Farouk commited on
Commit
52105fc
·
1 Parent(s): 6d53a28

Training in progress, step 8600

Browse files
adapter_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2850ca7abc398e417d125e6cdf46736a49bcc9bdae58fbb4b8474d6aba24cdcd
3
  size 319977229
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c9cf15b202832dcb1e3e4fa660effcba828a2ea3d11e26170a782cab3cf95bf
3
  size 319977229
checkpoint-4200/adapter_model/adapter_model/README.md CHANGED
@@ -224,6 +224,17 @@ The following `bitsandbytes` quantization config was used during training:
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
@@ -256,5 +267,6 @@ The following `bitsandbytes` quantization config was used during training:
256
  - PEFT 0.4.0
257
  - PEFT 0.4.0
258
  - PEFT 0.4.0
 
259
 
260
  - PEFT 0.4.0
 
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
230
+ - llm_int8_threshold: 6.0
231
+ - llm_int8_skip_modules: None
232
+ - llm_int8_enable_fp32_cpu_offload: False
233
+ - llm_int8_has_fp16_weight: False
234
+ - bnb_4bit_quant_type: nf4
235
+ - bnb_4bit_use_double_quant: True
236
+ - bnb_4bit_compute_dtype: bfloat16
237
+
238
  The following `bitsandbytes` quantization config was used during training:
239
  - load_in_8bit: False
240
  - load_in_4bit: True
 
267
  - PEFT 0.4.0
268
  - PEFT 0.4.0
269
  - PEFT 0.4.0
270
+ - PEFT 0.4.0
271
 
272
  - PEFT 0.4.0
checkpoint-4200/adapter_model/adapter_model/adapter_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:8a6df0ef79b7ff8706d61e9c9a32866e9e30c73c3ccff7215e8c50a981940d8d
3
  size 319977229
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2850ca7abc398e417d125e6cdf46736a49bcc9bdae58fbb4b8474d6aba24cdcd
3
  size 319977229
{checkpoint-6600 → checkpoint-8600}/README.md RENAMED
File without changes
{checkpoint-6600 → checkpoint-8600}/adapter_config.json RENAMED
File without changes
{checkpoint-6600 → checkpoint-8600}/adapter_model.bin RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2ff1c7776c69f3db26b9bf7859c8dc83d8d0f41ba573c31f5f38a8c0189dda16
3
  size 319977229
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c9cf15b202832dcb1e3e4fa660effcba828a2ea3d11e26170a782cab3cf95bf
3
  size 319977229
{checkpoint-6600 → checkpoint-8600}/added_tokens.json RENAMED
File without changes
{checkpoint-6600 → checkpoint-8600}/optimizer.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:fb5f932f2b86e3aeacce9f5b820c3abae46b4823d3be4ba69f0a45f973760aa6
3
  size 1279539973
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:20765a12051ef03eaceadc86825c9ac05481fc1bc8a42fde471dec525c1984c5
3
  size 1279539973
{checkpoint-6600 → checkpoint-8600}/rng_state.pth RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:ad0d74a1a0112b7f8a05adb4d95abdbdfeda4bc400546867e7d3babf9a670e57
3
  size 14511
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:00595a8c01568da2382e369e7e77062ca6c0a0b7cf7400adbe0a636e564c0c3f
3
  size 14511
{checkpoint-6600 → checkpoint-8600}/scheduler.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:aa19c433c8c029403e57118df2ab52631b3fc535294c01cab201bdeb198ed0f4
3
  size 627
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:58bb158dc8e2249ddde152cf1ccc6f5a31e448a10bd61b537efe520ccb7eb273
3
  size 627
{checkpoint-6600 → checkpoint-8600}/special_tokens_map.json RENAMED
File without changes
{checkpoint-6600 → checkpoint-8600}/tokenizer.model RENAMED
File without changes
{checkpoint-6600 → checkpoint-8600}/tokenizer_config.json RENAMED
File without changes
{checkpoint-6600 → checkpoint-8600}/trainer_state.json RENAMED
@@ -1,8 +1,8 @@
1
  {
2
  "best_metric": 1.0316325426101685,
3
  "best_model_checkpoint": "experts/expert-7/checkpoint-4200",
4
- "epoch": 2.9877772747849707,
5
- "global_step": 6600,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
@@ -6309,11 +6309,1921 @@
6309
  "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
6310
  "mmlu_loss": 1.2967368322704231,
6311
  "step": 6600
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6312
  }
6313
  ],
6314
  "max_steps": 10000,
6315
  "num_train_epochs": 5,
6316
- "total_flos": 2.3608750548006666e+18,
6317
  "trial_name": null,
6318
  "trial_params": null
6319
  }
 
1
  {
2
  "best_metric": 1.0316325426101685,
3
  "best_model_checkpoint": "experts/expert-7/checkpoint-4200",
4
+ "epoch": 3.893164327750113,
5
+ "global_step": 8600,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
 
6309
  "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
6310
  "mmlu_loss": 1.2967368322704231,
6311
  "step": 6600
6312
+ },
6313
+ {
6314
+ "epoch": 2.99,
6315
+ "learning_rate": 0.0002,
6316
+ "loss": 0.8544,
6317
+ "step": 6610
6318
+ },
6319
+ {
6320
+ "epoch": 3.0,
6321
+ "learning_rate": 0.0002,
6322
+ "loss": 0.8509,
6323
+ "step": 6620
6324
+ },
6325
+ {
6326
+ "epoch": 3.0,
6327
+ "learning_rate": 0.0002,
6328
+ "loss": 0.8302,
6329
+ "step": 6630
6330
+ },
6331
+ {
6332
+ "epoch": 3.01,
6333
+ "learning_rate": 0.0002,
6334
+ "loss": 0.7187,
6335
+ "step": 6640
6336
+ },
6337
+ {
6338
+ "epoch": 3.01,
6339
+ "learning_rate": 0.0002,
6340
+ "loss": 0.6596,
6341
+ "step": 6650
6342
+ },
6343
+ {
6344
+ "epoch": 3.01,
6345
+ "learning_rate": 0.0002,
6346
+ "loss": 0.7053,
6347
+ "step": 6660
6348
+ },
6349
+ {
6350
+ "epoch": 3.02,
6351
+ "learning_rate": 0.0002,
6352
+ "loss": 0.7569,
6353
+ "step": 6670
6354
+ },
6355
+ {
6356
+ "epoch": 3.02,
6357
+ "learning_rate": 0.0002,
6358
+ "loss": 0.6137,
6359
+ "step": 6680
6360
+ },
6361
+ {
6362
+ "epoch": 3.03,
6363
+ "learning_rate": 0.0002,
6364
+ "loss": 0.7059,
6365
+ "step": 6690
6366
+ },
6367
+ {
6368
+ "epoch": 3.03,
6369
+ "learning_rate": 0.0002,
6370
+ "loss": 0.6916,
6371
+ "step": 6700
6372
+ },
6373
+ {
6374
+ "epoch": 3.04,
6375
+ "learning_rate": 0.0002,
6376
+ "loss": 0.7222,
6377
+ "step": 6710
6378
+ },
6379
+ {
6380
+ "epoch": 3.04,
6381
+ "learning_rate": 0.0002,
6382
+ "loss": 0.7082,
6383
+ "step": 6720
6384
+ },
6385
+ {
6386
+ "epoch": 3.05,
6387
+ "learning_rate": 0.0002,
6388
+ "loss": 0.6925,
6389
+ "step": 6730
6390
+ },
6391
+ {
6392
+ "epoch": 3.05,
6393
+ "learning_rate": 0.0002,
6394
+ "loss": 0.7745,
6395
+ "step": 6740
6396
+ },
6397
+ {
6398
+ "epoch": 3.06,
6399
+ "learning_rate": 0.0002,
6400
+ "loss": 0.723,
6401
+ "step": 6750
6402
+ },
6403
+ {
6404
+ "epoch": 3.06,
6405
+ "learning_rate": 0.0002,
6406
+ "loss": 0.7076,
6407
+ "step": 6760
6408
+ },
6409
+ {
6410
+ "epoch": 3.06,
6411
+ "learning_rate": 0.0002,
6412
+ "loss": 0.7711,
6413
+ "step": 6770
6414
+ },
6415
+ {
6416
+ "epoch": 3.07,
6417
+ "learning_rate": 0.0002,
6418
+ "loss": 0.6491,
6419
+ "step": 6780
6420
+ },
6421
+ {
6422
+ "epoch": 3.07,
6423
+ "learning_rate": 0.0002,
6424
+ "loss": 0.6895,
6425
+ "step": 6790
6426
+ },
6427
+ {
6428
+ "epoch": 3.08,
6429
+ "learning_rate": 0.0002,
6430
+ "loss": 0.6633,
6431
+ "step": 6800
6432
+ },
6433
+ {
6434
+ "epoch": 3.08,
6435
+ "eval_loss": 1.100313425064087,
6436
+ "eval_runtime": 130.2464,
6437
+ "eval_samples_per_second": 7.678,
6438
+ "eval_steps_per_second": 3.839,
6439
+ "step": 6800
6440
+ },
6441
+ {
6442
+ "epoch": 3.08,
6443
+ "mmlu_eval_accuracy": 0.505369863742698,
6444
+ "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
6445
+ "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
6446
+ "mmlu_eval_accuracy_astronomy": 0.5,
6447
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
6448
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
6449
+ "mmlu_eval_accuracy_college_biology": 0.375,
6450
+ "mmlu_eval_accuracy_college_chemistry": 0.375,
6451
+ "mmlu_eval_accuracy_college_computer_science": 0.5454545454545454,
6452
+ "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
6453
+ "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
6454
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
6455
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
6456
+ "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
6457
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
6458
+ "mmlu_eval_accuracy_electrical_engineering": 0.5,
6459
+ "mmlu_eval_accuracy_elementary_mathematics": 0.4146341463414634,
6460
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
6461
+ "mmlu_eval_accuracy_global_facts": 0.3,
6462
+ "mmlu_eval_accuracy_high_school_biology": 0.5,
6463
+ "mmlu_eval_accuracy_high_school_chemistry": 0.45454545454545453,
6464
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
6465
+ "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
6466
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
6467
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.7142857142857143,
6468
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
6469
+ "mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
6470
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5769230769230769,
6471
+ "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
6472
+ "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
6473
+ "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
6474
+ "mmlu_eval_accuracy_high_school_us_history": 0.7727272727272727,
6475
+ "mmlu_eval_accuracy_high_school_world_history": 0.6153846153846154,
6476
+ "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
6477
+ "mmlu_eval_accuracy_human_sexuality": 0.5,
6478
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
6479
+ "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
6480
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
6481
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
6482
+ "mmlu_eval_accuracy_management": 0.7272727272727273,
6483
+ "mmlu_eval_accuracy_marketing": 0.8,
6484
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
6485
+ "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
6486
+ "mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
6487
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
6488
+ "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
6489
+ "mmlu_eval_accuracy_philosophy": 0.5,
6490
+ "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
6491
+ "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
6492
+ "mmlu_eval_accuracy_professional_law": 0.31176470588235294,
6493
+ "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
6494
+ "mmlu_eval_accuracy_professional_psychology": 0.4782608695652174,
6495
+ "mmlu_eval_accuracy_public_relations": 0.4166666666666667,
6496
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
6497
+ "mmlu_eval_accuracy_sociology": 0.7272727272727273,
6498
+ "mmlu_eval_accuracy_us_foreign_policy": 0.8181818181818182,
6499
+ "mmlu_eval_accuracy_virology": 0.4444444444444444,
6500
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
6501
+ "mmlu_loss": 1.3002990203722338,
6502
+ "step": 6800
6503
+ },
6504
+ {
6505
+ "epoch": 3.08,
6506
+ "learning_rate": 0.0002,
6507
+ "loss": 0.7095,
6508
+ "step": 6810
6509
+ },
6510
+ {
6511
+ "epoch": 3.09,
6512
+ "learning_rate": 0.0002,
6513
+ "loss": 0.693,
6514
+ "step": 6820
6515
+ },
6516
+ {
6517
+ "epoch": 3.09,
6518
+ "learning_rate": 0.0002,
6519
+ "loss": 0.6981,
6520
+ "step": 6830
6521
+ },
6522
+ {
6523
+ "epoch": 3.1,
6524
+ "learning_rate": 0.0002,
6525
+ "loss": 0.7349,
6526
+ "step": 6840
6527
+ },
6528
+ {
6529
+ "epoch": 3.1,
6530
+ "learning_rate": 0.0002,
6531
+ "loss": 0.7171,
6532
+ "step": 6850
6533
+ },
6534
+ {
6535
+ "epoch": 3.11,
6536
+ "learning_rate": 0.0002,
6537
+ "loss": 0.7711,
6538
+ "step": 6860
6539
+ },
6540
+ {
6541
+ "epoch": 3.11,
6542
+ "learning_rate": 0.0002,
6543
+ "loss": 0.6911,
6544
+ "step": 6870
6545
+ },
6546
+ {
6547
+ "epoch": 3.11,
6548
+ "learning_rate": 0.0002,
6549
+ "loss": 0.6943,
6550
+ "step": 6880
6551
+ },
6552
+ {
6553
+ "epoch": 3.12,
6554
+ "learning_rate": 0.0002,
6555
+ "loss": 0.7191,
6556
+ "step": 6890
6557
+ },
6558
+ {
6559
+ "epoch": 3.12,
6560
+ "learning_rate": 0.0002,
6561
+ "loss": 0.7128,
6562
+ "step": 6900
6563
+ },
6564
+ {
6565
+ "epoch": 3.13,
6566
+ "learning_rate": 0.0002,
6567
+ "loss": 0.7612,
6568
+ "step": 6910
6569
+ },
6570
+ {
6571
+ "epoch": 3.13,
6572
+ "learning_rate": 0.0002,
6573
+ "loss": 0.6915,
6574
+ "step": 6920
6575
+ },
6576
+ {
6577
+ "epoch": 3.14,
6578
+ "learning_rate": 0.0002,
6579
+ "loss": 0.6533,
6580
+ "step": 6930
6581
+ },
6582
+ {
6583
+ "epoch": 3.14,
6584
+ "learning_rate": 0.0002,
6585
+ "loss": 0.6817,
6586
+ "step": 6940
6587
+ },
6588
+ {
6589
+ "epoch": 3.15,
6590
+ "learning_rate": 0.0002,
6591
+ "loss": 0.7635,
6592
+ "step": 6950
6593
+ },
6594
+ {
6595
+ "epoch": 3.15,
6596
+ "learning_rate": 0.0002,
6597
+ "loss": 0.7887,
6598
+ "step": 6960
6599
+ },
6600
+ {
6601
+ "epoch": 3.16,
6602
+ "learning_rate": 0.0002,
6603
+ "loss": 0.6873,
6604
+ "step": 6970
6605
+ },
6606
+ {
6607
+ "epoch": 3.16,
6608
+ "learning_rate": 0.0002,
6609
+ "loss": 0.742,
6610
+ "step": 6980
6611
+ },
6612
+ {
6613
+ "epoch": 3.16,
6614
+ "learning_rate": 0.0002,
6615
+ "loss": 0.7462,
6616
+ "step": 6990
6617
+ },
6618
+ {
6619
+ "epoch": 3.17,
6620
+ "learning_rate": 0.0002,
6621
+ "loss": 0.7373,
6622
+ "step": 7000
6623
+ },
6624
+ {
6625
+ "epoch": 3.17,
6626
+ "eval_loss": 1.1085585355758667,
6627
+ "eval_runtime": 130.3553,
6628
+ "eval_samples_per_second": 7.671,
6629
+ "eval_steps_per_second": 3.836,
6630
+ "step": 7000
6631
+ },
6632
+ {
6633
+ "epoch": 3.17,
6634
+ "mmlu_eval_accuracy": 0.5070960205427749,
6635
+ "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
6636
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
6637
+ "mmlu_eval_accuracy_astronomy": 0.5,
6638
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
6639
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5862068965517241,
6640
+ "mmlu_eval_accuracy_college_biology": 0.4375,
6641
+ "mmlu_eval_accuracy_college_chemistry": 0.5,
6642
+ "mmlu_eval_accuracy_college_computer_science": 0.5454545454545454,
6643
+ "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
6644
+ "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
6645
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
6646
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
6647
+ "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
6648
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
6649
+ "mmlu_eval_accuracy_electrical_engineering": 0.375,
6650
+ "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
6651
+ "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
6652
+ "mmlu_eval_accuracy_global_facts": 0.4,
6653
+ "mmlu_eval_accuracy_high_school_biology": 0.5,
6654
+ "mmlu_eval_accuracy_high_school_chemistry": 0.45454545454545453,
6655
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
6656
+ "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
6657
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
6658
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.7142857142857143,
6659
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
6660
+ "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
6661
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
6662
+ "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
6663
+ "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
6664
+ "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
6665
+ "mmlu_eval_accuracy_high_school_us_history": 0.8181818181818182,
6666
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
6667
+ "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
6668
+ "mmlu_eval_accuracy_human_sexuality": 0.5,
6669
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
6670
+ "mmlu_eval_accuracy_jurisprudence": 0.6363636363636364,
6671
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
6672
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
6673
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
6674
+ "mmlu_eval_accuracy_marketing": 0.76,
6675
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
6676
+ "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
6677
+ "mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
6678
+ "mmlu_eval_accuracy_moral_scenarios": 0.22,
6679
+ "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
6680
+ "mmlu_eval_accuracy_philosophy": 0.5588235294117647,
6681
+ "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
6682
+ "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
6683
+ "mmlu_eval_accuracy_professional_law": 0.3176470588235294,
6684
+ "mmlu_eval_accuracy_professional_medicine": 0.4838709677419355,
6685
+ "mmlu_eval_accuracy_professional_psychology": 0.43478260869565216,
6686
+ "mmlu_eval_accuracy_public_relations": 0.4166666666666667,
6687
+ "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
6688
+ "mmlu_eval_accuracy_sociology": 0.7727272727272727,
6689
+ "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
6690
+ "mmlu_eval_accuracy_virology": 0.4444444444444444,
6691
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
6692
+ "mmlu_loss": 1.3408084753754868,
6693
+ "step": 7000
6694
+ },
6695
+ {
6696
+ "epoch": 3.17,
6697
+ "learning_rate": 0.0002,
6698
+ "loss": 0.7427,
6699
+ "step": 7010
6700
+ },
6701
+ {
6702
+ "epoch": 3.18,
6703
+ "learning_rate": 0.0002,
6704
+ "loss": 0.7485,
6705
+ "step": 7020
6706
+ },
6707
+ {
6708
+ "epoch": 3.18,
6709
+ "learning_rate": 0.0002,
6710
+ "loss": 0.705,
6711
+ "step": 7030
6712
+ },
6713
+ {
6714
+ "epoch": 3.19,
6715
+ "learning_rate": 0.0002,
6716
+ "loss": 0.7201,
6717
+ "step": 7040
6718
+ },
6719
+ {
6720
+ "epoch": 3.19,
6721
+ "learning_rate": 0.0002,
6722
+ "loss": 0.8002,
6723
+ "step": 7050
6724
+ },
6725
+ {
6726
+ "epoch": 3.2,
6727
+ "learning_rate": 0.0002,
6728
+ "loss": 0.6925,
6729
+ "step": 7060
6730
+ },
6731
+ {
6732
+ "epoch": 3.2,
6733
+ "learning_rate": 0.0002,
6734
+ "loss": 0.6757,
6735
+ "step": 7070
6736
+ },
6737
+ {
6738
+ "epoch": 3.21,
6739
+ "learning_rate": 0.0002,
6740
+ "loss": 0.6883,
6741
+ "step": 7080
6742
+ },
6743
+ {
6744
+ "epoch": 3.21,
6745
+ "learning_rate": 0.0002,
6746
+ "loss": 0.7801,
6747
+ "step": 7090
6748
+ },
6749
+ {
6750
+ "epoch": 3.21,
6751
+ "learning_rate": 0.0002,
6752
+ "loss": 0.6952,
6753
+ "step": 7100
6754
+ },
6755
+ {
6756
+ "epoch": 3.22,
6757
+ "learning_rate": 0.0002,
6758
+ "loss": 0.6992,
6759
+ "step": 7110
6760
+ },
6761
+ {
6762
+ "epoch": 3.22,
6763
+ "learning_rate": 0.0002,
6764
+ "loss": 0.7473,
6765
+ "step": 7120
6766
+ },
6767
+ {
6768
+ "epoch": 3.23,
6769
+ "learning_rate": 0.0002,
6770
+ "loss": 0.7346,
6771
+ "step": 7130
6772
+ },
6773
+ {
6774
+ "epoch": 3.23,
6775
+ "learning_rate": 0.0002,
6776
+ "loss": 0.7418,
6777
+ "step": 7140
6778
+ },
6779
+ {
6780
+ "epoch": 3.24,
6781
+ "learning_rate": 0.0002,
6782
+ "loss": 0.6946,
6783
+ "step": 7150
6784
+ },
6785
+ {
6786
+ "epoch": 3.24,
6787
+ "learning_rate": 0.0002,
6788
+ "loss": 0.7411,
6789
+ "step": 7160
6790
+ },
6791
+ {
6792
+ "epoch": 3.25,
6793
+ "learning_rate": 0.0002,
6794
+ "loss": 0.7398,
6795
+ "step": 7170
6796
+ },
6797
+ {
6798
+ "epoch": 3.25,
6799
+ "learning_rate": 0.0002,
6800
+ "loss": 0.7302,
6801
+ "step": 7180
6802
+ },
6803
+ {
6804
+ "epoch": 3.25,
6805
+ "learning_rate": 0.0002,
6806
+ "loss": 0.7543,
6807
+ "step": 7190
6808
+ },
6809
+ {
6810
+ "epoch": 3.26,
6811
+ "learning_rate": 0.0002,
6812
+ "loss": 0.7539,
6813
+ "step": 7200
6814
+ },
6815
+ {
6816
+ "epoch": 3.26,
6817
+ "eval_loss": 1.0998623371124268,
6818
+ "eval_runtime": 130.5608,
6819
+ "eval_samples_per_second": 7.659,
6820
+ "eval_steps_per_second": 3.83,
6821
+ "step": 7200
6822
+ },
6823
+ {
6824
+ "epoch": 3.26,
6825
+ "mmlu_eval_accuracy": 0.5053221169369858,
6826
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
6827
+ "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
6828
+ "mmlu_eval_accuracy_astronomy": 0.4375,
6829
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
6830
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5862068965517241,
6831
+ "mmlu_eval_accuracy_college_biology": 0.4375,
6832
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
6833
+ "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
6834
+ "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
6835
+ "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
6836
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
6837
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
6838
+ "mmlu_eval_accuracy_conceptual_physics": 0.5,
6839
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
6840
+ "mmlu_eval_accuracy_electrical_engineering": 0.5,
6841
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
6842
+ "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
6843
+ "mmlu_eval_accuracy_global_facts": 0.3,
6844
+ "mmlu_eval_accuracy_high_school_biology": 0.46875,
6845
+ "mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
6846
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
6847
+ "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
6848
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
6849
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.7142857142857143,
6850
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
6851
+ "mmlu_eval_accuracy_high_school_mathematics": 0.3103448275862069,
6852
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
6853
+ "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
6854
+ "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
6855
+ "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
6856
+ "mmlu_eval_accuracy_high_school_us_history": 0.8181818181818182,
6857
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
6858
+ "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
6859
+ "mmlu_eval_accuracy_human_sexuality": 0.5,
6860
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
6861
+ "mmlu_eval_accuracy_jurisprudence": 0.5454545454545454,
6862
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
6863
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
6864
+ "mmlu_eval_accuracy_management": 0.7272727272727273,
6865
+ "mmlu_eval_accuracy_marketing": 0.8,
6866
+ "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
6867
+ "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
6868
+ "mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
6869
+ "mmlu_eval_accuracy_moral_scenarios": 0.23,
6870
+ "mmlu_eval_accuracy_nutrition": 0.696969696969697,
6871
+ "mmlu_eval_accuracy_philosophy": 0.5,
6872
+ "mmlu_eval_accuracy_prehistory": 0.5142857142857142,
6873
+ "mmlu_eval_accuracy_professional_accounting": 0.25806451612903225,
6874
+ "mmlu_eval_accuracy_professional_law": 0.34705882352941175,
6875
+ "mmlu_eval_accuracy_professional_medicine": 0.45161290322580644,
6876
+ "mmlu_eval_accuracy_professional_psychology": 0.4782608695652174,
6877
+ "mmlu_eval_accuracy_public_relations": 0.4166666666666667,
6878
+ "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
6879
+ "mmlu_eval_accuracy_sociology": 0.7727272727272727,
6880
+ "mmlu_eval_accuracy_us_foreign_policy": 0.8181818181818182,
6881
+ "mmlu_eval_accuracy_virology": 0.4444444444444444,
6882
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
6883
+ "mmlu_loss": 1.3234099702758826,
6884
+ "step": 7200
6885
+ },
6886
+ {
6887
+ "epoch": 3.26,
6888
+ "learning_rate": 0.0002,
6889
+ "loss": 0.6862,
6890
+ "step": 7210
6891
+ },
6892
+ {
6893
+ "epoch": 3.27,
6894
+ "learning_rate": 0.0002,
6895
+ "loss": 0.7972,
6896
+ "step": 7220
6897
+ },
6898
+ {
6899
+ "epoch": 3.27,
6900
+ "learning_rate": 0.0002,
6901
+ "loss": 0.726,
6902
+ "step": 7230
6903
+ },
6904
+ {
6905
+ "epoch": 3.28,
6906
+ "learning_rate": 0.0002,
6907
+ "loss": 0.7776,
6908
+ "step": 7240
6909
+ },
6910
+ {
6911
+ "epoch": 3.28,
6912
+ "learning_rate": 0.0002,
6913
+ "loss": 0.6825,
6914
+ "step": 7250
6915
+ },
6916
+ {
6917
+ "epoch": 3.29,
6918
+ "learning_rate": 0.0002,
6919
+ "loss": 0.6939,
6920
+ "step": 7260
6921
+ },
6922
+ {
6923
+ "epoch": 3.29,
6924
+ "learning_rate": 0.0002,
6925
+ "loss": 0.7239,
6926
+ "step": 7270
6927
+ },
6928
+ {
6929
+ "epoch": 3.3,
6930
+ "learning_rate": 0.0002,
6931
+ "loss": 0.6697,
6932
+ "step": 7280
6933
+ },
6934
+ {
6935
+ "epoch": 3.3,
6936
+ "learning_rate": 0.0002,
6937
+ "loss": 0.698,
6938
+ "step": 7290
6939
+ },
6940
+ {
6941
+ "epoch": 3.3,
6942
+ "learning_rate": 0.0002,
6943
+ "loss": 0.7927,
6944
+ "step": 7300
6945
+ },
6946
+ {
6947
+ "epoch": 3.31,
6948
+ "learning_rate": 0.0002,
6949
+ "loss": 0.6807,
6950
+ "step": 7310
6951
+ },
6952
+ {
6953
+ "epoch": 3.31,
6954
+ "learning_rate": 0.0002,
6955
+ "loss": 0.7725,
6956
+ "step": 7320
6957
+ },
6958
+ {
6959
+ "epoch": 3.32,
6960
+ "learning_rate": 0.0002,
6961
+ "loss": 0.7064,
6962
+ "step": 7330
6963
+ },
6964
+ {
6965
+ "epoch": 3.32,
6966
+ "learning_rate": 0.0002,
6967
+ "loss": 0.7148,
6968
+ "step": 7340
6969
+ },
6970
+ {
6971
+ "epoch": 3.33,
6972
+ "learning_rate": 0.0002,
6973
+ "loss": 0.6924,
6974
+ "step": 7350
6975
+ },
6976
+ {
6977
+ "epoch": 3.33,
6978
+ "learning_rate": 0.0002,
6979
+ "loss": 0.7719,
6980
+ "step": 7360
6981
+ },
6982
+ {
6983
+ "epoch": 3.34,
6984
+ "learning_rate": 0.0002,
6985
+ "loss": 0.7478,
6986
+ "step": 7370
6987
+ },
6988
+ {
6989
+ "epoch": 3.34,
6990
+ "learning_rate": 0.0002,
6991
+ "loss": 0.7487,
6992
+ "step": 7380
6993
+ },
6994
+ {
6995
+ "epoch": 3.35,
6996
+ "learning_rate": 0.0002,
6997
+ "loss": 0.7798,
6998
+ "step": 7390
6999
+ },
7000
+ {
7001
+ "epoch": 3.35,
7002
+ "learning_rate": 0.0002,
7003
+ "loss": 0.7839,
7004
+ "step": 7400
7005
+ },
7006
+ {
7007
+ "epoch": 3.35,
7008
+ "eval_loss": 1.097154974937439,
7009
+ "eval_runtime": 130.2825,
7010
+ "eval_samples_per_second": 7.676,
7011
+ "eval_steps_per_second": 3.838,
7012
+ "step": 7400
7013
+ },
7014
+ {
7015
+ "epoch": 3.35,
7016
+ "mmlu_eval_accuracy": 0.4937458297927038,
7017
+ "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
7018
+ "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
7019
+ "mmlu_eval_accuracy_astronomy": 0.5,
7020
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
7021
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
7022
+ "mmlu_eval_accuracy_college_biology": 0.3125,
7023
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
7024
+ "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
7025
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
7026
+ "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
7027
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
7028
+ "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
7029
+ "mmlu_eval_accuracy_conceptual_physics": 0.5,
7030
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
7031
+ "mmlu_eval_accuracy_electrical_engineering": 0.4375,
7032
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
7033
+ "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
7034
+ "mmlu_eval_accuracy_global_facts": 0.4,
7035
+ "mmlu_eval_accuracy_high_school_biology": 0.4375,
7036
+ "mmlu_eval_accuracy_high_school_chemistry": 0.45454545454545453,
7037
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
7038
+ "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
7039
+ "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
7040
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
7041
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
7042
+ "mmlu_eval_accuracy_high_school_mathematics": 0.3103448275862069,
7043
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.38461538461538464,
7044
+ "mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
7045
+ "mmlu_eval_accuracy_high_school_psychology": 0.7833333333333333,
7046
+ "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
7047
+ "mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
7048
+ "mmlu_eval_accuracy_high_school_world_history": 0.6153846153846154,
7049
+ "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
7050
+ "mmlu_eval_accuracy_human_sexuality": 0.5,
7051
+ "mmlu_eval_accuracy_international_law": 0.7692307692307693,
7052
+ "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
7053
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
7054
+ "mmlu_eval_accuracy_machine_learning": 0.09090909090909091,
7055
+ "mmlu_eval_accuracy_management": 0.7272727272727273,
7056
+ "mmlu_eval_accuracy_marketing": 0.72,
7057
+ "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
7058
+ "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
7059
+ "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
7060
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
7061
+ "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
7062
+ "mmlu_eval_accuracy_philosophy": 0.5,
7063
+ "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
7064
+ "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
7065
+ "mmlu_eval_accuracy_professional_law": 0.3,
7066
+ "mmlu_eval_accuracy_professional_medicine": 0.45161290322580644,
7067
+ "mmlu_eval_accuracy_professional_psychology": 0.4927536231884058,
7068
+ "mmlu_eval_accuracy_public_relations": 0.5,
7069
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
7070
+ "mmlu_eval_accuracy_sociology": 0.7727272727272727,
7071
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
7072
+ "mmlu_eval_accuracy_virology": 0.5,
7073
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
7074
+ "mmlu_loss": 1.3399821356639228,
7075
+ "step": 7400
7076
+ },
7077
+ {
7078
+ "epoch": 3.35,
7079
+ "learning_rate": 0.0002,
7080
+ "loss": 0.6669,
7081
+ "step": 7410
7082
+ },
7083
+ {
7084
+ "epoch": 3.36,
7085
+ "learning_rate": 0.0002,
7086
+ "loss": 0.7012,
7087
+ "step": 7420
7088
+ },
7089
+ {
7090
+ "epoch": 3.36,
7091
+ "learning_rate": 0.0002,
7092
+ "loss": 0.7149,
7093
+ "step": 7430
7094
+ },
7095
+ {
7096
+ "epoch": 3.37,
7097
+ "learning_rate": 0.0002,
7098
+ "loss": 0.6914,
7099
+ "step": 7440
7100
+ },
7101
+ {
7102
+ "epoch": 3.37,
7103
+ "learning_rate": 0.0002,
7104
+ "loss": 0.8015,
7105
+ "step": 7450
7106
+ },
7107
+ {
7108
+ "epoch": 3.38,
7109
+ "learning_rate": 0.0002,
7110
+ "loss": 0.7466,
7111
+ "step": 7460
7112
+ },
7113
+ {
7114
+ "epoch": 3.38,
7115
+ "learning_rate": 0.0002,
7116
+ "loss": 0.6992,
7117
+ "step": 7470
7118
+ },
7119
+ {
7120
+ "epoch": 3.39,
7121
+ "learning_rate": 0.0002,
7122
+ "loss": 0.7613,
7123
+ "step": 7480
7124
+ },
7125
+ {
7126
+ "epoch": 3.39,
7127
+ "learning_rate": 0.0002,
7128
+ "loss": 0.7227,
7129
+ "step": 7490
7130
+ },
7131
+ {
7132
+ "epoch": 3.4,
7133
+ "learning_rate": 0.0002,
7134
+ "loss": 0.7034,
7135
+ "step": 7500
7136
+ },
7137
+ {
7138
+ "epoch": 3.4,
7139
+ "learning_rate": 0.0002,
7140
+ "loss": 0.7285,
7141
+ "step": 7510
7142
+ },
7143
+ {
7144
+ "epoch": 3.4,
7145
+ "learning_rate": 0.0002,
7146
+ "loss": 0.7611,
7147
+ "step": 7520
7148
+ },
7149
+ {
7150
+ "epoch": 3.41,
7151
+ "learning_rate": 0.0002,
7152
+ "loss": 0.7201,
7153
+ "step": 7530
7154
+ },
7155
+ {
7156
+ "epoch": 3.41,
7157
+ "learning_rate": 0.0002,
7158
+ "loss": 0.7754,
7159
+ "step": 7540
7160
+ },
7161
+ {
7162
+ "epoch": 3.42,
7163
+ "learning_rate": 0.0002,
7164
+ "loss": 0.7218,
7165
+ "step": 7550
7166
+ },
7167
+ {
7168
+ "epoch": 3.42,
7169
+ "learning_rate": 0.0002,
7170
+ "loss": 0.7054,
7171
+ "step": 7560
7172
+ },
7173
+ {
7174
+ "epoch": 3.43,
7175
+ "learning_rate": 0.0002,
7176
+ "loss": 0.6354,
7177
+ "step": 7570
7178
+ },
7179
+ {
7180
+ "epoch": 3.43,
7181
+ "learning_rate": 0.0002,
7182
+ "loss": 0.6886,
7183
+ "step": 7580
7184
+ },
7185
+ {
7186
+ "epoch": 3.44,
7187
+ "learning_rate": 0.0002,
7188
+ "loss": 0.7139,
7189
+ "step": 7590
7190
+ },
7191
+ {
7192
+ "epoch": 3.44,
7193
+ "learning_rate": 0.0002,
7194
+ "loss": 0.7604,
7195
+ "step": 7600
7196
+ },
7197
+ {
7198
+ "epoch": 3.44,
7199
+ "eval_loss": 1.1017097234725952,
7200
+ "eval_runtime": 130.4446,
7201
+ "eval_samples_per_second": 7.666,
7202
+ "eval_steps_per_second": 3.833,
7203
+ "step": 7600
7204
+ },
7205
+ {
7206
+ "epoch": 3.44,
7207
+ "mmlu_eval_accuracy": 0.4983807903457635,
7208
+ "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
7209
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
7210
+ "mmlu_eval_accuracy_astronomy": 0.4375,
7211
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
7212
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5862068965517241,
7213
+ "mmlu_eval_accuracy_college_biology": 0.4375,
7214
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
7215
+ "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
7216
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
7217
+ "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
7218
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
7219
+ "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
7220
+ "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
7221
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
7222
+ "mmlu_eval_accuracy_electrical_engineering": 0.375,
7223
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
7224
+ "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
7225
+ "mmlu_eval_accuracy_global_facts": 0.5,
7226
+ "mmlu_eval_accuracy_high_school_biology": 0.5,
7227
+ "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
7228
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
7229
+ "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
7230
+ "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
7231
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
7232
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
7233
+ "mmlu_eval_accuracy_high_school_mathematics": 0.3103448275862069,
7234
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
7235
+ "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
7236
+ "mmlu_eval_accuracy_high_school_psychology": 0.8166666666666667,
7237
+ "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
7238
+ "mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
7239
+ "mmlu_eval_accuracy_high_school_world_history": 0.6153846153846154,
7240
+ "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
7241
+ "mmlu_eval_accuracy_human_sexuality": 0.5,
7242
+ "mmlu_eval_accuracy_international_law": 0.7692307692307693,
7243
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
7244
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
7245
+ "mmlu_eval_accuracy_machine_learning": 0.09090909090909091,
7246
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
7247
+ "mmlu_eval_accuracy_marketing": 0.76,
7248
+ "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
7249
+ "mmlu_eval_accuracy_miscellaneous": 0.6395348837209303,
7250
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
7251
+ "mmlu_eval_accuracy_moral_scenarios": 0.23,
7252
+ "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
7253
+ "mmlu_eval_accuracy_philosophy": 0.5588235294117647,
7254
+ "mmlu_eval_accuracy_prehistory": 0.5142857142857142,
7255
+ "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
7256
+ "mmlu_eval_accuracy_professional_law": 0.3235294117647059,
7257
+ "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
7258
+ "mmlu_eval_accuracy_professional_psychology": 0.4927536231884058,
7259
+ "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
7260
+ "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
7261
+ "mmlu_eval_accuracy_sociology": 0.8181818181818182,
7262
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
7263
+ "mmlu_eval_accuracy_virology": 0.5,
7264
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
7265
+ "mmlu_loss": 1.291005614726724,
7266
+ "step": 7600
7267
+ },
7268
+ {
7269
+ "epoch": 3.44,
7270
+ "learning_rate": 0.0002,
7271
+ "loss": 0.7426,
7272
+ "step": 7610
7273
+ },
7274
+ {
7275
+ "epoch": 3.45,
7276
+ "learning_rate": 0.0002,
7277
+ "loss": 0.6424,
7278
+ "step": 7620
7279
+ },
7280
+ {
7281
+ "epoch": 3.45,
7282
+ "learning_rate": 0.0002,
7283
+ "loss": 0.8097,
7284
+ "step": 7630
7285
+ },
7286
+ {
7287
+ "epoch": 3.46,
7288
+ "learning_rate": 0.0002,
7289
+ "loss": 0.7112,
7290
+ "step": 7640
7291
+ },
7292
+ {
7293
+ "epoch": 3.46,
7294
+ "learning_rate": 0.0002,
7295
+ "loss": 0.7544,
7296
+ "step": 7650
7297
+ },
7298
+ {
7299
+ "epoch": 3.47,
7300
+ "learning_rate": 0.0002,
7301
+ "loss": 0.7586,
7302
+ "step": 7660
7303
+ },
7304
+ {
7305
+ "epoch": 3.47,
7306
+ "learning_rate": 0.0002,
7307
+ "loss": 0.7623,
7308
+ "step": 7670
7309
+ },
7310
+ {
7311
+ "epoch": 3.48,
7312
+ "learning_rate": 0.0002,
7313
+ "loss": 0.6811,
7314
+ "step": 7680
7315
+ },
7316
+ {
7317
+ "epoch": 3.48,
7318
+ "learning_rate": 0.0002,
7319
+ "loss": 0.7471,
7320
+ "step": 7690
7321
+ },
7322
+ {
7323
+ "epoch": 3.49,
7324
+ "learning_rate": 0.0002,
7325
+ "loss": 0.7105,
7326
+ "step": 7700
7327
+ },
7328
+ {
7329
+ "epoch": 3.49,
7330
+ "learning_rate": 0.0002,
7331
+ "loss": 0.7718,
7332
+ "step": 7710
7333
+ },
7334
+ {
7335
+ "epoch": 3.49,
7336
+ "learning_rate": 0.0002,
7337
+ "loss": 0.6904,
7338
+ "step": 7720
7339
+ },
7340
+ {
7341
+ "epoch": 3.5,
7342
+ "learning_rate": 0.0002,
7343
+ "loss": 0.6791,
7344
+ "step": 7730
7345
+ },
7346
+ {
7347
+ "epoch": 3.5,
7348
+ "learning_rate": 0.0002,
7349
+ "loss": 0.6911,
7350
+ "step": 7740
7351
+ },
7352
+ {
7353
+ "epoch": 3.51,
7354
+ "learning_rate": 0.0002,
7355
+ "loss": 0.7351,
7356
+ "step": 7750
7357
+ },
7358
+ {
7359
+ "epoch": 3.51,
7360
+ "learning_rate": 0.0002,
7361
+ "loss": 0.7143,
7362
+ "step": 7760
7363
+ },
7364
+ {
7365
+ "epoch": 3.52,
7366
+ "learning_rate": 0.0002,
7367
+ "loss": 0.7219,
7368
+ "step": 7770
7369
+ },
7370
+ {
7371
+ "epoch": 3.52,
7372
+ "learning_rate": 0.0002,
7373
+ "loss": 0.6999,
7374
+ "step": 7780
7375
+ },
7376
+ {
7377
+ "epoch": 3.53,
7378
+ "learning_rate": 0.0002,
7379
+ "loss": 0.6962,
7380
+ "step": 7790
7381
+ },
7382
+ {
7383
+ "epoch": 3.53,
7384
+ "learning_rate": 0.0002,
7385
+ "loss": 0.7188,
7386
+ "step": 7800
7387
+ },
7388
+ {
7389
+ "epoch": 3.53,
7390
+ "eval_loss": 1.0986956357955933,
7391
+ "eval_runtime": 130.3908,
7392
+ "eval_samples_per_second": 7.669,
7393
+ "eval_steps_per_second": 3.835,
7394
+ "step": 7800
7395
+ },
7396
+ {
7397
+ "epoch": 3.53,
7398
+ "mmlu_eval_accuracy": 0.5129068178429581,
7399
+ "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
7400
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
7401
+ "mmlu_eval_accuracy_astronomy": 0.625,
7402
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
7403
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5862068965517241,
7404
+ "mmlu_eval_accuracy_college_biology": 0.375,
7405
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
7406
+ "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
7407
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
7408
+ "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
7409
+ "mmlu_eval_accuracy_college_physics": 0.5454545454545454,
7410
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
7411
+ "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
7412
+ "mmlu_eval_accuracy_econometrics": 0.25,
7413
+ "mmlu_eval_accuracy_electrical_engineering": 0.375,
7414
+ "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
7415
+ "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
7416
+ "mmlu_eval_accuracy_global_facts": 0.4,
7417
+ "mmlu_eval_accuracy_high_school_biology": 0.53125,
7418
+ "mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
7419
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
7420
+ "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
7421
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
7422
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
7423
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
7424
+ "mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
7425
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
7426
+ "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
7427
+ "mmlu_eval_accuracy_high_school_psychology": 0.8,
7428
+ "mmlu_eval_accuracy_high_school_statistics": 0.4782608695652174,
7429
+ "mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
7430
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
7431
+ "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
7432
+ "mmlu_eval_accuracy_human_sexuality": 0.5,
7433
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
7434
+ "mmlu_eval_accuracy_jurisprudence": 0.5454545454545454,
7435
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
7436
+ "mmlu_eval_accuracy_machine_learning": 0.09090909090909091,
7437
+ "mmlu_eval_accuracy_management": 0.7272727272727273,
7438
+ "mmlu_eval_accuracy_marketing": 0.8,
7439
+ "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
7440
+ "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
7441
+ "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
7442
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
7443
+ "mmlu_eval_accuracy_nutrition": 0.7272727272727273,
7444
+ "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
7445
+ "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
7446
+ "mmlu_eval_accuracy_professional_accounting": 0.25806451612903225,
7447
+ "mmlu_eval_accuracy_professional_law": 0.32941176470588235,
7448
+ "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
7449
+ "mmlu_eval_accuracy_professional_psychology": 0.463768115942029,
7450
+ "mmlu_eval_accuracy_public_relations": 0.5,
7451
+ "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
7452
+ "mmlu_eval_accuracy_sociology": 0.7727272727272727,
7453
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
7454
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
7455
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
7456
+ "mmlu_loss": 1.1720976933019902,
7457
+ "step": 7800
7458
+ },
7459
+ {
7460
+ "epoch": 3.54,
7461
+ "learning_rate": 0.0002,
7462
+ "loss": 0.7186,
7463
+ "step": 7810
7464
+ },
7465
+ {
7466
+ "epoch": 3.54,
7467
+ "learning_rate": 0.0002,
7468
+ "loss": 0.6645,
7469
+ "step": 7820
7470
+ },
7471
+ {
7472
+ "epoch": 3.54,
7473
+ "learning_rate": 0.0002,
7474
+ "loss": 0.7719,
7475
+ "step": 7830
7476
+ },
7477
+ {
7478
+ "epoch": 3.55,
7479
+ "learning_rate": 0.0002,
7480
+ "loss": 0.7619,
7481
+ "step": 7840
7482
+ },
7483
+ {
7484
+ "epoch": 3.55,
7485
+ "learning_rate": 0.0002,
7486
+ "loss": 0.7767,
7487
+ "step": 7850
7488
+ },
7489
+ {
7490
+ "epoch": 3.56,
7491
+ "learning_rate": 0.0002,
7492
+ "loss": 0.7032,
7493
+ "step": 7860
7494
+ },
7495
+ {
7496
+ "epoch": 3.56,
7497
+ "learning_rate": 0.0002,
7498
+ "loss": 0.7684,
7499
+ "step": 7870
7500
+ },
7501
+ {
7502
+ "epoch": 3.57,
7503
+ "learning_rate": 0.0002,
7504
+ "loss": 0.6622,
7505
+ "step": 7880
7506
+ },
7507
+ {
7508
+ "epoch": 3.57,
7509
+ "learning_rate": 0.0002,
7510
+ "loss": 0.6993,
7511
+ "step": 7890
7512
+ },
7513
+ {
7514
+ "epoch": 3.58,
7515
+ "learning_rate": 0.0002,
7516
+ "loss": 0.6745,
7517
+ "step": 7900
7518
+ },
7519
+ {
7520
+ "epoch": 3.58,
7521
+ "learning_rate": 0.0002,
7522
+ "loss": 0.7236,
7523
+ "step": 7910
7524
+ },
7525
+ {
7526
+ "epoch": 3.59,
7527
+ "learning_rate": 0.0002,
7528
+ "loss": 0.7068,
7529
+ "step": 7920
7530
+ },
7531
+ {
7532
+ "epoch": 3.59,
7533
+ "learning_rate": 0.0002,
7534
+ "loss": 0.7284,
7535
+ "step": 7930
7536
+ },
7537
+ {
7538
+ "epoch": 3.59,
7539
+ "learning_rate": 0.0002,
7540
+ "loss": 0.7565,
7541
+ "step": 7940
7542
+ },
7543
+ {
7544
+ "epoch": 3.6,
7545
+ "learning_rate": 0.0002,
7546
+ "loss": 0.7618,
7547
+ "step": 7950
7548
+ },
7549
+ {
7550
+ "epoch": 3.6,
7551
+ "learning_rate": 0.0002,
7552
+ "loss": 0.6919,
7553
+ "step": 7960
7554
+ },
7555
+ {
7556
+ "epoch": 3.61,
7557
+ "learning_rate": 0.0002,
7558
+ "loss": 0.7632,
7559
+ "step": 7970
7560
+ },
7561
+ {
7562
+ "epoch": 3.61,
7563
+ "learning_rate": 0.0002,
7564
+ "loss": 0.7072,
7565
+ "step": 7980
7566
+ },
7567
+ {
7568
+ "epoch": 3.62,
7569
+ "learning_rate": 0.0002,
7570
+ "loss": 0.7988,
7571
+ "step": 7990
7572
+ },
7573
+ {
7574
+ "epoch": 3.62,
7575
+ "learning_rate": 0.0002,
7576
+ "loss": 0.6931,
7577
+ "step": 8000
7578
+ },
7579
+ {
7580
+ "epoch": 3.62,
7581
+ "eval_loss": 1.0916101932525635,
7582
+ "eval_runtime": 130.4702,
7583
+ "eval_samples_per_second": 7.665,
7584
+ "eval_steps_per_second": 3.832,
7585
+ "step": 8000
7586
+ },
7587
+ {
7588
+ "epoch": 3.62,
7589
+ "mmlu_eval_accuracy": 0.5105854802224947,
7590
+ "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
7591
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
7592
+ "mmlu_eval_accuracy_astronomy": 0.5,
7593
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
7594
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5862068965517241,
7595
+ "mmlu_eval_accuracy_college_biology": 0.375,
7596
+ "mmlu_eval_accuracy_college_chemistry": 0.625,
7597
+ "mmlu_eval_accuracy_college_computer_science": 0.5454545454545454,
7598
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
7599
+ "mmlu_eval_accuracy_college_medicine": 0.5,
7600
+ "mmlu_eval_accuracy_college_physics": 0.5454545454545454,
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": 0.375,
7605
+ "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
7606
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
7607
+ "mmlu_eval_accuracy_global_facts": 0.3,
7608
+ "mmlu_eval_accuracy_high_school_biology": 0.5,
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": 0.6666666666666666,
7612
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
7613
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
7614
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.37209302325581395,
7615
+ "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
7616
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
7617
+ "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
7618
+ "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
7619
+ "mmlu_eval_accuracy_high_school_statistics": 0.4782608695652174,
7620
+ "mmlu_eval_accuracy_high_school_us_history": 0.7727272727272727,
7621
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
7622
+ "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
7623
+ "mmlu_eval_accuracy_human_sexuality": 0.5,
7624
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
7625
+ "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
7626
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
7627
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
7628
+ "mmlu_eval_accuracy_management": 0.7272727272727273,
7629
+ "mmlu_eval_accuracy_marketing": 0.8,
7630
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
7631
+ "mmlu_eval_accuracy_miscellaneous": 0.6395348837209303,
7632
+ "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
7633
+ "mmlu_eval_accuracy_moral_scenarios": 0.23,
7634
+ "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
7635
+ "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
7636
+ "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
7637
+ "mmlu_eval_accuracy_professional_accounting": 0.25806451612903225,
7638
+ "mmlu_eval_accuracy_professional_law": 0.3411764705882353,
7639
+ "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
7640
+ "mmlu_eval_accuracy_professional_psychology": 0.4492753623188406,
7641
+ "mmlu_eval_accuracy_public_relations": 0.4166666666666667,
7642
+ "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
7643
+ "mmlu_eval_accuracy_sociology": 0.8181818181818182,
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": 1.4539437902113788,
7648
+ "step": 8000
7649
+ },
7650
+ {
7651
+ "epoch": 3.63,
7652
+ "learning_rate": 0.0002,
7653
+ "loss": 0.6682,
7654
+ "step": 8010
7655
+ },
7656
+ {
7657
+ "epoch": 3.63,
7658
+ "learning_rate": 0.0002,
7659
+ "loss": 0.7483,
7660
+ "step": 8020
7661
+ },
7662
+ {
7663
+ "epoch": 3.64,
7664
+ "learning_rate": 0.0002,
7665
+ "loss": 0.697,
7666
+ "step": 8030
7667
+ },
7668
+ {
7669
+ "epoch": 3.64,
7670
+ "learning_rate": 0.0002,
7671
+ "loss": 0.7493,
7672
+ "step": 8040
7673
+ },
7674
+ {
7675
+ "epoch": 3.64,
7676
+ "learning_rate": 0.0002,
7677
+ "loss": 0.8147,
7678
+ "step": 8050
7679
+ },
7680
+ {
7681
+ "epoch": 3.65,
7682
+ "learning_rate": 0.0002,
7683
+ "loss": 0.7484,
7684
+ "step": 8060
7685
+ },
7686
+ {
7687
+ "epoch": 3.65,
7688
+ "learning_rate": 0.0002,
7689
+ "loss": 0.7828,
7690
+ "step": 8070
7691
+ },
7692
+ {
7693
+ "epoch": 3.66,
7694
+ "learning_rate": 0.0002,
7695
+ "loss": 0.6756,
7696
+ "step": 8080
7697
+ },
7698
+ {
7699
+ "epoch": 3.66,
7700
+ "learning_rate": 0.0002,
7701
+ "loss": 0.8711,
7702
+ "step": 8090
7703
+ },
7704
+ {
7705
+ "epoch": 3.67,
7706
+ "learning_rate": 0.0002,
7707
+ "loss": 0.7307,
7708
+ "step": 8100
7709
+ },
7710
+ {
7711
+ "epoch": 3.67,
7712
+ "learning_rate": 0.0002,
7713
+ "loss": 0.7542,
7714
+ "step": 8110
7715
+ },
7716
+ {
7717
+ "epoch": 3.68,
7718
+ "learning_rate": 0.0002,
7719
+ "loss": 0.722,
7720
+ "step": 8120
7721
+ },
7722
+ {
7723
+ "epoch": 3.68,
7724
+ "learning_rate": 0.0002,
7725
+ "loss": 0.7708,
7726
+ "step": 8130
7727
+ },
7728
+ {
7729
+ "epoch": 3.68,
7730
+ "learning_rate": 0.0002,
7731
+ "loss": 0.6998,
7732
+ "step": 8140
7733
+ },
7734
+ {
7735
+ "epoch": 3.69,
7736
+ "learning_rate": 0.0002,
7737
+ "loss": 0.7725,
7738
+ "step": 8150
7739
+ },
7740
+ {
7741
+ "epoch": 3.69,
7742
+ "learning_rate": 0.0002,
7743
+ "loss": 0.765,
7744
+ "step": 8160
7745
+ },
7746
+ {
7747
+ "epoch": 3.7,
7748
+ "learning_rate": 0.0002,
7749
+ "loss": 0.7523,
7750
+ "step": 8170
7751
+ },
7752
+ {
7753
+ "epoch": 3.7,
7754
+ "learning_rate": 0.0002,
7755
+ "loss": 0.7689,
7756
+ "step": 8180
7757
+ },
7758
+ {
7759
+ "epoch": 3.71,
7760
+ "learning_rate": 0.0002,
7761
+ "loss": 0.7549,
7762
+ "step": 8190
7763
+ },
7764
+ {
7765
+ "epoch": 3.71,
7766
+ "learning_rate": 0.0002,
7767
+ "loss": 0.7585,
7768
+ "step": 8200
7769
+ },
7770
+ {
7771
+ "epoch": 3.71,
7772
+ "eval_loss": 1.0938563346862793,
7773
+ "eval_runtime": 130.749,
7774
+ "eval_samples_per_second": 7.648,
7775
+ "eval_steps_per_second": 3.824,
7776
+ "step": 8200
7777
+ },
7778
+ {
7779
+ "epoch": 3.71,
7780
+ "mmlu_eval_accuracy": 0.509832805467066,
7781
+ "mmlu_eval_accuracy_abstract_algebra": 0.18181818181818182,
7782
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
7783
+ "mmlu_eval_accuracy_astronomy": 0.5,
7784
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
7785
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5862068965517241,
7786
+ "mmlu_eval_accuracy_college_biology": 0.3125,
7787
+ "mmlu_eval_accuracy_college_chemistry": 0.625,
7788
+ "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
7789
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
7790
+ "mmlu_eval_accuracy_college_medicine": 0.5,
7791
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
7792
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
7793
+ "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
7794
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
7795
+ "mmlu_eval_accuracy_electrical_engineering": 0.375,
7796
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
7797
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
7798
+ "mmlu_eval_accuracy_global_facts": 0.3,
7799
+ "mmlu_eval_accuracy_high_school_biology": 0.46875,
7800
+ "mmlu_eval_accuracy_high_school_chemistry": 0.45454545454545453,
7801
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
7802
+ "mmlu_eval_accuracy_high_school_european_history": 0.7222222222222222,
7803
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
7804
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
7805
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
7806
+ "mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
7807
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
7808
+ "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
7809
+ "mmlu_eval_accuracy_high_school_psychology": 0.85,
7810
+ "mmlu_eval_accuracy_high_school_statistics": 0.4782608695652174,
7811
+ "mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
7812
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
7813
+ "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
7814
+ "mmlu_eval_accuracy_human_sexuality": 0.5,
7815
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
7816
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
7817
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
7818
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
7819
+ "mmlu_eval_accuracy_management": 0.8181818181818182,
7820
+ "mmlu_eval_accuracy_marketing": 0.8,
7821
+ "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
7822
+ "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
7823
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
7824
+ "mmlu_eval_accuracy_moral_scenarios": 0.25,
7825
+ "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
7826
+ "mmlu_eval_accuracy_philosophy": 0.5,
7827
+ "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
7828
+ "mmlu_eval_accuracy_professional_accounting": 0.25806451612903225,
7829
+ "mmlu_eval_accuracy_professional_law": 0.34705882352941175,
7830
+ "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
7831
+ "mmlu_eval_accuracy_professional_psychology": 0.4927536231884058,
7832
+ "mmlu_eval_accuracy_public_relations": 0.4166666666666667,
7833
+ "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
7834
+ "mmlu_eval_accuracy_sociology": 0.8181818181818182,
7835
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
7836
+ "mmlu_eval_accuracy_virology": 0.5,
7837
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
7838
+ "mmlu_loss": 1.2966852889398681,
7839
+ "step": 8200
7840
+ },
7841
+ {
7842
+ "epoch": 3.72,
7843
+ "learning_rate": 0.0002,
7844
+ "loss": 0.757,
7845
+ "step": 8210
7846
+ },
7847
+ {
7848
+ "epoch": 3.72,
7849
+ "learning_rate": 0.0002,
7850
+ "loss": 0.7398,
7851
+ "step": 8220
7852
+ },
7853
+ {
7854
+ "epoch": 3.73,
7855
+ "learning_rate": 0.0002,
7856
+ "loss": 0.7923,
7857
+ "step": 8230
7858
+ },
7859
+ {
7860
+ "epoch": 3.73,
7861
+ "learning_rate": 0.0002,
7862
+ "loss": 0.7271,
7863
+ "step": 8240
7864
+ },
7865
+ {
7866
+ "epoch": 3.73,
7867
+ "learning_rate": 0.0002,
7868
+ "loss": 0.7932,
7869
+ "step": 8250
7870
+ },
7871
+ {
7872
+ "epoch": 3.74,
7873
+ "learning_rate": 0.0002,
7874
+ "loss": 0.7686,
7875
+ "step": 8260
7876
+ },
7877
+ {
7878
+ "epoch": 3.74,
7879
+ "learning_rate": 0.0002,
7880
+ "loss": 0.6725,
7881
+ "step": 8270
7882
+ },
7883
+ {
7884
+ "epoch": 3.75,
7885
+ "learning_rate": 0.0002,
7886
+ "loss": 0.7114,
7887
+ "step": 8280
7888
+ },
7889
+ {
7890
+ "epoch": 3.75,
7891
+ "learning_rate": 0.0002,
7892
+ "loss": 0.7855,
7893
+ "step": 8290
7894
+ },
7895
+ {
7896
+ "epoch": 3.76,
7897
+ "learning_rate": 0.0002,
7898
+ "loss": 0.7489,
7899
+ "step": 8300
7900
+ },
7901
+ {
7902
+ "epoch": 3.76,
7903
+ "learning_rate": 0.0002,
7904
+ "loss": 0.7611,
7905
+ "step": 8310
7906
+ },
7907
+ {
7908
+ "epoch": 3.77,
7909
+ "learning_rate": 0.0002,
7910
+ "loss": 0.7051,
7911
+ "step": 8320
7912
+ },
7913
+ {
7914
+ "epoch": 3.77,
7915
+ "learning_rate": 0.0002,
7916
+ "loss": 0.7394,
7917
+ "step": 8330
7918
+ },
7919
+ {
7920
+ "epoch": 3.78,
7921
+ "learning_rate": 0.0002,
7922
+ "loss": 0.7712,
7923
+ "step": 8340
7924
+ },
7925
+ {
7926
+ "epoch": 3.78,
7927
+ "learning_rate": 0.0002,
7928
+ "loss": 0.745,
7929
+ "step": 8350
7930
+ },
7931
+ {
7932
+ "epoch": 3.78,
7933
+ "learning_rate": 0.0002,
7934
+ "loss": 0.6815,
7935
+ "step": 8360
7936
+ },
7937
+ {
7938
+ "epoch": 3.79,
7939
+ "learning_rate": 0.0002,
7940
+ "loss": 0.6954,
7941
+ "step": 8370
7942
+ },
7943
+ {
7944
+ "epoch": 3.79,
7945
+ "learning_rate": 0.0002,
7946
+ "loss": 0.6684,
7947
+ "step": 8380
7948
+ },
7949
+ {
7950
+ "epoch": 3.8,
7951
+ "learning_rate": 0.0002,
7952
+ "loss": 0.7217,
7953
+ "step": 8390
7954
+ },
7955
+ {
7956
+ "epoch": 3.8,
7957
+ "learning_rate": 0.0002,
7958
+ "loss": 0.7681,
7959
+ "step": 8400
7960
+ },
7961
+ {
7962
+ "epoch": 3.8,
7963
+ "eval_loss": 1.101223111152649,
7964
+ "eval_runtime": 131.7525,
7965
+ "eval_samples_per_second": 7.59,
7966
+ "eval_steps_per_second": 3.795,
7967
+ "step": 8400
7968
+ },
7969
+ {
7970
+ "epoch": 3.8,
7971
+ "mmlu_eval_accuracy": 0.5146294321294028,
7972
+ "mmlu_eval_accuracy_abstract_algebra": 0.18181818181818182,
7973
+ "mmlu_eval_accuracy_anatomy": 0.5,
7974
+ "mmlu_eval_accuracy_astronomy": 0.5,
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": 0.625,
7979
+ "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
7980
+ "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
7981
+ "mmlu_eval_accuracy_college_medicine": 0.5454545454545454,
7982
+ "mmlu_eval_accuracy_college_physics": 0.5454545454545454,
7983
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
7984
+ "mmlu_eval_accuracy_conceptual_physics": 0.5,
7985
+ "mmlu_eval_accuracy_econometrics": 0.25,
7986
+ "mmlu_eval_accuracy_electrical_engineering": 0.375,
7987
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
7988
+ "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
7989
+ "mmlu_eval_accuracy_global_facts": 0.4,
7990
+ "mmlu_eval_accuracy_high_school_biology": 0.5,
7991
+ "mmlu_eval_accuracy_high_school_chemistry": 0.5,
7992
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
7993
+ "mmlu_eval_accuracy_high_school_european_history": 0.7222222222222222,
7994
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
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": 0.2413793103448276,
7998
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5769230769230769,
7999
+ "mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
8000
+ "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
8001
+ "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
8002
+ "mmlu_eval_accuracy_high_school_us_history": 0.7727272727272727,
8003
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
8004
+ "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
8005
+ "mmlu_eval_accuracy_human_sexuality": 0.5,
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": 0.7272727272727273,
8011
+ "mmlu_eval_accuracy_marketing": 0.84,
8012
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
8013
+ "mmlu_eval_accuracy_miscellaneous": 0.627906976744186,
8014
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
8015
+ "mmlu_eval_accuracy_moral_scenarios": 0.23,
8016
+ "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
8017
+ "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
8018
+ "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
8019
+ "mmlu_eval_accuracy_professional_accounting": 0.16129032258064516,
8020
+ "mmlu_eval_accuracy_professional_law": 0.32941176470588235,
8021
+ "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
8022
+ "mmlu_eval_accuracy_professional_psychology": 0.463768115942029,
8023
+ "mmlu_eval_accuracy_public_relations": 0.4166666666666667,
8024
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
8025
+ "mmlu_eval_accuracy_sociology": 0.8181818181818182,
8026
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
8027
+ "mmlu_eval_accuracy_virology": 0.4444444444444444,
8028
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
8029
+ "mmlu_loss": 1.40192161761127,
8030
+ "step": 8400
8031
+ },
8032
+ {
8033
+ "epoch": 3.81,
8034
+ "learning_rate": 0.0002,
8035
+ "loss": 0.7766,
8036
+ "step": 8410
8037
+ },
8038
+ {
8039
+ "epoch": 3.81,
8040
+ "learning_rate": 0.0002,
8041
+ "loss": 0.6984,
8042
+ "step": 8420
8043
+ },
8044
+ {
8045
+ "epoch": 3.82,
8046
+ "learning_rate": 0.0002,
8047
+ "loss": 0.7283,
8048
+ "step": 8430
8049
+ },
8050
+ {
8051
+ "epoch": 3.82,
8052
+ "learning_rate": 0.0002,
8053
+ "loss": 0.7122,
8054
+ "step": 8440
8055
+ },
8056
+ {
8057
+ "epoch": 3.83,
8058
+ "learning_rate": 0.0002,
8059
+ "loss": 0.7026,
8060
+ "step": 8450
8061
+ },
8062
+ {
8063
+ "epoch": 3.83,
8064
+ "learning_rate": 0.0002,
8065
+ "loss": 0.7262,
8066
+ "step": 8460
8067
+ },
8068
+ {
8069
+ "epoch": 3.83,
8070
+ "learning_rate": 0.0002,
8071
+ "loss": 0.692,
8072
+ "step": 8470
8073
+ },
8074
+ {
8075
+ "epoch": 3.84,
8076
+ "learning_rate": 0.0002,
8077
+ "loss": 0.7611,
8078
+ "step": 8480
8079
+ },
8080
+ {
8081
+ "epoch": 3.84,
8082
+ "learning_rate": 0.0002,
8083
+ "loss": 0.7602,
8084
+ "step": 8490
8085
+ },
8086
+ {
8087
+ "epoch": 3.85,
8088
+ "learning_rate": 0.0002,
8089
+ "loss": 0.728,
8090
+ "step": 8500
8091
+ },
8092
+ {
8093
+ "epoch": 3.85,
8094
+ "learning_rate": 0.0002,
8095
+ "loss": 0.6521,
8096
+ "step": 8510
8097
+ },
8098
+ {
8099
+ "epoch": 3.86,
8100
+ "learning_rate": 0.0002,
8101
+ "loss": 0.7307,
8102
+ "step": 8520
8103
+ },
8104
+ {
8105
+ "epoch": 3.86,
8106
+ "learning_rate": 0.0002,
8107
+ "loss": 0.706,
8108
+ "step": 8530
8109
+ },
8110
+ {
8111
+ "epoch": 3.87,
8112
+ "learning_rate": 0.0002,
8113
+ "loss": 0.7543,
8114
+ "step": 8540
8115
+ },
8116
+ {
8117
+ "epoch": 3.87,
8118
+ "learning_rate": 0.0002,
8119
+ "loss": 0.7194,
8120
+ "step": 8550
8121
+ },
8122
+ {
8123
+ "epoch": 3.88,
8124
+ "learning_rate": 0.0002,
8125
+ "loss": 0.7664,
8126
+ "step": 8560
8127
+ },
8128
+ {
8129
+ "epoch": 3.88,
8130
+ "learning_rate": 0.0002,
8131
+ "loss": 0.7834,
8132
+ "step": 8570
8133
+ },
8134
+ {
8135
+ "epoch": 3.88,
8136
+ "learning_rate": 0.0002,
8137
+ "loss": 0.7056,
8138
+ "step": 8580
8139
+ },
8140
+ {
8141
+ "epoch": 3.89,
8142
+ "learning_rate": 0.0002,
8143
+ "loss": 0.7865,
8144
+ "step": 8590
8145
+ },
8146
+ {
8147
+ "epoch": 3.89,
8148
+ "learning_rate": 0.0002,
8149
+ "loss": 0.7447,
8150
+ "step": 8600
8151
+ },
8152
+ {
8153
+ "epoch": 3.89,
8154
+ "eval_loss": 1.0886354446411133,
8155
+ "eval_runtime": 131.9847,
8156
+ "eval_samples_per_second": 7.577,
8157
+ "eval_steps_per_second": 3.788,
8158
+ "step": 8600
8159
+ },
8160
+ {
8161
+ "epoch": 3.89,
8162
+ "mmlu_eval_accuracy": 0.5103698915592155,
8163
+ "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
8164
+ "mmlu_eval_accuracy_anatomy": 0.5,
8165
+ "mmlu_eval_accuracy_astronomy": 0.5,
8166
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
8167
+ "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
8168
+ "mmlu_eval_accuracy_college_biology": 0.3125,
8169
+ "mmlu_eval_accuracy_college_chemistry": 0.5,
8170
+ "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
8171
+ "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
8172
+ "mmlu_eval_accuracy_college_medicine": 0.5,
8173
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
8174
+ "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
8175
+ "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
8176
+ "mmlu_eval_accuracy_econometrics": 0.25,
8177
+ "mmlu_eval_accuracy_electrical_engineering": 0.4375,
8178
+ "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
8179
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
8180
+ "mmlu_eval_accuracy_global_facts": 0.4,
8181
+ "mmlu_eval_accuracy_high_school_biology": 0.5,
8182
+ "mmlu_eval_accuracy_high_school_chemistry": 0.5,
8183
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
8184
+ "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
8185
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
8186
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
8187
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
8188
+ "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
8189
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5769230769230769,
8190
+ "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
8191
+ "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
8192
+ "mmlu_eval_accuracy_high_school_statistics": 0.4782608695652174,
8193
+ "mmlu_eval_accuracy_high_school_us_history": 0.7727272727272727,
8194
+ "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
8195
+ "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
8196
+ "mmlu_eval_accuracy_human_sexuality": 0.5,
8197
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
8198
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
8199
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
8200
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
8201
+ "mmlu_eval_accuracy_management": 0.8181818181818182,
8202
+ "mmlu_eval_accuracy_marketing": 0.8,
8203
+ "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
8204
+ "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
8205
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
8206
+ "mmlu_eval_accuracy_moral_scenarios": 0.23,
8207
+ "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
8208
+ "mmlu_eval_accuracy_philosophy": 0.5588235294117647,
8209
+ "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
8210
+ "mmlu_eval_accuracy_professional_accounting": 0.25806451612903225,
8211
+ "mmlu_eval_accuracy_professional_law": 0.3411764705882353,
8212
+ "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
8213
+ "mmlu_eval_accuracy_professional_psychology": 0.463768115942029,
8214
+ "mmlu_eval_accuracy_public_relations": 0.4166666666666667,
8215
+ "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
8216
+ "mmlu_eval_accuracy_sociology": 0.6818181818181818,
8217
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
8218
+ "mmlu_eval_accuracy_virology": 0.5,
8219
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
8220
+ "mmlu_loss": 1.1779243290307628,
8221
+ "step": 8600
8222
  }
8223
  ],
8224
  "max_steps": 10000,
8225
  "num_train_epochs": 5,
8226
+ "total_flos": 3.070850251305812e+18,
8227
  "trial_name": null,
8228
  "trial_params": null
8229
  }
{checkpoint-6600 → checkpoint-8600}/training_args.bin RENAMED
File without changes