nvan15 commited on
Commit
f750623
·
verified ·
1 Parent(s): 6c9152c

Batch upload part 6

Browse files
Files changed (50) hide show
  1. nl_tasks/exps/run_ex17_3ep/ft/adapter_config.json +18 -0
  2. nl_tasks/exps/run_ex17_3ep/ft/added_tokens.json +3 -0
  3. nl_tasks/exps/run_ex17_3ep/ft/special_tokens_map.json +30 -0
  4. nl_tasks/exps/run_ex17_3ep/ft/tokenizer.json +0 -0
  5. nl_tasks/exps/run_ex17_3ep/ft/tokenizer.model +3 -0
  6. nl_tasks/exps/run_ex17_3ep/ft/tokenizer_config.json +51 -0
  7. nl_tasks/exps/run_ex17_3ep/ft2/adapter_config.json +18 -0
  8. nl_tasks/exps/run_ex17_3ep/ft2/adapter_model.bin +3 -0
  9. nl_tasks/exps/run_ex17_3ep/trainer_state.json +1209 -0
  10. nl_tasks/exps/run_ex18_2ep/ft/adapter_config.json +18 -0
  11. nl_tasks/exps/run_ex18_2ep/ft/added_tokens.json +3 -0
  12. nl_tasks/exps/run_ex18_2ep/ft/special_tokens_map.json +30 -0
  13. nl_tasks/exps/run_ex18_2ep/ft/tokenizer.json +0 -0
  14. nl_tasks/exps/run_ex18_2ep/ft/tokenizer.model +3 -0
  15. nl_tasks/exps/run_ex18_2ep/ft/tokenizer_config.json +51 -0
  16. nl_tasks/exps/run_ex18_2ep/ft2/adapter_config.json +18 -0
  17. nl_tasks/exps/run_ex18_2ep/ft2/adapter_model.bin +3 -0
  18. nl_tasks/exps/run_ex18_2ep/trainer_state.json +818 -0
  19. nl_tasks/exps/run_ex19/ft/adapter_config.json +18 -0
  20. nl_tasks/exps/run_ex19/ft/special_tokens_map.json +24 -0
  21. nl_tasks/exps/run_ex19/ft/tokenizer.json +0 -0
  22. nl_tasks/exps/run_ex19/ft/tokenizer.model +3 -0
  23. nl_tasks/exps/run_ex19/ft/tokenizer_config.json +43 -0
  24. nl_tasks/exps/run_ex19/ft2/adapter_config.json +18 -0
  25. nl_tasks/exps/run_ex19/ft2/adapter_model.bin +3 -0
  26. nl_tasks/exps/run_ex19/trainer_state.json +1149 -0
  27. nl_tasks/exps/run_ex19_2ep/ft/adapter_config.json +18 -0
  28. nl_tasks/exps/run_ex19_2ep/ft/added_tokens.json +3 -0
  29. nl_tasks/exps/run_ex19_2ep/ft/special_tokens_map.json +30 -0
  30. nl_tasks/exps/run_ex19_2ep/ft/tokenizer.json +0 -0
  31. nl_tasks/exps/run_ex19_2ep/ft/tokenizer.model +3 -0
  32. nl_tasks/exps/run_ex19_2ep/ft/tokenizer_config.json +51 -0
  33. nl_tasks/exps/run_ex19_2ep/ft2/adapter_config.json +18 -0
  34. nl_tasks/exps/run_ex19_2ep/ft2/adapter_model.bin +3 -0
  35. nl_tasks/exps/run_ex19_2ep/trainer_state.json +818 -0
  36. nl_tasks/exps/run_ex20/ft/adapter_config.json +18 -0
  37. nl_tasks/exps/run_ex20/ft/special_tokens_map.json +24 -0
  38. nl_tasks/exps/run_ex20/ft/tokenizer.json +0 -0
  39. nl_tasks/exps/run_ex20/ft/tokenizer.model +3 -0
  40. nl_tasks/exps/run_ex20/ft/tokenizer_config.json +43 -0
  41. nl_tasks/exps/run_ex20/ft2/adapter_config.json +18 -0
  42. nl_tasks/exps/run_ex20/ft2/adapter_model.bin +3 -0
  43. nl_tasks/exps/run_ex20/trainer_state.json +1149 -0
  44. nl_tasks/exps/run_ex21/ft/adapter_config.json +18 -0
  45. nl_tasks/exps/run_ex21/ft/special_tokens_map.json +24 -0
  46. nl_tasks/exps/run_ex21/ft/tokenizer.json +0 -0
  47. nl_tasks/exps/run_ex21/ft/tokenizer.model +3 -0
  48. nl_tasks/exps/run_ex21/ft/tokenizer_config.json +43 -0
  49. nl_tasks/exps/run_ex21/ft2/adapter_model.bin +3 -0
  50. nl_tasks/exps/run_ex21/trainer_state.json +783 -0
nl_tasks/exps/run_ex17_3ep/ft/adapter_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "inference_mode": false,
6
+ "layers_to_transform": null,
7
+ "modules_to_save": null,
8
+ "num_rotations": 1,
9
+ "peft_type": "ROTATION",
10
+ "r": 16,
11
+ "revision": null,
12
+ "target_modules": [
13
+ "q_proj",
14
+ "v_proj"
15
+ ],
16
+ "target_modules_to_skip": null,
17
+ "task_type": "CAUSAL_LM"
18
+ }
nl_tasks/exps/run_ex17_3ep/ft/added_tokens.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ {
2
+ "[PAD]": 32000
3
+ }
nl_tasks/exps/run_ex17_3ep/ft/special_tokens_map.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": {
17
+ "content": "[PAD]",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ },
23
+ "unk_token": {
24
+ "content": "</s>",
25
+ "lstrip": false,
26
+ "normalized": false,
27
+ "rstrip": false,
28
+ "single_word": false
29
+ }
30
+ }
nl_tasks/exps/run_ex17_3ep/ft/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
nl_tasks/exps/run_ex17_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/exps/run_ex17_3ep/ft/tokenizer_config.json ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "32000": {
31
+ "content": "[PAD]",
32
+ "lstrip": false,
33
+ "normalized": false,
34
+ "rstrip": false,
35
+ "single_word": false,
36
+ "special": true
37
+ }
38
+ },
39
+ "bos_token": "</s>",
40
+ "clean_up_tokenization_spaces": false,
41
+ "eos_token": "</s>",
42
+ "extra_special_tokens": {},
43
+ "legacy": false,
44
+ "model_max_length": 512,
45
+ "pad_token": "[PAD]",
46
+ "padding_side": "right",
47
+ "sp_model_kwargs": {},
48
+ "tokenizer_class": "LlamaTokenizer",
49
+ "unk_token": "</s>",
50
+ "use_default_system_prompt": false
51
+ }
nl_tasks/exps/run_ex17_3ep/ft2/adapter_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "inference_mode": true,
6
+ "layers_to_transform": null,
7
+ "modules_to_save": null,
8
+ "num_rotations": 1,
9
+ "peft_type": "ROTATION",
10
+ "r": 16,
11
+ "revision": null,
12
+ "target_modules": [
13
+ "q_proj",
14
+ "v_proj"
15
+ ],
16
+ "target_modules_to_skip": null,
17
+ "task_type": "CAUSAL_LM"
18
+ }
nl_tasks/exps/run_ex17_3ep/ft2/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5edb668d4c472ad12e6c3ccd8518ccac91ed6b829aa348e3201be6285d4eaed
3
+ size 33602915
nl_tasks/exps/run_ex17_3ep/trainer_state.json ADDED
@@ -0,0 +1,1209 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 3.0,
6
+ "eval_steps": 200,
7
+ "global_step": 3657,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.020508613617719443,
14
+ "grad_norm": 0.21541807055473328,
15
+ "learning_rate": 0.00013114754098360657,
16
+ "loss": 0.7241,
17
+ "step": 25
18
+ },
19
+ {
20
+ "epoch": 0.04101722723543889,
21
+ "grad_norm": 0.1841340959072113,
22
+ "learning_rate": 0.00026775956284153007,
23
+ "loss": 0.4236,
24
+ "step": 50
25
+ },
26
+ {
27
+ "epoch": 0.06152584085315833,
28
+ "grad_norm": 0.16279272735118866,
29
+ "learning_rate": 0.00040437158469945356,
30
+ "loss": 0.3816,
31
+ "step": 75
32
+ },
33
+ {
34
+ "epoch": 0.08203445447087777,
35
+ "grad_norm": 0.17813508212566376,
36
+ "learning_rate": 0.0005409836065573771,
37
+ "loss": 0.3441,
38
+ "step": 100
39
+ },
40
+ {
41
+ "epoch": 0.10254306808859721,
42
+ "grad_norm": 0.17836959660053253,
43
+ "learning_rate": 0.0006775956284153005,
44
+ "loss": 0.3495,
45
+ "step": 125
46
+ },
47
+ {
48
+ "epoch": 0.12305168170631665,
49
+ "grad_norm": 0.197165846824646,
50
+ "learning_rate": 0.0008142076502732241,
51
+ "loss": 0.3331,
52
+ "step": 150
53
+ },
54
+ {
55
+ "epoch": 0.1435602953240361,
56
+ "grad_norm": 0.18293458223342896,
57
+ "learning_rate": 0.0009508196721311475,
58
+ "loss": 0.3245,
59
+ "step": 175
60
+ },
61
+ {
62
+ "epoch": 0.16406890894175555,
63
+ "grad_norm": 0.24918632209300995,
64
+ "learning_rate": 0.001087431693989071,
65
+ "loss": 0.338,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.16406890894175555,
70
+ "eval_loss": 0.3094765841960907,
71
+ "eval_runtime": 19.508,
72
+ "eval_samples_per_second": 51.261,
73
+ "eval_steps_per_second": 0.82,
74
+ "step": 200
75
+ },
76
+ {
77
+ "epoch": 0.184577522559475,
78
+ "grad_norm": 0.23847968876361847,
79
+ "learning_rate": 0.0012240437158469945,
80
+ "loss": 0.3379,
81
+ "step": 225
82
+ },
83
+ {
84
+ "epoch": 0.20508613617719443,
85
+ "grad_norm": 0.3514845073223114,
86
+ "learning_rate": 0.001360655737704918,
87
+ "loss": 0.3206,
88
+ "step": 250
89
+ },
90
+ {
91
+ "epoch": 0.22559474979491387,
92
+ "grad_norm": 0.5564894676208496,
93
+ "learning_rate": 0.0014972677595628417,
94
+ "loss": 0.3344,
95
+ "step": 275
96
+ },
97
+ {
98
+ "epoch": 0.2461033634126333,
99
+ "grad_norm": 0.4419553875923157,
100
+ "learning_rate": 0.001633879781420765,
101
+ "loss": 0.3377,
102
+ "step": 300
103
+ },
104
+ {
105
+ "epoch": 0.2666119770303528,
106
+ "grad_norm": 0.38893619179725647,
107
+ "learning_rate": 0.0017704918032786885,
108
+ "loss": 0.3412,
109
+ "step": 325
110
+ },
111
+ {
112
+ "epoch": 0.2871205906480722,
113
+ "grad_norm": 0.25933361053466797,
114
+ "learning_rate": 0.0019071038251366119,
115
+ "loss": 0.3324,
116
+ "step": 350
117
+ },
118
+ {
119
+ "epoch": 0.30762920426579166,
120
+ "grad_norm": 0.27627459168434143,
121
+ "learning_rate": 0.0019999708397093505,
122
+ "loss": 0.3461,
123
+ "step": 375
124
+ },
125
+ {
126
+ "epoch": 0.3281378178835111,
127
+ "grad_norm": 0.2933349311351776,
128
+ "learning_rate": 0.0019995038580495014,
129
+ "loss": 0.3258,
130
+ "step": 400
131
+ },
132
+ {
133
+ "epoch": 0.3281378178835111,
134
+ "eval_loss": 0.3039913475513458,
135
+ "eval_runtime": 19.3205,
136
+ "eval_samples_per_second": 51.758,
137
+ "eval_steps_per_second": 0.828,
138
+ "step": 400
139
+ },
140
+ {
141
+ "epoch": 0.34864643150123054,
142
+ "grad_norm": 0.2349589616060257,
143
+ "learning_rate": 0.001998467646284801,
144
+ "loss": 0.3302,
145
+ "step": 425
146
+ },
147
+ {
148
+ "epoch": 0.36915504511895,
149
+ "grad_norm": 0.2064511477947235,
150
+ "learning_rate": 0.001996862794550972,
151
+ "loss": 0.3209,
152
+ "step": 450
153
+ },
154
+ {
155
+ "epoch": 0.3896636587366694,
156
+ "grad_norm": 0.36041706800460815,
157
+ "learning_rate": 0.0019946902168314013,
158
+ "loss": 0.3033,
159
+ "step": 475
160
+ },
161
+ {
162
+ "epoch": 0.41017227235438886,
163
+ "grad_norm": 0.15045902132987976,
164
+ "learning_rate": 0.0019919511504366124,
165
+ "loss": 0.2835,
166
+ "step": 500
167
+ },
168
+ {
169
+ "epoch": 0.4306808859721083,
170
+ "grad_norm": 0.19791097939014435,
171
+ "learning_rate": 0.0019886471552996053,
172
+ "loss": 0.2832,
173
+ "step": 525
174
+ },
175
+ {
176
+ "epoch": 0.45118949958982774,
177
+ "grad_norm": 0.19202283024787903,
178
+ "learning_rate": 0.0019847801130874524,
179
+ "loss": 0.2923,
180
+ "step": 550
181
+ },
182
+ {
183
+ "epoch": 0.4716981132075472,
184
+ "grad_norm": 0.18097373843193054,
185
+ "learning_rate": 0.0019803522261296652,
186
+ "loss": 0.2828,
187
+ "step": 575
188
+ },
189
+ {
190
+ "epoch": 0.4922067268252666,
191
+ "grad_norm": 0.1983017474412918,
192
+ "learning_rate": 0.0019753660161639405,
193
+ "loss": 0.2989,
194
+ "step": 600
195
+ },
196
+ {
197
+ "epoch": 0.4922067268252666,
198
+ "eval_loss": 0.2664524018764496,
199
+ "eval_runtime": 19.3377,
200
+ "eval_samples_per_second": 51.712,
201
+ "eval_steps_per_second": 0.827,
202
+ "step": 600
203
+ },
204
+ {
205
+ "epoch": 0.5127153404429861,
206
+ "grad_norm": 0.2155098021030426,
207
+ "learning_rate": 0.0019698243229,
208
+ "loss": 0.277,
209
+ "step": 625
210
+ },
211
+ {
212
+ "epoch": 0.5332239540607056,
213
+ "grad_norm": 0.15264636278152466,
214
+ "learning_rate": 0.001963730302402337,
215
+ "loss": 0.2687,
216
+ "step": 650
217
+ },
218
+ {
219
+ "epoch": 0.5537325676784249,
220
+ "grad_norm": 0.1524903029203415,
221
+ "learning_rate": 0.001957087425292799,
222
+ "loss": 0.2784,
223
+ "step": 675
224
+ },
225
+ {
226
+ "epoch": 0.5742411812961444,
227
+ "grad_norm": 0.18971478939056396,
228
+ "learning_rate": 0.0019498994747740258,
229
+ "loss": 0.2782,
230
+ "step": 700
231
+ },
232
+ {
233
+ "epoch": 0.5947497949138638,
234
+ "grad_norm": 0.13587556779384613,
235
+ "learning_rate": 0.0019421705444748652,
236
+ "loss": 0.2737,
237
+ "step": 725
238
+ },
239
+ {
240
+ "epoch": 0.6152584085315833,
241
+ "grad_norm": 0.13490721583366394,
242
+ "learning_rate": 0.0019339050361190018,
243
+ "loss": 0.2702,
244
+ "step": 750
245
+ },
246
+ {
247
+ "epoch": 0.6357670221493027,
248
+ "grad_norm": 0.13660351932048798,
249
+ "learning_rate": 0.0019251076570181186,
250
+ "loss": 0.2674,
251
+ "step": 775
252
+ },
253
+ {
254
+ "epoch": 0.6562756357670222,
255
+ "grad_norm": 0.14963610470294952,
256
+ "learning_rate": 0.0019157834173910233,
257
+ "loss": 0.2714,
258
+ "step": 800
259
+ },
260
+ {
261
+ "epoch": 0.6562756357670222,
262
+ "eval_loss": 0.2525370419025421,
263
+ "eval_runtime": 19.3205,
264
+ "eval_samples_per_second": 51.758,
265
+ "eval_steps_per_second": 0.828,
266
+ "step": 800
267
+ },
268
+ {
269
+ "epoch": 0.6767842493847416,
270
+ "grad_norm": 0.1333588808774948,
271
+ "learning_rate": 0.0019059376275102657,
272
+ "loss": 0.2633,
273
+ "step": 825
274
+ },
275
+ {
276
+ "epoch": 0.6972928630024611,
277
+ "grad_norm": 0.12624366581439972,
278
+ "learning_rate": 0.001895575894677869,
279
+ "loss": 0.266,
280
+ "step": 850
281
+ },
282
+ {
283
+ "epoch": 0.7178014766201805,
284
+ "grad_norm": 0.11346879601478577,
285
+ "learning_rate": 0.0018847041200319027,
286
+ "loss": 0.2538,
287
+ "step": 875
288
+ },
289
+ {
290
+ "epoch": 0.7383100902379,
291
+ "grad_norm": 0.15679576992988586,
292
+ "learning_rate": 0.001873328495185707,
293
+ "loss": 0.2594,
294
+ "step": 900
295
+ },
296
+ {
297
+ "epoch": 0.7588187038556193,
298
+ "grad_norm": 0.12098339200019836,
299
+ "learning_rate": 0.0018614554987016931,
300
+ "loss": 0.2629,
301
+ "step": 925
302
+ },
303
+ {
304
+ "epoch": 0.7793273174733388,
305
+ "grad_norm": 0.11968964338302612,
306
+ "learning_rate": 0.0018490918924017174,
307
+ "loss": 0.25,
308
+ "step": 950
309
+ },
310
+ {
311
+ "epoch": 0.7998359310910582,
312
+ "grad_norm": 0.14618395268917084,
313
+ "learning_rate": 0.0018362447175161404,
314
+ "loss": 0.2577,
315
+ "step": 975
316
+ },
317
+ {
318
+ "epoch": 0.8203445447087777,
319
+ "grad_norm": 0.13215681910514832,
320
+ "learning_rate": 0.0018229212906737556,
321
+ "loss": 0.2499,
322
+ "step": 1000
323
+ },
324
+ {
325
+ "epoch": 0.8203445447087777,
326
+ "eval_loss": 0.24081531167030334,
327
+ "eval_runtime": 19.3284,
328
+ "eval_samples_per_second": 51.737,
329
+ "eval_steps_per_second": 0.828,
330
+ "step": 1000
331
+ },
332
+ {
333
+ "epoch": 0.8408531583264971,
334
+ "grad_norm": 0.13577589392662048,
335
+ "learning_rate": 0.0018091291997348775,
336
+ "loss": 0.2494,
337
+ "step": 1025
338
+ },
339
+ {
340
+ "epoch": 0.8613617719442166,
341
+ "grad_norm": 0.17443382740020752,
342
+ "learning_rate": 0.001794876299469958,
343
+ "loss": 0.2488,
344
+ "step": 1050
345
+ },
346
+ {
347
+ "epoch": 0.881870385561936,
348
+ "grad_norm": 0.15822888910770416,
349
+ "learning_rate": 0.0017801707070861973,
350
+ "loss": 0.2555,
351
+ "step": 1075
352
+ },
353
+ {
354
+ "epoch": 0.9023789991796555,
355
+ "grad_norm": 0.11527194827795029,
356
+ "learning_rate": 0.0017650207976046884,
357
+ "loss": 0.2544,
358
+ "step": 1100
359
+ },
360
+ {
361
+ "epoch": 0.9228876127973749,
362
+ "grad_norm": 0.11024443805217743,
363
+ "learning_rate": 0.001749435199090739,
364
+ "loss": 0.2482,
365
+ "step": 1125
366
+ },
367
+ {
368
+ "epoch": 0.9433962264150944,
369
+ "grad_norm": 0.09550724923610687,
370
+ "learning_rate": 0.001733422787740079,
371
+ "loss": 0.247,
372
+ "step": 1150
373
+ },
374
+ {
375
+ "epoch": 0.9639048400328137,
376
+ "grad_norm": 0.10925162583589554,
377
+ "learning_rate": 0.001716992682823753,
378
+ "loss": 0.2526,
379
+ "step": 1175
380
+ },
381
+ {
382
+ "epoch": 0.9844134536505332,
383
+ "grad_norm": 0.1117289662361145,
384
+ "learning_rate": 0.0017001542414945814,
385
+ "loss": 0.2591,
386
+ "step": 1200
387
+ },
388
+ {
389
+ "epoch": 0.9844134536505332,
390
+ "eval_loss": 0.23284786939620972,
391
+ "eval_runtime": 19.3101,
392
+ "eval_samples_per_second": 51.786,
393
+ "eval_steps_per_second": 0.829,
394
+ "step": 1200
395
+ },
396
+ {
397
+ "epoch": 1.0049220672682526,
398
+ "grad_norm": 0.11111273616552353,
399
+ "learning_rate": 0.001682917053458143,
400
+ "loss": 0.2391,
401
+ "step": 1225
402
+ },
403
+ {
404
+ "epoch": 1.0254306808859721,
405
+ "grad_norm": 0.10441102832555771,
406
+ "learning_rate": 0.0016652909355113154,
407
+ "loss": 0.2211,
408
+ "step": 1250
409
+ },
410
+ {
411
+ "epoch": 1.0459392945036916,
412
+ "grad_norm": 0.11000514775514603,
413
+ "learning_rate": 0.0016472859259514863,
414
+ "loss": 0.2238,
415
+ "step": 1275
416
+ },
417
+ {
418
+ "epoch": 1.066447908121411,
419
+ "grad_norm": 0.09721962362527847,
420
+ "learning_rate": 0.001628912278859615,
421
+ "loss": 0.2259,
422
+ "step": 1300
423
+ },
424
+ {
425
+ "epoch": 1.0869565217391304,
426
+ "grad_norm": 0.1097666546702385,
427
+ "learning_rate": 0.0016101804582604037,
428
+ "loss": 0.2248,
429
+ "step": 1325
430
+ },
431
+ {
432
+ "epoch": 1.1074651353568499,
433
+ "grad_norm": 0.08520916104316711,
434
+ "learning_rate": 0.001591101132162903,
435
+ "loss": 0.2164,
436
+ "step": 1350
437
+ },
438
+ {
439
+ "epoch": 1.1279737489745694,
440
+ "grad_norm": 0.10846058279275894,
441
+ "learning_rate": 0.0015716851664849454,
442
+ "loss": 0.2194,
443
+ "step": 1375
444
+ },
445
+ {
446
+ "epoch": 1.1484823625922886,
447
+ "grad_norm": 0.1124526858329773,
448
+ "learning_rate": 0.001551943618864868,
449
+ "loss": 0.2131,
450
+ "step": 1400
451
+ },
452
+ {
453
+ "epoch": 1.1484823625922886,
454
+ "eval_loss": 0.2294142097234726,
455
+ "eval_runtime": 19.3502,
456
+ "eval_samples_per_second": 51.679,
457
+ "eval_steps_per_second": 0.827,
458
+ "step": 1400
459
+ },
460
+ {
461
+ "epoch": 1.1689909762100081,
462
+ "grad_norm": 0.10251163691282272,
463
+ "learning_rate": 0.001531887732364047,
464
+ "loss": 0.224,
465
+ "step": 1425
466
+ },
467
+ {
468
+ "epoch": 1.1894995898277276,
469
+ "grad_norm": 0.11376211792230606,
470
+ "learning_rate": 0.0015115289290638313,
471
+ "loss": 0.2253,
472
+ "step": 1450
473
+ },
474
+ {
475
+ "epoch": 1.2100082034454471,
476
+ "grad_norm": 0.1140202060341835,
477
+ "learning_rate": 0.0014908788035605245,
478
+ "loss": 0.2216,
479
+ "step": 1475
480
+ },
481
+ {
482
+ "epoch": 1.2305168170631666,
483
+ "grad_norm": 0.1042315736413002,
484
+ "learning_rate": 0.001469949116362113,
485
+ "loss": 0.2313,
486
+ "step": 1500
487
+ },
488
+ {
489
+ "epoch": 1.251025430680886,
490
+ "grad_norm": 0.11596186459064484,
491
+ "learning_rate": 0.0014487517871905095,
492
+ "loss": 0.2254,
493
+ "step": 1525
494
+ },
495
+ {
496
+ "epoch": 1.2715340442986054,
497
+ "grad_norm": 0.10951562225818634,
498
+ "learning_rate": 0.0014272988881931191,
499
+ "loss": 0.2186,
500
+ "step": 1550
501
+ },
502
+ {
503
+ "epoch": 1.2920426579163249,
504
+ "grad_norm": 0.08798109740018845,
505
+ "learning_rate": 0.0014056026370676002,
506
+ "loss": 0.221,
507
+ "step": 1575
508
+ },
509
+ {
510
+ "epoch": 1.3125512715340442,
511
+ "grad_norm": 0.0925319492816925,
512
+ "learning_rate": 0.00138367539010373,
513
+ "loss": 0.2147,
514
+ "step": 1600
515
+ },
516
+ {
517
+ "epoch": 1.3125512715340442,
518
+ "eval_loss": 0.22312961518764496,
519
+ "eval_runtime": 19.3308,
520
+ "eval_samples_per_second": 51.731,
521
+ "eval_steps_per_second": 0.828,
522
+ "step": 1600
523
+ },
524
+ {
525
+ "epoch": 1.3330598851517639,
526
+ "grad_norm": 0.09203176945447922,
527
+ "learning_rate": 0.00136152963514634,
528
+ "loss": 0.2168,
529
+ "step": 1625
530
+ },
531
+ {
532
+ "epoch": 1.3535684987694832,
533
+ "grad_norm": 0.09899195283651352,
534
+ "learning_rate": 0.0013391779844833319,
535
+ "loss": 0.2183,
536
+ "step": 1650
537
+ },
538
+ {
539
+ "epoch": 1.3740771123872026,
540
+ "grad_norm": 0.10101519525051117,
541
+ "learning_rate": 0.001316633167662819,
542
+ "loss": 0.2151,
543
+ "step": 1675
544
+ },
545
+ {
546
+ "epoch": 1.3945857260049221,
547
+ "grad_norm": 0.09377351403236389,
548
+ "learning_rate": 0.0012939080242434875,
549
+ "loss": 0.2149,
550
+ "step": 1700
551
+ },
552
+ {
553
+ "epoch": 1.4150943396226414,
554
+ "grad_norm": 0.09053958207368851,
555
+ "learning_rate": 0.0012710154964823053,
556
+ "loss": 0.2134,
557
+ "step": 1725
558
+ },
559
+ {
560
+ "epoch": 1.435602953240361,
561
+ "grad_norm": 0.09443776309490204,
562
+ "learning_rate": 0.0012479686219637432,
563
+ "loss": 0.2123,
564
+ "step": 1750
565
+ },
566
+ {
567
+ "epoch": 1.4561115668580804,
568
+ "grad_norm": 0.09800270199775696,
569
+ "learning_rate": 0.0012247805261747043,
570
+ "loss": 0.2196,
571
+ "step": 1775
572
+ },
573
+ {
574
+ "epoch": 1.4766201804758,
575
+ "grad_norm": 0.0898151770234108,
576
+ "learning_rate": 0.0012014644150293936,
577
+ "loss": 0.2095,
578
+ "step": 1800
579
+ },
580
+ {
581
+ "epoch": 1.4766201804758,
582
+ "eval_loss": 0.21844735741615295,
583
+ "eval_runtime": 19.3418,
584
+ "eval_samples_per_second": 51.701,
585
+ "eval_steps_per_second": 0.827,
586
+ "step": 1800
587
+ },
588
+ {
589
+ "epoch": 1.4971287940935194,
590
+ "grad_norm": 0.08816877007484436,
591
+ "learning_rate": 0.0011780335673483838,
592
+ "loss": 0.2077,
593
+ "step": 1825
594
+ },
595
+ {
596
+ "epoch": 1.5176374077112387,
597
+ "grad_norm": 0.08615899085998535,
598
+ "learning_rate": 0.0011545013272961562,
599
+ "loss": 0.2125,
600
+ "step": 1850
601
+ },
602
+ {
603
+ "epoch": 1.5381460213289582,
604
+ "grad_norm": 0.08997204899787903,
605
+ "learning_rate": 0.0011308810967814325,
606
+ "loss": 0.2121,
607
+ "step": 1875
608
+ },
609
+ {
610
+ "epoch": 1.5586546349466777,
611
+ "grad_norm": 0.09502812474966049,
612
+ "learning_rate": 0.0011071863278246166,
613
+ "loss": 0.215,
614
+ "step": 1900
615
+ },
616
+ {
617
+ "epoch": 1.579163248564397,
618
+ "grad_norm": 0.09312719106674194,
619
+ "learning_rate": 0.0010834305148967,
620
+ "loss": 0.2172,
621
+ "step": 1925
622
+ },
623
+ {
624
+ "epoch": 1.5996718621821167,
625
+ "grad_norm": 0.10026481747627258,
626
+ "learning_rate": 0.0010596271872339875,
627
+ "loss": 0.2159,
628
+ "step": 1950
629
+ },
630
+ {
631
+ "epoch": 1.620180475799836,
632
+ "grad_norm": 0.0816514864563942,
633
+ "learning_rate": 0.001035789901133028,
634
+ "loss": 0.2091,
635
+ "step": 1975
636
+ },
637
+ {
638
+ "epoch": 1.6406890894175554,
639
+ "grad_norm": 0.10148680955171585,
640
+ "learning_rate": 0.001011932232230131,
641
+ "loss": 0.2161,
642
+ "step": 2000
643
+ },
644
+ {
645
+ "epoch": 1.6406890894175554,
646
+ "eval_loss": 0.21507860720157623,
647
+ "eval_runtime": 19.3438,
648
+ "eval_samples_per_second": 51.696,
649
+ "eval_steps_per_second": 0.827,
650
+ "step": 2000
651
+ },
652
+ {
653
+ "epoch": 1.661197703035275,
654
+ "grad_norm": 0.08850109577178955,
655
+ "learning_rate": 0.0009880677677698694,
656
+ "loss": 0.2068,
657
+ "step": 2025
658
+ },
659
+ {
660
+ "epoch": 1.6817063166529942,
661
+ "grad_norm": 0.09822677820920944,
662
+ "learning_rate": 0.0009642100988669722,
663
+ "loss": 0.2091,
664
+ "step": 2050
665
+ },
666
+ {
667
+ "epoch": 1.7022149302707137,
668
+ "grad_norm": 0.10029079020023346,
669
+ "learning_rate": 0.0009403728127660128,
670
+ "loss": 0.2134,
671
+ "step": 2075
672
+ },
673
+ {
674
+ "epoch": 1.7227235438884332,
675
+ "grad_norm": 0.10123023390769958,
676
+ "learning_rate": 0.0009165694851033002,
677
+ "loss": 0.2043,
678
+ "step": 2100
679
+ },
680
+ {
681
+ "epoch": 1.7432321575061525,
682
+ "grad_norm": 0.07858412712812424,
683
+ "learning_rate": 0.0008928136721753837,
684
+ "loss": 0.2078,
685
+ "step": 2125
686
+ },
687
+ {
688
+ "epoch": 1.7637407711238722,
689
+ "grad_norm": 0.07627654820680618,
690
+ "learning_rate": 0.0008691189032185676,
691
+ "loss": 0.2055,
692
+ "step": 2150
693
+ },
694
+ {
695
+ "epoch": 1.7842493847415914,
696
+ "grad_norm": 0.07735332101583481,
697
+ "learning_rate": 0.0008454986727038442,
698
+ "loss": 0.2138,
699
+ "step": 2175
700
+ },
701
+ {
702
+ "epoch": 1.804757998359311,
703
+ "grad_norm": 0.08670951426029205,
704
+ "learning_rate": 0.0008219664326516164,
705
+ "loss": 0.2065,
706
+ "step": 2200
707
+ },
708
+ {
709
+ "epoch": 1.804757998359311,
710
+ "eval_loss": 0.21065279841423035,
711
+ "eval_runtime": 19.3209,
712
+ "eval_samples_per_second": 51.757,
713
+ "eval_steps_per_second": 0.828,
714
+ "step": 2200
715
+ },
716
+ {
717
+ "epoch": 1.8252666119770304,
718
+ "grad_norm": 0.09341748058795929,
719
+ "learning_rate": 0.0007985355849706064,
720
+ "loss": 0.2075,
721
+ "step": 2225
722
+ },
723
+ {
724
+ "epoch": 1.8457752255947497,
725
+ "grad_norm": 0.08657963573932648,
726
+ "learning_rate": 0.0007752194738252958,
727
+ "loss": 0.2079,
728
+ "step": 2250
729
+ },
730
+ {
731
+ "epoch": 1.8662838392124692,
732
+ "grad_norm": 0.08066408336162567,
733
+ "learning_rate": 0.0007520313780362566,
734
+ "loss": 0.2136,
735
+ "step": 2275
736
+ },
737
+ {
738
+ "epoch": 1.8867924528301887,
739
+ "grad_norm": 0.0823824554681778,
740
+ "learning_rate": 0.0007289845035176948,
741
+ "loss": 0.2063,
742
+ "step": 2300
743
+ },
744
+ {
745
+ "epoch": 1.907301066447908,
746
+ "grad_norm": 0.09431551396846771,
747
+ "learning_rate": 0.0007060919757565126,
748
+ "loss": 0.2149,
749
+ "step": 2325
750
+ },
751
+ {
752
+ "epoch": 1.9278096800656277,
753
+ "grad_norm": 0.08885450661182404,
754
+ "learning_rate": 0.0006833668323371813,
755
+ "loss": 0.1984,
756
+ "step": 2350
757
+ },
758
+ {
759
+ "epoch": 1.948318293683347,
760
+ "grad_norm": 0.07753252983093262,
761
+ "learning_rate": 0.0006608220155166682,
762
+ "loss": 0.2023,
763
+ "step": 2375
764
+ },
765
+ {
766
+ "epoch": 1.9688269073010665,
767
+ "grad_norm": 0.07865183055400848,
768
+ "learning_rate": 0.00063847036485366,
769
+ "loss": 0.2038,
770
+ "step": 2400
771
+ },
772
+ {
773
+ "epoch": 1.9688269073010665,
774
+ "eval_loss": 0.2072540819644928,
775
+ "eval_runtime": 19.3555,
776
+ "eval_samples_per_second": 51.665,
777
+ "eval_steps_per_second": 0.827,
778
+ "step": 2400
779
+ },
780
+ {
781
+ "epoch": 1.989335520918786,
782
+ "grad_norm": 0.07822025567293167,
783
+ "learning_rate": 0.00061632460989627,
784
+ "loss": 0.2016,
785
+ "step": 2425
786
+ },
787
+ {
788
+ "epoch": 2.0098441345365052,
789
+ "grad_norm": 0.09284231066703796,
790
+ "learning_rate": 0.0005943973629323999,
791
+ "loss": 0.1887,
792
+ "step": 2450
793
+ },
794
+ {
795
+ "epoch": 2.030352748154225,
796
+ "grad_norm": 0.08939560502767563,
797
+ "learning_rate": 0.0005727011118068812,
798
+ "loss": 0.1763,
799
+ "step": 2475
800
+ },
801
+ {
802
+ "epoch": 2.0508613617719442,
803
+ "grad_norm": 0.09618934243917465,
804
+ "learning_rate": 0.0005512482128094912,
805
+ "loss": 0.178,
806
+ "step": 2500
807
+ },
808
+ {
809
+ "epoch": 2.0713699753896635,
810
+ "grad_norm": 0.09260787069797516,
811
+ "learning_rate": 0.0005300508836378875,
812
+ "loss": 0.1726,
813
+ "step": 2525
814
+ },
815
+ {
816
+ "epoch": 2.091878589007383,
817
+ "grad_norm": 0.09563781321048737,
818
+ "learning_rate": 0.0005091211964394757,
819
+ "loss": 0.181,
820
+ "step": 2550
821
+ },
822
+ {
823
+ "epoch": 2.1123872026251025,
824
+ "grad_norm": 0.08169485628604889,
825
+ "learning_rate": 0.0004884710709361685,
826
+ "loss": 0.1741,
827
+ "step": 2575
828
+ },
829
+ {
830
+ "epoch": 2.132895816242822,
831
+ "grad_norm": 0.09133849292993546,
832
+ "learning_rate": 0.0004681122676359533,
833
+ "loss": 0.1663,
834
+ "step": 2600
835
+ },
836
+ {
837
+ "epoch": 2.132895816242822,
838
+ "eval_loss": 0.2086760252714157,
839
+ "eval_runtime": 19.3159,
840
+ "eval_samples_per_second": 51.771,
841
+ "eval_steps_per_second": 0.828,
842
+ "step": 2600
843
+ },
844
+ {
845
+ "epoch": 2.1534044298605415,
846
+ "grad_norm": 0.09362433105707169,
847
+ "learning_rate": 0.00044805638113513203,
848
+ "loss": 0.171,
849
+ "step": 2625
850
+ },
851
+ {
852
+ "epoch": 2.1739130434782608,
853
+ "grad_norm": 0.08513038605451584,
854
+ "learning_rate": 0.00042831483351505483,
855
+ "loss": 0.1696,
856
+ "step": 2650
857
+ },
858
+ {
859
+ "epoch": 2.1944216570959805,
860
+ "grad_norm": 0.0917547419667244,
861
+ "learning_rate": 0.0004088988678370973,
862
+ "loss": 0.1816,
863
+ "step": 2675
864
+ },
865
+ {
866
+ "epoch": 2.2149302707136997,
867
+ "grad_norm": 0.07697644084692001,
868
+ "learning_rate": 0.00038981954173959645,
869
+ "loss": 0.1779,
870
+ "step": 2700
871
+ },
872
+ {
873
+ "epoch": 2.235438884331419,
874
+ "grad_norm": 0.09827437996864319,
875
+ "learning_rate": 0.000371087721140385,
876
+ "loss": 0.1668,
877
+ "step": 2725
878
+ },
879
+ {
880
+ "epoch": 2.2559474979491387,
881
+ "grad_norm": 0.08829747885465622,
882
+ "learning_rate": 0.00035271407404851364,
883
+ "loss": 0.179,
884
+ "step": 2750
885
+ },
886
+ {
887
+ "epoch": 2.276456111566858,
888
+ "grad_norm": 0.09802579879760742,
889
+ "learning_rate": 0.0003347090644886848,
890
+ "loss": 0.1757,
891
+ "step": 2775
892
+ },
893
+ {
894
+ "epoch": 2.2969647251845773,
895
+ "grad_norm": 0.09801331907510757,
896
+ "learning_rate": 0.0003170829465418571,
897
+ "loss": 0.1719,
898
+ "step": 2800
899
+ },
900
+ {
901
+ "epoch": 2.2969647251845773,
902
+ "eval_loss": 0.2065618932247162,
903
+ "eval_runtime": 19.3412,
904
+ "eval_samples_per_second": 51.703,
905
+ "eval_steps_per_second": 0.827,
906
+ "step": 2800
907
+ },
908
+ {
909
+ "epoch": 2.317473338802297,
910
+ "grad_norm": 0.10572776198387146,
911
+ "learning_rate": 0.0002998457585054187,
912
+ "loss": 0.1679,
913
+ "step": 2825
914
+ },
915
+ {
916
+ "epoch": 2.3379819524200163,
917
+ "grad_norm": 0.09379872679710388,
918
+ "learning_rate": 0.00028300731717624707,
919
+ "loss": 0.1712,
920
+ "step": 2850
921
+ },
922
+ {
923
+ "epoch": 2.358490566037736,
924
+ "grad_norm": 0.07830008864402771,
925
+ "learning_rate": 0.00026657721225992106,
926
+ "loss": 0.1718,
927
+ "step": 2875
928
+ },
929
+ {
930
+ "epoch": 2.3789991796554553,
931
+ "grad_norm": 0.1549721360206604,
932
+ "learning_rate": 0.00025056480090926094,
933
+ "loss": 0.1758,
934
+ "step": 2900
935
+ },
936
+ {
937
+ "epoch": 2.399507793273175,
938
+ "grad_norm": 0.10021212697029114,
939
+ "learning_rate": 0.0002349792023953119,
940
+ "loss": 0.1731,
941
+ "step": 2925
942
+ },
943
+ {
944
+ "epoch": 2.4200164068908943,
945
+ "grad_norm": 0.1022850051522255,
946
+ "learning_rate": 0.00021982929291380283,
947
+ "loss": 0.176,
948
+ "step": 2950
949
+ },
950
+ {
951
+ "epoch": 2.4405250205086135,
952
+ "grad_norm": 0.08516191691160202,
953
+ "learning_rate": 0.00020512370053004191,
954
+ "loss": 0.1722,
955
+ "step": 2975
956
+ },
957
+ {
958
+ "epoch": 2.4610336341263332,
959
+ "grad_norm": 0.10133907198905945,
960
+ "learning_rate": 0.00019087080026512272,
961
+ "loss": 0.168,
962
+ "step": 3000
963
+ },
964
+ {
965
+ "epoch": 2.4610336341263332,
966
+ "eval_loss": 0.20472551882266998,
967
+ "eval_runtime": 19.3251,
968
+ "eval_samples_per_second": 51.746,
969
+ "eval_steps_per_second": 0.828,
970
+ "step": 3000
971
+ },
972
+ {
973
+ "epoch": 2.4815422477440525,
974
+ "grad_norm": 0.08802944421768188,
975
+ "learning_rate": 0.00017707870932624447,
976
+ "loss": 0.1721,
977
+ "step": 3025
978
+ },
979
+ {
980
+ "epoch": 2.502050861361772,
981
+ "grad_norm": 0.07769055664539337,
982
+ "learning_rate": 0.00016375528248385962,
983
+ "loss": 0.1664,
984
+ "step": 3050
985
+ },
986
+ {
987
+ "epoch": 2.5225594749794915,
988
+ "grad_norm": 0.09567312151193619,
989
+ "learning_rate": 0.00015090810759828277,
990
+ "loss": 0.163,
991
+ "step": 3075
992
+ },
993
+ {
994
+ "epoch": 2.543068088597211,
995
+ "grad_norm": 0.0900384858250618,
996
+ "learning_rate": 0.00013854450129830698,
997
+ "loss": 0.1743,
998
+ "step": 3100
999
+ },
1000
+ {
1001
+ "epoch": 2.56357670221493,
1002
+ "grad_norm": 0.09516915678977966,
1003
+ "learning_rate": 0.00012667150481429312,
1004
+ "loss": 0.163,
1005
+ "step": 3125
1006
+ },
1007
+ {
1008
+ "epoch": 2.5840853158326498,
1009
+ "grad_norm": 0.10980784893035889,
1010
+ "learning_rate": 0.00011529587996809753,
1011
+ "loss": 0.1676,
1012
+ "step": 3150
1013
+ },
1014
+ {
1015
+ "epoch": 2.604593929450369,
1016
+ "grad_norm": 0.08579864352941513,
1017
+ "learning_rate": 0.00010442410532213131,
1018
+ "loss": 0.1691,
1019
+ "step": 3175
1020
+ },
1021
+ {
1022
+ "epoch": 2.6251025430680883,
1023
+ "grad_norm": 0.08947206288576126,
1024
+ "learning_rate": 9.40623724897347e-05,
1025
+ "loss": 0.173,
1026
+ "step": 3200
1027
+ },
1028
+ {
1029
+ "epoch": 2.6251025430680883,
1030
+ "eval_loss": 0.20346350967884064,
1031
+ "eval_runtime": 19.3319,
1032
+ "eval_samples_per_second": 51.728,
1033
+ "eval_steps_per_second": 0.828,
1034
+ "step": 3200
1035
+ },
1036
+ {
1037
+ "epoch": 2.645611156685808,
1038
+ "grad_norm": 0.10664288699626923,
1039
+ "learning_rate": 8.421658260897668e-05,
1040
+ "loss": 0.18,
1041
+ "step": 3225
1042
+ },
1043
+ {
1044
+ "epoch": 2.6661197703035278,
1045
+ "grad_norm": 0.13291819393634796,
1046
+ "learning_rate": 7.489234298188141e-05,
1047
+ "loss": 0.1664,
1048
+ "step": 3250
1049
+ },
1050
+ {
1051
+ "epoch": 2.686628383921247,
1052
+ "grad_norm": 0.09891302138566971,
1053
+ "learning_rate": 6.609496388099811e-05,
1054
+ "loss": 0.1713,
1055
+ "step": 3275
1056
+ },
1057
+ {
1058
+ "epoch": 2.7071369975389663,
1059
+ "grad_norm": 0.08029094338417053,
1060
+ "learning_rate": 5.7829455525134965e-05,
1061
+ "loss": 0.1649,
1062
+ "step": 3300
1063
+ },
1064
+ {
1065
+ "epoch": 2.727645611156686,
1066
+ "grad_norm": 0.09289735555648804,
1067
+ "learning_rate": 5.010052522597441e-05,
1068
+ "loss": 0.1649,
1069
+ "step": 3325
1070
+ },
1071
+ {
1072
+ "epoch": 2.7481542247744053,
1073
+ "grad_norm": 0.08359911292791367,
1074
+ "learning_rate": 4.2912574707201e-05,
1075
+ "loss": 0.1601,
1076
+ "step": 3350
1077
+ },
1078
+ {
1079
+ "epoch": 2.7686628383921246,
1080
+ "grad_norm": 0.09924525767564774,
1081
+ "learning_rate": 3.6269697597663074e-05,
1082
+ "loss": 0.1781,
1083
+ "step": 3375
1084
+ },
1085
+ {
1086
+ "epoch": 2.7891714520098443,
1087
+ "grad_norm": 0.08123107254505157,
1088
+ "learning_rate": 3.017567709999991e-05,
1089
+ "loss": 0.1658,
1090
+ "step": 3400
1091
+ },
1092
+ {
1093
+ "epoch": 2.7891714520098443,
1094
+ "eval_loss": 0.20290200412273407,
1095
+ "eval_runtime": 19.3369,
1096
+ "eval_samples_per_second": 51.715,
1097
+ "eval_steps_per_second": 0.827,
1098
+ "step": 3400
1099
+ },
1100
+ {
1101
+ "epoch": 2.8096800656275636,
1102
+ "grad_norm": 0.08416769653558731,
1103
+ "learning_rate": 2.4633983836059305e-05,
1104
+ "loss": 0.1739,
1105
+ "step": 3425
1106
+ },
1107
+ {
1108
+ "epoch": 2.830188679245283,
1109
+ "grad_norm": 0.08879395574331284,
1110
+ "learning_rate": 1.9647773870335005e-05,
1111
+ "loss": 0.1659,
1112
+ "step": 3450
1113
+ },
1114
+ {
1115
+ "epoch": 2.8506972928630026,
1116
+ "grad_norm": 0.09227652102708817,
1117
+ "learning_rate": 1.521988691254772e-05,
1118
+ "loss": 0.1723,
1119
+ "step": 3475
1120
+ },
1121
+ {
1122
+ "epoch": 2.871205906480722,
1123
+ "grad_norm": 0.0831235870718956,
1124
+ "learning_rate": 1.135284470039466e-05,
1125
+ "loss": 0.1643,
1126
+ "step": 3500
1127
+ },
1128
+ {
1129
+ "epoch": 2.891714520098441,
1130
+ "grad_norm": 0.09819941967725754,
1131
+ "learning_rate": 8.048849563387584e-06,
1132
+ "loss": 0.1665,
1133
+ "step": 3525
1134
+ },
1135
+ {
1136
+ "epoch": 2.912223133716161,
1137
+ "grad_norm": 0.09480398893356323,
1138
+ "learning_rate": 5.309783168598759e-06,
1139
+ "loss": 0.167,
1140
+ "step": 3550
1141
+ },
1142
+ {
1143
+ "epoch": 2.93273174733388,
1144
+ "grad_norm": 0.10071857273578644,
1145
+ "learning_rate": 3.1372054490276247e-06,
1146
+ "loss": 0.1644,
1147
+ "step": 3575
1148
+ },
1149
+ {
1150
+ "epoch": 2.9532403609516,
1151
+ "grad_norm": 0.08304011821746826,
1152
+ "learning_rate": 1.5323537151988864e-06,
1153
+ "loss": 0.1704,
1154
+ "step": 3600
1155
+ },
1156
+ {
1157
+ "epoch": 2.9532403609516,
1158
+ "eval_loss": 0.2025250941514969,
1159
+ "eval_runtime": 19.322,
1160
+ "eval_samples_per_second": 51.754,
1161
+ "eval_steps_per_second": 0.828,
1162
+ "step": 3600
1163
+ },
1164
+ {
1165
+ "epoch": 2.973748974569319,
1166
+ "grad_norm": 0.08062180876731873,
1167
+ "learning_rate": 4.961419504987408e-07,
1168
+ "loss": 0.1663,
1169
+ "step": 3625
1170
+ },
1171
+ {
1172
+ "epoch": 2.994257588187039,
1173
+ "grad_norm": 0.08830997347831726,
1174
+ "learning_rate": 2.916029064969905e-08,
1175
+ "loss": 0.1645,
1176
+ "step": 3650
1177
+ },
1178
+ {
1179
+ "epoch": 3.0,
1180
+ "step": 3657,
1181
+ "total_flos": 2.37785441107968e+18,
1182
+ "train_loss": 0.2289935703449703,
1183
+ "train_runtime": 3736.472,
1184
+ "train_samples_per_second": 31.313,
1185
+ "train_steps_per_second": 0.979
1186
+ }
1187
+ ],
1188
+ "logging_steps": 25,
1189
+ "max_steps": 3657,
1190
+ "num_input_tokens_seen": 0,
1191
+ "num_train_epochs": 3,
1192
+ "save_steps": 0,
1193
+ "stateful_callbacks": {
1194
+ "TrainerControl": {
1195
+ "args": {
1196
+ "should_epoch_stop": false,
1197
+ "should_evaluate": false,
1198
+ "should_log": false,
1199
+ "should_save": true,
1200
+ "should_training_stop": true
1201
+ },
1202
+ "attributes": {}
1203
+ }
1204
+ },
1205
+ "total_flos": 2.37785441107968e+18,
1206
+ "train_batch_size": 32,
1207
+ "trial_name": null,
1208
+ "trial_params": null
1209
+ }
nl_tasks/exps/run_ex18_2ep/ft/adapter_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "inference_mode": false,
6
+ "layers_to_transform": null,
7
+ "modules_to_save": null,
8
+ "num_rotations": 1,
9
+ "peft_type": "ROTATION",
10
+ "r": 16,
11
+ "revision": null,
12
+ "target_modules": [
13
+ "q_proj",
14
+ "v_proj"
15
+ ],
16
+ "target_modules_to_skip": null,
17
+ "task_type": "CAUSAL_LM"
18
+ }
nl_tasks/exps/run_ex18_2ep/ft/added_tokens.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ {
2
+ "[PAD]": 32000
3
+ }
nl_tasks/exps/run_ex18_2ep/ft/special_tokens_map.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": {
17
+ "content": "[PAD]",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ },
23
+ "unk_token": {
24
+ "content": "</s>",
25
+ "lstrip": false,
26
+ "normalized": false,
27
+ "rstrip": false,
28
+ "single_word": false
29
+ }
30
+ }
nl_tasks/exps/run_ex18_2ep/ft/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
nl_tasks/exps/run_ex18_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/exps/run_ex18_2ep/ft/tokenizer_config.json ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "32000": {
31
+ "content": "[PAD]",
32
+ "lstrip": false,
33
+ "normalized": false,
34
+ "rstrip": false,
35
+ "single_word": false,
36
+ "special": true
37
+ }
38
+ },
39
+ "bos_token": "</s>",
40
+ "clean_up_tokenization_spaces": false,
41
+ "eos_token": "</s>",
42
+ "extra_special_tokens": {},
43
+ "legacy": false,
44
+ "model_max_length": 512,
45
+ "pad_token": "[PAD]",
46
+ "padding_side": "right",
47
+ "sp_model_kwargs": {},
48
+ "tokenizer_class": "LlamaTokenizer",
49
+ "unk_token": "</s>",
50
+ "use_default_system_prompt": false
51
+ }
nl_tasks/exps/run_ex18_2ep/ft2/adapter_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "inference_mode": true,
6
+ "layers_to_transform": null,
7
+ "modules_to_save": null,
8
+ "num_rotations": 1,
9
+ "peft_type": "ROTATION",
10
+ "r": 16,
11
+ "revision": null,
12
+ "target_modules": [
13
+ "q_proj",
14
+ "v_proj"
15
+ ],
16
+ "target_modules_to_skip": null,
17
+ "task_type": "CAUSAL_LM"
18
+ }
nl_tasks/exps/run_ex18_2ep/ft2/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7844b9b079728faaf196fb3ab0aa37bff543c2815eab16a02d813ee7764c4d73
3
+ size 33602915
nl_tasks/exps/run_ex18_2ep/trainer_state.json ADDED
@@ -0,0 +1,818 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 2.0,
6
+ "eval_steps": 200,
7
+ "global_step": 2438,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.020508613617719443,
14
+ "grad_norm": 0.4331222474575043,
15
+ "learning_rate": 9.836065573770491e-05,
16
+ "loss": 0.8296,
17
+ "step": 25
18
+ },
19
+ {
20
+ "epoch": 0.04101722723543889,
21
+ "grad_norm": 0.1884848177433014,
22
+ "learning_rate": 0.00020081967213114754,
23
+ "loss": 0.4641,
24
+ "step": 50
25
+ },
26
+ {
27
+ "epoch": 0.06152584085315833,
28
+ "grad_norm": 0.1777927428483963,
29
+ "learning_rate": 0.0003032786885245902,
30
+ "loss": 0.4054,
31
+ "step": 75
32
+ },
33
+ {
34
+ "epoch": 0.08203445447087777,
35
+ "grad_norm": 0.19921301305294037,
36
+ "learning_rate": 0.0004057377049180328,
37
+ "loss": 0.3637,
38
+ "step": 100
39
+ },
40
+ {
41
+ "epoch": 0.10254306808859721,
42
+ "grad_norm": 0.28275343775749207,
43
+ "learning_rate": 0.0005081967213114754,
44
+ "loss": 0.3683,
45
+ "step": 125
46
+ },
47
+ {
48
+ "epoch": 0.12305168170631665,
49
+ "grad_norm": 0.18318778276443481,
50
+ "learning_rate": 0.000610655737704918,
51
+ "loss": 0.348,
52
+ "step": 150
53
+ },
54
+ {
55
+ "epoch": 0.1435602953240361,
56
+ "grad_norm": 0.1784578114748001,
57
+ "learning_rate": 0.0007131147540983607,
58
+ "loss": 0.3356,
59
+ "step": 175
60
+ },
61
+ {
62
+ "epoch": 0.16406890894175555,
63
+ "grad_norm": 0.2444526106119156,
64
+ "learning_rate": 0.0008155737704918033,
65
+ "loss": 0.3463,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.16406890894175555,
70
+ "eval_loss": 0.3061140477657318,
71
+ "eval_runtime": 19.6138,
72
+ "eval_samples_per_second": 50.984,
73
+ "eval_steps_per_second": 0.816,
74
+ "step": 200
75
+ },
76
+ {
77
+ "epoch": 0.184577522559475,
78
+ "grad_norm": 0.1917794942855835,
79
+ "learning_rate": 0.0009180327868852459,
80
+ "loss": 0.3399,
81
+ "step": 225
82
+ },
83
+ {
84
+ "epoch": 0.20508613617719443,
85
+ "grad_norm": 0.2342771738767624,
86
+ "learning_rate": 0.0009999871854116063,
87
+ "loss": 0.3197,
88
+ "step": 250
89
+ },
90
+ {
91
+ "epoch": 0.22559474979491387,
92
+ "grad_norm": 0.2321336567401886,
93
+ "learning_rate": 0.0009995387437838027,
94
+ "loss": 0.3206,
95
+ "step": 275
96
+ },
97
+ {
98
+ "epoch": 0.2461033634126333,
99
+ "grad_norm": 0.23871086537837982,
100
+ "learning_rate": 0.000998450229439693,
101
+ "loss": 0.3225,
102
+ "step": 300
103
+ },
104
+ {
105
+ "epoch": 0.2666119770303528,
106
+ "grad_norm": 0.2566436231136322,
107
+ "learning_rate": 0.000996723037122612,
108
+ "loss": 0.3253,
109
+ "step": 325
110
+ },
111
+ {
112
+ "epoch": 0.2871205906480722,
113
+ "grad_norm": 0.2306467741727829,
114
+ "learning_rate": 0.0009943593799315263,
115
+ "loss": 0.3155,
116
+ "step": 350
117
+ },
118
+ {
119
+ "epoch": 0.30762920426579166,
120
+ "grad_norm": 0.24065160751342773,
121
+ "learning_rate": 0.0009913622864853324,
122
+ "loss": 0.329,
123
+ "step": 375
124
+ },
125
+ {
126
+ "epoch": 0.3281378178835111,
127
+ "grad_norm": 0.26398640871047974,
128
+ "learning_rate": 0.0009877355970422024,
129
+ "loss": 0.3106,
130
+ "step": 400
131
+ },
132
+ {
133
+ "epoch": 0.3281378178835111,
134
+ "eval_loss": 0.2841337323188782,
135
+ "eval_runtime": 19.337,
136
+ "eval_samples_per_second": 51.714,
137
+ "eval_steps_per_second": 0.827,
138
+ "step": 400
139
+ },
140
+ {
141
+ "epoch": 0.34864643150123054,
142
+ "grad_norm": 0.2208418995141983,
143
+ "learning_rate": 0.0009834839585789559,
144
+ "loss": 0.3204,
145
+ "step": 425
146
+ },
147
+ {
148
+ "epoch": 0.36915504511895,
149
+ "grad_norm": 0.3796643018722534,
150
+ "learning_rate": 0.000978612818836762,
151
+ "loss": 0.3153,
152
+ "step": 450
153
+ },
154
+ {
155
+ "epoch": 0.3896636587366694,
156
+ "grad_norm": 0.23547984659671783,
157
+ "learning_rate": 0.0009731284193407981,
158
+ "loss": 0.3021,
159
+ "step": 475
160
+ },
161
+ {
162
+ "epoch": 0.41017227235438886,
163
+ "grad_norm": 0.18797898292541504,
164
+ "learning_rate": 0.0009670377874028117,
165
+ "loss": 0.2888,
166
+ "step": 500
167
+ },
168
+ {
169
+ "epoch": 0.4306808859721083,
170
+ "grad_norm": 0.21823042631149292,
171
+ "learning_rate": 0.0009603487271168336,
172
+ "loss": 0.2876,
173
+ "step": 525
174
+ },
175
+ {
176
+ "epoch": 0.45118949958982774,
177
+ "grad_norm": 0.23094353079795837,
178
+ "learning_rate": 0.0009530698093595781,
179
+ "loss": 0.2981,
180
+ "step": 550
181
+ },
182
+ {
183
+ "epoch": 0.4716981132075472,
184
+ "grad_norm": 0.24919581413269043,
185
+ "learning_rate": 0.0009452103608083418,
186
+ "loss": 0.288,
187
+ "step": 575
188
+ },
189
+ {
190
+ "epoch": 0.4922067268252666,
191
+ "grad_norm": 0.2262776792049408,
192
+ "learning_rate": 0.0009367804519904775,
193
+ "loss": 0.3066,
194
+ "step": 600
195
+ },
196
+ {
197
+ "epoch": 0.4922067268252666,
198
+ "eval_loss": 0.2666381597518921,
199
+ "eval_runtime": 19.3034,
200
+ "eval_samples_per_second": 51.804,
201
+ "eval_steps_per_second": 0.829,
202
+ "step": 600
203
+ },
204
+ {
205
+ "epoch": 0.5127153404429861,
206
+ "grad_norm": 0.2199597805738449,
207
+ "learning_rate": 0.0009277908843797492,
208
+ "loss": 0.283,
209
+ "step": 625
210
+ },
211
+ {
212
+ "epoch": 0.5332239540607056,
213
+ "grad_norm": 0.2111976146697998,
214
+ "learning_rate": 0.0009182531765561084,
215
+ "loss": 0.2757,
216
+ "step": 650
217
+ },
218
+ {
219
+ "epoch": 0.5537325676784249,
220
+ "grad_norm": 0.22439832985401154,
221
+ "learning_rate": 0.0009081795494466201,
222
+ "loss": 0.2871,
223
+ "step": 675
224
+ },
225
+ {
226
+ "epoch": 0.5742411812961444,
227
+ "grad_norm": 0.1950651854276657,
228
+ "learning_rate": 0.0008975829106664539,
229
+ "loss": 0.2862,
230
+ "step": 700
231
+ },
232
+ {
233
+ "epoch": 0.5947497949138638,
234
+ "grad_norm": 0.21456119418144226,
235
+ "learning_rate": 0.0008864768379800017,
236
+ "loss": 0.2826,
237
+ "step": 725
238
+ },
239
+ {
240
+ "epoch": 0.6152584085315833,
241
+ "grad_norm": 0.22092773020267487,
242
+ "learning_rate": 0.0008748755619033153,
243
+ "loss": 0.2803,
244
+ "step": 750
245
+ },
246
+ {
247
+ "epoch": 0.6357670221493027,
248
+ "grad_norm": 0.19732265174388885,
249
+ "learning_rate": 0.000862793947470155,
250
+ "loss": 0.2765,
251
+ "step": 775
252
+ },
253
+ {
254
+ "epoch": 0.6562756357670222,
255
+ "grad_norm": 0.20570369064807892,
256
+ "learning_rate": 0.0008502474751850142,
257
+ "loss": 0.2794,
258
+ "step": 800
259
+ },
260
+ {
261
+ "epoch": 0.6562756357670222,
262
+ "eval_loss": 0.2556016445159912,
263
+ "eval_runtime": 19.3023,
264
+ "eval_samples_per_second": 51.807,
265
+ "eval_steps_per_second": 0.829,
266
+ "step": 800
267
+ },
268
+ {
269
+ "epoch": 0.6767842493847416,
270
+ "grad_norm": 0.20175480842590332,
271
+ "learning_rate": 0.0008372522211875224,
272
+ "loss": 0.2718,
273
+ "step": 825
274
+ },
275
+ {
276
+ "epoch": 0.6972928630024611,
277
+ "grad_norm": 0.19627583026885986,
278
+ "learning_rate": 0.0008238248366536473,
279
+ "loss": 0.2746,
280
+ "step": 850
281
+ },
282
+ {
283
+ "epoch": 0.7178014766201805,
284
+ "grad_norm": 0.18298354744911194,
285
+ "learning_rate": 0.0008099825264600842,
286
+ "loss": 0.2627,
287
+ "step": 875
288
+ },
289
+ {
290
+ "epoch": 0.7383100902379,
291
+ "grad_norm": 0.2175873965024948,
292
+ "learning_rate": 0.0007957430271391761,
293
+ "loss": 0.2677,
294
+ "step": 900
295
+ },
296
+ {
297
+ "epoch": 0.7588187038556193,
298
+ "grad_norm": 0.19386309385299683,
299
+ "learning_rate": 0.0007811245841526062,
300
+ "loss": 0.2701,
301
+ "step": 925
302
+ },
303
+ {
304
+ "epoch": 0.7793273174733388,
305
+ "grad_norm": 0.19052158296108246,
306
+ "learning_rate": 0.0007661459285129879,
307
+ "loss": 0.2574,
308
+ "step": 950
309
+ },
310
+ {
311
+ "epoch": 0.7998359310910582,
312
+ "grad_norm": 0.24320314824581146,
313
+ "learning_rate": 0.0007508262527833029,
314
+ "loss": 0.2643,
315
+ "step": 975
316
+ },
317
+ {
318
+ "epoch": 0.8203445447087777,
319
+ "grad_norm": 0.1941242814064026,
320
+ "learning_rate": 0.000735185186484943,
321
+ "loss": 0.257,
322
+ "step": 1000
323
+ },
324
+ {
325
+ "epoch": 0.8203445447087777,
326
+ "eval_loss": 0.24103032052516937,
327
+ "eval_runtime": 19.3071,
328
+ "eval_samples_per_second": 51.794,
329
+ "eval_steps_per_second": 0.829,
330
+ "step": 1000
331
+ },
332
+ {
333
+ "epoch": 0.8408531583264971,
334
+ "grad_norm": 0.20557403564453125,
335
+ "learning_rate": 0.0007192427709458656,
336
+ "loss": 0.2568,
337
+ "step": 1025
338
+ },
339
+ {
340
+ "epoch": 0.8613617719442166,
341
+ "grad_norm": 0.2422879934310913,
342
+ "learning_rate": 0.0007030194336210887,
343
+ "loss": 0.2537,
344
+ "step": 1050
345
+ },
346
+ {
347
+ "epoch": 0.881870385561936,
348
+ "grad_norm": 0.1749105304479599,
349
+ "learning_rate": 0.0006865359619184331,
350
+ "loss": 0.2591,
351
+ "step": 1075
352
+ },
353
+ {
354
+ "epoch": 0.9023789991796555,
355
+ "grad_norm": 0.19880717992782593,
356
+ "learning_rate": 0.0006698134765630434,
357
+ "loss": 0.2587,
358
+ "step": 1100
359
+ },
360
+ {
361
+ "epoch": 0.9228876127973749,
362
+ "grad_norm": 0.16422006487846375,
363
+ "learning_rate": 0.0006528734045348248,
364
+ "loss": 0.2532,
365
+ "step": 1125
366
+ },
367
+ {
368
+ "epoch": 0.9433962264150944,
369
+ "grad_norm": 0.16592341661453247,
370
+ "learning_rate": 0.0006357374516134643,
371
+ "loss": 0.2524,
372
+ "step": 1150
373
+ },
374
+ {
375
+ "epoch": 0.9639048400328137,
376
+ "grad_norm": 0.18438687920570374,
377
+ "learning_rate": 0.0006184275745662179,
378
+ "loss": 0.2583,
379
+ "step": 1175
380
+ },
381
+ {
382
+ "epoch": 0.9844134536505332,
383
+ "grad_norm": 0.17041641473770142,
384
+ "learning_rate": 0.0006009659530141031,
385
+ "loss": 0.2642,
386
+ "step": 1200
387
+ },
388
+ {
389
+ "epoch": 0.9844134536505332,
390
+ "eval_loss": 0.23156137764453888,
391
+ "eval_runtime": 19.2999,
392
+ "eval_samples_per_second": 51.814,
393
+ "eval_steps_per_second": 0.829,
394
+ "step": 1200
395
+ },
396
+ {
397
+ "epoch": 1.0049220672682526,
398
+ "grad_norm": 0.18642404675483704,
399
+ "learning_rate": 0.0005833749610125402,
400
+ "loss": 0.2442,
401
+ "step": 1225
402
+ },
403
+ {
404
+ "epoch": 1.0254306808859721,
405
+ "grad_norm": 0.16450554132461548,
406
+ "learning_rate": 0.0005656771383828602,
407
+ "loss": 0.2293,
408
+ "step": 1250
409
+ },
410
+ {
411
+ "epoch": 1.0459392945036916,
412
+ "grad_norm": 0.19716863334178925,
413
+ "learning_rate": 0.0005478951618314134,
414
+ "loss": 0.2313,
415
+ "step": 1275
416
+ },
417
+ {
418
+ "epoch": 1.066447908121411,
419
+ "grad_norm": 0.16655565798282623,
420
+ "learning_rate": 0.0005300518158932815,
421
+ "loss": 0.2336,
422
+ "step": 1300
423
+ },
424
+ {
425
+ "epoch": 1.0869565217391304,
426
+ "grad_norm": 0.18710236251354218,
427
+ "learning_rate": 0.0005121699637378282,
428
+ "loss": 0.2286,
429
+ "step": 1325
430
+ },
431
+ {
432
+ "epoch": 1.1074651353568499,
433
+ "grad_norm": 0.15258723497390747,
434
+ "learning_rate": 0.0004942725178734903,
435
+ "loss": 0.2211,
436
+ "step": 1350
437
+ },
438
+ {
439
+ "epoch": 1.1279737489745694,
440
+ "grad_norm": 0.2019195556640625,
441
+ "learning_rate": 0.00047638241078935324,
442
+ "loss": 0.2249,
443
+ "step": 1375
444
+ },
445
+ {
446
+ "epoch": 1.1484823625922886,
447
+ "grad_norm": 0.17765839397907257,
448
+ "learning_rate": 0.000458522565571121,
449
+ "loss": 0.2152,
450
+ "step": 1400
451
+ },
452
+ {
453
+ "epoch": 1.1484823625922886,
454
+ "eval_loss": 0.2268209159374237,
455
+ "eval_runtime": 19.3134,
456
+ "eval_samples_per_second": 51.777,
457
+ "eval_steps_per_second": 0.828,
458
+ "step": 1400
459
+ },
460
+ {
461
+ "epoch": 1.1689909762100081,
462
+ "grad_norm": 0.18442010879516602,
463
+ "learning_rate": 0.00044071586652913767,
464
+ "loss": 0.227,
465
+ "step": 1425
466
+ },
467
+ {
468
+ "epoch": 1.1894995898277276,
469
+ "grad_norm": 0.18356731534004211,
470
+ "learning_rate": 0.0004229851298760915,
471
+ "loss": 0.2282,
472
+ "step": 1450
473
+ },
474
+ {
475
+ "epoch": 1.2100082034454471,
476
+ "grad_norm": 0.1986420750617981,
477
+ "learning_rate": 0.0004053530744919749,
478
+ "loss": 0.2237,
479
+ "step": 1475
480
+ },
481
+ {
482
+ "epoch": 1.2305168170631666,
483
+ "grad_norm": 0.19423463940620422,
484
+ "learning_rate": 0.0003878422928137597,
485
+ "loss": 0.2335,
486
+ "step": 1500
487
+ },
488
+ {
489
+ "epoch": 1.251025430680886,
490
+ "grad_norm": 0.2016667127609253,
491
+ "learning_rate": 0.0003704752218870861,
492
+ "loss": 0.2278,
493
+ "step": 1525
494
+ },
495
+ {
496
+ "epoch": 1.2715340442986054,
497
+ "grad_norm": 0.15723592042922974,
498
+ "learning_rate": 0.00035327411461706025,
499
+ "loss": 0.2209,
500
+ "step": 1550
501
+ },
502
+ {
503
+ "epoch": 1.2920426579163249,
504
+ "grad_norm": 0.16182096302509308,
505
+ "learning_rate": 0.00033626101125499555,
506
+ "loss": 0.2223,
507
+ "step": 1575
508
+ },
509
+ {
510
+ "epoch": 1.3125512715340442,
511
+ "grad_norm": 0.15991301834583282,
512
+ "learning_rate": 0.0003194577111576333,
513
+ "loss": 0.2141,
514
+ "step": 1600
515
+ },
516
+ {
517
+ "epoch": 1.3125512715340442,
518
+ "eval_loss": 0.22021952271461487,
519
+ "eval_runtime": 19.3138,
520
+ "eval_samples_per_second": 51.776,
521
+ "eval_steps_per_second": 0.828,
522
+ "step": 1600
523
+ },
524
+ {
525
+ "epoch": 1.3330598851517639,
526
+ "grad_norm": 0.1668630689382553,
527
+ "learning_rate": 0.00030288574485502756,
528
+ "loss": 0.2188,
529
+ "step": 1625
530
+ },
531
+ {
532
+ "epoch": 1.3535684987694832,
533
+ "grad_norm": 0.17311203479766846,
534
+ "learning_rate": 0.00028656634646288565,
535
+ "loss": 0.2199,
536
+ "step": 1650
537
+ },
538
+ {
539
+ "epoch": 1.3740771123872026,
540
+ "grad_norm": 0.18106333911418915,
541
+ "learning_rate": 0.00027052042647471254,
542
+ "loss": 0.215,
543
+ "step": 1675
544
+ },
545
+ {
546
+ "epoch": 1.3945857260049221,
547
+ "grad_norm": 0.15484604239463806,
548
+ "learning_rate": 0.0002547685449686206,
549
+ "loss": 0.2155,
550
+ "step": 1700
551
+ },
552
+ {
553
+ "epoch": 1.4150943396226414,
554
+ "grad_norm": 0.16587196290493011,
555
+ "learning_rate": 0.0002393308852631373,
556
+ "loss": 0.2137,
557
+ "step": 1725
558
+ },
559
+ {
560
+ "epoch": 1.435602953240361,
561
+ "grad_norm": 0.16898871958255768,
562
+ "learning_rate": 0.0002242272280557645,
563
+ "loss": 0.2138,
564
+ "step": 1750
565
+ },
566
+ {
567
+ "epoch": 1.4561115668580804,
568
+ "grad_norm": 0.16651856899261475,
569
+ "learning_rate": 0.0002094769260774262,
570
+ "loss": 0.2191,
571
+ "step": 1775
572
+ },
573
+ {
574
+ "epoch": 1.4766201804758,
575
+ "grad_norm": 0.1638668179512024,
576
+ "learning_rate": 0.00019509887929528458,
577
+ "loss": 0.209,
578
+ "step": 1800
579
+ },
580
+ {
581
+ "epoch": 1.4766201804758,
582
+ "eval_loss": 0.21638154983520508,
583
+ "eval_runtime": 19.3132,
584
+ "eval_samples_per_second": 51.778,
585
+ "eval_steps_per_second": 0.828,
586
+ "step": 1800
587
+ },
588
+ {
589
+ "epoch": 1.4971287940935194,
590
+ "grad_norm": 0.1516977846622467,
591
+ "learning_rate": 0.0001811115106956918,
592
+ "loss": 0.2074,
593
+ "step": 1825
594
+ },
595
+ {
596
+ "epoch": 1.5176374077112387,
597
+ "grad_norm": 0.1567605584859848,
598
+ "learning_rate": 0.00016753274267831115,
599
+ "loss": 0.211,
600
+ "step": 1850
601
+ },
602
+ {
603
+ "epoch": 1.5381460213289582,
604
+ "grad_norm": 0.15604251623153687,
605
+ "learning_rate": 0.00015437997409165478,
606
+ "loss": 0.2123,
607
+ "step": 1875
608
+ },
609
+ {
610
+ "epoch": 1.5586546349466777,
611
+ "grad_norm": 0.1690610945224762,
612
+ "learning_rate": 0.00014167005793946035,
613
+ "loss": 0.2141,
614
+ "step": 1900
615
+ },
616
+ {
617
+ "epoch": 1.579163248564397,
618
+ "grad_norm": 0.17786969244480133,
619
+ "learning_rate": 0.00012941927978647527,
620
+ "loss": 0.2157,
621
+ "step": 1925
622
+ },
623
+ {
624
+ "epoch": 1.5996718621821167,
625
+ "grad_norm": 0.16428180038928986,
626
+ "learning_rate": 0.00011764333689131385,
627
+ "loss": 0.2146,
628
+ "step": 1950
629
+ },
630
+ {
631
+ "epoch": 1.620180475799836,
632
+ "grad_norm": 0.15552306175231934,
633
+ "learning_rate": 0.00010635731809312993,
634
+ "loss": 0.2084,
635
+ "step": 1975
636
+ },
637
+ {
638
+ "epoch": 1.6406890894175554,
639
+ "grad_norm": 0.1638088822364807,
640
+ "learning_rate": 9.557568447787201e-05,
641
+ "loss": 0.2156,
642
+ "step": 2000
643
+ },
644
+ {
645
+ "epoch": 1.6406890894175554,
646
+ "eval_loss": 0.21138574182987213,
647
+ "eval_runtime": 19.294,
648
+ "eval_samples_per_second": 51.83,
649
+ "eval_steps_per_second": 0.829,
650
+ "step": 2000
651
+ },
652
+ {
653
+ "epoch": 1.661197703035275,
654
+ "grad_norm": 0.15038058161735535,
655
+ "learning_rate": 8.531225084889654e-05,
656
+ "loss": 0.2053,
657
+ "step": 2025
658
+ },
659
+ {
660
+ "epoch": 1.6817063166529942,
661
+ "grad_norm": 0.18371206521987915,
662
+ "learning_rate": 7.558016802568091e-05,
663
+ "loss": 0.208,
664
+ "step": 2050
665
+ },
666
+ {
667
+ "epoch": 1.7022149302707137,
668
+ "grad_norm": 0.19614985585212708,
669
+ "learning_rate": 6.639190599331746e-05,
670
+ "loss": 0.2136,
671
+ "step": 2075
672
+ },
673
+ {
674
+ "epoch": 1.7227235438884332,
675
+ "grad_norm": 0.16417619585990906,
676
+ "learning_rate": 5.775923792437865e-05,
677
+ "loss": 0.2044,
678
+ "step": 2100
679
+ },
680
+ {
681
+ "epoch": 1.7432321575061525,
682
+ "grad_norm": 0.15051694214344025,
683
+ "learning_rate": 4.9693225093627616e-05,
684
+ "loss": 0.2077,
685
+ "step": 2125
686
+ },
687
+ {
688
+ "epoch": 1.7637407711238722,
689
+ "grad_norm": 0.1453014612197876,
690
+ "learning_rate": 4.220420270490294e-05,
691
+ "loss": 0.2053,
692
+ "step": 2150
693
+ },
694
+ {
695
+ "epoch": 1.7842493847415914,
696
+ "grad_norm": 0.15306603908538818,
697
+ "learning_rate": 3.530176664833834e-05,
698
+ "loss": 0.2151,
699
+ "step": 2175
700
+ },
701
+ {
702
+ "epoch": 1.804757998359311,
703
+ "grad_norm": 0.14955438673496246,
704
+ "learning_rate": 2.8994761204884756e-05,
705
+ "loss": 0.2073,
706
+ "step": 2200
707
+ },
708
+ {
709
+ "epoch": 1.804757998359311,
710
+ "eval_loss": 0.20938299596309662,
711
+ "eval_runtime": 19.293,
712
+ "eval_samples_per_second": 51.832,
713
+ "eval_steps_per_second": 0.829,
714
+ "step": 2200
715
+ },
716
+ {
717
+ "epoch": 1.8252666119770304,
718
+ "grad_norm": 0.1693127155303955,
719
+ "learning_rate": 2.329126771388995e-05,
720
+ "loss": 0.2081,
721
+ "step": 2225
722
+ },
723
+ {
724
+ "epoch": 1.8457752255947497,
725
+ "grad_norm": 0.15966977179050446,
726
+ "learning_rate": 1.8198594218256815e-05,
727
+ "loss": 0.2097,
728
+ "step": 2250
729
+ },
730
+ {
731
+ "epoch": 1.8662838392124692,
732
+ "grad_norm": 0.15011626482009888,
733
+ "learning_rate": 1.3723266100447052e-05,
734
+ "loss": 0.2147,
735
+ "step": 2275
736
+ },
737
+ {
738
+ "epoch": 1.8867924528301887,
739
+ "grad_norm": 0.14668996632099152,
740
+ "learning_rate": 9.871017721329201e-06,
741
+ "loss": 0.2086,
742
+ "step": 2300
743
+ },
744
+ {
745
+ "epoch": 1.907301066447908,
746
+ "grad_norm": 0.1699514389038086,
747
+ "learning_rate": 6.646785072584871e-06,
748
+ "loss": 0.2165,
749
+ "step": 2325
750
+ },
751
+ {
752
+ "epoch": 1.9278096800656277,
753
+ "grad_norm": 0.15601767599582672,
754
+ "learning_rate": 4.054699452086641e-06,
755
+ "loss": 0.1986,
756
+ "step": 2350
757
+ },
758
+ {
759
+ "epoch": 1.948318293683347,
760
+ "grad_norm": 0.14943112432956696,
761
+ "learning_rate": 2.0980821703527886e-06,
762
+ "loss": 0.205,
763
+ "step": 2375
764
+ },
765
+ {
766
+ "epoch": 1.9688269073010665,
767
+ "grad_norm": 0.15471364557743073,
768
+ "learning_rate": 7.794402948607671e-07,
769
+ "loss": 0.2078,
770
+ "step": 2400
771
+ },
772
+ {
773
+ "epoch": 1.9688269073010665,
774
+ "eval_loss": 0.20906180143356323,
775
+ "eval_runtime": 19.3168,
776
+ "eval_samples_per_second": 51.768,
777
+ "eval_steps_per_second": 0.828,
778
+ "step": 2400
779
+ },
780
+ {
781
+ "epoch": 1.989335520918786,
782
+ "grad_norm": 0.15639156103134155,
783
+ "learning_rate": 1.0046343767294853e-07,
784
+ "loss": 0.2032,
785
+ "step": 2425
786
+ },
787
+ {
788
+ "epoch": 2.0,
789
+ "step": 2438,
790
+ "total_flos": 1.58523627405312e+18,
791
+ "train_loss": 0.26224943425833347,
792
+ "train_runtime": 2666.0866,
793
+ "train_samples_per_second": 29.256,
794
+ "train_steps_per_second": 0.914
795
+ }
796
+ ],
797
+ "logging_steps": 25,
798
+ "max_steps": 2438,
799
+ "num_input_tokens_seen": 0,
800
+ "num_train_epochs": 2,
801
+ "save_steps": 0,
802
+ "stateful_callbacks": {
803
+ "TrainerControl": {
804
+ "args": {
805
+ "should_epoch_stop": false,
806
+ "should_evaluate": false,
807
+ "should_log": false,
808
+ "should_save": true,
809
+ "should_training_stop": true
810
+ },
811
+ "attributes": {}
812
+ }
813
+ },
814
+ "total_flos": 1.58523627405312e+18,
815
+ "train_batch_size": 32,
816
+ "trial_name": null,
817
+ "trial_params": null
818
+ }
nl_tasks/exps/run_ex19/ft/adapter_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "inference_mode": false,
6
+ "layers_to_transform": null,
7
+ "modules_to_save": null,
8
+ "num_rotations": 1,
9
+ "peft_type": "ROTATION",
10
+ "r": 16,
11
+ "revision": null,
12
+ "target_modules": [
13
+ "v_proj",
14
+ "q_proj"
15
+ ],
16
+ "target_modules_to_skip": null,
17
+ "task_type": "CAUSAL_LM"
18
+ }
nl_tasks/exps/run_ex19/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/exps/run_ex19/ft/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
nl_tasks/exps/run_ex19/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/exps/run_ex19/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/exps/run_ex19/ft2/adapter_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "inference_mode": true,
6
+ "layers_to_transform": null,
7
+ "modules_to_save": null,
8
+ "num_rotations": 1,
9
+ "peft_type": "ROTATION",
10
+ "r": 16,
11
+ "revision": null,
12
+ "target_modules": [
13
+ "v_proj",
14
+ "q_proj"
15
+ ],
16
+ "target_modules_to_skip": null,
17
+ "task_type": "CAUSAL_LM"
18
+ }
nl_tasks/exps/run_ex19/ft2/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cad4979930182e59cfa6282f33803329c4b46ad70d215a6128c14752c6f104d5
3
+ size 33602915
nl_tasks/exps/run_ex19/trainer_state.json ADDED
@@ -0,0 +1,1149 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 3.0,
6
+ "eval_steps": 500,
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": 0.5312772393226624,
15
+ "learning_rate": 0.000128,
16
+ "loss": 0.7982,
17
+ "step": 25
18
+ },
19
+ {
20
+ "epoch": 0.04,
21
+ "grad_norm": 0.260868102312088,
22
+ "learning_rate": 0.0002613333333333333,
23
+ "loss": 0.4362,
24
+ "step": 50
25
+ },
26
+ {
27
+ "epoch": 0.06,
28
+ "grad_norm": 0.35636189579963684,
29
+ "learning_rate": 0.0003946666666666667,
30
+ "loss": 0.3934,
31
+ "step": 75
32
+ },
33
+ {
34
+ "epoch": 0.08,
35
+ "grad_norm": 0.17371909320354462,
36
+ "learning_rate": 0.000528,
37
+ "loss": 0.3598,
38
+ "step": 100
39
+ },
40
+ {
41
+ "epoch": 0.1,
42
+ "grad_norm": 0.3316778242588043,
43
+ "learning_rate": 0.0006613333333333333,
44
+ "loss": 0.3567,
45
+ "step": 125
46
+ },
47
+ {
48
+ "epoch": 0.12,
49
+ "grad_norm": 0.2653660774230957,
50
+ "learning_rate": 0.0007946666666666666,
51
+ "loss": 0.3452,
52
+ "step": 150
53
+ },
54
+ {
55
+ "epoch": 0.14,
56
+ "grad_norm": 0.22913236916065216,
57
+ "learning_rate": 0.0009280000000000001,
58
+ "loss": 0.3448,
59
+ "step": 175
60
+ },
61
+ {
62
+ "epoch": 0.16,
63
+ "grad_norm": 0.22667862474918365,
64
+ "learning_rate": 0.0010613333333333332,
65
+ "loss": 0.3358,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.18,
70
+ "grad_norm": 0.2627682387828827,
71
+ "learning_rate": 0.0011946666666666668,
72
+ "loss": 0.3385,
73
+ "step": 225
74
+ },
75
+ {
76
+ "epoch": 0.2,
77
+ "grad_norm": 0.3162098824977875,
78
+ "learning_rate": 0.0013280000000000002,
79
+ "loss": 0.3478,
80
+ "step": 250
81
+ },
82
+ {
83
+ "epoch": 0.22,
84
+ "grad_norm": 0.3067827820777893,
85
+ "learning_rate": 0.0014613333333333334,
86
+ "loss": 0.3414,
87
+ "step": 275
88
+ },
89
+ {
90
+ "epoch": 0.24,
91
+ "grad_norm": 0.23997268080711365,
92
+ "learning_rate": 0.0015946666666666668,
93
+ "loss": 0.3377,
94
+ "step": 300
95
+ },
96
+ {
97
+ "epoch": 0.26,
98
+ "grad_norm": 0.3886880874633789,
99
+ "learning_rate": 0.001728,
100
+ "loss": 0.3381,
101
+ "step": 325
102
+ },
103
+ {
104
+ "epoch": 0.28,
105
+ "grad_norm": 0.3570641279220581,
106
+ "learning_rate": 0.0018613333333333333,
107
+ "loss": 0.3364,
108
+ "step": 350
109
+ },
110
+ {
111
+ "epoch": 0.3,
112
+ "grad_norm": 0.32272225618362427,
113
+ "learning_rate": 0.0019946666666666667,
114
+ "loss": 0.4012,
115
+ "step": 375
116
+ },
117
+ {
118
+ "epoch": 0.32,
119
+ "grad_norm": 0.3963306248188019,
120
+ "learning_rate": 0.0019997504677881223,
121
+ "loss": 0.3541,
122
+ "step": 400
123
+ },
124
+ {
125
+ "epoch": 0.34,
126
+ "grad_norm": 1.8844726085662842,
127
+ "learning_rate": 0.001998959986294895,
128
+ "loss": 0.323,
129
+ "step": 425
130
+ },
131
+ {
132
+ "epoch": 0.36,
133
+ "grad_norm": 0.21251291036605835,
134
+ "learning_rate": 0.001997628550346359,
135
+ "loss": 0.3313,
136
+ "step": 450
137
+ },
138
+ {
139
+ "epoch": 0.38,
140
+ "grad_norm": 0.2612377405166626,
141
+ "learning_rate": 0.0019957568809385695,
142
+ "loss": 0.3189,
143
+ "step": 475
144
+ },
145
+ {
146
+ "epoch": 0.4,
147
+ "grad_norm": 0.18152573704719543,
148
+ "learning_rate": 0.0019933459916135275,
149
+ "loss": 0.3032,
150
+ "step": 500
151
+ },
152
+ {
153
+ "epoch": 0.4,
154
+ "eval_loss": 0.2950391173362732,
155
+ "eval_runtime": 18.9126,
156
+ "eval_samples_per_second": 54.091,
157
+ "eval_steps_per_second": 0.846,
158
+ "step": 500
159
+ },
160
+ {
161
+ "epoch": 0.42,
162
+ "grad_norm": 0.1786627620458603,
163
+ "learning_rate": 0.0019903971879103345,
164
+ "loss": 0.3023,
165
+ "step": 525
166
+ },
167
+ {
168
+ "epoch": 0.44,
169
+ "grad_norm": 0.16251839697360992,
170
+ "learning_rate": 0.0019869120666582154,
171
+ "loss": 0.2824,
172
+ "step": 550
173
+ },
174
+ {
175
+ "epoch": 0.46,
176
+ "grad_norm": 0.18590645492076874,
177
+ "learning_rate": 0.001982892515111809,
178
+ "loss": 0.291,
179
+ "step": 575
180
+ },
181
+ {
182
+ "epoch": 0.48,
183
+ "grad_norm": 0.17091675102710724,
184
+ "learning_rate": 0.001978340709929188,
185
+ "loss": 0.2847,
186
+ "step": 600
187
+ },
188
+ {
189
+ "epoch": 0.5,
190
+ "grad_norm": 0.168325737118721,
191
+ "learning_rate": 0.0019732591159931563,
192
+ "loss": 0.2827,
193
+ "step": 625
194
+ },
195
+ {
196
+ "epoch": 0.52,
197
+ "grad_norm": 0.15048232674598694,
198
+ "learning_rate": 0.001967650485076476,
199
+ "loss": 0.2876,
200
+ "step": 650
201
+ },
202
+ {
203
+ "epoch": 0.54,
204
+ "grad_norm": 0.178253635764122,
205
+ "learning_rate": 0.0019615178543517303,
206
+ "loss": 0.2953,
207
+ "step": 675
208
+ },
209
+ {
210
+ "epoch": 0.56,
211
+ "grad_norm": 0.15488530695438385,
212
+ "learning_rate": 0.001954864544746643,
213
+ "loss": 0.2824,
214
+ "step": 700
215
+ },
216
+ {
217
+ "epoch": 0.58,
218
+ "grad_norm": 0.3576168715953827,
219
+ "learning_rate": 0.001947694159145737,
220
+ "loss": 0.2915,
221
+ "step": 725
222
+ },
223
+ {
224
+ "epoch": 0.6,
225
+ "grad_norm": 0.15641844272613525,
226
+ "learning_rate": 0.001940010580439308,
227
+ "loss": 0.2858,
228
+ "step": 750
229
+ },
230
+ {
231
+ "epoch": 0.62,
232
+ "grad_norm": 0.14580459892749786,
233
+ "learning_rate": 0.0019318179694207726,
234
+ "loss": 0.268,
235
+ "step": 775
236
+ },
237
+ {
238
+ "epoch": 0.64,
239
+ "grad_norm": 0.1843678057193756,
240
+ "learning_rate": 0.0019231207625335235,
241
+ "loss": 0.2799,
242
+ "step": 800
243
+ },
244
+ {
245
+ "epoch": 0.66,
246
+ "grad_norm": 0.14439673721790314,
247
+ "learning_rate": 0.0019139236694685185,
248
+ "loss": 0.2669,
249
+ "step": 825
250
+ },
251
+ {
252
+ "epoch": 0.68,
253
+ "grad_norm": 0.1365605741739273,
254
+ "learning_rate": 0.0019042316706138988,
255
+ "loss": 0.2717,
256
+ "step": 850
257
+ },
258
+ {
259
+ "epoch": 0.7,
260
+ "grad_norm": 0.13377751410007477,
261
+ "learning_rate": 0.0018940500143580215,
262
+ "loss": 0.2741,
263
+ "step": 875
264
+ },
265
+ {
266
+ "epoch": 0.72,
267
+ "grad_norm": 0.12897183001041412,
268
+ "learning_rate": 0.0018833842142473643,
269
+ "loss": 0.2807,
270
+ "step": 900
271
+ },
272
+ {
273
+ "epoch": 0.74,
274
+ "grad_norm": 0.11778058111667633,
275
+ "learning_rate": 0.0018722400460008438,
276
+ "loss": 0.2624,
277
+ "step": 925
278
+ },
279
+ {
280
+ "epoch": 0.76,
281
+ "grad_norm": 0.13213050365447998,
282
+ "learning_rate": 0.0018606235443821603,
283
+ "loss": 0.2733,
284
+ "step": 950
285
+ },
286
+ {
287
+ "epoch": 0.78,
288
+ "grad_norm": 0.11900192499160767,
289
+ "learning_rate": 0.0018485409999318679,
290
+ "loss": 0.2705,
291
+ "step": 975
292
+ },
293
+ {
294
+ "epoch": 0.8,
295
+ "grad_norm": 0.11382659524679184,
296
+ "learning_rate": 0.0018359989555609354,
297
+ "loss": 0.2715,
298
+ "step": 1000
299
+ },
300
+ {
301
+ "epoch": 0.8,
302
+ "eval_loss": 0.2514421045780182,
303
+ "eval_runtime": 18.7269,
304
+ "eval_samples_per_second": 54.627,
305
+ "eval_steps_per_second": 0.854,
306
+ "step": 1000
307
+ },
308
+ {
309
+ "epoch": 0.82,
310
+ "grad_norm": 0.1073298379778862,
311
+ "learning_rate": 0.0018230042030076452,
312
+ "loss": 0.254,
313
+ "step": 1025
314
+ },
315
+ {
316
+ "epoch": 0.84,
317
+ "grad_norm": 0.15638203918933868,
318
+ "learning_rate": 0.0018095637791597462,
319
+ "loss": 0.2639,
320
+ "step": 1050
321
+ },
322
+ {
323
+ "epoch": 0.86,
324
+ "grad_norm": 0.13015539944171906,
325
+ "learning_rate": 0.0017956849622438554,
326
+ "loss": 0.2581,
327
+ "step": 1075
328
+ },
329
+ {
330
+ "epoch": 0.88,
331
+ "grad_norm": 0.11949340254068375,
332
+ "learning_rate": 0.0017813752678841701,
333
+ "loss": 0.2608,
334
+ "step": 1100
335
+ },
336
+ {
337
+ "epoch": 0.9,
338
+ "grad_norm": 0.13293173909187317,
339
+ "learning_rate": 0.0017666424450326248,
340
+ "loss": 0.2747,
341
+ "step": 1125
342
+ },
343
+ {
344
+ "epoch": 0.92,
345
+ "grad_norm": 0.13074743747711182,
346
+ "learning_rate": 0.0017514944717726961,
347
+ "loss": 0.2503,
348
+ "step": 1150
349
+ },
350
+ {
351
+ "epoch": 0.94,
352
+ "grad_norm": 0.10643952339887619,
353
+ "learning_rate": 0.001735939550999131,
354
+ "loss": 0.2562,
355
+ "step": 1175
356
+ },
357
+ {
358
+ "epoch": 0.96,
359
+ "grad_norm": 0.11634570360183716,
360
+ "learning_rate": 0.0017199861059759337,
361
+ "loss": 0.2431,
362
+ "step": 1200
363
+ },
364
+ {
365
+ "epoch": 0.98,
366
+ "grad_norm": 0.1033262088894844,
367
+ "learning_rate": 0.0017036427757750204,
368
+ "loss": 0.2548,
369
+ "step": 1225
370
+ },
371
+ {
372
+ "epoch": 1.0,
373
+ "grad_norm": 0.11224926263093948,
374
+ "learning_rate": 0.0016869184105980089,
375
+ "loss": 0.2474,
376
+ "step": 1250
377
+ },
378
+ {
379
+ "epoch": 1.02,
380
+ "grad_norm": 0.11112752556800842,
381
+ "learning_rate": 0.0016698220669836782,
382
+ "loss": 0.2237,
383
+ "step": 1275
384
+ },
385
+ {
386
+ "epoch": 1.04,
387
+ "grad_norm": 0.10078830271959305,
388
+ "learning_rate": 0.0016523630029036931,
389
+ "loss": 0.2191,
390
+ "step": 1300
391
+ },
392
+ {
393
+ "epoch": 1.06,
394
+ "grad_norm": 0.1166379451751709,
395
+ "learning_rate": 0.00163455067274925,
396
+ "loss": 0.2438,
397
+ "step": 1325
398
+ },
399
+ {
400
+ "epoch": 1.08,
401
+ "grad_norm": 0.10196326673030853,
402
+ "learning_rate": 0.0016163947222113566,
403
+ "loss": 0.2165,
404
+ "step": 1350
405
+ },
406
+ {
407
+ "epoch": 1.1,
408
+ "grad_norm": 0.11595246195793152,
409
+ "learning_rate": 0.0015979049830575189,
410
+ "loss": 0.2368,
411
+ "step": 1375
412
+ },
413
+ {
414
+ "epoch": 1.12,
415
+ "grad_norm": 0.10186369717121124,
416
+ "learning_rate": 0.001579091467807668,
417
+ "loss": 0.2324,
418
+ "step": 1400
419
+ },
420
+ {
421
+ "epoch": 1.1400000000000001,
422
+ "grad_norm": 0.10030511766672134,
423
+ "learning_rate": 0.0015599643643122022,
424
+ "loss": 0.2242,
425
+ "step": 1425
426
+ },
427
+ {
428
+ "epoch": 1.16,
429
+ "grad_norm": 0.10598330944776535,
430
+ "learning_rate": 0.0015405340302350869,
431
+ "loss": 0.2354,
432
+ "step": 1450
433
+ },
434
+ {
435
+ "epoch": 1.18,
436
+ "grad_norm": 0.11982858180999756,
437
+ "learning_rate": 0.0015208109874449978,
438
+ "loss": 0.2271,
439
+ "step": 1475
440
+ },
441
+ {
442
+ "epoch": 1.2,
443
+ "grad_norm": 0.10632237792015076,
444
+ "learning_rate": 0.00150080591631754,
445
+ "loss": 0.2312,
446
+ "step": 1500
447
+ },
448
+ {
449
+ "epoch": 1.2,
450
+ "eval_loss": 0.2400166392326355,
451
+ "eval_runtime": 18.7225,
452
+ "eval_samples_per_second": 54.64,
453
+ "eval_steps_per_second": 0.855,
454
+ "step": 1500
455
+ },
456
+ {
457
+ "epoch": 1.22,
458
+ "grad_norm": 0.09122207015752792,
459
+ "learning_rate": 0.0014805296499516406,
460
+ "loss": 0.229,
461
+ "step": 1525
462
+ },
463
+ {
464
+ "epoch": 1.24,
465
+ "grad_norm": 0.0960044115781784,
466
+ "learning_rate": 0.0014599931683032326,
467
+ "loss": 0.2244,
468
+ "step": 1550
469
+ },
470
+ {
471
+ "epoch": 1.26,
472
+ "grad_norm": 0.12922610342502594,
473
+ "learning_rate": 0.0014392075922394132,
474
+ "loss": 0.2299,
475
+ "step": 1575
476
+ },
477
+ {
478
+ "epoch": 1.28,
479
+ "grad_norm": 0.10358211398124695,
480
+ "learning_rate": 0.0014181841775163014,
481
+ "loss": 0.2342,
482
+ "step": 1600
483
+ },
484
+ {
485
+ "epoch": 1.3,
486
+ "grad_norm": 0.11759316176176071,
487
+ "learning_rate": 0.0013969343086838457,
488
+ "loss": 0.2159,
489
+ "step": 1625
490
+ },
491
+ {
492
+ "epoch": 1.32,
493
+ "grad_norm": 0.1028522178530693,
494
+ "learning_rate": 0.001375469492920889,
495
+ "loss": 0.241,
496
+ "step": 1650
497
+ },
498
+ {
499
+ "epoch": 1.34,
500
+ "grad_norm": 0.08108657598495483,
501
+ "learning_rate": 0.0013538013538038295,
502
+ "loss": 0.2211,
503
+ "step": 1675
504
+ },
505
+ {
506
+ "epoch": 1.3599999999999999,
507
+ "grad_norm": 0.08613426983356476,
508
+ "learning_rate": 0.0013319416250122482,
509
+ "loss": 0.2282,
510
+ "step": 1700
511
+ },
512
+ {
513
+ "epoch": 1.38,
514
+ "grad_norm": 0.09871073067188263,
515
+ "learning_rate": 0.0013099021439749155,
516
+ "loss": 0.2327,
517
+ "step": 1725
518
+ },
519
+ {
520
+ "epoch": 1.4,
521
+ "grad_norm": 0.11070626229047775,
522
+ "learning_rate": 0.001287694845459613,
523
+ "loss": 0.2293,
524
+ "step": 1750
525
+ },
526
+ {
527
+ "epoch": 1.42,
528
+ "grad_norm": 0.0942375436425209,
529
+ "learning_rate": 0.001265331755110247,
530
+ "loss": 0.2227,
531
+ "step": 1775
532
+ },
533
+ {
534
+ "epoch": 1.44,
535
+ "grad_norm": 0.11163079738616943,
536
+ "learning_rate": 0.001242824982934751,
537
+ "loss": 0.2314,
538
+ "step": 1800
539
+ },
540
+ {
541
+ "epoch": 1.46,
542
+ "grad_norm": 0.11482635140419006,
543
+ "learning_rate": 0.0012201867167473015,
544
+ "loss": 0.23,
545
+ "step": 1825
546
+ },
547
+ {
548
+ "epoch": 1.48,
549
+ "grad_norm": 0.0959840789437294,
550
+ "learning_rate": 0.001197429215568403,
551
+ "loss": 0.2149,
552
+ "step": 1850
553
+ },
554
+ {
555
+ "epoch": 1.5,
556
+ "grad_norm": 0.09649599343538284,
557
+ "learning_rate": 0.00117456480298641,
558
+ "loss": 0.2272,
559
+ "step": 1875
560
+ },
561
+ {
562
+ "epoch": 1.52,
563
+ "grad_norm": 0.09136765450239182,
564
+ "learning_rate": 0.0011516058604840891,
565
+ "loss": 0.2214,
566
+ "step": 1900
567
+ },
568
+ {
569
+ "epoch": 1.54,
570
+ "grad_norm": 0.09074031561613083,
571
+ "learning_rate": 0.001128564820733825,
572
+ "loss": 0.2243,
573
+ "step": 1925
574
+ },
575
+ {
576
+ "epoch": 1.56,
577
+ "grad_norm": 0.07815123349428177,
578
+ "learning_rate": 0.001105454160865112,
579
+ "loss": 0.2127,
580
+ "step": 1950
581
+ },
582
+ {
583
+ "epoch": 1.58,
584
+ "grad_norm": 0.09617511183023453,
585
+ "learning_rate": 0.0010822863957079655,
586
+ "loss": 0.2193,
587
+ "step": 1975
588
+ },
589
+ {
590
+ "epoch": 1.6,
591
+ "grad_norm": 0.11400031298398972,
592
+ "learning_rate": 0.0010590740710159229,
593
+ "loss": 0.2229,
594
+ "step": 2000
595
+ },
596
+ {
597
+ "epoch": 1.6,
598
+ "eval_loss": 0.22801487147808075,
599
+ "eval_runtime": 18.7228,
600
+ "eval_samples_per_second": 54.639,
601
+ "eval_steps_per_second": 0.855,
602
+ "step": 2000
603
+ },
604
+ {
605
+ "epoch": 1.62,
606
+ "grad_norm": 0.09799627214670181,
607
+ "learning_rate": 0.0010358297566722945,
608
+ "loss": 0.2112,
609
+ "step": 2025
610
+ },
611
+ {
612
+ "epoch": 1.6400000000000001,
613
+ "grad_norm": 0.09368785470724106,
614
+ "learning_rate": 0.0010125660398833528,
615
+ "loss": 0.2203,
616
+ "step": 2050
617
+ },
618
+ {
619
+ "epoch": 1.6600000000000001,
620
+ "grad_norm": 0.09056692570447922,
621
+ "learning_rate": 0.0009892955183621354,
622
+ "loss": 0.2241,
623
+ "step": 2075
624
+ },
625
+ {
626
+ "epoch": 1.6800000000000002,
627
+ "grad_norm": 0.1016368642449379,
628
+ "learning_rate": 0.0009660307935065639,
629
+ "loss": 0.2265,
630
+ "step": 2100
631
+ },
632
+ {
633
+ "epoch": 1.7,
634
+ "grad_norm": 0.11028288304805756,
635
+ "learning_rate": 0.000942784463575562,
636
+ "loss": 0.2247,
637
+ "step": 2125
638
+ },
639
+ {
640
+ "epoch": 1.72,
641
+ "grad_norm": 0.08315608650445938,
642
+ "learning_rate": 0.0009195691168668773,
643
+ "loss": 0.2187,
644
+ "step": 2150
645
+ },
646
+ {
647
+ "epoch": 1.74,
648
+ "grad_norm": 0.07800303399562836,
649
+ "learning_rate": 0.0008963973249002967,
650
+ "loss": 0.214,
651
+ "step": 2175
652
+ },
653
+ {
654
+ "epoch": 1.76,
655
+ "grad_norm": 0.09412352740764618,
656
+ "learning_rate": 0.0008732816356099456,
657
+ "loss": 0.2179,
658
+ "step": 2200
659
+ },
660
+ {
661
+ "epoch": 1.78,
662
+ "grad_norm": 0.09773926436901093,
663
+ "learning_rate": 0.0008502345665493598,
664
+ "loss": 0.2217,
665
+ "step": 2225
666
+ },
667
+ {
668
+ "epoch": 1.8,
669
+ "grad_norm": 0.08439651131629944,
670
+ "learning_rate": 0.0008272685981130109,
671
+ "loss": 0.2148,
672
+ "step": 2250
673
+ },
674
+ {
675
+ "epoch": 1.8199999999999998,
676
+ "grad_norm": 0.0858495905995369,
677
+ "learning_rate": 0.0008043961667779519,
678
+ "loss": 0.2126,
679
+ "step": 2275
680
+ },
681
+ {
682
+ "epoch": 1.8399999999999999,
683
+ "grad_norm": 0.09474071115255356,
684
+ "learning_rate": 0.0007816296583692448,
685
+ "loss": 0.2072,
686
+ "step": 2300
687
+ },
688
+ {
689
+ "epoch": 1.8599999999999999,
690
+ "grad_norm": 0.0823824405670166,
691
+ "learning_rate": 0.0007589814013528185,
692
+ "loss": 0.2188,
693
+ "step": 2325
694
+ },
695
+ {
696
+ "epoch": 1.88,
697
+ "grad_norm": 0.08434902131557465,
698
+ "learning_rate": 0.0007364636601593875,
699
+ "loss": 0.2054,
700
+ "step": 2350
701
+ },
702
+ {
703
+ "epoch": 1.9,
704
+ "grad_norm": 0.09648870676755905,
705
+ "learning_rate": 0.0007140886285430447,
706
+ "loss": 0.2118,
707
+ "step": 2375
708
+ },
709
+ {
710
+ "epoch": 1.92,
711
+ "grad_norm": 0.09174276888370514,
712
+ "learning_rate": 0.0006918684229781305,
713
+ "loss": 0.2101,
714
+ "step": 2400
715
+ },
716
+ {
717
+ "epoch": 1.94,
718
+ "grad_norm": 0.08786734938621521,
719
+ "learning_rate": 0.0006698150760979462,
720
+ "loss": 0.2118,
721
+ "step": 2425
722
+ },
723
+ {
724
+ "epoch": 1.96,
725
+ "grad_norm": 0.08638143539428711,
726
+ "learning_rate": 0.0006479405301788728,
727
+ "loss": 0.1985,
728
+ "step": 2450
729
+ },
730
+ {
731
+ "epoch": 1.98,
732
+ "grad_norm": 0.0800042524933815,
733
+ "learning_rate": 0.0006262566306734155,
734
+ "loss": 0.2081,
735
+ "step": 2475
736
+ },
737
+ {
738
+ "epoch": 2.0,
739
+ "grad_norm": 0.08018375933170319,
740
+ "learning_rate": 0.0006047751197956838,
741
+ "loss": 0.2093,
742
+ "step": 2500
743
+ },
744
+ {
745
+ "epoch": 2.0,
746
+ "eval_loss": 0.21793007850646973,
747
+ "eval_runtime": 18.7244,
748
+ "eval_samples_per_second": 54.634,
749
+ "eval_steps_per_second": 0.854,
750
+ "step": 2500
751
+ },
752
+ {
753
+ "epoch": 2.02,
754
+ "grad_norm": 0.08357514441013336,
755
+ "learning_rate": 0.0005835076301627752,
756
+ "loss": 0.1784,
757
+ "step": 2525
758
+ },
759
+ {
760
+ "epoch": 2.04,
761
+ "grad_norm": 0.09539826959371567,
762
+ "learning_rate": 0.0005624656784955071,
763
+ "loss": 0.1823,
764
+ "step": 2550
765
+ },
766
+ {
767
+ "epoch": 2.06,
768
+ "grad_norm": 0.09116104990243912,
769
+ "learning_rate": 0.0005416606593819101,
770
+ "loss": 0.186,
771
+ "step": 2575
772
+ },
773
+ {
774
+ "epoch": 2.08,
775
+ "grad_norm": 0.08753357082605362,
776
+ "learning_rate": 0.0005211038391068576,
777
+ "loss": 0.1847,
778
+ "step": 2600
779
+ },
780
+ {
781
+ "epoch": 2.1,
782
+ "grad_norm": 0.08703944832086563,
783
+ "learning_rate": 0.0005008063495511723,
784
+ "loss": 0.1885,
785
+ "step": 2625
786
+ },
787
+ {
788
+ "epoch": 2.12,
789
+ "grad_norm": 0.0975218117237091,
790
+ "learning_rate": 0.00048077918216351856,
791
+ "loss": 0.184,
792
+ "step": 2650
793
+ },
794
+ {
795
+ "epoch": 2.14,
796
+ "grad_norm": 0.08066797256469727,
797
+ "learning_rate": 0.0004610331820083378,
798
+ "loss": 0.1782,
799
+ "step": 2675
800
+ },
801
+ {
802
+ "epoch": 2.16,
803
+ "grad_norm": 0.12133917212486267,
804
+ "learning_rate": 0.00044157904189305407,
805
+ "loss": 0.1857,
806
+ "step": 2700
807
+ },
808
+ {
809
+ "epoch": 2.18,
810
+ "grad_norm": 0.08209951967000961,
811
+ "learning_rate": 0.0004224272965777326,
812
+ "loss": 0.1839,
813
+ "step": 2725
814
+ },
815
+ {
816
+ "epoch": 2.2,
817
+ "grad_norm": 0.08961104601621628,
818
+ "learning_rate": 0.00040358831707031816,
819
+ "loss": 0.1787,
820
+ "step": 2750
821
+ },
822
+ {
823
+ "epoch": 2.22,
824
+ "grad_norm": 0.09573759883642197,
825
+ "learning_rate": 0.00038507230501055477,
826
+ "loss": 0.1851,
827
+ "step": 2775
828
+ },
829
+ {
830
+ "epoch": 2.24,
831
+ "grad_norm": 0.08124317228794098,
832
+ "learning_rate": 0.00036688928714561443,
833
+ "loss": 0.1849,
834
+ "step": 2800
835
+ },
836
+ {
837
+ "epoch": 2.26,
838
+ "grad_norm": 0.0962013229727745,
839
+ "learning_rate": 0.00034904910990043927,
840
+ "loss": 0.1869,
841
+ "step": 2825
842
+ },
843
+ {
844
+ "epoch": 2.2800000000000002,
845
+ "grad_norm": 0.08949823677539825,
846
+ "learning_rate": 0.0003315614340457279,
847
+ "loss": 0.18,
848
+ "step": 2850
849
+ },
850
+ {
851
+ "epoch": 2.3,
852
+ "grad_norm": 0.09404879063367844,
853
+ "learning_rate": 0.0003144357294664565,
854
+ "loss": 0.1823,
855
+ "step": 2875
856
+ },
857
+ {
858
+ "epoch": 2.32,
859
+ "grad_norm": 0.08220802247524261,
860
+ "learning_rate": 0.0002976812700337708,
861
+ "loss": 0.1889,
862
+ "step": 2900
863
+ },
864
+ {
865
+ "epoch": 2.34,
866
+ "grad_norm": 0.08680754154920578,
867
+ "learning_rate": 0.00028130712858302065,
868
+ "loss": 0.1888,
869
+ "step": 2925
870
+ },
871
+ {
872
+ "epoch": 2.36,
873
+ "grad_norm": 0.08079078048467636,
874
+ "learning_rate": 0.0002653221720006586,
875
+ "loss": 0.1852,
876
+ "step": 2950
877
+ },
878
+ {
879
+ "epoch": 2.38,
880
+ "grad_norm": 0.08778364211320877,
881
+ "learning_rate": 0.00024973505642266624,
882
+ "loss": 0.179,
883
+ "step": 2975
884
+ },
885
+ {
886
+ "epoch": 2.4,
887
+ "grad_norm": 0.10735012590885162,
888
+ "learning_rate": 0.00023455422254710235,
889
+ "loss": 0.1785,
890
+ "step": 3000
891
+ },
892
+ {
893
+ "epoch": 2.4,
894
+ "eval_loss": 0.2148195505142212,
895
+ "eval_runtime": 18.7355,
896
+ "eval_samples_per_second": 54.602,
897
+ "eval_steps_per_second": 0.854,
898
+ "step": 3000
899
+ },
900
+ {
901
+ "epoch": 2.42,
902
+ "grad_norm": 0.1003650426864624,
903
+ "learning_rate": 0.00021978789106331665,
904
+ "loss": 0.1831,
905
+ "step": 3025
906
+ },
907
+ {
908
+ "epoch": 2.44,
909
+ "grad_norm": 0.0836385041475296,
910
+ "learning_rate": 0.00020544405820030164,
911
+ "loss": 0.1744,
912
+ "step": 3050
913
+ },
914
+ {
915
+ "epoch": 2.46,
916
+ "grad_norm": 0.08322740346193314,
917
+ "learning_rate": 0.00019153049139659383,
918
+ "loss": 0.1854,
919
+ "step": 3075
920
+ },
921
+ {
922
+ "epoch": 2.48,
923
+ "grad_norm": 0.0786130279302597,
924
+ "learning_rate": 0.00017805472509406696,
925
+ "loss": 0.1855,
926
+ "step": 3100
927
+ },
928
+ {
929
+ "epoch": 2.5,
930
+ "grad_norm": 0.08832788467407227,
931
+ "learning_rate": 0.00016502405665790132,
932
+ "loss": 0.1769,
933
+ "step": 3125
934
+ },
935
+ {
936
+ "epoch": 2.52,
937
+ "grad_norm": 0.07723420858383179,
938
+ "learning_rate": 0.00015244554242493024,
939
+ "loss": 0.181,
940
+ "step": 3150
941
+ },
942
+ {
943
+ "epoch": 2.54,
944
+ "grad_norm": 0.09589297324419022,
945
+ "learning_rate": 0.000140325993882509,
946
+ "loss": 0.1873,
947
+ "step": 3175
948
+ },
949
+ {
950
+ "epoch": 2.56,
951
+ "grad_norm": 0.07765637338161469,
952
+ "learning_rate": 0.00012867197397997533,
953
+ "loss": 0.1785,
954
+ "step": 3200
955
+ },
956
+ {
957
+ "epoch": 2.58,
958
+ "grad_norm": 0.08880658447742462,
959
+ "learning_rate": 0.00011748979357469624,
960
+ "loss": 0.1755,
961
+ "step": 3225
962
+ },
963
+ {
964
+ "epoch": 2.6,
965
+ "grad_norm": 0.09514382481575012,
966
+ "learning_rate": 0.00010678550801462661,
967
+ "loss": 0.1858,
968
+ "step": 3250
969
+ },
970
+ {
971
+ "epoch": 2.62,
972
+ "grad_norm": 0.08952795714139938,
973
+ "learning_rate": 9.656491385923216e-05,
974
+ "loss": 0.1704,
975
+ "step": 3275
976
+ },
977
+ {
978
+ "epoch": 2.64,
979
+ "grad_norm": 0.08624137192964554,
980
+ "learning_rate": 8.683354574055058e-05,
981
+ "loss": 0.1758,
982
+ "step": 3300
983
+ },
984
+ {
985
+ "epoch": 2.66,
986
+ "grad_norm": 0.09443361312150955,
987
+ "learning_rate": 7.759667336609011e-05,
988
+ "loss": 0.1839,
989
+ "step": 3325
990
+ },
991
+ {
992
+ "epoch": 2.68,
993
+ "grad_norm": 0.10429251939058304,
994
+ "learning_rate": 6.885929866518948e-05,
995
+ "loss": 0.1722,
996
+ "step": 3350
997
+ },
998
+ {
999
+ "epoch": 2.7,
1000
+ "grad_norm": 0.09328500181436539,
1001
+ "learning_rate": 6.062615308038599e-05,
1002
+ "loss": 0.1818,
1003
+ "step": 3375
1004
+ },
1005
+ {
1006
+ "epoch": 2.7199999999999998,
1007
+ "grad_norm": 0.09164253622293472,
1008
+ "learning_rate": 5.290169500525577e-05,
1009
+ "loss": 0.1803,
1010
+ "step": 3400
1011
+ },
1012
+ {
1013
+ "epoch": 2.74,
1014
+ "grad_norm": 0.08848071843385696,
1015
+ "learning_rate": 4.569010737011381e-05,
1016
+ "loss": 0.1833,
1017
+ "step": 3425
1018
+ },
1019
+ {
1020
+ "epoch": 2.76,
1021
+ "grad_norm": 0.08395551890134811,
1022
+ "learning_rate": 3.8995295376884954e-05,
1023
+ "loss": 0.1827,
1024
+ "step": 3450
1025
+ },
1026
+ {
1027
+ "epoch": 2.7800000000000002,
1028
+ "grad_norm": 0.10054022073745728,
1029
+ "learning_rate": 3.2820884384367145e-05,
1030
+ "loss": 0.1826,
1031
+ "step": 3475
1032
+ },
1033
+ {
1034
+ "epoch": 2.8,
1035
+ "grad_norm": 0.08368613570928574,
1036
+ "learning_rate": 2.7170217945036156e-05,
1037
+ "loss": 0.182,
1038
+ "step": 3500
1039
+ },
1040
+ {
1041
+ "epoch": 2.8,
1042
+ "eval_loss": 0.2119080126285553,
1043
+ "eval_runtime": 18.7325,
1044
+ "eval_samples_per_second": 54.611,
1045
+ "eval_steps_per_second": 0.854,
1046
+ "step": 3500
1047
+ },
1048
+ {
1049
+ "epoch": 2.82,
1050
+ "grad_norm": 0.0736178457736969,
1051
+ "learning_rate": 2.2046355994452594e-05,
1052
+ "loss": 0.1769,
1053
+ "step": 3525
1054
+ },
1055
+ {
1056
+ "epoch": 2.84,
1057
+ "grad_norm": 0.09487678110599518,
1058
+ "learning_rate": 1.745207319425324e-05,
1059
+ "loss": 0.1819,
1060
+ "step": 3550
1061
+ },
1062
+ {
1063
+ "epoch": 2.86,
1064
+ "grad_norm": 0.07659710943698883,
1065
+ "learning_rate": 1.338985742962251e-05,
1066
+ "loss": 0.1793,
1067
+ "step": 3575
1068
+ },
1069
+ {
1070
+ "epoch": 2.88,
1071
+ "grad_norm": 0.09316933900117874,
1072
+ "learning_rate": 9.8619084620587e-06,
1073
+ "loss": 0.1817,
1074
+ "step": 3600
1075
+ },
1076
+ {
1077
+ "epoch": 2.9,
1078
+ "grad_norm": 0.10243932902812958,
1079
+ "learning_rate": 6.870136738164612e-06,
1080
+ "loss": 0.177,
1081
+ "step": 3625
1082
+ },
1083
+ {
1084
+ "epoch": 2.92,
1085
+ "grad_norm": 0.07816595584154129,
1086
+ "learning_rate": 4.416162355107001e-06,
1087
+ "loss": 0.1879,
1088
+ "step": 3650
1089
+ },
1090
+ {
1091
+ "epoch": 2.94,
1092
+ "grad_norm": 0.09459181874990463,
1093
+ "learning_rate": 2.5013141833046815e-06,
1094
+ "loss": 0.1786,
1095
+ "step": 3675
1096
+ },
1097
+ {
1098
+ "epoch": 2.96,
1099
+ "grad_norm": 0.10340872406959534,
1100
+ "learning_rate": 1.1266291468229327e-06,
1101
+ "loss": 0.184,
1102
+ "step": 3700
1103
+ },
1104
+ {
1105
+ "epoch": 2.98,
1106
+ "grad_norm": 0.09642766416072845,
1107
+ "learning_rate": 2.928516618599941e-07,
1108
+ "loss": 0.179,
1109
+ "step": 3725
1110
+ },
1111
+ {
1112
+ "epoch": 3.0,
1113
+ "grad_norm": 0.08058501780033112,
1114
+ "learning_rate": 4.332336324130992e-10,
1115
+ "loss": 0.1793,
1116
+ "step": 3750
1117
+ },
1118
+ {
1119
+ "epoch": 3.0,
1120
+ "step": 3750,
1121
+ "total_flos": 2.43882352705536e+18,
1122
+ "train_loss": 0.23855862986246745,
1123
+ "train_runtime": 3611.8553,
1124
+ "train_samples_per_second": 33.224,
1125
+ "train_steps_per_second": 1.038
1126
+ }
1127
+ ],
1128
+ "logging_steps": 25,
1129
+ "max_steps": 3750,
1130
+ "num_input_tokens_seen": 0,
1131
+ "num_train_epochs": 3,
1132
+ "save_steps": 0,
1133
+ "stateful_callbacks": {
1134
+ "TrainerControl": {
1135
+ "args": {
1136
+ "should_epoch_stop": false,
1137
+ "should_evaluate": false,
1138
+ "should_log": false,
1139
+ "should_save": true,
1140
+ "should_training_stop": true
1141
+ },
1142
+ "attributes": {}
1143
+ }
1144
+ },
1145
+ "total_flos": 2.43882352705536e+18,
1146
+ "train_batch_size": 32,
1147
+ "trial_name": null,
1148
+ "trial_params": null
1149
+ }
nl_tasks/exps/run_ex19_2ep/ft/adapter_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "inference_mode": false,
6
+ "layers_to_transform": null,
7
+ "modules_to_save": null,
8
+ "num_rotations": 1,
9
+ "peft_type": "ROTATION",
10
+ "r": 16,
11
+ "revision": null,
12
+ "target_modules": [
13
+ "v_proj",
14
+ "q_proj"
15
+ ],
16
+ "target_modules_to_skip": null,
17
+ "task_type": "CAUSAL_LM"
18
+ }
nl_tasks/exps/run_ex19_2ep/ft/added_tokens.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ {
2
+ "[PAD]": 32000
3
+ }
nl_tasks/exps/run_ex19_2ep/ft/special_tokens_map.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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": {
17
+ "content": "[PAD]",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ },
23
+ "unk_token": {
24
+ "content": "</s>",
25
+ "lstrip": false,
26
+ "normalized": false,
27
+ "rstrip": false,
28
+ "single_word": false
29
+ }
30
+ }
nl_tasks/exps/run_ex19_2ep/ft/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
nl_tasks/exps/run_ex19_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/exps/run_ex19_2ep/ft/tokenizer_config.json ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "32000": {
31
+ "content": "[PAD]",
32
+ "lstrip": false,
33
+ "normalized": false,
34
+ "rstrip": false,
35
+ "single_word": false,
36
+ "special": true
37
+ }
38
+ },
39
+ "bos_token": "</s>",
40
+ "clean_up_tokenization_spaces": false,
41
+ "eos_token": "</s>",
42
+ "extra_special_tokens": {},
43
+ "legacy": false,
44
+ "model_max_length": 512,
45
+ "pad_token": "[PAD]",
46
+ "padding_side": "right",
47
+ "sp_model_kwargs": {},
48
+ "tokenizer_class": "LlamaTokenizer",
49
+ "unk_token": "</s>",
50
+ "use_default_system_prompt": false
51
+ }
nl_tasks/exps/run_ex19_2ep/ft2/adapter_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "inference_mode": true,
6
+ "layers_to_transform": null,
7
+ "modules_to_save": null,
8
+ "num_rotations": 1,
9
+ "peft_type": "ROTATION",
10
+ "r": 16,
11
+ "revision": null,
12
+ "target_modules": [
13
+ "v_proj",
14
+ "q_proj"
15
+ ],
16
+ "target_modules_to_skip": null,
17
+ "task_type": "CAUSAL_LM"
18
+ }
nl_tasks/exps/run_ex19_2ep/ft2/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0322f9d7a2de8012fd1e2c5d523e1e4e08865632a8c5a5608e70a4cf9128edef
3
+ size 33602915
nl_tasks/exps/run_ex19_2ep/trainer_state.json ADDED
@@ -0,0 +1,818 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 2.0,
6
+ "eval_steps": 200,
7
+ "global_step": 2438,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.020508613617719443,
14
+ "grad_norm": 0.27057042717933655,
15
+ "learning_rate": 0.0004918032786885246,
16
+ "loss": 0.6665,
17
+ "step": 25
18
+ },
19
+ {
20
+ "epoch": 0.04101722723543889,
21
+ "grad_norm": 0.23511329293251038,
22
+ "learning_rate": 0.0010040983606557376,
23
+ "loss": 0.3958,
24
+ "step": 50
25
+ },
26
+ {
27
+ "epoch": 0.06152584085315833,
28
+ "grad_norm": 0.19983653724193573,
29
+ "learning_rate": 0.0015163934426229509,
30
+ "loss": 0.3823,
31
+ "step": 75
32
+ },
33
+ {
34
+ "epoch": 0.08203445447087777,
35
+ "grad_norm": 0.30811649560928345,
36
+ "learning_rate": 0.002028688524590164,
37
+ "loss": 0.6459,
38
+ "step": 100
39
+ },
40
+ {
41
+ "epoch": 0.10254306808859721,
42
+ "grad_norm": 0.4009745419025421,
43
+ "learning_rate": 0.002540983606557377,
44
+ "loss": 0.385,
45
+ "step": 125
46
+ },
47
+ {
48
+ "epoch": 0.12305168170631665,
49
+ "grad_norm": 0.4340912103652954,
50
+ "learning_rate": 0.00305327868852459,
51
+ "loss": 0.3897,
52
+ "step": 150
53
+ },
54
+ {
55
+ "epoch": 0.1435602953240361,
56
+ "grad_norm": 0.39641642570495605,
57
+ "learning_rate": 0.0035655737704918034,
58
+ "loss": 0.3933,
59
+ "step": 175
60
+ },
61
+ {
62
+ "epoch": 0.16406890894175555,
63
+ "grad_norm": 0.3010708689689636,
64
+ "learning_rate": 0.004077868852459017,
65
+ "loss": 0.397,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.16406890894175555,
70
+ "eval_loss": 0.3470916450023651,
71
+ "eval_runtime": 19.5729,
72
+ "eval_samples_per_second": 51.091,
73
+ "eval_steps_per_second": 0.817,
74
+ "step": 200
75
+ },
76
+ {
77
+ "epoch": 0.184577522559475,
78
+ "grad_norm": 0.47758591175079346,
79
+ "learning_rate": 0.0045901639344262295,
80
+ "loss": 0.3753,
81
+ "step": 225
82
+ },
83
+ {
84
+ "epoch": 0.20508613617719443,
85
+ "grad_norm": 0.24603353440761566,
86
+ "learning_rate": 0.0049999359270580315,
87
+ "loss": 0.3476,
88
+ "step": 250
89
+ },
90
+ {
91
+ "epoch": 0.22559474979491387,
92
+ "grad_norm": 0.2857294976711273,
93
+ "learning_rate": 0.004997693718919013,
94
+ "loss": 0.3364,
95
+ "step": 275
96
+ },
97
+ {
98
+ "epoch": 0.2461033634126333,
99
+ "grad_norm": 0.19567205011844635,
100
+ "learning_rate": 0.004992251147198466,
101
+ "loss": 0.3277,
102
+ "step": 300
103
+ },
104
+ {
105
+ "epoch": 0.2666119770303528,
106
+ "grad_norm": 0.15630540251731873,
107
+ "learning_rate": 0.00498361518561306,
108
+ "loss": 0.3244,
109
+ "step": 325
110
+ },
111
+ {
112
+ "epoch": 0.2871205906480722,
113
+ "grad_norm": 0.129176065325737,
114
+ "learning_rate": 0.004971796899657632,
115
+ "loss": 0.3123,
116
+ "step": 350
117
+ },
118
+ {
119
+ "epoch": 0.30762920426579166,
120
+ "grad_norm": 0.1556542068719864,
121
+ "learning_rate": 0.004956811432426662,
122
+ "loss": 0.3202,
123
+ "step": 375
124
+ },
125
+ {
126
+ "epoch": 0.3281378178835111,
127
+ "grad_norm": 0.12940339744091034,
128
+ "learning_rate": 0.0049386779852110116,
129
+ "loss": 0.3027,
130
+ "step": 400
131
+ },
132
+ {
133
+ "epoch": 0.3281378178835111,
134
+ "eval_loss": 0.2784655690193176,
135
+ "eval_runtime": 19.3407,
136
+ "eval_samples_per_second": 51.704,
137
+ "eval_steps_per_second": 0.827,
138
+ "step": 400
139
+ },
140
+ {
141
+ "epoch": 0.34864643150123054,
142
+ "grad_norm": 0.11640695482492447,
143
+ "learning_rate": 0.0049174197928947795,
144
+ "loss": 0.3085,
145
+ "step": 425
146
+ },
147
+ {
148
+ "epoch": 0.36915504511895,
149
+ "grad_norm": 0.10215829312801361,
150
+ "learning_rate": 0.004893064094183811,
151
+ "loss": 0.3027,
152
+ "step": 450
153
+ },
154
+ {
155
+ "epoch": 0.3896636587366694,
156
+ "grad_norm": 0.12192098051309586,
157
+ "learning_rate": 0.00486564209670399,
158
+ "loss": 0.2895,
159
+ "step": 475
160
+ },
161
+ {
162
+ "epoch": 0.41017227235438886,
163
+ "grad_norm": 0.08577418327331543,
164
+ "learning_rate": 0.004835188937014058,
165
+ "loss": 0.2745,
166
+ "step": 500
167
+ },
168
+ {
169
+ "epoch": 0.4306808859721083,
170
+ "grad_norm": 0.10127752274274826,
171
+ "learning_rate": 0.0048017436355841684,
172
+ "loss": 0.2758,
173
+ "step": 525
174
+ },
175
+ {
176
+ "epoch": 0.45118949958982774,
177
+ "grad_norm": 0.1058003306388855,
178
+ "learning_rate": 0.0047653490467978905,
179
+ "loss": 0.2852,
180
+ "step": 550
181
+ },
182
+ {
183
+ "epoch": 0.4716981132075472,
184
+ "grad_norm": 0.1132282242178917,
185
+ "learning_rate": 0.004726051804041709,
186
+ "loss": 0.2769,
187
+ "step": 575
188
+ },
189
+ {
190
+ "epoch": 0.4922067268252666,
191
+ "grad_norm": 0.0941806212067604,
192
+ "learning_rate": 0.004683902259952387,
193
+ "loss": 0.2936,
194
+ "step": 600
195
+ },
196
+ {
197
+ "epoch": 0.4922067268252666,
198
+ "eval_loss": 0.2574584484100342,
199
+ "eval_runtime": 19.3378,
200
+ "eval_samples_per_second": 51.712,
201
+ "eval_steps_per_second": 0.827,
202
+ "step": 600
203
+ },
204
+ {
205
+ "epoch": 0.5127153404429861,
206
+ "grad_norm": 0.08521772176027298,
207
+ "learning_rate": 0.004638954421898746,
208
+ "loss": 0.2737,
209
+ "step": 625
210
+ },
211
+ {
212
+ "epoch": 0.5332239540607056,
213
+ "grad_norm": 0.08367811143398285,
214
+ "learning_rate": 0.004591265882780542,
215
+ "loss": 0.2657,
216
+ "step": 650
217
+ },
218
+ {
219
+ "epoch": 0.5537325676784249,
220
+ "grad_norm": 0.07922013849020004,
221
+ "learning_rate": 0.0045408977472331,
222
+ "loss": 0.2789,
223
+ "step": 675
224
+ },
225
+ {
226
+ "epoch": 0.5742411812961444,
227
+ "grad_norm": 0.0850289985537529,
228
+ "learning_rate": 0.00448791455333227,
229
+ "loss": 0.2763,
230
+ "step": 700
231
+ },
232
+ {
233
+ "epoch": 0.5947497949138638,
234
+ "grad_norm": 0.09892452508211136,
235
+ "learning_rate": 0.004432384189900008,
236
+ "loss": 0.2745,
237
+ "step": 725
238
+ },
239
+ {
240
+ "epoch": 0.6152584085315833,
241
+ "grad_norm": 0.0870669037103653,
242
+ "learning_rate": 0.004374377809516577,
243
+ "loss": 0.2722,
244
+ "step": 750
245
+ },
246
+ {
247
+ "epoch": 0.6357670221493027,
248
+ "grad_norm": 0.0841507762670517,
249
+ "learning_rate": 0.004313969737350775,
250
+ "loss": 0.2676,
251
+ "step": 775
252
+ },
253
+ {
254
+ "epoch": 0.6562756357670222,
255
+ "grad_norm": 0.08125782757997513,
256
+ "learning_rate": 0.004251237375925071,
257
+ "loss": 0.2724,
258
+ "step": 800
259
+ },
260
+ {
261
+ "epoch": 0.6562756357670222,
262
+ "eval_loss": 0.24871860444545746,
263
+ "eval_runtime": 19.3235,
264
+ "eval_samples_per_second": 51.75,
265
+ "eval_steps_per_second": 0.828,
266
+ "step": 800
267
+ },
268
+ {
269
+ "epoch": 0.6767842493847416,
270
+ "grad_norm": 0.07026949524879456,
271
+ "learning_rate": 0.004186261105937612,
272
+ "loss": 0.265,
273
+ "step": 825
274
+ },
275
+ {
276
+ "epoch": 0.6972928630024611,
277
+ "grad_norm": 0.08101508021354675,
278
+ "learning_rate": 0.004119124183268237,
279
+ "loss": 0.2674,
280
+ "step": 850
281
+ },
282
+ {
283
+ "epoch": 0.7178014766201805,
284
+ "grad_norm": 0.09817145019769669,
285
+ "learning_rate": 0.004049912632300421,
286
+ "loss": 0.2556,
287
+ "step": 875
288
+ },
289
+ {
290
+ "epoch": 0.7383100902379,
291
+ "grad_norm": 0.07607579976320267,
292
+ "learning_rate": 0.003978715135695881,
293
+ "loss": 0.2604,
294
+ "step": 900
295
+ },
296
+ {
297
+ "epoch": 0.7588187038556193,
298
+ "grad_norm": 0.08425214886665344,
299
+ "learning_rate": 0.003905622920763031,
300
+ "loss": 0.2645,
301
+ "step": 925
302
+ },
303
+ {
304
+ "epoch": 0.7793273174733388,
305
+ "grad_norm": 0.08241744339466095,
306
+ "learning_rate": 0.00383072964256494,
307
+ "loss": 0.2519,
308
+ "step": 950
309
+ },
310
+ {
311
+ "epoch": 0.7998359310910582,
312
+ "grad_norm": 0.07475872337818146,
313
+ "learning_rate": 0.0037541312639165145,
314
+ "loss": 0.2594,
315
+ "step": 975
316
+ },
317
+ {
318
+ "epoch": 0.8203445447087777,
319
+ "grad_norm": 0.0758645310997963,
320
+ "learning_rate": 0.003675925932424715,
321
+ "loss": 0.2529,
322
+ "step": 1000
323
+ },
324
+ {
325
+ "epoch": 0.8203445447087777,
326
+ "eval_loss": 0.2370542287826538,
327
+ "eval_runtime": 19.3412,
328
+ "eval_samples_per_second": 51.703,
329
+ "eval_steps_per_second": 0.827,
330
+ "step": 1000
331
+ },
332
+ {
333
+ "epoch": 0.8408531583264971,
334
+ "grad_norm": 0.11314820498228073,
335
+ "learning_rate": 0.0035962138547293277,
336
+ "loss": 0.2526,
337
+ "step": 1025
338
+ },
339
+ {
340
+ "epoch": 0.8613617719442166,
341
+ "grad_norm": 0.0737948939204216,
342
+ "learning_rate": 0.003515097168105444,
343
+ "loss": 0.249,
344
+ "step": 1050
345
+ },
346
+ {
347
+ "epoch": 0.881870385561936,
348
+ "grad_norm": 0.06914939731359482,
349
+ "learning_rate": 0.003432679809592165,
350
+ "loss": 0.2542,
351
+ "step": 1075
352
+ },
353
+ {
354
+ "epoch": 0.9023789991796555,
355
+ "grad_norm": 0.0717969536781311,
356
+ "learning_rate": 0.003349067382815217,
357
+ "loss": 0.2551,
358
+ "step": 1100
359
+ },
360
+ {
361
+ "epoch": 0.9228876127973749,
362
+ "grad_norm": 0.07602035254240036,
363
+ "learning_rate": 0.003264367022674124,
364
+ "loss": 0.2528,
365
+ "step": 1125
366
+ },
367
+ {
368
+ "epoch": 0.9433962264150944,
369
+ "grad_norm": 0.05798409879207611,
370
+ "learning_rate": 0.0031786872580673213,
371
+ "loss": 0.2496,
372
+ "step": 1150
373
+ },
374
+ {
375
+ "epoch": 0.9639048400328137,
376
+ "grad_norm": 0.05903997644782066,
377
+ "learning_rate": 0.0030921378728310895,
378
+ "loss": 0.2547,
379
+ "step": 1175
380
+ },
381
+ {
382
+ "epoch": 0.9844134536505332,
383
+ "grad_norm": 0.0637289434671402,
384
+ "learning_rate": 0.0030048297650705157,
385
+ "loss": 0.2607,
386
+ "step": 1200
387
+ },
388
+ {
389
+ "epoch": 0.9844134536505332,
390
+ "eval_loss": 0.22966955602169037,
391
+ "eval_runtime": 19.3215,
392
+ "eval_samples_per_second": 51.756,
393
+ "eval_steps_per_second": 0.828,
394
+ "step": 1200
395
+ },
396
+ {
397
+ "epoch": 1.0049220672682526,
398
+ "grad_norm": 0.06701506674289703,
399
+ "learning_rate": 0.002916874805062701,
400
+ "loss": 0.2402,
401
+ "step": 1225
402
+ },
403
+ {
404
+ "epoch": 1.0254306808859721,
405
+ "grad_norm": 0.06756694614887238,
406
+ "learning_rate": 0.0028283856919143007,
407
+ "loss": 0.2235,
408
+ "step": 1250
409
+ },
410
+ {
411
+ "epoch": 1.0459392945036916,
412
+ "grad_norm": 0.0699879452586174,
413
+ "learning_rate": 0.0027394758091570663,
414
+ "loss": 0.2245,
415
+ "step": 1275
416
+ },
417
+ {
418
+ "epoch": 1.066447908121411,
419
+ "grad_norm": 0.05907871574163437,
420
+ "learning_rate": 0.0026502590794664073,
421
+ "loss": 0.2277,
422
+ "step": 1300
423
+ },
424
+ {
425
+ "epoch": 1.0869565217391304,
426
+ "grad_norm": 0.06299999356269836,
427
+ "learning_rate": 0.002560849818689141,
428
+ "loss": 0.2233,
429
+ "step": 1325
430
+ },
431
+ {
432
+ "epoch": 1.1074651353568499,
433
+ "grad_norm": 0.05034913495182991,
434
+ "learning_rate": 0.002471362589367452,
435
+ "loss": 0.2174,
436
+ "step": 1350
437
+ },
438
+ {
439
+ "epoch": 1.1279737489745694,
440
+ "grad_norm": 0.06710750609636307,
441
+ "learning_rate": 0.0023819120539467662,
442
+ "loss": 0.2205,
443
+ "step": 1375
444
+ },
445
+ {
446
+ "epoch": 1.1484823625922886,
447
+ "grad_norm": 0.05786602199077606,
448
+ "learning_rate": 0.002292612827855605,
449
+ "loss": 0.2123,
450
+ "step": 1400
451
+ },
452
+ {
453
+ "epoch": 1.1484823625922886,
454
+ "eval_loss": 0.22506149113178253,
455
+ "eval_runtime": 19.316,
456
+ "eval_samples_per_second": 51.771,
457
+ "eval_steps_per_second": 0.828,
458
+ "step": 1400
459
+ },
460
+ {
461
+ "epoch": 1.1689909762100081,
462
+ "grad_norm": 0.06614014506340027,
463
+ "learning_rate": 0.002203579332645688,
464
+ "loss": 0.2238,
465
+ "step": 1425
466
+ },
467
+ {
468
+ "epoch": 1.1894995898277276,
469
+ "grad_norm": 0.06342432647943497,
470
+ "learning_rate": 0.0021149256493804575,
471
+ "loss": 0.2251,
472
+ "step": 1450
473
+ },
474
+ {
475
+ "epoch": 1.2100082034454471,
476
+ "grad_norm": 0.06356894224882126,
477
+ "learning_rate": 0.0020267653724598744,
478
+ "loss": 0.2213,
479
+ "step": 1475
480
+ },
481
+ {
482
+ "epoch": 1.2305168170631666,
483
+ "grad_norm": 0.06422504037618637,
484
+ "learning_rate": 0.0019392114640687983,
485
+ "loss": 0.2299,
486
+ "step": 1500
487
+ },
488
+ {
489
+ "epoch": 1.251025430680886,
490
+ "grad_norm": 0.07064881175756454,
491
+ "learning_rate": 0.0018523761094354305,
492
+ "loss": 0.2232,
493
+ "step": 1525
494
+ },
495
+ {
496
+ "epoch": 1.2715340442986054,
497
+ "grad_norm": 0.06022433191537857,
498
+ "learning_rate": 0.0017663705730853014,
499
+ "loss": 0.2165,
500
+ "step": 1550
501
+ },
502
+ {
503
+ "epoch": 1.2920426579163249,
504
+ "grad_norm": 0.052290238440036774,
505
+ "learning_rate": 0.0016813050562749776,
506
+ "loss": 0.219,
507
+ "step": 1575
508
+ },
509
+ {
510
+ "epoch": 1.3125512715340442,
511
+ "grad_norm": 0.05720488727092743,
512
+ "learning_rate": 0.0015972885557881667,
513
+ "loss": 0.2108,
514
+ "step": 1600
515
+ },
516
+ {
517
+ "epoch": 1.3125512715340442,
518
+ "eval_loss": 0.21939675509929657,
519
+ "eval_runtime": 19.2766,
520
+ "eval_samples_per_second": 51.876,
521
+ "eval_steps_per_second": 0.83,
522
+ "step": 1600
523
+ },
524
+ {
525
+ "epoch": 1.3330598851517639,
526
+ "grad_norm": 0.05397438257932663,
527
+ "learning_rate": 0.0015144287242751379,
528
+ "loss": 0.2161,
529
+ "step": 1625
530
+ },
531
+ {
532
+ "epoch": 1.3535684987694832,
533
+ "grad_norm": 0.060338784009218216,
534
+ "learning_rate": 0.0014328317323144284,
535
+ "loss": 0.2151,
536
+ "step": 1650
537
+ },
538
+ {
539
+ "epoch": 1.3740771123872026,
540
+ "grad_norm": 0.06446268409490585,
541
+ "learning_rate": 0.0013526021323735627,
542
+ "loss": 0.2127,
543
+ "step": 1675
544
+ },
545
+ {
546
+ "epoch": 1.3945857260049221,
547
+ "grad_norm": 0.05473887547850609,
548
+ "learning_rate": 0.0012738427248431027,
549
+ "loss": 0.212,
550
+ "step": 1700
551
+ },
552
+ {
553
+ "epoch": 1.4150943396226414,
554
+ "grad_norm": 0.054772548377513885,
555
+ "learning_rate": 0.0011966544263156864,
556
+ "loss": 0.2093,
557
+ "step": 1725
558
+ },
559
+ {
560
+ "epoch": 1.435602953240361,
561
+ "grad_norm": 0.07580874860286713,
562
+ "learning_rate": 0.0011211361402788226,
563
+ "loss": 0.2085,
564
+ "step": 1750
565
+ },
566
+ {
567
+ "epoch": 1.4561115668580804,
568
+ "grad_norm": 0.05953690782189369,
569
+ "learning_rate": 0.001047384630387131,
570
+ "loss": 0.2157,
571
+ "step": 1775
572
+ },
573
+ {
574
+ "epoch": 1.4766201804758,
575
+ "grad_norm": 0.0596901997923851,
576
+ "learning_rate": 0.0009754943964764229,
577
+ "loss": 0.2058,
578
+ "step": 1800
579
+ },
580
+ {
581
+ "epoch": 1.4766201804758,
582
+ "eval_loss": 0.21527747809886932,
583
+ "eval_runtime": 19.3456,
584
+ "eval_samples_per_second": 51.691,
585
+ "eval_steps_per_second": 0.827,
586
+ "step": 1800
587
+ },
588
+ {
589
+ "epoch": 1.4971287940935194,
590
+ "grad_norm": 0.0607760027050972,
591
+ "learning_rate": 0.0009055575534784591,
592
+ "loss": 0.205,
593
+ "step": 1825
594
+ },
595
+ {
596
+ "epoch": 1.5176374077112387,
597
+ "grad_norm": 0.05277996510267258,
598
+ "learning_rate": 0.0008376637133915557,
599
+ "loss": 0.2074,
600
+ "step": 1850
601
+ },
602
+ {
603
+ "epoch": 1.5381460213289582,
604
+ "grad_norm": 0.06398818641901016,
605
+ "learning_rate": 0.0007718998704582739,
606
+ "loss": 0.2078,
607
+ "step": 1875
608
+ },
609
+ {
610
+ "epoch": 1.5586546349466777,
611
+ "grad_norm": 0.06612731516361237,
612
+ "learning_rate": 0.0007083502896973018,
613
+ "loss": 0.2113,
614
+ "step": 1900
615
+ },
616
+ {
617
+ "epoch": 1.579163248564397,
618
+ "grad_norm": 0.058860503137111664,
619
+ "learning_rate": 0.0006470963989323764,
620
+ "loss": 0.2123,
621
+ "step": 1925
622
+ },
623
+ {
624
+ "epoch": 1.5996718621821167,
625
+ "grad_norm": 0.06213616952300072,
626
+ "learning_rate": 0.0005882166844565692,
627
+ "loss": 0.2097,
628
+ "step": 1950
629
+ },
630
+ {
631
+ "epoch": 1.620180475799836,
632
+ "grad_norm": 0.05753476545214653,
633
+ "learning_rate": 0.0005317865904656496,
634
+ "loss": 0.205,
635
+ "step": 1975
636
+ },
637
+ {
638
+ "epoch": 1.6406890894175554,
639
+ "grad_norm": 0.06556696444749832,
640
+ "learning_rate": 0.0004778784223893601,
641
+ "loss": 0.2126,
642
+ "step": 2000
643
+ },
644
+ {
645
+ "epoch": 1.6406890894175554,
646
+ "eval_loss": 0.20999926328659058,
647
+ "eval_runtime": 19.3526,
648
+ "eval_samples_per_second": 51.673,
649
+ "eval_steps_per_second": 0.827,
650
+ "step": 2000
651
+ },
652
+ {
653
+ "epoch": 1.661197703035275,
654
+ "grad_norm": 0.05554095283150673,
655
+ "learning_rate": 0.0004265612542444827,
656
+ "loss": 0.2026,
657
+ "step": 2025
658
+ },
659
+ {
660
+ "epoch": 1.6817063166529942,
661
+ "grad_norm": 0.06666585803031921,
662
+ "learning_rate": 0.00037790084012840453,
663
+ "loss": 0.205,
664
+ "step": 2050
665
+ },
666
+ {
667
+ "epoch": 1.7022149302707137,
668
+ "grad_norm": 0.06906694173812866,
669
+ "learning_rate": 0.00033195952996658727,
670
+ "loss": 0.2104,
671
+ "step": 2075
672
+ },
673
+ {
674
+ "epoch": 1.7227235438884332,
675
+ "grad_norm": 0.05908932536840439,
676
+ "learning_rate": 0.00028879618962189324,
677
+ "loss": 0.2004,
678
+ "step": 2100
679
+ },
680
+ {
681
+ "epoch": 1.7432321575061525,
682
+ "grad_norm": 0.05455790087580681,
683
+ "learning_rate": 0.0002484661254681381,
684
+ "loss": 0.2038,
685
+ "step": 2125
686
+ },
687
+ {
688
+ "epoch": 1.7637407711238722,
689
+ "grad_norm": 0.049193527549505234,
690
+ "learning_rate": 0.0002110210135245147,
691
+ "loss": 0.2013,
692
+ "step": 2150
693
+ },
694
+ {
695
+ "epoch": 1.7842493847415914,
696
+ "grad_norm": 0.11325162649154663,
697
+ "learning_rate": 0.0001765088332416917,
698
+ "loss": 0.21,
699
+ "step": 2175
700
+ },
701
+ {
702
+ "epoch": 1.804757998359311,
703
+ "grad_norm": 0.05528680980205536,
704
+ "learning_rate": 0.00014497380602442378,
705
+ "loss": 0.2028,
706
+ "step": 2200
707
+ },
708
+ {
709
+ "epoch": 1.804757998359311,
710
+ "eval_loss": 0.2075502574443817,
711
+ "eval_runtime": 19.3246,
712
+ "eval_samples_per_second": 51.748,
713
+ "eval_steps_per_second": 0.828,
714
+ "step": 2200
715
+ },
716
+ {
717
+ "epoch": 1.8252666119770304,
718
+ "grad_norm": 0.06102382019162178,
719
+ "learning_rate": 0.00011645633856944976,
720
+ "loss": 0.204,
721
+ "step": 2225
722
+ },
723
+ {
724
+ "epoch": 1.8457752255947497,
725
+ "grad_norm": 0.05555664747953415,
726
+ "learning_rate": 9.099297109128407e-05,
727
+ "loss": 0.2051,
728
+ "step": 2250
729
+ },
730
+ {
731
+ "epoch": 1.8662838392124692,
732
+ "grad_norm": 0.05285027623176575,
733
+ "learning_rate": 6.861633050223526e-05,
734
+ "loss": 0.2115,
735
+ "step": 2275
736
+ },
737
+ {
738
+ "epoch": 1.8867924528301887,
739
+ "grad_norm": 0.052628323435783386,
740
+ "learning_rate": 4.9355088606646005e-05,
741
+ "loss": 0.2034,
742
+ "step": 2300
743
+ },
744
+ {
745
+ "epoch": 1.907301066447908,
746
+ "grad_norm": 0.06317585706710815,
747
+ "learning_rate": 3.323392536292435e-05,
748
+ "loss": 0.2119,
749
+ "step": 2325
750
+ },
751
+ {
752
+ "epoch": 1.9278096800656277,
753
+ "grad_norm": 0.05837605521082878,
754
+ "learning_rate": 2.0273497260433205e-05,
755
+ "loss": 0.1951,
756
+ "step": 2350
757
+ },
758
+ {
759
+ "epoch": 1.948318293683347,
760
+ "grad_norm": 0.053302984684705734,
761
+ "learning_rate": 1.0490410851763942e-05,
762
+ "loss": 0.2007,
763
+ "step": 2375
764
+ },
765
+ {
766
+ "epoch": 1.9688269073010665,
767
+ "grad_norm": 0.06087959557771683,
768
+ "learning_rate": 3.8972014743038356e-06,
769
+ "loss": 0.2028,
770
+ "step": 2400
771
+ },
772
+ {
773
+ "epoch": 1.9688269073010665,
774
+ "eval_loss": 0.2069627046585083,
775
+ "eval_runtime": 19.3313,
776
+ "eval_samples_per_second": 51.73,
777
+ "eval_steps_per_second": 0.828,
778
+ "step": 2400
779
+ },
780
+ {
781
+ "epoch": 1.989335520918786,
782
+ "grad_norm": 0.056488387286663055,
783
+ "learning_rate": 5.023171883647426e-07,
784
+ "loss": 0.1994,
785
+ "step": 2425
786
+ },
787
+ {
788
+ "epoch": 2.0,
789
+ "step": 2438,
790
+ "total_flos": 1.58523627405312e+18,
791
+ "train_loss": 0.2603811172308542,
792
+ "train_runtime": 2509.455,
793
+ "train_samples_per_second": 31.082,
794
+ "train_steps_per_second": 0.972
795
+ }
796
+ ],
797
+ "logging_steps": 25,
798
+ "max_steps": 2438,
799
+ "num_input_tokens_seen": 0,
800
+ "num_train_epochs": 2,
801
+ "save_steps": 0,
802
+ "stateful_callbacks": {
803
+ "TrainerControl": {
804
+ "args": {
805
+ "should_epoch_stop": false,
806
+ "should_evaluate": false,
807
+ "should_log": false,
808
+ "should_save": true,
809
+ "should_training_stop": true
810
+ },
811
+ "attributes": {}
812
+ }
813
+ },
814
+ "total_flos": 1.58523627405312e+18,
815
+ "train_batch_size": 32,
816
+ "trial_name": null,
817
+ "trial_params": null
818
+ }
nl_tasks/exps/run_ex20/ft/adapter_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "inference_mode": false,
6
+ "layers_to_transform": null,
7
+ "modules_to_save": null,
8
+ "num_rotations": 1,
9
+ "peft_type": "ROTATION",
10
+ "r": 16,
11
+ "revision": null,
12
+ "target_modules": [
13
+ "v_proj",
14
+ "q_proj"
15
+ ],
16
+ "target_modules_to_skip": null,
17
+ "task_type": "CAUSAL_LM"
18
+ }
nl_tasks/exps/run_ex20/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/exps/run_ex20/ft/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
nl_tasks/exps/run_ex20/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/exps/run_ex20/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/exps/run_ex20/ft2/adapter_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "inference_mode": true,
6
+ "layers_to_transform": null,
7
+ "modules_to_save": null,
8
+ "num_rotations": 1,
9
+ "peft_type": "ROTATION",
10
+ "r": 16,
11
+ "revision": null,
12
+ "target_modules": [
13
+ "v_proj",
14
+ "q_proj"
15
+ ],
16
+ "target_modules_to_skip": null,
17
+ "task_type": "CAUSAL_LM"
18
+ }
nl_tasks/exps/run_ex20/ft2/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96217e29e790c0bb6d7c878f02228e5aaf258e8ee684934cc1cbbf6118b110a1
3
+ size 33602915
nl_tasks/exps/run_ex20/trainer_state.json ADDED
@@ -0,0 +1,1149 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 3.0,
6
+ "eval_steps": 500,
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": 0.9810560345649719,
15
+ "learning_rate": 5.1200000000000004e-05,
16
+ "loss": 0.8888,
17
+ "step": 25
18
+ },
19
+ {
20
+ "epoch": 0.04,
21
+ "grad_norm": 0.4568330645561218,
22
+ "learning_rate": 0.00010453333333333333,
23
+ "loss": 0.53,
24
+ "step": 50
25
+ },
26
+ {
27
+ "epoch": 0.06,
28
+ "grad_norm": 0.28402772545814514,
29
+ "learning_rate": 0.00015786666666666666,
30
+ "loss": 0.4263,
31
+ "step": 75
32
+ },
33
+ {
34
+ "epoch": 0.08,
35
+ "grad_norm": 0.2093808352947235,
36
+ "learning_rate": 0.0002112,
37
+ "loss": 0.3795,
38
+ "step": 100
39
+ },
40
+ {
41
+ "epoch": 0.1,
42
+ "grad_norm": 0.1766759306192398,
43
+ "learning_rate": 0.00026453333333333333,
44
+ "loss": 0.3715,
45
+ "step": 125
46
+ },
47
+ {
48
+ "epoch": 0.12,
49
+ "grad_norm": 0.20399300754070282,
50
+ "learning_rate": 0.0003178666666666667,
51
+ "loss": 0.3579,
52
+ "step": 150
53
+ },
54
+ {
55
+ "epoch": 0.14,
56
+ "grad_norm": 0.36051347851753235,
57
+ "learning_rate": 0.0003712,
58
+ "loss": 0.3542,
59
+ "step": 175
60
+ },
61
+ {
62
+ "epoch": 0.16,
63
+ "grad_norm": 0.1855020970106125,
64
+ "learning_rate": 0.0004245333333333333,
65
+ "loss": 0.3391,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.18,
70
+ "grad_norm": 0.17372916638851166,
71
+ "learning_rate": 0.0004778666666666667,
72
+ "loss": 0.3345,
73
+ "step": 225
74
+ },
75
+ {
76
+ "epoch": 0.2,
77
+ "grad_norm": 0.16113638877868652,
78
+ "learning_rate": 0.0005312,
79
+ "loss": 0.3366,
80
+ "step": 250
81
+ },
82
+ {
83
+ "epoch": 0.22,
84
+ "grad_norm": 0.18627846240997314,
85
+ "learning_rate": 0.0005845333333333334,
86
+ "loss": 0.3255,
87
+ "step": 275
88
+ },
89
+ {
90
+ "epoch": 0.24,
91
+ "grad_norm": 0.19450171291828156,
92
+ "learning_rate": 0.0006378666666666667,
93
+ "loss": 0.3201,
94
+ "step": 300
95
+ },
96
+ {
97
+ "epoch": 0.26,
98
+ "grad_norm": 0.4323269724845886,
99
+ "learning_rate": 0.0006912,
100
+ "loss": 0.3186,
101
+ "step": 325
102
+ },
103
+ {
104
+ "epoch": 0.28,
105
+ "grad_norm": 0.17046867311000824,
106
+ "learning_rate": 0.0007445333333333334,
107
+ "loss": 0.3136,
108
+ "step": 350
109
+ },
110
+ {
111
+ "epoch": 0.3,
112
+ "grad_norm": 0.20966920256614685,
113
+ "learning_rate": 0.0007978666666666667,
114
+ "loss": 0.3057,
115
+ "step": 375
116
+ },
117
+ {
118
+ "epoch": 0.32,
119
+ "grad_norm": 0.2461271584033966,
120
+ "learning_rate": 0.000799900187115249,
121
+ "loss": 0.3246,
122
+ "step": 400
123
+ },
124
+ {
125
+ "epoch": 0.34,
126
+ "grad_norm": 0.20508386194705963,
127
+ "learning_rate": 0.000799583994517958,
128
+ "loss": 0.3017,
129
+ "step": 425
130
+ },
131
+ {
132
+ "epoch": 0.36,
133
+ "grad_norm": 0.21382732689380646,
134
+ "learning_rate": 0.0007990514201385437,
135
+ "loss": 0.3176,
136
+ "step": 450
137
+ },
138
+ {
139
+ "epoch": 0.38,
140
+ "grad_norm": 0.23454499244689941,
141
+ "learning_rate": 0.0007983027523754278,
142
+ "loss": 0.3111,
143
+ "step": 475
144
+ },
145
+ {
146
+ "epoch": 0.4,
147
+ "grad_norm": 0.2275204211473465,
148
+ "learning_rate": 0.000797338396645411,
149
+ "loss": 0.3007,
150
+ "step": 500
151
+ },
152
+ {
153
+ "epoch": 0.4,
154
+ "eval_loss": 0.28894639015197754,
155
+ "eval_runtime": 18.9105,
156
+ "eval_samples_per_second": 54.097,
157
+ "eval_steps_per_second": 0.846,
158
+ "step": 500
159
+ },
160
+ {
161
+ "epoch": 0.42,
162
+ "grad_norm": 0.2434164434671402,
163
+ "learning_rate": 0.0007961588751641338,
164
+ "loss": 0.2989,
165
+ "step": 525
166
+ },
167
+ {
168
+ "epoch": 0.44,
169
+ "grad_norm": 0.24518902599811554,
170
+ "learning_rate": 0.0007947648266632861,
171
+ "loss": 0.2843,
172
+ "step": 550
173
+ },
174
+ {
175
+ "epoch": 0.46,
176
+ "grad_norm": 0.25589656829833984,
177
+ "learning_rate": 0.0007931570060447236,
178
+ "loss": 0.2934,
179
+ "step": 575
180
+ },
181
+ {
182
+ "epoch": 0.48,
183
+ "grad_norm": 0.3379192650318146,
184
+ "learning_rate": 0.0007913362839716751,
185
+ "loss": 0.2896,
186
+ "step": 600
187
+ },
188
+ {
189
+ "epoch": 0.5,
190
+ "grad_norm": 0.33558326959609985,
191
+ "learning_rate": 0.0007893036463972624,
192
+ "loss": 0.286,
193
+ "step": 625
194
+ },
195
+ {
196
+ "epoch": 0.52,
197
+ "grad_norm": 0.21904699504375458,
198
+ "learning_rate": 0.0007870601940305903,
199
+ "loss": 0.2929,
200
+ "step": 650
201
+ },
202
+ {
203
+ "epoch": 0.54,
204
+ "grad_norm": 0.2434883713722229,
205
+ "learning_rate": 0.0007846071417406921,
206
+ "loss": 0.3006,
207
+ "step": 675
208
+ },
209
+ {
210
+ "epoch": 0.56,
211
+ "grad_norm": 0.23831988871097565,
212
+ "learning_rate": 0.0007819458178986573,
213
+ "loss": 0.2874,
214
+ "step": 700
215
+ },
216
+ {
217
+ "epoch": 0.58,
218
+ "grad_norm": 0.2200782746076584,
219
+ "learning_rate": 0.0007790776636582948,
220
+ "loss": 0.2972,
221
+ "step": 725
222
+ },
223
+ {
224
+ "epoch": 0.6,
225
+ "grad_norm": 0.21287283301353455,
226
+ "learning_rate": 0.0007760042321757233,
227
+ "loss": 0.292,
228
+ "step": 750
229
+ },
230
+ {
231
+ "epoch": 0.62,
232
+ "grad_norm": 0.20402580499649048,
233
+ "learning_rate": 0.000772727187768309,
234
+ "loss": 0.2755,
235
+ "step": 775
236
+ },
237
+ {
238
+ "epoch": 0.64,
239
+ "grad_norm": 0.2204650640487671,
240
+ "learning_rate": 0.0007692483050134094,
241
+ "loss": 0.2868,
242
+ "step": 800
243
+ },
244
+ {
245
+ "epoch": 0.66,
246
+ "grad_norm": 0.22145523130893707,
247
+ "learning_rate": 0.0007655694677874075,
248
+ "loss": 0.2737,
249
+ "step": 825
250
+ },
251
+ {
252
+ "epoch": 0.68,
253
+ "grad_norm": 0.1988125890493393,
254
+ "learning_rate": 0.0007616926682455595,
255
+ "loss": 0.2784,
256
+ "step": 850
257
+ },
258
+ {
259
+ "epoch": 0.7,
260
+ "grad_norm": 0.22155867516994476,
261
+ "learning_rate": 0.0007576200057432086,
262
+ "loss": 0.2797,
263
+ "step": 875
264
+ },
265
+ {
266
+ "epoch": 0.72,
267
+ "grad_norm": 0.21928289532661438,
268
+ "learning_rate": 0.0007533536856989458,
269
+ "loss": 0.2882,
270
+ "step": 900
271
+ },
272
+ {
273
+ "epoch": 0.74,
274
+ "grad_norm": 0.19247911870479584,
275
+ "learning_rate": 0.0007488960184003375,
276
+ "loss": 0.2682,
277
+ "step": 925
278
+ },
279
+ {
280
+ "epoch": 0.76,
281
+ "grad_norm": 0.20903678238391876,
282
+ "learning_rate": 0.0007442494177528641,
283
+ "loss": 0.2792,
284
+ "step": 950
285
+ },
286
+ {
287
+ "epoch": 0.78,
288
+ "grad_norm": 0.20451922714710236,
289
+ "learning_rate": 0.0007394163999727472,
290
+ "loss": 0.277,
291
+ "step": 975
292
+ },
293
+ {
294
+ "epoch": 0.8,
295
+ "grad_norm": 0.23223024606704712,
296
+ "learning_rate": 0.0007343995822243741,
297
+ "loss": 0.2799,
298
+ "step": 1000
299
+ },
300
+ {
301
+ "epoch": 0.8,
302
+ "eval_loss": 0.2589196264743805,
303
+ "eval_runtime": 18.712,
304
+ "eval_samples_per_second": 54.671,
305
+ "eval_steps_per_second": 0.855,
306
+ "step": 1000
307
+ },
308
+ {
309
+ "epoch": 0.82,
310
+ "grad_norm": 0.1852971762418747,
311
+ "learning_rate": 0.0007292016812030581,
312
+ "loss": 0.2607,
313
+ "step": 1025
314
+ },
315
+ {
316
+ "epoch": 0.84,
317
+ "grad_norm": 0.2267303317785263,
318
+ "learning_rate": 0.0007238255116638985,
319
+ "loss": 0.2715,
320
+ "step": 1050
321
+ },
322
+ {
323
+ "epoch": 0.86,
324
+ "grad_norm": 0.1889960914850235,
325
+ "learning_rate": 0.0007182739848975422,
326
+ "loss": 0.2657,
327
+ "step": 1075
328
+ },
329
+ {
330
+ "epoch": 0.88,
331
+ "grad_norm": 0.22209246456623077,
332
+ "learning_rate": 0.000712550107153668,
333
+ "loss": 0.2654,
334
+ "step": 1100
335
+ },
336
+ {
337
+ "epoch": 0.9,
338
+ "grad_norm": 0.22123408317565918,
339
+ "learning_rate": 0.0007066569780130499,
340
+ "loss": 0.2815,
341
+ "step": 1125
342
+ },
343
+ {
344
+ "epoch": 0.92,
345
+ "grad_norm": 0.2009263038635254,
346
+ "learning_rate": 0.0007005977887090785,
347
+ "loss": 0.2552,
348
+ "step": 1150
349
+ },
350
+ {
351
+ "epoch": 0.94,
352
+ "grad_norm": 0.17973631620407104,
353
+ "learning_rate": 0.0006943758203996524,
354
+ "loss": 0.262,
355
+ "step": 1175
356
+ },
357
+ {
358
+ "epoch": 0.96,
359
+ "grad_norm": 0.1837298572063446,
360
+ "learning_rate": 0.0006879944423903735,
361
+ "loss": 0.2489,
362
+ "step": 1200
363
+ },
364
+ {
365
+ "epoch": 0.98,
366
+ "grad_norm": 0.2094447761774063,
367
+ "learning_rate": 0.0006814571103100082,
368
+ "loss": 0.2588,
369
+ "step": 1225
370
+ },
371
+ {
372
+ "epoch": 1.0,
373
+ "grad_norm": 0.1988164782524109,
374
+ "learning_rate": 0.0006747673642392035,
375
+ "loss": 0.2516,
376
+ "step": 1250
377
+ },
378
+ {
379
+ "epoch": 1.02,
380
+ "grad_norm": 0.18271949887275696,
381
+ "learning_rate": 0.0006679288267934713,
382
+ "loss": 0.2334,
383
+ "step": 1275
384
+ },
385
+ {
386
+ "epoch": 1.04,
387
+ "grad_norm": 0.22138486802577972,
388
+ "learning_rate": 0.0006609452011614773,
389
+ "loss": 0.2274,
390
+ "step": 1300
391
+ },
392
+ {
393
+ "epoch": 1.06,
394
+ "grad_norm": 0.20442257821559906,
395
+ "learning_rate": 0.0006538202690997,
396
+ "loss": 0.2525,
397
+ "step": 1325
398
+ },
399
+ {
400
+ "epoch": 1.08,
401
+ "grad_norm": 0.16748201847076416,
402
+ "learning_rate": 0.0006465578888845427,
403
+ "loss": 0.2235,
404
+ "step": 1350
405
+ },
406
+ {
407
+ "epoch": 1.1,
408
+ "grad_norm": 0.19223178923130035,
409
+ "learning_rate": 0.0006391619932230076,
410
+ "loss": 0.2439,
411
+ "step": 1375
412
+ },
413
+ {
414
+ "epoch": 1.12,
415
+ "grad_norm": 0.19608688354492188,
416
+ "learning_rate": 0.0006316365871230673,
417
+ "loss": 0.2397,
418
+ "step": 1400
419
+ },
420
+ {
421
+ "epoch": 1.1400000000000001,
422
+ "grad_norm": 0.18464374542236328,
423
+ "learning_rate": 0.0006239857457248809,
424
+ "loss": 0.2306,
425
+ "step": 1425
426
+ },
427
+ {
428
+ "epoch": 1.16,
429
+ "grad_norm": 0.21416647732257843,
430
+ "learning_rate": 0.0006162136120940349,
431
+ "loss": 0.2416,
432
+ "step": 1450
433
+ },
434
+ {
435
+ "epoch": 1.18,
436
+ "grad_norm": 0.19346030056476593,
437
+ "learning_rate": 0.0006083243949779991,
438
+ "loss": 0.234,
439
+ "step": 1475
440
+ },
441
+ {
442
+ "epoch": 1.2,
443
+ "grad_norm": 0.20999440550804138,
444
+ "learning_rate": 0.000600322366527016,
445
+ "loss": 0.2392,
446
+ "step": 1500
447
+ },
448
+ {
449
+ "epoch": 1.2,
450
+ "eval_loss": 0.24392135441303253,
451
+ "eval_runtime": 18.7114,
452
+ "eval_samples_per_second": 54.672,
453
+ "eval_steps_per_second": 0.855,
454
+ "step": 1500
455
+ },
456
+ {
457
+ "epoch": 1.22,
458
+ "grad_norm": 0.1713901311159134,
459
+ "learning_rate": 0.0005922118599806563,
460
+ "loss": 0.2351,
461
+ "step": 1525
462
+ },
463
+ {
464
+ "epoch": 1.24,
465
+ "grad_norm": 0.16384203732013702,
466
+ "learning_rate": 0.0005839972673212931,
467
+ "loss": 0.2292,
468
+ "step": 1550
469
+ },
470
+ {
471
+ "epoch": 1.26,
472
+ "grad_norm": 0.21365433931350708,
473
+ "learning_rate": 0.0005756830368957653,
474
+ "loss": 0.2347,
475
+ "step": 1575
476
+ },
477
+ {
478
+ "epoch": 1.28,
479
+ "grad_norm": 0.1989300400018692,
480
+ "learning_rate": 0.0005672736710065205,
481
+ "loss": 0.2395,
482
+ "step": 1600
483
+ },
484
+ {
485
+ "epoch": 1.3,
486
+ "grad_norm": 0.2063118815422058,
487
+ "learning_rate": 0.0005587737234735383,
488
+ "loss": 0.2216,
489
+ "step": 1625
490
+ },
491
+ {
492
+ "epoch": 1.32,
493
+ "grad_norm": 0.17772799730300903,
494
+ "learning_rate": 0.0005501877971683557,
495
+ "loss": 0.2441,
496
+ "step": 1650
497
+ },
498
+ {
499
+ "epoch": 1.34,
500
+ "grad_norm": 0.14975492656230927,
501
+ "learning_rate": 0.0005415205415215318,
502
+ "loss": 0.2272,
503
+ "step": 1675
504
+ },
505
+ {
506
+ "epoch": 1.3599999999999999,
507
+ "grad_norm": 0.1504419595003128,
508
+ "learning_rate": 0.0005327766500048993,
509
+ "loss": 0.2319,
510
+ "step": 1700
511
+ },
512
+ {
513
+ "epoch": 1.38,
514
+ "grad_norm": 0.18854503333568573,
515
+ "learning_rate": 0.0005239608575899662,
516
+ "loss": 0.2381,
517
+ "step": 1725
518
+ },
519
+ {
520
+ "epoch": 1.4,
521
+ "grad_norm": 0.20569413900375366,
522
+ "learning_rate": 0.0005150779381838451,
523
+ "loss": 0.2332,
524
+ "step": 1750
525
+ },
526
+ {
527
+ "epoch": 1.42,
528
+ "grad_norm": 0.1786058247089386,
529
+ "learning_rate": 0.0005061327020440988,
530
+ "loss": 0.2276,
531
+ "step": 1775
532
+ },
533
+ {
534
+ "epoch": 1.44,
535
+ "grad_norm": 0.21534062922000885,
536
+ "learning_rate": 0.0004971299931739003,
537
+ "loss": 0.2351,
538
+ "step": 1800
539
+ },
540
+ {
541
+ "epoch": 1.46,
542
+ "grad_norm": 0.1800985336303711,
543
+ "learning_rate": 0.0004880746866989206,
544
+ "loss": 0.2349,
545
+ "step": 1825
546
+ },
547
+ {
548
+ "epoch": 1.48,
549
+ "grad_norm": 0.1841922551393509,
550
+ "learning_rate": 0.0004789716862273612,
551
+ "loss": 0.2188,
552
+ "step": 1850
553
+ },
554
+ {
555
+ "epoch": 1.5,
556
+ "grad_norm": 0.17767614126205444,
557
+ "learning_rate": 0.00046982592119456404,
558
+ "loss": 0.2312,
559
+ "step": 1875
560
+ },
561
+ {
562
+ "epoch": 1.52,
563
+ "grad_norm": 0.1671237200498581,
564
+ "learning_rate": 0.00046064234419363565,
565
+ "loss": 0.2247,
566
+ "step": 1900
567
+ },
568
+ {
569
+ "epoch": 1.54,
570
+ "grad_norm": 0.1720089465379715,
571
+ "learning_rate": 0.00045142592829353,
572
+ "loss": 0.2284,
573
+ "step": 1925
574
+ },
575
+ {
576
+ "epoch": 1.56,
577
+ "grad_norm": 0.13443343341350555,
578
+ "learning_rate": 0.00044218166434604485,
579
+ "loss": 0.2157,
580
+ "step": 1950
581
+ },
582
+ {
583
+ "epoch": 1.58,
584
+ "grad_norm": 0.17765246331691742,
585
+ "learning_rate": 0.00043291455828318624,
586
+ "loss": 0.2224,
587
+ "step": 1975
588
+ },
589
+ {
590
+ "epoch": 1.6,
591
+ "grad_norm": 0.18341372907161713,
592
+ "learning_rate": 0.0004236296284063691,
593
+ "loss": 0.2264,
594
+ "step": 2000
595
+ },
596
+ {
597
+ "epoch": 1.6,
598
+ "eval_loss": 0.22886164486408234,
599
+ "eval_runtime": 18.7265,
600
+ "eval_samples_per_second": 54.628,
601
+ "eval_steps_per_second": 0.854,
602
+ "step": 2000
603
+ },
604
+ {
605
+ "epoch": 1.62,
606
+ "grad_norm": 0.16667109727859497,
607
+ "learning_rate": 0.0004143319026689178,
608
+ "loss": 0.2152,
609
+ "step": 2025
610
+ },
611
+ {
612
+ "epoch": 1.6400000000000001,
613
+ "grad_norm": 0.16984052956104279,
614
+ "learning_rate": 0.0004050264159533411,
615
+ "loss": 0.2228,
616
+ "step": 2050
617
+ },
618
+ {
619
+ "epoch": 1.6600000000000001,
620
+ "grad_norm": 0.164903461933136,
621
+ "learning_rate": 0.0003957182073448542,
622
+ "loss": 0.2278,
623
+ "step": 2075
624
+ },
625
+ {
626
+ "epoch": 1.6800000000000002,
627
+ "grad_norm": 0.18476568162441254,
628
+ "learning_rate": 0.0003864123174026256,
629
+ "loss": 0.2286,
630
+ "step": 2100
631
+ },
632
+ {
633
+ "epoch": 1.7,
634
+ "grad_norm": 0.19166547060012817,
635
+ "learning_rate": 0.00037711378543022476,
636
+ "loss": 0.2282,
637
+ "step": 2125
638
+ },
639
+ {
640
+ "epoch": 1.72,
641
+ "grad_norm": 0.16085763275623322,
642
+ "learning_rate": 0.0003678276467467509,
643
+ "loss": 0.2226,
644
+ "step": 2150
645
+ },
646
+ {
647
+ "epoch": 1.74,
648
+ "grad_norm": 0.1440000832080841,
649
+ "learning_rate": 0.0003585589299601187,
650
+ "loss": 0.2168,
651
+ "step": 2175
652
+ },
653
+ {
654
+ "epoch": 1.76,
655
+ "grad_norm": 0.18244686722755432,
656
+ "learning_rate": 0.0003493126542439782,
657
+ "loss": 0.2209,
658
+ "step": 2200
659
+ },
660
+ {
661
+ "epoch": 1.78,
662
+ "grad_norm": 0.1721377968788147,
663
+ "learning_rate": 0.0003400938266197439,
664
+ "loss": 0.2236,
665
+ "step": 2225
666
+ },
667
+ {
668
+ "epoch": 1.8,
669
+ "grad_norm": 0.16125494241714478,
670
+ "learning_rate": 0.00033090743924520436,
671
+ "loss": 0.219,
672
+ "step": 2250
673
+ },
674
+ {
675
+ "epoch": 1.8199999999999998,
676
+ "grad_norm": 0.16669046878814697,
677
+ "learning_rate": 0.0003217584667111808,
678
+ "loss": 0.2151,
679
+ "step": 2275
680
+ },
681
+ {
682
+ "epoch": 1.8399999999999999,
683
+ "grad_norm": 0.1670670360326767,
684
+ "learning_rate": 0.00031265186334769796,
685
+ "loss": 0.2101,
686
+ "step": 2300
687
+ },
688
+ {
689
+ "epoch": 1.8599999999999999,
690
+ "grad_norm": 0.15344752371311188,
691
+ "learning_rate": 0.00030359256054112745,
692
+ "loss": 0.2215,
693
+ "step": 2325
694
+ },
695
+ {
696
+ "epoch": 1.88,
697
+ "grad_norm": 0.1619005799293518,
698
+ "learning_rate": 0.000294585464063755,
699
+ "loss": 0.2078,
700
+ "step": 2350
701
+ },
702
+ {
703
+ "epoch": 1.9,
704
+ "grad_norm": 0.18533560633659363,
705
+ "learning_rate": 0.0002856354514172179,
706
+ "loss": 0.213,
707
+ "step": 2375
708
+ },
709
+ {
710
+ "epoch": 1.92,
711
+ "grad_norm": 0.16015726327896118,
712
+ "learning_rate": 0.0002767473691912522,
713
+ "loss": 0.2115,
714
+ "step": 2400
715
+ },
716
+ {
717
+ "epoch": 1.94,
718
+ "grad_norm": 0.16600584983825684,
719
+ "learning_rate": 0.0002679260304391785,
720
+ "loss": 0.2135,
721
+ "step": 2425
722
+ },
723
+ {
724
+ "epoch": 1.96,
725
+ "grad_norm": 0.16894803941249847,
726
+ "learning_rate": 0.00025917621207154914,
727
+ "loss": 0.2015,
728
+ "step": 2450
729
+ },
730
+ {
731
+ "epoch": 1.98,
732
+ "grad_norm": 0.16149814426898956,
733
+ "learning_rate": 0.00025050265226936617,
734
+ "loss": 0.2114,
735
+ "step": 2475
736
+ },
737
+ {
738
+ "epoch": 2.0,
739
+ "grad_norm": 0.13848654925823212,
740
+ "learning_rate": 0.0002419100479182735,
741
+ "loss": 0.2125,
742
+ "step": 2500
743
+ },
744
+ {
745
+ "epoch": 2.0,
746
+ "eval_loss": 0.21840418875217438,
747
+ "eval_runtime": 18.7178,
748
+ "eval_samples_per_second": 54.654,
749
+ "eval_steps_per_second": 0.855,
750
+ "step": 2500
751
+ },
752
+ {
753
+ "epoch": 2.02,
754
+ "grad_norm": 0.1629616618156433,
755
+ "learning_rate": 0.0002334030520651101,
756
+ "loss": 0.1818,
757
+ "step": 2525
758
+ },
759
+ {
760
+ "epoch": 2.04,
761
+ "grad_norm": 0.16421562433242798,
762
+ "learning_rate": 0.00022498627139820287,
763
+ "loss": 0.186,
764
+ "step": 2550
765
+ },
766
+ {
767
+ "epoch": 2.06,
768
+ "grad_norm": 0.16993357241153717,
769
+ "learning_rate": 0.00021666426375276405,
770
+ "loss": 0.1894,
771
+ "step": 2575
772
+ },
773
+ {
774
+ "epoch": 2.08,
775
+ "grad_norm": 0.1614699810743332,
776
+ "learning_rate": 0.00020844153564274307,
777
+ "loss": 0.1889,
778
+ "step": 2600
779
+ },
780
+ {
781
+ "epoch": 2.1,
782
+ "grad_norm": 0.16604189574718475,
783
+ "learning_rate": 0.00020032253982046894,
784
+ "loss": 0.192,
785
+ "step": 2625
786
+ },
787
+ {
788
+ "epoch": 2.12,
789
+ "grad_norm": 0.186918705701828,
790
+ "learning_rate": 0.00019231167286540743,
791
+ "loss": 0.1873,
792
+ "step": 2650
793
+ },
794
+ {
795
+ "epoch": 2.14,
796
+ "grad_norm": 0.14999671280384064,
797
+ "learning_rate": 0.00018441327280333512,
798
+ "loss": 0.1801,
799
+ "step": 2675
800
+ },
801
+ {
802
+ "epoch": 2.16,
803
+ "grad_norm": 0.17705227434635162,
804
+ "learning_rate": 0.00017663161675722164,
805
+ "loss": 0.189,
806
+ "step": 2700
807
+ },
808
+ {
809
+ "epoch": 2.18,
810
+ "grad_norm": 0.15323910117149353,
811
+ "learning_rate": 0.00016897091863109305,
812
+ "loss": 0.1862,
813
+ "step": 2725
814
+ },
815
+ {
816
+ "epoch": 2.2,
817
+ "grad_norm": 0.16227515041828156,
818
+ "learning_rate": 0.00016143532682812728,
819
+ "loss": 0.1812,
820
+ "step": 2750
821
+ },
822
+ {
823
+ "epoch": 2.22,
824
+ "grad_norm": 0.17776158452033997,
825
+ "learning_rate": 0.00015402892200422192,
826
+ "loss": 0.1865,
827
+ "step": 2775
828
+ },
829
+ {
830
+ "epoch": 2.24,
831
+ "grad_norm": 0.1510823518037796,
832
+ "learning_rate": 0.00014675571485824577,
833
+ "loss": 0.187,
834
+ "step": 2800
835
+ },
836
+ {
837
+ "epoch": 2.26,
838
+ "grad_norm": 0.17457172274589539,
839
+ "learning_rate": 0.00013961964396017572,
840
+ "loss": 0.1889,
841
+ "step": 2825
842
+ },
843
+ {
844
+ "epoch": 2.2800000000000002,
845
+ "grad_norm": 0.16321052610874176,
846
+ "learning_rate": 0.00013262457361829118,
847
+ "loss": 0.1828,
848
+ "step": 2850
849
+ },
850
+ {
851
+ "epoch": 2.3,
852
+ "grad_norm": 0.1677648425102234,
853
+ "learning_rate": 0.0001257742917865826,
854
+ "loss": 0.1863,
855
+ "step": 2875
856
+ },
857
+ {
858
+ "epoch": 2.32,
859
+ "grad_norm": 0.16572973132133484,
860
+ "learning_rate": 0.00011907250801350831,
861
+ "loss": 0.1912,
862
+ "step": 2900
863
+ },
864
+ {
865
+ "epoch": 2.34,
866
+ "grad_norm": 0.15434414148330688,
867
+ "learning_rate": 0.00011252285143320827,
868
+ "loss": 0.1912,
869
+ "step": 2925
870
+ },
871
+ {
872
+ "epoch": 2.36,
873
+ "grad_norm": 0.14998921751976013,
874
+ "learning_rate": 0.00010612886880026343,
875
+ "loss": 0.1884,
876
+ "step": 2950
877
+ },
878
+ {
879
+ "epoch": 2.38,
880
+ "grad_norm": 0.16378054022789001,
881
+ "learning_rate": 9.98940225690665e-05,
882
+ "loss": 0.1831,
883
+ "step": 2975
884
+ },
885
+ {
886
+ "epoch": 2.4,
887
+ "grad_norm": 0.20425669848918915,
888
+ "learning_rate": 9.382168901884095e-05,
889
+ "loss": 0.1805,
890
+ "step": 3000
891
+ },
892
+ {
893
+ "epoch": 2.4,
894
+ "eval_loss": 0.21560731530189514,
895
+ "eval_runtime": 18.7254,
896
+ "eval_samples_per_second": 54.632,
897
+ "eval_steps_per_second": 0.854,
898
+ "step": 3000
899
+ },
900
+ {
901
+ "epoch": 2.42,
902
+ "grad_norm": 0.22065085172653198,
903
+ "learning_rate": 8.791515642532666e-05,
904
+ "loss": 0.1858,
905
+ "step": 3025
906
+ },
907
+ {
908
+ "epoch": 2.44,
909
+ "grad_norm": 0.1566791832447052,
910
+ "learning_rate": 8.217762328012067e-05,
911
+ "loss": 0.1774,
912
+ "step": 3050
913
+ },
914
+ {
915
+ "epoch": 2.46,
916
+ "grad_norm": 0.16523855924606323,
917
+ "learning_rate": 7.661219655863753e-05,
918
+ "loss": 0.1878,
919
+ "step": 3075
920
+ },
921
+ {
922
+ "epoch": 2.48,
923
+ "grad_norm": 0.1506161093711853,
924
+ "learning_rate": 7.122189003762678e-05,
925
+ "loss": 0.1892,
926
+ "step": 3100
927
+ },
928
+ {
929
+ "epoch": 2.5,
930
+ "grad_norm": 0.16826751828193665,
931
+ "learning_rate": 6.600962266316053e-05,
932
+ "loss": 0.1789,
933
+ "step": 3125
934
+ },
935
+ {
936
+ "epoch": 2.52,
937
+ "grad_norm": 0.1406136155128479,
938
+ "learning_rate": 6.0978216969972104e-05,
939
+ "loss": 0.183,
940
+ "step": 3150
941
+ },
942
+ {
943
+ "epoch": 2.54,
944
+ "grad_norm": 0.17880959808826447,
945
+ "learning_rate": 5.61303975530036e-05,
946
+ "loss": 0.1901,
947
+ "step": 3175
948
+ },
949
+ {
950
+ "epoch": 2.56,
951
+ "grad_norm": 0.16088391840457916,
952
+ "learning_rate": 5.146878959199013e-05,
953
+ "loss": 0.1829,
954
+ "step": 3200
955
+ },
956
+ {
957
+ "epoch": 2.58,
958
+ "grad_norm": 0.16157594323158264,
959
+ "learning_rate": 4.69959174298785e-05,
960
+ "loss": 0.1779,
961
+ "step": 3225
962
+ },
963
+ {
964
+ "epoch": 2.6,
965
+ "grad_norm": 0.1672293096780777,
966
+ "learning_rate": 4.271420320585064e-05,
967
+ "loss": 0.189,
968
+ "step": 3250
969
+ },
970
+ {
971
+ "epoch": 2.62,
972
+ "grad_norm": 0.17222677171230316,
973
+ "learning_rate": 3.862596554369287e-05,
974
+ "loss": 0.1731,
975
+ "step": 3275
976
+ },
977
+ {
978
+ "epoch": 2.64,
979
+ "grad_norm": 0.15946194529533386,
980
+ "learning_rate": 3.473341829622023e-05,
981
+ "loss": 0.1787,
982
+ "step": 3300
983
+ },
984
+ {
985
+ "epoch": 2.66,
986
+ "grad_norm": 0.17513757944107056,
987
+ "learning_rate": 3.103866934643604e-05,
988
+ "loss": 0.1869,
989
+ "step": 3325
990
+ },
991
+ {
992
+ "epoch": 2.68,
993
+ "grad_norm": 0.18372221291065216,
994
+ "learning_rate": 2.754371946607579e-05,
995
+ "loss": 0.1751,
996
+ "step": 3350
997
+ },
998
+ {
999
+ "epoch": 2.7,
1000
+ "grad_norm": 0.17666997015476227,
1001
+ "learning_rate": 2.4250461232154398e-05,
1002
+ "loss": 0.1846,
1003
+ "step": 3375
1004
+ },
1005
+ {
1006
+ "epoch": 2.7199999999999998,
1007
+ "grad_norm": 0.1679190993309021,
1008
+ "learning_rate": 2.1160678002102307e-05,
1009
+ "loss": 0.184,
1010
+ "step": 3400
1011
+ },
1012
+ {
1013
+ "epoch": 2.74,
1014
+ "grad_norm": 0.16586284339427948,
1015
+ "learning_rate": 1.8276042948045523e-05,
1016
+ "loss": 0.1861,
1017
+ "step": 3425
1018
+ },
1019
+ {
1020
+ "epoch": 2.76,
1021
+ "grad_norm": 0.16841112077236176,
1022
+ "learning_rate": 1.5598118150753984e-05,
1023
+ "loss": 0.1866,
1024
+ "step": 3450
1025
+ },
1026
+ {
1027
+ "epoch": 2.7800000000000002,
1028
+ "grad_norm": 0.18072092533111572,
1029
+ "learning_rate": 1.3128353753746858e-05,
1030
+ "loss": 0.1851,
1031
+ "step": 3475
1032
+ },
1033
+ {
1034
+ "epoch": 2.8,
1035
+ "grad_norm": 0.15921686589717865,
1036
+ "learning_rate": 1.0868087178014463e-05,
1037
+ "loss": 0.1847,
1038
+ "step": 3500
1039
+ },
1040
+ {
1041
+ "epoch": 2.8,
1042
+ "eval_loss": 0.21289591491222382,
1043
+ "eval_runtime": 18.7421,
1044
+ "eval_samples_per_second": 54.583,
1045
+ "eval_steps_per_second": 0.854,
1046
+ "step": 3500
1047
+ },
1048
+ {
1049
+ "epoch": 2.82,
1050
+ "grad_norm": 0.16726893186569214,
1051
+ "learning_rate": 8.818542397781038e-06,
1052
+ "loss": 0.1788,
1053
+ "step": 3525
1054
+ },
1055
+ {
1056
+ "epoch": 2.84,
1057
+ "grad_norm": 0.169475719332695,
1058
+ "learning_rate": 6.980829277701295e-06,
1059
+ "loss": 0.1844,
1060
+ "step": 3550
1061
+ },
1062
+ {
1063
+ "epoch": 2.86,
1064
+ "grad_norm": 0.1342073529958725,
1065
+ "learning_rate": 5.355942971849004e-06,
1066
+ "loss": 0.1819,
1067
+ "step": 3575
1068
+ },
1069
+ {
1070
+ "epoch": 2.88,
1071
+ "grad_norm": 0.15730345249176025,
1072
+ "learning_rate": 3.9447633848234796e-06,
1073
+ "loss": 0.1856,
1074
+ "step": 3600
1075
+ },
1076
+ {
1077
+ "epoch": 2.9,
1078
+ "grad_norm": 0.1773420125246048,
1079
+ "learning_rate": 2.7480546952658448e-06,
1080
+ "loss": 0.1798,
1081
+ "step": 3625
1082
+ },
1083
+ {
1084
+ "epoch": 2.92,
1085
+ "grad_norm": 0.13879047334194183,
1086
+ "learning_rate": 1.7664649420428004e-06,
1087
+ "loss": 0.1906,
1088
+ "step": 3650
1089
+ },
1090
+ {
1091
+ "epoch": 2.94,
1092
+ "grad_norm": 0.16108667850494385,
1093
+ "learning_rate": 1.0005256733218726e-06,
1094
+ "loss": 0.1818,
1095
+ "step": 3675
1096
+ },
1097
+ {
1098
+ "epoch": 2.96,
1099
+ "grad_norm": 0.17699511349201202,
1100
+ "learning_rate": 4.5065165872917314e-07,
1101
+ "loss": 0.1861,
1102
+ "step": 3700
1103
+ },
1104
+ {
1105
+ "epoch": 2.98,
1106
+ "grad_norm": 0.17515534162521362,
1107
+ "learning_rate": 1.1714066474399765e-07,
1108
+ "loss": 0.1825,
1109
+ "step": 3725
1110
+ },
1111
+ {
1112
+ "epoch": 3.0,
1113
+ "grad_norm": 0.15442979335784912,
1114
+ "learning_rate": 1.732934529652397e-10,
1115
+ "loss": 0.1817,
1116
+ "step": 3750
1117
+ },
1118
+ {
1119
+ "epoch": 3.0,
1120
+ "step": 3750,
1121
+ "total_flos": 2.43882352705536e+18,
1122
+ "train_loss": 0.24216656379699708,
1123
+ "train_runtime": 3619.9124,
1124
+ "train_samples_per_second": 33.15,
1125
+ "train_steps_per_second": 1.036
1126
+ }
1127
+ ],
1128
+ "logging_steps": 25,
1129
+ "max_steps": 3750,
1130
+ "num_input_tokens_seen": 0,
1131
+ "num_train_epochs": 3,
1132
+ "save_steps": 0,
1133
+ "stateful_callbacks": {
1134
+ "TrainerControl": {
1135
+ "args": {
1136
+ "should_epoch_stop": false,
1137
+ "should_evaluate": false,
1138
+ "should_log": false,
1139
+ "should_save": true,
1140
+ "should_training_stop": true
1141
+ },
1142
+ "attributes": {}
1143
+ }
1144
+ },
1145
+ "total_flos": 2.43882352705536e+18,
1146
+ "train_batch_size": 32,
1147
+ "trial_name": null,
1148
+ "trial_params": null
1149
+ }
nl_tasks/exps/run_ex21/ft/adapter_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "T": 1.0,
3
+ "base_model_name_or_path": "meta-llama/Llama-2-7b-hf",
4
+ "bias": "none",
5
+ "inference_mode": false,
6
+ "layers_to_transform": null,
7
+ "modules_to_save": null,
8
+ "num_rotations": 1,
9
+ "peft_type": "ROTATION",
10
+ "r": 16,
11
+ "revision": null,
12
+ "target_modules": [
13
+ "q_proj",
14
+ "v_proj"
15
+ ],
16
+ "target_modules_to_skip": null,
17
+ "task_type": "CAUSAL_LM"
18
+ }
nl_tasks/exps/run_ex21/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/exps/run_ex21/ft/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
nl_tasks/exps/run_ex21/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/exps/run_ex21/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/exps/run_ex21/ft2/adapter_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f116a6cd4eec4e07751d0d7bcd8913ccc364f6a2c5bc1d8d9345f568919f757
3
+ size 33602915
nl_tasks/exps/run_ex21/trainer_state.json ADDED
@@ -0,0 +1,783 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 2.0,
6
+ "eval_steps": 500,
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": 1.0315879583358765,
15
+ "learning_rate": 9.6e-05,
16
+ "loss": 0.8325,
17
+ "step": 25
18
+ },
19
+ {
20
+ "epoch": 0.04,
21
+ "grad_norm": 0.20474183559417725,
22
+ "learning_rate": 0.00019600000000000002,
23
+ "loss": 0.4556,
24
+ "step": 50
25
+ },
26
+ {
27
+ "epoch": 0.06,
28
+ "grad_norm": 0.22688248753547668,
29
+ "learning_rate": 0.000296,
30
+ "loss": 0.4015,
31
+ "step": 75
32
+ },
33
+ {
34
+ "epoch": 0.08,
35
+ "grad_norm": 0.19172126054763794,
36
+ "learning_rate": 0.00039600000000000003,
37
+ "loss": 0.3645,
38
+ "step": 100
39
+ },
40
+ {
41
+ "epoch": 0.1,
42
+ "grad_norm": 0.15229463577270508,
43
+ "learning_rate": 0.000496,
44
+ "loss": 0.3602,
45
+ "step": 125
46
+ },
47
+ {
48
+ "epoch": 0.12,
49
+ "grad_norm": 0.20351602137088776,
50
+ "learning_rate": 0.000596,
51
+ "loss": 0.3475,
52
+ "step": 150
53
+ },
54
+ {
55
+ "epoch": 0.14,
56
+ "grad_norm": 0.19812420010566711,
57
+ "learning_rate": 0.000696,
58
+ "loss": 0.345,
59
+ "step": 175
60
+ },
61
+ {
62
+ "epoch": 0.16,
63
+ "grad_norm": 0.18347425758838654,
64
+ "learning_rate": 0.000796,
65
+ "loss": 0.3327,
66
+ "step": 200
67
+ },
68
+ {
69
+ "epoch": 0.18,
70
+ "grad_norm": 0.2595237195491791,
71
+ "learning_rate": 0.000896,
72
+ "loss": 0.3302,
73
+ "step": 225
74
+ },
75
+ {
76
+ "epoch": 0.2,
77
+ "grad_norm": 0.2259773463010788,
78
+ "learning_rate": 0.000996,
79
+ "loss": 0.3392,
80
+ "step": 250
81
+ },
82
+ {
83
+ "epoch": 0.22,
84
+ "grad_norm": 0.2506314516067505,
85
+ "learning_rate": 0.0009997192908557321,
86
+ "loss": 0.328,
87
+ "step": 275
88
+ },
89
+ {
90
+ "epoch": 0.24,
91
+ "grad_norm": 0.22288699448108673,
92
+ "learning_rate": 0.000998830238119205,
93
+ "loss": 0.3248,
94
+ "step": 300
95
+ },
96
+ {
97
+ "epoch": 0.26,
98
+ "grad_norm": 0.29342198371887207,
99
+ "learning_rate": 0.000997333437576437,
100
+ "loss": 0.3206,
101
+ "step": 325
102
+ },
103
+ {
104
+ "epoch": 0.28,
105
+ "grad_norm": 0.23638910055160522,
106
+ "learning_rate": 0.0009952307128483257,
107
+ "loss": 0.3119,
108
+ "step": 350
109
+ },
110
+ {
111
+ "epoch": 0.3,
112
+ "grad_norm": 0.20430760085582733,
113
+ "learning_rate": 0.0009925246257810518,
114
+ "loss": 0.3029,
115
+ "step": 375
116
+ },
117
+ {
118
+ "epoch": 0.32,
119
+ "grad_norm": 0.2517499327659607,
120
+ "learning_rate": 0.0009892184733248665,
121
+ "loss": 0.3218,
122
+ "step": 400
123
+ },
124
+ {
125
+ "epoch": 0.34,
126
+ "grad_norm": 0.2284773737192154,
127
+ "learning_rate": 0.0009853162835172637,
128
+ "loss": 0.2969,
129
+ "step": 425
130
+ },
131
+ {
132
+ "epoch": 0.36,
133
+ "grad_norm": 0.26538360118865967,
134
+ "learning_rate": 0.0009808228105754376,
135
+ "loss": 0.3134,
136
+ "step": 450
137
+ },
138
+ {
139
+ "epoch": 0.38,
140
+ "grad_norm": 0.24021762609481812,
141
+ "learning_rate": 0.0009757435291040016,
142
+ "loss": 0.3076,
143
+ "step": 475
144
+ },
145
+ {
146
+ "epoch": 0.4,
147
+ "grad_norm": 0.27008429169654846,
148
+ "learning_rate": 0.0009700846274250251,
149
+ "loss": 0.2964,
150
+ "step": 500
151
+ },
152
+ {
153
+ "epoch": 0.4,
154
+ "eval_loss": 0.2879575788974762,
155
+ "eval_runtime": 18.9021,
156
+ "eval_samples_per_second": 54.121,
157
+ "eval_steps_per_second": 0.846,
158
+ "step": 500
159
+ },
160
+ {
161
+ "epoch": 0.42,
162
+ "grad_norm": 0.2501317858695984,
163
+ "learning_rate": 0.000963853000038517,
164
+ "loss": 0.2981,
165
+ "step": 525
166
+ },
167
+ {
168
+ "epoch": 0.44,
169
+ "grad_norm": 0.19082719087600708,
170
+ "learning_rate": 0.0009570562392225395,
171
+ "loss": 0.2809,
172
+ "step": 550
173
+ },
174
+ {
175
+ "epoch": 0.46,
176
+ "grad_norm": 0.24488045275211334,
177
+ "learning_rate": 0.0009497026257831855,
178
+ "loss": 0.291,
179
+ "step": 575
180
+ },
181
+ {
182
+ "epoch": 0.48,
183
+ "grad_norm": 0.21422015130519867,
184
+ "learning_rate": 0.0009418011189656941,
185
+ "loss": 0.2851,
186
+ "step": 600
187
+ },
188
+ {
189
+ "epoch": 0.5,
190
+ "grad_norm": 0.19914306700229645,
191
+ "learning_rate": 0.0009333613455389882,
192
+ "loss": 0.2841,
193
+ "step": 625
194
+ },
195
+ {
196
+ "epoch": 0.52,
197
+ "grad_norm": 0.2367725670337677,
198
+ "learning_rate": 0.000924393588066941,
199
+ "loss": 0.2879,
200
+ "step": 650
201
+ },
202
+ {
203
+ "epoch": 0.54,
204
+ "grad_norm": 0.24288254976272583,
205
+ "learning_rate": 0.0009149087723806549,
206
+ "loss": 0.2986,
207
+ "step": 675
208
+ },
209
+ {
210
+ "epoch": 0.56,
211
+ "grad_norm": 0.22942474484443665,
212
+ "learning_rate": 0.0009049184542670199,
213
+ "loss": 0.2842,
214
+ "step": 700
215
+ },
216
+ {
217
+ "epoch": 0.58,
218
+ "grad_norm": 0.23413586616516113,
219
+ "learning_rate": 0.0008944348053897671,
220
+ "loss": 0.2925,
221
+ "step": 725
222
+ },
223
+ {
224
+ "epoch": 0.6,
225
+ "grad_norm": 0.22139714658260345,
226
+ "learning_rate": 0.0008834705984601709,
227
+ "loss": 0.2873,
228
+ "step": 750
229
+ },
230
+ {
231
+ "epoch": 0.62,
232
+ "grad_norm": 0.2082182914018631,
233
+ "learning_rate": 0.0008720391916754683,
234
+ "loss": 0.2712,
235
+ "step": 775
236
+ },
237
+ {
238
+ "epoch": 0.64,
239
+ "grad_norm": 0.24570266902446747,
240
+ "learning_rate": 0.0008601545124439535,
241
+ "loss": 0.2813,
242
+ "step": 800
243
+ },
244
+ {
245
+ "epoch": 0.66,
246
+ "grad_norm": 0.20145930349826813,
247
+ "learning_rate": 0.0008478310404165754,
248
+ "loss": 0.2678,
249
+ "step": 825
250
+ },
251
+ {
252
+ "epoch": 0.68,
253
+ "grad_norm": 0.1796109676361084,
254
+ "learning_rate": 0.0008350837898457143,
255
+ "loss": 0.2732,
256
+ "step": 850
257
+ },
258
+ {
259
+ "epoch": 0.7,
260
+ "grad_norm": 0.22070881724357605,
261
+ "learning_rate": 0.0008219282912926269,
262
+ "loss": 0.2744,
263
+ "step": 875
264
+ },
265
+ {
266
+ "epoch": 0.72,
267
+ "grad_norm": 0.18907737731933594,
268
+ "learning_rate": 0.0008083805727058513,
269
+ "loss": 0.2834,
270
+ "step": 900
271
+ },
272
+ {
273
+ "epoch": 0.74,
274
+ "grad_norm": 0.1631525754928589,
275
+ "learning_rate": 0.0007944571398936193,
276
+ "loss": 0.2615,
277
+ "step": 925
278
+ },
279
+ {
280
+ "epoch": 0.76,
281
+ "grad_norm": 0.1881309598684311,
282
+ "learning_rate": 0.0007801749564140723,
283
+ "loss": 0.2732,
284
+ "step": 950
285
+ },
286
+ {
287
+ "epoch": 0.78,
288
+ "grad_norm": 0.19089096784591675,
289
+ "learning_rate": 0.0007655514229077783,
290
+ "loss": 0.2705,
291
+ "step": 975
292
+ },
293
+ {
294
+ "epoch": 0.8,
295
+ "grad_norm": 0.17820408940315247,
296
+ "learning_rate": 0.0007506043558977322,
297
+ "loss": 0.2716,
298
+ "step": 1000
299
+ },
300
+ {
301
+ "epoch": 0.8,
302
+ "eval_loss": 0.2520262598991394,
303
+ "eval_runtime": 18.7237,
304
+ "eval_samples_per_second": 54.637,
305
+ "eval_steps_per_second": 0.855,
306
+ "step": 1000
307
+ },
308
+ {
309
+ "epoch": 0.82,
310
+ "grad_norm": 0.1798166185617447,
311
+ "learning_rate": 0.0007353519660826664,
312
+ "loss": 0.2546,
313
+ "step": 1025
314
+ },
315
+ {
316
+ "epoch": 0.84,
317
+ "grad_norm": 0.21515925228595734,
318
+ "learning_rate": 0.00071981283615012,
319
+ "loss": 0.2647,
320
+ "step": 1050
321
+ },
322
+ {
323
+ "epoch": 0.86,
324
+ "grad_norm": 0.1794830709695816,
325
+ "learning_rate": 0.0007040058981362964,
326
+ "loss": 0.2577,
327
+ "step": 1075
328
+ },
329
+ {
330
+ "epoch": 0.88,
331
+ "grad_norm": 0.18966132402420044,
332
+ "learning_rate": 0.0006879504103602934,
333
+ "loss": 0.2594,
334
+ "step": 1100
335
+ },
336
+ {
337
+ "epoch": 0.9,
338
+ "grad_norm": 0.21163836121559143,
339
+ "learning_rate": 0.0006716659339608077,
340
+ "loss": 0.2725,
341
+ "step": 1125
342
+ },
343
+ {
344
+ "epoch": 0.92,
345
+ "grad_norm": 0.17575646936893463,
346
+ "learning_rate": 0.0006551723090639006,
347
+ "loss": 0.2483,
348
+ "step": 1150
349
+ },
350
+ {
351
+ "epoch": 0.94,
352
+ "grad_norm": 0.17414060235023499,
353
+ "learning_rate": 0.0006384896306108612,
354
+ "loss": 0.2555,
355
+ "step": 1175
356
+ },
357
+ {
358
+ "epoch": 0.96,
359
+ "grad_norm": 0.16540151834487915,
360
+ "learning_rate": 0.0006216382238756146,
361
+ "loss": 0.2427,
362
+ "step": 1200
363
+ },
364
+ {
365
+ "epoch": 0.98,
366
+ "grad_norm": 0.16461016237735748,
367
+ "learning_rate": 0.0006046386197015076,
368
+ "loss": 0.251,
369
+ "step": 1225
370
+ },
371
+ {
372
+ "epoch": 1.0,
373
+ "grad_norm": 0.16811122000217438,
374
+ "learning_rate": 0.0005875115294876381,
375
+ "loss": 0.2451,
376
+ "step": 1250
377
+ },
378
+ {
379
+ "epoch": 1.02,
380
+ "grad_norm": 0.17021119594573975,
381
+ "learning_rate": 0.0005702778199552054,
382
+ "loss": 0.2227,
383
+ "step": 1275
384
+ },
385
+ {
386
+ "epoch": 1.04,
387
+ "grad_norm": 0.16884687542915344,
388
+ "learning_rate": 0.000552958487724626,
389
+ "loss": 0.2164,
390
+ "step": 1300
391
+ },
392
+ {
393
+ "epoch": 1.06,
394
+ "grad_norm": 0.17195427417755127,
395
+ "learning_rate": 0.0005355746337343836,
396
+ "loss": 0.2406,
397
+ "step": 1325
398
+ },
399
+ {
400
+ "epoch": 1.08,
401
+ "grad_norm": 0.15978871285915375,
402
+ "learning_rate": 0.0005181474375327879,
403
+ "loss": 0.2138,
404
+ "step": 1350
405
+ },
406
+ {
407
+ "epoch": 1.1,
408
+ "grad_norm": 0.1938290148973465,
409
+ "learning_rate": 0.0005006981314739573,
410
+ "loss": 0.2322,
411
+ "step": 1375
412
+ },
413
+ {
414
+ "epoch": 1.12,
415
+ "grad_norm": 0.1653754562139511,
416
+ "learning_rate": 0.00048324797484946424,
417
+ "loss": 0.2285,
418
+ "step": 1400
419
+ },
420
+ {
421
+ "epoch": 1.1400000000000001,
422
+ "grad_norm": 0.16597716510295868,
423
+ "learning_rate": 0.0004658182279871657,
424
+ "loss": 0.2195,
425
+ "step": 1425
426
+ },
427
+ {
428
+ "epoch": 1.16,
429
+ "grad_norm": 0.1771618276834488,
430
+ "learning_rate": 0.00044843012634876645,
431
+ "loss": 0.2311,
432
+ "step": 1450
433
+ },
434
+ {
435
+ "epoch": 1.18,
436
+ "grad_norm": 0.1888245791196823,
437
+ "learning_rate": 0.000431104854657681,
438
+ "loss": 0.2224,
439
+ "step": 1475
440
+ },
441
+ {
442
+ "epoch": 1.2,
443
+ "grad_norm": 0.19004148244857788,
444
+ "learning_rate": 0.0004138635210887117,
445
+ "loss": 0.2261,
446
+ "step": 1500
447
+ },
448
+ {
449
+ "epoch": 1.2,
450
+ "eval_loss": 0.23447048664093018,
451
+ "eval_runtime": 18.7321,
452
+ "eval_samples_per_second": 54.612,
453
+ "eval_steps_per_second": 0.854,
454
+ "step": 1500
455
+ },
456
+ {
457
+ "epoch": 1.22,
458
+ "grad_norm": 0.15263979136943817,
459
+ "learning_rate": 0.0003967271315509884,
460
+ "loss": 0.2227,
461
+ "step": 1525
462
+ },
463
+ {
464
+ "epoch": 1.24,
465
+ "grad_norm": 0.16230420768260956,
466
+ "learning_rate": 0.0003797165640955041,
467
+ "loss": 0.218,
468
+ "step": 1550
469
+ },
470
+ {
471
+ "epoch": 1.26,
472
+ "grad_norm": 0.1780393123626709,
473
+ "learning_rate": 0.0003628525434784268,
474
+ "loss": 0.2228,
475
+ "step": 1575
476
+ },
477
+ {
478
+ "epoch": 1.28,
479
+ "grad_norm": 0.1682538390159607,
480
+ "learning_rate": 0.0003461556159111748,
481
+ "loss": 0.2274,
482
+ "step": 1600
483
+ },
484
+ {
485
+ "epoch": 1.3,
486
+ "grad_norm": 0.17813384532928467,
487
+ "learning_rate": 0.0003296461240280242,
488
+ "loss": 0.2095,
489
+ "step": 1625
490
+ },
491
+ {
492
+ "epoch": 1.32,
493
+ "grad_norm": 0.17569252848625183,
494
+ "learning_rate": 0.00031334418210174266,
495
+ "loss": 0.2325,
496
+ "step": 1650
497
+ },
498
+ {
499
+ "epoch": 1.34,
500
+ "grad_norm": 0.14148516952991486,
501
+ "learning_rate": 0.0002972696515374455,
502
+ "loss": 0.2152,
503
+ "step": 1675
504
+ },
505
+ {
506
+ "epoch": 1.3599999999999999,
507
+ "grad_norm": 0.14034417271614075,
508
+ "learning_rate": 0.00028144211667453366,
509
+ "loss": 0.2189,
510
+ "step": 1700
511
+ },
512
+ {
513
+ "epoch": 1.38,
514
+ "grad_norm": 0.16050834953784943,
515
+ "learning_rate": 0.00026588086092619277,
516
+ "loss": 0.2252,
517
+ "step": 1725
518
+ },
519
+ {
520
+ "epoch": 1.4,
521
+ "grad_norm": 0.1896321177482605,
522
+ "learning_rate": 0.00025060484328552466,
523
+ "loss": 0.2219,
524
+ "step": 1750
525
+ },
526
+ {
527
+ "epoch": 1.42,
528
+ "grad_norm": 0.16782040894031525,
529
+ "learning_rate": 0.00023563267522693415,
530
+ "loss": 0.2139,
531
+ "step": 1775
532
+ },
533
+ {
534
+ "epoch": 1.44,
535
+ "grad_norm": 0.1926344931125641,
536
+ "learning_rate": 0.0002209825980309151,
537
+ "loss": 0.2211,
538
+ "step": 1800
539
+ },
540
+ {
541
+ "epoch": 1.46,
542
+ "grad_norm": 0.16422989964485168,
543
+ "learning_rate": 0.00020667246055985938,
544
+ "loss": 0.2215,
545
+ "step": 1825
546
+ },
547
+ {
548
+ "epoch": 1.48,
549
+ "grad_norm": 0.16321636736392975,
550
+ "learning_rate": 0.00019271969751196778,
551
+ "loss": 0.2066,
552
+ "step": 1850
553
+ },
554
+ {
555
+ "epoch": 1.5,
556
+ "grad_norm": 0.15386155247688293,
557
+ "learning_rate": 0.00017914130817975592,
558
+ "loss": 0.2179,
559
+ "step": 1875
560
+ },
561
+ {
562
+ "epoch": 1.52,
563
+ "grad_norm": 0.1478998363018036,
564
+ "learning_rate": 0.00016595383573903412,
565
+ "loss": 0.2117,
566
+ "step": 1900
567
+ },
568
+ {
569
+ "epoch": 1.54,
570
+ "grad_norm": 0.16174481809139252,
571
+ "learning_rate": 0.0001531733470935976,
572
+ "loss": 0.2149,
573
+ "step": 1925
574
+ },
575
+ {
576
+ "epoch": 1.56,
577
+ "grad_norm": 0.13468210399150848,
578
+ "learning_rate": 0.00014081541330017704,
579
+ "loss": 0.2037,
580
+ "step": 1950
581
+ },
582
+ {
583
+ "epoch": 1.58,
584
+ "grad_norm": 0.1549725979566574,
585
+ "learning_rate": 0.00012889509059750602,
586
+ "loss": 0.2096,
587
+ "step": 1975
588
+ },
589
+ {
590
+ "epoch": 1.6,
591
+ "grad_norm": 0.15235114097595215,
592
+ "learning_rate": 0.00011742690206261292,
593
+ "loss": 0.215,
594
+ "step": 2000
595
+ },
596
+ {
597
+ "epoch": 1.6,
598
+ "eval_loss": 0.22170303761959076,
599
+ "eval_runtime": 18.729,
600
+ "eval_samples_per_second": 54.621,
601
+ "eval_steps_per_second": 0.854,
602
+ "step": 2000
603
+ },
604
+ {
605
+ "epoch": 1.62,
606
+ "grad_norm": 0.16818706691265106,
607
+ "learning_rate": 0.0001064248199166884,
608
+ "loss": 0.2024,
609
+ "step": 2025
610
+ },
611
+ {
612
+ "epoch": 1.6400000000000001,
613
+ "grad_norm": 0.1615205854177475,
614
+ "learning_rate": 9.590224850208645e-05,
615
+ "loss": 0.2116,
616
+ "step": 2050
617
+ },
618
+ {
619
+ "epoch": 1.6600000000000001,
620
+ "grad_norm": 0.15136006474494934,
621
+ "learning_rate": 8.587200795119792e-05,
622
+ "loss": 0.2159,
623
+ "step": 2075
624
+ },
625
+ {
626
+ "epoch": 1.6800000000000002,
627
+ "grad_norm": 0.17629875242710114,
628
+ "learning_rate": 7.634631856709389e-05,
629
+ "loss": 0.2173,
630
+ "step": 2100
631
+ },
632
+ {
633
+ "epoch": 1.7,
634
+ "grad_norm": 0.17757442593574524,
635
+ "learning_rate": 6.733678593496901e-05,
636
+ "loss": 0.2165,
637
+ "step": 2125
638
+ },
639
+ {
640
+ "epoch": 1.72,
641
+ "grad_norm": 0.17096401751041412,
642
+ "learning_rate": 5.885438678252342e-05,
643
+ "loss": 0.2104,
644
+ "step": 2150
645
+ },
646
+ {
647
+ "epoch": 1.74,
648
+ "grad_norm": 0.1552375704050064,
649
+ "learning_rate": 5.0909455606510726e-05,
650
+ "loss": 0.2065,
651
+ "step": 2175
652
+ },
653
+ {
654
+ "epoch": 1.76,
655
+ "grad_norm": 0.2247905433177948,
656
+ "learning_rate": 4.3511672081746386e-05,
657
+ "loss": 0.211,
658
+ "step": 2200
659
+ },
660
+ {
661
+ "epoch": 1.78,
662
+ "grad_norm": 0.15804603695869446,
663
+ "learning_rate": 3.667004926791395e-05,
664
+ "loss": 0.2135,
665
+ "step": 2225
666
+ },
667
+ {
668
+ "epoch": 1.8,
669
+ "grad_norm": 0.15289010107517242,
670
+ "learning_rate": 3.0392922628540875e-05,
671
+ "loss": 0.2079,
672
+ "step": 2250
673
+ },
674
+ {
675
+ "epoch": 1.8199999999999998,
676
+ "grad_norm": 0.148426815867424,
677
+ "learning_rate": 2.468793987551998e-05,
678
+ "loss": 0.2048,
679
+ "step": 2275
680
+ },
681
+ {
682
+ "epoch": 1.8399999999999999,
683
+ "grad_norm": 0.1801871359348297,
684
+ "learning_rate": 1.9562051651550784e-05,
685
+ "loss": 0.1998,
686
+ "step": 2300
687
+ },
688
+ {
689
+ "epoch": 1.8599999999999999,
690
+ "grad_norm": 0.15381543338298798,
691
+ "learning_rate": 1.5021503061851349e-05,
692
+ "loss": 0.2127,
693
+ "step": 2325
694
+ },
695
+ {
696
+ "epoch": 1.88,
697
+ "grad_norm": 0.1602175533771515,
698
+ "learning_rate": 1.1071826065460589e-05,
699
+ "loss": 0.1991,
700
+ "step": 2350
701
+ },
702
+ {
703
+ "epoch": 1.9,
704
+ "grad_norm": 0.17019180953502655,
705
+ "learning_rate": 7.717832735397334e-06,
706
+ "loss": 0.2036,
707
+ "step": 2375
708
+ },
709
+ {
710
+ "epoch": 1.92,
711
+ "grad_norm": 0.19569824635982513,
712
+ "learning_rate": 4.963609395891299e-06,
713
+ "loss": 0.2034,
714
+ "step": 2400
715
+ },
716
+ {
717
+ "epoch": 1.94,
718
+ "grad_norm": 0.15439940989017487,
719
+ "learning_rate": 2.81251164382601e-06,
720
+ "loss": 0.2063,
721
+ "step": 2425
722
+ },
723
+ {
724
+ "epoch": 1.96,
725
+ "grad_norm": 0.1678876131772995,
726
+ "learning_rate": 1.267160260461253e-06,
727
+ "loss": 0.1939,
728
+ "step": 2450
729
+ },
730
+ {
731
+ "epoch": 1.98,
732
+ "grad_norm": 0.14893606305122375,
733
+ "learning_rate": 3.2943801841439634e-07,
734
+ "loss": 0.203,
735
+ "step": 2475
736
+ },
737
+ {
738
+ "epoch": 2.0,
739
+ "grad_norm": 0.13442523777484894,
740
+ "learning_rate": 4.873877924582715e-10,
741
+ "loss": 0.2036,
742
+ "step": 2500
743
+ },
744
+ {
745
+ "epoch": 2.0,
746
+ "eval_loss": 0.21816210448741913,
747
+ "eval_runtime": 18.7265,
748
+ "eval_samples_per_second": 54.629,
749
+ "eval_steps_per_second": 0.854,
750
+ "step": 2500
751
+ },
752
+ {
753
+ "epoch": 2.0,
754
+ "step": 2500,
755
+ "total_flos": 1.62588235137024e+18,
756
+ "train_loss": 0.26149349727630616,
757
+ "train_runtime": 2422.0842,
758
+ "train_samples_per_second": 33.029,
759
+ "train_steps_per_second": 1.032
760
+ }
761
+ ],
762
+ "logging_steps": 25,
763
+ "max_steps": 2500,
764
+ "num_input_tokens_seen": 0,
765
+ "num_train_epochs": 2,
766
+ "save_steps": 0,
767
+ "stateful_callbacks": {
768
+ "TrainerControl": {
769
+ "args": {
770
+ "should_epoch_stop": false,
771
+ "should_evaluate": false,
772
+ "should_log": false,
773
+ "should_save": true,
774
+ "should_training_stop": true
775
+ },
776
+ "attributes": {}
777
+ }
778
+ },
779
+ "total_flos": 1.62588235137024e+18,
780
+ "train_batch_size": 32,
781
+ "trial_name": null,
782
+ "trial_params": null
783
+ }