Hugging Face's logo

HydraLM
/
mistralic-expert-4

Model card Files
xet
 Community
pharaouk commited on
Commit
743b248
·
1 Parent(s): 7cc1912

Training in progress, step 1000, checkpoint

Browse files
Files changed (15) hide show
  1. checkpoint-1000/README.md +21 -0
  2. checkpoint-1000/adapter_config.json +26 -0
  3. checkpoint-1000/adapter_model.bin +3 -0
  4. checkpoint-1000/adapter_model/adapter_model/README.md +21 -0
  5. checkpoint-1000/adapter_model/adapter_model/adapter_config.json +26 -0
  6. checkpoint-1000/adapter_model/adapter_model/adapter_model.bin +3 -0
  7. checkpoint-1000/added_tokens.json +8 -0
  8. checkpoint-1000/optimizer.pt +3 -0
  9. checkpoint-1000/rng_state.pth +3 -0
  10. checkpoint-1000/scheduler.pt +3 -0
  11. checkpoint-1000/special_tokens_map.json +11 -0
  12. checkpoint-1000/tokenizer.model +3 -0
  13. checkpoint-1000/tokenizer_config.json +74 -0
  14. checkpoint-1000/trainer_state.json +974 -0
  15. checkpoint-1000/training_args.bin +3 -0
checkpoint-1000/README.md ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: peft
3
+ ---
4
+ ## Training procedure
5
+
6
+
7
+ The following `bitsandbytes` quantization config was used during training:
8
+ - quant_method: bitsandbytes
9
+ - load_in_8bit: False
10
+ - load_in_4bit: True
11
+ - llm_int8_threshold: 6.0
12
+ - llm_int8_skip_modules: None
13
+ - llm_int8_enable_fp32_cpu_offload: False
14
+ - llm_int8_has_fp16_weight: False
15
+ - bnb_4bit_quant_type: nf4
16
+ - bnb_4bit_use_double_quant: True
17
+ - bnb_4bit_compute_dtype: bfloat16
18
+ ### Framework versions
19
+
20
+
21
+ - PEFT 0.4.0
checkpoint-1000/adapter_config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "auto_mapping": null,
3
+ "base_model_name_or_path": "SkunkworksAI/Mistralic-7B-1",
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
+ "v_proj",
18
+ "k_proj",
19
+ "up_proj",
20
+ "q_proj",
21
+ "gate_proj",
22
+ "o_proj",
23
+ "down_proj"
24
+ ],
25
+ "task_type": "CAUSAL_LM"
26
+ }
checkpoint-1000/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:16baf9eedf3c113a3e09ea4f160b1c45d9b5c575d9211874bfac6803dc7d8861
3
+ size 335706314
checkpoint-1000/adapter_model/adapter_model/README.md ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: peft
3
+ ---
4
+ ## Training procedure
5
+
6
+
7
+ The following `bitsandbytes` quantization config was used during training:
8
+ - quant_method: bitsandbytes
9
+ - load_in_8bit: False
10
+ - load_in_4bit: True
11
+ - llm_int8_threshold: 6.0
12
+ - llm_int8_skip_modules: None
13
+ - llm_int8_enable_fp32_cpu_offload: False
14
+ - llm_int8_has_fp16_weight: False
15
+ - bnb_4bit_quant_type: nf4
16
+ - bnb_4bit_use_double_quant: True
17
+ - bnb_4bit_compute_dtype: bfloat16
18
+ ### Framework versions
19
+
20
+
21
+ - PEFT 0.4.0
checkpoint-1000/adapter_model/adapter_model/adapter_config.json ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "auto_mapping": null,
3
+ "base_model_name_or_path": "SkunkworksAI/Mistralic-7B-1",
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
+ "v_proj",
18
+ "k_proj",
19
+ "up_proj",
20
+ "q_proj",
21
+ "gate_proj",
22
+ "o_proj",
23
+ "down_proj"
24
+ ],
25
+ "task_type": "CAUSAL_LM"
26
+ }
checkpoint-1000/adapter_model/adapter_model/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:16baf9eedf3c113a3e09ea4f160b1c45d9b5c575d9211874bfac6803dc7d8861
3
+ size 335706314
checkpoint-1000/added_tokens.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</s>": 2,
3
+ "<s>": 1,
4
+ "<unk>": 0,
5
+ "<|im_end|>": 32000,
6
+ "<|im_start|>": 32001,
7
+ "[PAD]": 32002
8
+ }
checkpoint-1000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:181c115a43ed563427c93010fb75786b3038153a7d49c1a61a0d2fcb4e549253
3
+ size 1342453434
checkpoint-1000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d9a5f64f979f39a6db8a0c45a9c034bb2173bde293916499ee0373a79d0575a4
3
+ size 14180
checkpoint-1000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:19bcbcfd505eb4fc1a12c7d832a77268d82b4a9b82157e01d48b5fb1cf5fbcc1
3
+ size 1064
checkpoint-1000/special_tokens_map.json ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<unk>",
4
+ "<s>",
5
+ "</s>"
6
+ ],
7
+ "bos_token": "<s>",
8
+ "eos_token": "</s>",
9
+ "pad_token": "[PAD]",
10
+ "unk_token": "<unk>"
11
+ }
checkpoint-1000/tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dadfd56d766715c61d2ef780a525ab43b8e6da4de6865bda3d95fdef5e134055
3
+ size 493443
checkpoint-1000/tokenizer_config.json ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": true,
3
+ "add_eos_token": false,
4
+ "added_tokens_decoder": {
5
+ "0": {
6
+ "content": "<unk>",
7
+ "lstrip": true,
8
+ "normalized": false,
9
+ "rstrip": true,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "1": {
14
+ "content": "<s>",
15
+ "lstrip": true,
16
+ "normalized": false,
17
+ "rstrip": true,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "2": {
22
+ "content": "</s>",
23
+ "lstrip": true,
24
+ "normalized": false,
25
+ "rstrip": true,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "32000": {
30
+ "content": "<|im_end|>",
31
+ "lstrip": true,
32
+ "normalized": true,
33
+ "rstrip": true,
34
+ "single_word": false,
35
+ "special": false
36
+ },
37
+ "32001": {
38
+ "content": "<|im_start|>",
39
+ "lstrip": true,
40
+ "normalized": true,
41
+ "rstrip": true,
42
+ "single_word": false,
43
+ "special": false
44
+ },
45
+ "32002": {
46
+ "content": "[PAD]",
47
+ "lstrip": true,
48
+ "normalized": false,
49
+ "rstrip": true,
50
+ "single_word": false,
51
+ "special": true
52
+ }
53
+ },
54
+ "additional_special_tokens": [
55
+ "<unk>",
56
+ "<s>",
57
+ "</s>"
58
+ ],
59
+ "bos_token": "<s>",
60
+ "clean_up_tokenization_spaces": false,
61
+ "eos_token": "</s>",
62
+ "legacy": true,
63
+ "model_max_length": 1000000000000000019884624838656,
64
+ "pad_token": "[PAD]",
65
+ "padding_side": "right",
66
+ "sp_model_kwargs": {},
67
+ "spaces_between_special_tokens": false,
68
+ "tokenizer_class": "LlamaTokenizer",
69
+ "tokenizer_file": null,
70
+ "trust_remote_code": false,
71
+ "unk_token": "<unk>",
72
+ "use_default_system_prompt": true,
73
+ "use_fast": true
74
+ }
checkpoint-1000/trainer_state.json ADDED
@@ -0,0 +1,974 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.7314385771751404,
3
+ "best_model_checkpoint": "experts/mistralic-expert-4/checkpoint-1000",
4
+ "epoch": 0.22357609971494047,
5
+ "eval_steps": 200,
6
+ "global_step": 1000,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.0,
13
+ "learning_rate": 0.0002,
14
+ "loss": 0.8828,
15
+ "step": 10
16
+ },
17
+ {
18
+ "epoch": 0.0,
19
+ "learning_rate": 0.0002,
20
+ "loss": 0.8642,
21
+ "step": 20
22
+ },
23
+ {
24
+ "epoch": 0.01,
25
+ "learning_rate": 0.0002,
26
+ "loss": 0.8519,
27
+ "step": 30
28
+ },
29
+ {
30
+ "epoch": 0.01,
31
+ "learning_rate": 0.0002,
32
+ "loss": 0.7916,
33
+ "step": 40
34
+ },
35
+ {
36
+ "epoch": 0.01,
37
+ "learning_rate": 0.0002,
38
+ "loss": 0.8655,
39
+ "step": 50
40
+ },
41
+ {
42
+ "epoch": 0.01,
43
+ "learning_rate": 0.0002,
44
+ "loss": 0.8312,
45
+ "step": 60
46
+ },
47
+ {
48
+ "epoch": 0.02,
49
+ "learning_rate": 0.0002,
50
+ "loss": 0.8108,
51
+ "step": 70
52
+ },
53
+ {
54
+ "epoch": 0.02,
55
+ "learning_rate": 0.0002,
56
+ "loss": 0.8403,
57
+ "step": 80
58
+ },
59
+ {
60
+ "epoch": 0.02,
61
+ "learning_rate": 0.0002,
62
+ "loss": 0.8013,
63
+ "step": 90
64
+ },
65
+ {
66
+ "epoch": 0.02,
67
+ "learning_rate": 0.0002,
68
+ "loss": 0.8937,
69
+ "step": 100
70
+ },
71
+ {
72
+ "epoch": 0.02,
73
+ "learning_rate": 0.0002,
74
+ "loss": 0.8408,
75
+ "step": 110
76
+ },
77
+ {
78
+ "epoch": 0.03,
79
+ "learning_rate": 0.0002,
80
+ "loss": 0.8257,
81
+ "step": 120
82
+ },
83
+ {
84
+ "epoch": 0.03,
85
+ "learning_rate": 0.0002,
86
+ "loss": 0.8461,
87
+ "step": 130
88
+ },
89
+ {
90
+ "epoch": 0.03,
91
+ "learning_rate": 0.0002,
92
+ "loss": 0.8367,
93
+ "step": 140
94
+ },
95
+ {
96
+ "epoch": 0.03,
97
+ "learning_rate": 0.0002,
98
+ "loss": 0.7979,
99
+ "step": 150
100
+ },
101
+ {
102
+ "epoch": 0.04,
103
+ "learning_rate": 0.0002,
104
+ "loss": 0.843,
105
+ "step": 160
106
+ },
107
+ {
108
+ "epoch": 0.04,
109
+ "learning_rate": 0.0002,
110
+ "loss": 0.8144,
111
+ "step": 170
112
+ },
113
+ {
114
+ "epoch": 0.04,
115
+ "learning_rate": 0.0002,
116
+ "loss": 0.8484,
117
+ "step": 180
118
+ },
119
+ {
120
+ "epoch": 0.04,
121
+ "learning_rate": 0.0002,
122
+ "loss": 0.799,
123
+ "step": 190
124
+ },
125
+ {
126
+ "epoch": 0.04,
127
+ "learning_rate": 0.0002,
128
+ "loss": 0.8573,
129
+ "step": 200
130
+ },
131
+ {
132
+ "epoch": 0.04,
133
+ "eval_loss": 0.7452409863471985,
134
+ "eval_runtime": 137.1753,
135
+ "eval_samples_per_second": 7.29,
136
+ "eval_steps_per_second": 3.645,
137
+ "step": 200
138
+ },
139
+ {
140
+ "epoch": 0.04,
141
+ "mmlu_eval_accuracy": 0.5996444862522003,
142
+ "mmlu_eval_accuracy_abstract_algebra": 0.18181818181818182,
143
+ "mmlu_eval_accuracy_anatomy": 0.5,
144
+ "mmlu_eval_accuracy_astronomy": 0.8125,
145
+ "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
146
+ "mmlu_eval_accuracy_clinical_knowledge": 0.6206896551724138,
147
+ "mmlu_eval_accuracy_college_biology": 0.6875,
148
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
149
+ "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
150
+ "mmlu_eval_accuracy_college_mathematics": 0.5454545454545454,
151
+ "mmlu_eval_accuracy_college_medicine": 0.6363636363636364,
152
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
153
+ "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
154
+ "mmlu_eval_accuracy_conceptual_physics": 0.5384615384615384,
155
+ "mmlu_eval_accuracy_econometrics": 0.5,
156
+ "mmlu_eval_accuracy_electrical_engineering": 0.5625,
157
+ "mmlu_eval_accuracy_elementary_mathematics": 0.5365853658536586,
158
+ "mmlu_eval_accuracy_formal_logic": 0.07142857142857142,
159
+ "mmlu_eval_accuracy_global_facts": 0.4,
160
+ "mmlu_eval_accuracy_high_school_biology": 0.625,
161
+ "mmlu_eval_accuracy_high_school_chemistry": 0.45454545454545453,
162
+ "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
163
+ "mmlu_eval_accuracy_high_school_european_history": 0.7222222222222222,
164
+ "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
165
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.7619047619047619,
166
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.627906976744186,
167
+ "mmlu_eval_accuracy_high_school_mathematics": 0.41379310344827586,
168
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.6538461538461539,
169
+ "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
170
+ "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
171
+ "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
172
+ "mmlu_eval_accuracy_high_school_us_history": 0.7727272727272727,
173
+ "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
174
+ "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
175
+ "mmlu_eval_accuracy_human_sexuality": 0.5833333333333334,
176
+ "mmlu_eval_accuracy_international_law": 1.0,
177
+ "mmlu_eval_accuracy_jurisprudence": 0.6363636363636364,
178
+ "mmlu_eval_accuracy_logical_fallacies": 0.7222222222222222,
179
+ "mmlu_eval_accuracy_machine_learning": 0.45454545454545453,
180
+ "mmlu_eval_accuracy_management": 0.9090909090909091,
181
+ "mmlu_eval_accuracy_marketing": 0.88,
182
+ "mmlu_eval_accuracy_medical_genetics": 1.0,
183
+ "mmlu_eval_accuracy_miscellaneous": 0.7441860465116279,
184
+ "mmlu_eval_accuracy_moral_disputes": 0.6578947368421053,
185
+ "mmlu_eval_accuracy_moral_scenarios": 0.33,
186
+ "mmlu_eval_accuracy_nutrition": 0.7575757575757576,
187
+ "mmlu_eval_accuracy_philosophy": 0.6764705882352942,
188
+ "mmlu_eval_accuracy_prehistory": 0.6,
189
+ "mmlu_eval_accuracy_professional_accounting": 0.4838709677419355,
190
+ "mmlu_eval_accuracy_professional_law": 0.4,
191
+ "mmlu_eval_accuracy_professional_medicine": 0.6451612903225806,
192
+ "mmlu_eval_accuracy_professional_psychology": 0.5942028985507246,
193
+ "mmlu_eval_accuracy_public_relations": 0.5,
194
+ "mmlu_eval_accuracy_security_studies": 0.7037037037037037,
195
+ "mmlu_eval_accuracy_sociology": 1.0,
196
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
197
+ "mmlu_eval_accuracy_virology": 0.5,
198
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
199
+ "mmlu_loss": 1.1377019190834956,
200
+ "step": 200
201
+ },
202
+ {
203
+ "epoch": 0.05,
204
+ "learning_rate": 0.0002,
205
+ "loss": 0.8106,
206
+ "step": 210
207
+ },
208
+ {
209
+ "epoch": 0.05,
210
+ "learning_rate": 0.0002,
211
+ "loss": 0.789,
212
+ "step": 220
213
+ },
214
+ {
215
+ "epoch": 0.05,
216
+ "learning_rate": 0.0002,
217
+ "loss": 0.8575,
218
+ "step": 230
219
+ },
220
+ {
221
+ "epoch": 0.05,
222
+ "learning_rate": 0.0002,
223
+ "loss": 0.7618,
224
+ "step": 240
225
+ },
226
+ {
227
+ "epoch": 0.06,
228
+ "learning_rate": 0.0002,
229
+ "loss": 0.778,
230
+ "step": 250
231
+ },
232
+ {
233
+ "epoch": 0.06,
234
+ "learning_rate": 0.0002,
235
+ "loss": 0.8824,
236
+ "step": 260
237
+ },
238
+ {
239
+ "epoch": 0.06,
240
+ "learning_rate": 0.0002,
241
+ "loss": 0.8281,
242
+ "step": 270
243
+ },
244
+ {
245
+ "epoch": 0.06,
246
+ "learning_rate": 0.0002,
247
+ "loss": 0.8067,
248
+ "step": 280
249
+ },
250
+ {
251
+ "epoch": 0.06,
252
+ "learning_rate": 0.0002,
253
+ "loss": 0.7467,
254
+ "step": 290
255
+ },
256
+ {
257
+ "epoch": 0.07,
258
+ "learning_rate": 0.0002,
259
+ "loss": 0.8204,
260
+ "step": 300
261
+ },
262
+ {
263
+ "epoch": 0.07,
264
+ "learning_rate": 0.0002,
265
+ "loss": 0.7581,
266
+ "step": 310
267
+ },
268
+ {
269
+ "epoch": 0.07,
270
+ "learning_rate": 0.0002,
271
+ "loss": 0.7942,
272
+ "step": 320
273
+ },
274
+ {
275
+ "epoch": 0.07,
276
+ "learning_rate": 0.0002,
277
+ "loss": 0.8976,
278
+ "step": 330
279
+ },
280
+ {
281
+ "epoch": 0.08,
282
+ "learning_rate": 0.0002,
283
+ "loss": 0.8094,
284
+ "step": 340
285
+ },
286
+ {
287
+ "epoch": 0.08,
288
+ "learning_rate": 0.0002,
289
+ "loss": 0.8201,
290
+ "step": 350
291
+ },
292
+ {
293
+ "epoch": 0.08,
294
+ "learning_rate": 0.0002,
295
+ "loss": 0.8922,
296
+ "step": 360
297
+ },
298
+ {
299
+ "epoch": 0.08,
300
+ "learning_rate": 0.0002,
301
+ "loss": 0.7782,
302
+ "step": 370
303
+ },
304
+ {
305
+ "epoch": 0.08,
306
+ "learning_rate": 0.0002,
307
+ "loss": 0.6944,
308
+ "step": 380
309
+ },
310
+ {
311
+ "epoch": 0.09,
312
+ "learning_rate": 0.0002,
313
+ "loss": 0.8458,
314
+ "step": 390
315
+ },
316
+ {
317
+ "epoch": 0.09,
318
+ "learning_rate": 0.0002,
319
+ "loss": 0.7986,
320
+ "step": 400
321
+ },
322
+ {
323
+ "epoch": 0.09,
324
+ "eval_loss": 0.7476009130477905,
325
+ "eval_runtime": 136.7729,
326
+ "eval_samples_per_second": 7.311,
327
+ "eval_steps_per_second": 3.656,
328
+ "step": 400
329
+ },
330
+ {
331
+ "epoch": 0.09,
332
+ "mmlu_eval_accuracy": 0.5776334992003169,
333
+ "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
334
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
335
+ "mmlu_eval_accuracy_astronomy": 0.5625,
336
+ "mmlu_eval_accuracy_business_ethics": 0.7272727272727273,
337
+ "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
338
+ "mmlu_eval_accuracy_college_biology": 0.625,
339
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
340
+ "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
341
+ "mmlu_eval_accuracy_college_mathematics": 0.5454545454545454,
342
+ "mmlu_eval_accuracy_college_medicine": 0.6363636363636364,
343
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
344
+ "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
345
+ "mmlu_eval_accuracy_conceptual_physics": 0.5769230769230769,
346
+ "mmlu_eval_accuracy_econometrics": 0.4166666666666667,
347
+ "mmlu_eval_accuracy_electrical_engineering": 0.5,
348
+ "mmlu_eval_accuracy_elementary_mathematics": 0.4146341463414634,
349
+ "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
350
+ "mmlu_eval_accuracy_global_facts": 0.3,
351
+ "mmlu_eval_accuracy_high_school_biology": 0.46875,
352
+ "mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
353
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
354
+ "mmlu_eval_accuracy_high_school_european_history": 0.8333333333333334,
355
+ "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
356
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.7142857142857143,
357
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.5813953488372093,
358
+ "mmlu_eval_accuracy_high_school_mathematics": 0.3448275862068966,
359
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.6923076923076923,
360
+ "mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
361
+ "mmlu_eval_accuracy_high_school_psychology": 0.8166666666666667,
362
+ "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
363
+ "mmlu_eval_accuracy_high_school_us_history": 0.8181818181818182,
364
+ "mmlu_eval_accuracy_high_school_world_history": 0.6153846153846154,
365
+ "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
366
+ "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
367
+ "mmlu_eval_accuracy_international_law": 0.9230769230769231,
368
+ "mmlu_eval_accuracy_jurisprudence": 0.6363636363636364,
369
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
370
+ "mmlu_eval_accuracy_machine_learning": 0.45454545454545453,
371
+ "mmlu_eval_accuracy_management": 0.8181818181818182,
372
+ "mmlu_eval_accuracy_marketing": 0.92,
373
+ "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
374
+ "mmlu_eval_accuracy_miscellaneous": 0.7558139534883721,
375
+ "mmlu_eval_accuracy_moral_disputes": 0.5526315789473685,
376
+ "mmlu_eval_accuracy_moral_scenarios": 0.27,
377
+ "mmlu_eval_accuracy_nutrition": 0.696969696969697,
378
+ "mmlu_eval_accuracy_philosophy": 0.7647058823529411,
379
+ "mmlu_eval_accuracy_prehistory": 0.5428571428571428,
380
+ "mmlu_eval_accuracy_professional_accounting": 0.5483870967741935,
381
+ "mmlu_eval_accuracy_professional_law": 0.36470588235294116,
382
+ "mmlu_eval_accuracy_professional_medicine": 0.6129032258064516,
383
+ "mmlu_eval_accuracy_professional_psychology": 0.6231884057971014,
384
+ "mmlu_eval_accuracy_public_relations": 0.5,
385
+ "mmlu_eval_accuracy_security_studies": 0.7037037037037037,
386
+ "mmlu_eval_accuracy_sociology": 0.8636363636363636,
387
+ "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
388
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
389
+ "mmlu_eval_accuracy_world_religions": 0.8947368421052632,
390
+ "mmlu_loss": 1.1887153384346252,
391
+ "step": 400
392
+ },
393
+ {
394
+ "epoch": 0.09,
395
+ "learning_rate": 0.0002,
396
+ "loss": 0.7745,
397
+ "step": 410
398
+ },
399
+ {
400
+ "epoch": 0.09,
401
+ "learning_rate": 0.0002,
402
+ "loss": 0.784,
403
+ "step": 420
404
+ },
405
+ {
406
+ "epoch": 0.1,
407
+ "learning_rate": 0.0002,
408
+ "loss": 0.7849,
409
+ "step": 430
410
+ },
411
+ {
412
+ "epoch": 0.1,
413
+ "learning_rate": 0.0002,
414
+ "loss": 0.7558,
415
+ "step": 440
416
+ },
417
+ {
418
+ "epoch": 0.1,
419
+ "learning_rate": 0.0002,
420
+ "loss": 0.9095,
421
+ "step": 450
422
+ },
423
+ {
424
+ "epoch": 0.1,
425
+ "learning_rate": 0.0002,
426
+ "loss": 0.7935,
427
+ "step": 460
428
+ },
429
+ {
430
+ "epoch": 0.11,
431
+ "learning_rate": 0.0002,
432
+ "loss": 0.7563,
433
+ "step": 470
434
+ },
435
+ {
436
+ "epoch": 0.11,
437
+ "learning_rate": 0.0002,
438
+ "loss": 0.7224,
439
+ "step": 480
440
+ },
441
+ {
442
+ "epoch": 0.11,
443
+ "learning_rate": 0.0002,
444
+ "loss": 0.7191,
445
+ "step": 490
446
+ },
447
+ {
448
+ "epoch": 0.11,
449
+ "learning_rate": 0.0002,
450
+ "loss": 0.7099,
451
+ "step": 500
452
+ },
453
+ {
454
+ "epoch": 0.11,
455
+ "learning_rate": 0.0002,
456
+ "loss": 0.794,
457
+ "step": 510
458
+ },
459
+ {
460
+ "epoch": 0.12,
461
+ "learning_rate": 0.0002,
462
+ "loss": 0.8334,
463
+ "step": 520
464
+ },
465
+ {
466
+ "epoch": 0.12,
467
+ "learning_rate": 0.0002,
468
+ "loss": 0.7737,
469
+ "step": 530
470
+ },
471
+ {
472
+ "epoch": 0.12,
473
+ "learning_rate": 0.0002,
474
+ "loss": 0.7638,
475
+ "step": 540
476
+ },
477
+ {
478
+ "epoch": 0.12,
479
+ "learning_rate": 0.0002,
480
+ "loss": 0.7722,
481
+ "step": 550
482
+ },
483
+ {
484
+ "epoch": 0.13,
485
+ "learning_rate": 0.0002,
486
+ "loss": 0.7403,
487
+ "step": 560
488
+ },
489
+ {
490
+ "epoch": 0.13,
491
+ "learning_rate": 0.0002,
492
+ "loss": 0.9063,
493
+ "step": 570
494
+ },
495
+ {
496
+ "epoch": 0.13,
497
+ "learning_rate": 0.0002,
498
+ "loss": 0.7987,
499
+ "step": 580
500
+ },
501
+ {
502
+ "epoch": 0.13,
503
+ "learning_rate": 0.0002,
504
+ "loss": 0.8311,
505
+ "step": 590
506
+ },
507
+ {
508
+ "epoch": 0.13,
509
+ "learning_rate": 0.0002,
510
+ "loss": 0.8578,
511
+ "step": 600
512
+ },
513
+ {
514
+ "epoch": 0.13,
515
+ "eval_loss": 0.7366986274719238,
516
+ "eval_runtime": 137.0058,
517
+ "eval_samples_per_second": 7.299,
518
+ "eval_steps_per_second": 3.649,
519
+ "step": 600
520
+ },
521
+ {
522
+ "epoch": 0.13,
523
+ "mmlu_eval_accuracy": 0.5829466763905795,
524
+ "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
525
+ "mmlu_eval_accuracy_anatomy": 0.42857142857142855,
526
+ "mmlu_eval_accuracy_astronomy": 0.75,
527
+ "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
528
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
529
+ "mmlu_eval_accuracy_college_biology": 0.625,
530
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
531
+ "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
532
+ "mmlu_eval_accuracy_college_mathematics": 0.5454545454545454,
533
+ "mmlu_eval_accuracy_college_medicine": 0.5909090909090909,
534
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
535
+ "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
536
+ "mmlu_eval_accuracy_conceptual_physics": 0.5,
537
+ "mmlu_eval_accuracy_econometrics": 0.5,
538
+ "mmlu_eval_accuracy_electrical_engineering": 0.5,
539
+ "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
540
+ "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
541
+ "mmlu_eval_accuracy_global_facts": 0.4,
542
+ "mmlu_eval_accuracy_high_school_biology": 0.5625,
543
+ "mmlu_eval_accuracy_high_school_chemistry": 0.5,
544
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
545
+ "mmlu_eval_accuracy_high_school_european_history": 0.7777777777777778,
546
+ "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
547
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
548
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.6046511627906976,
549
+ "mmlu_eval_accuracy_high_school_mathematics": 0.4482758620689655,
550
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.6923076923076923,
551
+ "mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
552
+ "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
553
+ "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
554
+ "mmlu_eval_accuracy_high_school_us_history": 0.7727272727272727,
555
+ "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
556
+ "mmlu_eval_accuracy_human_aging": 0.782608695652174,
557
+ "mmlu_eval_accuracy_human_sexuality": 0.5833333333333334,
558
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
559
+ "mmlu_eval_accuracy_jurisprudence": 0.6363636363636364,
560
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
561
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
562
+ "mmlu_eval_accuracy_management": 0.9090909090909091,
563
+ "mmlu_eval_accuracy_marketing": 0.88,
564
+ "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
565
+ "mmlu_eval_accuracy_miscellaneous": 0.7209302325581395,
566
+ "mmlu_eval_accuracy_moral_disputes": 0.5526315789473685,
567
+ "mmlu_eval_accuracy_moral_scenarios": 0.25,
568
+ "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
569
+ "mmlu_eval_accuracy_philosophy": 0.7058823529411765,
570
+ "mmlu_eval_accuracy_prehistory": 0.6285714285714286,
571
+ "mmlu_eval_accuracy_professional_accounting": 0.6129032258064516,
572
+ "mmlu_eval_accuracy_professional_law": 0.3764705882352941,
573
+ "mmlu_eval_accuracy_professional_medicine": 0.5806451612903226,
574
+ "mmlu_eval_accuracy_professional_psychology": 0.6086956521739131,
575
+ "mmlu_eval_accuracy_public_relations": 0.5,
576
+ "mmlu_eval_accuracy_security_studies": 0.6666666666666666,
577
+ "mmlu_eval_accuracy_sociology": 0.8636363636363636,
578
+ "mmlu_eval_accuracy_us_foreign_policy": 0.8181818181818182,
579
+ "mmlu_eval_accuracy_virology": 0.5,
580
+ "mmlu_eval_accuracy_world_religions": 0.8947368421052632,
581
+ "mmlu_loss": 1.030824544944122,
582
+ "step": 600
583
+ },
584
+ {
585
+ "epoch": 0.14,
586
+ "learning_rate": 0.0002,
587
+ "loss": 0.7365,
588
+ "step": 610
589
+ },
590
+ {
591
+ "epoch": 0.14,
592
+ "learning_rate": 0.0002,
593
+ "loss": 0.7479,
594
+ "step": 620
595
+ },
596
+ {
597
+ "epoch": 0.14,
598
+ "learning_rate": 0.0002,
599
+ "loss": 0.8401,
600
+ "step": 630
601
+ },
602
+ {
603
+ "epoch": 0.14,
604
+ "learning_rate": 0.0002,
605
+ "loss": 0.755,
606
+ "step": 640
607
+ },
608
+ {
609
+ "epoch": 0.15,
610
+ "learning_rate": 0.0002,
611
+ "loss": 0.8119,
612
+ "step": 650
613
+ },
614
+ {
615
+ "epoch": 0.15,
616
+ "learning_rate": 0.0002,
617
+ "loss": 0.7592,
618
+ "step": 660
619
+ },
620
+ {
621
+ "epoch": 0.15,
622
+ "learning_rate": 0.0002,
623
+ "loss": 0.7654,
624
+ "step": 670
625
+ },
626
+ {
627
+ "epoch": 0.15,
628
+ "learning_rate": 0.0002,
629
+ "loss": 0.8567,
630
+ "step": 680
631
+ },
632
+ {
633
+ "epoch": 0.15,
634
+ "learning_rate": 0.0002,
635
+ "loss": 0.869,
636
+ "step": 690
637
+ },
638
+ {
639
+ "epoch": 0.16,
640
+ "learning_rate": 0.0002,
641
+ "loss": 0.849,
642
+ "step": 700
643
+ },
644
+ {
645
+ "epoch": 0.16,
646
+ "learning_rate": 0.0002,
647
+ "loss": 0.8629,
648
+ "step": 710
649
+ },
650
+ {
651
+ "epoch": 0.16,
652
+ "learning_rate": 0.0002,
653
+ "loss": 0.8072,
654
+ "step": 720
655
+ },
656
+ {
657
+ "epoch": 0.16,
658
+ "learning_rate": 0.0002,
659
+ "loss": 0.7881,
660
+ "step": 730
661
+ },
662
+ {
663
+ "epoch": 0.17,
664
+ "learning_rate": 0.0002,
665
+ "loss": 0.8063,
666
+ "step": 740
667
+ },
668
+ {
669
+ "epoch": 0.17,
670
+ "learning_rate": 0.0002,
671
+ "loss": 0.723,
672
+ "step": 750
673
+ },
674
+ {
675
+ "epoch": 0.17,
676
+ "learning_rate": 0.0002,
677
+ "loss": 0.7324,
678
+ "step": 760
679
+ },
680
+ {
681
+ "epoch": 0.17,
682
+ "learning_rate": 0.0002,
683
+ "loss": 0.7442,
684
+ "step": 770
685
+ },
686
+ {
687
+ "epoch": 0.17,
688
+ "learning_rate": 0.0002,
689
+ "loss": 0.7676,
690
+ "step": 780
691
+ },
692
+ {
693
+ "epoch": 0.18,
694
+ "learning_rate": 0.0002,
695
+ "loss": 0.8126,
696
+ "step": 790
697
+ },
698
+ {
699
+ "epoch": 0.18,
700
+ "learning_rate": 0.0002,
701
+ "loss": 0.7805,
702
+ "step": 800
703
+ },
704
+ {
705
+ "epoch": 0.18,
706
+ "eval_loss": 0.7362542152404785,
707
+ "eval_runtime": 137.0606,
708
+ "eval_samples_per_second": 7.296,
709
+ "eval_steps_per_second": 3.648,
710
+ "step": 800
711
+ },
712
+ {
713
+ "epoch": 0.18,
714
+ "mmlu_eval_accuracy": 0.5876508722590474,
715
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
716
+ "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
717
+ "mmlu_eval_accuracy_astronomy": 0.625,
718
+ "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
719
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
720
+ "mmlu_eval_accuracy_college_biology": 0.6875,
721
+ "mmlu_eval_accuracy_college_chemistry": 0.375,
722
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
723
+ "mmlu_eval_accuracy_college_mathematics": 0.5454545454545454,
724
+ "mmlu_eval_accuracy_college_medicine": 0.5909090909090909,
725
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
726
+ "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
727
+ "mmlu_eval_accuracy_conceptual_physics": 0.5769230769230769,
728
+ "mmlu_eval_accuracy_econometrics": 0.5833333333333334,
729
+ "mmlu_eval_accuracy_electrical_engineering": 0.5,
730
+ "mmlu_eval_accuracy_elementary_mathematics": 0.4146341463414634,
731
+ "mmlu_eval_accuracy_formal_logic": 0.42857142857142855,
732
+ "mmlu_eval_accuracy_global_facts": 0.4,
733
+ "mmlu_eval_accuracy_high_school_biology": 0.59375,
734
+ "mmlu_eval_accuracy_high_school_chemistry": 0.45454545454545453,
735
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
736
+ "mmlu_eval_accuracy_high_school_european_history": 0.7777777777777778,
737
+ "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
738
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.7142857142857143,
739
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.6046511627906976,
740
+ "mmlu_eval_accuracy_high_school_mathematics": 0.3103448275862069,
741
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.6153846153846154,
742
+ "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
743
+ "mmlu_eval_accuracy_high_school_psychology": 0.7833333333333333,
744
+ "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
745
+ "mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
746
+ "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
747
+ "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
748
+ "mmlu_eval_accuracy_human_sexuality": 0.6666666666666666,
749
+ "mmlu_eval_accuracy_international_law": 0.8461538461538461,
750
+ "mmlu_eval_accuracy_jurisprudence": 0.5454545454545454,
751
+ "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
752
+ "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
753
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
754
+ "mmlu_eval_accuracy_marketing": 0.88,
755
+ "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
756
+ "mmlu_eval_accuracy_miscellaneous": 0.7093023255813954,
757
+ "mmlu_eval_accuracy_moral_disputes": 0.6052631578947368,
758
+ "mmlu_eval_accuracy_moral_scenarios": 0.28,
759
+ "mmlu_eval_accuracy_nutrition": 0.7272727272727273,
760
+ "mmlu_eval_accuracy_philosophy": 0.5588235294117647,
761
+ "mmlu_eval_accuracy_prehistory": 0.5714285714285714,
762
+ "mmlu_eval_accuracy_professional_accounting": 0.5161290322580645,
763
+ "mmlu_eval_accuracy_professional_law": 0.37058823529411766,
764
+ "mmlu_eval_accuracy_professional_medicine": 0.6451612903225806,
765
+ "mmlu_eval_accuracy_professional_psychology": 0.6231884057971014,
766
+ "mmlu_eval_accuracy_public_relations": 0.3333333333333333,
767
+ "mmlu_eval_accuracy_security_studies": 0.7777777777777778,
768
+ "mmlu_eval_accuracy_sociology": 0.8181818181818182,
769
+ "mmlu_eval_accuracy_us_foreign_policy": 0.8181818181818182,
770
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
771
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
772
+ "mmlu_loss": 0.9463959685855061,
773
+ "step": 800
774
+ },
775
+ {
776
+ "epoch": 0.18,
777
+ "learning_rate": 0.0002,
778
+ "loss": 0.8164,
779
+ "step": 810
780
+ },
781
+ {
782
+ "epoch": 0.18,
783
+ "learning_rate": 0.0002,
784
+ "loss": 0.7922,
785
+ "step": 820
786
+ },
787
+ {
788
+ "epoch": 0.19,
789
+ "learning_rate": 0.0002,
790
+ "loss": 0.7332,
791
+ "step": 830
792
+ },
793
+ {
794
+ "epoch": 0.19,
795
+ "learning_rate": 0.0002,
796
+ "loss": 0.8391,
797
+ "step": 840
798
+ },
799
+ {
800
+ "epoch": 0.19,
801
+ "learning_rate": 0.0002,
802
+ "loss": 0.8028,
803
+ "step": 850
804
+ },
805
+ {
806
+ "epoch": 0.19,
807
+ "learning_rate": 0.0002,
808
+ "loss": 0.8506,
809
+ "step": 860
810
+ },
811
+ {
812
+ "epoch": 0.19,
813
+ "learning_rate": 0.0002,
814
+ "loss": 0.8054,
815
+ "step": 870
816
+ },
817
+ {
818
+ "epoch": 0.2,
819
+ "learning_rate": 0.0002,
820
+ "loss": 0.795,
821
+ "step": 880
822
+ },
823
+ {
824
+ "epoch": 0.2,
825
+ "learning_rate": 0.0002,
826
+ "loss": 0.8364,
827
+ "step": 890
828
+ },
829
+ {
830
+ "epoch": 0.2,
831
+ "learning_rate": 0.0002,
832
+ "loss": 0.8145,
833
+ "step": 900
834
+ },
835
+ {
836
+ "epoch": 0.2,
837
+ "learning_rate": 0.0002,
838
+ "loss": 0.8151,
839
+ "step": 910
840
+ },
841
+ {
842
+ "epoch": 0.21,
843
+ "learning_rate": 0.0002,
844
+ "loss": 0.7902,
845
+ "step": 920
846
+ },
847
+ {
848
+ "epoch": 0.21,
849
+ "learning_rate": 0.0002,
850
+ "loss": 0.8121,
851
+ "step": 930
852
+ },
853
+ {
854
+ "epoch": 0.21,
855
+ "learning_rate": 0.0002,
856
+ "loss": 0.6945,
857
+ "step": 940
858
+ },
859
+ {
860
+ "epoch": 0.21,
861
+ "learning_rate": 0.0002,
862
+ "loss": 0.7703,
863
+ "step": 950
864
+ },
865
+ {
866
+ "epoch": 0.21,
867
+ "learning_rate": 0.0002,
868
+ "loss": 0.797,
869
+ "step": 960
870
+ },
871
+ {
872
+ "epoch": 0.22,
873
+ "learning_rate": 0.0002,
874
+ "loss": 0.8324,
875
+ "step": 970
876
+ },
877
+ {
878
+ "epoch": 0.22,
879
+ "learning_rate": 0.0002,
880
+ "loss": 0.8072,
881
+ "step": 980
882
+ },
883
+ {
884
+ "epoch": 0.22,
885
+ "learning_rate": 0.0002,
886
+ "loss": 0.7871,
887
+ "step": 990
888
+ },
889
+ {
890
+ "epoch": 0.22,
891
+ "learning_rate": 0.0002,
892
+ "loss": 0.7199,
893
+ "step": 1000
894
+ },
895
+ {
896
+ "epoch": 0.22,
897
+ "eval_loss": 0.7314385771751404,
898
+ "eval_runtime": 137.2049,
899
+ "eval_samples_per_second": 7.288,
900
+ "eval_steps_per_second": 3.644,
901
+ "step": 1000
902
+ },
903
+ {
904
+ "epoch": 0.22,
905
+ "mmlu_eval_accuracy": 0.5727217823861395,
906
+ "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
907
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
908
+ "mmlu_eval_accuracy_astronomy": 0.625,
909
+ "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
910
+ "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
911
+ "mmlu_eval_accuracy_college_biology": 0.625,
912
+ "mmlu_eval_accuracy_college_chemistry": 0.5,
913
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
914
+ "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
915
+ "mmlu_eval_accuracy_college_medicine": 0.5,
916
+ "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
917
+ "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
918
+ "mmlu_eval_accuracy_conceptual_physics": 0.5,
919
+ "mmlu_eval_accuracy_econometrics": 0.5833333333333334,
920
+ "mmlu_eval_accuracy_electrical_engineering": 0.6875,
921
+ "mmlu_eval_accuracy_elementary_mathematics": 0.4634146341463415,
922
+ "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
923
+ "mmlu_eval_accuracy_global_facts": 0.3,
924
+ "mmlu_eval_accuracy_high_school_biology": 0.5625,
925
+ "mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
926
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
927
+ "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
928
+ "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
929
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
930
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.4883720930232558,
931
+ "mmlu_eval_accuracy_high_school_mathematics": 0.3793103448275862,
932
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
933
+ "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
934
+ "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
935
+ "mmlu_eval_accuracy_high_school_statistics": 0.5217391304347826,
936
+ "mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
937
+ "mmlu_eval_accuracy_high_school_world_history": 0.6153846153846154,
938
+ "mmlu_eval_accuracy_human_aging": 0.8260869565217391,
939
+ "mmlu_eval_accuracy_human_sexuality": 0.6666666666666666,
940
+ "mmlu_eval_accuracy_international_law": 0.7692307692307693,
941
+ "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
942
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
943
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
944
+ "mmlu_eval_accuracy_management": 0.8181818181818182,
945
+ "mmlu_eval_accuracy_marketing": 0.92,
946
+ "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
947
+ "mmlu_eval_accuracy_miscellaneous": 0.7093023255813954,
948
+ "mmlu_eval_accuracy_moral_disputes": 0.6052631578947368,
949
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
950
+ "mmlu_eval_accuracy_nutrition": 0.7575757575757576,
951
+ "mmlu_eval_accuracy_philosophy": 0.5882352941176471,
952
+ "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
953
+ "mmlu_eval_accuracy_professional_accounting": 0.5483870967741935,
954
+ "mmlu_eval_accuracy_professional_law": 0.3764705882352941,
955
+ "mmlu_eval_accuracy_professional_medicine": 0.6129032258064516,
956
+ "mmlu_eval_accuracy_professional_psychology": 0.6521739130434783,
957
+ "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
958
+ "mmlu_eval_accuracy_security_studies": 0.7407407407407407,
959
+ "mmlu_eval_accuracy_sociology": 0.9090909090909091,
960
+ "mmlu_eval_accuracy_us_foreign_policy": 0.8181818181818182,
961
+ "mmlu_eval_accuracy_virology": 0.6111111111111112,
962
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
963
+ "mmlu_loss": 1.1422579408703524,
964
+ "step": 1000
965
+ }
966
+ ],
967
+ "logging_steps": 10,
968
+ "max_steps": 13416,
969
+ "num_train_epochs": 3,
970
+ "save_steps": 200,
971
+ "total_flos": 4.963223739354317e+17,
972
+ "trial_name": null,
973
+ "trial_params": null
974
+ }
checkpoint-1000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6dd1696fde4e81c3df56a3128258d984d341b0dc475300123cd2379e16a096da
3
+ size 6456