Farouk commited on
Commit
5d0a8e4
·
1 Parent(s): 9f74e83

Training in progress, step 4000

Browse files
adapter_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1729484645601c6f411d93af1babae9d650991d95733f05ee9f95336be9e2528
3
  size 319977229
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ca27fcb1caa8cc1908f6cd4d9eaaadfcf65cc6aecfa1867be950a5c4da9bc745
3
  size 319977229
checkpoint-1800/adapter_model/adapter_model/README.md DELETED
@@ -1,32 +0,0 @@
1
- ---
2
- library_name: peft
3
- ---
4
- ## Training procedure
5
-
6
-
7
- The following `bitsandbytes` quantization config was used during training:
8
- - load_in_8bit: False
9
- - load_in_4bit: True
10
- - llm_int8_threshold: 6.0
11
- - llm_int8_skip_modules: None
12
- - llm_int8_enable_fp32_cpu_offload: False
13
- - llm_int8_has_fp16_weight: False
14
- - bnb_4bit_quant_type: nf4
15
- - bnb_4bit_use_double_quant: True
16
- - bnb_4bit_compute_dtype: bfloat16
17
-
18
- The following `bitsandbytes` quantization config was used during training:
19
- - load_in_8bit: False
20
- - load_in_4bit: True
21
- - llm_int8_threshold: 6.0
22
- - llm_int8_skip_modules: None
23
- - llm_int8_enable_fp32_cpu_offload: False
24
- - llm_int8_has_fp16_weight: False
25
- - bnb_4bit_quant_type: nf4
26
- - bnb_4bit_use_double_quant: True
27
- - bnb_4bit_compute_dtype: bfloat16
28
- ### Framework versions
29
-
30
- - PEFT 0.4.0
31
-
32
- - PEFT 0.4.0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
checkpoint-1800/adapter_model/adapter_model/adapter_config.json DELETED
@@ -1,26 +0,0 @@
1
- {
2
- "auto_mapping": null,
3
- "base_model_name_or_path": "pankajmathur/orca_mini_v3_7b",
4
- "bias": "none",
5
- "fan_in_fan_out": false,
6
- "inference_mode": true,
7
- "init_lora_weights": true,
8
- "layers_pattern": null,
9
- "layers_to_transform": null,
10
- "lora_alpha": 16.0,
11
- "lora_dropout": 0.1,
12
- "modules_to_save": null,
13
- "peft_type": "LORA",
14
- "r": 64,
15
- "revision": null,
16
- "target_modules": [
17
- "up_proj",
18
- "down_proj",
19
- "o_proj",
20
- "k_proj",
21
- "gate_proj",
22
- "q_proj",
23
- "v_proj"
24
- ],
25
- "task_type": "CAUSAL_LM"
26
- }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
checkpoint-1800/adapter_model/adapter_model/adapter_model.bin DELETED
@@ -1,3 +0,0 @@
1
- version https://git-lfs.github.com/spec/v1
2
- oid sha256:c28482802b48db4938d126d1ba697efcc6ddf1612c1a60d5a3feaaaafc55f542
3
- size 319977229
 
 
 
 
checkpoint-3600/adapter_model/adapter_model/README.md CHANGED
@@ -4,6 +4,17 @@ library_name: peft
4
  ## Training procedure
5
 
6
 
 
 
 
 
 
 
 
 
 
 
 
7
  The following `bitsandbytes` quantization config was used during training:
8
  - load_in_8bit: False
9
  - load_in_4bit: True
@@ -16,5 +27,6 @@ The following `bitsandbytes` quantization config was used during training:
16
  - bnb_4bit_compute_dtype: bfloat16
17
  ### Framework versions
18
 
 
19
 
20
  - PEFT 0.4.0
 
4
  ## Training procedure
5
 
6
 
7
+ The following `bitsandbytes` quantization config was used during training:
8
+ - load_in_8bit: False
9
+ - load_in_4bit: True
10
+ - llm_int8_threshold: 6.0
11
+ - llm_int8_skip_modules: None
12
+ - llm_int8_enable_fp32_cpu_offload: False
13
+ - llm_int8_has_fp16_weight: False
14
+ - bnb_4bit_quant_type: nf4
15
+ - bnb_4bit_use_double_quant: True
16
+ - bnb_4bit_compute_dtype: bfloat16
17
+
18
  The following `bitsandbytes` quantization config was used during training:
19
  - load_in_8bit: False
20
  - load_in_4bit: True
 
27
  - bnb_4bit_compute_dtype: bfloat16
28
  ### Framework versions
29
 
30
+ - PEFT 0.4.0
31
 
32
  - PEFT 0.4.0
checkpoint-3600/adapter_model/adapter_model/adapter_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:791269f88799b37760beb7acdc6b76b03f3ffdecc328d0ae043dbe349edc2d6f
3
  size 319977229
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1729484645601c6f411d93af1babae9d650991d95733f05ee9f95336be9e2528
3
  size 319977229
{checkpoint-1800 → checkpoint-4000}/README.md RENAMED
File without changes
{checkpoint-1800 → checkpoint-4000}/adapter_config.json RENAMED
File without changes
{checkpoint-1800 → checkpoint-4000}/adapter_model.bin RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:d4af70031687a0d66e1bd252ca378afbe209e26d666d1fcfa5c4d020d7f76e8f
3
  size 319977229
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ca27fcb1caa8cc1908f6cd4d9eaaadfcf65cc6aecfa1867be950a5c4da9bc745
3
  size 319977229
{checkpoint-1800 → checkpoint-4000}/added_tokens.json RENAMED
File without changes
{checkpoint-1800 → checkpoint-4000}/optimizer.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e68d4137d1cd2bfb681d6d56c64a4db6b5f6d8c342999bd01eeb491ea784758f
3
  size 1279539973
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de34c70c729d0746eb5f04fc9e96cb4cfdbcdc426396442999f2981a2fda7cd5
3
  size 1279539973
{checkpoint-1800 → checkpoint-4000}/rng_state.pth RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9570440de6c14d35c09438ff8743f0a14728fd20bd8ed5a80ff7138acf5633c1
3
  size 14511
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c8ce1c9f95fc279e8a184bbab19b124e8da0e24690333e01ea7c2b87a1e85aa
3
  size 14511
{checkpoint-1800 → checkpoint-4000}/scheduler.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:c0e0a1736a46fd1627af3c246e44261aaac909256abbd413b5ee5c968f6b2d8e
3
  size 627
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:593f43b0c83adb27a2db37a6418c2ef12a213bbc2a02f2dc881de6846a69a931
3
  size 627
{checkpoint-1800 → checkpoint-4000}/special_tokens_map.json RENAMED
File without changes
{checkpoint-1800 → checkpoint-4000}/tokenizer.model RENAMED
File without changes
{checkpoint-1800 → checkpoint-4000}/tokenizer_config.json RENAMED
File without changes
{checkpoint-1800 → checkpoint-4000}/trainer_state.json RENAMED
@@ -1,8 +1,8 @@
1
  {
2
- "best_metric": 0.7712340354919434,
3
- "best_model_checkpoint": "experts/expert-5/checkpoint-1800",
4
- "epoch": 0.4356243949661181,
5
- "global_step": 1800,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
@@ -1725,11 +1725,2112 @@
1725
  "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
1726
  "mmlu_loss": 1.2632950348657976,
1727
  "step": 1800
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1728
  }
1729
  ],
1730
  "max_steps": 10000,
1731
  "num_train_epochs": 3,
1732
- "total_flos": 1.98382028068651e+17,
1733
  "trial_name": null,
1734
  "trial_params": null
1735
  }
 
1
  {
2
+ "best_metric": 0.758108913898468,
3
+ "best_model_checkpoint": "experts/expert-5/checkpoint-4000",
4
+ "epoch": 0.968054211035818,
5
+ "global_step": 4000,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
 
1725
  "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
1726
  "mmlu_loss": 1.2632950348657976,
1727
  "step": 1800
1728
+ },
1729
+ {
1730
+ "epoch": 0.44,
1731
+ "learning_rate": 0.0002,
1732
+ "loss": 0.7185,
1733
+ "step": 1810
1734
+ },
1735
+ {
1736
+ "epoch": 0.44,
1737
+ "learning_rate": 0.0002,
1738
+ "loss": 0.6711,
1739
+ "step": 1820
1740
+ },
1741
+ {
1742
+ "epoch": 0.44,
1743
+ "learning_rate": 0.0002,
1744
+ "loss": 0.7649,
1745
+ "step": 1830
1746
+ },
1747
+ {
1748
+ "epoch": 0.45,
1749
+ "learning_rate": 0.0002,
1750
+ "loss": 0.8121,
1751
+ "step": 1840
1752
+ },
1753
+ {
1754
+ "epoch": 0.45,
1755
+ "learning_rate": 0.0002,
1756
+ "loss": 0.7883,
1757
+ "step": 1850
1758
+ },
1759
+ {
1760
+ "epoch": 0.45,
1761
+ "learning_rate": 0.0002,
1762
+ "loss": 0.7323,
1763
+ "step": 1860
1764
+ },
1765
+ {
1766
+ "epoch": 0.45,
1767
+ "learning_rate": 0.0002,
1768
+ "loss": 0.7326,
1769
+ "step": 1870
1770
+ },
1771
+ {
1772
+ "epoch": 0.45,
1773
+ "learning_rate": 0.0002,
1774
+ "loss": 0.7225,
1775
+ "step": 1880
1776
+ },
1777
+ {
1778
+ "epoch": 0.46,
1779
+ "learning_rate": 0.0002,
1780
+ "loss": 0.8117,
1781
+ "step": 1890
1782
+ },
1783
+ {
1784
+ "epoch": 0.46,
1785
+ "learning_rate": 0.0002,
1786
+ "loss": 0.8257,
1787
+ "step": 1900
1788
+ },
1789
+ {
1790
+ "epoch": 0.46,
1791
+ "learning_rate": 0.0002,
1792
+ "loss": 0.806,
1793
+ "step": 1910
1794
+ },
1795
+ {
1796
+ "epoch": 0.46,
1797
+ "learning_rate": 0.0002,
1798
+ "loss": 0.789,
1799
+ "step": 1920
1800
+ },
1801
+ {
1802
+ "epoch": 0.47,
1803
+ "learning_rate": 0.0002,
1804
+ "loss": 0.778,
1805
+ "step": 1930
1806
+ },
1807
+ {
1808
+ "epoch": 0.47,
1809
+ "learning_rate": 0.0002,
1810
+ "loss": 0.7417,
1811
+ "step": 1940
1812
+ },
1813
+ {
1814
+ "epoch": 0.47,
1815
+ "learning_rate": 0.0002,
1816
+ "loss": 0.8278,
1817
+ "step": 1950
1818
+ },
1819
+ {
1820
+ "epoch": 0.47,
1821
+ "learning_rate": 0.0002,
1822
+ "loss": 0.7784,
1823
+ "step": 1960
1824
+ },
1825
+ {
1826
+ "epoch": 0.48,
1827
+ "learning_rate": 0.0002,
1828
+ "loss": 0.791,
1829
+ "step": 1970
1830
+ },
1831
+ {
1832
+ "epoch": 0.48,
1833
+ "learning_rate": 0.0002,
1834
+ "loss": 0.8058,
1835
+ "step": 1980
1836
+ },
1837
+ {
1838
+ "epoch": 0.48,
1839
+ "learning_rate": 0.0002,
1840
+ "loss": 0.8396,
1841
+ "step": 1990
1842
+ },
1843
+ {
1844
+ "epoch": 0.48,
1845
+ "learning_rate": 0.0002,
1846
+ "loss": 0.7755,
1847
+ "step": 2000
1848
+ },
1849
+ {
1850
+ "epoch": 0.48,
1851
+ "eval_loss": 0.7712973952293396,
1852
+ "eval_runtime": 147.0365,
1853
+ "eval_samples_per_second": 6.801,
1854
+ "eval_steps_per_second": 3.401,
1855
+ "step": 2000
1856
+ },
1857
+ {
1858
+ "epoch": 0.48,
1859
+ "mmlu_eval_accuracy": 0.49357278366151575,
1860
+ "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
1861
+ "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
1862
+ "mmlu_eval_accuracy_astronomy": 0.5,
1863
+ "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
1864
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
1865
+ "mmlu_eval_accuracy_college_biology": 0.5,
1866
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
1867
+ "mmlu_eval_accuracy_college_computer_science": 0.09090909090909091,
1868
+ "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
1869
+ "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
1870
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
1871
+ "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
1872
+ "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
1873
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
1874
+ "mmlu_eval_accuracy_electrical_engineering": 0.375,
1875
+ "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
1876
+ "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
1877
+ "mmlu_eval_accuracy_global_facts": 0.4,
1878
+ "mmlu_eval_accuracy_high_school_biology": 0.34375,
1879
+ "mmlu_eval_accuracy_high_school_chemistry": 0.22727272727272727,
1880
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
1881
+ "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
1882
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
1883
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
1884
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
1885
+ "mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
1886
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
1887
+ "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
1888
+ "mmlu_eval_accuracy_high_school_psychology": 0.8833333333333333,
1889
+ "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
1890
+ "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
1891
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
1892
+ "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
1893
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
1894
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
1895
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
1896
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
1897
+ "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
1898
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
1899
+ "mmlu_eval_accuracy_marketing": 0.92,
1900
+ "mmlu_eval_accuracy_medical_genetics": 1.0,
1901
+ "mmlu_eval_accuracy_miscellaneous": 0.6395348837209303,
1902
+ "mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
1903
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
1904
+ "mmlu_eval_accuracy_nutrition": 0.7272727272727273,
1905
+ "mmlu_eval_accuracy_philosophy": 0.5,
1906
+ "mmlu_eval_accuracy_prehistory": 0.5142857142857142,
1907
+ "mmlu_eval_accuracy_professional_accounting": 0.41935483870967744,
1908
+ "mmlu_eval_accuracy_professional_law": 0.31176470588235294,
1909
+ "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
1910
+ "mmlu_eval_accuracy_professional_psychology": 0.4927536231884058,
1911
+ "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
1912
+ "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
1913
+ "mmlu_eval_accuracy_sociology": 0.6818181818181818,
1914
+ "mmlu_eval_accuracy_us_foreign_policy": 0.45454545454545453,
1915
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
1916
+ "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
1917
+ "mmlu_loss": 1.206857019087042,
1918
+ "step": 2000
1919
+ },
1920
+ {
1921
+ "epoch": 0.49,
1922
+ "learning_rate": 0.0002,
1923
+ "loss": 0.8252,
1924
+ "step": 2010
1925
+ },
1926
+ {
1927
+ "epoch": 0.49,
1928
+ "learning_rate": 0.0002,
1929
+ "loss": 0.7394,
1930
+ "step": 2020
1931
+ },
1932
+ {
1933
+ "epoch": 0.49,
1934
+ "learning_rate": 0.0002,
1935
+ "loss": 0.7894,
1936
+ "step": 2030
1937
+ },
1938
+ {
1939
+ "epoch": 0.49,
1940
+ "learning_rate": 0.0002,
1941
+ "loss": 0.7243,
1942
+ "step": 2040
1943
+ },
1944
+ {
1945
+ "epoch": 0.5,
1946
+ "learning_rate": 0.0002,
1947
+ "loss": 0.7644,
1948
+ "step": 2050
1949
+ },
1950
+ {
1951
+ "epoch": 0.5,
1952
+ "learning_rate": 0.0002,
1953
+ "loss": 0.7672,
1954
+ "step": 2060
1955
+ },
1956
+ {
1957
+ "epoch": 0.5,
1958
+ "learning_rate": 0.0002,
1959
+ "loss": 0.797,
1960
+ "step": 2070
1961
+ },
1962
+ {
1963
+ "epoch": 0.5,
1964
+ "learning_rate": 0.0002,
1965
+ "loss": 0.7323,
1966
+ "step": 2080
1967
+ },
1968
+ {
1969
+ "epoch": 0.51,
1970
+ "learning_rate": 0.0002,
1971
+ "loss": 0.7562,
1972
+ "step": 2090
1973
+ },
1974
+ {
1975
+ "epoch": 0.51,
1976
+ "learning_rate": 0.0002,
1977
+ "loss": 0.7778,
1978
+ "step": 2100
1979
+ },
1980
+ {
1981
+ "epoch": 0.51,
1982
+ "learning_rate": 0.0002,
1983
+ "loss": 0.7975,
1984
+ "step": 2110
1985
+ },
1986
+ {
1987
+ "epoch": 0.51,
1988
+ "learning_rate": 0.0002,
1989
+ "loss": 0.7042,
1990
+ "step": 2120
1991
+ },
1992
+ {
1993
+ "epoch": 0.52,
1994
+ "learning_rate": 0.0002,
1995
+ "loss": 0.7728,
1996
+ "step": 2130
1997
+ },
1998
+ {
1999
+ "epoch": 0.52,
2000
+ "learning_rate": 0.0002,
2001
+ "loss": 0.7602,
2002
+ "step": 2140
2003
+ },
2004
+ {
2005
+ "epoch": 0.52,
2006
+ "learning_rate": 0.0002,
2007
+ "loss": 0.824,
2008
+ "step": 2150
2009
+ },
2010
+ {
2011
+ "epoch": 0.52,
2012
+ "learning_rate": 0.0002,
2013
+ "loss": 0.8124,
2014
+ "step": 2160
2015
+ },
2016
+ {
2017
+ "epoch": 0.53,
2018
+ "learning_rate": 0.0002,
2019
+ "loss": 0.711,
2020
+ "step": 2170
2021
+ },
2022
+ {
2023
+ "epoch": 0.53,
2024
+ "learning_rate": 0.0002,
2025
+ "loss": 0.7865,
2026
+ "step": 2180
2027
+ },
2028
+ {
2029
+ "epoch": 0.53,
2030
+ "learning_rate": 0.0002,
2031
+ "loss": 0.7878,
2032
+ "step": 2190
2033
+ },
2034
+ {
2035
+ "epoch": 0.53,
2036
+ "learning_rate": 0.0002,
2037
+ "loss": 0.8129,
2038
+ "step": 2200
2039
+ },
2040
+ {
2041
+ "epoch": 0.53,
2042
+ "eval_loss": 0.7692970037460327,
2043
+ "eval_runtime": 147.118,
2044
+ "eval_samples_per_second": 6.797,
2045
+ "eval_steps_per_second": 3.399,
2046
+ "step": 2200
2047
+ },
2048
+ {
2049
+ "epoch": 0.53,
2050
+ "mmlu_eval_accuracy": 0.4957801049230341,
2051
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
2052
+ "mmlu_eval_accuracy_anatomy": 0.7857142857142857,
2053
+ "mmlu_eval_accuracy_astronomy": 0.5,
2054
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
2055
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
2056
+ "mmlu_eval_accuracy_college_biology": 0.4375,
2057
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
2058
+ "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
2059
+ "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
2060
+ "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
2061
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
2062
+ "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
2063
+ "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
2064
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
2065
+ "mmlu_eval_accuracy_electrical_engineering": 0.3125,
2066
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
2067
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
2068
+ "mmlu_eval_accuracy_global_facts": 0.4,
2069
+ "mmlu_eval_accuracy_high_school_biology": 0.40625,
2070
+ "mmlu_eval_accuracy_high_school_chemistry": 0.18181818181818182,
2071
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
2072
+ "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
2073
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
2074
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
2075
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
2076
+ "mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
2077
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
2078
+ "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
2079
+ "mmlu_eval_accuracy_high_school_psychology": 0.8833333333333333,
2080
+ "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
2081
+ "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
2082
+ "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
2083
+ "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
2084
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
2085
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
2086
+ "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
2087
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
2088
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
2089
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
2090
+ "mmlu_eval_accuracy_marketing": 0.84,
2091
+ "mmlu_eval_accuracy_medical_genetics": 1.0,
2092
+ "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
2093
+ "mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
2094
+ "mmlu_eval_accuracy_moral_scenarios": 0.25,
2095
+ "mmlu_eval_accuracy_nutrition": 0.696969696969697,
2096
+ "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
2097
+ "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
2098
+ "mmlu_eval_accuracy_professional_accounting": 0.3870967741935484,
2099
+ "mmlu_eval_accuracy_professional_law": 0.3352941176470588,
2100
+ "mmlu_eval_accuracy_professional_medicine": 0.6129032258064516,
2101
+ "mmlu_eval_accuracy_professional_psychology": 0.4927536231884058,
2102
+ "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
2103
+ "mmlu_eval_accuracy_security_studies": 0.4074074074074074,
2104
+ "mmlu_eval_accuracy_sociology": 0.7727272727272727,
2105
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
2106
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
2107
+ "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
2108
+ "mmlu_loss": 1.2995749724829477,
2109
+ "step": 2200
2110
+ },
2111
+ {
2112
+ "epoch": 0.53,
2113
+ "learning_rate": 0.0002,
2114
+ "loss": 0.7872,
2115
+ "step": 2210
2116
+ },
2117
+ {
2118
+ "epoch": 0.54,
2119
+ "learning_rate": 0.0002,
2120
+ "loss": 0.8147,
2121
+ "step": 2220
2122
+ },
2123
+ {
2124
+ "epoch": 0.54,
2125
+ "learning_rate": 0.0002,
2126
+ "loss": 0.7864,
2127
+ "step": 2230
2128
+ },
2129
+ {
2130
+ "epoch": 0.54,
2131
+ "learning_rate": 0.0002,
2132
+ "loss": 0.7246,
2133
+ "step": 2240
2134
+ },
2135
+ {
2136
+ "epoch": 0.54,
2137
+ "learning_rate": 0.0002,
2138
+ "loss": 0.803,
2139
+ "step": 2250
2140
+ },
2141
+ {
2142
+ "epoch": 0.55,
2143
+ "learning_rate": 0.0002,
2144
+ "loss": 0.8007,
2145
+ "step": 2260
2146
+ },
2147
+ {
2148
+ "epoch": 0.55,
2149
+ "learning_rate": 0.0002,
2150
+ "loss": 0.8184,
2151
+ "step": 2270
2152
+ },
2153
+ {
2154
+ "epoch": 0.55,
2155
+ "learning_rate": 0.0002,
2156
+ "loss": 0.7324,
2157
+ "step": 2280
2158
+ },
2159
+ {
2160
+ "epoch": 0.55,
2161
+ "learning_rate": 0.0002,
2162
+ "loss": 0.7753,
2163
+ "step": 2290
2164
+ },
2165
+ {
2166
+ "epoch": 0.56,
2167
+ "learning_rate": 0.0002,
2168
+ "loss": 0.7539,
2169
+ "step": 2300
2170
+ },
2171
+ {
2172
+ "epoch": 0.56,
2173
+ "learning_rate": 0.0002,
2174
+ "loss": 0.8573,
2175
+ "step": 2310
2176
+ },
2177
+ {
2178
+ "epoch": 0.56,
2179
+ "learning_rate": 0.0002,
2180
+ "loss": 0.7916,
2181
+ "step": 2320
2182
+ },
2183
+ {
2184
+ "epoch": 0.56,
2185
+ "learning_rate": 0.0002,
2186
+ "loss": 0.7771,
2187
+ "step": 2330
2188
+ },
2189
+ {
2190
+ "epoch": 0.57,
2191
+ "learning_rate": 0.0002,
2192
+ "loss": 0.7613,
2193
+ "step": 2340
2194
+ },
2195
+ {
2196
+ "epoch": 0.57,
2197
+ "learning_rate": 0.0002,
2198
+ "loss": 0.8023,
2199
+ "step": 2350
2200
+ },
2201
+ {
2202
+ "epoch": 0.57,
2203
+ "learning_rate": 0.0002,
2204
+ "loss": 0.7853,
2205
+ "step": 2360
2206
+ },
2207
+ {
2208
+ "epoch": 0.57,
2209
+ "learning_rate": 0.0002,
2210
+ "loss": 0.7421,
2211
+ "step": 2370
2212
+ },
2213
+ {
2214
+ "epoch": 0.58,
2215
+ "learning_rate": 0.0002,
2216
+ "loss": 0.7429,
2217
+ "step": 2380
2218
+ },
2219
+ {
2220
+ "epoch": 0.58,
2221
+ "learning_rate": 0.0002,
2222
+ "loss": 0.7937,
2223
+ "step": 2390
2224
+ },
2225
+ {
2226
+ "epoch": 0.58,
2227
+ "learning_rate": 0.0002,
2228
+ "loss": 0.7076,
2229
+ "step": 2400
2230
+ },
2231
+ {
2232
+ "epoch": 0.58,
2233
+ "eval_loss": 0.7674811482429504,
2234
+ "eval_runtime": 147.0136,
2235
+ "eval_samples_per_second": 6.802,
2236
+ "eval_steps_per_second": 3.401,
2237
+ "step": 2400
2238
+ },
2239
+ {
2240
+ "epoch": 0.58,
2241
+ "mmlu_eval_accuracy": 0.49364105387872714,
2242
+ "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
2243
+ "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
2244
+ "mmlu_eval_accuracy_astronomy": 0.4375,
2245
+ "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
2246
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
2247
+ "mmlu_eval_accuracy_college_biology": 0.5,
2248
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
2249
+ "mmlu_eval_accuracy_college_computer_science": 0.09090909090909091,
2250
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
2251
+ "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
2252
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
2253
+ "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
2254
+ "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
2255
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
2256
+ "mmlu_eval_accuracy_electrical_engineering": 0.3125,
2257
+ "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
2258
+ "mmlu_eval_accuracy_formal_logic": 0.07142857142857142,
2259
+ "mmlu_eval_accuracy_global_facts": 0.5,
2260
+ "mmlu_eval_accuracy_high_school_biology": 0.34375,
2261
+ "mmlu_eval_accuracy_high_school_chemistry": 0.22727272727272727,
2262
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
2263
+ "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
2264
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
2265
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
2266
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
2267
+ "mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
2268
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
2269
+ "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
2270
+ "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
2271
+ "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
2272
+ "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
2273
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
2274
+ "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
2275
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
2276
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
2277
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
2278
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
2279
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
2280
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
2281
+ "mmlu_eval_accuracy_marketing": 0.8,
2282
+ "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
2283
+ "mmlu_eval_accuracy_miscellaneous": 0.686046511627907,
2284
+ "mmlu_eval_accuracy_moral_disputes": 0.5789473684210527,
2285
+ "mmlu_eval_accuracy_moral_scenarios": 0.25,
2286
+ "mmlu_eval_accuracy_nutrition": 0.7272727272727273,
2287
+ "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
2288
+ "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
2289
+ "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
2290
+ "mmlu_eval_accuracy_professional_law": 0.32941176470588235,
2291
+ "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
2292
+ "mmlu_eval_accuracy_professional_psychology": 0.5362318840579711,
2293
+ "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
2294
+ "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
2295
+ "mmlu_eval_accuracy_sociology": 0.6818181818181818,
2296
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
2297
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
2298
+ "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
2299
+ "mmlu_loss": 1.2485307624694884,
2300
+ "step": 2400
2301
+ },
2302
+ {
2303
+ "epoch": 0.58,
2304
+ "learning_rate": 0.0002,
2305
+ "loss": 0.7945,
2306
+ "step": 2410
2307
+ },
2308
+ {
2309
+ "epoch": 0.59,
2310
+ "learning_rate": 0.0002,
2311
+ "loss": 0.7821,
2312
+ "step": 2420
2313
+ },
2314
+ {
2315
+ "epoch": 0.59,
2316
+ "learning_rate": 0.0002,
2317
+ "loss": 0.7279,
2318
+ "step": 2430
2319
+ },
2320
+ {
2321
+ "epoch": 0.59,
2322
+ "learning_rate": 0.0002,
2323
+ "loss": 0.7193,
2324
+ "step": 2440
2325
+ },
2326
+ {
2327
+ "epoch": 0.59,
2328
+ "learning_rate": 0.0002,
2329
+ "loss": 0.7915,
2330
+ "step": 2450
2331
+ },
2332
+ {
2333
+ "epoch": 0.6,
2334
+ "learning_rate": 0.0002,
2335
+ "loss": 0.7948,
2336
+ "step": 2460
2337
+ },
2338
+ {
2339
+ "epoch": 0.6,
2340
+ "learning_rate": 0.0002,
2341
+ "loss": 0.7497,
2342
+ "step": 2470
2343
+ },
2344
+ {
2345
+ "epoch": 0.6,
2346
+ "learning_rate": 0.0002,
2347
+ "loss": 0.774,
2348
+ "step": 2480
2349
+ },
2350
+ {
2351
+ "epoch": 0.6,
2352
+ "learning_rate": 0.0002,
2353
+ "loss": 0.7792,
2354
+ "step": 2490
2355
+ },
2356
+ {
2357
+ "epoch": 0.61,
2358
+ "learning_rate": 0.0002,
2359
+ "loss": 0.847,
2360
+ "step": 2500
2361
+ },
2362
+ {
2363
+ "epoch": 0.61,
2364
+ "learning_rate": 0.0002,
2365
+ "loss": 0.7087,
2366
+ "step": 2510
2367
+ },
2368
+ {
2369
+ "epoch": 0.61,
2370
+ "learning_rate": 0.0002,
2371
+ "loss": 0.9396,
2372
+ "step": 2520
2373
+ },
2374
+ {
2375
+ "epoch": 0.61,
2376
+ "learning_rate": 0.0002,
2377
+ "loss": 0.8027,
2378
+ "step": 2530
2379
+ },
2380
+ {
2381
+ "epoch": 0.61,
2382
+ "learning_rate": 0.0002,
2383
+ "loss": 0.7517,
2384
+ "step": 2540
2385
+ },
2386
+ {
2387
+ "epoch": 0.62,
2388
+ "learning_rate": 0.0002,
2389
+ "loss": 0.8338,
2390
+ "step": 2550
2391
+ },
2392
+ {
2393
+ "epoch": 0.62,
2394
+ "learning_rate": 0.0002,
2395
+ "loss": 0.7267,
2396
+ "step": 2560
2397
+ },
2398
+ {
2399
+ "epoch": 0.62,
2400
+ "learning_rate": 0.0002,
2401
+ "loss": 0.7424,
2402
+ "step": 2570
2403
+ },
2404
+ {
2405
+ "epoch": 0.62,
2406
+ "learning_rate": 0.0002,
2407
+ "loss": 0.8404,
2408
+ "step": 2580
2409
+ },
2410
+ {
2411
+ "epoch": 0.63,
2412
+ "learning_rate": 0.0002,
2413
+ "loss": 0.8795,
2414
+ "step": 2590
2415
+ },
2416
+ {
2417
+ "epoch": 0.63,
2418
+ "learning_rate": 0.0002,
2419
+ "loss": 0.7453,
2420
+ "step": 2600
2421
+ },
2422
+ {
2423
+ "epoch": 0.63,
2424
+ "eval_loss": 0.7636871337890625,
2425
+ "eval_runtime": 147.0736,
2426
+ "eval_samples_per_second": 6.799,
2427
+ "eval_steps_per_second": 3.4,
2428
+ "step": 2600
2429
+ },
2430
+ {
2431
+ "epoch": 0.63,
2432
+ "mmlu_eval_accuracy": 0.4929898018311285,
2433
+ "mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
2434
+ "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
2435
+ "mmlu_eval_accuracy_astronomy": 0.5,
2436
+ "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
2437
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
2438
+ "mmlu_eval_accuracy_college_biology": 0.5,
2439
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
2440
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
2441
+ "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
2442
+ "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
2443
+ "mmlu_eval_accuracy_college_physics": 0.2727272727272727,
2444
+ "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
2445
+ "mmlu_eval_accuracy_conceptual_physics": 0.34615384615384615,
2446
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
2447
+ "mmlu_eval_accuracy_electrical_engineering": 0.25,
2448
+ "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
2449
+ "mmlu_eval_accuracy_formal_logic": 0.07142857142857142,
2450
+ "mmlu_eval_accuracy_global_facts": 0.5,
2451
+ "mmlu_eval_accuracy_high_school_biology": 0.4375,
2452
+ "mmlu_eval_accuracy_high_school_chemistry": 0.18181818181818182,
2453
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
2454
+ "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
2455
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
2456
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
2457
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
2458
+ "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
2459
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
2460
+ "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
2461
+ "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
2462
+ "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
2463
+ "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
2464
+ "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
2465
+ "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
2466
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
2467
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
2468
+ "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
2469
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
2470
+ "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
2471
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
2472
+ "mmlu_eval_accuracy_marketing": 0.76,
2473
+ "mmlu_eval_accuracy_medical_genetics": 1.0,
2474
+ "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
2475
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
2476
+ "mmlu_eval_accuracy_moral_scenarios": 0.18,
2477
+ "mmlu_eval_accuracy_nutrition": 0.696969696969697,
2478
+ "mmlu_eval_accuracy_philosophy": 0.5,
2479
+ "mmlu_eval_accuracy_prehistory": 0.5428571428571428,
2480
+ "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
2481
+ "mmlu_eval_accuracy_professional_law": 0.36470588235294116,
2482
+ "mmlu_eval_accuracy_professional_medicine": 0.5806451612903226,
2483
+ "mmlu_eval_accuracy_professional_psychology": 0.4782608695652174,
2484
+ "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
2485
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
2486
+ "mmlu_eval_accuracy_sociology": 0.7272727272727273,
2487
+ "mmlu_eval_accuracy_us_foreign_policy": 0.8181818181818182,
2488
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
2489
+ "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
2490
+ "mmlu_loss": 1.240978464600934,
2491
+ "step": 2600
2492
+ },
2493
+ {
2494
+ "epoch": 0.63,
2495
+ "learning_rate": 0.0002,
2496
+ "loss": 0.7782,
2497
+ "step": 2610
2498
+ },
2499
+ {
2500
+ "epoch": 0.63,
2501
+ "learning_rate": 0.0002,
2502
+ "loss": 0.8294,
2503
+ "step": 2620
2504
+ },
2505
+ {
2506
+ "epoch": 0.64,
2507
+ "learning_rate": 0.0002,
2508
+ "loss": 0.7894,
2509
+ "step": 2630
2510
+ },
2511
+ {
2512
+ "epoch": 0.64,
2513
+ "learning_rate": 0.0002,
2514
+ "loss": 0.8196,
2515
+ "step": 2640
2516
+ },
2517
+ {
2518
+ "epoch": 0.64,
2519
+ "learning_rate": 0.0002,
2520
+ "loss": 0.8473,
2521
+ "step": 2650
2522
+ },
2523
+ {
2524
+ "epoch": 0.64,
2525
+ "learning_rate": 0.0002,
2526
+ "loss": 0.7501,
2527
+ "step": 2660
2528
+ },
2529
+ {
2530
+ "epoch": 0.65,
2531
+ "learning_rate": 0.0002,
2532
+ "loss": 0.804,
2533
+ "step": 2670
2534
+ },
2535
+ {
2536
+ "epoch": 0.65,
2537
+ "learning_rate": 0.0002,
2538
+ "loss": 0.8067,
2539
+ "step": 2680
2540
+ },
2541
+ {
2542
+ "epoch": 0.65,
2543
+ "learning_rate": 0.0002,
2544
+ "loss": 0.7634,
2545
+ "step": 2690
2546
+ },
2547
+ {
2548
+ "epoch": 0.65,
2549
+ "learning_rate": 0.0002,
2550
+ "loss": 0.7999,
2551
+ "step": 2700
2552
+ },
2553
+ {
2554
+ "epoch": 0.66,
2555
+ "learning_rate": 0.0002,
2556
+ "loss": 0.7885,
2557
+ "step": 2710
2558
+ },
2559
+ {
2560
+ "epoch": 0.66,
2561
+ "learning_rate": 0.0002,
2562
+ "loss": 0.7948,
2563
+ "step": 2720
2564
+ },
2565
+ {
2566
+ "epoch": 0.66,
2567
+ "learning_rate": 0.0002,
2568
+ "loss": 0.7307,
2569
+ "step": 2730
2570
+ },
2571
+ {
2572
+ "epoch": 0.66,
2573
+ "learning_rate": 0.0002,
2574
+ "loss": 0.7406,
2575
+ "step": 2740
2576
+ },
2577
+ {
2578
+ "epoch": 0.67,
2579
+ "learning_rate": 0.0002,
2580
+ "loss": 0.6913,
2581
+ "step": 2750
2582
+ },
2583
+ {
2584
+ "epoch": 0.67,
2585
+ "learning_rate": 0.0002,
2586
+ "loss": 0.7622,
2587
+ "step": 2760
2588
+ },
2589
+ {
2590
+ "epoch": 0.67,
2591
+ "learning_rate": 0.0002,
2592
+ "loss": 0.7793,
2593
+ "step": 2770
2594
+ },
2595
+ {
2596
+ "epoch": 0.67,
2597
+ "learning_rate": 0.0002,
2598
+ "loss": 0.802,
2599
+ "step": 2780
2600
+ },
2601
+ {
2602
+ "epoch": 0.68,
2603
+ "learning_rate": 0.0002,
2604
+ "loss": 0.7532,
2605
+ "step": 2790
2606
+ },
2607
+ {
2608
+ "epoch": 0.68,
2609
+ "learning_rate": 0.0002,
2610
+ "loss": 0.7541,
2611
+ "step": 2800
2612
+ },
2613
+ {
2614
+ "epoch": 0.68,
2615
+ "eval_loss": 0.7657595872879028,
2616
+ "eval_runtime": 146.8917,
2617
+ "eval_samples_per_second": 6.808,
2618
+ "eval_steps_per_second": 3.404,
2619
+ "step": 2800
2620
+ },
2621
+ {
2622
+ "epoch": 0.68,
2623
+ "mmlu_eval_accuracy": 0.5007086424249709,
2624
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
2625
+ "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
2626
+ "mmlu_eval_accuracy_astronomy": 0.5,
2627
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
2628
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
2629
+ "mmlu_eval_accuracy_college_biology": 0.4375,
2630
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
2631
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
2632
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
2633
+ "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
2634
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
2635
+ "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
2636
+ "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
2637
+ "mmlu_eval_accuracy_econometrics": 0.25,
2638
+ "mmlu_eval_accuracy_electrical_engineering": 0.25,
2639
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
2640
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
2641
+ "mmlu_eval_accuracy_global_facts": 0.6,
2642
+ "mmlu_eval_accuracy_high_school_biology": 0.4375,
2643
+ "mmlu_eval_accuracy_high_school_chemistry": 0.22727272727272727,
2644
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
2645
+ "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
2646
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
2647
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
2648
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
2649
+ "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
2650
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
2651
+ "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
2652
+ "mmlu_eval_accuracy_high_school_psychology": 0.8833333333333333,
2653
+ "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
2654
+ "mmlu_eval_accuracy_high_school_us_history": 0.5909090909090909,
2655
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
2656
+ "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
2657
+ "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
2658
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
2659
+ "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
2660
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
2661
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
2662
+ "mmlu_eval_accuracy_management": 0.5454545454545454,
2663
+ "mmlu_eval_accuracy_marketing": 0.76,
2664
+ "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
2665
+ "mmlu_eval_accuracy_miscellaneous": 0.6395348837209303,
2666
+ "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
2667
+ "mmlu_eval_accuracy_moral_scenarios": 0.25,
2668
+ "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
2669
+ "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
2670
+ "mmlu_eval_accuracy_prehistory": 0.6,
2671
+ "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
2672
+ "mmlu_eval_accuracy_professional_law": 0.3411764705882353,
2673
+ "mmlu_eval_accuracy_professional_medicine": 0.6129032258064516,
2674
+ "mmlu_eval_accuracy_professional_psychology": 0.5217391304347826,
2675
+ "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
2676
+ "mmlu_eval_accuracy_security_studies": 0.4444444444444444,
2677
+ "mmlu_eval_accuracy_sociology": 0.7727272727272727,
2678
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
2679
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
2680
+ "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
2681
+ "mmlu_loss": 1.196944424751222,
2682
+ "step": 2800
2683
+ },
2684
+ {
2685
+ "epoch": 0.68,
2686
+ "learning_rate": 0.0002,
2687
+ "loss": 0.7829,
2688
+ "step": 2810
2689
+ },
2690
+ {
2691
+ "epoch": 0.68,
2692
+ "learning_rate": 0.0002,
2693
+ "loss": 0.7969,
2694
+ "step": 2820
2695
+ },
2696
+ {
2697
+ "epoch": 0.68,
2698
+ "learning_rate": 0.0002,
2699
+ "loss": 0.7507,
2700
+ "step": 2830
2701
+ },
2702
+ {
2703
+ "epoch": 0.69,
2704
+ "learning_rate": 0.0002,
2705
+ "loss": 0.7748,
2706
+ "step": 2840
2707
+ },
2708
+ {
2709
+ "epoch": 0.69,
2710
+ "learning_rate": 0.0002,
2711
+ "loss": 0.7723,
2712
+ "step": 2850
2713
+ },
2714
+ {
2715
+ "epoch": 0.69,
2716
+ "learning_rate": 0.0002,
2717
+ "loss": 0.698,
2718
+ "step": 2860
2719
+ },
2720
+ {
2721
+ "epoch": 0.69,
2722
+ "learning_rate": 0.0002,
2723
+ "loss": 0.7199,
2724
+ "step": 2870
2725
+ },
2726
+ {
2727
+ "epoch": 0.7,
2728
+ "learning_rate": 0.0002,
2729
+ "loss": 0.778,
2730
+ "step": 2880
2731
+ },
2732
+ {
2733
+ "epoch": 0.7,
2734
+ "learning_rate": 0.0002,
2735
+ "loss": 0.7405,
2736
+ "step": 2890
2737
+ },
2738
+ {
2739
+ "epoch": 0.7,
2740
+ "learning_rate": 0.0002,
2741
+ "loss": 0.7006,
2742
+ "step": 2900
2743
+ },
2744
+ {
2745
+ "epoch": 0.7,
2746
+ "learning_rate": 0.0002,
2747
+ "loss": 0.7905,
2748
+ "step": 2910
2749
+ },
2750
+ {
2751
+ "epoch": 0.71,
2752
+ "learning_rate": 0.0002,
2753
+ "loss": 0.7681,
2754
+ "step": 2920
2755
+ },
2756
+ {
2757
+ "epoch": 0.71,
2758
+ "learning_rate": 0.0002,
2759
+ "loss": 0.8055,
2760
+ "step": 2930
2761
+ },
2762
+ {
2763
+ "epoch": 0.71,
2764
+ "learning_rate": 0.0002,
2765
+ "loss": 0.833,
2766
+ "step": 2940
2767
+ },
2768
+ {
2769
+ "epoch": 0.71,
2770
+ "learning_rate": 0.0002,
2771
+ "loss": 0.765,
2772
+ "step": 2950
2773
+ },
2774
+ {
2775
+ "epoch": 0.72,
2776
+ "learning_rate": 0.0002,
2777
+ "loss": 0.7558,
2778
+ "step": 2960
2779
+ },
2780
+ {
2781
+ "epoch": 0.72,
2782
+ "learning_rate": 0.0002,
2783
+ "loss": 0.76,
2784
+ "step": 2970
2785
+ },
2786
+ {
2787
+ "epoch": 0.72,
2788
+ "learning_rate": 0.0002,
2789
+ "loss": 0.772,
2790
+ "step": 2980
2791
+ },
2792
+ {
2793
+ "epoch": 0.72,
2794
+ "learning_rate": 0.0002,
2795
+ "loss": 0.7748,
2796
+ "step": 2990
2797
+ },
2798
+ {
2799
+ "epoch": 0.73,
2800
+ "learning_rate": 0.0002,
2801
+ "loss": 0.8289,
2802
+ "step": 3000
2803
+ },
2804
+ {
2805
+ "epoch": 0.73,
2806
+ "eval_loss": 0.763559103012085,
2807
+ "eval_runtime": 146.8702,
2808
+ "eval_samples_per_second": 6.809,
2809
+ "eval_steps_per_second": 3.404,
2810
+ "step": 3000
2811
+ },
2812
+ {
2813
+ "epoch": 0.73,
2814
+ "mmlu_eval_accuracy": 0.5046616393700263,
2815
+ "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
2816
+ "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
2817
+ "mmlu_eval_accuracy_astronomy": 0.5625,
2818
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
2819
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
2820
+ "mmlu_eval_accuracy_college_biology": 0.4375,
2821
+ "mmlu_eval_accuracy_college_chemistry": 0.125,
2822
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
2823
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
2824
+ "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
2825
+ "mmlu_eval_accuracy_college_physics": 0.5454545454545454,
2826
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
2827
+ "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
2828
+ "mmlu_eval_accuracy_econometrics": 0.25,
2829
+ "mmlu_eval_accuracy_electrical_engineering": 0.3125,
2830
+ "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
2831
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
2832
+ "mmlu_eval_accuracy_global_facts": 0.5,
2833
+ "mmlu_eval_accuracy_high_school_biology": 0.40625,
2834
+ "mmlu_eval_accuracy_high_school_chemistry": 0.22727272727272727,
2835
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
2836
+ "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
2837
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
2838
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
2839
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4883720930232558,
2840
+ "mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
2841
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5384615384615384,
2842
+ "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
2843
+ "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
2844
+ "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
2845
+ "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
2846
+ "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
2847
+ "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
2848
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
2849
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
2850
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
2851
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
2852
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
2853
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
2854
+ "mmlu_eval_accuracy_marketing": 0.84,
2855
+ "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
2856
+ "mmlu_eval_accuracy_miscellaneous": 0.6395348837209303,
2857
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
2858
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
2859
+ "mmlu_eval_accuracy_nutrition": 0.696969696969697,
2860
+ "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
2861
+ "mmlu_eval_accuracy_prehistory": 0.5142857142857142,
2862
+ "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
2863
+ "mmlu_eval_accuracy_professional_law": 0.3235294117647059,
2864
+ "mmlu_eval_accuracy_professional_medicine": 0.6129032258064516,
2865
+ "mmlu_eval_accuracy_professional_psychology": 0.4927536231884058,
2866
+ "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
2867
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
2868
+ "mmlu_eval_accuracy_sociology": 0.7272727272727273,
2869
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
2870
+ "mmlu_eval_accuracy_virology": 0.5,
2871
+ "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
2872
+ "mmlu_loss": 1.2194577517381848,
2873
+ "step": 3000
2874
+ },
2875
+ {
2876
+ "epoch": 0.73,
2877
+ "learning_rate": 0.0002,
2878
+ "loss": 0.8104,
2879
+ "step": 3010
2880
+ },
2881
+ {
2882
+ "epoch": 0.73,
2883
+ "learning_rate": 0.0002,
2884
+ "loss": 0.734,
2885
+ "step": 3020
2886
+ },
2887
+ {
2888
+ "epoch": 0.73,
2889
+ "learning_rate": 0.0002,
2890
+ "loss": 0.7242,
2891
+ "step": 3030
2892
+ },
2893
+ {
2894
+ "epoch": 0.74,
2895
+ "learning_rate": 0.0002,
2896
+ "loss": 0.7488,
2897
+ "step": 3040
2898
+ },
2899
+ {
2900
+ "epoch": 0.74,
2901
+ "learning_rate": 0.0002,
2902
+ "loss": 0.7684,
2903
+ "step": 3050
2904
+ },
2905
+ {
2906
+ "epoch": 0.74,
2907
+ "learning_rate": 0.0002,
2908
+ "loss": 0.7224,
2909
+ "step": 3060
2910
+ },
2911
+ {
2912
+ "epoch": 0.74,
2913
+ "learning_rate": 0.0002,
2914
+ "loss": 0.7356,
2915
+ "step": 3070
2916
+ },
2917
+ {
2918
+ "epoch": 0.75,
2919
+ "learning_rate": 0.0002,
2920
+ "loss": 0.7893,
2921
+ "step": 3080
2922
+ },
2923
+ {
2924
+ "epoch": 0.75,
2925
+ "learning_rate": 0.0002,
2926
+ "loss": 0.7264,
2927
+ "step": 3090
2928
+ },
2929
+ {
2930
+ "epoch": 0.75,
2931
+ "learning_rate": 0.0002,
2932
+ "loss": 0.7764,
2933
+ "step": 3100
2934
+ },
2935
+ {
2936
+ "epoch": 0.75,
2937
+ "learning_rate": 0.0002,
2938
+ "loss": 0.7459,
2939
+ "step": 3110
2940
+ },
2941
+ {
2942
+ "epoch": 0.76,
2943
+ "learning_rate": 0.0002,
2944
+ "loss": 0.8493,
2945
+ "step": 3120
2946
+ },
2947
+ {
2948
+ "epoch": 0.76,
2949
+ "learning_rate": 0.0002,
2950
+ "loss": 0.7896,
2951
+ "step": 3130
2952
+ },
2953
+ {
2954
+ "epoch": 0.76,
2955
+ "learning_rate": 0.0002,
2956
+ "loss": 0.7499,
2957
+ "step": 3140
2958
+ },
2959
+ {
2960
+ "epoch": 0.76,
2961
+ "learning_rate": 0.0002,
2962
+ "loss": 0.783,
2963
+ "step": 3150
2964
+ },
2965
+ {
2966
+ "epoch": 0.76,
2967
+ "learning_rate": 0.0002,
2968
+ "loss": 0.6757,
2969
+ "step": 3160
2970
+ },
2971
+ {
2972
+ "epoch": 0.77,
2973
+ "learning_rate": 0.0002,
2974
+ "loss": 0.8313,
2975
+ "step": 3170
2976
+ },
2977
+ {
2978
+ "epoch": 0.77,
2979
+ "learning_rate": 0.0002,
2980
+ "loss": 0.8387,
2981
+ "step": 3180
2982
+ },
2983
+ {
2984
+ "epoch": 0.77,
2985
+ "learning_rate": 0.0002,
2986
+ "loss": 0.7895,
2987
+ "step": 3190
2988
+ },
2989
+ {
2990
+ "epoch": 0.77,
2991
+ "learning_rate": 0.0002,
2992
+ "loss": 0.7867,
2993
+ "step": 3200
2994
+ },
2995
+ {
2996
+ "epoch": 0.77,
2997
+ "eval_loss": 0.7642560005187988,
2998
+ "eval_runtime": 146.9194,
2999
+ "eval_samples_per_second": 6.806,
3000
+ "eval_steps_per_second": 3.403,
3001
+ "step": 3200
3002
+ },
3003
+ {
3004
+ "epoch": 0.77,
3005
+ "mmlu_eval_accuracy": 0.49573529869125116,
3006
+ "mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
3007
+ "mmlu_eval_accuracy_anatomy": 0.7857142857142857,
3008
+ "mmlu_eval_accuracy_astronomy": 0.375,
3009
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
3010
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
3011
+ "mmlu_eval_accuracy_college_biology": 0.5,
3012
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
3013
+ "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
3014
+ "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
3015
+ "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
3016
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
3017
+ "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
3018
+ "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
3019
+ "mmlu_eval_accuracy_econometrics": 0.25,
3020
+ "mmlu_eval_accuracy_electrical_engineering": 0.1875,
3021
+ "mmlu_eval_accuracy_elementary_mathematics": 0.2926829268292683,
3022
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
3023
+ "mmlu_eval_accuracy_global_facts": 0.5,
3024
+ "mmlu_eval_accuracy_high_school_biology": 0.40625,
3025
+ "mmlu_eval_accuracy_high_school_chemistry": 0.22727272727272727,
3026
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
3027
+ "mmlu_eval_accuracy_high_school_european_history": 0.5,
3028
+ "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
3029
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
3030
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.46511627906976744,
3031
+ "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
3032
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5769230769230769,
3033
+ "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
3034
+ "mmlu_eval_accuracy_high_school_psychology": 0.85,
3035
+ "mmlu_eval_accuracy_high_school_statistics": 0.2608695652173913,
3036
+ "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
3037
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
3038
+ "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
3039
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
3040
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
3041
+ "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
3042
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
3043
+ "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
3044
+ "mmlu_eval_accuracy_management": 0.45454545454545453,
3045
+ "mmlu_eval_accuracy_marketing": 0.76,
3046
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
3047
+ "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
3048
+ "mmlu_eval_accuracy_moral_disputes": 0.5526315789473685,
3049
+ "mmlu_eval_accuracy_moral_scenarios": 0.28,
3050
+ "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
3051
+ "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
3052
+ "mmlu_eval_accuracy_prehistory": 0.6285714285714286,
3053
+ "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
3054
+ "mmlu_eval_accuracy_professional_law": 0.35294117647058826,
3055
+ "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
3056
+ "mmlu_eval_accuracy_professional_psychology": 0.5507246376811594,
3057
+ "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
3058
+ "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
3059
+ "mmlu_eval_accuracy_sociology": 0.7272727272727273,
3060
+ "mmlu_eval_accuracy_us_foreign_policy": 0.9090909090909091,
3061
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
3062
+ "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
3063
+ "mmlu_loss": 1.1026417209958905,
3064
+ "step": 3200
3065
+ },
3066
+ {
3067
+ "epoch": 0.78,
3068
+ "learning_rate": 0.0002,
3069
+ "loss": 0.7601,
3070
+ "step": 3210
3071
+ },
3072
+ {
3073
+ "epoch": 0.78,
3074
+ "learning_rate": 0.0002,
3075
+ "loss": 0.7931,
3076
+ "step": 3220
3077
+ },
3078
+ {
3079
+ "epoch": 0.78,
3080
+ "learning_rate": 0.0002,
3081
+ "loss": 0.6574,
3082
+ "step": 3230
3083
+ },
3084
+ {
3085
+ "epoch": 0.78,
3086
+ "learning_rate": 0.0002,
3087
+ "loss": 0.6949,
3088
+ "step": 3240
3089
+ },
3090
+ {
3091
+ "epoch": 0.79,
3092
+ "learning_rate": 0.0002,
3093
+ "loss": 0.8531,
3094
+ "step": 3250
3095
+ },
3096
+ {
3097
+ "epoch": 0.79,
3098
+ "learning_rate": 0.0002,
3099
+ "loss": 0.8129,
3100
+ "step": 3260
3101
+ },
3102
+ {
3103
+ "epoch": 0.79,
3104
+ "learning_rate": 0.0002,
3105
+ "loss": 0.8115,
3106
+ "step": 3270
3107
+ },
3108
+ {
3109
+ "epoch": 0.79,
3110
+ "learning_rate": 0.0002,
3111
+ "loss": 0.6962,
3112
+ "step": 3280
3113
+ },
3114
+ {
3115
+ "epoch": 0.8,
3116
+ "learning_rate": 0.0002,
3117
+ "loss": 0.7614,
3118
+ "step": 3290
3119
+ },
3120
+ {
3121
+ "epoch": 0.8,
3122
+ "learning_rate": 0.0002,
3123
+ "loss": 0.7047,
3124
+ "step": 3300
3125
+ },
3126
+ {
3127
+ "epoch": 0.8,
3128
+ "learning_rate": 0.0002,
3129
+ "loss": 0.7831,
3130
+ "step": 3310
3131
+ },
3132
+ {
3133
+ "epoch": 0.8,
3134
+ "learning_rate": 0.0002,
3135
+ "loss": 0.8018,
3136
+ "step": 3320
3137
+ },
3138
+ {
3139
+ "epoch": 0.81,
3140
+ "learning_rate": 0.0002,
3141
+ "loss": 0.79,
3142
+ "step": 3330
3143
+ },
3144
+ {
3145
+ "epoch": 0.81,
3146
+ "learning_rate": 0.0002,
3147
+ "loss": 0.7092,
3148
+ "step": 3340
3149
+ },
3150
+ {
3151
+ "epoch": 0.81,
3152
+ "learning_rate": 0.0002,
3153
+ "loss": 0.7116,
3154
+ "step": 3350
3155
+ },
3156
+ {
3157
+ "epoch": 0.81,
3158
+ "learning_rate": 0.0002,
3159
+ "loss": 0.8317,
3160
+ "step": 3360
3161
+ },
3162
+ {
3163
+ "epoch": 0.82,
3164
+ "learning_rate": 0.0002,
3165
+ "loss": 0.7386,
3166
+ "step": 3370
3167
+ },
3168
+ {
3169
+ "epoch": 0.82,
3170
+ "learning_rate": 0.0002,
3171
+ "loss": 0.7151,
3172
+ "step": 3380
3173
+ },
3174
+ {
3175
+ "epoch": 0.82,
3176
+ "learning_rate": 0.0002,
3177
+ "loss": 0.7603,
3178
+ "step": 3390
3179
+ },
3180
+ {
3181
+ "epoch": 0.82,
3182
+ "learning_rate": 0.0002,
3183
+ "loss": 0.7558,
3184
+ "step": 3400
3185
+ },
3186
+ {
3187
+ "epoch": 0.82,
3188
+ "eval_loss": 0.7633880376815796,
3189
+ "eval_runtime": 146.9077,
3190
+ "eval_samples_per_second": 6.807,
3191
+ "eval_steps_per_second": 3.403,
3192
+ "step": 3400
3193
+ },
3194
+ {
3195
+ "epoch": 0.82,
3196
+ "mmlu_eval_accuracy": 0.48873600158220964,
3197
+ "mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
3198
+ "mmlu_eval_accuracy_anatomy": 0.7857142857142857,
3199
+ "mmlu_eval_accuracy_astronomy": 0.375,
3200
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
3201
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5862068965517241,
3202
+ "mmlu_eval_accuracy_college_biology": 0.3125,
3203
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
3204
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
3205
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
3206
+ "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
3207
+ "mmlu_eval_accuracy_college_physics": 0.2727272727272727,
3208
+ "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
3209
+ "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
3210
+ "mmlu_eval_accuracy_econometrics": 0.08333333333333333,
3211
+ "mmlu_eval_accuracy_electrical_engineering": 0.1875,
3212
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
3213
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
3214
+ "mmlu_eval_accuracy_global_facts": 0.6,
3215
+ "mmlu_eval_accuracy_high_school_biology": 0.46875,
3216
+ "mmlu_eval_accuracy_high_school_chemistry": 0.22727272727272727,
3217
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
3218
+ "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
3219
+ "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
3220
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
3221
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.46511627906976744,
3222
+ "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
3223
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
3224
+ "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
3225
+ "mmlu_eval_accuracy_high_school_psychology": 0.85,
3226
+ "mmlu_eval_accuracy_high_school_statistics": 0.21739130434782608,
3227
+ "mmlu_eval_accuracy_high_school_us_history": 0.5909090909090909,
3228
+ "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
3229
+ "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
3230
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
3231
+ "mmlu_eval_accuracy_international_law": 0.9230769230769231,
3232
+ "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
3233
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
3234
+ "mmlu_eval_accuracy_machine_learning": 0.45454545454545453,
3235
+ "mmlu_eval_accuracy_management": 0.45454545454545453,
3236
+ "mmlu_eval_accuracy_marketing": 0.76,
3237
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
3238
+ "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
3239
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
3240
+ "mmlu_eval_accuracy_moral_scenarios": 0.22,
3241
+ "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
3242
+ "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
3243
+ "mmlu_eval_accuracy_prehistory": 0.6285714285714286,
3244
+ "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
3245
+ "mmlu_eval_accuracy_professional_law": 0.34705882352941175,
3246
+ "mmlu_eval_accuracy_professional_medicine": 0.4838709677419355,
3247
+ "mmlu_eval_accuracy_professional_psychology": 0.5217391304347826,
3248
+ "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
3249
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
3250
+ "mmlu_eval_accuracy_sociology": 0.7272727272727273,
3251
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
3252
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
3253
+ "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
3254
+ "mmlu_loss": 1.2204070659670443,
3255
+ "step": 3400
3256
+ },
3257
+ {
3258
+ "epoch": 0.83,
3259
+ "learning_rate": 0.0002,
3260
+ "loss": 0.7962,
3261
+ "step": 3410
3262
+ },
3263
+ {
3264
+ "epoch": 0.83,
3265
+ "learning_rate": 0.0002,
3266
+ "loss": 0.7581,
3267
+ "step": 3420
3268
+ },
3269
+ {
3270
+ "epoch": 0.83,
3271
+ "learning_rate": 0.0002,
3272
+ "loss": 0.7571,
3273
+ "step": 3430
3274
+ },
3275
+ {
3276
+ "epoch": 0.83,
3277
+ "learning_rate": 0.0002,
3278
+ "loss": 0.7761,
3279
+ "step": 3440
3280
+ },
3281
+ {
3282
+ "epoch": 0.83,
3283
+ "learning_rate": 0.0002,
3284
+ "loss": 0.8017,
3285
+ "step": 3450
3286
+ },
3287
+ {
3288
+ "epoch": 0.84,
3289
+ "learning_rate": 0.0002,
3290
+ "loss": 0.7293,
3291
+ "step": 3460
3292
+ },
3293
+ {
3294
+ "epoch": 0.84,
3295
+ "learning_rate": 0.0002,
3296
+ "loss": 0.7898,
3297
+ "step": 3470
3298
+ },
3299
+ {
3300
+ "epoch": 0.84,
3301
+ "learning_rate": 0.0002,
3302
+ "loss": 0.837,
3303
+ "step": 3480
3304
+ },
3305
+ {
3306
+ "epoch": 0.84,
3307
+ "learning_rate": 0.0002,
3308
+ "loss": 0.8751,
3309
+ "step": 3490
3310
+ },
3311
+ {
3312
+ "epoch": 0.85,
3313
+ "learning_rate": 0.0002,
3314
+ "loss": 0.7367,
3315
+ "step": 3500
3316
+ },
3317
+ {
3318
+ "epoch": 0.85,
3319
+ "learning_rate": 0.0002,
3320
+ "loss": 0.739,
3321
+ "step": 3510
3322
+ },
3323
+ {
3324
+ "epoch": 0.85,
3325
+ "learning_rate": 0.0002,
3326
+ "loss": 0.8002,
3327
+ "step": 3520
3328
+ },
3329
+ {
3330
+ "epoch": 0.85,
3331
+ "learning_rate": 0.0002,
3332
+ "loss": 0.8364,
3333
+ "step": 3530
3334
+ },
3335
+ {
3336
+ "epoch": 0.86,
3337
+ "learning_rate": 0.0002,
3338
+ "loss": 0.8192,
3339
+ "step": 3540
3340
+ },
3341
+ {
3342
+ "epoch": 0.86,
3343
+ "learning_rate": 0.0002,
3344
+ "loss": 0.8373,
3345
+ "step": 3550
3346
+ },
3347
+ {
3348
+ "epoch": 0.86,
3349
+ "learning_rate": 0.0002,
3350
+ "loss": 0.8157,
3351
+ "step": 3560
3352
+ },
3353
+ {
3354
+ "epoch": 0.86,
3355
+ "learning_rate": 0.0002,
3356
+ "loss": 0.7234,
3357
+ "step": 3570
3358
+ },
3359
+ {
3360
+ "epoch": 0.87,
3361
+ "learning_rate": 0.0002,
3362
+ "loss": 0.7886,
3363
+ "step": 3580
3364
+ },
3365
+ {
3366
+ "epoch": 0.87,
3367
+ "learning_rate": 0.0002,
3368
+ "loss": 0.7399,
3369
+ "step": 3590
3370
+ },
3371
+ {
3372
+ "epoch": 0.87,
3373
+ "learning_rate": 0.0002,
3374
+ "loss": 0.7283,
3375
+ "step": 3600
3376
+ },
3377
+ {
3378
+ "epoch": 0.87,
3379
+ "eval_loss": 0.7614111304283142,
3380
+ "eval_runtime": 146.7673,
3381
+ "eval_samples_per_second": 6.814,
3382
+ "eval_steps_per_second": 3.407,
3383
+ "step": 3600
3384
+ },
3385
+ {
3386
+ "epoch": 0.87,
3387
+ "mmlu_eval_accuracy": 0.4911723596011134,
3388
+ "mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
3389
+ "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
3390
+ "mmlu_eval_accuracy_astronomy": 0.3125,
3391
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
3392
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5862068965517241,
3393
+ "mmlu_eval_accuracy_college_biology": 0.375,
3394
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
3395
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
3396
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
3397
+ "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
3398
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
3399
+ "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
3400
+ "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
3401
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
3402
+ "mmlu_eval_accuracy_electrical_engineering": 0.25,
3403
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
3404
+ "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
3405
+ "mmlu_eval_accuracy_global_facts": 0.4,
3406
+ "mmlu_eval_accuracy_high_school_biology": 0.46875,
3407
+ "mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
3408
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
3409
+ "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
3410
+ "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
3411
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.47619047619047616,
3412
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.46511627906976744,
3413
+ "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
3414
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
3415
+ "mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
3416
+ "mmlu_eval_accuracy_high_school_psychology": 0.85,
3417
+ "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
3418
+ "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
3419
+ "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
3420
+ "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
3421
+ "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
3422
+ "mmlu_eval_accuracy_international_law": 0.9230769230769231,
3423
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
3424
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
3425
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
3426
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
3427
+ "mmlu_eval_accuracy_marketing": 0.76,
3428
+ "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
3429
+ "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
3430
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
3431
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
3432
+ "mmlu_eval_accuracy_nutrition": 0.696969696969697,
3433
+ "mmlu_eval_accuracy_philosophy": 0.5,
3434
+ "mmlu_eval_accuracy_prehistory": 0.6571428571428571,
3435
+ "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
3436
+ "mmlu_eval_accuracy_professional_law": 0.3176470588235294,
3437
+ "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
3438
+ "mmlu_eval_accuracy_professional_psychology": 0.463768115942029,
3439
+ "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
3440
+ "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
3441
+ "mmlu_eval_accuracy_sociology": 0.7272727272727273,
3442
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
3443
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
3444
+ "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
3445
+ "mmlu_loss": 1.19231010249794,
3446
+ "step": 3600
3447
+ },
3448
+ {
3449
+ "epoch": 0.87,
3450
+ "learning_rate": 0.0002,
3451
+ "loss": 0.7301,
3452
+ "step": 3610
3453
+ },
3454
+ {
3455
+ "epoch": 0.88,
3456
+ "learning_rate": 0.0002,
3457
+ "loss": 0.7471,
3458
+ "step": 3620
3459
+ },
3460
+ {
3461
+ "epoch": 0.88,
3462
+ "learning_rate": 0.0002,
3463
+ "loss": 0.7064,
3464
+ "step": 3630
3465
+ },
3466
+ {
3467
+ "epoch": 0.88,
3468
+ "learning_rate": 0.0002,
3469
+ "loss": 0.8056,
3470
+ "step": 3640
3471
+ },
3472
+ {
3473
+ "epoch": 0.88,
3474
+ "learning_rate": 0.0002,
3475
+ "loss": 0.7804,
3476
+ "step": 3650
3477
+ },
3478
+ {
3479
+ "epoch": 0.89,
3480
+ "learning_rate": 0.0002,
3481
+ "loss": 0.7989,
3482
+ "step": 3660
3483
+ },
3484
+ {
3485
+ "epoch": 0.89,
3486
+ "learning_rate": 0.0002,
3487
+ "loss": 0.8823,
3488
+ "step": 3670
3489
+ },
3490
+ {
3491
+ "epoch": 0.89,
3492
+ "learning_rate": 0.0002,
3493
+ "loss": 0.7778,
3494
+ "step": 3680
3495
+ },
3496
+ {
3497
+ "epoch": 0.89,
3498
+ "learning_rate": 0.0002,
3499
+ "loss": 0.8654,
3500
+ "step": 3690
3501
+ },
3502
+ {
3503
+ "epoch": 0.9,
3504
+ "learning_rate": 0.0002,
3505
+ "loss": 0.8275,
3506
+ "step": 3700
3507
+ },
3508
+ {
3509
+ "epoch": 0.9,
3510
+ "learning_rate": 0.0002,
3511
+ "loss": 0.7796,
3512
+ "step": 3710
3513
+ },
3514
+ {
3515
+ "epoch": 0.9,
3516
+ "learning_rate": 0.0002,
3517
+ "loss": 0.7013,
3518
+ "step": 3720
3519
+ },
3520
+ {
3521
+ "epoch": 0.9,
3522
+ "learning_rate": 0.0002,
3523
+ "loss": 0.829,
3524
+ "step": 3730
3525
+ },
3526
+ {
3527
+ "epoch": 0.91,
3528
+ "learning_rate": 0.0002,
3529
+ "loss": 0.7031,
3530
+ "step": 3740
3531
+ },
3532
+ {
3533
+ "epoch": 0.91,
3534
+ "learning_rate": 0.0002,
3535
+ "loss": 0.7213,
3536
+ "step": 3750
3537
+ },
3538
+ {
3539
+ "epoch": 0.91,
3540
+ "learning_rate": 0.0002,
3541
+ "loss": 0.8002,
3542
+ "step": 3760
3543
+ },
3544
+ {
3545
+ "epoch": 0.91,
3546
+ "learning_rate": 0.0002,
3547
+ "loss": 0.7743,
3548
+ "step": 3770
3549
+ },
3550
+ {
3551
+ "epoch": 0.91,
3552
+ "learning_rate": 0.0002,
3553
+ "loss": 0.7334,
3554
+ "step": 3780
3555
+ },
3556
+ {
3557
+ "epoch": 0.92,
3558
+ "learning_rate": 0.0002,
3559
+ "loss": 0.8116,
3560
+ "step": 3790
3561
+ },
3562
+ {
3563
+ "epoch": 0.92,
3564
+ "learning_rate": 0.0002,
3565
+ "loss": 0.7995,
3566
+ "step": 3800
3567
+ },
3568
+ {
3569
+ "epoch": 0.92,
3570
+ "eval_loss": 0.7618931531906128,
3571
+ "eval_runtime": 146.7645,
3572
+ "eval_samples_per_second": 6.814,
3573
+ "eval_steps_per_second": 3.407,
3574
+ "step": 3800
3575
+ },
3576
+ {
3577
+ "epoch": 0.92,
3578
+ "mmlu_eval_accuracy": 0.4786676171243828,
3579
+ "mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
3580
+ "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
3581
+ "mmlu_eval_accuracy_astronomy": 0.375,
3582
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
3583
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
3584
+ "mmlu_eval_accuracy_college_biology": 0.3125,
3585
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
3586
+ "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
3587
+ "mmlu_eval_accuracy_college_mathematics": 0.09090909090909091,
3588
+ "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
3589
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
3590
+ "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
3591
+ "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
3592
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
3593
+ "mmlu_eval_accuracy_electrical_engineering": 0.1875,
3594
+ "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
3595
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
3596
+ "mmlu_eval_accuracy_global_facts": 0.6,
3597
+ "mmlu_eval_accuracy_high_school_biology": 0.5,
3598
+ "mmlu_eval_accuracy_high_school_chemistry": 0.18181818181818182,
3599
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
3600
+ "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
3601
+ "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
3602
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.38095238095238093,
3603
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
3604
+ "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
3605
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.6153846153846154,
3606
+ "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
3607
+ "mmlu_eval_accuracy_high_school_psychology": 0.85,
3608
+ "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
3609
+ "mmlu_eval_accuracy_high_school_us_history": 0.5909090909090909,
3610
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
3611
+ "mmlu_eval_accuracy_human_aging": 0.5652173913043478,
3612
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
3613
+ "mmlu_eval_accuracy_international_law": 0.9230769230769231,
3614
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
3615
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
3616
+ "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
3617
+ "mmlu_eval_accuracy_management": 0.5454545454545454,
3618
+ "mmlu_eval_accuracy_marketing": 0.72,
3619
+ "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
3620
+ "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
3621
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
3622
+ "mmlu_eval_accuracy_moral_scenarios": 0.23,
3623
+ "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
3624
+ "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
3625
+ "mmlu_eval_accuracy_prehistory": 0.6,
3626
+ "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
3627
+ "mmlu_eval_accuracy_professional_law": 0.3352941176470588,
3628
+ "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
3629
+ "mmlu_eval_accuracy_professional_psychology": 0.4782608695652174,
3630
+ "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
3631
+ "mmlu_eval_accuracy_security_studies": 0.4444444444444444,
3632
+ "mmlu_eval_accuracy_sociology": 0.6363636363636364,
3633
+ "mmlu_eval_accuracy_us_foreign_policy": 0.8181818181818182,
3634
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
3635
+ "mmlu_eval_accuracy_world_religions": 0.631578947368421,
3636
+ "mmlu_loss": 1.2730830661933041,
3637
+ "step": 3800
3638
+ },
3639
+ {
3640
+ "epoch": 0.92,
3641
+ "learning_rate": 0.0002,
3642
+ "loss": 0.7851,
3643
+ "step": 3810
3644
+ },
3645
+ {
3646
+ "epoch": 0.92,
3647
+ "learning_rate": 0.0002,
3648
+ "loss": 0.7385,
3649
+ "step": 3820
3650
+ },
3651
+ {
3652
+ "epoch": 0.93,
3653
+ "learning_rate": 0.0002,
3654
+ "loss": 0.8477,
3655
+ "step": 3830
3656
+ },
3657
+ {
3658
+ "epoch": 0.93,
3659
+ "learning_rate": 0.0002,
3660
+ "loss": 0.8057,
3661
+ "step": 3840
3662
+ },
3663
+ {
3664
+ "epoch": 0.93,
3665
+ "learning_rate": 0.0002,
3666
+ "loss": 0.7247,
3667
+ "step": 3850
3668
+ },
3669
+ {
3670
+ "epoch": 0.93,
3671
+ "learning_rate": 0.0002,
3672
+ "loss": 0.741,
3673
+ "step": 3860
3674
+ },
3675
+ {
3676
+ "epoch": 0.94,
3677
+ "learning_rate": 0.0002,
3678
+ "loss": 0.7132,
3679
+ "step": 3870
3680
+ },
3681
+ {
3682
+ "epoch": 0.94,
3683
+ "learning_rate": 0.0002,
3684
+ "loss": 0.7317,
3685
+ "step": 3880
3686
+ },
3687
+ {
3688
+ "epoch": 0.94,
3689
+ "learning_rate": 0.0002,
3690
+ "loss": 0.8438,
3691
+ "step": 3890
3692
+ },
3693
+ {
3694
+ "epoch": 0.94,
3695
+ "learning_rate": 0.0002,
3696
+ "loss": 0.8944,
3697
+ "step": 3900
3698
+ },
3699
+ {
3700
+ "epoch": 0.95,
3701
+ "learning_rate": 0.0002,
3702
+ "loss": 0.7458,
3703
+ "step": 3910
3704
+ },
3705
+ {
3706
+ "epoch": 0.95,
3707
+ "learning_rate": 0.0002,
3708
+ "loss": 0.7914,
3709
+ "step": 3920
3710
+ },
3711
+ {
3712
+ "epoch": 0.95,
3713
+ "learning_rate": 0.0002,
3714
+ "loss": 0.7907,
3715
+ "step": 3930
3716
+ },
3717
+ {
3718
+ "epoch": 0.95,
3719
+ "learning_rate": 0.0002,
3720
+ "loss": 0.7812,
3721
+ "step": 3940
3722
+ },
3723
+ {
3724
+ "epoch": 0.96,
3725
+ "learning_rate": 0.0002,
3726
+ "loss": 0.7115,
3727
+ "step": 3950
3728
+ },
3729
+ {
3730
+ "epoch": 0.96,
3731
+ "learning_rate": 0.0002,
3732
+ "loss": 0.7997,
3733
+ "step": 3960
3734
+ },
3735
+ {
3736
+ "epoch": 0.96,
3737
+ "learning_rate": 0.0002,
3738
+ "loss": 0.782,
3739
+ "step": 3970
3740
+ },
3741
+ {
3742
+ "epoch": 0.96,
3743
+ "learning_rate": 0.0002,
3744
+ "loss": 0.727,
3745
+ "step": 3980
3746
+ },
3747
+ {
3748
+ "epoch": 0.97,
3749
+ "learning_rate": 0.0002,
3750
+ "loss": 0.702,
3751
+ "step": 3990
3752
+ },
3753
+ {
3754
+ "epoch": 0.97,
3755
+ "learning_rate": 0.0002,
3756
+ "loss": 0.7375,
3757
+ "step": 4000
3758
+ },
3759
+ {
3760
+ "epoch": 0.97,
3761
+ "eval_loss": 0.758108913898468,
3762
+ "eval_runtime": 146.8949,
3763
+ "eval_samples_per_second": 6.808,
3764
+ "eval_steps_per_second": 3.404,
3765
+ "step": 4000
3766
+ },
3767
+ {
3768
+ "epoch": 0.97,
3769
+ "mmlu_eval_accuracy": 0.4858725474885569,
3770
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
3771
+ "mmlu_eval_accuracy_anatomy": 0.7857142857142857,
3772
+ "mmlu_eval_accuracy_astronomy": 0.375,
3773
+ "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
3774
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
3775
+ "mmlu_eval_accuracy_college_biology": 0.375,
3776
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
3777
+ "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
3778
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
3779
+ "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
3780
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
3781
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
3782
+ "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
3783
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
3784
+ "mmlu_eval_accuracy_electrical_engineering": 0.25,
3785
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
3786
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
3787
+ "mmlu_eval_accuracy_global_facts": 0.6,
3788
+ "mmlu_eval_accuracy_high_school_biology": 0.46875,
3789
+ "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
3790
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
3791
+ "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
3792
+ "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
3793
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.42857142857142855,
3794
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
3795
+ "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
3796
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.6153846153846154,
3797
+ "mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
3798
+ "mmlu_eval_accuracy_high_school_psychology": 0.85,
3799
+ "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
3800
+ "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
3801
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
3802
+ "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
3803
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
3804
+ "mmlu_eval_accuracy_international_law": 0.9230769230769231,
3805
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
3806
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
3807
+ "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
3808
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
3809
+ "mmlu_eval_accuracy_marketing": 0.76,
3810
+ "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
3811
+ "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
3812
+ "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
3813
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
3814
+ "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
3815
+ "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
3816
+ "mmlu_eval_accuracy_prehistory": 0.5142857142857142,
3817
+ "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
3818
+ "mmlu_eval_accuracy_professional_law": 0.3352941176470588,
3819
+ "mmlu_eval_accuracy_professional_medicine": 0.4838709677419355,
3820
+ "mmlu_eval_accuracy_professional_psychology": 0.4927536231884058,
3821
+ "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
3822
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
3823
+ "mmlu_eval_accuracy_sociology": 0.6818181818181818,
3824
+ "mmlu_eval_accuracy_us_foreign_policy": 0.9090909090909091,
3825
+ "mmlu_eval_accuracy_virology": 0.4444444444444444,
3826
+ "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
3827
+ "mmlu_loss": 1.2080161583174613,
3828
+ "step": 4000
3829
  }
3830
  ],
3831
  "max_steps": 10000,
3832
  "num_train_epochs": 3,
3833
+ "total_flos": 4.43163637635711e+17,
3834
  "trial_name": null,
3835
  "trial_params": null
3836
  }
{checkpoint-1800 → checkpoint-4000}/training_args.bin RENAMED
File without changes