": 8670
8264
+ },
8265
+ {
8266
+ "epoch": 2.67,
8267
+ "learning_rate": 0.0002,
8268
+ "loss": 0.7938,
8269
+ "step": 8680
8270
+ },
8271
+ {
8272
+ "epoch": 2.67,
8273
+ "learning_rate": 0.0002,
8274
+ "loss": 0.7443,
8275
+ "step": 8690
8276
+ },
8277
+ {
8278
+ "epoch": 2.68,
8279
+ "learning_rate": 0.0002,
8280
+ "loss": 0.809,
8281
+ "step": 8700
8282
+ },
8283
+ {
8284
+ "epoch": 2.68,
8285
+ "learning_rate": 0.0002,
8286
+ "loss": 0.7389,
8287
+ "step": 8710
8288
+ },
8289
+ {
8290
+ "epoch": 2.68,
8291
+ "learning_rate": 0.0002,
8292
+ "loss": 0.7783,
8293
+ "step": 8720
8294
+ },
8295
+ {
8296
+ "epoch": 2.69,
8297
+ "learning_rate": 0.0002,
8298
+ "loss": 0.8125,
8299
+ "step": 8730
8300
+ },
8301
+ {
8302
+ "epoch": 2.69,
8303
+ "learning_rate": 0.0002,
8304
+ "loss": 0.7157,
8305
+ "step": 8740
8306
+ },
8307
+ {
8308
+ "epoch": 2.69,
8309
+ "learning_rate": 0.0002,
8310
+ "loss": 0.7714,
8311
+ "step": 8750
8312
+ },
8313
+ {
8314
+ "epoch": 2.69,
8315
+ "learning_rate": 0.0002,
8316
+ "loss": 0.8028,
8317
+ "step": 8760
8318
+ },
8319
+ {
8320
+ "epoch": 2.7,
8321
+ "learning_rate": 0.0002,
8322
+ "loss": 0.7408,
8323
+ "step": 8770
8324
+ },
8325
+ {
8326
+ "epoch": 2.7,
8327
+ "learning_rate": 0.0002,
8328
+ "loss": 0.7985,
8329
+ "step": 8780
8330
+ },
8331
+ {
8332
+ "epoch": 2.7,
8333
+ "learning_rate": 0.0002,
8334
+ "loss": 0.6885,
8335
+ "step": 8790
8336
+ },
8337
+ {
8338
+ "epoch": 2.71,
8339
+ "learning_rate": 0.0002,
8340
+ "loss": 0.7649,
8341
+ "step": 8800
8342
+ },
8343
+ {
8344
+ "epoch": 2.71,
8345
+ "eval_loss": 0.9492066502571106,
8346
+ "eval_runtime": 67.6841,
8347
+ "eval_samples_per_second": 14.775,
8348
+ "eval_steps_per_second": 7.387,
8349
+ "step": 8800
8350
+ },
8351
+ {
8352
+ "epoch": 2.71,
8353
+ "mmlu_eval_accuracy": 0.5004865823351183,
8354
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
8355
+ "mmlu_eval_accuracy_anatomy": 0.5,
8356
+ "mmlu_eval_accuracy_astronomy": 0.5,
8357
+ "mmlu_eval_accuracy_business_ethics": 0.36363636363636365,
8358
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
8359
+ "mmlu_eval_accuracy_college_biology": 0.4375,
8360
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
8361
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
8362
+ "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
8363
+ "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
8364
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
8365
+ "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
8366
+ "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
8367
+ "mmlu_eval_accuracy_econometrics": 0.25,
8368
+ "mmlu_eval_accuracy_electrical_engineering": 0.1875,
8369
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
8370
+ "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
8371
+ "mmlu_eval_accuracy_global_facts": 0.6,
8372
+ "mmlu_eval_accuracy_high_school_biology": 0.53125,
8373
+ "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
8374
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
8375
+ "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
8376
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
8377
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
8378
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
8379
+ "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
8380
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5769230769230769,
8381
+ "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
8382
+ "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
8383
+ "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
8384
+ "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
8385
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
8386
+ "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
8387
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
8388
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
8389
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
8390
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
8391
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
8392
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
8393
+ "mmlu_eval_accuracy_marketing": 0.8,
8394
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
8395
+ "mmlu_eval_accuracy_miscellaneous": 0.686046511627907,
8396
+ "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
8397
+ "mmlu_eval_accuracy_moral_scenarios": 0.25,
8398
+ "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
8399
+ "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
8400
+ "mmlu_eval_accuracy_prehistory": 0.6571428571428571,
8401
+ "mmlu_eval_accuracy_professional_accounting": 0.41935483870967744,
8402
+ "mmlu_eval_accuracy_professional_law": 0.3588235294117647,
8403
+ "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
8404
+ "mmlu_eval_accuracy_professional_psychology": 0.43478260869565216,
8405
+ "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
8406
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
8407
+ "mmlu_eval_accuracy_sociology": 0.7727272727272727,
8408
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
8409
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
8410
+ "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
8411
+ "mmlu_loss": 1.3377955840710871,
8412
+ "step": 8800
8413
+ },
8414
+ {
8415
+ "epoch": 2.71,
8416
+ "learning_rate": 0.0002,
8417
+ "loss": 0.7126,
8418
+ "step": 8810
8419
+ },
8420
+ {
8421
+ "epoch": 2.71,
8422
+ "learning_rate": 0.0002,
8423
+ "loss": 0.7481,
8424
+ "step": 8820
8425
+ },
8426
+ {
8427
+ "epoch": 2.72,
8428
+ "learning_rate": 0.0002,
8429
+ "loss": 0.7264,
8430
+ "step": 8830
8431
+ },
8432
+ {
8433
+ "epoch": 2.72,
8434
+ "learning_rate": 0.0002,
8435
+ "loss": 0.7501,
8436
+ "step": 8840
8437
+ },
8438
+ {
8439
+ "epoch": 2.72,
8440
+ "learning_rate": 0.0002,
8441
+ "loss": 0.7714,
8442
+ "step": 8850
8443
+ },
8444
+ {
8445
+ "epoch": 2.73,
8446
+ "learning_rate": 0.0002,
8447
+ "loss": 0.7933,
8448
+ "step": 8860
8449
+ },
8450
+ {
8451
+ "epoch": 2.73,
8452
+ "learning_rate": 0.0002,
8453
+ "loss": 0.7968,
8454
+ "step": 8870
8455
+ },
8456
+ {
8457
+ "epoch": 2.73,
8458
+ "learning_rate": 0.0002,
8459
+ "loss": 0.7458,
8460
+ "step": 8880
8461
+ },
8462
+ {
8463
+ "epoch": 2.73,
8464
+ "learning_rate": 0.0002,
8465
+ "loss": 0.7822,
8466
+ "step": 8890
8467
+ },
8468
+ {
8469
+ "epoch": 2.74,
8470
+ "learning_rate": 0.0002,
8471
+ "loss": 0.765,
8472
+ "step": 8900
8473
+ },
8474
+ {
8475
+ "epoch": 2.74,
8476
+ "learning_rate": 0.0002,
8477
+ "loss": 0.7829,
8478
+ "step": 8910
8479
+ },
8480
+ {
8481
+ "epoch": 2.74,
8482
+ "learning_rate": 0.0002,
8483
+ "loss": 0.7429,
8484
+ "step": 8920
8485
+ },
8486
+ {
8487
+ "epoch": 2.75,
8488
+ "learning_rate": 0.0002,
8489
+ "loss": 0.7886,
8490
+ "step": 8930
8491
+ },
8492
+ {
8493
+ "epoch": 2.75,
8494
+ "learning_rate": 0.0002,
8495
+ "loss": 0.7442,
8496
+ "step": 8940
8497
+ },
8498
+ {
8499
+ "epoch": 2.75,
8500
+ "learning_rate": 0.0002,
8501
+ "loss": 0.7958,
8502
+ "step": 8950
8503
+ },
8504
+ {
8505
+ "epoch": 2.76,
8506
+ "learning_rate": 0.0002,
8507
+ "loss": 0.7333,
8508
+ "step": 8960
8509
+ },
8510
+ {
8511
+ "epoch": 2.76,
8512
+ "learning_rate": 0.0002,
8513
+ "loss": 0.7775,
8514
+ "step": 8970
8515
+ },
8516
+ {
8517
+ "epoch": 2.76,
8518
+ "learning_rate": 0.0002,
8519
+ "loss": 0.8065,
8520
+ "step": 8980
8521
+ },
8522
+ {
8523
+ "epoch": 2.77,
8524
+ "learning_rate": 0.0002,
8525
+ "loss": 0.7737,
8526
+ "step": 8990
8527
+ },
8528
+ {
8529
+ "epoch": 2.77,
8530
+ "learning_rate": 0.0002,
8531
+ "loss": 0.7743,
8532
+ "step": 9000
8533
+ },
8534
+ {
8535
+ "epoch": 2.77,
8536
+ "eval_loss": 0.9460086226463318,
8537
+ "eval_runtime": 67.3893,
8538
+ "eval_samples_per_second": 14.839,
8539
+ "eval_steps_per_second": 7.42,
8540
+ "step": 9000
8541
+ },
8542
+ {
8543
+ "epoch": 2.77,
8544
+ "mmlu_eval_accuracy": 0.5000339664478387,
8545
+ "mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
8546
+ "mmlu_eval_accuracy_anatomy": 0.5,
8547
+ "mmlu_eval_accuracy_astronomy": 0.5,
8548
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
8549
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
8550
+ "mmlu_eval_accuracy_college_biology": 0.4375,
8551
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
8552
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
8553
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
8554
+ "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
8555
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
8556
+ "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
8557
+ "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
8558
+ "mmlu_eval_accuracy_econometrics": 0.25,
8559
+ "mmlu_eval_accuracy_electrical_engineering": 0.1875,
8560
+ "mmlu_eval_accuracy_elementary_mathematics": 0.2926829268292683,
8561
+ "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
8562
+ "mmlu_eval_accuracy_global_facts": 0.6,
8563
+ "mmlu_eval_accuracy_high_school_biology": 0.5,
8564
+ "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
8565
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
8566
+ "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
8567
+ "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
8568
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
8569
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
8570
+ "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
8571
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
8572
+ "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
8573
+ "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
8574
+ "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
8575
+ "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
8576
+ "mmlu_eval_accuracy_high_school_world_history": 0.7692307692307693,
8577
+ "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
8578
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
8579
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
8580
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
8581
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
8582
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
8583
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
8584
+ "mmlu_eval_accuracy_marketing": 0.84,
8585
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
8586
+ "mmlu_eval_accuracy_miscellaneous": 0.7093023255813954,
8587
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
8588
+ "mmlu_eval_accuracy_moral_scenarios": 0.21,
8589
+ "mmlu_eval_accuracy_nutrition": 0.5151515151515151,
8590
+ "mmlu_eval_accuracy_philosophy": 0.5,
8591
+ "mmlu_eval_accuracy_prehistory": 0.6857142857142857,
8592
+ "mmlu_eval_accuracy_professional_accounting": 0.41935483870967744,
8593
+ "mmlu_eval_accuracy_professional_law": 0.35294117647058826,
8594
+ "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
8595
+ "mmlu_eval_accuracy_professional_psychology": 0.4492753623188406,
8596
+ "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
8597
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
8598
+ "mmlu_eval_accuracy_sociology": 0.8181818181818182,
8599
+ "mmlu_eval_accuracy_us_foreign_policy": 0.8181818181818182,
8600
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
8601
+ "mmlu_eval_accuracy_world_religions": 0.631578947368421,
8602
+ "mmlu_loss": 1.3751481442320752,
8603
+ "step": 9000
8604
+ },
8605
+ {
8606
+ "epoch": 2.77,
8607
+ "learning_rate": 0.0002,
8608
+ "loss": 0.7736,
8609
+ "step": 9010
8610
+ },
8611
+ {
8612
+ "epoch": 2.77,
8613
+ "learning_rate": 0.0002,
8614
+ "loss": 0.7281,
8615
+ "step": 9020
8616
+ },
8617
+ {
8618
+ "epoch": 2.78,
8619
+ "learning_rate": 0.0002,
8620
+ "loss": 0.7571,
8621
+ "step": 9030
8622
+ },
8623
+ {
8624
+ "epoch": 2.78,
8625
+ "learning_rate": 0.0002,
8626
+ "loss": 0.7373,
8627
+ "step": 9040
8628
+ },
8629
+ {
8630
+ "epoch": 2.78,
8631
+ "learning_rate": 0.0002,
8632
+ "loss": 0.7117,
8633
+ "step": 9050
8634
+ },
8635
+ {
8636
+ "epoch": 2.79,
8637
+ "learning_rate": 0.0002,
8638
+ "loss": 0.7537,
8639
+ "step": 9060
8640
+ },
8641
+ {
8642
+ "epoch": 2.79,
8643
+ "learning_rate": 0.0002,
8644
+ "loss": 0.7267,
8645
+ "step": 9070
8646
+ },
8647
+ {
8648
+ "epoch": 2.79,
8649
+ "learning_rate": 0.0002,
8650
+ "loss": 0.7674,
8651
+ "step": 9080
8652
+ },
8653
+ {
8654
+ "epoch": 2.8,
8655
+ "learning_rate": 0.0002,
8656
+ "loss": 0.7515,
8657
+ "step": 9090
8658
+ },
8659
+ {
8660
+ "epoch": 2.8,
8661
+ "learning_rate": 0.0002,
8662
+ "loss": 0.7882,
8663
+ "step": 9100
8664
+ },
8665
+ {
8666
+ "epoch": 2.8,
8667
+ "learning_rate": 0.0002,
8668
+ "loss": 0.6996,
8669
+ "step": 9110
8670
+ },
8671
+ {
8672
+ "epoch": 2.81,
8673
+ "learning_rate": 0.0002,
8674
+ "loss": 0.7122,
8675
+ "step": 9120
8676
+ },
8677
+ {
8678
+ "epoch": 2.81,
8679
+ "learning_rate": 0.0002,
8680
+ "loss": 0.6944,
8681
+ "step": 9130
8682
+ },
8683
+ {
8684
+ "epoch": 2.81,
8685
+ "learning_rate": 0.0002,
8686
+ "loss": 0.7341,
8687
+ "step": 9140
8688
+ },
8689
+ {
8690
+ "epoch": 2.81,
8691
+ "learning_rate": 0.0002,
8692
+ "loss": 0.7631,
8693
+ "step": 9150
8694
+ },
8695
+ {
8696
+ "epoch": 2.82,
8697
+ "learning_rate": 0.0002,
8698
+ "loss": 0.7504,
8699
+ "step": 9160
8700
+ },
8701
+ {
8702
+ "epoch": 2.82,
8703
+ "learning_rate": 0.0002,
8704
+ "loss": 0.6992,
8705
+ "step": 9170
8706
+ },
8707
+ {
8708
+ "epoch": 2.82,
8709
+ "learning_rate": 0.0002,
8710
+ "loss": 0.768,
8711
+ "step": 9180
8712
+ },
8713
+ {
8714
+ "epoch": 2.83,
8715
+ "learning_rate": 0.0002,
8716
+ "loss": 0.7527,
8717
+ "step": 9190
8718
+ },
8719
+ {
8720
+ "epoch": 2.83,
8721
+ "learning_rate": 0.0002,
8722
+ "loss": 0.7673,
8723
+ "step": 9200
8724
+ },
8725
+ {
8726
+ "epoch": 2.83,
8727
+ "eval_loss": 0.9476088285446167,
8728
+ "eval_runtime": 67.523,
8729
+ "eval_samples_per_second": 14.81,
8730
+ "eval_steps_per_second": 7.405,
8731
+ "step": 9200
8732
+ },
8733
+ {
8734
+ "epoch": 2.83,
8735
+ "mmlu_eval_accuracy": 0.507623397777817,
8736
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
8737
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
8738
+ "mmlu_eval_accuracy_astronomy": 0.5625,
8739
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
8740
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
8741
+ "mmlu_eval_accuracy_college_biology": 0.4375,
8742
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
8743
+ "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
8744
+ "mmlu_eval_accuracy_college_mathematics": 0.45454545454545453,
8745
+ "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
8746
+ "mmlu_eval_accuracy_college_physics": 0.5454545454545454,
8747
+ "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
8748
+ "mmlu_eval_accuracy_conceptual_physics": 0.5,
8749
+ "mmlu_eval_accuracy_econometrics": 0.25,
8750
+ "mmlu_eval_accuracy_electrical_engineering": 0.1875,
8751
+ "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
8752
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
8753
+ "mmlu_eval_accuracy_global_facts": 0.5,
8754
+ "mmlu_eval_accuracy_high_school_biology": 0.5,
8755
+ "mmlu_eval_accuracy_high_school_chemistry": 0.22727272727272727,
8756
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
8757
+ "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
8758
+ "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
8759
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
8760
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
8761
+ "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
8762
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5769230769230769,
8763
+ "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
8764
+ "mmlu_eval_accuracy_high_school_psychology": 0.8833333333333333,
8765
+ "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
8766
+ "mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
8767
+ "mmlu_eval_accuracy_high_school_world_history": 0.7692307692307693,
8768
+ "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
8769
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
8770
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
8771
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
8772
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
8773
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
8774
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
8775
+ "mmlu_eval_accuracy_marketing": 0.8,
8776
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
8777
+ "mmlu_eval_accuracy_miscellaneous": 0.6976744186046512,
8778
+ "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
8779
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
8780
+ "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
8781
+ "mmlu_eval_accuracy_philosophy": 0.5,
8782
+ "mmlu_eval_accuracy_prehistory": 0.6571428571428571,
8783
+ "mmlu_eval_accuracy_professional_accounting": 0.45161290322580644,
8784
+ "mmlu_eval_accuracy_professional_law": 0.36470588235294116,
8785
+ "mmlu_eval_accuracy_professional_medicine": 0.5806451612903226,
8786
+ "mmlu_eval_accuracy_professional_psychology": 0.4492753623188406,
8787
+ "mmlu_eval_accuracy_public_relations": 0.5,
8788
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
8789
+ "mmlu_eval_accuracy_sociology": 0.7727272727272727,
8790
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
8791
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
8792
+ "mmlu_eval_accuracy_world_religions": 0.5789473684210527,
8793
+ "mmlu_loss": 1.2142367916225454,
8794
+ "step": 9200
8795
+ },
8796
+ {
8797
+ "epoch": 2.83,
8798
+ "learning_rate": 0.0002,
8799
+ "loss": 0.7534,
8800
+ "step": 9210
8801
+ },
8802
+ {
8803
+ "epoch": 2.84,
8804
+ "learning_rate": 0.0002,
8805
+ "loss": 0.7422,
8806
+ "step": 9220
8807
+ },
8808
+ {
8809
+ "epoch": 2.84,
8810
+ "learning_rate": 0.0002,
8811
+ "loss": 0.7734,
8812
+ "step": 9230
8813
+ },
8814
+ {
8815
+ "epoch": 2.84,
8816
+ "learning_rate": 0.0002,
8817
+ "loss": 0.7919,
8818
+ "step": 9240
8819
+ },
8820
+ {
8821
+ "epoch": 2.85,
8822
+ "learning_rate": 0.0002,
8823
+ "loss": 0.7599,
8824
+ "step": 9250
8825
+ },
8826
+ {
8827
+ "epoch": 2.85,
8828
+ "learning_rate": 0.0002,
8829
+ "loss": 0.7595,
8830
+ "step": 9260
8831
+ },
8832
+ {
8833
+ "epoch": 2.85,
8834
+ "learning_rate": 0.0002,
8835
+ "loss": 0.7703,
8836
+ "step": 9270
8837
+ },
8838
+ {
8839
+ "epoch": 2.85,
8840
+ "learning_rate": 0.0002,
8841
+ "loss": 0.7413,
8842
+ "step": 9280
8843
+ },
8844
+ {
8845
+ "epoch": 2.86,
8846
+ "learning_rate": 0.0002,
8847
+ "loss": 0.7631,
8848
+ "step": 9290
8849
+ },
8850
+ {
8851
+ "epoch": 2.86,
8852
+ "learning_rate": 0.0002,
8853
+ "loss": 0.7278,
8854
+ "step": 9300
8855
+ },
8856
+ {
8857
+ "epoch": 2.86,
8858
+ "learning_rate": 0.0002,
8859
+ "loss": 0.7493,
8860
+ "step": 9310
8861
+ },
8862
+ {
8863
+ "epoch": 2.87,
8864
+ "learning_rate": 0.0002,
8865
+ "loss": 0.7226,
8866
+ "step": 9320
8867
+ },
8868
+ {
8869
+ "epoch": 2.87,
8870
+ "learning_rate": 0.0002,
8871
+ "loss": 0.7773,
8872
+ "step": 9330
8873
+ },
8874
+ {
8875
+ "epoch": 2.87,
8876
+ "learning_rate": 0.0002,
8877
+ "loss": 0.7674,
8878
+ "step": 9340
8879
+ },
8880
+ {
8881
+ "epoch": 2.88,
8882
+ "learning_rate": 0.0002,
8883
+ "loss": 0.7739,
8884
+ "step": 9350
8885
+ },
8886
+ {
8887
+ "epoch": 2.88,
8888
+ "learning_rate": 0.0002,
8889
+ "loss": 0.7742,
8890
+ "step": 9360
8891
+ },
8892
+ {
8893
+ "epoch": 2.88,
8894
+ "learning_rate": 0.0002,
8895
+ "loss": 0.7224,
8896
+ "step": 9370
8897
+ },
8898
+ {
8899
+ "epoch": 2.89,
8900
+ "learning_rate": 0.0002,
8901
+ "loss": 0.7846,
8902
+ "step": 9380
8903
+ },
8904
+ {
8905
+ "epoch": 2.89,
8906
+ "learning_rate": 0.0002,
8907
+ "loss": 0.7982,
8908
+ "step": 9390
8909
+ },
8910
+ {
8911
+ "epoch": 2.89,
8912
+ "learning_rate": 0.0002,
8913
+ "loss": 0.7616,
8914
+ "step": 9400
8915
+ },
8916
+ {
8917
+ "epoch": 2.89,
8918
+ "eval_loss": 0.9453636407852173,
8919
+ "eval_runtime": 67.5275,
8920
+ "eval_samples_per_second": 14.809,
8921
+ "eval_steps_per_second": 7.404,
8922
+ "step": 9400
8923
+ },
8924
+ {
8925
+ "epoch": 2.89,
8926
+ "mmlu_eval_accuracy": 0.5016145076125647,
8927
+ "mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
8928
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
8929
+ "mmlu_eval_accuracy_astronomy": 0.5,
8930
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
8931
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
8932
+ "mmlu_eval_accuracy_college_biology": 0.4375,
8933
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
8934
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
8935
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
8936
+ "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
8937
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
8938
+ "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
8939
+ "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
8940
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
8941
+ "mmlu_eval_accuracy_electrical_engineering": 0.1875,
8942
+ "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
8943
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
8944
+ "mmlu_eval_accuracy_global_facts": 0.5,
8945
+ "mmlu_eval_accuracy_high_school_biology": 0.46875,
8946
+ "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
8947
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
8948
+ "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
8949
+ "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
8950
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
8951
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
8952
+ "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
8953
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
8954
+ "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
8955
+ "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
8956
+ "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
8957
+ "mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
8958
+ "mmlu_eval_accuracy_high_school_world_history": 0.7692307692307693,
8959
+ "mmlu_eval_accuracy_human_aging": 0.5652173913043478,
8960
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
8961
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
8962
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
8963
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
8964
+ "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
8965
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
8966
+ "mmlu_eval_accuracy_marketing": 0.72,
8967
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
8968
+ "mmlu_eval_accuracy_miscellaneous": 0.7093023255813954,
8969
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
8970
+ "mmlu_eval_accuracy_moral_scenarios": 0.23,
8971
+ "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
8972
+ "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
8973
+ "mmlu_eval_accuracy_prehistory": 0.6285714285714286,
8974
+ "mmlu_eval_accuracy_professional_accounting": 0.3870967741935484,
8975
+ "mmlu_eval_accuracy_professional_law": 0.3941176470588235,
8976
+ "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
8977
+ "mmlu_eval_accuracy_professional_psychology": 0.4927536231884058,
8978
+ "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
8979
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
8980
+ "mmlu_eval_accuracy_sociology": 0.6818181818181818,
8981
+ "mmlu_eval_accuracy_us_foreign_policy": 0.9090909090909091,
8982
+ "mmlu_eval_accuracy_virology": 0.6111111111111112,
8983
+ "mmlu_eval_accuracy_world_religions": 0.5789473684210527,
8984
+ "mmlu_loss": 1.2881674546330155,
8985
+ "step": 9400
8986
+ },
8987
+ {
8988
+ "epoch": 2.89,
8989
+ "learning_rate": 0.0002,
8990
+ "loss": 0.7788,
8991
+ "step": 9410
8992
+ },
8993
+ {
8994
+ "epoch": 2.9,
8995
+ "learning_rate": 0.0002,
8996
+ "loss": 0.7699,
8997
+ "step": 9420
8998
+ },
8999
+ {
9000
+ "epoch": 2.9,
9001
+ "learning_rate": 0.0002,
9002
+ "loss": 0.7344,
9003
+ "step": 9430
9004
+ },
9005
+ {
9006
+ "epoch": 2.9,
9007
+ "learning_rate": 0.0002,
9008
+ "loss": 0.7598,
9009
+ "step": 9440
9010
+ },
9011
+ {
9012
+ "epoch": 2.91,
9013
+ "learning_rate": 0.0002,
9014
+ "loss": 0.7718,
9015
+ "step": 9450
9016
+ },
9017
+ {
9018
+ "epoch": 2.91,
9019
+ "learning_rate": 0.0002,
9020
+ "loss": 0.7487,
9021
+ "step": 9460
9022
+ },
9023
+ {
9024
+ "epoch": 2.91,
9025
+ "learning_rate": 0.0002,
9026
+ "loss": 0.704,
9027
+ "step": 9470
9028
+ },
9029
+ {
9030
+ "epoch": 2.92,
9031
+ "learning_rate": 0.0002,
9032
+ "loss": 0.7448,
9033
+ "step": 9480
9034
+ },
9035
+ {
9036
+ "epoch": 2.92,
9037
+ "learning_rate": 0.0002,
9038
+ "loss": 0.8239,
9039
+ "step": 9490
9040
+ },
9041
+ {
9042
+ "epoch": 2.92,
9043
+ "learning_rate": 0.0002,
9044
+ "loss": 0.8094,
9045
+ "step": 9500
9046
+ },
9047
+ {
9048
+ "epoch": 2.93,
9049
+ "learning_rate": 0.0002,
9050
+ "loss": 0.7615,
9051
+ "step": 9510
9052
+ },
9053
+ {
9054
+ "epoch": 2.93,
9055
+ "learning_rate": 0.0002,
9056
+ "loss": 0.7667,
9057
+ "step": 9520
9058
+ },
9059
+ {
9060
+ "epoch": 2.93,
9061
+ "learning_rate": 0.0002,
9062
+ "loss": 0.8003,
9063
+ "step": 9530
9064
+ },
9065
+ {
9066
+ "epoch": 2.93,
9067
+ "learning_rate": 0.0002,
9068
+ "loss": 0.7862,
9069
+ "step": 9540
9070
+ },
9071
+ {
9072
+ "epoch": 2.94,
9073
+ "learning_rate": 0.0002,
9074
+ "loss": 0.7704,
9075
+ "step": 9550
9076
+ },
9077
+ {
9078
+ "epoch": 2.94,
9079
+ "learning_rate": 0.0002,
9080
+ "loss": 0.7521,
9081
+ "step": 9560
9082
+ },
9083
+ {
9084
+ "epoch": 2.94,
9085
+ "learning_rate": 0.0002,
9086
+ "loss": 0.7913,
9087
+ "step": 9570
9088
+ },
9089
+ {
9090
+ "epoch": 2.95,
9091
+ "learning_rate": 0.0002,
9092
+ "loss": 0.7307,
9093
+ "step": 9580
9094
+ },
9095
+ {
9096
+ "epoch": 2.95,
9097
+ "learning_rate": 0.0002,
9098
+ "loss": 0.7812,
9099
+ "step": 9590
9100
+ },
9101
+ {
9102
+ "epoch": 2.95,
9103
+ "learning_rate": 0.0002,
9104
+ "loss": 0.7629,
9105
+ "step": 9600
9106
+ },
9107
+ {
9108
+ "epoch": 2.95,
9109
+ "eval_loss": 0.9446858167648315,
9110
+ "eval_runtime": 67.9228,
9111
+ "eval_samples_per_second": 14.723,
9112
+ "eval_steps_per_second": 7.361,
9113
+ "step": 9600
9114
+ },
9115
+ {
9116
+ "epoch": 2.95,
9117
+ "mmlu_eval_accuracy": 0.4982738948537338,
9118
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
9119
+ "mmlu_eval_accuracy_anatomy": 0.5,
9120
+ "mmlu_eval_accuracy_astronomy": 0.5,
9121
+ "mmlu_eval_accuracy_business_ethics": 0.36363636363636365,
9122
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
9123
+ "mmlu_eval_accuracy_college_biology": 0.375,
9124
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
9125
+ "mmlu_eval_accuracy_college_computer_science": 0.5454545454545454,
9126
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
9127
+ "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
9128
+ "mmlu_eval_accuracy_college_physics": 0.2727272727272727,
9129
+ "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
9130
+ "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
9131
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
9132
+ "mmlu_eval_accuracy_electrical_engineering": 0.1875,
9133
+ "mmlu_eval_accuracy_elementary_mathematics": 0.2926829268292683,
9134
+ "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
9135
+ "mmlu_eval_accuracy_global_facts": 0.5,
9136
+ "mmlu_eval_accuracy_high_school_biology": 0.46875,
9137
+ "mmlu_eval_accuracy_high_school_chemistry": 0.22727272727272727,
9138
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
9139
+ "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
9140
+ "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
9141
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.7142857142857143,
9142
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
9143
+ "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
9144
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5769230769230769,
9145
+ "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
9146
+ "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
9147
+ "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
9148
+ "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
9149
+ "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
9150
+ "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
9151
+ "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
9152
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
9153
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
9154
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
9155
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
9156
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
9157
+ "mmlu_eval_accuracy_marketing": 0.8,
9158
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
9159
+ "mmlu_eval_accuracy_miscellaneous": 0.7093023255813954,
9160
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
9161
+ "mmlu_eval_accuracy_moral_scenarios": 0.22,
9162
+ "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
9163
+ "mmlu_eval_accuracy_philosophy": 0.5588235294117647,
9164
+ "mmlu_eval_accuracy_prehistory": 0.6857142857142857,
9165
+ "mmlu_eval_accuracy_professional_accounting": 0.41935483870967744,
9166
+ "mmlu_eval_accuracy_professional_law": 0.3235294117647059,
9167
+ "mmlu_eval_accuracy_professional_medicine": 0.4838709677419355,
9168
+ "mmlu_eval_accuracy_professional_psychology": 0.463768115942029,
9169
+ "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
9170
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
9171
+ "mmlu_eval_accuracy_sociology": 0.6363636363636364,
9172
+ "mmlu_eval_accuracy_us_foreign_policy": 0.8181818181818182,
9173
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
9174
+ "mmlu_eval_accuracy_world_religions": 0.5789473684210527,
9175
+ "mmlu_loss": 1.2640136218491174,
9176
+ "step": 9600
9177
+ },
9178
+ {
9179
+ "epoch": 2.96,
9180
+ "learning_rate": 0.0002,
9181
+ "loss": 0.758,
9182
+ "step": 9610
9183
+ },
9184
+ {
9185
+ "epoch": 2.96,
9186
+ "learning_rate": 0.0002,
9187
+ "loss": 0.7651,
9188
+ "step": 9620
9189
+ },
9190
+ {
9191
+ "epoch": 2.96,
9192
+ "learning_rate": 0.0002,
9193
+ "loss": 0.7488,
9194
+ "step": 9630
9195
+ },
9196
+ {
9197
+ "epoch": 2.97,
9198
+ "learning_rate": 0.0002,
9199
+ "loss": 0.7588,
9200
+ "step": 9640
9201
+ },
9202
+ {
9203
+ "epoch": 2.97,
9204
+ "learning_rate": 0.0002,
9205
+ "loss": 0.7455,
9206
+ "step": 9650
9207
+ },
9208
+ {
9209
+ "epoch": 2.97,
9210
+ "learning_rate": 0.0002,
9211
+ "loss": 0.8135,
9212
+ "step": 9660
9213
+ },
9214
+ {
9215
+ "epoch": 2.97,
9216
+ "learning_rate": 0.0002,
9217
+ "loss": 0.7772,
9218
+ "step": 9670
9219
+ },
9220
+ {
9221
+ "epoch": 2.98,
9222
+ "learning_rate": 0.0002,
9223
+ "loss": 0.7787,
9224
+ "step": 9680
9225
+ },
9226
+ {
9227
+ "epoch": 2.98,
9228
+ "learning_rate": 0.0002,
9229
+ "loss": 0.6935,
9230
+ "step": 9690
9231
+ },
9232
+ {
9233
+ "epoch": 2.98,
9234
+ "learning_rate": 0.0002,
9235
+ "loss": 0.7499,
9236
+ "step": 9700
9237
+ },
9238
+ {
9239
+ "epoch": 2.99,
9240
+ "learning_rate": 0.0002,
9241
+ "loss": 0.7031,
9242
+ "step": 9710
9243
+ },
9244
+ {
9245
+ "epoch": 2.99,
9246
+ "learning_rate": 0.0002,
9247
+ "loss": 0.79,
9248
+ "step": 9720
9249
+ },
9250
+ {
9251
+ "epoch": 2.99,
9252
+ "learning_rate": 0.0002,
9253
+ "loss": 0.7383,
9254
+ "step": 9730
9255
+ },
9256
+ {
9257
+ "epoch": 3.0,
9258
+ "learning_rate": 0.0002,
9259
+ "loss": 0.7068,
9260
+ "step": 9740
9261
+ },
9262
+ {
9263
+ "epoch": 3.0,
9264
+ "learning_rate": 0.0002,
9265
+ "loss": 0.7816,
9266
+ "step": 9750
9267
+ },
9268
+ {
9269
+ "epoch": 3.0,
9270
+ "learning_rate": 0.0002,
9271
+ "loss": 0.7056,
9272
+ "step": 9760
9273
+ },
9274
+ {
9275
+ "epoch": 3.0,
9276
+ "learning_rate": 0.0002,
9277
+ "loss": 0.5936,
9278
+ "step": 9770
9279
+ },
9280
+ {
9281
+ "epoch": 3.01,
9282
+ "learning_rate": 0.0002,
9283
+ "loss": 0.5878,
9284
+ "step": 9780
9285
+ },
9286
+ {
9287
+ "epoch": 3.01,
9288
+ "learning_rate": 0.0002,
9289
+ "loss": 0.6266,
9290
+ "step": 9790
9291
+ },
9292
+ {
9293
+ "epoch": 3.01,
9294
+ "learning_rate": 0.0002,
9295
+ "loss": 0.6211,
9296
+ "step": 9800
9297
+ },
9298
+ {
9299
+ "epoch": 3.01,
9300
+ "eval_loss": 1.001284122467041,
9301
+ "eval_runtime": 67.6787,
9302
+ "eval_samples_per_second": 14.776,
9303
+ "eval_steps_per_second": 7.388,
9304
+ "step": 9800
9305
+ },
9306
+ {
9307
+ "epoch": 3.01,
9308
+ "mmlu_eval_accuracy": 0.5029741003299407,
9309
+ "mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
9310
+ "mmlu_eval_accuracy_anatomy": 0.5,
9311
+ "mmlu_eval_accuracy_astronomy": 0.5,
9312
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
9313
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
9314
+ "mmlu_eval_accuracy_college_biology": 0.4375,
9315
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
9316
+ "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
9317
+ "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
9318
+ "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
9319
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
9320
+ "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
9321
+ "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
9322
+ "mmlu_eval_accuracy_econometrics": 0.25,
9323
+ "mmlu_eval_accuracy_electrical_engineering": 0.1875,
9324
+ "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
9325
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
9326
+ "mmlu_eval_accuracy_global_facts": 0.5,
9327
+ "mmlu_eval_accuracy_high_school_biology": 0.46875,
9328
+ "mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
9329
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
9330
+ "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
9331
+ "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
9332
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
9333
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
9334
+ "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
9335
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
9336
+ "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
9337
+ "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
9338
+ "mmlu_eval_accuracy_high_school_statistics": 0.4782608695652174,
9339
+ "mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
9340
+ "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
9341
+ "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
9342
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
9343
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
9344
+ "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
9345
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
9346
+ "mmlu_eval_accuracy_machine_learning": 0.45454545454545453,
9347
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
9348
+ "mmlu_eval_accuracy_marketing": 0.76,
9349
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
9350
+ "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
9351
+ "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
9352
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
9353
+ "mmlu_eval_accuracy_nutrition": 0.5454545454545454,
9354
+ "mmlu_eval_accuracy_philosophy": 0.5882352941176471,
9355
+ "mmlu_eval_accuracy_prehistory": 0.6285714285714286,
9356
+ "mmlu_eval_accuracy_professional_accounting": 0.4838709677419355,
9357
+ "mmlu_eval_accuracy_professional_law": 0.3588235294117647,
9358
+ "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
9359
+ "mmlu_eval_accuracy_professional_psychology": 0.43478260869565216,
9360
+ "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
9361
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
9362
+ "mmlu_eval_accuracy_sociology": 0.7272727272727273,
9363
+ "mmlu_eval_accuracy_us_foreign_policy": 1.0,
9364
+ "mmlu_eval_accuracy_virology": 0.6111111111111112,
9365
+ "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
9366
+ "mmlu_loss": 1.34189098376985,
9367
+ "step": 9800
9368
+ },
9369
+ {
9370
+ "epoch": 3.02,
9371
+ "learning_rate": 0.0002,
9372
+ "loss": 0.5533,
9373
+ "step": 9810
9374
+ },
9375
+ {
9376
+ "epoch": 3.02,
9377
+ "learning_rate": 0.0002,
9378
+ "loss": 0.6007,
9379
+ "step": 9820
9380
+ },
9381
+ {
9382
+ "epoch": 3.02,
9383
+ "learning_rate": 0.0002,
9384
+ "loss": 0.6974,
9385
+ "step": 9830
9386
+ },
9387
+ {
9388
+ "epoch": 3.03,
9389
+ "learning_rate": 0.0002,
9390
+ "loss": 0.6468,
9391
+ "step": 9840
9392
+ },
9393
+ {
9394
+ "epoch": 3.03,
9395
+ "learning_rate": 0.0002,
9396
+ "loss": 0.5845,
9397
+ "step": 9850
9398
+ },
9399
+ {
9400
+ "epoch": 3.03,
9401
+ "learning_rate": 0.0002,
9402
+ "loss": 0.6218,
9403
+ "step": 9860
9404
+ },
9405
+ {
9406
+ "epoch": 3.04,
9407
+ "learning_rate": 0.0002,
9408
+ "loss": 0.5846,
9409
+ "step": 9870
9410
+ },
9411
+ {
9412
+ "epoch": 3.04,
9413
+ "learning_rate": 0.0002,
9414
+ "loss": 0.6122,
9415
+ "step": 9880
9416
+ },
9417
+ {
9418
+ "epoch": 3.04,
9419
+ "learning_rate": 0.0002,
9420
+ "loss": 0.6213,
9421
+ "step": 9890
9422
+ },
9423
+ {
9424
+ "epoch": 3.04,
9425
+ "learning_rate": 0.0002,
9426
+ "loss": 0.5899,
9427
+ "step": 9900
9428
+ },
9429
+ {
9430
+ "epoch": 3.05,
9431
+ "learning_rate": 0.0002,
9432
+ "loss": 0.6461,
9433
+ "step": 9910
9434
+ },
9435
+ {
9436
+ "epoch": 3.05,
9437
+ "learning_rate": 0.0002,
9438
+ "loss": 0.5862,
9439
+ "step": 9920
9440
+ },
9441
+ {
9442
+ "epoch": 3.05,
9443
+ "learning_rate": 0.0002,
9444
+ "loss": 0.6331,
9445
+ "step": 9930
9446
+ },
9447
+ {
9448
+ "epoch": 3.06,
9449
+ "learning_rate": 0.0002,
9450
+ "loss": 0.6468,
9451
+ "step": 9940
9452
+ },
9453
+ {
9454
+ "epoch": 3.06,
9455
+ "learning_rate": 0.0002,
9456
+ "loss": 0.5971,
9457
+ "step": 9950
9458
+ },
9459
+ {
9460
+ "epoch": 3.06,
9461
+ "learning_rate": 0.0002,
9462
+ "loss": 0.6529,
9463
+ "step": 9960
9464
+ },
9465
+ {
9466
+ "epoch": 3.07,
9467
+ "learning_rate": 0.0002,
9468
+ "loss": 0.5753,
9469
+ "step": 9970
9470
+ },
9471
+ {
9472
+ "epoch": 3.07,
9473
+ "learning_rate": 0.0002,
9474
+ "loss": 0.6048,
9475
+ "step": 9980
9476
+ },
9477
+ {
9478
+ "epoch": 3.07,
9479
+ "learning_rate": 0.0002,
9480
+ "loss": 0.6149,
9481
+ "step": 9990
9482
+ },
9483
+ {
9484
+ "epoch": 3.08,
9485
+ "learning_rate": 0.0002,
9486
+ "loss": 0.586,
9487
+ "step": 10000
9488
+ },
9489
+ {
9490
+ "epoch": 3.08,
9491
+ "eval_loss": 1.0243921279907227,
9492
+ "eval_runtime": 68.559,
9493
+ "eval_samples_per_second": 14.586,
9494
+ "eval_steps_per_second": 7.293,
9495
+ "step": 10000
9496
+ },
9497
+ {
9498
+ "epoch": 3.08,
9499
+ "mmlu_eval_accuracy": 0.4904185441435523,
9500
+ "mmlu_eval_accuracy_abstract_algebra": 0.18181818181818182,
9501
+ "mmlu_eval_accuracy_anatomy": 0.5,
9502
+ "mmlu_eval_accuracy_astronomy": 0.5,
9503
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
9504
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
9505
+ "mmlu_eval_accuracy_college_biology": 0.4375,
9506
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
9507
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
9508
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
9509
+ "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
9510
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
9511
+ "mmlu_eval_accuracy_computer_security": 0.2727272727272727,
9512
+ "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
9513
+ "mmlu_eval_accuracy_econometrics": 0.3333333333333333,
9514
+ "mmlu_eval_accuracy_electrical_engineering": 0.25,
9515
+ "mmlu_eval_accuracy_elementary_mathematics": 0.2926829268292683,
9516
+ "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
9517
+ "mmlu_eval_accuracy_global_facts": 0.4,
9518
+ "mmlu_eval_accuracy_high_school_biology": 0.46875,
9519
+ "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
9520
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
9521
+ "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
9522
+ "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
9523
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
9524
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
9525
+ "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
9526
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.6153846153846154,
9527
+ "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
9528
+ "mmlu_eval_accuracy_high_school_psychology": 0.8166666666666667,
9529
+ "mmlu_eval_accuracy_high_school_statistics": 0.4782608695652174,
9530
+ "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
9531
+ "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
9532
+ "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
9533
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
9534
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
9535
+ "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
9536
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
9537
+ "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
9538
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
9539
+ "mmlu_eval_accuracy_marketing": 0.76,
9540
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
9541
+ "mmlu_eval_accuracy_miscellaneous": 0.6976744186046512,
9542
+ "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
9543
+ "mmlu_eval_accuracy_moral_scenarios": 0.22,
9544
+ "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
9545
+ "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
9546
+ "mmlu_eval_accuracy_prehistory": 0.6,
9547
+ "mmlu_eval_accuracy_professional_accounting": 0.3870967741935484,
9548
+ "mmlu_eval_accuracy_professional_law": 0.34705882352941175,
9549
+ "mmlu_eval_accuracy_professional_medicine": 0.4838709677419355,
9550
+ "mmlu_eval_accuracy_professional_psychology": 0.43478260869565216,
9551
+ "mmlu_eval_accuracy_public_relations": 0.5,
9552
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
9553
+ "mmlu_eval_accuracy_sociology": 0.5909090909090909,
9554
+ "mmlu_eval_accuracy_us_foreign_policy": 0.9090909090909091,
9555
+ "mmlu_eval_accuracy_virology": 0.6666666666666666,
9556
+ "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
9557
+ "mmlu_loss": 1.2799786642077076,
9558
+ "step": 10000
9559
  }
9560
  ],
9561
  "max_steps": 10000,
9562
  "num_train_epochs": 4,
9563
+ "total_flos": 1.5821185598032773e+18,
9564
  "trial_name": null,
9565
  "trial_params": null
9566
  }
{checkpoint-8000 β†’ checkpoint-10000}/training_args.bin RENAMED
File without changes
checkpoint-5800/adapter_model/adapter_model/README.md CHANGED
@@ -213,6 +213,17 @@ The following `bitsandbytes` quantization config was used during training:
213
  - bnb_4bit_use_double_quant: True
214
  - bnb_4bit_compute_dtype: bfloat16
215
 
 
 
 
 
 
 
 
 
 
 
 
216
  The following `bitsandbytes` quantization config was used during training:
217
  - load_in_8bit: False
218
  - load_in_4bit: True
@@ -244,5 +255,6 @@ The following `bitsandbytes` quantization config was used during training:
244
  - PEFT 0.4.0
245
  - PEFT 0.4.0
246
  - PEFT 0.4.0
 
247
 
248
  - PEFT 0.4.0
 
213
  - bnb_4bit_use_double_quant: True
214
  - bnb_4bit_compute_dtype: bfloat16
215
 
216
+ The following `bitsandbytes` quantization config was used during training:
217
+ - load_in_8bit: False
218
+ - load_in_4bit: True
219
+ - llm_int8_threshold: 6.0
220
+ - llm_int8_skip_modules: None
221
+ - llm_int8_enable_fp32_cpu_offload: False
222
+ - llm_int8_has_fp16_weight: False
223
+ - bnb_4bit_quant_type: nf4
224
+ - bnb_4bit_use_double_quant: True
225
+ - bnb_4bit_compute_dtype: bfloat16
226
+
227
  The following `bitsandbytes` quantization config was used during training:
228
  - load_in_8bit: False
229
  - load_in_4bit: True
 
255
  - PEFT 0.4.0
256
  - PEFT 0.4.0
257
  - PEFT 0.4.0
258
+ - PEFT 0.4.0
259
 
260
  - PEFT 0.4.0
checkpoint-5800/adapter_model/adapter_model/adapter_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:be0db4bdea6fac10d53baf4da110713459762efaa4b85b28e5d0711cda88a0b9
3
  size 319977229
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:832fe1bf4c1262a05e4e8abdb9b9fcf332b3d23a12789e0e35934df54e3ebf69
3
  size 319977229