nvan15 commited on
Commit
f96f8e0
·
verified ·
1 Parent(s): 5d64278

Batch upload part 13

Browse files
Files changed (50) hide show
  1. nl_tasks/exprep/run_ex24_3ep/ft/adapter_config.json +19 -0
  2. nl_tasks/exprep/run_ex24_3ep/ft/tokenizer.json +0 -0
  3. nl_tasks/exprep/run_ex24_3ep/ft/tokenizer.model +3 -0
  4. nl_tasks/exprep/run_ex24_3ep/ft2/adapter_config.json +19 -0
  5. nl_tasks/exprep/run_ex24_3ep/ft2/adapter_model.bin +3 -0
  6. nl_tasks/exprep/run_ex25_3ep/ft/adapter_config.json +19 -0
  7. nl_tasks/exprep/run_ex25_3ep/ft/special_tokens_map.json +24 -0
  8. nl_tasks/exprep/run_ex25_3ep/ft/tokenizer.json +0 -0
  9. nl_tasks/exprep/run_ex25_3ep/ft/tokenizer.model +3 -0
  10. nl_tasks/exprep/run_ex25_3ep/ft/tokenizer_config.json +43 -0
  11. nl_tasks/exprep/run_ex25_3ep/ft2/adapter_config.json +19 -0
  12. nl_tasks/exprep/run_ex25_3ep/ft2/adapter_model.bin +3 -0
  13. nl_tasks/exprep/run_ex25_3ep/output.txt +4 -0
  14. nl_tasks/exprep/run_ex25_3ep/trainer_state.json +1093 -0
  15. nl_tasks/exprep/run_ex26_2ep/ft/adapter_config.json +19 -0
  16. nl_tasks/exprep/run_ex26_2ep/ft/special_tokens_map.json +24 -0
  17. nl_tasks/exprep/run_ex26_2ep/ft/tokenizer.json +0 -0
  18. nl_tasks/exprep/run_ex26_2ep/ft/tokenizer.model +3 -0
  19. nl_tasks/exprep/run_ex26_2ep/ft/tokenizer_config.json +43 -0
  20. nl_tasks/exprep/run_ex26_2ep/ft2/adapter_config.json +19 -0
  21. nl_tasks/exprep/run_ex26_2ep/ft2/adapter_model.bin +3 -0
  22. nl_tasks/exprep/run_ex26_2ep/output.txt +4 -0
  23. nl_tasks/exprep/run_ex26_2ep/trainer_state.json +743 -0
  24. nl_tasks/exprep/run_ex27_2ep/ft/adapter_config.json +19 -0
  25. nl_tasks/exprep/run_ex27_2ep/ft/special_tokens_map.json +24 -0
  26. nl_tasks/exprep/run_ex27_2ep/ft/tokenizer.json +0 -0
  27. nl_tasks/exprep/run_ex27_2ep/ft/tokenizer.model +3 -0
  28. nl_tasks/exprep/run_ex27_2ep/ft/tokenizer_config.json +43 -0
  29. nl_tasks/exprep/run_ex27_2ep/ft2/adapter_config.json +19 -0
  30. nl_tasks/exprep/run_ex27_2ep/ft2/adapter_model.bin +3 -0
  31. nl_tasks/exprep/run_ex27_2ep/output.txt +4 -0
  32. nl_tasks/exprep/run_ex27_2ep/trainer_state.json +743 -0
  33. nl_tasks/exprep/run_ex28_2ep/ft/adapter_config.json +19 -0
  34. nl_tasks/exprep/run_ex28_2ep/ft/special_tokens_map.json +24 -0
  35. nl_tasks/exprep/run_ex28_2ep/ft/tokenizer.json +0 -0
  36. nl_tasks/exprep/run_ex28_2ep/ft/tokenizer.model +3 -0
  37. nl_tasks/exprep/run_ex28_2ep/ft/tokenizer_config.json +43 -0
  38. nl_tasks/exprep/run_ex28_2ep/ft2/adapter_config.json +19 -0
  39. nl_tasks/exprep/run_ex28_2ep/ft2/adapter_model.bin +3 -0
  40. nl_tasks/exprep/run_ex28_2ep/output.txt +4 -0
  41. nl_tasks/exprep/run_ex28_2ep/trainer_state.json +743 -0
  42. nl_tasks/exprep/run_ex29_2ep/ft/adapter_config.json +19 -0
  43. nl_tasks/exprep/run_ex29_2ep/ft/special_tokens_map.json +24 -0
  44. nl_tasks/exprep/run_ex29_2ep/ft/tokenizer.json +0 -0
  45. nl_tasks/exprep/run_ex29_2ep/ft/tokenizer.model +3 -0
  46. nl_tasks/exprep/run_ex29_2ep/ft/tokenizer_config.json +43 -0
  47. nl_tasks/exprep/run_ex29_2ep/ft2/adapter_config.json +19 -0
  48. nl_tasks/exprep/run_ex29_2ep/ft2/adapter_model.bin +3 -0
  49. nl_tasks/exprep/run_ex29_2ep/output.txt +4 -0
  50. nl_tasks/exprep/run_ex29_2ep/trainer_state.json +743 -0
nl_tasks/exprep/run_ex24_3ep/ft/adapter_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "drop_out": 0.0,
6
+ "inference_mode": false,
7
+ "layers_to_transform": null,
8
+ "modules_to_save": null,
9
+ "num_rotations": 1,
10
+ "peft_type": "ROTATION",
11
+ "r": 16,
12
+ "revision": null,
13
+ "target_modules": [
14
+ "q_proj",
15
+ "v_proj"
16
+ ],
17
+ "target_modules_to_skip": null,
18
+ "task_type": "CAUSAL_LM"
19
+ }
nl_tasks/exprep/run_ex24_3ep/ft/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
nl_tasks/exprep/run_ex24_3ep/ft/tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
3
+ size 499723
nl_tasks/exprep/run_ex24_3ep/ft2/adapter_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "drop_out": 0.0,
6
+ "inference_mode": true,
7
+ "layers_to_transform": null,
8
+ "modules_to_save": null,
9
+ "num_rotations": 1,
10
+ "peft_type": "ROTATION",
11
+ "r": 16,
12
+ "revision": null,
13
+ "target_modules": [
14
+ "q_proj",
15
+ "v_proj"
16
+ ],
17
+ "target_modules_to_skip": null,
18
+ "task_type": "CAUSAL_LM"
19
+ }
nl_tasks/exprep/run_ex24_3ep/ft2/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9653b8956d0725861333c1ead44ef4a25bd1e87d3abd768f7553cbf90c6d904e
3
+ size 33602915
nl_tasks/exprep/run_ex25_3ep/ft/adapter_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "drop_out": 0.0,
6
+ "inference_mode": false,
7
+ "layers_to_transform": null,
8
+ "modules_to_save": null,
9
+ "num_rotations": 1,
10
+ "peft_type": "ROTATION",
11
+ "r": 16,
12
+ "revision": null,
13
+ "target_modules": [
14
+ "q_proj",
15
+ "v_proj"
16
+ ],
17
+ "target_modules_to_skip": null,
18
+ "task_type": "CAUSAL_LM"
19
+ }
nl_tasks/exprep/run_ex25_3ep/ft/special_tokens_map.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "</s>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": "<unk>",
17
+ "unk_token": {
18
+ "content": "<unk>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ }
24
+ }
nl_tasks/exprep/run_ex25_3ep/ft/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
nl_tasks/exprep/run_ex25_3ep/ft/tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
3
+ size 499723
nl_tasks/exprep/run_ex25_3ep/ft/tokenizer_config.json ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": true,
3
+ "add_eos_token": false,
4
+ "add_prefix_space": null,
5
+ "added_tokens_decoder": {
6
+ "0": {
7
+ "content": "<unk>",
8
+ "lstrip": false,
9
+ "normalized": false,
10
+ "rstrip": false,
11
+ "single_word": false,
12
+ "special": true
13
+ },
14
+ "1": {
15
+ "content": "<s>",
16
+ "lstrip": false,
17
+ "normalized": false,
18
+ "rstrip": false,
19
+ "single_word": false,
20
+ "special": true
21
+ },
22
+ "2": {
23
+ "content": "</s>",
24
+ "lstrip": false,
25
+ "normalized": false,
26
+ "rstrip": false,
27
+ "single_word": false,
28
+ "special": true
29
+ }
30
+ },
31
+ "bos_token": "<s>",
32
+ "clean_up_tokenization_spaces": false,
33
+ "eos_token": "</s>",
34
+ "extra_special_tokens": {},
35
+ "legacy": false,
36
+ "model_max_length": 512,
37
+ "pad_token": "<unk>",
38
+ "padding_side": "right",
39
+ "sp_model_kwargs": {},
40
+ "tokenizer_class": "LlamaTokenizer",
41
+ "unk_token": "<unk>",
42
+ "use_default_system_prompt": false
43
+ }
nl_tasks/exprep/run_ex25_3ep/ft2/adapter_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "drop_out": 0.0,
6
+ "inference_mode": true,
7
+ "layers_to_transform": null,
8
+ "modules_to_save": null,
9
+ "num_rotations": 1,
10
+ "peft_type": "ROTATION",
11
+ "r": 16,
12
+ "revision": null,
13
+ "target_modules": [
14
+ "q_proj",
15
+ "v_proj"
16
+ ],
17
+ "target_modules_to_skip": null,
18
+ "task_type": "CAUSAL_LM"
19
+ }
nl_tasks/exprep/run_ex25_3ep/ft2/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf7991b5a3b77ff990b840b32e902d11bb53e1c3daaa95f1e14d3cc64a7dba83
3
+ size 33602915
nl_tasks/exprep/run_ex25_3ep/output.txt ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+
2
+ MATH math MAX TOKEN = 1408, length==== 5000, math acc %====, 7.739999999999999
3
+
4
+ gsm8k MAX TOKEN = 1024, length==== 1319, gsm8k acc %====, 49.58301743745261
nl_tasks/exprep/run_ex25_3ep/trainer_state.json ADDED
@@ -0,0 +1,1093 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 3.0,
6
+ "eval_steps": 100,
7
+ "global_step": 3750,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.02,
14
+ "grad_norm": 0.21462175250053406,
15
+ "learning_rate": 0.0009998989386555814,
16
+ "loss": 0.4539,
17
+ "step": 25
18
+ },
19
+ {
20
+ "epoch": 0.04,
21
+ "grad_norm": 0.20701375603675842,
22
+ "learning_rate": 0.0009995787805744778,
23
+ "loss": 0.3404,
24
+ "step": 50
25
+ },
26
+ {
27
+ "epoch": 0.06,
28
+ "grad_norm": 0.21636246144771576,
29
+ "learning_rate": 0.000999039490728981,
30
+ "loss": 0.3344,
31
+ "step": 75
32
+ },
33
+ {
34
+ "epoch": 0.08,
35
+ "grad_norm": 0.20059522986412048,
36
+ "learning_rate": 0.000998281305669441,
37
+ "loss": 0.3145,
38
+ "step": 100
39
+ },
40
+ {
41
+ "epoch": 0.1,
42
+ "grad_norm": 0.2221987545490265,
43
+ "learning_rate": 0.0009973045579608833,
44
+ "loss": 0.3076,
45
+ "step": 125
46
+ },
47
+ {
48
+ "epoch": 0.12,
49
+ "grad_norm": 0.1883535087108612,
50
+ "learning_rate": 0.0009961096760371347,
51
+ "loss": 0.3013,
52
+ "step": 150
53
+ },
54
+ {
55
+ "epoch": 0.14,
56
+ "grad_norm": 0.1912730187177658,
57
+ "learning_rate": 0.0009946971840128981,
58
+ "loss": 0.2981,
59
+ "step": 175
60
+ },
61
+ {
62
+ "epoch": 0.16,
63
+ "grad_norm": 0.2401607632637024,
64
+ "learning_rate": 0.0009930677014538588,
65
+ "loss": 0.3,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.18,
70
+ "grad_norm": 0.21345016360282898,
71
+ "learning_rate": 0.0009912219431049217,
72
+ "loss": 0.2952,
73
+ "step": 225
74
+ },
75
+ {
76
+ "epoch": 0.2,
77
+ "grad_norm": 0.22069482505321503,
78
+ "learning_rate": 0.0009891607185767018,
79
+ "loss": 0.2909,
80
+ "step": 250
81
+ },
82
+ {
83
+ "epoch": 0.22,
84
+ "grad_norm": 0.24048374593257904,
85
+ "learning_rate": 0.0009868849319904012,
86
+ "loss": 0.2903,
87
+ "step": 275
88
+ },
89
+ {
90
+ "epoch": 0.24,
91
+ "grad_norm": 0.2725104093551636,
92
+ "learning_rate": 0.000984395581581232,
93
+ "loss": 0.3004,
94
+ "step": 300
95
+ },
96
+ {
97
+ "epoch": 0.26,
98
+ "grad_norm": 0.18401336669921875,
99
+ "learning_rate": 0.000981693759260558,
100
+ "loss": 0.2924,
101
+ "step": 325
102
+ },
103
+ {
104
+ "epoch": 0.28,
105
+ "grad_norm": 0.24094586074352264,
106
+ "learning_rate": 0.0009787806501369446,
107
+ "loss": 0.2939,
108
+ "step": 350
109
+ },
110
+ {
111
+ "epoch": 0.3,
112
+ "grad_norm": 0.26356571912765503,
113
+ "learning_rate": 0.0009756575319963324,
114
+ "loss": 0.287,
115
+ "step": 375
116
+ },
117
+ {
118
+ "epoch": 0.32,
119
+ "grad_norm": 0.22905994951725006,
120
+ "learning_rate": 0.0009723257747415584,
121
+ "loss": 0.2858,
122
+ "step": 400
123
+ },
124
+ {
125
+ "epoch": 0.34,
126
+ "grad_norm": 0.19996008276939392,
127
+ "learning_rate": 0.00096878683979147,
128
+ "loss": 0.2765,
129
+ "step": 425
130
+ },
131
+ {
132
+ "epoch": 0.36,
133
+ "grad_norm": 0.2158837467432022,
134
+ "learning_rate": 0.000965042279439899,
135
+ "loss": 0.2771,
136
+ "step": 450
137
+ },
138
+ {
139
+ "epoch": 0.38,
140
+ "grad_norm": 0.20435525476932526,
141
+ "learning_rate": 0.0009610937361747747,
142
+ "loss": 0.2737,
143
+ "step": 475
144
+ },
145
+ {
146
+ "epoch": 0.4,
147
+ "grad_norm": 0.21901778876781464,
148
+ "learning_rate": 0.0009569429419576737,
149
+ "loss": 0.2721,
150
+ "step": 500
151
+ },
152
+ {
153
+ "epoch": 0.42,
154
+ "grad_norm": 0.18634380400180817,
155
+ "learning_rate": 0.0009525917174641245,
156
+ "loss": 0.2731,
157
+ "step": 525
158
+ },
159
+ {
160
+ "epoch": 0.44,
161
+ "grad_norm": 0.30839431285858154,
162
+ "learning_rate": 0.0009480419712849994,
163
+ "loss": 0.2611,
164
+ "step": 550
165
+ },
166
+ {
167
+ "epoch": 0.46,
168
+ "grad_norm": 0.2272380292415619,
169
+ "learning_rate": 0.0009432956990893433,
170
+ "loss": 0.2792,
171
+ "step": 575
172
+ },
173
+ {
174
+ "epoch": 0.48,
175
+ "grad_norm": 0.18616899847984314,
176
+ "learning_rate": 0.0009383549827490066,
177
+ "loss": 0.2653,
178
+ "step": 600
179
+ },
180
+ {
181
+ "epoch": 0.5,
182
+ "grad_norm": 0.20831988751888275,
183
+ "learning_rate": 0.0009332219894254686,
184
+ "loss": 0.278,
185
+ "step": 625
186
+ },
187
+ {
188
+ "epoch": 0.52,
189
+ "grad_norm": 0.17307333648204803,
190
+ "learning_rate": 0.0009278989706192479,
191
+ "loss": 0.2677,
192
+ "step": 650
193
+ },
194
+ {
195
+ "epoch": 0.54,
196
+ "grad_norm": 0.1601818948984146,
197
+ "learning_rate": 0.0009223882611823205,
198
+ "loss": 0.2578,
199
+ "step": 675
200
+ },
201
+ {
202
+ "epoch": 0.56,
203
+ "grad_norm": 0.21388870477676392,
204
+ "learning_rate": 0.0009166922782939757,
205
+ "loss": 0.2584,
206
+ "step": 700
207
+ },
208
+ {
209
+ "epoch": 0.58,
210
+ "grad_norm": 0.2094329297542572,
211
+ "learning_rate": 0.0009108135204005628,
212
+ "loss": 0.2647,
213
+ "step": 725
214
+ },
215
+ {
216
+ "epoch": 0.6,
217
+ "grad_norm": 0.17878609895706177,
218
+ "learning_rate": 0.0009047545661195884,
219
+ "loss": 0.261,
220
+ "step": 750
221
+ },
222
+ {
223
+ "epoch": 0.62,
224
+ "grad_norm": 0.1805742383003235,
225
+ "learning_rate": 0.0008985180731086505,
226
+ "loss": 0.2504,
227
+ "step": 775
228
+ },
229
+ {
230
+ "epoch": 0.64,
231
+ "grad_norm": 0.19105853140354156,
232
+ "learning_rate": 0.0008921067768997017,
233
+ "loss": 0.255,
234
+ "step": 800
235
+ },
236
+ {
237
+ "epoch": 0.66,
238
+ "grad_norm": 0.23210591077804565,
239
+ "learning_rate": 0.0008855234896991544,
240
+ "loss": 0.2513,
241
+ "step": 825
242
+ },
243
+ {
244
+ "epoch": 0.68,
245
+ "grad_norm": 0.20442335307598114,
246
+ "learning_rate": 0.0008787710991543547,
247
+ "loss": 0.251,
248
+ "step": 850
249
+ },
250
+ {
251
+ "epoch": 0.7,
252
+ "grad_norm": 0.16994322836399078,
253
+ "learning_rate": 0.0008718525670869639,
254
+ "loss": 0.2572,
255
+ "step": 875
256
+ },
257
+ {
258
+ "epoch": 0.72,
259
+ "grad_norm": 0.17936541140079498,
260
+ "learning_rate": 0.0008647709281938065,
261
+ "loss": 0.2541,
262
+ "step": 900
263
+ },
264
+ {
265
+ "epoch": 0.74,
266
+ "grad_norm": 0.188994899392128,
267
+ "learning_rate": 0.0008575292887157515,
268
+ "loss": 0.2495,
269
+ "step": 925
270
+ },
271
+ {
272
+ "epoch": 0.76,
273
+ "grad_norm": 0.17709816992282867,
274
+ "learning_rate": 0.0008501308250752123,
275
+ "loss": 0.2477,
276
+ "step": 950
277
+ },
278
+ {
279
+ "epoch": 0.78,
280
+ "grad_norm": 0.16799673438072205,
281
+ "learning_rate": 0.0008425787824828631,
282
+ "loss": 0.2436,
283
+ "step": 975
284
+ },
285
+ {
286
+ "epoch": 0.8,
287
+ "grad_norm": 0.23775896430015564,
288
+ "learning_rate": 0.0008348764735141823,
289
+ "loss": 0.2497,
290
+ "step": 1000
291
+ },
292
+ {
293
+ "epoch": 0.82,
294
+ "grad_norm": 0.21398940682411194,
295
+ "learning_rate": 0.0008270272766564472,
296
+ "loss": 0.2482,
297
+ "step": 1025
298
+ },
299
+ {
300
+ "epoch": 0.84,
301
+ "grad_norm": 0.17115604877471924,
302
+ "learning_rate": 0.000819034634826818,
303
+ "loss": 0.2394,
304
+ "step": 1050
305
+ },
306
+ {
307
+ "epoch": 0.86,
308
+ "grad_norm": 0.17867575585842133,
309
+ "learning_rate": 0.0008109020538621606,
310
+ "loss": 0.2357,
311
+ "step": 1075
312
+ },
313
+ {
314
+ "epoch": 0.88,
315
+ "grad_norm": 0.16491059958934784,
316
+ "learning_rate": 0.0008026331009812703,
317
+ "loss": 0.2355,
318
+ "step": 1100
319
+ },
320
+ {
321
+ "epoch": 0.9,
322
+ "grad_norm": 0.18323355913162231,
323
+ "learning_rate": 0.0007942314032201719,
324
+ "loss": 0.245,
325
+ "step": 1125
326
+ },
327
+ {
328
+ "epoch": 0.92,
329
+ "grad_norm": 0.1559755951166153,
330
+ "learning_rate": 0.0007857006458411826,
331
+ "loss": 0.2382,
332
+ "step": 1150
333
+ },
334
+ {
335
+ "epoch": 0.94,
336
+ "grad_norm": 0.1760474443435669,
337
+ "learning_rate": 0.0007770445707164325,
338
+ "loss": 0.2413,
339
+ "step": 1175
340
+ },
341
+ {
342
+ "epoch": 0.96,
343
+ "grad_norm": 0.14930203557014465,
344
+ "learning_rate": 0.0007682669746865577,
345
+ "loss": 0.2403,
346
+ "step": 1200
347
+ },
348
+ {
349
+ "epoch": 0.98,
350
+ "grad_norm": 0.1802491694688797,
351
+ "learning_rate": 0.0007593717078952787,
352
+ "loss": 0.2427,
353
+ "step": 1225
354
+ },
355
+ {
356
+ "epoch": 1.0,
357
+ "grad_norm": 0.17276862263679504,
358
+ "learning_rate": 0.0007503626721006019,
359
+ "loss": 0.2329,
360
+ "step": 1250
361
+ },
362
+ {
363
+ "epoch": 1.02,
364
+ "grad_norm": 0.17234313488006592,
365
+ "learning_rate": 0.0007412438189633781,
366
+ "loss": 0.2015,
367
+ "step": 1275
368
+ },
369
+ {
370
+ "epoch": 1.04,
371
+ "grad_norm": 0.13246309757232666,
372
+ "learning_rate": 0.0007320191483139742,
373
+ "loss": 0.2084,
374
+ "step": 1300
375
+ },
376
+ {
377
+ "epoch": 1.06,
378
+ "grad_norm": 0.1591002494096756,
379
+ "learning_rate": 0.0007226927063978153,
380
+ "loss": 0.2052,
381
+ "step": 1325
382
+ },
383
+ {
384
+ "epoch": 1.08,
385
+ "grad_norm": 0.1653255671262741,
386
+ "learning_rate": 0.0007132685841005674,
387
+ "loss": 0.2109,
388
+ "step": 1350
389
+ },
390
+ {
391
+ "epoch": 1.1,
392
+ "grad_norm": 0.1305050551891327,
393
+ "learning_rate": 0.0007037509151537404,
394
+ "loss": 0.2127,
395
+ "step": 1375
396
+ },
397
+ {
398
+ "epoch": 1.12,
399
+ "grad_norm": 0.15579962730407715,
400
+ "learning_rate": 0.0006941438743214963,
401
+ "loss": 0.2118,
402
+ "step": 1400
403
+ },
404
+ {
405
+ "epoch": 1.1400000000000001,
406
+ "grad_norm": 0.15080277621746063,
407
+ "learning_rate": 0.0006844516755694598,
408
+ "loss": 0.2035,
409
+ "step": 1425
410
+ },
411
+ {
412
+ "epoch": 1.16,
413
+ "grad_norm": 0.15345105528831482,
414
+ "learning_rate": 0.0006746785702163335,
415
+ "loss": 0.2102,
416
+ "step": 1450
417
+ },
418
+ {
419
+ "epoch": 1.18,
420
+ "grad_norm": 0.14529818296432495,
421
+ "learning_rate": 0.0006648288450691298,
422
+ "loss": 0.209,
423
+ "step": 1475
424
+ },
425
+ {
426
+ "epoch": 1.2,
427
+ "grad_norm": 0.18043482303619385,
428
+ "learning_rate": 0.0006549068205428343,
429
+ "loss": 0.2098,
430
+ "step": 1500
431
+ },
432
+ {
433
+ "epoch": 1.22,
434
+ "grad_norm": 0.15822406113147736,
435
+ "learning_rate": 0.0006449168487653305,
436
+ "loss": 0.2124,
437
+ "step": 1525
438
+ },
439
+ {
440
+ "epoch": 1.24,
441
+ "grad_norm": 0.1320612132549286,
442
+ "learning_rate": 0.0006348633116684117,
443
+ "loss": 0.204,
444
+ "step": 1550
445
+ },
446
+ {
447
+ "epoch": 1.26,
448
+ "grad_norm": 0.18705196678638458,
449
+ "learning_rate": 0.0006247506190657209,
450
+ "loss": 0.2043,
451
+ "step": 1575
452
+ },
453
+ {
454
+ "epoch": 1.28,
455
+ "grad_norm": 0.15128564834594727,
456
+ "learning_rate": 0.0006145832067184614,
457
+ "loss": 0.2004,
458
+ "step": 1600
459
+ },
460
+ {
461
+ "epoch": 1.3,
462
+ "grad_norm": 0.15656492114067078,
463
+ "learning_rate": 0.0006043655343897249,
464
+ "loss": 0.2012,
465
+ "step": 1625
466
+ },
467
+ {
468
+ "epoch": 1.32,
469
+ "grad_norm": 0.16500142216682434,
470
+ "learning_rate": 0.0005941020838882917,
471
+ "loss": 0.1951,
472
+ "step": 1650
473
+ },
474
+ {
475
+ "epoch": 1.34,
476
+ "grad_norm": 0.1325126588344574,
477
+ "learning_rate": 0.000583797357102762,
478
+ "loss": 0.1998,
479
+ "step": 1675
480
+ },
481
+ {
482
+ "epoch": 1.3599999999999999,
483
+ "grad_norm": 0.13856101036071777,
484
+ "learning_rate": 0.0005734558740268789,
485
+ "loss": 0.1982,
486
+ "step": 1700
487
+ },
488
+ {
489
+ "epoch": 1.38,
490
+ "grad_norm": 0.15071479976177216,
491
+ "learning_rate": 0.000563082170776908,
492
+ "loss": 0.1986,
493
+ "step": 1725
494
+ },
495
+ {
496
+ "epoch": 1.4,
497
+ "grad_norm": 0.15130354464054108,
498
+ "learning_rate": 0.0005526807976019493,
499
+ "loss": 0.199,
500
+ "step": 1750
501
+ },
502
+ {
503
+ "epoch": 1.42,
504
+ "grad_norm": 0.13519452512264252,
505
+ "learning_rate": 0.0005422563168880455,
506
+ "loss": 0.2007,
507
+ "step": 1775
508
+ },
509
+ {
510
+ "epoch": 1.44,
511
+ "grad_norm": 0.15829263627529144,
512
+ "learning_rate": 0.0005318133011569704,
513
+ "loss": 0.1887,
514
+ "step": 1800
515
+ },
516
+ {
517
+ "epoch": 1.46,
518
+ "grad_norm": 0.14695768058300018,
519
+ "learning_rate": 0.0005213563310605686,
520
+ "loss": 0.1945,
521
+ "step": 1825
522
+ },
523
+ {
524
+ "epoch": 1.48,
525
+ "grad_norm": 0.17777042090892792,
526
+ "learning_rate": 0.00051088999337153,
527
+ "loss": 0.1985,
528
+ "step": 1850
529
+ },
530
+ {
531
+ "epoch": 1.5,
532
+ "grad_norm": 0.14207887649536133,
533
+ "learning_rate": 0.0005004188789714811,
534
+ "loss": 0.1996,
535
+ "step": 1875
536
+ },
537
+ {
538
+ "epoch": 1.52,
539
+ "grad_norm": 0.15234456956386566,
540
+ "learning_rate": 0.0004899475808372714,
541
+ "loss": 0.1973,
542
+ "step": 1900
543
+ },
544
+ {
545
+ "epoch": 1.54,
546
+ "grad_norm": 0.1501147598028183,
547
+ "learning_rate": 0.0004794806920263417,
548
+ "loss": 0.1967,
549
+ "step": 1925
550
+ },
551
+ {
552
+ "epoch": 1.56,
553
+ "grad_norm": 0.13002285361289978,
554
+ "learning_rate": 0.0004690228036620589,
555
+ "loss": 0.1865,
556
+ "step": 1950
557
+ },
558
+ {
559
+ "epoch": 1.58,
560
+ "grad_norm": 0.1612955778837204,
561
+ "learning_rate": 0.0004585785029198959,
562
+ "loss": 0.1989,
563
+ "step": 1975
564
+ },
565
+ {
566
+ "epoch": 1.6,
567
+ "grad_norm": 0.14512036740779877,
568
+ "learning_rate": 0.00044815237101534535,
569
+ "loss": 0.1991,
570
+ "step": 2000
571
+ },
572
+ {
573
+ "epoch": 1.62,
574
+ "grad_norm": 0.12977199256420135,
575
+ "learning_rate": 0.0004377489811944478,
576
+ "loss": 0.185,
577
+ "step": 2025
578
+ },
579
+ {
580
+ "epoch": 1.6400000000000001,
581
+ "grad_norm": 0.13996095955371857,
582
+ "learning_rate": 0.00042737289672781367,
583
+ "loss": 0.1939,
584
+ "step": 2050
585
+ },
586
+ {
587
+ "epoch": 1.6600000000000001,
588
+ "grad_norm": 0.126466304063797,
589
+ "learning_rate": 0.0004170286689090228,
590
+ "loss": 0.1981,
591
+ "step": 2075
592
+ },
593
+ {
594
+ "epoch": 1.6800000000000002,
595
+ "grad_norm": 0.15216952562332153,
596
+ "learning_rate": 0.0004067208350582768,
597
+ "loss": 0.1956,
598
+ "step": 2100
599
+ },
600
+ {
601
+ "epoch": 1.7,
602
+ "grad_norm": 0.13788118958473206,
603
+ "learning_rate": 0.0003964539165321794,
604
+ "loss": 0.1937,
605
+ "step": 2125
606
+ },
607
+ {
608
+ "epoch": 1.72,
609
+ "grad_norm": 0.1397712081670761,
610
+ "learning_rate": 0.00038623241674052113,
611
+ "loss": 0.1862,
612
+ "step": 2150
613
+ },
614
+ {
615
+ "epoch": 1.74,
616
+ "grad_norm": 0.13192042708396912,
617
+ "learning_rate": 0.00037606081917093416,
618
+ "loss": 0.2014,
619
+ "step": 2175
620
+ },
621
+ {
622
+ "epoch": 1.76,
623
+ "grad_norm": 0.14153067767620087,
624
+ "learning_rate": 0.0003659435854222869,
625
+ "loss": 0.1925,
626
+ "step": 2200
627
+ },
628
+ {
629
+ "epoch": 1.78,
630
+ "grad_norm": 0.1448894590139389,
631
+ "learning_rate": 0.0003558851532476796,
632
+ "loss": 0.1861,
633
+ "step": 2225
634
+ },
635
+ {
636
+ "epoch": 1.8,
637
+ "grad_norm": 0.1368015706539154,
638
+ "learning_rate": 0.0003458899346078979,
639
+ "loss": 0.195,
640
+ "step": 2250
641
+ },
642
+ {
643
+ "epoch": 1.8199999999999998,
644
+ "grad_norm": 0.13369007408618927,
645
+ "learning_rate": 0.00033596231373618247,
646
+ "loss": 0.1869,
647
+ "step": 2275
648
+ },
649
+ {
650
+ "epoch": 1.8399999999999999,
651
+ "grad_norm": 0.14294853806495667,
652
+ "learning_rate": 0.0003261066452151587,
653
+ "loss": 0.1899,
654
+ "step": 2300
655
+ },
656
+ {
657
+ "epoch": 1.8599999999999999,
658
+ "grad_norm": 0.13691408932209015,
659
+ "learning_rate": 0.0003163272520667726,
660
+ "loss": 0.1942,
661
+ "step": 2325
662
+ },
663
+ {
664
+ "epoch": 1.88,
665
+ "grad_norm": 0.14375685155391693,
666
+ "learning_rate": 0.00030662842385607126,
667
+ "loss": 0.1857,
668
+ "step": 2350
669
+ },
670
+ {
671
+ "epoch": 1.9,
672
+ "grad_norm": 0.15597157180309296,
673
+ "learning_rate": 0.0002970144148096568,
674
+ "loss": 0.2074,
675
+ "step": 2375
676
+ },
677
+ {
678
+ "epoch": 1.92,
679
+ "grad_norm": 0.1424601674079895,
680
+ "learning_rate": 0.0002874894419496431,
681
+ "loss": 0.1926,
682
+ "step": 2400
683
+ },
684
+ {
685
+ "epoch": 1.94,
686
+ "grad_norm": 0.15565040707588196,
687
+ "learning_rate": 0.00027805768324393014,
688
+ "loss": 0.1851,
689
+ "step": 2425
690
+ },
691
+ {
692
+ "epoch": 1.96,
693
+ "grad_norm": 0.1320432424545288,
694
+ "learning_rate": 0.0002687232757736082,
695
+ "loss": 0.1866,
696
+ "step": 2450
697
+ },
698
+ {
699
+ "epoch": 1.98,
700
+ "grad_norm": 0.13733460009098053,
701
+ "learning_rate": 0.0002594903139182996,
702
+ "loss": 0.1924,
703
+ "step": 2475
704
+ },
705
+ {
706
+ "epoch": 2.0,
707
+ "grad_norm": 0.12571240961551666,
708
+ "learning_rate": 0.0002503628475602256,
709
+ "loss": 0.1885,
710
+ "step": 2500
711
+ },
712
+ {
713
+ "epoch": 2.02,
714
+ "grad_norm": 0.13903608918190002,
715
+ "learning_rate": 0.00024134488030779655,
716
+ "loss": 0.1598,
717
+ "step": 2525
718
+ },
719
+ {
720
+ "epoch": 2.04,
721
+ "grad_norm": 0.15173819661140442,
722
+ "learning_rate": 0.00023244036773949656,
723
+ "loss": 0.1522,
724
+ "step": 2550
725
+ },
726
+ {
727
+ "epoch": 2.06,
728
+ "grad_norm": 0.13790248334407806,
729
+ "learning_rate": 0.00022365321566883433,
730
+ "loss": 0.1584,
731
+ "step": 2575
732
+ },
733
+ {
734
+ "epoch": 2.08,
735
+ "grad_norm": 0.1581411212682724,
736
+ "learning_rate": 0.0002149872784311262,
737
+ "loss": 0.1508,
738
+ "step": 2600
739
+ },
740
+ {
741
+ "epoch": 2.1,
742
+ "grad_norm": 0.1376183032989502,
743
+ "learning_rate": 0.00020644635719285705,
744
+ "loss": 0.1573,
745
+ "step": 2625
746
+ },
747
+ {
748
+ "epoch": 2.12,
749
+ "grad_norm": 0.12709948420524597,
750
+ "learning_rate": 0.0001980341982843616,
751
+ "loss": 0.1511,
752
+ "step": 2650
753
+ },
754
+ {
755
+ "epoch": 2.14,
756
+ "grad_norm": 0.13753506541252136,
757
+ "learning_rate": 0.0001897544915565616,
758
+ "loss": 0.1585,
759
+ "step": 2675
760
+ },
761
+ {
762
+ "epoch": 2.16,
763
+ "grad_norm": 0.1560378223657608,
764
+ "learning_rate": 0.0001816108687624749,
765
+ "loss": 0.1551,
766
+ "step": 2700
767
+ },
768
+ {
769
+ "epoch": 2.18,
770
+ "grad_norm": 0.14048361778259277,
771
+ "learning_rate": 0.00017360690196420813,
772
+ "loss": 0.1593,
773
+ "step": 2725
774
+ },
775
+ {
776
+ "epoch": 2.2,
777
+ "grad_norm": 0.15227031707763672,
778
+ "learning_rate": 0.0001657461019661326,
779
+ "loss": 0.1561,
780
+ "step": 2750
781
+ },
782
+ {
783
+ "epoch": 2.22,
784
+ "grad_norm": 0.14169102907180786,
785
+ "learning_rate": 0.0001580319167749294,
786
+ "loss": 0.1516,
787
+ "step": 2775
788
+ },
789
+ {
790
+ "epoch": 2.24,
791
+ "grad_norm": 0.13188719749450684,
792
+ "learning_rate": 0.00015046773008717967,
793
+ "loss": 0.1536,
794
+ "step": 2800
795
+ },
796
+ {
797
+ "epoch": 2.26,
798
+ "grad_norm": 0.1357060670852661,
799
+ "learning_rate": 0.00014305685980516293,
800
+ "loss": 0.1598,
801
+ "step": 2825
802
+ },
803
+ {
804
+ "epoch": 2.2800000000000002,
805
+ "grad_norm": 0.13821901381015778,
806
+ "learning_rate": 0.00013580255658151685,
807
+ "loss": 0.1494,
808
+ "step": 2850
809
+ },
810
+ {
811
+ "epoch": 2.3,
812
+ "grad_norm": 0.13461646437644958,
813
+ "learning_rate": 0.00012870800239339237,
814
+ "loss": 0.1539,
815
+ "step": 2875
816
+ },
817
+ {
818
+ "epoch": 2.32,
819
+ "grad_norm": 0.1364993304014206,
820
+ "learning_rate": 0.00012177630914673327,
821
+ "loss": 0.1599,
822
+ "step": 2900
823
+ },
824
+ {
825
+ "epoch": 2.34,
826
+ "grad_norm": 0.12037210166454315,
827
+ "learning_rate": 0.00011501051731129224,
828
+ "loss": 0.1503,
829
+ "step": 2925
830
+ },
831
+ {
832
+ "epoch": 2.36,
833
+ "grad_norm": 0.15030299127101898,
834
+ "learning_rate": 0.00010841359458697985,
835
+ "loss": 0.156,
836
+ "step": 2950
837
+ },
838
+ {
839
+ "epoch": 2.38,
840
+ "grad_norm": 0.1375398337841034,
841
+ "learning_rate": 0.00010198843460213336,
842
+ "loss": 0.1505,
843
+ "step": 2975
844
+ },
845
+ {
846
+ "epoch": 2.4,
847
+ "grad_norm": 0.13607092201709747,
848
+ "learning_rate": 9.573785564427562e-05,
849
+ "loss": 0.1518,
850
+ "step": 3000
851
+ },
852
+ {
853
+ "epoch": 2.42,
854
+ "grad_norm": 0.1567993462085724,
855
+ "learning_rate": 8.966459942392108e-05,
856
+ "loss": 0.1532,
857
+ "step": 3025
858
+ },
859
+ {
860
+ "epoch": 2.44,
861
+ "grad_norm": 0.1382758766412735,
862
+ "learning_rate": 8.3771329871971e-05,
863
+ "loss": 0.1548,
864
+ "step": 3050
865
+ },
866
+ {
867
+ "epoch": 2.46,
868
+ "grad_norm": 0.1536836475133896,
869
+ "learning_rate": 7.806063197122521e-05,
870
+ "loss": 0.155,
871
+ "step": 3075
872
+ },
873
+ {
874
+ "epoch": 2.48,
875
+ "grad_norm": 0.15493731200695038,
876
+ "learning_rate": 7.253501062252338e-05,
877
+ "loss": 0.1506,
878
+ "step": 3100
879
+ },
880
+ {
881
+ "epoch": 2.5,
882
+ "grad_norm": 0.15662576258182526,
883
+ "learning_rate": 6.719688954601267e-05,
884
+ "loss": 0.1578,
885
+ "step": 3125
886
+ },
887
+ {
888
+ "epoch": 2.52,
889
+ "grad_norm": 0.143663227558136,
890
+ "learning_rate": 6.204861021802333e-05,
891
+ "loss": 0.1499,
892
+ "step": 3150
893
+ },
894
+ {
895
+ "epoch": 2.54,
896
+ "grad_norm": 0.15504980087280273,
897
+ "learning_rate": 5.709243084402127e-05,
898
+ "loss": 0.1543,
899
+ "step": 3175
900
+ },
901
+ {
902
+ "epoch": 2.56,
903
+ "grad_norm": 0.1305331438779831,
904
+ "learning_rate": 5.2330525368083193e-05,
905
+ "loss": 0.1523,
906
+ "step": 3200
907
+ },
908
+ {
909
+ "epoch": 2.58,
910
+ "grad_norm": 0.13434988260269165,
911
+ "learning_rate": 4.776498251933292e-05,
912
+ "loss": 0.1471,
913
+ "step": 3225
914
+ },
915
+ {
916
+ "epoch": 2.6,
917
+ "grad_norm": 0.15677686035633087,
918
+ "learning_rate": 4.3397804895756956e-05,
919
+ "loss": 0.1481,
920
+ "step": 3250
921
+ },
922
+ {
923
+ "epoch": 2.62,
924
+ "grad_norm": 0.1596074402332306,
925
+ "learning_rate": 3.923090808579727e-05,
926
+ "loss": 0.1503,
927
+ "step": 3275
928
+ },
929
+ {
930
+ "epoch": 2.64,
931
+ "grad_norm": 0.13074015080928802,
932
+ "learning_rate": 3.5266119828111955e-05,
933
+ "loss": 0.1499,
934
+ "step": 3300
935
+ },
936
+ {
937
+ "epoch": 2.66,
938
+ "grad_norm": 0.14830058813095093,
939
+ "learning_rate": 3.150517920986851e-05,
940
+ "loss": 0.1532,
941
+ "step": 3325
942
+ },
943
+ {
944
+ "epoch": 2.68,
945
+ "grad_norm": 0.15981647372245789,
946
+ "learning_rate": 2.794973590392219e-05,
947
+ "loss": 0.1559,
948
+ "step": 3350
949
+ },
950
+ {
951
+ "epoch": 2.7,
952
+ "grad_norm": 0.14149022102355957,
953
+ "learning_rate": 2.460134944521547e-05,
954
+ "loss": 0.1518,
955
+ "step": 3375
956
+ },
957
+ {
958
+ "epoch": 2.7199999999999998,
959
+ "grad_norm": 0.16715629398822784,
960
+ "learning_rate": 2.1461488546714426e-05,
961
+ "loss": 0.1609,
962
+ "step": 3400
963
+ },
964
+ {
965
+ "epoch": 2.74,
966
+ "grad_norm": 0.11692510545253754,
967
+ "learning_rate": 1.853153045518252e-05,
968
+ "loss": 0.1537,
969
+ "step": 3425
970
+ },
971
+ {
972
+ "epoch": 2.76,
973
+ "grad_norm": 0.1363067477941513,
974
+ "learning_rate": 1.581276034707463e-05,
975
+ "loss": 0.1446,
976
+ "step": 3450
977
+ },
978
+ {
979
+ "epoch": 2.7800000000000002,
980
+ "grad_norm": 0.1612735539674759,
981
+ "learning_rate": 1.3306370764816389e-05,
982
+ "loss": 0.1493,
983
+ "step": 3475
984
+ },
985
+ {
986
+ "epoch": 2.8,
987
+ "grad_norm": 0.1385759711265564,
988
+ "learning_rate": 1.1013461093715594e-05,
989
+ "loss": 0.1488,
990
+ "step": 3500
991
+ },
992
+ {
993
+ "epoch": 2.82,
994
+ "grad_norm": 0.14177751541137695,
995
+ "learning_rate": 8.935037079735309e-06,
996
+ "loss": 0.1513,
997
+ "step": 3525
998
+ },
999
+ {
1000
+ "epoch": 2.84,
1001
+ "grad_norm": 0.15858101844787598,
1002
+ "learning_rate": 7.072010388340655e-06,
1003
+ "loss": 0.1475,
1004
+ "step": 3550
1005
+ },
1006
+ {
1007
+ "epoch": 2.86,
1008
+ "grad_norm": 0.14567403495311737,
1009
+ "learning_rate": 5.425198204612069e-06,
1010
+ "loss": 0.146,
1011
+ "step": 3575
1012
+ },
1013
+ {
1014
+ "epoch": 2.88,
1015
+ "grad_norm": 0.14445094764232635,
1016
+ "learning_rate": 3.995322874800922e-06,
1017
+ "loss": 0.153,
1018
+ "step": 3600
1019
+ },
1020
+ {
1021
+ "epoch": 2.9,
1022
+ "grad_norm": 0.1478790044784546,
1023
+ "learning_rate": 2.7830115894847407e-06,
1024
+ "loss": 0.1532,
1025
+ "step": 3625
1026
+ },
1027
+ {
1028
+ "epoch": 2.92,
1029
+ "grad_norm": 0.1748945713043213,
1030
+ "learning_rate": 1.7887961084605553e-06,
1031
+ "loss": 0.151,
1032
+ "step": 3650
1033
+ },
1034
+ {
1035
+ "epoch": 2.94,
1036
+ "grad_norm": 0.13460545241832733,
1037
+ "learning_rate": 1.013112527497473e-06,
1038
+ "loss": 0.1467,
1039
+ "step": 3675
1040
+ },
1041
+ {
1042
+ "epoch": 2.96,
1043
+ "grad_norm": 0.14923390746116638,
1044
+ "learning_rate": 4.563010870506368e-07,
1045
+ "loss": 0.1487,
1046
+ "step": 3700
1047
+ },
1048
+ {
1049
+ "epoch": 2.98,
1050
+ "grad_norm": 0.12898091971874237,
1051
+ "learning_rate": 1.1860602302066203e-07,
1052
+ "loss": 0.1481,
1053
+ "step": 3725
1054
+ },
1055
+ {
1056
+ "epoch": 3.0,
1057
+ "grad_norm": 0.13318267464637756,
1058
+ "learning_rate": 1.7545962355258739e-10,
1059
+ "loss": 0.1523,
1060
+ "step": 3750
1061
+ },
1062
+ {
1063
+ "epoch": 3.0,
1064
+ "step": 3750,
1065
+ "total_flos": 2.43882352705536e+18,
1066
+ "train_loss": 0.2079261500676473,
1067
+ "train_runtime": 3290.8864,
1068
+ "train_samples_per_second": 36.464,
1069
+ "train_steps_per_second": 1.14
1070
+ }
1071
+ ],
1072
+ "logging_steps": 25,
1073
+ "max_steps": 3750,
1074
+ "num_input_tokens_seen": 0,
1075
+ "num_train_epochs": 3,
1076
+ "save_steps": 0,
1077
+ "stateful_callbacks": {
1078
+ "TrainerControl": {
1079
+ "args": {
1080
+ "should_epoch_stop": false,
1081
+ "should_evaluate": false,
1082
+ "should_log": false,
1083
+ "should_save": false,
1084
+ "should_training_stop": false
1085
+ },
1086
+ "attributes": {}
1087
+ }
1088
+ },
1089
+ "total_flos": 2.43882352705536e+18,
1090
+ "train_batch_size": 32,
1091
+ "trial_name": null,
1092
+ "trial_params": null
1093
+ }
nl_tasks/exprep/run_ex26_2ep/ft/adapter_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "drop_out": 0.0,
6
+ "inference_mode": false,
7
+ "layers_to_transform": null,
8
+ "modules_to_save": null,
9
+ "num_rotations": 1,
10
+ "peft_type": "ROTATION",
11
+ "r": 16,
12
+ "revision": null,
13
+ "target_modules": [
14
+ "q_proj",
15
+ "v_proj"
16
+ ],
17
+ "target_modules_to_skip": null,
18
+ "task_type": "CAUSAL_LM"
19
+ }
nl_tasks/exprep/run_ex26_2ep/ft/special_tokens_map.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "</s>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": "<unk>",
17
+ "unk_token": {
18
+ "content": "<unk>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ }
24
+ }
nl_tasks/exprep/run_ex26_2ep/ft/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
nl_tasks/exprep/run_ex26_2ep/ft/tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
3
+ size 499723
nl_tasks/exprep/run_ex26_2ep/ft/tokenizer_config.json ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": true,
3
+ "add_eos_token": false,
4
+ "add_prefix_space": null,
5
+ "added_tokens_decoder": {
6
+ "0": {
7
+ "content": "<unk>",
8
+ "lstrip": false,
9
+ "normalized": false,
10
+ "rstrip": false,
11
+ "single_word": false,
12
+ "special": true
13
+ },
14
+ "1": {
15
+ "content": "<s>",
16
+ "lstrip": false,
17
+ "normalized": false,
18
+ "rstrip": false,
19
+ "single_word": false,
20
+ "special": true
21
+ },
22
+ "2": {
23
+ "content": "</s>",
24
+ "lstrip": false,
25
+ "normalized": false,
26
+ "rstrip": false,
27
+ "single_word": false,
28
+ "special": true
29
+ }
30
+ },
31
+ "bos_token": "<s>",
32
+ "clean_up_tokenization_spaces": false,
33
+ "eos_token": "</s>",
34
+ "extra_special_tokens": {},
35
+ "legacy": false,
36
+ "model_max_length": 512,
37
+ "pad_token": "<unk>",
38
+ "padding_side": "right",
39
+ "sp_model_kwargs": {},
40
+ "tokenizer_class": "LlamaTokenizer",
41
+ "unk_token": "<unk>",
42
+ "use_default_system_prompt": false
43
+ }
nl_tasks/exprep/run_ex26_2ep/ft2/adapter_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "drop_out": 0.0,
6
+ "inference_mode": true,
7
+ "layers_to_transform": null,
8
+ "modules_to_save": null,
9
+ "num_rotations": 1,
10
+ "peft_type": "ROTATION",
11
+ "r": 16,
12
+ "revision": null,
13
+ "target_modules": [
14
+ "q_proj",
15
+ "v_proj"
16
+ ],
17
+ "target_modules_to_skip": null,
18
+ "task_type": "CAUSAL_LM"
19
+ }
nl_tasks/exprep/run_ex26_2ep/ft2/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d0b154b76fa19caa8c6c9624749ab9c2b7b22f7e4d1358967eb95a56eb605cca
3
+ size 33602915
nl_tasks/exprep/run_ex26_2ep/output.txt ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+
2
+ MATH math MAX TOKEN = 1408, length==== 5000, math acc %====, 7.5200000000000005
3
+
4
+ gsm8k MAX TOKEN = 1024, length==== 1319, gsm8k acc %====, 49.58301743745261
nl_tasks/exprep/run_ex26_2ep/trainer_state.json ADDED
@@ -0,0 +1,743 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 2.0,
6
+ "eval_steps": 100,
7
+ "global_step": 2500,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.02,
14
+ "grad_norm": 0.18889440596103668,
15
+ "learning_rate": 0.0007998180972402738,
16
+ "loss": 0.44,
17
+ "step": 25
18
+ },
19
+ {
20
+ "epoch": 0.04,
21
+ "grad_norm": 0.2018483579158783,
22
+ "learning_rate": 0.0007992419381164945,
23
+ "loss": 0.3401,
24
+ "step": 50
25
+ },
26
+ {
27
+ "epoch": 0.06,
28
+ "grad_norm": 0.19704411923885345,
29
+ "learning_rate": 0.0007982717754008577,
30
+ "loss": 0.3321,
31
+ "step": 75
32
+ },
33
+ {
34
+ "epoch": 0.08,
35
+ "grad_norm": 0.18659791350364685,
36
+ "learning_rate": 0.0007969085665268344,
37
+ "loss": 0.3118,
38
+ "step": 100
39
+ },
40
+ {
41
+ "epoch": 0.1,
42
+ "grad_norm": 34.316856384277344,
43
+ "learning_rate": 0.0007951536568170009,
44
+ "loss": 0.3062,
45
+ "step": 125
46
+ },
47
+ {
48
+ "epoch": 0.12,
49
+ "grad_norm": 0.2079719454050064,
50
+ "learning_rate": 0.0007930087781553683,
51
+ "loss": 0.2998,
52
+ "step": 150
53
+ },
54
+ {
55
+ "epoch": 0.14,
56
+ "grad_norm": 0.2102925032377243,
57
+ "learning_rate": 0.0007904760472782212,
58
+ "loss": 0.297,
59
+ "step": 175
60
+ },
61
+ {
62
+ "epoch": 0.16,
63
+ "grad_norm": 0.23445385694503784,
64
+ "learning_rate": 0.0007875579636851548,
65
+ "loss": 0.2981,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.18,
70
+ "grad_norm": 0.21835000813007355,
71
+ "learning_rate": 0.0007842574071723712,
72
+ "loss": 0.2925,
73
+ "step": 225
74
+ },
75
+ {
76
+ "epoch": 0.2,
77
+ "grad_norm": 0.2103710025548935,
78
+ "learning_rate": 0.0007805776349906676,
79
+ "loss": 0.2875,
80
+ "step": 250
81
+ },
82
+ {
83
+ "epoch": 0.22,
84
+ "grad_norm": 0.20724424719810486,
85
+ "learning_rate": 0.0007765222786309228,
86
+ "loss": 0.2869,
87
+ "step": 275
88
+ },
89
+ {
90
+ "epoch": 0.24,
91
+ "grad_norm": 0.21245749294757843,
92
+ "learning_rate": 0.0007720953402402549,
93
+ "loss": 0.295,
94
+ "step": 300
95
+ },
96
+ {
97
+ "epoch": 0.26,
98
+ "grad_norm": 0.18448853492736816,
99
+ "learning_rate": 0.0007673011886723848,
100
+ "loss": 0.2878,
101
+ "step": 325
102
+ },
103
+ {
104
+ "epoch": 0.28,
105
+ "grad_norm": 0.21692252159118652,
106
+ "learning_rate": 0.0007621445551761047,
107
+ "loss": 0.2898,
108
+ "step": 350
109
+ },
110
+ {
111
+ "epoch": 0.3,
112
+ "grad_norm": 0.21624179184436798,
113
+ "learning_rate": 0.0007566305287261081,
114
+ "loss": 0.2804,
115
+ "step": 375
116
+ },
117
+ {
118
+ "epoch": 0.32,
119
+ "grad_norm": 0.23646491765975952,
120
+ "learning_rate": 0.0007507645510007842,
121
+ "loss": 0.2824,
122
+ "step": 400
123
+ },
124
+ {
125
+ "epoch": 0.34,
126
+ "grad_norm": 0.1886572539806366,
127
+ "learning_rate": 0.00074455241101194,
128
+ "loss": 0.2713,
129
+ "step": 425
130
+ },
131
+ {
132
+ "epoch": 0.36,
133
+ "grad_norm": 0.2086557298898697,
134
+ "learning_rate": 0.0007380002393917437,
135
+ "loss": 0.273,
136
+ "step": 450
137
+ },
138
+ {
139
+ "epoch": 0.38,
140
+ "grad_norm": 0.19460327923297882,
141
+ "learning_rate": 0.0007311145023425311,
142
+ "loss": 0.2705,
143
+ "step": 475
144
+ },
145
+ {
146
+ "epoch": 0.4,
147
+ "grad_norm": 0.2240498960018158,
148
+ "learning_rate": 0.000723901995255445,
149
+ "loss": 0.2684,
150
+ "step": 500
151
+ },
152
+ {
153
+ "epoch": 0.42,
154
+ "grad_norm": 0.1837693303823471,
155
+ "learning_rate": 0.0007163698360042034,
156
+ "loss": 0.27,
157
+ "step": 525
158
+ },
159
+ {
160
+ "epoch": 0.44,
161
+ "grad_norm": 0.2206038236618042,
162
+ "learning_rate": 0.0007085254579206188,
163
+ "loss": 0.2584,
164
+ "step": 550
165
+ },
166
+ {
167
+ "epoch": 0.46,
168
+ "grad_norm": 0.21168336272239685,
169
+ "learning_rate": 0.0007003766024587967,
170
+ "loss": 0.2754,
171
+ "step": 575
172
+ },
173
+ {
174
+ "epoch": 0.48,
175
+ "grad_norm": 0.26030775904655457,
176
+ "learning_rate": 0.0006919313115552542,
177
+ "loss": 0.2601,
178
+ "step": 600
179
+ },
180
+ {
181
+ "epoch": 0.5,
182
+ "grad_norm": 0.21076390147209167,
183
+ "learning_rate": 0.0006831979196925012,
184
+ "loss": 0.275,
185
+ "step": 625
186
+ },
187
+ {
188
+ "epoch": 0.52,
189
+ "grad_norm": 0.1726297289133072,
190
+ "learning_rate": 0.0006741850456739107,
191
+ "loss": 0.2641,
192
+ "step": 650
193
+ },
194
+ {
195
+ "epoch": 0.54,
196
+ "grad_norm": 0.16293871402740479,
197
+ "learning_rate": 0.0006649015841180021,
198
+ "loss": 0.2553,
199
+ "step": 675
200
+ },
201
+ {
202
+ "epoch": 0.56,
203
+ "grad_norm": 0.22282379865646362,
204
+ "learning_rate": 0.0006553566966805246,
205
+ "loss": 0.2563,
206
+ "step": 700
207
+ },
208
+ {
209
+ "epoch": 0.58,
210
+ "grad_norm": 0.22788006067276,
211
+ "learning_rate": 0.0006455598030130111,
212
+ "loss": 0.2628,
213
+ "step": 725
214
+ },
215
+ {
216
+ "epoch": 0.6,
217
+ "grad_norm": 0.18745820224285126,
218
+ "learning_rate": 0.000635520571466718,
219
+ "loss": 0.2581,
220
+ "step": 750
221
+ },
222
+ {
223
+ "epoch": 0.62,
224
+ "grad_norm": 0.19868342578411102,
225
+ "learning_rate": 0.0006252489095511306,
226
+ "loss": 0.2469,
227
+ "step": 775
228
+ },
229
+ {
230
+ "epoch": 0.64,
231
+ "grad_norm": 0.2159603387117386,
232
+ "learning_rate": 0.0006147549541564466,
233
+ "loss": 0.2521,
234
+ "step": 800
235
+ },
236
+ {
237
+ "epoch": 0.66,
238
+ "grad_norm": 0.22038054466247559,
239
+ "learning_rate": 0.0006040490615496899,
240
+ "loss": 0.2445,
241
+ "step": 825
242
+ },
243
+ {
244
+ "epoch": 0.68,
245
+ "grad_norm": 0.23632705211639404,
246
+ "learning_rate": 0.0005931417971543255,
247
+ "loss": 0.2473,
248
+ "step": 850
249
+ },
250
+ {
251
+ "epoch": 0.7,
252
+ "grad_norm": 0.17541927099227905,
253
+ "learning_rate": 0.0005820439251234615,
254
+ "loss": 0.255,
255
+ "step": 875
256
+ },
257
+ {
258
+ "epoch": 0.72,
259
+ "grad_norm": 0.1904209852218628,
260
+ "learning_rate": 0.0005707663977169301,
261
+ "loss": 0.2492,
262
+ "step": 900
263
+ },
264
+ {
265
+ "epoch": 0.74,
266
+ "grad_norm": 0.22302572429180145,
267
+ "learning_rate": 0.0005593203444927291,
268
+ "loss": 0.2462,
269
+ "step": 925
270
+ },
271
+ {
272
+ "epoch": 0.76,
273
+ "grad_norm": 0.20162935554981232,
274
+ "learning_rate": 0.0005477170613234922,
275
+ "loss": 0.2432,
276
+ "step": 950
277
+ },
278
+ {
279
+ "epoch": 0.78,
280
+ "grad_norm": 0.1888890117406845,
281
+ "learning_rate": 0.0005359679992488259,
282
+ "loss": 0.2394,
283
+ "step": 975
284
+ },
285
+ {
286
+ "epoch": 0.8,
287
+ "grad_norm": 0.21808800101280212,
288
+ "learning_rate": 0.0005240847531745164,
289
+ "loss": 0.2475,
290
+ "step": 1000
291
+ },
292
+ {
293
+ "epoch": 0.82,
294
+ "grad_norm": 0.20709307491779327,
295
+ "learning_rate": 0.0005120790504297575,
296
+ "loss": 0.247,
297
+ "step": 1025
298
+ },
299
+ {
300
+ "epoch": 0.84,
301
+ "grad_norm": 0.19248826801776886,
302
+ "learning_rate": 0.0004999627391936922,
303
+ "loss": 0.2362,
304
+ "step": 1050
305
+ },
306
+ {
307
+ "epoch": 0.86,
308
+ "grad_norm": 0.19905374944210052,
309
+ "learning_rate": 0.00048774777680269044,
310
+ "loss": 0.2323,
311
+ "step": 1075
312
+ },
313
+ {
314
+ "epoch": 0.88,
315
+ "grad_norm": 0.1787562519311905,
316
+ "learning_rate": 0.00047544621794990197,
317
+ "loss": 0.2323,
318
+ "step": 1100
319
+ },
320
+ {
321
+ "epoch": 0.9,
322
+ "grad_norm": 0.20622661709785461,
323
+ "learning_rate": 0.0004630702027887291,
324
+ "loss": 0.2418,
325
+ "step": 1125
326
+ },
327
+ {
328
+ "epoch": 0.92,
329
+ "grad_norm": 0.18993164598941803,
330
+ "learning_rate": 0.0004506319449519617,
331
+ "loss": 0.2352,
332
+ "step": 1150
333
+ },
334
+ {
335
+ "epoch": 0.94,
336
+ "grad_norm": 0.19909675419330597,
337
+ "learning_rate": 0.00043814371949839543,
338
+ "loss": 0.2378,
339
+ "step": 1175
340
+ },
341
+ {
342
+ "epoch": 0.96,
343
+ "grad_norm": 0.17602947354316711,
344
+ "learning_rate": 0.0004256178507988314,
345
+ "loss": 0.237,
346
+ "step": 1200
347
+ },
348
+ {
349
+ "epoch": 0.98,
350
+ "grad_norm": 0.19191932678222656,
351
+ "learning_rate": 0.00041306670037340957,
352
+ "loss": 0.2377,
353
+ "step": 1225
354
+ },
355
+ {
356
+ "epoch": 1.0,
357
+ "grad_norm": 0.202675461769104,
358
+ "learning_rate": 0.00040050265469228103,
359
+ "loss": 0.2278,
360
+ "step": 1250
361
+ },
362
+ {
363
+ "epoch": 1.02,
364
+ "grad_norm": 0.1944100260734558,
365
+ "learning_rate": 0.00038793811295165725,
366
+ "loss": 0.1968,
367
+ "step": 1275
368
+ },
369
+ {
370
+ "epoch": 1.04,
371
+ "grad_norm": 0.16131027042865753,
372
+ "learning_rate": 0.0003753854748373012,
373
+ "loss": 0.2042,
374
+ "step": 1300
375
+ },
376
+ {
377
+ "epoch": 1.06,
378
+ "grad_norm": 0.1991313099861145,
379
+ "learning_rate": 0.00036285712828753295,
380
+ "loss": 0.2005,
381
+ "step": 1325
382
+ },
383
+ {
384
+ "epoch": 1.08,
385
+ "grad_norm": 0.21017074584960938,
386
+ "learning_rate": 0.0003503654372678317,
387
+ "loss": 0.2048,
388
+ "step": 1350
389
+ },
390
+ {
391
+ "epoch": 1.1,
392
+ "grad_norm": 0.15863581001758575,
393
+ "learning_rate": 0.0003379227295690936,
394
+ "loss": 0.2086,
395
+ "step": 1375
396
+ },
397
+ {
398
+ "epoch": 1.12,
399
+ "grad_norm": 0.2023034393787384,
400
+ "learning_rate": 0.0003255412846415912,
401
+ "loss": 0.2055,
402
+ "step": 1400
403
+ },
404
+ {
405
+ "epoch": 1.1400000000000001,
406
+ "grad_norm": 0.19162552058696747,
407
+ "learning_rate": 0.0003132333214766379,
408
+ "loss": 0.1978,
409
+ "step": 1425
410
+ },
411
+ {
412
+ "epoch": 1.16,
413
+ "grad_norm": 0.17940473556518555,
414
+ "learning_rate": 0.0003010109865479191,
415
+ "loss": 0.2039,
416
+ "step": 1450
417
+ },
418
+ {
419
+ "epoch": 1.18,
420
+ "grad_norm": 0.16584932804107666,
421
+ "learning_rate": 0.0002888863418243891,
422
+ "loss": 0.2014,
423
+ "step": 1475
424
+ },
425
+ {
426
+ "epoch": 1.2,
427
+ "grad_norm": 0.20522530376911163,
428
+ "learning_rate": 0.0002768713528665639,
429
+ "loss": 0.2019,
430
+ "step": 1500
431
+ },
432
+ {
433
+ "epoch": 1.22,
434
+ "grad_norm": 0.18001499772071838,
435
+ "learning_rate": 0.0002649778770179578,
436
+ "loss": 0.2049,
437
+ "step": 1525
438
+ },
439
+ {
440
+ "epoch": 1.24,
441
+ "grad_norm": 0.16104567050933838,
442
+ "learning_rate": 0.00025321765170331634,
443
+ "loss": 0.1971,
444
+ "step": 1550
445
+ },
446
+ {
447
+ "epoch": 1.26,
448
+ "grad_norm": 0.17344725131988525,
449
+ "learning_rate": 0.00024160228284519604,
450
+ "loss": 0.1958,
451
+ "step": 1575
452
+ },
453
+ {
454
+ "epoch": 1.28,
455
+ "grad_norm": 0.16607116162776947,
456
+ "learning_rate": 0.0002301432334103195,
457
+ "loss": 0.1917,
458
+ "step": 1600
459
+ },
460
+ {
461
+ "epoch": 1.3,
462
+ "grad_norm": 0.19144991040229797,
463
+ "learning_rate": 0.00021885181209701105,
464
+ "loss": 0.1921,
465
+ "step": 1625
466
+ },
467
+ {
468
+ "epoch": 1.32,
469
+ "grad_norm": 0.1864110231399536,
470
+ "learning_rate": 0.0002077391621748769,
471
+ "loss": 0.1871,
472
+ "step": 1650
473
+ },
474
+ {
475
+ "epoch": 1.34,
476
+ "grad_norm": 0.1492071896791458,
477
+ "learning_rate": 0.0001968162504877441,
478
+ "loss": 0.1911,
479
+ "step": 1675
480
+ },
481
+ {
482
+ "epoch": 1.3599999999999999,
483
+ "grad_norm": 0.17169590294361115,
484
+ "learning_rate": 0.0001860938566307083,
485
+ "loss": 0.1895,
486
+ "step": 1700
487
+ },
488
+ {
489
+ "epoch": 1.38,
490
+ "grad_norm": 0.18199872970581055,
491
+ "learning_rate": 0.00017558256231197655,
492
+ "loss": 0.1909,
493
+ "step": 1725
494
+ },
495
+ {
496
+ "epoch": 1.4,
497
+ "grad_norm": 0.19598472118377686,
498
+ "learning_rate": 0.00016529274090999708,
499
+ "loss": 0.1898,
500
+ "step": 1750
501
+ },
502
+ {
503
+ "epoch": 1.42,
504
+ "grad_norm": 0.18632934987545013,
505
+ "learning_rate": 0.00015523454723618882,
506
+ "loss": 0.1912,
507
+ "step": 1775
508
+ },
509
+ {
510
+ "epoch": 1.44,
511
+ "grad_norm": 0.194894939661026,
512
+ "learning_rate": 0.0001454179075133671,
513
+ "loss": 0.1796,
514
+ "step": 1800
515
+ },
516
+ {
517
+ "epoch": 1.46,
518
+ "grad_norm": 0.1669997125864029,
519
+ "learning_rate": 0.00013585250957976128,
520
+ "loss": 0.1864,
521
+ "step": 1825
522
+ },
523
+ {
524
+ "epoch": 1.48,
525
+ "grad_norm": 0.2236054241657257,
526
+ "learning_rate": 0.00012654779332828725,
527
+ "loss": 0.1897,
528
+ "step": 1850
529
+ },
530
+ {
531
+ "epoch": 1.5,
532
+ "grad_norm": 0.1877262443304062,
533
+ "learning_rate": 0.00011751294139051308,
534
+ "loss": 0.1898,
535
+ "step": 1875
536
+ },
537
+ {
538
+ "epoch": 1.52,
539
+ "grad_norm": 0.18614347279071808,
540
+ "learning_rate": 0.00010875687007451065,
541
+ "loss": 0.1874,
542
+ "step": 1900
543
+ },
544
+ {
545
+ "epoch": 1.54,
546
+ "grad_norm": 0.20270414650440216,
547
+ "learning_rate": 0.00010028822056553551,
548
+ "loss": 0.1865,
549
+ "step": 1925
550
+ },
551
+ {
552
+ "epoch": 1.56,
553
+ "grad_norm": 0.16489075124263763,
554
+ "learning_rate": 9.211535039822043e-05,
555
+ "loss": 0.1773,
556
+ "step": 1950
557
+ },
558
+ {
559
+ "epoch": 1.58,
560
+ "grad_norm": 0.20095045864582062,
561
+ "learning_rate": 8.424632520869823e-05,
562
+ "loss": 0.1903,
563
+ "step": 1975
564
+ },
565
+ {
566
+ "epoch": 1.6,
567
+ "grad_norm": 0.17656490206718445,
568
+ "learning_rate": 7.668891077479186e-05,
569
+ "loss": 0.1886,
570
+ "step": 2000
571
+ },
572
+ {
573
+ "epoch": 1.62,
574
+ "grad_norm": 0.16787678003311157,
575
+ "learning_rate": 6.945056535212984e-05,
576
+ "loss": 0.1759,
577
+ "step": 2025
578
+ },
579
+ {
580
+ "epoch": 1.6400000000000001,
581
+ "grad_norm": 0.18357592821121216,
582
+ "learning_rate": 6.253843231374847e-05,
583
+ "loss": 0.1842,
584
+ "step": 2050
585
+ },
586
+ {
587
+ "epoch": 1.6600000000000001,
588
+ "grad_norm": 0.1719401627779007,
589
+ "learning_rate": 5.5959333100444563e-05,
590
+ "loss": 0.1885,
591
+ "step": 2075
592
+ },
593
+ {
594
+ "epoch": 1.6800000000000002,
595
+ "grad_norm": 0.18339356780052185,
596
+ "learning_rate": 4.971976048883749e-05,
597
+ "loss": 0.1856,
598
+ "step": 2100
599
+ },
600
+ {
601
+ "epoch": 1.7,
602
+ "grad_norm": 0.17477916181087494,
603
+ "learning_rate": 4.3825872183782046e-05,
604
+ "loss": 0.1835,
605
+ "step": 2125
606
+ },
607
+ {
608
+ "epoch": 1.72,
609
+ "grad_norm": 0.18121196329593658,
610
+ "learning_rate": 3.8283484741457754e-05,
611
+ "loss": 0.1765,
612
+ "step": 2150
613
+ },
614
+ {
615
+ "epoch": 1.74,
616
+ "grad_norm": 0.1621008664369583,
617
+ "learning_rate": 3.3098067829129405e-05,
618
+ "loss": 0.1931,
619
+ "step": 2175
620
+ },
621
+ {
622
+ "epoch": 1.76,
623
+ "grad_norm": 0.17131492495536804,
624
+ "learning_rate": 2.827473882724667e-05,
625
+ "loss": 0.184,
626
+ "step": 2200
627
+ },
628
+ {
629
+ "epoch": 1.78,
630
+ "grad_norm": 0.17526541650295258,
631
+ "learning_rate": 2.381825777920681e-05,
632
+ "loss": 0.1785,
633
+ "step": 2225
634
+ },
635
+ {
636
+ "epoch": 1.8,
637
+ "grad_norm": 0.17301619052886963,
638
+ "learning_rate": 1.9733022693766732e-05,
639
+ "loss": 0.1859,
640
+ "step": 2250
641
+ },
642
+ {
643
+ "epoch": 1.8199999999999998,
644
+ "grad_norm": 0.1691872924566269,
645
+ "learning_rate": 1.602306520473933e-05,
646
+ "loss": 0.1779,
647
+ "step": 2275
648
+ },
649
+ {
650
+ "epoch": 1.8399999999999999,
651
+ "grad_norm": 0.1832149177789688,
652
+ "learning_rate": 1.2692046592257711e-05,
653
+ "loss": 0.182,
654
+ "step": 2300
655
+ },
656
+ {
657
+ "epoch": 1.8599999999999999,
658
+ "grad_norm": 0.1830042004585266,
659
+ "learning_rate": 9.743254169533612e-06,
660
+ "loss": 0.1856,
661
+ "step": 2325
662
+ },
663
+ {
664
+ "epoch": 1.88,
665
+ "grad_norm": 0.17799071967601776,
666
+ "learning_rate": 7.1795980386763166e-06,
667
+ "loss": 0.1765,
668
+ "step": 2350
669
+ },
670
+ {
671
+ "epoch": 1.9,
672
+ "grad_norm": 0.20449566841125488,
673
+ "learning_rate": 5.003608218772949e-06,
674
+ "loss": 0.1993,
675
+ "step": 2375
676
+ },
677
+ {
678
+ "epoch": 1.92,
679
+ "grad_norm": 0.18933053314685822,
680
+ "learning_rate": 3.2174321490654023e-06,
681
+ "loss": 0.1857,
682
+ "step": 2400
683
+ },
684
+ {
685
+ "epoch": 1.94,
686
+ "grad_norm": 0.20368848741054535,
687
+ "learning_rate": 1.8228325696867388e-06,
688
+ "loss": 0.1792,
689
+ "step": 2425
690
+ },
691
+ {
692
+ "epoch": 1.96,
693
+ "grad_norm": 0.16929568350315094,
694
+ "learning_rate": 8.211857820497936e-07,
695
+ "loss": 0.1815,
696
+ "step": 2450
697
+ },
698
+ {
699
+ "epoch": 1.98,
700
+ "grad_norm": 0.1676410287618637,
701
+ "learning_rate": 2.134802906036626e-07,
702
+ "loss": 0.1858,
703
+ "step": 2475
704
+ },
705
+ {
706
+ "epoch": 2.0,
707
+ "grad_norm": 0.1708236187696457,
708
+ "learning_rate": 3.158272992909161e-10,
709
+ "loss": 0.1826,
710
+ "step": 2500
711
+ },
712
+ {
713
+ "epoch": 2.0,
714
+ "step": 2500,
715
+ "total_flos": 1.62588235137024e+18,
716
+ "train_loss": 0.22976668643951417,
717
+ "train_runtime": 2191.9604,
718
+ "train_samples_per_second": 36.497,
719
+ "train_steps_per_second": 1.141
720
+ }
721
+ ],
722
+ "logging_steps": 25,
723
+ "max_steps": 2500,
724
+ "num_input_tokens_seen": 0,
725
+ "num_train_epochs": 2,
726
+ "save_steps": 0,
727
+ "stateful_callbacks": {
728
+ "TrainerControl": {
729
+ "args": {
730
+ "should_epoch_stop": false,
731
+ "should_evaluate": false,
732
+ "should_log": false,
733
+ "should_save": false,
734
+ "should_training_stop": false
735
+ },
736
+ "attributes": {}
737
+ }
738
+ },
739
+ "total_flos": 1.62588235137024e+18,
740
+ "train_batch_size": 32,
741
+ "trial_name": null,
742
+ "trial_params": null
743
+ }
nl_tasks/exprep/run_ex27_2ep/ft/adapter_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "drop_out": 0.0,
6
+ "inference_mode": false,
7
+ "layers_to_transform": null,
8
+ "modules_to_save": null,
9
+ "num_rotations": 1,
10
+ "peft_type": "ROTATION",
11
+ "r": 16,
12
+ "revision": null,
13
+ "target_modules": [
14
+ "v_proj",
15
+ "q_proj"
16
+ ],
17
+ "target_modules_to_skip": null,
18
+ "task_type": "CAUSAL_LM"
19
+ }
nl_tasks/exprep/run_ex27_2ep/ft/special_tokens_map.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "</s>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": "<unk>",
17
+ "unk_token": {
18
+ "content": "<unk>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ }
24
+ }
nl_tasks/exprep/run_ex27_2ep/ft/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
nl_tasks/exprep/run_ex27_2ep/ft/tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
3
+ size 499723
nl_tasks/exprep/run_ex27_2ep/ft/tokenizer_config.json ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": true,
3
+ "add_eos_token": false,
4
+ "add_prefix_space": null,
5
+ "added_tokens_decoder": {
6
+ "0": {
7
+ "content": "<unk>",
8
+ "lstrip": false,
9
+ "normalized": false,
10
+ "rstrip": false,
11
+ "single_word": false,
12
+ "special": true
13
+ },
14
+ "1": {
15
+ "content": "<s>",
16
+ "lstrip": false,
17
+ "normalized": false,
18
+ "rstrip": false,
19
+ "single_word": false,
20
+ "special": true
21
+ },
22
+ "2": {
23
+ "content": "</s>",
24
+ "lstrip": false,
25
+ "normalized": false,
26
+ "rstrip": false,
27
+ "single_word": false,
28
+ "special": true
29
+ }
30
+ },
31
+ "bos_token": "<s>",
32
+ "clean_up_tokenization_spaces": false,
33
+ "eos_token": "</s>",
34
+ "extra_special_tokens": {},
35
+ "legacy": false,
36
+ "model_max_length": 512,
37
+ "pad_token": "<unk>",
38
+ "padding_side": "right",
39
+ "sp_model_kwargs": {},
40
+ "tokenizer_class": "LlamaTokenizer",
41
+ "unk_token": "<unk>",
42
+ "use_default_system_prompt": false
43
+ }
nl_tasks/exprep/run_ex27_2ep/ft2/adapter_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "drop_out": 0.0,
6
+ "inference_mode": true,
7
+ "layers_to_transform": null,
8
+ "modules_to_save": null,
9
+ "num_rotations": 1,
10
+ "peft_type": "ROTATION",
11
+ "r": 16,
12
+ "revision": null,
13
+ "target_modules": [
14
+ "v_proj",
15
+ "q_proj"
16
+ ],
17
+ "target_modules_to_skip": null,
18
+ "task_type": "CAUSAL_LM"
19
+ }
nl_tasks/exprep/run_ex27_2ep/ft2/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d2dbf019ad86e8c456c2e0832674f6eaca05fe021c9e496b0f14db596e72a8c0
3
+ size 33602915
nl_tasks/exprep/run_ex27_2ep/output.txt ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+
2
+ MATH math MAX TOKEN = 1408, length==== 5000, math acc %====, 8.260000000000002
3
+
4
+ gsm8k MAX TOKEN = 1024, length==== 1319, gsm8k acc %====, 50.49279757391963
nl_tasks/exprep/run_ex27_2ep/trainer_state.json ADDED
@@ -0,0 +1,743 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 2.0,
6
+ "eval_steps": 100,
7
+ "global_step": 2500,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.02,
14
+ "grad_norm": 0.19473141431808472,
15
+ "learning_rate": 0.0007998180972402738,
16
+ "loss": 0.4451,
17
+ "step": 25
18
+ },
19
+ {
20
+ "epoch": 0.04,
21
+ "grad_norm": 0.17338153719902039,
22
+ "learning_rate": 0.0007992419381164945,
23
+ "loss": 0.3417,
24
+ "step": 50
25
+ },
26
+ {
27
+ "epoch": 0.06,
28
+ "grad_norm": 0.19947905838489532,
29
+ "learning_rate": 0.0007982717754008577,
30
+ "loss": 0.3335,
31
+ "step": 75
32
+ },
33
+ {
34
+ "epoch": 0.08,
35
+ "grad_norm": 0.1866132766008377,
36
+ "learning_rate": 0.0007969085665268344,
37
+ "loss": 0.3141,
38
+ "step": 100
39
+ },
40
+ {
41
+ "epoch": 0.1,
42
+ "grad_norm": 0.20007456839084625,
43
+ "learning_rate": 0.0007951536568170009,
44
+ "loss": 0.306,
45
+ "step": 125
46
+ },
47
+ {
48
+ "epoch": 0.12,
49
+ "grad_norm": 0.2290574461221695,
50
+ "learning_rate": 0.0007930087781553683,
51
+ "loss": 0.3002,
52
+ "step": 150
53
+ },
54
+ {
55
+ "epoch": 0.14,
56
+ "grad_norm": 0.17895342409610748,
57
+ "learning_rate": 0.0007904760472782212,
58
+ "loss": 0.2969,
59
+ "step": 175
60
+ },
61
+ {
62
+ "epoch": 0.16,
63
+ "grad_norm": 0.22589446604251862,
64
+ "learning_rate": 0.0007875579636851548,
65
+ "loss": 0.2967,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.18,
70
+ "grad_norm": 0.21161246299743652,
71
+ "learning_rate": 0.0007842574071723712,
72
+ "loss": 0.2925,
73
+ "step": 225
74
+ },
75
+ {
76
+ "epoch": 0.2,
77
+ "grad_norm": 0.19765077531337738,
78
+ "learning_rate": 0.0007805776349906676,
79
+ "loss": 0.2876,
80
+ "step": 250
81
+ },
82
+ {
83
+ "epoch": 0.22,
84
+ "grad_norm": 0.2728215456008911,
85
+ "learning_rate": 0.0007765222786309228,
86
+ "loss": 0.2866,
87
+ "step": 275
88
+ },
89
+ {
90
+ "epoch": 0.24,
91
+ "grad_norm": 0.21330168843269348,
92
+ "learning_rate": 0.0007720953402402549,
93
+ "loss": 0.2975,
94
+ "step": 300
95
+ },
96
+ {
97
+ "epoch": 0.26,
98
+ "grad_norm": 0.19791795313358307,
99
+ "learning_rate": 0.0007673011886723848,
100
+ "loss": 0.288,
101
+ "step": 325
102
+ },
103
+ {
104
+ "epoch": 0.28,
105
+ "grad_norm": 0.20638969540596008,
106
+ "learning_rate": 0.0007621445551761047,
107
+ "loss": 0.2898,
108
+ "step": 350
109
+ },
110
+ {
111
+ "epoch": 0.3,
112
+ "grad_norm": 0.20914600789546967,
113
+ "learning_rate": 0.0007566305287261081,
114
+ "loss": 0.2813,
115
+ "step": 375
116
+ },
117
+ {
118
+ "epoch": 0.32,
119
+ "grad_norm": 0.22533443570137024,
120
+ "learning_rate": 0.0007507645510007842,
121
+ "loss": 0.2829,
122
+ "step": 400
123
+ },
124
+ {
125
+ "epoch": 0.34,
126
+ "grad_norm": 0.18723896145820618,
127
+ "learning_rate": 0.00074455241101194,
128
+ "loss": 0.2718,
129
+ "step": 425
130
+ },
131
+ {
132
+ "epoch": 0.36,
133
+ "grad_norm": 0.21526457369327545,
134
+ "learning_rate": 0.0007380002393917437,
135
+ "loss": 0.2748,
136
+ "step": 450
137
+ },
138
+ {
139
+ "epoch": 0.38,
140
+ "grad_norm": 0.19950586557388306,
141
+ "learning_rate": 0.0007311145023425311,
142
+ "loss": 0.271,
143
+ "step": 475
144
+ },
145
+ {
146
+ "epoch": 0.4,
147
+ "grad_norm": 0.21014024317264557,
148
+ "learning_rate": 0.000723901995255445,
149
+ "loss": 0.2696,
150
+ "step": 500
151
+ },
152
+ {
153
+ "epoch": 0.42,
154
+ "grad_norm": 0.20397445559501648,
155
+ "learning_rate": 0.0007163698360042034,
156
+ "loss": 0.2701,
157
+ "step": 525
158
+ },
159
+ {
160
+ "epoch": 0.44,
161
+ "grad_norm": 0.22161701321601868,
162
+ "learning_rate": 0.0007085254579206188,
163
+ "loss": 0.2585,
164
+ "step": 550
165
+ },
166
+ {
167
+ "epoch": 0.46,
168
+ "grad_norm": 0.21265128254890442,
169
+ "learning_rate": 0.0007003766024587967,
170
+ "loss": 0.2745,
171
+ "step": 575
172
+ },
173
+ {
174
+ "epoch": 0.48,
175
+ "grad_norm": 0.2174677550792694,
176
+ "learning_rate": 0.0006919313115552542,
177
+ "loss": 0.2616,
178
+ "step": 600
179
+ },
180
+ {
181
+ "epoch": 0.5,
182
+ "grad_norm": 0.20031920075416565,
183
+ "learning_rate": 0.0006831979196925012,
184
+ "loss": 0.2751,
185
+ "step": 625
186
+ },
187
+ {
188
+ "epoch": 0.52,
189
+ "grad_norm": 0.19264648854732513,
190
+ "learning_rate": 0.0006741850456739107,
191
+ "loss": 0.2656,
192
+ "step": 650
193
+ },
194
+ {
195
+ "epoch": 0.54,
196
+ "grad_norm": 0.16887356340885162,
197
+ "learning_rate": 0.0006649015841180021,
198
+ "loss": 0.2556,
199
+ "step": 675
200
+ },
201
+ {
202
+ "epoch": 0.56,
203
+ "grad_norm": 0.22922997176647186,
204
+ "learning_rate": 0.0006553566966805246,
205
+ "loss": 0.2563,
206
+ "step": 700
207
+ },
208
+ {
209
+ "epoch": 0.58,
210
+ "grad_norm": 0.22202101349830627,
211
+ "learning_rate": 0.0006455598030130111,
212
+ "loss": 0.2626,
213
+ "step": 725
214
+ },
215
+ {
216
+ "epoch": 0.6,
217
+ "grad_norm": 0.17986278235912323,
218
+ "learning_rate": 0.000635520571466718,
219
+ "loss": 0.258,
220
+ "step": 750
221
+ },
222
+ {
223
+ "epoch": 0.62,
224
+ "grad_norm": 0.20846596360206604,
225
+ "learning_rate": 0.0006252489095511306,
226
+ "loss": 0.2472,
227
+ "step": 775
228
+ },
229
+ {
230
+ "epoch": 0.64,
231
+ "grad_norm": 0.201336070895195,
232
+ "learning_rate": 0.0006147549541564466,
233
+ "loss": 0.2522,
234
+ "step": 800
235
+ },
236
+ {
237
+ "epoch": 0.66,
238
+ "grad_norm": 0.22574113309383392,
239
+ "learning_rate": 0.0006040490615496899,
240
+ "loss": 0.245,
241
+ "step": 825
242
+ },
243
+ {
244
+ "epoch": 0.68,
245
+ "grad_norm": 0.2243126779794693,
246
+ "learning_rate": 0.0005931417971543255,
247
+ "loss": 0.2474,
248
+ "step": 850
249
+ },
250
+ {
251
+ "epoch": 0.7,
252
+ "grad_norm": 0.16449490189552307,
253
+ "learning_rate": 0.0005820439251234615,
254
+ "loss": 0.2554,
255
+ "step": 875
256
+ },
257
+ {
258
+ "epoch": 0.72,
259
+ "grad_norm": 0.19276930391788483,
260
+ "learning_rate": 0.0005707663977169301,
261
+ "loss": 0.2505,
262
+ "step": 900
263
+ },
264
+ {
265
+ "epoch": 0.74,
266
+ "grad_norm": 0.20434655249118805,
267
+ "learning_rate": 0.0005593203444927291,
268
+ "loss": 0.2465,
269
+ "step": 925
270
+ },
271
+ {
272
+ "epoch": 0.76,
273
+ "grad_norm": 0.18825042247772217,
274
+ "learning_rate": 0.0005477170613234922,
275
+ "loss": 0.2444,
276
+ "step": 950
277
+ },
278
+ {
279
+ "epoch": 0.78,
280
+ "grad_norm": 0.17940199375152588,
281
+ "learning_rate": 0.0005359679992488259,
282
+ "loss": 0.2402,
283
+ "step": 975
284
+ },
285
+ {
286
+ "epoch": 0.8,
287
+ "grad_norm": 0.22762835025787354,
288
+ "learning_rate": 0.0005240847531745164,
289
+ "loss": 0.2475,
290
+ "step": 1000
291
+ },
292
+ {
293
+ "epoch": 0.82,
294
+ "grad_norm": 0.21415109932422638,
295
+ "learning_rate": 0.0005120790504297575,
296
+ "loss": 0.2457,
297
+ "step": 1025
298
+ },
299
+ {
300
+ "epoch": 0.84,
301
+ "grad_norm": 0.18646469712257385,
302
+ "learning_rate": 0.0004999627391936922,
303
+ "loss": 0.2366,
304
+ "step": 1050
305
+ },
306
+ {
307
+ "epoch": 0.86,
308
+ "grad_norm": 0.1955793797969818,
309
+ "learning_rate": 0.00048774777680269044,
310
+ "loss": 0.2331,
311
+ "step": 1075
312
+ },
313
+ {
314
+ "epoch": 0.88,
315
+ "grad_norm": 0.18219004571437836,
316
+ "learning_rate": 0.00047544621794990197,
317
+ "loss": 0.2327,
318
+ "step": 1100
319
+ },
320
+ {
321
+ "epoch": 0.9,
322
+ "grad_norm": 0.19894939661026,
323
+ "learning_rate": 0.0004630702027887291,
324
+ "loss": 0.2423,
325
+ "step": 1125
326
+ },
327
+ {
328
+ "epoch": 0.92,
329
+ "grad_norm": 0.1845511943101883,
330
+ "learning_rate": 0.0004506319449519617,
331
+ "loss": 0.235,
332
+ "step": 1150
333
+ },
334
+ {
335
+ "epoch": 0.94,
336
+ "grad_norm": 0.18997769057750702,
337
+ "learning_rate": 0.00043814371949839543,
338
+ "loss": 0.2381,
339
+ "step": 1175
340
+ },
341
+ {
342
+ "epoch": 0.96,
343
+ "grad_norm": 0.17170192301273346,
344
+ "learning_rate": 0.0004256178507988314,
345
+ "loss": 0.2367,
346
+ "step": 1200
347
+ },
348
+ {
349
+ "epoch": 0.98,
350
+ "grad_norm": 0.19611263275146484,
351
+ "learning_rate": 0.00041306670037340957,
352
+ "loss": 0.238,
353
+ "step": 1225
354
+ },
355
+ {
356
+ "epoch": 1.0,
357
+ "grad_norm": 0.1940995752811432,
358
+ "learning_rate": 0.00040050265469228103,
359
+ "loss": 0.2292,
360
+ "step": 1250
361
+ },
362
+ {
363
+ "epoch": 1.02,
364
+ "grad_norm": 0.19600820541381836,
365
+ "learning_rate": 0.00038793811295165725,
366
+ "loss": 0.1978,
367
+ "step": 1275
368
+ },
369
+ {
370
+ "epoch": 1.04,
371
+ "grad_norm": 0.15749067068099976,
372
+ "learning_rate": 0.0003753854748373012,
373
+ "loss": 0.2043,
374
+ "step": 1300
375
+ },
376
+ {
377
+ "epoch": 1.06,
378
+ "grad_norm": 0.18851898610591888,
379
+ "learning_rate": 0.00036285712828753295,
380
+ "loss": 0.2007,
381
+ "step": 1325
382
+ },
383
+ {
384
+ "epoch": 1.08,
385
+ "grad_norm": 0.19308072328567505,
386
+ "learning_rate": 0.0003503654372678317,
387
+ "loss": 0.2053,
388
+ "step": 1350
389
+ },
390
+ {
391
+ "epoch": 1.1,
392
+ "grad_norm": 0.17407938838005066,
393
+ "learning_rate": 0.0003379227295690936,
394
+ "loss": 0.2089,
395
+ "step": 1375
396
+ },
397
+ {
398
+ "epoch": 1.12,
399
+ "grad_norm": 0.18805305659770966,
400
+ "learning_rate": 0.0003255412846415912,
401
+ "loss": 0.2062,
402
+ "step": 1400
403
+ },
404
+ {
405
+ "epoch": 1.1400000000000001,
406
+ "grad_norm": 0.175668865442276,
407
+ "learning_rate": 0.0003132333214766379,
408
+ "loss": 0.1981,
409
+ "step": 1425
410
+ },
411
+ {
412
+ "epoch": 1.16,
413
+ "grad_norm": 0.18495599925518036,
414
+ "learning_rate": 0.0003010109865479191,
415
+ "loss": 0.2045,
416
+ "step": 1450
417
+ },
418
+ {
419
+ "epoch": 1.18,
420
+ "grad_norm": 0.17436009645462036,
421
+ "learning_rate": 0.0002888863418243891,
422
+ "loss": 0.2016,
423
+ "step": 1475
424
+ },
425
+ {
426
+ "epoch": 1.2,
427
+ "grad_norm": 0.20739738643169403,
428
+ "learning_rate": 0.0002768713528665639,
429
+ "loss": 0.202,
430
+ "step": 1500
431
+ },
432
+ {
433
+ "epoch": 1.22,
434
+ "grad_norm": 0.1784922331571579,
435
+ "learning_rate": 0.0002649778770179578,
436
+ "loss": 0.2052,
437
+ "step": 1525
438
+ },
439
+ {
440
+ "epoch": 1.24,
441
+ "grad_norm": 0.16647133231163025,
442
+ "learning_rate": 0.00025321765170331634,
443
+ "loss": 0.1973,
444
+ "step": 1550
445
+ },
446
+ {
447
+ "epoch": 1.26,
448
+ "grad_norm": 0.16744956374168396,
449
+ "learning_rate": 0.00024160228284519604,
450
+ "loss": 0.1958,
451
+ "step": 1575
452
+ },
453
+ {
454
+ "epoch": 1.28,
455
+ "grad_norm": 0.16644257307052612,
456
+ "learning_rate": 0.0002301432334103195,
457
+ "loss": 0.1922,
458
+ "step": 1600
459
+ },
460
+ {
461
+ "epoch": 1.3,
462
+ "grad_norm": 0.1884928047657013,
463
+ "learning_rate": 0.00021885181209701105,
464
+ "loss": 0.1927,
465
+ "step": 1625
466
+ },
467
+ {
468
+ "epoch": 1.32,
469
+ "grad_norm": 0.18612010776996613,
470
+ "learning_rate": 0.0002077391621748769,
471
+ "loss": 0.1873,
472
+ "step": 1650
473
+ },
474
+ {
475
+ "epoch": 1.34,
476
+ "grad_norm": 0.15592561662197113,
477
+ "learning_rate": 0.0001968162504877441,
478
+ "loss": 0.191,
479
+ "step": 1675
480
+ },
481
+ {
482
+ "epoch": 1.3599999999999999,
483
+ "grad_norm": 0.1704113483428955,
484
+ "learning_rate": 0.0001860938566307083,
485
+ "loss": 0.1896,
486
+ "step": 1700
487
+ },
488
+ {
489
+ "epoch": 1.38,
490
+ "grad_norm": 0.18215148150920868,
491
+ "learning_rate": 0.00017558256231197655,
492
+ "loss": 0.1909,
493
+ "step": 1725
494
+ },
495
+ {
496
+ "epoch": 1.4,
497
+ "grad_norm": 0.19163167476654053,
498
+ "learning_rate": 0.00016529274090999708,
499
+ "loss": 0.1912,
500
+ "step": 1750
501
+ },
502
+ {
503
+ "epoch": 1.42,
504
+ "grad_norm": 0.17857863008975983,
505
+ "learning_rate": 0.00015523454723618882,
506
+ "loss": 0.1914,
507
+ "step": 1775
508
+ },
509
+ {
510
+ "epoch": 1.44,
511
+ "grad_norm": 0.19024500250816345,
512
+ "learning_rate": 0.0001454179075133671,
513
+ "loss": 0.1801,
514
+ "step": 1800
515
+ },
516
+ {
517
+ "epoch": 1.46,
518
+ "grad_norm": 0.17564091086387634,
519
+ "learning_rate": 0.00013585250957976128,
520
+ "loss": 0.186,
521
+ "step": 1825
522
+ },
523
+ {
524
+ "epoch": 1.48,
525
+ "grad_norm": 0.207962304353714,
526
+ "learning_rate": 0.00012654779332828725,
527
+ "loss": 0.1899,
528
+ "step": 1850
529
+ },
530
+ {
531
+ "epoch": 1.5,
532
+ "grad_norm": 0.18071043491363525,
533
+ "learning_rate": 0.00011751294139051308,
534
+ "loss": 0.1898,
535
+ "step": 1875
536
+ },
537
+ {
538
+ "epoch": 1.52,
539
+ "grad_norm": 0.1892896443605423,
540
+ "learning_rate": 0.00010875687007451065,
541
+ "loss": 0.1879,
542
+ "step": 1900
543
+ },
544
+ {
545
+ "epoch": 1.54,
546
+ "grad_norm": 0.2063281238079071,
547
+ "learning_rate": 0.00010028822056553551,
548
+ "loss": 0.1872,
549
+ "step": 1925
550
+ },
551
+ {
552
+ "epoch": 1.56,
553
+ "grad_norm": 0.15895141661167145,
554
+ "learning_rate": 9.211535039822043e-05,
555
+ "loss": 0.1774,
556
+ "step": 1950
557
+ },
558
+ {
559
+ "epoch": 1.58,
560
+ "grad_norm": 0.19671915471553802,
561
+ "learning_rate": 8.424632520869823e-05,
562
+ "loss": 0.1905,
563
+ "step": 1975
564
+ },
565
+ {
566
+ "epoch": 1.6,
567
+ "grad_norm": 0.18066942691802979,
568
+ "learning_rate": 7.668891077479186e-05,
569
+ "loss": 0.1878,
570
+ "step": 2000
571
+ },
572
+ {
573
+ "epoch": 1.62,
574
+ "grad_norm": 0.17278794944286346,
575
+ "learning_rate": 6.945056535212984e-05,
576
+ "loss": 0.1756,
577
+ "step": 2025
578
+ },
579
+ {
580
+ "epoch": 1.6400000000000001,
581
+ "grad_norm": 0.17625869810581207,
582
+ "learning_rate": 6.253843231374847e-05,
583
+ "loss": 0.185,
584
+ "step": 2050
585
+ },
586
+ {
587
+ "epoch": 1.6600000000000001,
588
+ "grad_norm": 0.17790623009204865,
589
+ "learning_rate": 5.5959333100444563e-05,
590
+ "loss": 0.1883,
591
+ "step": 2075
592
+ },
593
+ {
594
+ "epoch": 1.6800000000000002,
595
+ "grad_norm": 0.18333546817302704,
596
+ "learning_rate": 4.971976048883749e-05,
597
+ "loss": 0.1858,
598
+ "step": 2100
599
+ },
600
+ {
601
+ "epoch": 1.7,
602
+ "grad_norm": 0.1687062680721283,
603
+ "learning_rate": 4.3825872183782046e-05,
604
+ "loss": 0.1836,
605
+ "step": 2125
606
+ },
607
+ {
608
+ "epoch": 1.72,
609
+ "grad_norm": 0.1715196669101715,
610
+ "learning_rate": 3.8283484741457754e-05,
611
+ "loss": 0.178,
612
+ "step": 2150
613
+ },
614
+ {
615
+ "epoch": 1.74,
616
+ "grad_norm": 0.16999520361423492,
617
+ "learning_rate": 3.3098067829129405e-05,
618
+ "loss": 0.1933,
619
+ "step": 2175
620
+ },
621
+ {
622
+ "epoch": 1.76,
623
+ "grad_norm": 0.18033349514007568,
624
+ "learning_rate": 2.827473882724667e-05,
625
+ "loss": 0.1832,
626
+ "step": 2200
627
+ },
628
+ {
629
+ "epoch": 1.78,
630
+ "grad_norm": 0.18522976338863373,
631
+ "learning_rate": 2.381825777920681e-05,
632
+ "loss": 0.1781,
633
+ "step": 2225
634
+ },
635
+ {
636
+ "epoch": 1.8,
637
+ "grad_norm": 0.1742217093706131,
638
+ "learning_rate": 1.9733022693766732e-05,
639
+ "loss": 0.1864,
640
+ "step": 2250
641
+ },
642
+ {
643
+ "epoch": 1.8199999999999998,
644
+ "grad_norm": 0.17098219692707062,
645
+ "learning_rate": 1.602306520473933e-05,
646
+ "loss": 0.1785,
647
+ "step": 2275
648
+ },
649
+ {
650
+ "epoch": 1.8399999999999999,
651
+ "grad_norm": 0.19406339526176453,
652
+ "learning_rate": 1.2692046592257711e-05,
653
+ "loss": 0.1825,
654
+ "step": 2300
655
+ },
656
+ {
657
+ "epoch": 1.8599999999999999,
658
+ "grad_norm": 0.18150834739208221,
659
+ "learning_rate": 9.743254169533612e-06,
660
+ "loss": 0.186,
661
+ "step": 2325
662
+ },
663
+ {
664
+ "epoch": 1.88,
665
+ "grad_norm": 0.1748575121164322,
666
+ "learning_rate": 7.1795980386763166e-06,
667
+ "loss": 0.1771,
668
+ "step": 2350
669
+ },
670
+ {
671
+ "epoch": 1.9,
672
+ "grad_norm": 0.20150317251682281,
673
+ "learning_rate": 5.003608218772949e-06,
674
+ "loss": 0.1998,
675
+ "step": 2375
676
+ },
677
+ {
678
+ "epoch": 1.92,
679
+ "grad_norm": 0.18875426054000854,
680
+ "learning_rate": 3.2174321490654023e-06,
681
+ "loss": 0.1858,
682
+ "step": 2400
683
+ },
684
+ {
685
+ "epoch": 1.94,
686
+ "grad_norm": 0.2131820172071457,
687
+ "learning_rate": 1.8228325696867388e-06,
688
+ "loss": 0.1797,
689
+ "step": 2425
690
+ },
691
+ {
692
+ "epoch": 1.96,
693
+ "grad_norm": 0.16796648502349854,
694
+ "learning_rate": 8.211857820497936e-07,
695
+ "loss": 0.181,
696
+ "step": 2450
697
+ },
698
+ {
699
+ "epoch": 1.98,
700
+ "grad_norm": 0.17758920788764954,
701
+ "learning_rate": 2.134802906036626e-07,
702
+ "loss": 0.1866,
703
+ "step": 2475
704
+ },
705
+ {
706
+ "epoch": 2.0,
707
+ "grad_norm": 0.16435496509075165,
708
+ "learning_rate": 3.158272992909161e-10,
709
+ "loss": 0.1833,
710
+ "step": 2500
711
+ },
712
+ {
713
+ "epoch": 2.0,
714
+ "step": 2500,
715
+ "total_flos": 1.62588235137024e+18,
716
+ "train_loss": 0.23017547645568848,
717
+ "train_runtime": 2191.7204,
718
+ "train_samples_per_second": 36.501,
719
+ "train_steps_per_second": 1.141
720
+ }
721
+ ],
722
+ "logging_steps": 25,
723
+ "max_steps": 2500,
724
+ "num_input_tokens_seen": 0,
725
+ "num_train_epochs": 2,
726
+ "save_steps": 0,
727
+ "stateful_callbacks": {
728
+ "TrainerControl": {
729
+ "args": {
730
+ "should_epoch_stop": false,
731
+ "should_evaluate": false,
732
+ "should_log": false,
733
+ "should_save": false,
734
+ "should_training_stop": false
735
+ },
736
+ "attributes": {}
737
+ }
738
+ },
739
+ "total_flos": 1.62588235137024e+18,
740
+ "train_batch_size": 32,
741
+ "trial_name": null,
742
+ "trial_params": null
743
+ }
nl_tasks/exprep/run_ex28_2ep/ft/adapter_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "drop_out": 0.0,
6
+ "inference_mode": false,
7
+ "layers_to_transform": null,
8
+ "modules_to_save": null,
9
+ "num_rotations": 1,
10
+ "peft_type": "ROTATION",
11
+ "r": 16,
12
+ "revision": null,
13
+ "target_modules": [
14
+ "q_proj",
15
+ "v_proj"
16
+ ],
17
+ "target_modules_to_skip": null,
18
+ "task_type": "CAUSAL_LM"
19
+ }
nl_tasks/exprep/run_ex28_2ep/ft/special_tokens_map.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "</s>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": "<unk>",
17
+ "unk_token": {
18
+ "content": "<unk>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ }
24
+ }
nl_tasks/exprep/run_ex28_2ep/ft/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
nl_tasks/exprep/run_ex28_2ep/ft/tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
3
+ size 499723
nl_tasks/exprep/run_ex28_2ep/ft/tokenizer_config.json ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": true,
3
+ "add_eos_token": false,
4
+ "add_prefix_space": null,
5
+ "added_tokens_decoder": {
6
+ "0": {
7
+ "content": "<unk>",
8
+ "lstrip": false,
9
+ "normalized": false,
10
+ "rstrip": false,
11
+ "single_word": false,
12
+ "special": true
13
+ },
14
+ "1": {
15
+ "content": "<s>",
16
+ "lstrip": false,
17
+ "normalized": false,
18
+ "rstrip": false,
19
+ "single_word": false,
20
+ "special": true
21
+ },
22
+ "2": {
23
+ "content": "</s>",
24
+ "lstrip": false,
25
+ "normalized": false,
26
+ "rstrip": false,
27
+ "single_word": false,
28
+ "special": true
29
+ }
30
+ },
31
+ "bos_token": "<s>",
32
+ "clean_up_tokenization_spaces": false,
33
+ "eos_token": "</s>",
34
+ "extra_special_tokens": {},
35
+ "legacy": false,
36
+ "model_max_length": 512,
37
+ "pad_token": "<unk>",
38
+ "padding_side": "right",
39
+ "sp_model_kwargs": {},
40
+ "tokenizer_class": "LlamaTokenizer",
41
+ "unk_token": "<unk>",
42
+ "use_default_system_prompt": false
43
+ }
nl_tasks/exprep/run_ex28_2ep/ft2/adapter_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "drop_out": 0.0,
6
+ "inference_mode": true,
7
+ "layers_to_transform": null,
8
+ "modules_to_save": null,
9
+ "num_rotations": 1,
10
+ "peft_type": "ROTATION",
11
+ "r": 16,
12
+ "revision": null,
13
+ "target_modules": [
14
+ "q_proj",
15
+ "v_proj"
16
+ ],
17
+ "target_modules_to_skip": null,
18
+ "task_type": "CAUSAL_LM"
19
+ }
nl_tasks/exprep/run_ex28_2ep/ft2/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:669aff8dc4f7131c4029d1add932e4275ae4df370bbab465e38c930b7a75b725
3
+ size 33602915
nl_tasks/exprep/run_ex28_2ep/output.txt ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+
2
+ MATH math MAX TOKEN = 1408, length==== 5000, math acc %====, 7.5200000000000005
3
+
4
+ gsm8k MAX TOKEN = 1024, length==== 1319, gsm8k acc %====, 50.11372251705838
nl_tasks/exprep/run_ex28_2ep/trainer_state.json ADDED
@@ -0,0 +1,743 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 2.0,
6
+ "eval_steps": 100,
7
+ "global_step": 2500,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.02,
14
+ "grad_norm": 0.24684618413448334,
15
+ "learning_rate": 0.0019995452431006843,
16
+ "loss": 0.5979,
17
+ "step": 25
18
+ },
19
+ {
20
+ "epoch": 0.04,
21
+ "grad_norm": 0.2515503466129303,
22
+ "learning_rate": 0.0019981048452912364,
23
+ "loss": 0.3524,
24
+ "step": 50
25
+ },
26
+ {
27
+ "epoch": 0.06,
28
+ "grad_norm": 0.27393826842308044,
29
+ "learning_rate": 0.001995679438502144,
30
+ "loss": 0.3483,
31
+ "step": 75
32
+ },
33
+ {
34
+ "epoch": 0.08,
35
+ "grad_norm": 0.2420358955860138,
36
+ "learning_rate": 0.0019922714163170857,
37
+ "loss": 0.3297,
38
+ "step": 100
39
+ },
40
+ {
41
+ "epoch": 0.1,
42
+ "grad_norm": 0.25277575850486755,
43
+ "learning_rate": 0.0019878841420425024,
44
+ "loss": 0.323,
45
+ "step": 125
46
+ },
47
+ {
48
+ "epoch": 0.12,
49
+ "grad_norm": 0.2367502897977829,
50
+ "learning_rate": 0.001982521945388421,
51
+ "loss": 0.3193,
52
+ "step": 150
53
+ },
54
+ {
55
+ "epoch": 0.14,
56
+ "grad_norm": 0.2569390535354614,
57
+ "learning_rate": 0.001976190118195553,
58
+ "loss": 0.3183,
59
+ "step": 175
60
+ },
61
+ {
62
+ "epoch": 0.16,
63
+ "grad_norm": 1.0411403179168701,
64
+ "learning_rate": 0.001968894909212887,
65
+ "loss": 0.3193,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.18,
70
+ "grad_norm": 0.40176984667778015,
71
+ "learning_rate": 0.001960643517930928,
72
+ "loss": 0.3185,
73
+ "step": 225
74
+ },
75
+ {
76
+ "epoch": 0.2,
77
+ "grad_norm": 0.2491682916879654,
78
+ "learning_rate": 0.001951444087476669,
79
+ "loss": 0.3117,
80
+ "step": 250
81
+ },
82
+ {
83
+ "epoch": 0.22,
84
+ "grad_norm": 0.25479909777641296,
85
+ "learning_rate": 0.0019413056965773068,
86
+ "loss": 0.3016,
87
+ "step": 275
88
+ },
89
+ {
90
+ "epoch": 0.24,
91
+ "grad_norm": 0.21606914699077606,
92
+ "learning_rate": 0.0019302383506006372,
93
+ "loss": 0.3082,
94
+ "step": 300
95
+ },
96
+ {
97
+ "epoch": 0.26,
98
+ "grad_norm": 0.17797593772411346,
99
+ "learning_rate": 0.0019182529716809619,
100
+ "loss": 0.3015,
101
+ "step": 325
102
+ },
103
+ {
104
+ "epoch": 0.28,
105
+ "grad_norm": 0.19606435298919678,
106
+ "learning_rate": 0.001905361387940262,
107
+ "loss": 0.2993,
108
+ "step": 350
109
+ },
110
+ {
111
+ "epoch": 0.3,
112
+ "grad_norm": 0.19274692237377167,
113
+ "learning_rate": 0.0018915763218152702,
114
+ "loss": 0.2875,
115
+ "step": 375
116
+ },
117
+ {
118
+ "epoch": 0.32,
119
+ "grad_norm": 0.1828506737947464,
120
+ "learning_rate": 0.0018769113775019605,
121
+ "loss": 0.2862,
122
+ "step": 400
123
+ },
124
+ {
125
+ "epoch": 0.34,
126
+ "grad_norm": 0.14898830652236938,
127
+ "learning_rate": 0.00186138102752985,
128
+ "loss": 0.2726,
129
+ "step": 425
130
+ },
131
+ {
132
+ "epoch": 0.36,
133
+ "grad_norm": 0.1651427447795868,
134
+ "learning_rate": 0.0018450005984793592,
135
+ "loss": 0.2755,
136
+ "step": 450
137
+ },
138
+ {
139
+ "epoch": 0.38,
140
+ "grad_norm": 0.15676464140415192,
141
+ "learning_rate": 0.001827786255856328,
142
+ "loss": 0.2717,
143
+ "step": 475
144
+ },
145
+ {
146
+ "epoch": 0.4,
147
+ "grad_norm": 0.16891220211982727,
148
+ "learning_rate": 0.0018097549881386124,
149
+ "loss": 0.2672,
150
+ "step": 500
151
+ },
152
+ {
153
+ "epoch": 0.42,
154
+ "grad_norm": 0.13077136874198914,
155
+ "learning_rate": 0.0017909245900105085,
156
+ "loss": 0.2684,
157
+ "step": 525
158
+ },
159
+ {
160
+ "epoch": 0.44,
161
+ "grad_norm": 0.15448248386383057,
162
+ "learning_rate": 0.001771313644801547,
163
+ "loss": 0.2554,
164
+ "step": 550
165
+ },
166
+ {
167
+ "epoch": 0.46,
168
+ "grad_norm": 0.20096218585968018,
169
+ "learning_rate": 0.0017509415061469915,
170
+ "loss": 0.2731,
171
+ "step": 575
172
+ },
173
+ {
174
+ "epoch": 0.48,
175
+ "grad_norm": 0.13531051576137543,
176
+ "learning_rate": 0.0017298282788881353,
177
+ "loss": 0.2596,
178
+ "step": 600
179
+ },
180
+ {
181
+ "epoch": 0.5,
182
+ "grad_norm": 0.14265109598636627,
183
+ "learning_rate": 0.0017079947992312528,
184
+ "loss": 0.2722,
185
+ "step": 625
186
+ },
187
+ {
188
+ "epoch": 0.52,
189
+ "grad_norm": 0.12629778683185577,
190
+ "learning_rate": 0.0016854626141847768,
191
+ "loss": 0.2632,
192
+ "step": 650
193
+ },
194
+ {
195
+ "epoch": 0.54,
196
+ "grad_norm": 0.11075589060783386,
197
+ "learning_rate": 0.001662253960295005,
198
+ "loss": 0.2516,
199
+ "step": 675
200
+ },
201
+ {
202
+ "epoch": 0.56,
203
+ "grad_norm": 0.18264581263065338,
204
+ "learning_rate": 0.0016383917417013114,
205
+ "loss": 0.2527,
206
+ "step": 700
207
+ },
208
+ {
209
+ "epoch": 0.58,
210
+ "grad_norm": 0.1566222757101059,
211
+ "learning_rate": 0.0016138995075325275,
212
+ "loss": 0.2596,
213
+ "step": 725
214
+ },
215
+ {
216
+ "epoch": 0.6,
217
+ "grad_norm": 0.1463090479373932,
218
+ "learning_rate": 0.0015888014286667951,
219
+ "loss": 0.2551,
220
+ "step": 750
221
+ },
222
+ {
223
+ "epoch": 0.62,
224
+ "grad_norm": 0.1279381960630417,
225
+ "learning_rate": 0.0015631222738778266,
226
+ "loss": 0.2424,
227
+ "step": 775
228
+ },
229
+ {
230
+ "epoch": 0.64,
231
+ "grad_norm": 0.14039047062397003,
232
+ "learning_rate": 0.0015368873853911164,
233
+ "loss": 0.2489,
234
+ "step": 800
235
+ },
236
+ {
237
+ "epoch": 0.66,
238
+ "grad_norm": 0.14027215540409088,
239
+ "learning_rate": 0.0015101226538742248,
240
+ "loss": 0.2429,
241
+ "step": 825
242
+ },
243
+ {
244
+ "epoch": 0.68,
245
+ "grad_norm": 0.1330851912498474,
246
+ "learning_rate": 0.0014828544928858136,
247
+ "loss": 0.2451,
248
+ "step": 850
249
+ },
250
+ {
251
+ "epoch": 0.7,
252
+ "grad_norm": 0.1147708147764206,
253
+ "learning_rate": 0.0014551098128086537,
254
+ "loss": 0.251,
255
+ "step": 875
256
+ },
257
+ {
258
+ "epoch": 0.72,
259
+ "grad_norm": 0.11136401444673538,
260
+ "learning_rate": 0.0014269159942923253,
261
+ "loss": 0.2472,
262
+ "step": 900
263
+ },
264
+ {
265
+ "epoch": 0.74,
266
+ "grad_norm": 0.12382576614618301,
267
+ "learning_rate": 0.001398300861231823,
268
+ "loss": 0.2437,
269
+ "step": 925
270
+ },
271
+ {
272
+ "epoch": 0.76,
273
+ "grad_norm": 0.1270427405834198,
274
+ "learning_rate": 0.0013692926533087305,
275
+ "loss": 0.2416,
276
+ "step": 950
277
+ },
278
+ {
279
+ "epoch": 0.78,
280
+ "grad_norm": 0.11768626421689987,
281
+ "learning_rate": 0.0013399199981220646,
282
+ "loss": 0.2384,
283
+ "step": 975
284
+ },
285
+ {
286
+ "epoch": 0.8,
287
+ "grad_norm": 0.1460709571838379,
288
+ "learning_rate": 0.001310211882936291,
289
+ "loss": 0.2455,
290
+ "step": 1000
291
+ },
292
+ {
293
+ "epoch": 0.82,
294
+ "grad_norm": 0.13237980008125305,
295
+ "learning_rate": 0.0012801976260743938,
296
+ "loss": 0.2427,
297
+ "step": 1025
298
+ },
299
+ {
300
+ "epoch": 0.84,
301
+ "grad_norm": 0.11956711858510971,
302
+ "learning_rate": 0.0012499068479842306,
303
+ "loss": 0.2334,
304
+ "step": 1050
305
+ },
306
+ {
307
+ "epoch": 0.86,
308
+ "grad_norm": 0.11264153569936752,
309
+ "learning_rate": 0.001219369442006726,
310
+ "loss": 0.2304,
311
+ "step": 1075
312
+ },
313
+ {
314
+ "epoch": 0.88,
315
+ "grad_norm": 0.11026793718338013,
316
+ "learning_rate": 0.001188615544874755,
317
+ "loss": 0.2311,
318
+ "step": 1100
319
+ },
320
+ {
321
+ "epoch": 0.9,
322
+ "grad_norm": 0.12786801159381866,
323
+ "learning_rate": 0.0011576755069718227,
324
+ "loss": 0.2387,
325
+ "step": 1125
326
+ },
327
+ {
328
+ "epoch": 0.92,
329
+ "grad_norm": 0.1032777950167656,
330
+ "learning_rate": 0.0011265798623799042,
331
+ "loss": 0.2332,
332
+ "step": 1150
333
+ },
334
+ {
335
+ "epoch": 0.94,
336
+ "grad_norm": 0.12246724963188171,
337
+ "learning_rate": 0.0010953592987459886,
338
+ "loss": 0.2363,
339
+ "step": 1175
340
+ },
341
+ {
342
+ "epoch": 0.96,
343
+ "grad_norm": 0.1136309951543808,
344
+ "learning_rate": 0.0010640446269970785,
345
+ "loss": 0.2348,
346
+ "step": 1200
347
+ },
348
+ {
349
+ "epoch": 0.98,
350
+ "grad_norm": 0.12908512353897095,
351
+ "learning_rate": 0.001032666750933524,
352
+ "loss": 0.2367,
353
+ "step": 1225
354
+ },
355
+ {
356
+ "epoch": 1.0,
357
+ "grad_norm": 0.12008602172136307,
358
+ "learning_rate": 0.0010012566367307025,
359
+ "loss": 0.2268,
360
+ "step": 1250
361
+ },
362
+ {
363
+ "epoch": 1.02,
364
+ "grad_norm": 0.11756276339292526,
365
+ "learning_rate": 0.0009698452823791432,
366
+ "loss": 0.1931,
367
+ "step": 1275
368
+ },
369
+ {
370
+ "epoch": 1.04,
371
+ "grad_norm": 0.09515941143035889,
372
+ "learning_rate": 0.000938463687093253,
373
+ "loss": 0.2001,
374
+ "step": 1300
375
+ },
376
+ {
377
+ "epoch": 1.06,
378
+ "grad_norm": 0.12476788461208344,
379
+ "learning_rate": 0.0009071428207188324,
380
+ "loss": 0.1954,
381
+ "step": 1325
382
+ },
383
+ {
384
+ "epoch": 1.08,
385
+ "grad_norm": 0.10795003920793533,
386
+ "learning_rate": 0.0008759135931695792,
387
+ "loss": 0.2017,
388
+ "step": 1350
389
+ },
390
+ {
391
+ "epoch": 1.1,
392
+ "grad_norm": 0.10403226315975189,
393
+ "learning_rate": 0.000844806823922734,
394
+ "loss": 0.2046,
395
+ "step": 1375
396
+ },
397
+ {
398
+ "epoch": 1.12,
399
+ "grad_norm": 0.10814425349235535,
400
+ "learning_rate": 0.0008138532116039781,
401
+ "loss": 0.2031,
402
+ "step": 1400
403
+ },
404
+ {
405
+ "epoch": 1.1400000000000001,
406
+ "grad_norm": 0.10225003957748413,
407
+ "learning_rate": 0.0007830833036915947,
408
+ "loss": 0.1953,
409
+ "step": 1425
410
+ },
411
+ {
412
+ "epoch": 1.16,
413
+ "grad_norm": 0.1085972711443901,
414
+ "learning_rate": 0.0007525274663697977,
415
+ "loss": 0.201,
416
+ "step": 1450
417
+ },
418
+ {
419
+ "epoch": 1.18,
420
+ "grad_norm": 0.09524306654930115,
421
+ "learning_rate": 0.0007222158545609727,
422
+ "loss": 0.1981,
423
+ "step": 1475
424
+ },
425
+ {
426
+ "epoch": 1.2,
427
+ "grad_norm": 0.1145230382680893,
428
+ "learning_rate": 0.0006921783821664097,
429
+ "loss": 0.1992,
430
+ "step": 1500
431
+ },
432
+ {
433
+ "epoch": 1.22,
434
+ "grad_norm": 0.10256361961364746,
435
+ "learning_rate": 0.0006624446925448944,
436
+ "loss": 0.2024,
437
+ "step": 1525
438
+ },
439
+ {
440
+ "epoch": 1.24,
441
+ "grad_norm": 0.09765516221523285,
442
+ "learning_rate": 0.0006330441292582908,
443
+ "loss": 0.1941,
444
+ "step": 1550
445
+ },
446
+ {
447
+ "epoch": 1.26,
448
+ "grad_norm": 0.10076688230037689,
449
+ "learning_rate": 0.00060400570711299,
450
+ "loss": 0.194,
451
+ "step": 1575
452
+ },
453
+ {
454
+ "epoch": 1.28,
455
+ "grad_norm": 0.10018528252840042,
456
+ "learning_rate": 0.0005753580835257988,
457
+ "loss": 0.1892,
458
+ "step": 1600
459
+ },
460
+ {
461
+ "epoch": 1.3,
462
+ "grad_norm": 0.10300764441490173,
463
+ "learning_rate": 0.0005471295302425277,
464
+ "loss": 0.1902,
465
+ "step": 1625
466
+ },
467
+ {
468
+ "epoch": 1.32,
469
+ "grad_norm": 0.11492912471294403,
470
+ "learning_rate": 0.0005193479054371922,
471
+ "loss": 0.185,
472
+ "step": 1650
473
+ },
474
+ {
475
+ "epoch": 1.34,
476
+ "grad_norm": 0.08382758498191833,
477
+ "learning_rate": 0.0004920406262193602,
478
+ "loss": 0.1888,
479
+ "step": 1675
480
+ },
481
+ {
482
+ "epoch": 1.3599999999999999,
483
+ "grad_norm": 0.09236115962266922,
484
+ "learning_rate": 0.00046523464157677074,
485
+ "loss": 0.1877,
486
+ "step": 1700
487
+ },
488
+ {
489
+ "epoch": 1.38,
490
+ "grad_norm": 0.12412581592798233,
491
+ "learning_rate": 0.00043895640577994134,
492
+ "loss": 0.1881,
493
+ "step": 1725
494
+ },
495
+ {
496
+ "epoch": 1.4,
497
+ "grad_norm": 0.1180960163474083,
498
+ "learning_rate": 0.00041323185227499273,
499
+ "loss": 0.1884,
500
+ "step": 1750
501
+ },
502
+ {
503
+ "epoch": 1.42,
504
+ "grad_norm": 0.10159043967723846,
505
+ "learning_rate": 0.00038808636809047205,
506
+ "loss": 0.1883,
507
+ "step": 1775
508
+ },
509
+ {
510
+ "epoch": 1.44,
511
+ "grad_norm": 0.11169745773077011,
512
+ "learning_rate": 0.00036354476878341766,
513
+ "loss": 0.1781,
514
+ "step": 1800
515
+ },
516
+ {
517
+ "epoch": 1.46,
518
+ "grad_norm": 0.10570152848958969,
519
+ "learning_rate": 0.00033963127394940316,
520
+ "loss": 0.1843,
521
+ "step": 1825
522
+ },
523
+ {
524
+ "epoch": 1.48,
525
+ "grad_norm": 0.12019435316324234,
526
+ "learning_rate": 0.00031636948332071813,
527
+ "loss": 0.1877,
528
+ "step": 1850
529
+ },
530
+ {
531
+ "epoch": 1.5,
532
+ "grad_norm": 0.10546473413705826,
533
+ "learning_rate": 0.00029378235347628267,
534
+ "loss": 0.1873,
535
+ "step": 1875
536
+ },
537
+ {
538
+ "epoch": 1.52,
539
+ "grad_norm": 0.11661262810230255,
540
+ "learning_rate": 0.00027189217518627665,
541
+ "loss": 0.186,
542
+ "step": 1900
543
+ },
544
+ {
545
+ "epoch": 1.54,
546
+ "grad_norm": 0.10733507573604584,
547
+ "learning_rate": 0.00025072055141383876,
548
+ "loss": 0.1844,
549
+ "step": 1925
550
+ },
551
+ {
552
+ "epoch": 1.56,
553
+ "grad_norm": 0.09054522961378098,
554
+ "learning_rate": 0.0002302883759955511,
555
+ "loss": 0.174,
556
+ "step": 1950
557
+ },
558
+ {
559
+ "epoch": 1.58,
560
+ "grad_norm": 0.11164789646863937,
561
+ "learning_rate": 0.0002106158130217456,
562
+ "loss": 0.1871,
563
+ "step": 1975
564
+ },
565
+ {
566
+ "epoch": 1.6,
567
+ "grad_norm": 0.09507589787244797,
568
+ "learning_rate": 0.00019172227693697963,
569
+ "loss": 0.1849,
570
+ "step": 2000
571
+ },
572
+ {
573
+ "epoch": 1.62,
574
+ "grad_norm": 0.09963341057300568,
575
+ "learning_rate": 0.0001736264133803246,
576
+ "loss": 0.1739,
577
+ "step": 2025
578
+ },
579
+ {
580
+ "epoch": 1.6400000000000001,
581
+ "grad_norm": 0.10520156472921371,
582
+ "learning_rate": 0.0001563460807843712,
583
+ "loss": 0.1823,
584
+ "step": 2050
585
+ },
586
+ {
587
+ "epoch": 1.6600000000000001,
588
+ "grad_norm": 0.09740244597196579,
589
+ "learning_rate": 0.00013989833275111142,
590
+ "loss": 0.1859,
591
+ "step": 2075
592
+ },
593
+ {
594
+ "epoch": 1.6800000000000002,
595
+ "grad_norm": 0.10571368783712387,
596
+ "learning_rate": 0.0001242994012220937,
597
+ "loss": 0.1832,
598
+ "step": 2100
599
+ },
600
+ {
601
+ "epoch": 1.7,
602
+ "grad_norm": 0.09903980791568756,
603
+ "learning_rate": 0.00010956468045945511,
604
+ "loss": 0.1811,
605
+ "step": 2125
606
+ },
607
+ {
608
+ "epoch": 1.72,
609
+ "grad_norm": 0.09760060906410217,
610
+ "learning_rate": 9.570871185364438e-05,
611
+ "loss": 0.1743,
612
+ "step": 2150
613
+ },
614
+ {
615
+ "epoch": 1.74,
616
+ "grad_norm": 0.09394048899412155,
617
+ "learning_rate": 8.274516957282352e-05,
618
+ "loss": 0.1907,
619
+ "step": 2175
620
+ },
621
+ {
622
+ "epoch": 1.76,
623
+ "grad_norm": 0.10015013068914413,
624
+ "learning_rate": 7.068684706811668e-05,
625
+ "loss": 0.1809,
626
+ "step": 2200
627
+ },
628
+ {
629
+ "epoch": 1.78,
630
+ "grad_norm": 0.10288713872432709,
631
+ "learning_rate": 5.9545644448017025e-05,
632
+ "loss": 0.1753,
633
+ "step": 2225
634
+ },
635
+ {
636
+ "epoch": 1.8,
637
+ "grad_norm": 0.1012328565120697,
638
+ "learning_rate": 4.9332556734416836e-05,
639
+ "loss": 0.1848,
640
+ "step": 2250
641
+ },
642
+ {
643
+ "epoch": 1.8199999999999998,
644
+ "grad_norm": 0.10257886350154877,
645
+ "learning_rate": 4.0057663011848324e-05,
646
+ "loss": 0.1764,
647
+ "step": 2275
648
+ },
649
+ {
650
+ "epoch": 1.8399999999999999,
651
+ "grad_norm": 0.11384698003530502,
652
+ "learning_rate": 3.173011648064428e-05,
653
+ "loss": 0.1797,
654
+ "step": 2300
655
+ },
656
+ {
657
+ "epoch": 1.8599999999999999,
658
+ "grad_norm": 0.10323520004749298,
659
+ "learning_rate": 2.435813542383403e-05,
660
+ "loss": 0.1828,
661
+ "step": 2325
662
+ },
663
+ {
664
+ "epoch": 1.88,
665
+ "grad_norm": 0.10576365888118744,
666
+ "learning_rate": 1.794899509669079e-05,
667
+ "loss": 0.1738,
668
+ "step": 2350
669
+ },
670
+ {
671
+ "epoch": 1.9,
672
+ "grad_norm": 0.12268764525651932,
673
+ "learning_rate": 1.250902054693237e-05,
674
+ "loss": 0.1956,
675
+ "step": 2375
676
+ },
677
+ {
678
+ "epoch": 1.92,
679
+ "grad_norm": 0.09913565963506699,
680
+ "learning_rate": 8.043580372663506e-06,
681
+ "loss": 0.1835,
682
+ "step": 2400
683
+ },
684
+ {
685
+ "epoch": 1.94,
686
+ "grad_norm": 0.12227698415517807,
687
+ "learning_rate": 4.557081424216847e-06,
688
+ "loss": 0.1767,
689
+ "step": 2425
690
+ },
691
+ {
692
+ "epoch": 1.96,
693
+ "grad_norm": 0.10373125225305557,
694
+ "learning_rate": 2.052964455124484e-06,
695
+ "loss": 0.1786,
696
+ "step": 2450
697
+ },
698
+ {
699
+ "epoch": 1.98,
700
+ "grad_norm": 0.10990923643112183,
701
+ "learning_rate": 5.337007265091564e-07,
702
+ "loss": 0.1839,
703
+ "step": 2475
704
+ },
705
+ {
706
+ "epoch": 2.0,
707
+ "grad_norm": 0.0938950777053833,
708
+ "learning_rate": 7.895682482272904e-10,
709
+ "loss": 0.1806,
710
+ "step": 2500
711
+ },
712
+ {
713
+ "epoch": 2.0,
714
+ "step": 2500,
715
+ "total_flos": 1.62588235137024e+18,
716
+ "train_loss": 0.231773907661438,
717
+ "train_runtime": 2194.5903,
718
+ "train_samples_per_second": 36.453,
719
+ "train_steps_per_second": 1.139
720
+ }
721
+ ],
722
+ "logging_steps": 25,
723
+ "max_steps": 2500,
724
+ "num_input_tokens_seen": 0,
725
+ "num_train_epochs": 2,
726
+ "save_steps": 0,
727
+ "stateful_callbacks": {
728
+ "TrainerControl": {
729
+ "args": {
730
+ "should_epoch_stop": false,
731
+ "should_evaluate": false,
732
+ "should_log": false,
733
+ "should_save": false,
734
+ "should_training_stop": false
735
+ },
736
+ "attributes": {}
737
+ }
738
+ },
739
+ "total_flos": 1.62588235137024e+18,
740
+ "train_batch_size": 32,
741
+ "trial_name": null,
742
+ "trial_params": null
743
+ }
nl_tasks/exprep/run_ex29_2ep/ft/adapter_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "drop_out": 0.0,
6
+ "inference_mode": false,
7
+ "layers_to_transform": null,
8
+ "modules_to_save": null,
9
+ "num_rotations": 1,
10
+ "peft_type": "ROTATION",
11
+ "r": 16,
12
+ "revision": null,
13
+ "target_modules": [
14
+ "v_proj",
15
+ "q_proj"
16
+ ],
17
+ "target_modules_to_skip": null,
18
+ "task_type": "CAUSAL_LM"
19
+ }
nl_tasks/exprep/run_ex29_2ep/ft/special_tokens_map.json ADDED
@@ -0,0 +1,24 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<s>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "</s>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": "<unk>",
17
+ "unk_token": {
18
+ "content": "<unk>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ }
24
+ }
nl_tasks/exprep/run_ex29_2ep/ft/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
nl_tasks/exprep/run_ex29_2ep/ft/tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
3
+ size 499723
nl_tasks/exprep/run_ex29_2ep/ft/tokenizer_config.json ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": true,
3
+ "add_eos_token": false,
4
+ "add_prefix_space": null,
5
+ "added_tokens_decoder": {
6
+ "0": {
7
+ "content": "<unk>",
8
+ "lstrip": false,
9
+ "normalized": false,
10
+ "rstrip": false,
11
+ "single_word": false,
12
+ "special": true
13
+ },
14
+ "1": {
15
+ "content": "<s>",
16
+ "lstrip": false,
17
+ "normalized": false,
18
+ "rstrip": false,
19
+ "single_word": false,
20
+ "special": true
21
+ },
22
+ "2": {
23
+ "content": "</s>",
24
+ "lstrip": false,
25
+ "normalized": false,
26
+ "rstrip": false,
27
+ "single_word": false,
28
+ "special": true
29
+ }
30
+ },
31
+ "bos_token": "<s>",
32
+ "clean_up_tokenization_spaces": false,
33
+ "eos_token": "</s>",
34
+ "extra_special_tokens": {},
35
+ "legacy": false,
36
+ "model_max_length": 512,
37
+ "pad_token": "<unk>",
38
+ "padding_side": "right",
39
+ "sp_model_kwargs": {},
40
+ "tokenizer_class": "LlamaTokenizer",
41
+ "unk_token": "<unk>",
42
+ "use_default_system_prompt": false
43
+ }
nl_tasks/exprep/run_ex29_2ep/ft2/adapter_config.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "drop_out": 0.0,
6
+ "inference_mode": true,
7
+ "layers_to_transform": null,
8
+ "modules_to_save": null,
9
+ "num_rotations": 1,
10
+ "peft_type": "ROTATION",
11
+ "r": 16,
12
+ "revision": null,
13
+ "target_modules": [
14
+ "v_proj",
15
+ "q_proj"
16
+ ],
17
+ "target_modules_to_skip": null,
18
+ "task_type": "CAUSAL_LM"
19
+ }
nl_tasks/exprep/run_ex29_2ep/ft2/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5322da057fb194de004bf09f6c1c5ce09a8486879bd122bbb0073fdc2507d1d9
3
+ size 33602915
nl_tasks/exprep/run_ex29_2ep/output.txt ADDED
@@ -0,0 +1,4 @@
 
 
 
 
 
1
+
2
+ MATH math MAX TOKEN = 1408, length==== 5000, math acc %====, 7.3999999999999995
3
+
4
+ gsm8k MAX TOKEN = 1024, length==== 1319, gsm8k acc %====, 48.14253222137984
nl_tasks/exprep/run_ex29_2ep/trainer_state.json ADDED
@@ -0,0 +1,743 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 2.0,
6
+ "eval_steps": 100,
7
+ "global_step": 2500,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.02,
14
+ "grad_norm": 3.0515835285186768,
15
+ "learning_rate": 0.0019995452431006843,
16
+ "loss": 6.1405,
17
+ "step": 25
18
+ },
19
+ {
20
+ "epoch": 0.04,
21
+ "grad_norm": 0.34508204460144043,
22
+ "learning_rate": 0.0019981048452912364,
23
+ "loss": 2.2112,
24
+ "step": 50
25
+ },
26
+ {
27
+ "epoch": 0.06,
28
+ "grad_norm": 0.3236846923828125,
29
+ "learning_rate": 0.001995679438502144,
30
+ "loss": 0.3908,
31
+ "step": 75
32
+ },
33
+ {
34
+ "epoch": 0.08,
35
+ "grad_norm": 0.2487092912197113,
36
+ "learning_rate": 0.0019922714163170857,
37
+ "loss": 0.3499,
38
+ "step": 100
39
+ },
40
+ {
41
+ "epoch": 0.1,
42
+ "grad_norm": 0.2623882293701172,
43
+ "learning_rate": 0.0019878841420425024,
44
+ "loss": 0.3354,
45
+ "step": 125
46
+ },
47
+ {
48
+ "epoch": 0.12,
49
+ "grad_norm": 0.23623116314411163,
50
+ "learning_rate": 0.001982521945388421,
51
+ "loss": 0.3235,
52
+ "step": 150
53
+ },
54
+ {
55
+ "epoch": 0.14,
56
+ "grad_norm": 0.19215604662895203,
57
+ "learning_rate": 0.001976190118195553,
58
+ "loss": 0.3197,
59
+ "step": 175
60
+ },
61
+ {
62
+ "epoch": 0.16,
63
+ "grad_norm": 0.22971579432487488,
64
+ "learning_rate": 0.001968894909212887,
65
+ "loss": 0.3166,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.18,
70
+ "grad_norm": 0.22463415563106537,
71
+ "learning_rate": 0.001960643517930928,
72
+ "loss": 0.308,
73
+ "step": 225
74
+ },
75
+ {
76
+ "epoch": 0.2,
77
+ "grad_norm": 0.21010400354862213,
78
+ "learning_rate": 0.001951444087476669,
79
+ "loss": 0.3036,
80
+ "step": 250
81
+ },
82
+ {
83
+ "epoch": 0.22,
84
+ "grad_norm": 0.2055121213197708,
85
+ "learning_rate": 0.0019413056965773068,
86
+ "loss": 0.3,
87
+ "step": 275
88
+ },
89
+ {
90
+ "epoch": 0.24,
91
+ "grad_norm": 0.22161348164081573,
92
+ "learning_rate": 0.0019302383506006372,
93
+ "loss": 0.3085,
94
+ "step": 300
95
+ },
96
+ {
97
+ "epoch": 0.26,
98
+ "grad_norm": 0.1612371802330017,
99
+ "learning_rate": 0.0019182529716809619,
100
+ "loss": 0.2952,
101
+ "step": 325
102
+ },
103
+ {
104
+ "epoch": 0.28,
105
+ "grad_norm": 0.17721955478191376,
106
+ "learning_rate": 0.001905361387940262,
107
+ "loss": 0.2971,
108
+ "step": 350
109
+ },
110
+ {
111
+ "epoch": 0.3,
112
+ "grad_norm": 0.1718919575214386,
113
+ "learning_rate": 0.0018915763218152702,
114
+ "loss": 0.2887,
115
+ "step": 375
116
+ },
117
+ {
118
+ "epoch": 0.32,
119
+ "grad_norm": 0.20089927315711975,
120
+ "learning_rate": 0.0018769113775019605,
121
+ "loss": 0.2882,
122
+ "step": 400
123
+ },
124
+ {
125
+ "epoch": 0.34,
126
+ "grad_norm": 0.1430523544549942,
127
+ "learning_rate": 0.00186138102752985,
128
+ "loss": 0.2762,
129
+ "step": 425
130
+ },
131
+ {
132
+ "epoch": 0.36,
133
+ "grad_norm": 0.16578099131584167,
134
+ "learning_rate": 0.0018450005984793592,
135
+ "loss": 0.2781,
136
+ "step": 450
137
+ },
138
+ {
139
+ "epoch": 0.38,
140
+ "grad_norm": 0.1863383799791336,
141
+ "learning_rate": 0.001827786255856328,
142
+ "loss": 0.2741,
143
+ "step": 475
144
+ },
145
+ {
146
+ "epoch": 0.4,
147
+ "grad_norm": 0.17521248757839203,
148
+ "learning_rate": 0.0018097549881386124,
149
+ "loss": 0.271,
150
+ "step": 500
151
+ },
152
+ {
153
+ "epoch": 0.42,
154
+ "grad_norm": 0.14374853670597076,
155
+ "learning_rate": 0.0017909245900105085,
156
+ "loss": 0.2725,
157
+ "step": 525
158
+ },
159
+ {
160
+ "epoch": 0.44,
161
+ "grad_norm": 0.16397333145141602,
162
+ "learning_rate": 0.001771313644801547,
163
+ "loss": 0.2598,
164
+ "step": 550
165
+ },
166
+ {
167
+ "epoch": 0.46,
168
+ "grad_norm": 0.15914176404476166,
169
+ "learning_rate": 0.0017509415061469915,
170
+ "loss": 0.277,
171
+ "step": 575
172
+ },
173
+ {
174
+ "epoch": 0.48,
175
+ "grad_norm": 0.16967792809009552,
176
+ "learning_rate": 0.0017298282788881353,
177
+ "loss": 0.262,
178
+ "step": 600
179
+ },
180
+ {
181
+ "epoch": 0.5,
182
+ "grad_norm": 0.18250292539596558,
183
+ "learning_rate": 0.0017079947992312528,
184
+ "loss": 0.2761,
185
+ "step": 625
186
+ },
187
+ {
188
+ "epoch": 0.52,
189
+ "grad_norm": 0.1378776878118515,
190
+ "learning_rate": 0.0016854626141847768,
191
+ "loss": 0.2657,
192
+ "step": 650
193
+ },
194
+ {
195
+ "epoch": 0.54,
196
+ "grad_norm": 0.12245441973209381,
197
+ "learning_rate": 0.001662253960295005,
198
+ "loss": 0.2561,
199
+ "step": 675
200
+ },
201
+ {
202
+ "epoch": 0.56,
203
+ "grad_norm": 0.15305551886558533,
204
+ "learning_rate": 0.0016383917417013114,
205
+ "loss": 0.2558,
206
+ "step": 700
207
+ },
208
+ {
209
+ "epoch": 0.58,
210
+ "grad_norm": 0.16352495551109314,
211
+ "learning_rate": 0.0016138995075325275,
212
+ "loss": 0.2625,
213
+ "step": 725
214
+ },
215
+ {
216
+ "epoch": 0.6,
217
+ "grad_norm": 0.14869992434978485,
218
+ "learning_rate": 0.0015888014286667951,
219
+ "loss": 0.2592,
220
+ "step": 750
221
+ },
222
+ {
223
+ "epoch": 0.62,
224
+ "grad_norm": 0.15682505071163177,
225
+ "learning_rate": 0.0015631222738778266,
226
+ "loss": 0.2466,
227
+ "step": 775
228
+ },
229
+ {
230
+ "epoch": 0.64,
231
+ "grad_norm": 0.15803146362304688,
232
+ "learning_rate": 0.0015368873853911164,
233
+ "loss": 0.2537,
234
+ "step": 800
235
+ },
236
+ {
237
+ "epoch": 0.66,
238
+ "grad_norm": 0.14613056182861328,
239
+ "learning_rate": 0.0015101226538742248,
240
+ "loss": 0.245,
241
+ "step": 825
242
+ },
243
+ {
244
+ "epoch": 0.68,
245
+ "grad_norm": 0.1405636966228485,
246
+ "learning_rate": 0.0014828544928858136,
247
+ "loss": 0.2478,
248
+ "step": 850
249
+ },
250
+ {
251
+ "epoch": 0.7,
252
+ "grad_norm": 0.1207919493317604,
253
+ "learning_rate": 0.0014551098128086537,
254
+ "loss": 0.254,
255
+ "step": 875
256
+ },
257
+ {
258
+ "epoch": 0.72,
259
+ "grad_norm": 0.13489678502082825,
260
+ "learning_rate": 0.0014269159942923253,
261
+ "loss": 0.2493,
262
+ "step": 900
263
+ },
264
+ {
265
+ "epoch": 0.74,
266
+ "grad_norm": 0.14353223145008087,
267
+ "learning_rate": 0.001398300861231823,
268
+ "loss": 0.2465,
269
+ "step": 925
270
+ },
271
+ {
272
+ "epoch": 0.76,
273
+ "grad_norm": 0.13418437540531158,
274
+ "learning_rate": 0.0013692926533087305,
275
+ "loss": 0.2442,
276
+ "step": 950
277
+ },
278
+ {
279
+ "epoch": 0.78,
280
+ "grad_norm": 0.12505637109279633,
281
+ "learning_rate": 0.0013399199981220646,
282
+ "loss": 0.2411,
283
+ "step": 975
284
+ },
285
+ {
286
+ "epoch": 0.8,
287
+ "grad_norm": 0.15944132208824158,
288
+ "learning_rate": 0.001310211882936291,
289
+ "loss": 0.2481,
290
+ "step": 1000
291
+ },
292
+ {
293
+ "epoch": 0.82,
294
+ "grad_norm": 0.16288740932941437,
295
+ "learning_rate": 0.0012801976260743938,
296
+ "loss": 0.2466,
297
+ "step": 1025
298
+ },
299
+ {
300
+ "epoch": 0.84,
301
+ "grad_norm": 0.13887740671634674,
302
+ "learning_rate": 0.0012499068479842306,
303
+ "loss": 0.2361,
304
+ "step": 1050
305
+ },
306
+ {
307
+ "epoch": 0.86,
308
+ "grad_norm": 0.1796649843454361,
309
+ "learning_rate": 0.001219369442006726,
310
+ "loss": 0.2324,
311
+ "step": 1075
312
+ },
313
+ {
314
+ "epoch": 0.88,
315
+ "grad_norm": 0.1486123949289322,
316
+ "learning_rate": 0.001188615544874755,
317
+ "loss": 0.2337,
318
+ "step": 1100
319
+ },
320
+ {
321
+ "epoch": 0.9,
322
+ "grad_norm": 0.1394972950220108,
323
+ "learning_rate": 0.0011576755069718227,
324
+ "loss": 0.2429,
325
+ "step": 1125
326
+ },
327
+ {
328
+ "epoch": 0.92,
329
+ "grad_norm": 0.21638131141662598,
330
+ "learning_rate": 0.0011265798623799042,
331
+ "loss": 0.2356,
332
+ "step": 1150
333
+ },
334
+ {
335
+ "epoch": 0.94,
336
+ "grad_norm": 0.15221650898456573,
337
+ "learning_rate": 0.0010953592987459886,
338
+ "loss": 0.2397,
339
+ "step": 1175
340
+ },
341
+ {
342
+ "epoch": 0.96,
343
+ "grad_norm": 0.1171698272228241,
344
+ "learning_rate": 0.0010640446269970785,
345
+ "loss": 0.2372,
346
+ "step": 1200
347
+ },
348
+ {
349
+ "epoch": 0.98,
350
+ "grad_norm": 0.1338423192501068,
351
+ "learning_rate": 0.001032666750933524,
352
+ "loss": 0.2383,
353
+ "step": 1225
354
+ },
355
+ {
356
+ "epoch": 1.0,
357
+ "grad_norm": 0.1542067527770996,
358
+ "learning_rate": 0.0010012566367307025,
359
+ "loss": 0.2291,
360
+ "step": 1250
361
+ },
362
+ {
363
+ "epoch": 1.02,
364
+ "grad_norm": 0.13059671223163605,
365
+ "learning_rate": 0.0009698452823791432,
366
+ "loss": 0.1972,
367
+ "step": 1275
368
+ },
369
+ {
370
+ "epoch": 1.04,
371
+ "grad_norm": 0.10279057919979095,
372
+ "learning_rate": 0.000938463687093253,
373
+ "loss": 0.2048,
374
+ "step": 1300
375
+ },
376
+ {
377
+ "epoch": 1.06,
378
+ "grad_norm": 0.17310836911201477,
379
+ "learning_rate": 0.0009071428207188324,
380
+ "loss": 0.1992,
381
+ "step": 1325
382
+ },
383
+ {
384
+ "epoch": 1.08,
385
+ "grad_norm": 0.1401149034500122,
386
+ "learning_rate": 0.0008759135931695792,
387
+ "loss": 0.205,
388
+ "step": 1350
389
+ },
390
+ {
391
+ "epoch": 1.1,
392
+ "grad_norm": 0.10482831299304962,
393
+ "learning_rate": 0.000844806823922734,
394
+ "loss": 0.2084,
395
+ "step": 1375
396
+ },
397
+ {
398
+ "epoch": 1.12,
399
+ "grad_norm": 0.12018042802810669,
400
+ "learning_rate": 0.0008138532116039781,
401
+ "loss": 0.2065,
402
+ "step": 1400
403
+ },
404
+ {
405
+ "epoch": 1.1400000000000001,
406
+ "grad_norm": 0.1350913643836975,
407
+ "learning_rate": 0.0007830833036915947,
408
+ "loss": 0.1979,
409
+ "step": 1425
410
+ },
411
+ {
412
+ "epoch": 1.16,
413
+ "grad_norm": 0.12790362536907196,
414
+ "learning_rate": 0.0007525274663697977,
415
+ "loss": 0.206,
416
+ "step": 1450
417
+ },
418
+ {
419
+ "epoch": 1.18,
420
+ "grad_norm": 0.11059988290071487,
421
+ "learning_rate": 0.0007222158545609727,
422
+ "loss": 0.2017,
423
+ "step": 1475
424
+ },
425
+ {
426
+ "epoch": 1.2,
427
+ "grad_norm": 0.13595543801784515,
428
+ "learning_rate": 0.0006921783821664097,
429
+ "loss": 0.2025,
430
+ "step": 1500
431
+ },
432
+ {
433
+ "epoch": 1.22,
434
+ "grad_norm": 0.12031928449869156,
435
+ "learning_rate": 0.0006624446925448944,
436
+ "loss": 0.205,
437
+ "step": 1525
438
+ },
439
+ {
440
+ "epoch": 1.24,
441
+ "grad_norm": 0.10558769851922989,
442
+ "learning_rate": 0.0006330441292582908,
443
+ "loss": 0.1972,
444
+ "step": 1550
445
+ },
446
+ {
447
+ "epoch": 1.26,
448
+ "grad_norm": 0.11100303381681442,
449
+ "learning_rate": 0.00060400570711299,
450
+ "loss": 0.1976,
451
+ "step": 1575
452
+ },
453
+ {
454
+ "epoch": 1.28,
455
+ "grad_norm": 0.11664096266031265,
456
+ "learning_rate": 0.0005753580835257988,
457
+ "loss": 0.1926,
458
+ "step": 1600
459
+ },
460
+ {
461
+ "epoch": 1.3,
462
+ "grad_norm": 0.13128218054771423,
463
+ "learning_rate": 0.0005471295302425277,
464
+ "loss": 0.1926,
465
+ "step": 1625
466
+ },
467
+ {
468
+ "epoch": 1.32,
469
+ "grad_norm": 0.1304730474948883,
470
+ "learning_rate": 0.0005193479054371922,
471
+ "loss": 0.1872,
472
+ "step": 1650
473
+ },
474
+ {
475
+ "epoch": 1.34,
476
+ "grad_norm": 0.09425242990255356,
477
+ "learning_rate": 0.0004920406262193602,
478
+ "loss": 0.1914,
479
+ "step": 1675
480
+ },
481
+ {
482
+ "epoch": 1.3599999999999999,
483
+ "grad_norm": 0.11022331565618515,
484
+ "learning_rate": 0.00046523464157677074,
485
+ "loss": 0.1902,
486
+ "step": 1700
487
+ },
488
+ {
489
+ "epoch": 1.38,
490
+ "grad_norm": 0.11479215323925018,
491
+ "learning_rate": 0.00043895640577994134,
492
+ "loss": 0.1908,
493
+ "step": 1725
494
+ },
495
+ {
496
+ "epoch": 1.4,
497
+ "grad_norm": 0.13587336242198944,
498
+ "learning_rate": 0.00041323185227499273,
499
+ "loss": 0.1906,
500
+ "step": 1750
501
+ },
502
+ {
503
+ "epoch": 1.42,
504
+ "grad_norm": 0.10993529111146927,
505
+ "learning_rate": 0.00038808636809047205,
506
+ "loss": 0.1909,
507
+ "step": 1775
508
+ },
509
+ {
510
+ "epoch": 1.44,
511
+ "grad_norm": 0.15703265368938446,
512
+ "learning_rate": 0.00036354476878341766,
513
+ "loss": 0.1796,
514
+ "step": 1800
515
+ },
516
+ {
517
+ "epoch": 1.46,
518
+ "grad_norm": 0.10287538170814514,
519
+ "learning_rate": 0.00033963127394940316,
520
+ "loss": 0.1861,
521
+ "step": 1825
522
+ },
523
+ {
524
+ "epoch": 1.48,
525
+ "grad_norm": 0.13588131964206696,
526
+ "learning_rate": 0.00031636948332071813,
527
+ "loss": 0.1904,
528
+ "step": 1850
529
+ },
530
+ {
531
+ "epoch": 1.5,
532
+ "grad_norm": 0.11725416779518127,
533
+ "learning_rate": 0.00029378235347628267,
534
+ "loss": 0.1914,
535
+ "step": 1875
536
+ },
537
+ {
538
+ "epoch": 1.52,
539
+ "grad_norm": 0.12543147802352905,
540
+ "learning_rate": 0.00027189217518627665,
541
+ "loss": 0.1876,
542
+ "step": 1900
543
+ },
544
+ {
545
+ "epoch": 1.54,
546
+ "grad_norm": 0.12276057153940201,
547
+ "learning_rate": 0.00025072055141383876,
548
+ "loss": 0.1871,
549
+ "step": 1925
550
+ },
551
+ {
552
+ "epoch": 1.56,
553
+ "grad_norm": 0.10937570035457611,
554
+ "learning_rate": 0.0002302883759955511,
555
+ "loss": 0.1784,
556
+ "step": 1950
557
+ },
558
+ {
559
+ "epoch": 1.58,
560
+ "grad_norm": 0.11975866556167603,
561
+ "learning_rate": 0.0002106158130217456,
562
+ "loss": 0.1895,
563
+ "step": 1975
564
+ },
565
+ {
566
+ "epoch": 1.6,
567
+ "grad_norm": 0.11259443312883377,
568
+ "learning_rate": 0.00019172227693697963,
569
+ "loss": 0.1877,
570
+ "step": 2000
571
+ },
572
+ {
573
+ "epoch": 1.62,
574
+ "grad_norm": 0.10356508195400238,
575
+ "learning_rate": 0.0001736264133803246,
576
+ "loss": 0.1759,
577
+ "step": 2025
578
+ },
579
+ {
580
+ "epoch": 1.6400000000000001,
581
+ "grad_norm": 0.11048847436904907,
582
+ "learning_rate": 0.0001563460807843712,
583
+ "loss": 0.1842,
584
+ "step": 2050
585
+ },
586
+ {
587
+ "epoch": 1.6600000000000001,
588
+ "grad_norm": 0.10486841946840286,
589
+ "learning_rate": 0.00013989833275111142,
590
+ "loss": 0.1885,
591
+ "step": 2075
592
+ },
593
+ {
594
+ "epoch": 1.6800000000000002,
595
+ "grad_norm": 0.11607607454061508,
596
+ "learning_rate": 0.0001242994012220937,
597
+ "loss": 0.1861,
598
+ "step": 2100
599
+ },
600
+ {
601
+ "epoch": 1.7,
602
+ "grad_norm": 0.10461049526929855,
603
+ "learning_rate": 0.00010956468045945511,
604
+ "loss": 0.1828,
605
+ "step": 2125
606
+ },
607
+ {
608
+ "epoch": 1.72,
609
+ "grad_norm": 0.12405046820640564,
610
+ "learning_rate": 9.570871185364438e-05,
611
+ "loss": 0.1768,
612
+ "step": 2150
613
+ },
614
+ {
615
+ "epoch": 1.74,
616
+ "grad_norm": 0.10572119057178497,
617
+ "learning_rate": 8.274516957282352e-05,
618
+ "loss": 0.1925,
619
+ "step": 2175
620
+ },
621
+ {
622
+ "epoch": 1.76,
623
+ "grad_norm": 0.11157557368278503,
624
+ "learning_rate": 7.068684706811668e-05,
625
+ "loss": 0.1836,
626
+ "step": 2200
627
+ },
628
+ {
629
+ "epoch": 1.78,
630
+ "grad_norm": 0.11235346645116806,
631
+ "learning_rate": 5.9545644448017025e-05,
632
+ "loss": 0.1774,
633
+ "step": 2225
634
+ },
635
+ {
636
+ "epoch": 1.8,
637
+ "grad_norm": 0.10231998562812805,
638
+ "learning_rate": 4.9332556734416836e-05,
639
+ "loss": 0.1856,
640
+ "step": 2250
641
+ },
642
+ {
643
+ "epoch": 1.8199999999999998,
644
+ "grad_norm": 0.10948530584573746,
645
+ "learning_rate": 4.0057663011848324e-05,
646
+ "loss": 0.1778,
647
+ "step": 2275
648
+ },
649
+ {
650
+ "epoch": 1.8399999999999999,
651
+ "grad_norm": 0.12041065096855164,
652
+ "learning_rate": 3.173011648064428e-05,
653
+ "loss": 0.1815,
654
+ "step": 2300
655
+ },
656
+ {
657
+ "epoch": 1.8599999999999999,
658
+ "grad_norm": 0.11477901041507721,
659
+ "learning_rate": 2.435813542383403e-05,
660
+ "loss": 0.1845,
661
+ "step": 2325
662
+ },
663
+ {
664
+ "epoch": 1.88,
665
+ "grad_norm": 0.11077318340539932,
666
+ "learning_rate": 1.794899509669079e-05,
667
+ "loss": 0.1766,
668
+ "step": 2350
669
+ },
670
+ {
671
+ "epoch": 1.9,
672
+ "grad_norm": 0.12476975470781326,
673
+ "learning_rate": 1.250902054693237e-05,
674
+ "loss": 0.1992,
675
+ "step": 2375
676
+ },
677
+ {
678
+ "epoch": 1.92,
679
+ "grad_norm": 0.13025566935539246,
680
+ "learning_rate": 8.043580372663506e-06,
681
+ "loss": 0.1848,
682
+ "step": 2400
683
+ },
684
+ {
685
+ "epoch": 1.94,
686
+ "grad_norm": 0.1307641863822937,
687
+ "learning_rate": 4.557081424216847e-06,
688
+ "loss": 0.1784,
689
+ "step": 2425
690
+ },
691
+ {
692
+ "epoch": 1.96,
693
+ "grad_norm": 0.10733317583799362,
694
+ "learning_rate": 2.052964455124484e-06,
695
+ "loss": 0.1803,
696
+ "step": 2450
697
+ },
698
+ {
699
+ "epoch": 1.98,
700
+ "grad_norm": 0.11906908452510834,
701
+ "learning_rate": 5.337007265091564e-07,
702
+ "loss": 0.1856,
703
+ "step": 2475
704
+ },
705
+ {
706
+ "epoch": 2.0,
707
+ "grad_norm": 0.1013617143034935,
708
+ "learning_rate": 7.895682482272904e-10,
709
+ "loss": 0.1829,
710
+ "step": 2500
711
+ },
712
+ {
713
+ "epoch": 2.0,
714
+ "step": 2500,
715
+ "total_flos": 1.62588235137024e+18,
716
+ "train_loss": 0.3086915157318115,
717
+ "train_runtime": 2194.1953,
718
+ "train_samples_per_second": 36.46,
719
+ "train_steps_per_second": 1.139
720
+ }
721
+ ],
722
+ "logging_steps": 25,
723
+ "max_steps": 2500,
724
+ "num_input_tokens_seen": 0,
725
+ "num_train_epochs": 2,
726
+ "save_steps": 0,
727
+ "stateful_callbacks": {
728
+ "TrainerControl": {
729
+ "args": {
730
+ "should_epoch_stop": false,
731
+ "should_evaluate": false,
732
+ "should_log": false,
733
+ "should_save": false,
734
+ "should_training_stop": false
735
+ },
736
+ "attributes": {}
737
+ }
738
+ },
739
+ "total_flos": 1.62588235137024e+18,
740
+ "train_batch_size": 32,
741
+ "trial_name": null,
742
+ "trial_params": null
743
+ }