Farouk commited on
Commit
368567d
·
1 Parent(s): edff642

Training in progress, step 9200

Browse files
adapter_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5a3e6f5650fc22dbdb81ae971b3bb8fe6439b5f2b6101483f958db82024c669e
3
  size 319977229
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87f87e7f7d2175b326b550f9e309f8074663f7b3e501e003c8256e7d136ff432
3
  size 319977229
checkpoint-4000/adapter_model/adapter_model/README.md CHANGED
@@ -268,6 +268,17 @@ The following `bitsandbytes` quantization config was used during training:
268
  - bnb_4bit_use_double_quant: True
269
  - bnb_4bit_compute_dtype: bfloat16
270
 
 
 
 
 
 
 
 
 
 
 
 
271
  The following `bitsandbytes` quantization config was used during training:
272
  - load_in_8bit: False
273
  - load_in_4bit: True
@@ -304,5 +315,6 @@ The following `bitsandbytes` quantization config was used during training:
304
  - PEFT 0.4.0
305
  - PEFT 0.4.0
306
  - PEFT 0.4.0
 
307
 
308
  - PEFT 0.4.0
 
268
  - bnb_4bit_use_double_quant: True
269
  - bnb_4bit_compute_dtype: bfloat16
270
 
271
+ The following `bitsandbytes` quantization config was used during training:
272
+ - load_in_8bit: False
273
+ - load_in_4bit: True
274
+ - llm_int8_threshold: 6.0
275
+ - llm_int8_skip_modules: None
276
+ - llm_int8_enable_fp32_cpu_offload: False
277
+ - llm_int8_has_fp16_weight: False
278
+ - bnb_4bit_quant_type: nf4
279
+ - bnb_4bit_use_double_quant: True
280
+ - bnb_4bit_compute_dtype: bfloat16
281
+
282
  The following `bitsandbytes` quantization config was used during training:
283
  - load_in_8bit: False
284
  - load_in_4bit: True
 
315
  - PEFT 0.4.0
316
  - PEFT 0.4.0
317
  - PEFT 0.4.0
318
+ - PEFT 0.4.0
319
 
320
  - PEFT 0.4.0
checkpoint-4000/adapter_model/adapter_model/adapter_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:307f5730ad8f130fc3f8470de8759bf97fde4d7ced350b57e2c75cd290271047
3
  size 319977229
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5a3e6f5650fc22dbdb81ae971b3bb8fe6439b5f2b6101483f958db82024c669e
3
  size 319977229
{checkpoint-7200 → checkpoint-9200}/README.md RENAMED
File without changes
{checkpoint-7200 → checkpoint-9200}/adapter_config.json RENAMED
File without changes
{checkpoint-7200 → checkpoint-9200}/adapter_model.bin RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:7a432637143d79e71eaa7001020004b31ea27645f5b0370fb79e201988ec28be
3
  size 319977229
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87f87e7f7d2175b326b550f9e309f8074663f7b3e501e003c8256e7d136ff432
3
  size 319977229
{checkpoint-7200 → checkpoint-9200}/added_tokens.json RENAMED
File without changes
{checkpoint-7200 → checkpoint-9200}/optimizer.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d5b5f1e66aec7f76a5cc239e093bbd64ae415fabba764d2cb2e1f8d3ea0d814d
3
  size 1279539973
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:233a8b6446310c3e2bd7593aa1a25aca3829ac3efd8e3844f2f1284e3d153fde
3
  size 1279539973
{checkpoint-7200 → checkpoint-9200}/rng_state.pth RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b5133cdb814cf705adae88284f7502516bce89e6c896b5f3996ea5b10852b3e5
3
  size 14511
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:26e65a12a4953df74d75e86905ee6a8775af21406c87aff7b4425db759434833
3
  size 14511
{checkpoint-7200 → checkpoint-9200}/scheduler.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d4da7c28a3a535a6e57cf63c765db0f804917f5f393e9593c49a21cba57192b2
3
  size 627
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8cf1be3c846f4a44286600d19f1ec9319c4fc765ad8c1c089d3918122af94a49
3
  size 627
{checkpoint-7200 → checkpoint-9200}/special_tokens_map.json RENAMED
File without changes
{checkpoint-7200 → checkpoint-9200}/tokenizer.model RENAMED
File without changes
{checkpoint-7200 → checkpoint-9200}/tokenizer_config.json RENAMED
File without changes
{checkpoint-7200 → checkpoint-9200}/trainer_state.json RENAMED
@@ -1,8 +1,8 @@
1
  {
2
  "best_metric": 0.6935309767723083,
3
  "best_model_checkpoint": "experts/expert-9/checkpoint-4000",
4
- "epoch": 3.465286969077127,
5
- "global_step": 7200,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
@@ -6882,11 +6882,1921 @@
6882
  "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
6883
  "mmlu_loss": 1.116465316487199,
6884
  "step": 7200
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6885
  }
6886
  ],
6887
  "max_steps": 10000,
6888
  "num_train_epochs": 5,
6889
- "total_flos": 1.900445705122775e+18,
6890
  "trial_name": null,
6891
  "trial_params": null
6892
  }
 
1
  {
2
  "best_metric": 0.6935309767723083,
3
  "best_model_checkpoint": "experts/expert-9/checkpoint-4000",
4
+ "epoch": 4.427866682709662,
5
+ "global_step": 9200,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
 
6882
  "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
6883
  "mmlu_loss": 1.116465316487199,
6884
  "step": 7200
6885
+ },
6886
+ {
6887
+ "epoch": 3.47,
6888
+ "learning_rate": 0.0002,
6889
+ "loss": 0.3925,
6890
+ "step": 7210
6891
+ },
6892
+ {
6893
+ "epoch": 3.47,
6894
+ "learning_rate": 0.0002,
6895
+ "loss": 0.4807,
6896
+ "step": 7220
6897
+ },
6898
+ {
6899
+ "epoch": 3.48,
6900
+ "learning_rate": 0.0002,
6901
+ "loss": 0.4891,
6902
+ "step": 7230
6903
+ },
6904
+ {
6905
+ "epoch": 3.48,
6906
+ "learning_rate": 0.0002,
6907
+ "loss": 0.4816,
6908
+ "step": 7240
6909
+ },
6910
+ {
6911
+ "epoch": 3.49,
6912
+ "learning_rate": 0.0002,
6913
+ "loss": 0.4269,
6914
+ "step": 7250
6915
+ },
6916
+ {
6917
+ "epoch": 3.49,
6918
+ "learning_rate": 0.0002,
6919
+ "loss": 0.4331,
6920
+ "step": 7260
6921
+ },
6922
+ {
6923
+ "epoch": 3.5,
6924
+ "learning_rate": 0.0002,
6925
+ "loss": 0.4448,
6926
+ "step": 7270
6927
+ },
6928
+ {
6929
+ "epoch": 3.5,
6930
+ "learning_rate": 0.0002,
6931
+ "loss": 0.4801,
6932
+ "step": 7280
6933
+ },
6934
+ {
6935
+ "epoch": 3.51,
6936
+ "learning_rate": 0.0002,
6937
+ "loss": 0.4581,
6938
+ "step": 7290
6939
+ },
6940
+ {
6941
+ "epoch": 3.51,
6942
+ "learning_rate": 0.0002,
6943
+ "loss": 0.463,
6944
+ "step": 7300
6945
+ },
6946
+ {
6947
+ "epoch": 3.52,
6948
+ "learning_rate": 0.0002,
6949
+ "loss": 0.4134,
6950
+ "step": 7310
6951
+ },
6952
+ {
6953
+ "epoch": 3.52,
6954
+ "learning_rate": 0.0002,
6955
+ "loss": 0.503,
6956
+ "step": 7320
6957
+ },
6958
+ {
6959
+ "epoch": 3.53,
6960
+ "learning_rate": 0.0002,
6961
+ "loss": 0.4586,
6962
+ "step": 7330
6963
+ },
6964
+ {
6965
+ "epoch": 3.53,
6966
+ "learning_rate": 0.0002,
6967
+ "loss": 0.4563,
6968
+ "step": 7340
6969
+ },
6970
+ {
6971
+ "epoch": 3.54,
6972
+ "learning_rate": 0.0002,
6973
+ "loss": 0.4111,
6974
+ "step": 7350
6975
+ },
6976
+ {
6977
+ "epoch": 3.54,
6978
+ "learning_rate": 0.0002,
6979
+ "loss": 0.4922,
6980
+ "step": 7360
6981
+ },
6982
+ {
6983
+ "epoch": 3.55,
6984
+ "learning_rate": 0.0002,
6985
+ "loss": 0.431,
6986
+ "step": 7370
6987
+ },
6988
+ {
6989
+ "epoch": 3.55,
6990
+ "learning_rate": 0.0002,
6991
+ "loss": 0.4517,
6992
+ "step": 7380
6993
+ },
6994
+ {
6995
+ "epoch": 3.56,
6996
+ "learning_rate": 0.0002,
6997
+ "loss": 0.4292,
6998
+ "step": 7390
6999
+ },
7000
+ {
7001
+ "epoch": 3.56,
7002
+ "learning_rate": 0.0002,
7003
+ "loss": 0.4961,
7004
+ "step": 7400
7005
+ },
7006
+ {
7007
+ "epoch": 3.56,
7008
+ "eval_loss": 0.7399595975875854,
7009
+ "eval_runtime": 96.7334,
7010
+ "eval_samples_per_second": 10.338,
7011
+ "eval_steps_per_second": 5.169,
7012
+ "step": 7400
7013
+ },
7014
+ {
7015
+ "epoch": 3.56,
7016
+ "mmlu_eval_accuracy": 0.4827946170832024,
7017
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
7018
+ "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
7019
+ "mmlu_eval_accuracy_astronomy": 0.4375,
7020
+ "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
7021
+ "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
7022
+ "mmlu_eval_accuracy_college_biology": 0.5,
7023
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
7024
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
7025
+ "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
7026
+ "mmlu_eval_accuracy_college_medicine": 0.5454545454545454,
7027
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
7028
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
7029
+ "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
7030
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
7031
+ "mmlu_eval_accuracy_electrical_engineering": 0.3125,
7032
+ "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
7033
+ "mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
7034
+ "mmlu_eval_accuracy_global_facts": 0.4,
7035
+ "mmlu_eval_accuracy_high_school_biology": 0.4375,
7036
+ "mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
7037
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
7038
+ "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
7039
+ "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
7040
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.42857142857142855,
7041
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.5116279069767442,
7042
+ "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
7043
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
7044
+ "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
7045
+ "mmlu_eval_accuracy_high_school_psychology": 0.7666666666666667,
7046
+ "mmlu_eval_accuracy_high_school_statistics": 0.2608695652173913,
7047
+ "mmlu_eval_accuracy_high_school_us_history": 0.7727272727272727,
7048
+ "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
7049
+ "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
7050
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
7051
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
7052
+ "mmlu_eval_accuracy_jurisprudence": 0.18181818181818182,
7053
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
7054
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
7055
+ "mmlu_eval_accuracy_management": 0.5454545454545454,
7056
+ "mmlu_eval_accuracy_marketing": 0.8,
7057
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
7058
+ "mmlu_eval_accuracy_miscellaneous": 0.627906976744186,
7059
+ "mmlu_eval_accuracy_moral_disputes": 0.5526315789473685,
7060
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
7061
+ "mmlu_eval_accuracy_nutrition": 0.696969696969697,
7062
+ "mmlu_eval_accuracy_philosophy": 0.5,
7063
+ "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
7064
+ "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
7065
+ "mmlu_eval_accuracy_professional_law": 0.34705882352941175,
7066
+ "mmlu_eval_accuracy_professional_medicine": 0.6129032258064516,
7067
+ "mmlu_eval_accuracy_professional_psychology": 0.42028985507246375,
7068
+ "mmlu_eval_accuracy_public_relations": 0.5,
7069
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
7070
+ "mmlu_eval_accuracy_sociology": 0.6818181818181818,
7071
+ "mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
7072
+ "mmlu_eval_accuracy_virology": 0.4444444444444444,
7073
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
7074
+ "mmlu_loss": 1.0798200886920288,
7075
+ "step": 7400
7076
+ },
7077
+ {
7078
+ "epoch": 3.57,
7079
+ "learning_rate": 0.0002,
7080
+ "loss": 0.4556,
7081
+ "step": 7410
7082
+ },
7083
+ {
7084
+ "epoch": 3.57,
7085
+ "learning_rate": 0.0002,
7086
+ "loss": 0.4559,
7087
+ "step": 7420
7088
+ },
7089
+ {
7090
+ "epoch": 3.58,
7091
+ "learning_rate": 0.0002,
7092
+ "loss": 0.4844,
7093
+ "step": 7430
7094
+ },
7095
+ {
7096
+ "epoch": 3.58,
7097
+ "learning_rate": 0.0002,
7098
+ "loss": 0.4668,
7099
+ "step": 7440
7100
+ },
7101
+ {
7102
+ "epoch": 3.59,
7103
+ "learning_rate": 0.0002,
7104
+ "loss": 0.5052,
7105
+ "step": 7450
7106
+ },
7107
+ {
7108
+ "epoch": 3.59,
7109
+ "learning_rate": 0.0002,
7110
+ "loss": 0.4001,
7111
+ "step": 7460
7112
+ },
7113
+ {
7114
+ "epoch": 3.6,
7115
+ "learning_rate": 0.0002,
7116
+ "loss": 0.4503,
7117
+ "step": 7470
7118
+ },
7119
+ {
7120
+ "epoch": 3.6,
7121
+ "learning_rate": 0.0002,
7122
+ "loss": 0.451,
7123
+ "step": 7480
7124
+ },
7125
+ {
7126
+ "epoch": 3.6,
7127
+ "learning_rate": 0.0002,
7128
+ "loss": 0.4258,
7129
+ "step": 7490
7130
+ },
7131
+ {
7132
+ "epoch": 3.61,
7133
+ "learning_rate": 0.0002,
7134
+ "loss": 0.4727,
7135
+ "step": 7500
7136
+ },
7137
+ {
7138
+ "epoch": 3.61,
7139
+ "learning_rate": 0.0002,
7140
+ "loss": 0.4347,
7141
+ "step": 7510
7142
+ },
7143
+ {
7144
+ "epoch": 3.62,
7145
+ "learning_rate": 0.0002,
7146
+ "loss": 0.4837,
7147
+ "step": 7520
7148
+ },
7149
+ {
7150
+ "epoch": 3.62,
7151
+ "learning_rate": 0.0002,
7152
+ "loss": 0.4274,
7153
+ "step": 7530
7154
+ },
7155
+ {
7156
+ "epoch": 3.63,
7157
+ "learning_rate": 0.0002,
7158
+ "loss": 0.422,
7159
+ "step": 7540
7160
+ },
7161
+ {
7162
+ "epoch": 3.63,
7163
+ "learning_rate": 0.0002,
7164
+ "loss": 0.4856,
7165
+ "step": 7550
7166
+ },
7167
+ {
7168
+ "epoch": 3.64,
7169
+ "learning_rate": 0.0002,
7170
+ "loss": 0.5021,
7171
+ "step": 7560
7172
+ },
7173
+ {
7174
+ "epoch": 3.64,
7175
+ "learning_rate": 0.0002,
7176
+ "loss": 0.4716,
7177
+ "step": 7570
7178
+ },
7179
+ {
7180
+ "epoch": 3.65,
7181
+ "learning_rate": 0.0002,
7182
+ "loss": 0.4615,
7183
+ "step": 7580
7184
+ },
7185
+ {
7186
+ "epoch": 3.65,
7187
+ "learning_rate": 0.0002,
7188
+ "loss": 0.4498,
7189
+ "step": 7590
7190
+ },
7191
+ {
7192
+ "epoch": 3.66,
7193
+ "learning_rate": 0.0002,
7194
+ "loss": 0.4309,
7195
+ "step": 7600
7196
+ },
7197
+ {
7198
+ "epoch": 3.66,
7199
+ "eval_loss": 0.7468517422676086,
7200
+ "eval_runtime": 96.6541,
7201
+ "eval_samples_per_second": 10.346,
7202
+ "eval_steps_per_second": 5.173,
7203
+ "step": 7600
7204
+ },
7205
+ {
7206
+ "epoch": 3.66,
7207
+ "mmlu_eval_accuracy": 0.4964692181052332,
7208
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
7209
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
7210
+ "mmlu_eval_accuracy_astronomy": 0.375,
7211
+ "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
7212
+ "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
7213
+ "mmlu_eval_accuracy_college_biology": 0.4375,
7214
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
7215
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
7216
+ "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
7217
+ "mmlu_eval_accuracy_college_medicine": 0.5909090909090909,
7218
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
7219
+ "mmlu_eval_accuracy_computer_security": 0.6363636363636364,
7220
+ "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
7221
+ "mmlu_eval_accuracy_econometrics": 0.25,
7222
+ "mmlu_eval_accuracy_electrical_engineering": 0.25,
7223
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
7224
+ "mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
7225
+ "mmlu_eval_accuracy_global_facts": 0.5,
7226
+ "mmlu_eval_accuracy_high_school_biology": 0.4375,
7227
+ "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
7228
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
7229
+ "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
7230
+ "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
7231
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.47619047619047616,
7232
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.5116279069767442,
7233
+ "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
7234
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
7235
+ "mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
7236
+ "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
7237
+ "mmlu_eval_accuracy_high_school_statistics": 0.2608695652173913,
7238
+ "mmlu_eval_accuracy_high_school_us_history": 0.7727272727272727,
7239
+ "mmlu_eval_accuracy_high_school_world_history": 0.7692307692307693,
7240
+ "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
7241
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
7242
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
7243
+ "mmlu_eval_accuracy_jurisprudence": 0.18181818181818182,
7244
+ "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
7245
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
7246
+ "mmlu_eval_accuracy_management": 0.5454545454545454,
7247
+ "mmlu_eval_accuracy_marketing": 0.88,
7248
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
7249
+ "mmlu_eval_accuracy_miscellaneous": 0.6395348837209303,
7250
+ "mmlu_eval_accuracy_moral_disputes": 0.5789473684210527,
7251
+ "mmlu_eval_accuracy_moral_scenarios": 0.29,
7252
+ "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
7253
+ "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
7254
+ "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
7255
+ "mmlu_eval_accuracy_professional_accounting": 0.3870967741935484,
7256
+ "mmlu_eval_accuracy_professional_law": 0.3235294117647059,
7257
+ "mmlu_eval_accuracy_professional_medicine": 0.6451612903225806,
7258
+ "mmlu_eval_accuracy_professional_psychology": 0.4492753623188406,
7259
+ "mmlu_eval_accuracy_public_relations": 0.5,
7260
+ "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
7261
+ "mmlu_eval_accuracy_sociology": 0.6818181818181818,
7262
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
7263
+ "mmlu_eval_accuracy_virology": 0.5,
7264
+ "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
7265
+ "mmlu_loss": 1.0270358441402412,
7266
+ "step": 7600
7267
+ },
7268
+ {
7269
+ "epoch": 3.66,
7270
+ "learning_rate": 0.0002,
7271
+ "loss": 0.444,
7272
+ "step": 7610
7273
+ },
7274
+ {
7275
+ "epoch": 3.67,
7276
+ "learning_rate": 0.0002,
7277
+ "loss": 0.4273,
7278
+ "step": 7620
7279
+ },
7280
+ {
7281
+ "epoch": 3.67,
7282
+ "learning_rate": 0.0002,
7283
+ "loss": 0.4883,
7284
+ "step": 7630
7285
+ },
7286
+ {
7287
+ "epoch": 3.68,
7288
+ "learning_rate": 0.0002,
7289
+ "loss": 0.4807,
7290
+ "step": 7640
7291
+ },
7292
+ {
7293
+ "epoch": 3.68,
7294
+ "learning_rate": 0.0002,
7295
+ "loss": 0.4485,
7296
+ "step": 7650
7297
+ },
7298
+ {
7299
+ "epoch": 3.69,
7300
+ "learning_rate": 0.0002,
7301
+ "loss": 0.452,
7302
+ "step": 7660
7303
+ },
7304
+ {
7305
+ "epoch": 3.69,
7306
+ "learning_rate": 0.0002,
7307
+ "loss": 0.4342,
7308
+ "step": 7670
7309
+ },
7310
+ {
7311
+ "epoch": 3.7,
7312
+ "learning_rate": 0.0002,
7313
+ "loss": 0.4677,
7314
+ "step": 7680
7315
+ },
7316
+ {
7317
+ "epoch": 3.7,
7318
+ "learning_rate": 0.0002,
7319
+ "loss": 0.4522,
7320
+ "step": 7690
7321
+ },
7322
+ {
7323
+ "epoch": 3.71,
7324
+ "learning_rate": 0.0002,
7325
+ "loss": 0.4342,
7326
+ "step": 7700
7327
+ },
7328
+ {
7329
+ "epoch": 3.71,
7330
+ "learning_rate": 0.0002,
7331
+ "loss": 0.4008,
7332
+ "step": 7710
7333
+ },
7334
+ {
7335
+ "epoch": 3.72,
7336
+ "learning_rate": 0.0002,
7337
+ "loss": 0.4661,
7338
+ "step": 7720
7339
+ },
7340
+ {
7341
+ "epoch": 3.72,
7342
+ "learning_rate": 0.0002,
7343
+ "loss": 0.4148,
7344
+ "step": 7730
7345
+ },
7346
+ {
7347
+ "epoch": 3.73,
7348
+ "learning_rate": 0.0002,
7349
+ "loss": 0.4199,
7350
+ "step": 7740
7351
+ },
7352
+ {
7353
+ "epoch": 3.73,
7354
+ "learning_rate": 0.0002,
7355
+ "loss": 0.4507,
7356
+ "step": 7750
7357
+ },
7358
+ {
7359
+ "epoch": 3.73,
7360
+ "learning_rate": 0.0002,
7361
+ "loss": 0.4365,
7362
+ "step": 7760
7363
+ },
7364
+ {
7365
+ "epoch": 3.74,
7366
+ "learning_rate": 0.0002,
7367
+ "loss": 0.458,
7368
+ "step": 7770
7369
+ },
7370
+ {
7371
+ "epoch": 3.74,
7372
+ "learning_rate": 0.0002,
7373
+ "loss": 0.4213,
7374
+ "step": 7780
7375
+ },
7376
+ {
7377
+ "epoch": 3.75,
7378
+ "learning_rate": 0.0002,
7379
+ "loss": 0.4138,
7380
+ "step": 7790
7381
+ },
7382
+ {
7383
+ "epoch": 3.75,
7384
+ "learning_rate": 0.0002,
7385
+ "loss": 0.4995,
7386
+ "step": 7800
7387
+ },
7388
+ {
7389
+ "epoch": 3.75,
7390
+ "eval_loss": 0.7374092936515808,
7391
+ "eval_runtime": 96.6333,
7392
+ "eval_samples_per_second": 10.348,
7393
+ "eval_steps_per_second": 5.174,
7394
+ "step": 7800
7395
+ },
7396
+ {
7397
+ "epoch": 3.75,
7398
+ "mmlu_eval_accuracy": 0.49141637575450603,
7399
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
7400
+ "mmlu_eval_accuracy_anatomy": 0.5,
7401
+ "mmlu_eval_accuracy_astronomy": 0.375,
7402
+ "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
7403
+ "mmlu_eval_accuracy_clinical_knowledge": 0.4482758620689655,
7404
+ "mmlu_eval_accuracy_college_biology": 0.5,
7405
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
7406
+ "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
7407
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
7408
+ "mmlu_eval_accuracy_college_medicine": 0.5454545454545454,
7409
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
7410
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
7411
+ "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
7412
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
7413
+ "mmlu_eval_accuracy_electrical_engineering": 0.3125,
7414
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
7415
+ "mmlu_eval_accuracy_formal_logic": 0.42857142857142855,
7416
+ "mmlu_eval_accuracy_global_facts": 0.4,
7417
+ "mmlu_eval_accuracy_high_school_biology": 0.5,
7418
+ "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
7419
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
7420
+ "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
7421
+ "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
7422
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.47619047619047616,
7423
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4883720930232558,
7424
+ "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
7425
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.3076923076923077,
7426
+ "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
7427
+ "mmlu_eval_accuracy_high_school_psychology": 0.8166666666666667,
7428
+ "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
7429
+ "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
7430
+ "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
7431
+ "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
7432
+ "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
7433
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
7434
+ "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
7435
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
7436
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
7437
+ "mmlu_eval_accuracy_management": 0.5454545454545454,
7438
+ "mmlu_eval_accuracy_marketing": 0.84,
7439
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
7440
+ "mmlu_eval_accuracy_miscellaneous": 0.6395348837209303,
7441
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
7442
+ "mmlu_eval_accuracy_moral_scenarios": 0.23,
7443
+ "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
7444
+ "mmlu_eval_accuracy_philosophy": 0.5,
7445
+ "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
7446
+ "mmlu_eval_accuracy_professional_accounting": 0.3870967741935484,
7447
+ "mmlu_eval_accuracy_professional_law": 0.3235294117647059,
7448
+ "mmlu_eval_accuracy_professional_medicine": 0.6451612903225806,
7449
+ "mmlu_eval_accuracy_professional_psychology": 0.43478260869565216,
7450
+ "mmlu_eval_accuracy_public_relations": 0.5,
7451
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
7452
+ "mmlu_eval_accuracy_sociology": 0.6818181818181818,
7453
+ "mmlu_eval_accuracy_us_foreign_policy": 0.8181818181818182,
7454
+ "mmlu_eval_accuracy_virology": 0.5,
7455
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
7456
+ "mmlu_loss": 1.1190382100486103,
7457
+ "step": 7800
7458
+ },
7459
+ {
7460
+ "epoch": 3.76,
7461
+ "learning_rate": 0.0002,
7462
+ "loss": 0.5019,
7463
+ "step": 7810
7464
+ },
7465
+ {
7466
+ "epoch": 3.76,
7467
+ "learning_rate": 0.0002,
7468
+ "loss": 0.4398,
7469
+ "step": 7820
7470
+ },
7471
+ {
7472
+ "epoch": 3.77,
7473
+ "learning_rate": 0.0002,
7474
+ "loss": 0.4795,
7475
+ "step": 7830
7476
+ },
7477
+ {
7478
+ "epoch": 3.77,
7479
+ "learning_rate": 0.0002,
7480
+ "loss": 0.4937,
7481
+ "step": 7840
7482
+ },
7483
+ {
7484
+ "epoch": 3.78,
7485
+ "learning_rate": 0.0002,
7486
+ "loss": 0.4307,
7487
+ "step": 7850
7488
+ },
7489
+ {
7490
+ "epoch": 3.78,
7491
+ "learning_rate": 0.0002,
7492
+ "loss": 0.4578,
7493
+ "step": 7860
7494
+ },
7495
+ {
7496
+ "epoch": 3.79,
7497
+ "learning_rate": 0.0002,
7498
+ "loss": 0.4428,
7499
+ "step": 7870
7500
+ },
7501
+ {
7502
+ "epoch": 3.79,
7503
+ "learning_rate": 0.0002,
7504
+ "loss": 0.5122,
7505
+ "step": 7880
7506
+ },
7507
+ {
7508
+ "epoch": 3.8,
7509
+ "learning_rate": 0.0002,
7510
+ "loss": 0.4255,
7511
+ "step": 7890
7512
+ },
7513
+ {
7514
+ "epoch": 3.8,
7515
+ "learning_rate": 0.0002,
7516
+ "loss": 0.4854,
7517
+ "step": 7900
7518
+ },
7519
+ {
7520
+ "epoch": 3.81,
7521
+ "learning_rate": 0.0002,
7522
+ "loss": 0.4914,
7523
+ "step": 7910
7524
+ },
7525
+ {
7526
+ "epoch": 3.81,
7527
+ "learning_rate": 0.0002,
7528
+ "loss": 0.4744,
7529
+ "step": 7920
7530
+ },
7531
+ {
7532
+ "epoch": 3.82,
7533
+ "learning_rate": 0.0002,
7534
+ "loss": 0.4631,
7535
+ "step": 7930
7536
+ },
7537
+ {
7538
+ "epoch": 3.82,
7539
+ "learning_rate": 0.0002,
7540
+ "loss": 0.5065,
7541
+ "step": 7940
7542
+ },
7543
+ {
7544
+ "epoch": 3.83,
7545
+ "learning_rate": 0.0002,
7546
+ "loss": 0.4729,
7547
+ "step": 7950
7548
+ },
7549
+ {
7550
+ "epoch": 3.83,
7551
+ "learning_rate": 0.0002,
7552
+ "loss": 0.4387,
7553
+ "step": 7960
7554
+ },
7555
+ {
7556
+ "epoch": 3.84,
7557
+ "learning_rate": 0.0002,
7558
+ "loss": 0.4813,
7559
+ "step": 7970
7560
+ },
7561
+ {
7562
+ "epoch": 3.84,
7563
+ "learning_rate": 0.0002,
7564
+ "loss": 0.4584,
7565
+ "step": 7980
7566
+ },
7567
+ {
7568
+ "epoch": 3.85,
7569
+ "learning_rate": 0.0002,
7570
+ "loss": 0.4704,
7571
+ "step": 7990
7572
+ },
7573
+ {
7574
+ "epoch": 3.85,
7575
+ "learning_rate": 0.0002,
7576
+ "loss": 0.5204,
7577
+ "step": 8000
7578
+ },
7579
+ {
7580
+ "epoch": 3.85,
7581
+ "eval_loss": 0.7310526371002197,
7582
+ "eval_runtime": 96.7143,
7583
+ "eval_samples_per_second": 10.34,
7584
+ "eval_steps_per_second": 5.17,
7585
+ "step": 8000
7586
+ },
7587
+ {
7588
+ "epoch": 3.85,
7589
+ "mmlu_eval_accuracy": 0.4895440752854303,
7590
+ "mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
7591
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
7592
+ "mmlu_eval_accuracy_astronomy": 0.4375,
7593
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
7594
+ "mmlu_eval_accuracy_clinical_knowledge": 0.4482758620689655,
7595
+ "mmlu_eval_accuracy_college_biology": 0.5625,
7596
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
7597
+ "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
7598
+ "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
7599
+ "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
7600
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
7601
+ "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
7602
+ "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
7603
+ "mmlu_eval_accuracy_econometrics": 0.25,
7604
+ "mmlu_eval_accuracy_electrical_engineering": 0.3125,
7605
+ "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
7606
+ "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
7607
+ "mmlu_eval_accuracy_global_facts": 0.5,
7608
+ "mmlu_eval_accuracy_high_school_biology": 0.46875,
7609
+ "mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
7610
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
7611
+ "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
7612
+ "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
7613
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.47619047619047616,
7614
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.46511627906976744,
7615
+ "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
7616
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.38461538461538464,
7617
+ "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
7618
+ "mmlu_eval_accuracy_high_school_psychology": 0.7833333333333333,
7619
+ "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
7620
+ "mmlu_eval_accuracy_high_school_us_history": 0.7727272727272727,
7621
+ "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
7622
+ "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
7623
+ "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
7624
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
7625
+ "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
7626
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
7627
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
7628
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
7629
+ "mmlu_eval_accuracy_marketing": 0.84,
7630
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
7631
+ "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
7632
+ "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
7633
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
7634
+ "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
7635
+ "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
7636
+ "mmlu_eval_accuracy_prehistory": 0.4,
7637
+ "mmlu_eval_accuracy_professional_accounting": 0.3870967741935484,
7638
+ "mmlu_eval_accuracy_professional_law": 0.31176470588235294,
7639
+ "mmlu_eval_accuracy_professional_medicine": 0.6451612903225806,
7640
+ "mmlu_eval_accuracy_professional_psychology": 0.4492753623188406,
7641
+ "mmlu_eval_accuracy_public_relations": 0.5,
7642
+ "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
7643
+ "mmlu_eval_accuracy_sociology": 0.5909090909090909,
7644
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
7645
+ "mmlu_eval_accuracy_virology": 0.4444444444444444,
7646
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
7647
+ "mmlu_loss": 1.043521719379229,
7648
+ "step": 8000
7649
+ },
7650
+ {
7651
+ "epoch": 3.86,
7652
+ "learning_rate": 0.0002,
7653
+ "loss": 0.4623,
7654
+ "step": 8010
7655
+ },
7656
+ {
7657
+ "epoch": 3.86,
7658
+ "learning_rate": 0.0002,
7659
+ "loss": 0.4379,
7660
+ "step": 8020
7661
+ },
7662
+ {
7663
+ "epoch": 3.86,
7664
+ "learning_rate": 0.0002,
7665
+ "loss": 0.4636,
7666
+ "step": 8030
7667
+ },
7668
+ {
7669
+ "epoch": 3.87,
7670
+ "learning_rate": 0.0002,
7671
+ "loss": 0.4402,
7672
+ "step": 8040
7673
+ },
7674
+ {
7675
+ "epoch": 3.87,
7676
+ "learning_rate": 0.0002,
7677
+ "loss": 0.4416,
7678
+ "step": 8050
7679
+ },
7680
+ {
7681
+ "epoch": 3.88,
7682
+ "learning_rate": 0.0002,
7683
+ "loss": 0.4826,
7684
+ "step": 8060
7685
+ },
7686
+ {
7687
+ "epoch": 3.88,
7688
+ "learning_rate": 0.0002,
7689
+ "loss": 0.4179,
7690
+ "step": 8070
7691
+ },
7692
+ {
7693
+ "epoch": 3.89,
7694
+ "learning_rate": 0.0002,
7695
+ "loss": 0.4354,
7696
+ "step": 8080
7697
+ },
7698
+ {
7699
+ "epoch": 3.89,
7700
+ "learning_rate": 0.0002,
7701
+ "loss": 0.4763,
7702
+ "step": 8090
7703
+ },
7704
+ {
7705
+ "epoch": 3.9,
7706
+ "learning_rate": 0.0002,
7707
+ "loss": 0.4451,
7708
+ "step": 8100
7709
+ },
7710
+ {
7711
+ "epoch": 3.9,
7712
+ "learning_rate": 0.0002,
7713
+ "loss": 0.4896,
7714
+ "step": 8110
7715
+ },
7716
+ {
7717
+ "epoch": 3.91,
7718
+ "learning_rate": 0.0002,
7719
+ "loss": 0.4496,
7720
+ "step": 8120
7721
+ },
7722
+ {
7723
+ "epoch": 3.91,
7724
+ "learning_rate": 0.0002,
7725
+ "loss": 0.4513,
7726
+ "step": 8130
7727
+ },
7728
+ {
7729
+ "epoch": 3.92,
7730
+ "learning_rate": 0.0002,
7731
+ "loss": 0.4528,
7732
+ "step": 8140
7733
+ },
7734
+ {
7735
+ "epoch": 3.92,
7736
+ "learning_rate": 0.0002,
7737
+ "loss": 0.4633,
7738
+ "step": 8150
7739
+ },
7740
+ {
7741
+ "epoch": 3.93,
7742
+ "learning_rate": 0.0002,
7743
+ "loss": 0.4697,
7744
+ "step": 8160
7745
+ },
7746
+ {
7747
+ "epoch": 3.93,
7748
+ "learning_rate": 0.0002,
7749
+ "loss": 0.476,
7750
+ "step": 8170
7751
+ },
7752
+ {
7753
+ "epoch": 3.94,
7754
+ "learning_rate": 0.0002,
7755
+ "loss": 0.5351,
7756
+ "step": 8180
7757
+ },
7758
+ {
7759
+ "epoch": 3.94,
7760
+ "learning_rate": 0.0002,
7761
+ "loss": 0.4989,
7762
+ "step": 8190
7763
+ },
7764
+ {
7765
+ "epoch": 3.95,
7766
+ "learning_rate": 0.0002,
7767
+ "loss": 0.4607,
7768
+ "step": 8200
7769
+ },
7770
+ {
7771
+ "epoch": 3.95,
7772
+ "eval_loss": 0.735032320022583,
7773
+ "eval_runtime": 96.6471,
7774
+ "eval_samples_per_second": 10.347,
7775
+ "eval_steps_per_second": 5.173,
7776
+ "step": 8200
7777
+ },
7778
+ {
7779
+ "epoch": 3.95,
7780
+ "mmlu_eval_accuracy": 0.4851314842694453,
7781
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
7782
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
7783
+ "mmlu_eval_accuracy_astronomy": 0.4375,
7784
+ "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
7785
+ "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
7786
+ "mmlu_eval_accuracy_college_biology": 0.4375,
7787
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
7788
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
7789
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
7790
+ "mmlu_eval_accuracy_college_medicine": 0.5454545454545454,
7791
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
7792
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
7793
+ "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
7794
+ "mmlu_eval_accuracy_econometrics": 0.3333333333333333,
7795
+ "mmlu_eval_accuracy_electrical_engineering": 0.3125,
7796
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
7797
+ "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
7798
+ "mmlu_eval_accuracy_global_facts": 0.3,
7799
+ "mmlu_eval_accuracy_high_school_biology": 0.40625,
7800
+ "mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
7801
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
7802
+ "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
7803
+ "mmlu_eval_accuracy_high_school_geography": 0.6818181818181818,
7804
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
7805
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
7806
+ "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
7807
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
7808
+ "mmlu_eval_accuracy_high_school_physics": 0.4117647058823529,
7809
+ "mmlu_eval_accuracy_high_school_psychology": 0.7666666666666667,
7810
+ "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
7811
+ "mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
7812
+ "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
7813
+ "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
7814
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
7815
+ "mmlu_eval_accuracy_international_law": 0.7692307692307693,
7816
+ "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
7817
+ "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
7818
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
7819
+ "mmlu_eval_accuracy_management": 0.5454545454545454,
7820
+ "mmlu_eval_accuracy_marketing": 0.84,
7821
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
7822
+ "mmlu_eval_accuracy_miscellaneous": 0.627906976744186,
7823
+ "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
7824
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
7825
+ "mmlu_eval_accuracy_nutrition": 0.696969696969697,
7826
+ "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
7827
+ "mmlu_eval_accuracy_prehistory": 0.4,
7828
+ "mmlu_eval_accuracy_professional_accounting": 0.3870967741935484,
7829
+ "mmlu_eval_accuracy_professional_law": 0.35294117647058826,
7830
+ "mmlu_eval_accuracy_professional_medicine": 0.6774193548387096,
7831
+ "mmlu_eval_accuracy_professional_psychology": 0.4057971014492754,
7832
+ "mmlu_eval_accuracy_public_relations": 0.5,
7833
+ "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
7834
+ "mmlu_eval_accuracy_sociology": 0.6363636363636364,
7835
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
7836
+ "mmlu_eval_accuracy_virology": 0.4444444444444444,
7837
+ "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
7838
+ "mmlu_loss": 1.0230581155800633,
7839
+ "step": 8200
7840
+ },
7841
+ {
7842
+ "epoch": 3.95,
7843
+ "learning_rate": 0.0002,
7844
+ "loss": 0.442,
7845
+ "step": 8210
7846
+ },
7847
+ {
7848
+ "epoch": 3.96,
7849
+ "learning_rate": 0.0002,
7850
+ "loss": 0.4797,
7851
+ "step": 8220
7852
+ },
7853
+ {
7854
+ "epoch": 3.96,
7855
+ "learning_rate": 0.0002,
7856
+ "loss": 0.4999,
7857
+ "step": 8230
7858
+ },
7859
+ {
7860
+ "epoch": 3.97,
7861
+ "learning_rate": 0.0002,
7862
+ "loss": 0.4891,
7863
+ "step": 8240
7864
+ },
7865
+ {
7866
+ "epoch": 3.97,
7867
+ "learning_rate": 0.0002,
7868
+ "loss": 0.5017,
7869
+ "step": 8250
7870
+ },
7871
+ {
7872
+ "epoch": 3.98,
7873
+ "learning_rate": 0.0002,
7874
+ "loss": 0.461,
7875
+ "step": 8260
7876
+ },
7877
+ {
7878
+ "epoch": 3.98,
7879
+ "learning_rate": 0.0002,
7880
+ "loss": 0.4616,
7881
+ "step": 8270
7882
+ },
7883
+ {
7884
+ "epoch": 3.99,
7885
+ "learning_rate": 0.0002,
7886
+ "loss": 0.4803,
7887
+ "step": 8280
7888
+ },
7889
+ {
7890
+ "epoch": 3.99,
7891
+ "learning_rate": 0.0002,
7892
+ "loss": 0.4601,
7893
+ "step": 8290
7894
+ },
7895
+ {
7896
+ "epoch": 3.99,
7897
+ "learning_rate": 0.0002,
7898
+ "loss": 0.4645,
7899
+ "step": 8300
7900
+ },
7901
+ {
7902
+ "epoch": 4.0,
7903
+ "learning_rate": 0.0002,
7904
+ "loss": 0.4591,
7905
+ "step": 8310
7906
+ },
7907
+ {
7908
+ "epoch": 4.0,
7909
+ "learning_rate": 0.0002,
7910
+ "loss": 0.3653,
7911
+ "step": 8320
7912
+ },
7913
+ {
7914
+ "epoch": 4.01,
7915
+ "learning_rate": 0.0002,
7916
+ "loss": 0.3672,
7917
+ "step": 8330
7918
+ },
7919
+ {
7920
+ "epoch": 4.01,
7921
+ "learning_rate": 0.0002,
7922
+ "loss": 0.3526,
7923
+ "step": 8340
7924
+ },
7925
+ {
7926
+ "epoch": 4.02,
7927
+ "learning_rate": 0.0002,
7928
+ "loss": 0.3478,
7929
+ "step": 8350
7930
+ },
7931
+ {
7932
+ "epoch": 4.02,
7933
+ "learning_rate": 0.0002,
7934
+ "loss": 0.3005,
7935
+ "step": 8360
7936
+ },
7937
+ {
7938
+ "epoch": 4.03,
7939
+ "learning_rate": 0.0002,
7940
+ "loss": 0.3679,
7941
+ "step": 8370
7942
+ },
7943
+ {
7944
+ "epoch": 4.03,
7945
+ "learning_rate": 0.0002,
7946
+ "loss": 0.3759,
7947
+ "step": 8380
7948
+ },
7949
+ {
7950
+ "epoch": 4.04,
7951
+ "learning_rate": 0.0002,
7952
+ "loss": 0.2821,
7953
+ "step": 8390
7954
+ },
7955
+ {
7956
+ "epoch": 4.04,
7957
+ "learning_rate": 0.0002,
7958
+ "loss": 0.3584,
7959
+ "step": 8400
7960
+ },
7961
+ {
7962
+ "epoch": 4.04,
7963
+ "eval_loss": 0.8006640672683716,
7964
+ "eval_runtime": 96.5052,
7965
+ "eval_samples_per_second": 10.362,
7966
+ "eval_steps_per_second": 5.181,
7967
+ "step": 8400
7968
+ },
7969
+ {
7970
+ "epoch": 4.04,
7971
+ "mmlu_eval_accuracy": 0.4863456433402196,
7972
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
7973
+ "mmlu_eval_accuracy_anatomy": 0.5,
7974
+ "mmlu_eval_accuracy_astronomy": 0.4375,
7975
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
7976
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
7977
+ "mmlu_eval_accuracy_college_biology": 0.4375,
7978
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
7979
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
7980
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
7981
+ "mmlu_eval_accuracy_college_medicine": 0.5909090909090909,
7982
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
7983
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
7984
+ "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
7985
+ "mmlu_eval_accuracy_econometrics": 0.3333333333333333,
7986
+ "mmlu_eval_accuracy_electrical_engineering": 0.3125,
7987
+ "mmlu_eval_accuracy_elementary_mathematics": 0.2682926829268293,
7988
+ "mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
7989
+ "mmlu_eval_accuracy_global_facts": 0.5,
7990
+ "mmlu_eval_accuracy_high_school_biology": 0.4375,
7991
+ "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
7992
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
7993
+ "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
7994
+ "mmlu_eval_accuracy_high_school_geography": 0.7272727272727273,
7995
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
7996
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
7997
+ "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
7998
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.38461538461538464,
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": 0.34782608695652173,
8002
+ "mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
8003
+ "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
8004
+ "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
8005
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
8006
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
8007
+ "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
8008
+ "mmlu_eval_accuracy_logical_fallacies": 0.7222222222222222,
8009
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
8010
+ "mmlu_eval_accuracy_management": 0.5454545454545454,
8011
+ "mmlu_eval_accuracy_marketing": 0.84,
8012
+ "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
8013
+ "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
8014
+ "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
8015
+ "mmlu_eval_accuracy_moral_scenarios": 0.23,
8016
+ "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
8017
+ "mmlu_eval_accuracy_philosophy": 0.5,
8018
+ "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
8019
+ "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
8020
+ "mmlu_eval_accuracy_professional_law": 0.36470588235294116,
8021
+ "mmlu_eval_accuracy_professional_medicine": 0.6774193548387096,
8022
+ "mmlu_eval_accuracy_professional_psychology": 0.42028985507246375,
8023
+ "mmlu_eval_accuracy_public_relations": 0.4166666666666667,
8024
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
8025
+ "mmlu_eval_accuracy_sociology": 0.5454545454545454,
8026
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
8027
+ "mmlu_eval_accuracy_virology": 0.5,
8028
+ "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
8029
+ "mmlu_loss": 1.2135406544349367,
8030
+ "step": 8400
8031
+ },
8032
+ {
8033
+ "epoch": 4.05,
8034
+ "learning_rate": 0.0002,
8035
+ "loss": 0.3367,
8036
+ "step": 8410
8037
+ },
8038
+ {
8039
+ "epoch": 4.05,
8040
+ "learning_rate": 0.0002,
8041
+ "loss": 0.3606,
8042
+ "step": 8420
8043
+ },
8044
+ {
8045
+ "epoch": 4.06,
8046
+ "learning_rate": 0.0002,
8047
+ "loss": 0.3457,
8048
+ "step": 8430
8049
+ },
8050
+ {
8051
+ "epoch": 4.06,
8052
+ "learning_rate": 0.0002,
8053
+ "loss": 0.3287,
8054
+ "step": 8440
8055
+ },
8056
+ {
8057
+ "epoch": 4.07,
8058
+ "learning_rate": 0.0002,
8059
+ "loss": 0.3029,
8060
+ "step": 8450
8061
+ },
8062
+ {
8063
+ "epoch": 4.07,
8064
+ "learning_rate": 0.0002,
8065
+ "loss": 0.3176,
8066
+ "step": 8460
8067
+ },
8068
+ {
8069
+ "epoch": 4.08,
8070
+ "learning_rate": 0.0002,
8071
+ "loss": 0.3528,
8072
+ "step": 8470
8073
+ },
8074
+ {
8075
+ "epoch": 4.08,
8076
+ "learning_rate": 0.0002,
8077
+ "loss": 0.3381,
8078
+ "step": 8480
8079
+ },
8080
+ {
8081
+ "epoch": 4.09,
8082
+ "learning_rate": 0.0002,
8083
+ "loss": 0.3168,
8084
+ "step": 8490
8085
+ },
8086
+ {
8087
+ "epoch": 4.09,
8088
+ "learning_rate": 0.0002,
8089
+ "loss": 0.3466,
8090
+ "step": 8500
8091
+ },
8092
+ {
8093
+ "epoch": 4.1,
8094
+ "learning_rate": 0.0002,
8095
+ "loss": 0.3619,
8096
+ "step": 8510
8097
+ },
8098
+ {
8099
+ "epoch": 4.1,
8100
+ "learning_rate": 0.0002,
8101
+ "loss": 0.3334,
8102
+ "step": 8520
8103
+ },
8104
+ {
8105
+ "epoch": 4.11,
8106
+ "learning_rate": 0.0002,
8107
+ "loss": 0.3264,
8108
+ "step": 8530
8109
+ },
8110
+ {
8111
+ "epoch": 4.11,
8112
+ "learning_rate": 0.0002,
8113
+ "loss": 0.3306,
8114
+ "step": 8540
8115
+ },
8116
+ {
8117
+ "epoch": 4.12,
8118
+ "learning_rate": 0.0002,
8119
+ "loss": 0.3747,
8120
+ "step": 8550
8121
+ },
8122
+ {
8123
+ "epoch": 4.12,
8124
+ "learning_rate": 0.0002,
8125
+ "loss": 0.3643,
8126
+ "step": 8560
8127
+ },
8128
+ {
8129
+ "epoch": 4.12,
8130
+ "learning_rate": 0.0002,
8131
+ "loss": 0.3901,
8132
+ "step": 8570
8133
+ },
8134
+ {
8135
+ "epoch": 4.13,
8136
+ "learning_rate": 0.0002,
8137
+ "loss": 0.3495,
8138
+ "step": 8580
8139
+ },
8140
+ {
8141
+ "epoch": 4.13,
8142
+ "learning_rate": 0.0002,
8143
+ "loss": 0.3507,
8144
+ "step": 8590
8145
+ },
8146
+ {
8147
+ "epoch": 4.14,
8148
+ "learning_rate": 0.0002,
8149
+ "loss": 0.3618,
8150
+ "step": 8600
8151
+ },
8152
+ {
8153
+ "epoch": 4.14,
8154
+ "eval_loss": 0.8057907819747925,
8155
+ "eval_runtime": 96.4887,
8156
+ "eval_samples_per_second": 10.364,
8157
+ "eval_steps_per_second": 5.182,
8158
+ "step": 8600
8159
+ },
8160
+ {
8161
+ "epoch": 4.14,
8162
+ "mmlu_eval_accuracy": 0.4814356853985624,
8163
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
8164
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
8165
+ "mmlu_eval_accuracy_astronomy": 0.375,
8166
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
8167
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
8168
+ "mmlu_eval_accuracy_college_biology": 0.5,
8169
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
8170
+ "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
8171
+ "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
8172
+ "mmlu_eval_accuracy_college_medicine": 0.5909090909090909,
8173
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
8174
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
8175
+ "mmlu_eval_accuracy_conceptual_physics": 0.34615384615384615,
8176
+ "mmlu_eval_accuracy_econometrics": 0.3333333333333333,
8177
+ "mmlu_eval_accuracy_electrical_engineering": 0.3125,
8178
+ "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
8179
+ "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
8180
+ "mmlu_eval_accuracy_global_facts": 0.5,
8181
+ "mmlu_eval_accuracy_high_school_biology": 0.40625,
8182
+ "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
8183
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
8184
+ "mmlu_eval_accuracy_high_school_european_history": 0.5,
8185
+ "mmlu_eval_accuracy_high_school_geography": 0.7272727272727273,
8186
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.47619047619047616,
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": 0.4230769230769231,
8190
+ "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
8191
+ "mmlu_eval_accuracy_high_school_psychology": 0.8,
8192
+ "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
8193
+ "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
8194
+ "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
8195
+ "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
8196
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
8197
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
8198
+ "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
8199
+ "mmlu_eval_accuracy_logical_fallacies": 0.7222222222222222,
8200
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
8201
+ "mmlu_eval_accuracy_management": 0.5454545454545454,
8202
+ "mmlu_eval_accuracy_marketing": 0.88,
8203
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
8204
+ "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
8205
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
8206
+ "mmlu_eval_accuracy_moral_scenarios": 0.25,
8207
+ "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
8208
+ "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
8209
+ "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
8210
+ "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
8211
+ "mmlu_eval_accuracy_professional_law": 0.3352941176470588,
8212
+ "mmlu_eval_accuracy_professional_medicine": 0.6129032258064516,
8213
+ "mmlu_eval_accuracy_professional_psychology": 0.43478260869565216,
8214
+ "mmlu_eval_accuracy_public_relations": 0.5,
8215
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
8216
+ "mmlu_eval_accuracy_sociology": 0.5909090909090909,
8217
+ "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
8218
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
8219
+ "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
8220
+ "mmlu_loss": 1.0640197027153664,
8221
+ "step": 8600
8222
+ },
8223
+ {
8224
+ "epoch": 4.14,
8225
+ "learning_rate": 0.0002,
8226
+ "loss": 0.3428,
8227
+ "step": 8610
8228
+ },
8229
+ {
8230
+ "epoch": 4.15,
8231
+ "learning_rate": 0.0002,
8232
+ "loss": 0.3608,
8233
+ "step": 8620
8234
+ },
8235
+ {
8236
+ "epoch": 4.15,
8237
+ "learning_rate": 0.0002,
8238
+ "loss": 0.3496,
8239
+ "step": 8630
8240
+ },
8241
+ {
8242
+ "epoch": 4.16,
8243
+ "learning_rate": 0.0002,
8244
+ "loss": 0.3777,
8245
+ "step": 8640
8246
+ },
8247
+ {
8248
+ "epoch": 4.16,
8249
+ "learning_rate": 0.0002,
8250
+ "loss": 0.3261,
8251
+ "step": 8650
8252
+ },
8253
+ {
8254
+ "epoch": 4.17,
8255
+ "learning_rate": 0.0002,
8256
+ "loss": 0.3967,
8257
+ "step": 8660
8258
+ },
8259
+ {
8260
+ "epoch": 4.17,
8261
+ "learning_rate": 0.0002,
8262
+ "loss": 0.3231,
8263
+ "step": 8670
8264
+ },
8265
+ {
8266
+ "epoch": 4.18,
8267
+ "learning_rate": 0.0002,
8268
+ "loss": 0.3721,
8269
+ "step": 8680
8270
+ },
8271
+ {
8272
+ "epoch": 4.18,
8273
+ "learning_rate": 0.0002,
8274
+ "loss": 0.3766,
8275
+ "step": 8690
8276
+ },
8277
+ {
8278
+ "epoch": 4.19,
8279
+ "learning_rate": 0.0002,
8280
+ "loss": 0.3365,
8281
+ "step": 8700
8282
+ },
8283
+ {
8284
+ "epoch": 4.19,
8285
+ "learning_rate": 0.0002,
8286
+ "loss": 0.3603,
8287
+ "step": 8710
8288
+ },
8289
+ {
8290
+ "epoch": 4.2,
8291
+ "learning_rate": 0.0002,
8292
+ "loss": 0.3776,
8293
+ "step": 8720
8294
+ },
8295
+ {
8296
+ "epoch": 4.2,
8297
+ "learning_rate": 0.0002,
8298
+ "loss": 0.3407,
8299
+ "step": 8730
8300
+ },
8301
+ {
8302
+ "epoch": 4.21,
8303
+ "learning_rate": 0.0002,
8304
+ "loss": 0.3657,
8305
+ "step": 8740
8306
+ },
8307
+ {
8308
+ "epoch": 4.21,
8309
+ "learning_rate": 0.0002,
8310
+ "loss": 0.3541,
8311
+ "step": 8750
8312
+ },
8313
+ {
8314
+ "epoch": 4.22,
8315
+ "learning_rate": 0.0002,
8316
+ "loss": 0.4092,
8317
+ "step": 8760
8318
+ },
8319
+ {
8320
+ "epoch": 4.22,
8321
+ "learning_rate": 0.0002,
8322
+ "loss": 0.3455,
8323
+ "step": 8770
8324
+ },
8325
+ {
8326
+ "epoch": 4.23,
8327
+ "learning_rate": 0.0002,
8328
+ "loss": 0.3471,
8329
+ "step": 8780
8330
+ },
8331
+ {
8332
+ "epoch": 4.23,
8333
+ "learning_rate": 0.0002,
8334
+ "loss": 0.3615,
8335
+ "step": 8790
8336
+ },
8337
+ {
8338
+ "epoch": 4.24,
8339
+ "learning_rate": 0.0002,
8340
+ "loss": 0.295,
8341
+ "step": 8800
8342
+ },
8343
+ {
8344
+ "epoch": 4.24,
8345
+ "eval_loss": 0.8001754283905029,
8346
+ "eval_runtime": 96.445,
8347
+ "eval_samples_per_second": 10.369,
8348
+ "eval_steps_per_second": 5.184,
8349
+ "step": 8800
8350
+ },
8351
+ {
8352
+ "epoch": 4.24,
8353
+ "mmlu_eval_accuracy": 0.4763952795669252,
8354
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
8355
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
8356
+ "mmlu_eval_accuracy_astronomy": 0.375,
8357
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
8358
+ "mmlu_eval_accuracy_clinical_knowledge": 0.41379310344827586,
8359
+ "mmlu_eval_accuracy_college_biology": 0.5,
8360
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
8361
+ "mmlu_eval_accuracy_college_computer_science": 0.18181818181818182,
8362
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
8363
+ "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
8364
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
8365
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
8366
+ "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
8367
+ "mmlu_eval_accuracy_econometrics": 0.25,
8368
+ "mmlu_eval_accuracy_electrical_engineering": 0.3125,
8369
+ "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
8370
+ "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
8371
+ "mmlu_eval_accuracy_global_facts": 0.5,
8372
+ "mmlu_eval_accuracy_high_school_biology": 0.59375,
8373
+ "mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
8374
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
8375
+ "mmlu_eval_accuracy_high_school_european_history": 0.5,
8376
+ "mmlu_eval_accuracy_high_school_geography": 0.7272727272727273,
8377
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.47619047619047616,
8378
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.46511627906976744,
8379
+ "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
8380
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.38461538461538464,
8381
+ "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
8382
+ "mmlu_eval_accuracy_high_school_psychology": 0.8,
8383
+ "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
8384
+ "mmlu_eval_accuracy_high_school_us_history": 0.7727272727272727,
8385
+ "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
8386
+ "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
8387
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
8388
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
8389
+ "mmlu_eval_accuracy_jurisprudence": 0.18181818181818182,
8390
+ "mmlu_eval_accuracy_logical_fallacies": 0.7222222222222222,
8391
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
8392
+ "mmlu_eval_accuracy_management": 0.5454545454545454,
8393
+ "mmlu_eval_accuracy_marketing": 0.8,
8394
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
8395
+ "mmlu_eval_accuracy_miscellaneous": 0.627906976744186,
8396
+ "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
8397
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
8398
+ "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
8399
+ "mmlu_eval_accuracy_philosophy": 0.4117647058823529,
8400
+ "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
8401
+ "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
8402
+ "mmlu_eval_accuracy_professional_law": 0.3411764705882353,
8403
+ "mmlu_eval_accuracy_professional_medicine": 0.5806451612903226,
8404
+ "mmlu_eval_accuracy_professional_psychology": 0.42028985507246375,
8405
+ "mmlu_eval_accuracy_public_relations": 0.5,
8406
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
8407
+ "mmlu_eval_accuracy_sociology": 0.6363636363636364,
8408
+ "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
8409
+ "mmlu_eval_accuracy_virology": 0.4444444444444444,
8410
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
8411
+ "mmlu_loss": 1.0888570746596582,
8412
+ "step": 8800
8413
+ },
8414
+ {
8415
+ "epoch": 4.24,
8416
+ "learning_rate": 0.0002,
8417
+ "loss": 0.329,
8418
+ "step": 8810
8419
+ },
8420
+ {
8421
+ "epoch": 4.24,
8422
+ "learning_rate": 0.0002,
8423
+ "loss": 0.3518,
8424
+ "step": 8820
8425
+ },
8426
+ {
8427
+ "epoch": 4.25,
8428
+ "learning_rate": 0.0002,
8429
+ "loss": 0.3712,
8430
+ "step": 8830
8431
+ },
8432
+ {
8433
+ "epoch": 4.25,
8434
+ "learning_rate": 0.0002,
8435
+ "loss": 0.3756,
8436
+ "step": 8840
8437
+ },
8438
+ {
8439
+ "epoch": 4.26,
8440
+ "learning_rate": 0.0002,
8441
+ "loss": 0.3558,
8442
+ "step": 8850
8443
+ },
8444
+ {
8445
+ "epoch": 4.26,
8446
+ "learning_rate": 0.0002,
8447
+ "loss": 0.3594,
8448
+ "step": 8860
8449
+ },
8450
+ {
8451
+ "epoch": 4.27,
8452
+ "learning_rate": 0.0002,
8453
+ "loss": 0.3776,
8454
+ "step": 8870
8455
+ },
8456
+ {
8457
+ "epoch": 4.27,
8458
+ "learning_rate": 0.0002,
8459
+ "loss": 0.3714,
8460
+ "step": 8880
8461
+ },
8462
+ {
8463
+ "epoch": 4.28,
8464
+ "learning_rate": 0.0002,
8465
+ "loss": 0.4023,
8466
+ "step": 8890
8467
+ },
8468
+ {
8469
+ "epoch": 4.28,
8470
+ "learning_rate": 0.0002,
8471
+ "loss": 0.3722,
8472
+ "step": 8900
8473
+ },
8474
+ {
8475
+ "epoch": 4.29,
8476
+ "learning_rate": 0.0002,
8477
+ "loss": 0.3751,
8478
+ "step": 8910
8479
+ },
8480
+ {
8481
+ "epoch": 4.29,
8482
+ "learning_rate": 0.0002,
8483
+ "loss": 0.3676,
8484
+ "step": 8920
8485
+ },
8486
+ {
8487
+ "epoch": 4.3,
8488
+ "learning_rate": 0.0002,
8489
+ "loss": 0.3397,
8490
+ "step": 8930
8491
+ },
8492
+ {
8493
+ "epoch": 4.3,
8494
+ "learning_rate": 0.0002,
8495
+ "loss": 0.3854,
8496
+ "step": 8940
8497
+ },
8498
+ {
8499
+ "epoch": 4.31,
8500
+ "learning_rate": 0.0002,
8501
+ "loss": 0.381,
8502
+ "step": 8950
8503
+ },
8504
+ {
8505
+ "epoch": 4.31,
8506
+ "learning_rate": 0.0002,
8507
+ "loss": 0.3463,
8508
+ "step": 8960
8509
+ },
8510
+ {
8511
+ "epoch": 4.32,
8512
+ "learning_rate": 0.0002,
8513
+ "loss": 0.3571,
8514
+ "step": 8970
8515
+ },
8516
+ {
8517
+ "epoch": 4.32,
8518
+ "learning_rate": 0.0002,
8519
+ "loss": 0.3736,
8520
+ "step": 8980
8521
+ },
8522
+ {
8523
+ "epoch": 4.33,
8524
+ "learning_rate": 0.0002,
8525
+ "loss": 0.391,
8526
+ "step": 8990
8527
+ },
8528
+ {
8529
+ "epoch": 4.33,
8530
+ "learning_rate": 0.0002,
8531
+ "loss": 0.3377,
8532
+ "step": 9000
8533
+ },
8534
+ {
8535
+ "epoch": 4.33,
8536
+ "eval_loss": 0.8083432912826538,
8537
+ "eval_runtime": 96.5387,
8538
+ "eval_samples_per_second": 10.359,
8539
+ "eval_steps_per_second": 5.179,
8540
+ "step": 9000
8541
+ },
8542
+ {
8543
+ "epoch": 4.33,
8544
+ "mmlu_eval_accuracy": 0.49746173965555757,
8545
+ "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
8546
+ "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
8547
+ "mmlu_eval_accuracy_astronomy": 0.375,
8548
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
8549
+ "mmlu_eval_accuracy_clinical_knowledge": 0.4482758620689655,
8550
+ "mmlu_eval_accuracy_college_biology": 0.5,
8551
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
8552
+ "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
8553
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
8554
+ "mmlu_eval_accuracy_college_medicine": 0.5454545454545454,
8555
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
8556
+ "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
8557
+ "mmlu_eval_accuracy_conceptual_physics": 0.34615384615384615,
8558
+ "mmlu_eval_accuracy_econometrics": 0.25,
8559
+ "mmlu_eval_accuracy_electrical_engineering": 0.3125,
8560
+ "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
8561
+ "mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
8562
+ "mmlu_eval_accuracy_global_facts": 0.5,
8563
+ "mmlu_eval_accuracy_high_school_biology": 0.53125,
8564
+ "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
8565
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
8566
+ "mmlu_eval_accuracy_high_school_european_history": 0.5,
8567
+ "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
8568
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
8569
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4883720930232558,
8570
+ "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
8571
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
8572
+ "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
8573
+ "mmlu_eval_accuracy_high_school_psychology": 0.8,
8574
+ "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
8575
+ "mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
8576
+ "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
8577
+ "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
8578
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
8579
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
8580
+ "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
8581
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
8582
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
8583
+ "mmlu_eval_accuracy_management": 0.5454545454545454,
8584
+ "mmlu_eval_accuracy_marketing": 0.84,
8585
+ "mmlu_eval_accuracy_medical_genetics": 1.0,
8586
+ "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
8587
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
8588
+ "mmlu_eval_accuracy_moral_scenarios": 0.27,
8589
+ "mmlu_eval_accuracy_nutrition": 0.7272727272727273,
8590
+ "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
8591
+ "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
8592
+ "mmlu_eval_accuracy_professional_accounting": 0.4838709677419355,
8593
+ "mmlu_eval_accuracy_professional_law": 0.3352941176470588,
8594
+ "mmlu_eval_accuracy_professional_medicine": 0.6129032258064516,
8595
+ "mmlu_eval_accuracy_professional_psychology": 0.463768115942029,
8596
+ "mmlu_eval_accuracy_public_relations": 0.4166666666666667,
8597
+ "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
8598
+ "mmlu_eval_accuracy_sociology": 0.7272727272727273,
8599
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
8600
+ "mmlu_eval_accuracy_virology": 0.4444444444444444,
8601
+ "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
8602
+ "mmlu_loss": 1.1287356225905,
8603
+ "step": 9000
8604
+ },
8605
+ {
8606
+ "epoch": 4.34,
8607
+ "learning_rate": 0.0002,
8608
+ "loss": 0.3486,
8609
+ "step": 9010
8610
+ },
8611
+ {
8612
+ "epoch": 4.34,
8613
+ "learning_rate": 0.0002,
8614
+ "loss": 0.4132,
8615
+ "step": 9020
8616
+ },
8617
+ {
8618
+ "epoch": 4.35,
8619
+ "learning_rate": 0.0002,
8620
+ "loss": 0.3642,
8621
+ "step": 9030
8622
+ },
8623
+ {
8624
+ "epoch": 4.35,
8625
+ "learning_rate": 0.0002,
8626
+ "loss": 0.3397,
8627
+ "step": 9040
8628
+ },
8629
+ {
8630
+ "epoch": 4.36,
8631
+ "learning_rate": 0.0002,
8632
+ "loss": 0.3457,
8633
+ "step": 9050
8634
+ },
8635
+ {
8636
+ "epoch": 4.36,
8637
+ "learning_rate": 0.0002,
8638
+ "loss": 0.3627,
8639
+ "step": 9060
8640
+ },
8641
+ {
8642
+ "epoch": 4.37,
8643
+ "learning_rate": 0.0002,
8644
+ "loss": 0.3524,
8645
+ "step": 9070
8646
+ },
8647
+ {
8648
+ "epoch": 4.37,
8649
+ "learning_rate": 0.0002,
8650
+ "loss": 0.3969,
8651
+ "step": 9080
8652
+ },
8653
+ {
8654
+ "epoch": 4.37,
8655
+ "learning_rate": 0.0002,
8656
+ "loss": 0.3271,
8657
+ "step": 9090
8658
+ },
8659
+ {
8660
+ "epoch": 4.38,
8661
+ "learning_rate": 0.0002,
8662
+ "loss": 0.4009,
8663
+ "step": 9100
8664
+ },
8665
+ {
8666
+ "epoch": 4.38,
8667
+ "learning_rate": 0.0002,
8668
+ "loss": 0.3668,
8669
+ "step": 9110
8670
+ },
8671
+ {
8672
+ "epoch": 4.39,
8673
+ "learning_rate": 0.0002,
8674
+ "loss": 0.3493,
8675
+ "step": 9120
8676
+ },
8677
+ {
8678
+ "epoch": 4.39,
8679
+ "learning_rate": 0.0002,
8680
+ "loss": 0.3959,
8681
+ "step": 9130
8682
+ },
8683
+ {
8684
+ "epoch": 4.4,
8685
+ "learning_rate": 0.0002,
8686
+ "loss": 0.3286,
8687
+ "step": 9140
8688
+ },
8689
+ {
8690
+ "epoch": 4.4,
8691
+ "learning_rate": 0.0002,
8692
+ "loss": 0.3556,
8693
+ "step": 9150
8694
+ },
8695
+ {
8696
+ "epoch": 4.41,
8697
+ "learning_rate": 0.0002,
8698
+ "loss": 0.3549,
8699
+ "step": 9160
8700
+ },
8701
+ {
8702
+ "epoch": 4.41,
8703
+ "learning_rate": 0.0002,
8704
+ "loss": 0.388,
8705
+ "step": 9170
8706
+ },
8707
+ {
8708
+ "epoch": 4.42,
8709
+ "learning_rate": 0.0002,
8710
+ "loss": 0.3434,
8711
+ "step": 9180
8712
+ },
8713
+ {
8714
+ "epoch": 4.42,
8715
+ "learning_rate": 0.0002,
8716
+ "loss": 0.3751,
8717
+ "step": 9190
8718
+ },
8719
+ {
8720
+ "epoch": 4.43,
8721
+ "learning_rate": 0.0002,
8722
+ "loss": 0.3987,
8723
+ "step": 9200
8724
+ },
8725
+ {
8726
+ "epoch": 4.43,
8727
+ "eval_loss": 0.7953082323074341,
8728
+ "eval_runtime": 96.5341,
8729
+ "eval_samples_per_second": 10.359,
8730
+ "eval_steps_per_second": 5.18,
8731
+ "step": 9200
8732
+ },
8733
+ {
8734
+ "epoch": 4.43,
8735
+ "mmlu_eval_accuracy": 0.4821778515325232,
8736
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
8737
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
8738
+ "mmlu_eval_accuracy_astronomy": 0.375,
8739
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
8740
+ "mmlu_eval_accuracy_clinical_knowledge": 0.4482758620689655,
8741
+ "mmlu_eval_accuracy_college_biology": 0.5,
8742
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
8743
+ "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
8744
+ "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
8745
+ "mmlu_eval_accuracy_college_medicine": 0.5454545454545454,
8746
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
8747
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
8748
+ "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
8749
+ "mmlu_eval_accuracy_econometrics": 0.25,
8750
+ "mmlu_eval_accuracy_electrical_engineering": 0.3125,
8751
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
8752
+ "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
8753
+ "mmlu_eval_accuracy_global_facts": 0.5,
8754
+ "mmlu_eval_accuracy_high_school_biology": 0.53125,
8755
+ "mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
8756
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
8757
+ "mmlu_eval_accuracy_high_school_european_history": 0.5,
8758
+ "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
8759
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.42857142857142855,
8760
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.46511627906976744,
8761
+ "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
8762
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.34615384615384615,
8763
+ "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
8764
+ "mmlu_eval_accuracy_high_school_psychology": 0.8,
8765
+ "mmlu_eval_accuracy_high_school_statistics": 0.2608695652173913,
8766
+ "mmlu_eval_accuracy_high_school_us_history": 0.7727272727272727,
8767
+ "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
8768
+ "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
8769
+ "mmlu_eval_accuracy_human_sexuality": 0.25,
8770
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
8771
+ "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
8772
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
8773
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
8774
+ "mmlu_eval_accuracy_management": 0.5454545454545454,
8775
+ "mmlu_eval_accuracy_marketing": 0.76,
8776
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
8777
+ "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
8778
+ "mmlu_eval_accuracy_moral_disputes": 0.5789473684210527,
8779
+ "mmlu_eval_accuracy_moral_scenarios": 0.25,
8780
+ "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
8781
+ "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
8782
+ "mmlu_eval_accuracy_prehistory": 0.5142857142857142,
8783
+ "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
8784
+ "mmlu_eval_accuracy_professional_law": 0.3235294117647059,
8785
+ "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
8786
+ "mmlu_eval_accuracy_professional_psychology": 0.4492753623188406,
8787
+ "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
8788
+ "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
8789
+ "mmlu_eval_accuracy_sociology": 0.7272727272727273,
8790
+ "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
8791
+ "mmlu_eval_accuracy_virology": 0.5,
8792
+ "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
8793
+ "mmlu_loss": 1.2391800577156538,
8794
+ "step": 9200
8795
  }
8796
  ],
8797
  "max_steps": 10000,
8798
  "num_train_epochs": 5,
8799
+ "total_flos": 2.4301525803626004e+18,
8800
  "trial_name": null,
8801
  "trial_params": null
8802
  }
{checkpoint-7200 → checkpoint-9200}/training_args.bin RENAMED
File without changes