Muqeeth commited on
Commit
b472ffb
·
verified ·
1 Parent(s): 756ec40

Upload files

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. checkpoint-104/config.json +31 -0
  2. checkpoint-104/generation_config.json +6 -0
  3. checkpoint-104/model.safetensors +3 -0
  4. checkpoint-104/optimizer.pt +3 -0
  5. checkpoint-104/rng_state.pth +3 -0
  6. checkpoint-104/scheduler.pt +3 -0
  7. checkpoint-104/special_tokens_map.json +30 -0
  8. checkpoint-104/tokenizer.json +0 -0
  9. checkpoint-104/tokenizer_config.json +220 -0
  10. checkpoint-104/trainer_state.json +2219 -0
  11. checkpoint-104/training_args.bin +3 -0
  12. checkpoint-156/config.json +31 -0
  13. checkpoint-156/generation_config.json +6 -0
  14. checkpoint-156/model.safetensors +3 -0
  15. checkpoint-156/optimizer.pt +3 -0
  16. checkpoint-156/rng_state.pth +3 -0
  17. checkpoint-156/scheduler.pt +3 -0
  18. checkpoint-156/special_tokens_map.json +30 -0
  19. checkpoint-156/tokenizer.json +0 -0
  20. checkpoint-156/tokenizer_config.json +220 -0
  21. checkpoint-156/trainer_state.json +0 -0
  22. checkpoint-156/training_args.bin +3 -0
  23. checkpoint-208/config.json +31 -0
  24. checkpoint-208/generation_config.json +6 -0
  25. checkpoint-208/model.safetensors +3 -0
  26. checkpoint-208/optimizer.pt +3 -0
  27. checkpoint-208/rng_state.pth +3 -0
  28. checkpoint-208/scheduler.pt +3 -0
  29. checkpoint-208/special_tokens_map.json +30 -0
  30. checkpoint-208/tokenizer.json +0 -0
  31. checkpoint-208/tokenizer_config.json +220 -0
  32. checkpoint-208/trainer_state.json +0 -0
  33. checkpoint-208/training_args.bin +3 -0
  34. checkpoint-256/config.json +31 -0
  35. checkpoint-256/generation_config.json +6 -0
  36. checkpoint-256/model.safetensors +3 -0
  37. checkpoint-256/optimizer.pt +3 -0
  38. checkpoint-256/rng_state.pth +3 -0
  39. checkpoint-256/scheduler.pt +3 -0
  40. checkpoint-256/special_tokens_map.json +30 -0
  41. checkpoint-256/tokenizer.json +0 -0
  42. checkpoint-256/tokenizer_config.json +220 -0
  43. checkpoint-256/trainer_state.json +0 -0
  44. checkpoint-256/training_args.bin +3 -0
  45. checkpoint-52/config.json +31 -0
  46. checkpoint-52/generation_config.json +6 -0
  47. checkpoint-52/model.safetensors +3 -0
  48. checkpoint-52/optimizer.pt +3 -0
  49. checkpoint-52/rng_state.pth +3 -0
  50. checkpoint-52/scheduler.pt +3 -0
checkpoint-104/config.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "GPTNeoXForCausalLM"
4
+ ],
5
+ "attention_bias": true,
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 0,
8
+ "classifier_dropout": 0.1,
9
+ "eos_token_id": 0,
10
+ "hidden_act": "gelu",
11
+ "hidden_dropout": 0.0,
12
+ "hidden_size": 1024,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 4096,
15
+ "layer_norm_eps": 1e-05,
16
+ "max_position_embeddings": 2048,
17
+ "model_type": "gpt_neox",
18
+ "num_attention_heads": 16,
19
+ "num_hidden_layers": 24,
20
+ "partial_rotary_factor": 0.25,
21
+ "rope_scaling": null,
22
+ "rope_theta": 10000,
23
+ "rotary_emb_base": 10000,
24
+ "rotary_pct": 0.25,
25
+ "tie_word_embeddings": false,
26
+ "torch_dtype": "float32",
27
+ "transformers_version": "4.50.3",
28
+ "use_cache": true,
29
+ "use_parallel_residual": true,
30
+ "vocab_size": 50304
31
+ }
checkpoint-104/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 0,
4
+ "eos_token_id": 0,
5
+ "transformers_version": "4.50.3"
6
+ }
checkpoint-104/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78f77069daa4fa582da971016f8687d142f8f55f574d0c2e1f1c2adf49fc48ee
3
+ size 1621370224
checkpoint-104/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ae8a8a6d62d6f48b04c609511b0d4859c714b5bb80ba27390013590cedfcbd95
3
+ size 6073758958
checkpoint-104/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e52511727eee9d2c7c0e7e3020960ecaa28db92355b722b98ce455ebc54ded65
3
+ size 14308
checkpoint-104/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:460b88bee87df6269083a417c3dc33b510203846d08a30e05e6a13825374d6b6
3
+ size 1064
checkpoint-104/special_tokens_map.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<|endoftext|>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "<|endoftext|>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": {
17
+ "content": "<|padding|>",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ },
23
+ "unk_token": {
24
+ "content": "<|endoftext|>",
25
+ "lstrip": false,
26
+ "normalized": false,
27
+ "rstrip": false,
28
+ "single_word": false
29
+ }
30
+ }
checkpoint-104/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-104/tokenizer_config.json ADDED
@@ -0,0 +1,220 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_eos_token": false,
4
+ "add_prefix_space": false,
5
+ "added_tokens_decoder": {
6
+ "0": {
7
+ "content": "<|endoftext|>",
8
+ "lstrip": false,
9
+ "normalized": false,
10
+ "rstrip": false,
11
+ "single_word": false,
12
+ "special": true
13
+ },
14
+ "1": {
15
+ "content": "<|padding|>",
16
+ "lstrip": false,
17
+ "normalized": false,
18
+ "rstrip": false,
19
+ "single_word": false,
20
+ "special": true
21
+ },
22
+ "50254": {
23
+ "content": " ",
24
+ "lstrip": false,
25
+ "normalized": true,
26
+ "rstrip": false,
27
+ "single_word": false,
28
+ "special": false
29
+ },
30
+ "50255": {
31
+ "content": " ",
32
+ "lstrip": false,
33
+ "normalized": true,
34
+ "rstrip": false,
35
+ "single_word": false,
36
+ "special": false
37
+ },
38
+ "50256": {
39
+ "content": " ",
40
+ "lstrip": false,
41
+ "normalized": true,
42
+ "rstrip": false,
43
+ "single_word": false,
44
+ "special": false
45
+ },
46
+ "50257": {
47
+ "content": " ",
48
+ "lstrip": false,
49
+ "normalized": true,
50
+ "rstrip": false,
51
+ "single_word": false,
52
+ "special": false
53
+ },
54
+ "50258": {
55
+ "content": " ",
56
+ "lstrip": false,
57
+ "normalized": true,
58
+ "rstrip": false,
59
+ "single_word": false,
60
+ "special": false
61
+ },
62
+ "50259": {
63
+ "content": " ",
64
+ "lstrip": false,
65
+ "normalized": true,
66
+ "rstrip": false,
67
+ "single_word": false,
68
+ "special": false
69
+ },
70
+ "50260": {
71
+ "content": " ",
72
+ "lstrip": false,
73
+ "normalized": true,
74
+ "rstrip": false,
75
+ "single_word": false,
76
+ "special": false
77
+ },
78
+ "50261": {
79
+ "content": " ",
80
+ "lstrip": false,
81
+ "normalized": true,
82
+ "rstrip": false,
83
+ "single_word": false,
84
+ "special": false
85
+ },
86
+ "50262": {
87
+ "content": " ",
88
+ "lstrip": false,
89
+ "normalized": true,
90
+ "rstrip": false,
91
+ "single_word": false,
92
+ "special": false
93
+ },
94
+ "50263": {
95
+ "content": " ",
96
+ "lstrip": false,
97
+ "normalized": true,
98
+ "rstrip": false,
99
+ "single_word": false,
100
+ "special": false
101
+ },
102
+ "50264": {
103
+ "content": " ",
104
+ "lstrip": false,
105
+ "normalized": true,
106
+ "rstrip": false,
107
+ "single_word": false,
108
+ "special": false
109
+ },
110
+ "50265": {
111
+ "content": " ",
112
+ "lstrip": false,
113
+ "normalized": true,
114
+ "rstrip": false,
115
+ "single_word": false,
116
+ "special": false
117
+ },
118
+ "50266": {
119
+ "content": " ",
120
+ "lstrip": false,
121
+ "normalized": true,
122
+ "rstrip": false,
123
+ "single_word": false,
124
+ "special": false
125
+ },
126
+ "50267": {
127
+ "content": " ",
128
+ "lstrip": false,
129
+ "normalized": true,
130
+ "rstrip": false,
131
+ "single_word": false,
132
+ "special": false
133
+ },
134
+ "50268": {
135
+ "content": " ",
136
+ "lstrip": false,
137
+ "normalized": true,
138
+ "rstrip": false,
139
+ "single_word": false,
140
+ "special": false
141
+ },
142
+ "50269": {
143
+ "content": " ",
144
+ "lstrip": false,
145
+ "normalized": true,
146
+ "rstrip": false,
147
+ "single_word": false,
148
+ "special": false
149
+ },
150
+ "50270": {
151
+ "content": " ",
152
+ "lstrip": false,
153
+ "normalized": true,
154
+ "rstrip": false,
155
+ "single_word": false,
156
+ "special": false
157
+ },
158
+ "50271": {
159
+ "content": " ",
160
+ "lstrip": false,
161
+ "normalized": true,
162
+ "rstrip": false,
163
+ "single_word": false,
164
+ "special": false
165
+ },
166
+ "50272": {
167
+ "content": " ",
168
+ "lstrip": false,
169
+ "normalized": true,
170
+ "rstrip": false,
171
+ "single_word": false,
172
+ "special": false
173
+ },
174
+ "50273": {
175
+ "content": " ",
176
+ "lstrip": false,
177
+ "normalized": true,
178
+ "rstrip": false,
179
+ "single_word": false,
180
+ "special": false
181
+ },
182
+ "50274": {
183
+ "content": " ",
184
+ "lstrip": false,
185
+ "normalized": true,
186
+ "rstrip": false,
187
+ "single_word": false,
188
+ "special": false
189
+ },
190
+ "50275": {
191
+ "content": " ",
192
+ "lstrip": false,
193
+ "normalized": true,
194
+ "rstrip": false,
195
+ "single_word": false,
196
+ "special": false
197
+ },
198
+ "50276": {
199
+ "content": " ",
200
+ "lstrip": false,
201
+ "normalized": true,
202
+ "rstrip": false,
203
+ "single_word": false,
204
+ "special": false
205
+ }
206
+ },
207
+ "bos_token": "<|endoftext|>",
208
+ "clean_up_tokenization_spaces": true,
209
+ "eos_token": "<|endoftext|>",
210
+ "extra_special_tokens": {},
211
+ "max_length": 580,
212
+ "model_max_length": 1000000000000000019884624838656,
213
+ "pad_token": "<|padding|>",
214
+ "padding_side": "left",
215
+ "stride": 0,
216
+ "tokenizer_class": "GPTNeoXTokenizer",
217
+ "truncation_side": "right",
218
+ "truncation_strategy": "longest_first",
219
+ "unk_token": "<|endoftext|>"
220
+ }
checkpoint-104/trainer_state.json ADDED
@@ -0,0 +1,2219 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "episode": 53248,
6
+ "epoch": 0.45620287868403014,
7
+ "eval_steps": 500,
8
+ "global_step": 104,
9
+ "is_hyper_param_search": false,
10
+ "is_local_process_zero": true,
11
+ "is_world_process_zero": true,
12
+ "log_history": [
13
+ {
14
+ "episode": 512,
15
+ "epoch": 0.004386566141192597,
16
+ "eps": 5,
17
+ "loss/policy_avg": -0.0064637912437319756,
18
+ "loss/value_avg": 2.062100887298584,
19
+ "lr": 3e-06,
20
+ "objective/entropy": -56.19151306152344,
21
+ "objective/kl": 2.391636371612549e-06,
22
+ "objective/non_score_reward": -1.1958181289628556e-07,
23
+ "objective/rlhf_reward": 0.2761589586734772,
24
+ "objective/scores": 0.2761590778827667,
25
+ "policy/approxkl_avg": 0.001525376457720995,
26
+ "policy/clipfrac_avg": 0.02386678196489811,
27
+ "policy/entropy_avg": 1.2976278066635132,
28
+ "step": 1,
29
+ "val/clipfrac_avg": 0.19838498532772064,
30
+ "val/num_eos_tokens": 13845,
31
+ "val/ratio": 0.9997893571853638,
32
+ "val/ratio_var": 3.1534598292637384e-06
33
+ },
34
+ {
35
+ "episode": 1024,
36
+ "epoch": 0.008773132282385195,
37
+ "eps": 6,
38
+ "loss/policy_avg": 0.0009075724519789219,
39
+ "loss/value_avg": 0.9430553317070007,
40
+ "lr": 2.9882812500000002e-06,
41
+ "objective/entropy": 11.115456581115723,
42
+ "objective/kl": 0.711035966873169,
43
+ "objective/non_score_reward": -0.03555179759860039,
44
+ "objective/rlhf_reward": 0.052655573934316635,
45
+ "objective/scores": 0.08820737153291702,
46
+ "policy/approxkl_avg": 0.003618996124714613,
47
+ "policy/clipfrac_avg": 0.027872614562511444,
48
+ "policy/entropy_avg": 0.9095284938812256,
49
+ "step": 2,
50
+ "val/clipfrac_avg": 0.09195894002914429,
51
+ "val/num_eos_tokens": 8329,
52
+ "val/ratio": 0.9998072981834412,
53
+ "val/ratio_var": 2.803345341817476e-05
54
+ },
55
+ {
56
+ "episode": 1536,
57
+ "epoch": 0.013159698423577794,
58
+ "eps": 6,
59
+ "loss/policy_avg": 0.006346262060105801,
60
+ "loss/value_avg": 1.5911083221435547,
61
+ "lr": 2.9765625e-06,
62
+ "objective/entropy": 7.955699443817139,
63
+ "objective/kl": 0.8145447969436646,
64
+ "objective/non_score_reward": -0.040727242827415466,
65
+ "objective/rlhf_reward": 0.27417320013046265,
66
+ "objective/scores": 0.3149004280567169,
67
+ "policy/approxkl_avg": 0.0033809328451752663,
68
+ "policy/clipfrac_avg": 0.027064774185419083,
69
+ "policy/entropy_avg": 0.9011605978012085,
70
+ "step": 3,
71
+ "val/clipfrac_avg": 0.08764837682247162,
72
+ "val/num_eos_tokens": 7341,
73
+ "val/ratio": 0.9998272061347961,
74
+ "val/ratio_var": 1.4393113815458491e-05
75
+ },
76
+ {
77
+ "episode": 2048,
78
+ "epoch": 0.01754626456477039,
79
+ "eps": 6,
80
+ "loss/policy_avg": 0.00562187097966671,
81
+ "loss/value_avg": 1.058276653289795,
82
+ "lr": 2.96484375e-06,
83
+ "objective/entropy": 4.988356590270996,
84
+ "objective/kl": 1.2054615020751953,
85
+ "objective/non_score_reward": -0.060273073613643646,
86
+ "objective/rlhf_reward": 0.5730378031730652,
87
+ "objective/scores": 0.633310854434967,
88
+ "policy/approxkl_avg": 0.0034487333614379168,
89
+ "policy/clipfrac_avg": 0.024757273495197296,
90
+ "policy/entropy_avg": 1.0609033107757568,
91
+ "step": 4,
92
+ "val/clipfrac_avg": 0.07027730345726013,
93
+ "val/num_eos_tokens": 8861,
94
+ "val/ratio": 1.000706434249878,
95
+ "val/ratio_var": 1.794960189727135e-05
96
+ },
97
+ {
98
+ "episode": 2560,
99
+ "epoch": 0.02193283070596299,
100
+ "eps": 6,
101
+ "loss/policy_avg": 0.00019519031047821045,
102
+ "loss/value_avg": 0.5995841026306152,
103
+ "lr": 2.953125e-06,
104
+ "objective/entropy": -47.6590690612793,
105
+ "objective/kl": 3.0212581157684326,
106
+ "objective/non_score_reward": -0.15106290578842163,
107
+ "objective/rlhf_reward": 0.886078417301178,
108
+ "objective/scores": 1.0371413230895996,
109
+ "policy/approxkl_avg": 0.0020040306262671947,
110
+ "policy/clipfrac_avg": 0.02877683937549591,
111
+ "policy/entropy_avg": 1.821481466293335,
112
+ "step": 5,
113
+ "val/clipfrac_avg": 0.06764362007379532,
114
+ "val/num_eos_tokens": 16649,
115
+ "val/ratio": 1.0005488395690918,
116
+ "val/ratio_var": 4.007795268989867e-06
117
+ },
118
+ {
119
+ "episode": 3072,
120
+ "epoch": 0.026319396847155587,
121
+ "eps": 6,
122
+ "loss/policy_avg": -0.0008811755105853081,
123
+ "loss/value_avg": 0.5564485192298889,
124
+ "lr": 2.94140625e-06,
125
+ "objective/entropy": -7.285348892211914,
126
+ "objective/kl": 7.259183406829834,
127
+ "objective/non_score_reward": -0.3629591464996338,
128
+ "objective/rlhf_reward": 1.0025866031646729,
129
+ "objective/scores": 1.3655457496643066,
130
+ "policy/approxkl_avg": 0.0025039517786353827,
131
+ "policy/clipfrac_avg": 0.02588822692632675,
132
+ "policy/entropy_avg": 1.90171480178833,
133
+ "step": 6,
134
+ "val/clipfrac_avg": 0.05784378945827484,
135
+ "val/num_eos_tokens": 14701,
136
+ "val/ratio": 0.9996325969696045,
137
+ "val/ratio_var": 4.991092737327563e-06
138
+ },
139
+ {
140
+ "episode": 3584,
141
+ "epoch": 0.030705962988348184,
142
+ "eps": 6,
143
+ "loss/policy_avg": -0.002672974020242691,
144
+ "loss/value_avg": 0.6466186046600342,
145
+ "lr": 2.9296875e-06,
146
+ "objective/entropy": -19.351350784301758,
147
+ "objective/kl": 13.451505661010742,
148
+ "objective/non_score_reward": -0.6725753545761108,
149
+ "objective/rlhf_reward": 1.2054418325424194,
150
+ "objective/scores": 1.8780171871185303,
151
+ "policy/approxkl_avg": 0.0032266275957226753,
152
+ "policy/clipfrac_avg": 0.029281172901391983,
153
+ "policy/entropy_avg": 2.318066358566284,
154
+ "step": 7,
155
+ "val/clipfrac_avg": 0.043010011315345764,
156
+ "val/num_eos_tokens": 23664,
157
+ "val/ratio": 1.0000731945037842,
158
+ "val/ratio_var": 4.646569323085714e-06
159
+ },
160
+ {
161
+ "episode": 4096,
162
+ "epoch": 0.03509252912954078,
163
+ "eps": 5,
164
+ "loss/policy_avg": -0.004132507368922234,
165
+ "loss/value_avg": 0.7284023761749268,
166
+ "lr": 2.91796875e-06,
167
+ "objective/entropy": -18.51163101196289,
168
+ "objective/kl": 16.813966751098633,
169
+ "objective/non_score_reward": -0.8406983613967896,
170
+ "objective/rlhf_reward": 1.5278810262680054,
171
+ "objective/scores": 2.368579387664795,
172
+ "policy/approxkl_avg": 0.002941709943115711,
173
+ "policy/clipfrac_avg": 0.03002079948782921,
174
+ "policy/entropy_avg": 2.2803053855895996,
175
+ "step": 8,
176
+ "val/clipfrac_avg": 0.0415257066488266,
177
+ "val/num_eos_tokens": 29999,
178
+ "val/ratio": 0.9997649192810059,
179
+ "val/ratio_var": 3.7887116377532948e-06
180
+ },
181
+ {
182
+ "episode": 4608,
183
+ "epoch": 0.03947909527073338,
184
+ "eps": 5,
185
+ "loss/policy_avg": -0.00987590104341507,
186
+ "loss/value_avg": 0.6991376876831055,
187
+ "lr": 2.90625e-06,
188
+ "objective/entropy": -28.510631561279297,
189
+ "objective/kl": 16.568471908569336,
190
+ "objective/non_score_reward": -0.8284236192703247,
191
+ "objective/rlhf_reward": 1.9216898679733276,
192
+ "objective/scores": 2.7501134872436523,
193
+ "policy/approxkl_avg": 0.0029653366655111313,
194
+ "policy/clipfrac_avg": 0.034570105373859406,
195
+ "policy/entropy_avg": 2.3752803802490234,
196
+ "step": 9,
197
+ "val/clipfrac_avg": 0.03531229496002197,
198
+ "val/num_eos_tokens": 25112,
199
+ "val/ratio": 0.9997652769088745,
200
+ "val/ratio_var": 5.58018882657052e-06
201
+ },
202
+ {
203
+ "episode": 5120,
204
+ "epoch": 0.04386566141192598,
205
+ "eps": 5,
206
+ "loss/policy_avg": -0.01345974113792181,
207
+ "loss/value_avg": 0.5584598779678345,
208
+ "lr": 2.89453125e-06,
209
+ "objective/entropy": -38.812477111816406,
210
+ "objective/kl": 15.714573860168457,
211
+ "objective/non_score_reward": -0.7857286930084229,
212
+ "objective/rlhf_reward": 1.9641573429107666,
213
+ "objective/scores": 2.7498860359191895,
214
+ "policy/approxkl_avg": 0.0030483785085380077,
215
+ "policy/clipfrac_avg": 0.03399527072906494,
216
+ "policy/entropy_avg": 2.4547998905181885,
217
+ "step": 10,
218
+ "val/clipfrac_avg": 0.04387196898460388,
219
+ "val/num_eos_tokens": 26266,
220
+ "val/ratio": 0.9999704360961914,
221
+ "val/ratio_var": 4.795186669070972e-06
222
+ },
223
+ {
224
+ "episode": 5632,
225
+ "epoch": 0.048252227553118573,
226
+ "eps": 5,
227
+ "loss/policy_avg": -0.016908088698983192,
228
+ "loss/value_avg": 0.5334082841873169,
229
+ "lr": 2.8828125e-06,
230
+ "objective/entropy": -43.84290313720703,
231
+ "objective/kl": 15.399422645568848,
232
+ "objective/non_score_reward": -0.7699711322784424,
233
+ "objective/rlhf_reward": 1.745584487915039,
234
+ "objective/scores": 2.5155556201934814,
235
+ "policy/approxkl_avg": 0.003285747254267335,
236
+ "policy/clipfrac_avg": 0.031903285533189774,
237
+ "policy/entropy_avg": 2.4688422679901123,
238
+ "step": 11,
239
+ "val/clipfrac_avg": 0.04632589966058731,
240
+ "val/num_eos_tokens": 32294,
241
+ "val/ratio": 1.0003118515014648,
242
+ "val/ratio_var": 1.00084889709251e-05
243
+ },
244
+ {
245
+ "episode": 6144,
246
+ "epoch": 0.052638793694311174,
247
+ "eps": 5,
248
+ "loss/policy_avg": -0.03192070871591568,
249
+ "loss/value_avg": 0.6824725866317749,
250
+ "lr": 2.87109375e-06,
251
+ "objective/entropy": -45.36285400390625,
252
+ "objective/kl": 15.74565315246582,
253
+ "objective/non_score_reward": -0.7872827053070068,
254
+ "objective/rlhf_reward": 1.465782880783081,
255
+ "objective/scores": 2.253065586090088,
256
+ "policy/approxkl_avg": 0.002630905481055379,
257
+ "policy/clipfrac_avg": 0.03375859186053276,
258
+ "policy/entropy_avg": 2.431551933288574,
259
+ "step": 12,
260
+ "val/clipfrac_avg": 0.04146707057952881,
261
+ "val/num_eos_tokens": 40074,
262
+ "val/ratio": 0.9995942115783691,
263
+ "val/ratio_var": 4.198932401777711e-06
264
+ },
265
+ {
266
+ "episode": 6656,
267
+ "epoch": 0.05702535983550377,
268
+ "eps": 5,
269
+ "loss/policy_avg": -0.016745194792747498,
270
+ "loss/value_avg": 0.6150881052017212,
271
+ "lr": 2.859375e-06,
272
+ "objective/entropy": -50.91920471191406,
273
+ "objective/kl": 15.028314590454102,
274
+ "objective/non_score_reward": -0.7514157891273499,
275
+ "objective/rlhf_reward": 1.7634849548339844,
276
+ "objective/scores": 2.5149006843566895,
277
+ "policy/approxkl_avg": 0.002676167292520404,
278
+ "policy/clipfrac_avg": 0.033921413123607635,
279
+ "policy/entropy_avg": 2.5404675006866455,
280
+ "step": 13,
281
+ "val/clipfrac_avg": 0.04346344619989395,
282
+ "val/num_eos_tokens": 35364,
283
+ "val/ratio": 1.0000118017196655,
284
+ "val/ratio_var": 7.1822228164819535e-06
285
+ },
286
+ {
287
+ "episode": 7168,
288
+ "epoch": 0.06141192597669637,
289
+ "eps": 5,
290
+ "loss/policy_avg": -0.01955597847700119,
291
+ "loss/value_avg": 0.6850756406784058,
292
+ "lr": 2.84765625e-06,
293
+ "objective/entropy": -50.385826110839844,
294
+ "objective/kl": 15.855907440185547,
295
+ "objective/non_score_reward": -0.7927954196929932,
296
+ "objective/rlhf_reward": 1.653841257095337,
297
+ "objective/scores": 2.44663667678833,
298
+ "policy/approxkl_avg": 0.0030123821925371885,
299
+ "policy/clipfrac_avg": 0.03439797833561897,
300
+ "policy/entropy_avg": 2.517819404602051,
301
+ "step": 14,
302
+ "val/clipfrac_avg": 0.04387962073087692,
303
+ "val/num_eos_tokens": 37451,
304
+ "val/ratio": 0.9994909167289734,
305
+ "val/ratio_var": 5.496693120221607e-06
306
+ },
307
+ {
308
+ "episode": 7680,
309
+ "epoch": 0.06579849211788896,
310
+ "eps": 5,
311
+ "loss/policy_avg": -0.006825360003858805,
312
+ "loss/value_avg": 0.5871850252151489,
313
+ "lr": 2.8359375e-06,
314
+ "objective/entropy": -53.392860412597656,
315
+ "objective/kl": 15.780671119689941,
316
+ "objective/non_score_reward": -0.7890335321426392,
317
+ "objective/rlhf_reward": 1.9988459348678589,
318
+ "objective/scores": 2.787879467010498,
319
+ "policy/approxkl_avg": 0.0029939599335193634,
320
+ "policy/clipfrac_avg": 0.035068489611148834,
321
+ "policy/entropy_avg": 2.510554075241089,
322
+ "step": 15,
323
+ "val/clipfrac_avg": 0.03168816119432449,
324
+ "val/num_eos_tokens": 27865,
325
+ "val/ratio": 0.9999009370803833,
326
+ "val/ratio_var": 6.548730652866652e-06
327
+ },
328
+ {
329
+ "episode": 8192,
330
+ "epoch": 0.07018505825908156,
331
+ "eps": 5,
332
+ "loss/policy_avg": -0.0055807908065617085,
333
+ "loss/value_avg": 0.4809805154800415,
334
+ "lr": 2.82421875e-06,
335
+ "objective/entropy": -53.630775451660156,
336
+ "objective/kl": 16.142826080322266,
337
+ "objective/non_score_reward": -0.8071413040161133,
338
+ "objective/rlhf_reward": 2.224621295928955,
339
+ "objective/scores": 3.0317625999450684,
340
+ "policy/approxkl_avg": 0.0036894569639116526,
341
+ "policy/clipfrac_avg": 0.034702934324741364,
342
+ "policy/entropy_avg": 2.5526700019836426,
343
+ "step": 16,
344
+ "val/clipfrac_avg": 0.03549609333276749,
345
+ "val/num_eos_tokens": 28237,
346
+ "val/ratio": 0.999828577041626,
347
+ "val/ratio_var": 4.990592060494237e-06
348
+ },
349
+ {
350
+ "episode": 8704,
351
+ "epoch": 0.07457162440027416,
352
+ "eps": 5,
353
+ "loss/policy_avg": -0.011602696031332016,
354
+ "loss/value_avg": 0.42486828565597534,
355
+ "lr": 2.8125e-06,
356
+ "objective/entropy": -49.62677764892578,
357
+ "objective/kl": 17.828514099121094,
358
+ "objective/non_score_reward": -0.8914257287979126,
359
+ "objective/rlhf_reward": 2.028501033782959,
360
+ "objective/scores": 2.919926881790161,
361
+ "policy/approxkl_avg": 0.0037828730419278145,
362
+ "policy/clipfrac_avg": 0.03432994708418846,
363
+ "policy/entropy_avg": 2.481055498123169,
364
+ "step": 17,
365
+ "val/clipfrac_avg": 0.030638843774795532,
366
+ "val/num_eos_tokens": 26426,
367
+ "val/ratio": 0.9998823404312134,
368
+ "val/ratio_var": 4.263031314621912e-06
369
+ },
370
+ {
371
+ "episode": 9216,
372
+ "epoch": 0.07895819054146676,
373
+ "eps": 5,
374
+ "loss/policy_avg": -0.0081618782132864,
375
+ "loss/value_avg": 0.3735288381576538,
376
+ "lr": 2.80078125e-06,
377
+ "objective/entropy": -45.896141052246094,
378
+ "objective/kl": 19.944303512573242,
379
+ "objective/non_score_reward": -0.9972151517868042,
380
+ "objective/rlhf_reward": 2.1970882415771484,
381
+ "objective/scores": 3.194303512573242,
382
+ "policy/approxkl_avg": 0.0032740728929638863,
383
+ "policy/clipfrac_avg": 0.03566484898328781,
384
+ "policy/entropy_avg": 2.453451156616211,
385
+ "step": 18,
386
+ "val/clipfrac_avg": 0.031092893332242966,
387
+ "val/num_eos_tokens": 26621,
388
+ "val/ratio": 0.9999486207962036,
389
+ "val/ratio_var": 6.638757895416347e-06
390
+ },
391
+ {
392
+ "episode": 9728,
393
+ "epoch": 0.08334475668265935,
394
+ "eps": 5,
395
+ "loss/policy_avg": -0.0033360933884978294,
396
+ "loss/value_avg": 0.3661920130252838,
397
+ "lr": 2.7890625e-06,
398
+ "objective/entropy": -47.260589599609375,
399
+ "objective/kl": 20.071331024169922,
400
+ "objective/non_score_reward": -1.0035665035247803,
401
+ "objective/rlhf_reward": 2.2624642848968506,
402
+ "objective/scores": 3.266030788421631,
403
+ "policy/approxkl_avg": 0.002959758508950472,
404
+ "policy/clipfrac_avg": 0.03424590453505516,
405
+ "policy/entropy_avg": 2.4731497764587402,
406
+ "step": 19,
407
+ "val/clipfrac_avg": 0.02276640385389328,
408
+ "val/num_eos_tokens": 22059,
409
+ "val/ratio": 1.0003962516784668,
410
+ "val/ratio_var": 5.304964815877611e-06
411
+ },
412
+ {
413
+ "episode": 10240,
414
+ "epoch": 0.08773132282385196,
415
+ "eps": 5,
416
+ "loss/policy_avg": -0.0007621180266141891,
417
+ "loss/value_avg": 0.3776611387729645,
418
+ "lr": 2.77734375e-06,
419
+ "objective/entropy": -44.47180938720703,
420
+ "objective/kl": 21.4638614654541,
421
+ "objective/non_score_reward": -1.073193073272705,
422
+ "objective/rlhf_reward": 2.3129146099090576,
423
+ "objective/scores": 3.3861076831817627,
424
+ "policy/approxkl_avg": 0.003540318924933672,
425
+ "policy/clipfrac_avg": 0.03623297065496445,
426
+ "policy/entropy_avg": 2.428302764892578,
427
+ "step": 20,
428
+ "val/clipfrac_avg": 0.018316645175218582,
429
+ "val/num_eos_tokens": 24058,
430
+ "val/ratio": 1.0002745389938354,
431
+ "val/ratio_var": 4.784485554409912e-06
432
+ },
433
+ {
434
+ "episode": 10752,
435
+ "epoch": 0.09211788896504455,
436
+ "eps": 5,
437
+ "loss/policy_avg": -0.006207556463778019,
438
+ "loss/value_avg": 0.3847636580467224,
439
+ "lr": 2.765625e-06,
440
+ "objective/entropy": -44.75438690185547,
441
+ "objective/kl": 22.54155731201172,
442
+ "objective/non_score_reward": -1.1270779371261597,
443
+ "objective/rlhf_reward": 2.391925811767578,
444
+ "objective/scores": 3.5190038681030273,
445
+ "policy/approxkl_avg": 0.0036967298947274685,
446
+ "policy/clipfrac_avg": 0.0338059663772583,
447
+ "policy/entropy_avg": 2.4630565643310547,
448
+ "step": 21,
449
+ "val/clipfrac_avg": 0.019349252805113792,
450
+ "val/num_eos_tokens": 18028,
451
+ "val/ratio": 1.0002729892730713,
452
+ "val/ratio_var": 8.466742656310089e-06
453
+ },
454
+ {
455
+ "episode": 11264,
456
+ "epoch": 0.09650445510623715,
457
+ "eps": 5,
458
+ "loss/policy_avg": -0.003522126004099846,
459
+ "loss/value_avg": 0.37042468786239624,
460
+ "lr": 2.75390625e-06,
461
+ "objective/entropy": -44.410377502441406,
462
+ "objective/kl": 21.86150360107422,
463
+ "objective/non_score_reward": -1.0930752754211426,
464
+ "objective/rlhf_reward": 2.342880964279175,
465
+ "objective/scores": 3.4359562397003174,
466
+ "policy/approxkl_avg": 0.0028353030793368816,
467
+ "policy/clipfrac_avg": 0.0346149206161499,
468
+ "policy/entropy_avg": 2.4011077880859375,
469
+ "step": 22,
470
+ "val/clipfrac_avg": 0.01756729930639267,
471
+ "val/num_eos_tokens": 21896,
472
+ "val/ratio": 1.0002131462097168,
473
+ "val/ratio_var": 5.494702691066777e-06
474
+ },
475
+ {
476
+ "episode": 11776,
477
+ "epoch": 0.10089102124742974,
478
+ "eps": 5,
479
+ "loss/policy_avg": -0.0007715318351984024,
480
+ "loss/value_avg": 0.3489900827407837,
481
+ "lr": 2.7421875e-06,
482
+ "objective/entropy": -43.98833084106445,
483
+ "objective/kl": 22.233755111694336,
484
+ "objective/non_score_reward": -1.1116878986358643,
485
+ "objective/rlhf_reward": 2.2883124351501465,
486
+ "objective/scores": 3.4000003337860107,
487
+ "policy/approxkl_avg": 0.003675229847431183,
488
+ "policy/clipfrac_avg": 0.03448382019996643,
489
+ "policy/entropy_avg": 2.383819103240967,
490
+ "step": 23,
491
+ "val/clipfrac_avg": 0.01657916232943535,
492
+ "val/num_eos_tokens": 22430,
493
+ "val/ratio": 0.9994198679924011,
494
+ "val/ratio_var": 5.044780209573219e-06
495
+ },
496
+ {
497
+ "episode": 12288,
498
+ "epoch": 0.10527758738862235,
499
+ "eps": 5,
500
+ "loss/policy_avg": -0.0016354843974113464,
501
+ "loss/value_avg": 0.3431737422943115,
502
+ "lr": 2.73046875e-06,
503
+ "objective/entropy": -44.860107421875,
504
+ "objective/kl": 22.525632858276367,
505
+ "objective/non_score_reward": -1.1262816190719604,
506
+ "objective/rlhf_reward": 2.484004020690918,
507
+ "objective/scores": 3.610285758972168,
508
+ "policy/approxkl_avg": 0.002885159570723772,
509
+ "policy/clipfrac_avg": 0.03467557206749916,
510
+ "policy/entropy_avg": 2.435291051864624,
511
+ "step": 24,
512
+ "val/clipfrac_avg": 0.017246956005692482,
513
+ "val/num_eos_tokens": 17173,
514
+ "val/ratio": 1.0001412630081177,
515
+ "val/ratio_var": 3.7197830806690035e-06
516
+ },
517
+ {
518
+ "episode": 12800,
519
+ "epoch": 0.10966415352981494,
520
+ "eps": 5,
521
+ "loss/policy_avg": 0.0005915295332670212,
522
+ "loss/value_avg": 0.33912354707717896,
523
+ "lr": 2.71875e-06,
524
+ "objective/entropy": -45.409812927246094,
525
+ "objective/kl": 22.53536605834961,
526
+ "objective/non_score_reward": -1.1267682313919067,
527
+ "objective/rlhf_reward": 2.32200288772583,
528
+ "objective/scores": 3.4487709999084473,
529
+ "policy/approxkl_avg": 0.0032442878000438213,
530
+ "policy/clipfrac_avg": 0.03321682661771774,
531
+ "policy/entropy_avg": 2.3514833450317383,
532
+ "step": 25,
533
+ "val/clipfrac_avg": 0.018273092806339264,
534
+ "val/num_eos_tokens": 22397,
535
+ "val/ratio": 0.9998367428779602,
536
+ "val/ratio_var": 6.378372745530214e-06
537
+ },
538
+ {
539
+ "episode": 13312,
540
+ "epoch": 0.11405071967100754,
541
+ "eps": 5,
542
+ "loss/policy_avg": -0.003818085417151451,
543
+ "loss/value_avg": 0.3161803185939789,
544
+ "lr": 2.70703125e-06,
545
+ "objective/entropy": -46.711700439453125,
546
+ "objective/kl": 22.636919021606445,
547
+ "objective/non_score_reward": -1.1318459510803223,
548
+ "objective/rlhf_reward": 2.1982736587524414,
549
+ "objective/scores": 3.3301196098327637,
550
+ "policy/approxkl_avg": 0.0031049251556396484,
551
+ "policy/clipfrac_avg": 0.035522766411304474,
552
+ "policy/entropy_avg": 2.4093337059020996,
553
+ "step": 26,
554
+ "val/clipfrac_avg": 0.021533971652388573,
555
+ "val/num_eos_tokens": 18959,
556
+ "val/ratio": 1.000077486038208,
557
+ "val/ratio_var": 5.0870248742285185e-06
558
+ },
559
+ {
560
+ "episode": 13824,
561
+ "epoch": 0.11843728581220014,
562
+ "eps": 5,
563
+ "loss/policy_avg": 0.0028609121218323708,
564
+ "loss/value_avg": 0.3064606785774231,
565
+ "lr": 2.6953125e-06,
566
+ "objective/entropy": -18.036178588867188,
567
+ "objective/kl": 22.373611450195312,
568
+ "objective/non_score_reward": -1.118680477142334,
569
+ "objective/rlhf_reward": 2.4143335819244385,
570
+ "objective/scores": 3.5330140590667725,
571
+ "policy/approxkl_avg": 0.004590032622218132,
572
+ "policy/clipfrac_avg": 0.03282628580927849,
573
+ "policy/entropy_avg": 2.060394287109375,
574
+ "step": 27,
575
+ "val/clipfrac_avg": 0.01763969659805298,
576
+ "val/num_eos_tokens": 16526,
577
+ "val/ratio": 1.0000416040420532,
578
+ "val/ratio_var": 1.0717338227550499e-05
579
+ },
580
+ {
581
+ "episode": 14336,
582
+ "epoch": 0.12282385195339274,
583
+ "eps": 5,
584
+ "loss/policy_avg": 0.0013361498713493347,
585
+ "loss/value_avg": 0.3187541961669922,
586
+ "lr": 2.68359375e-06,
587
+ "objective/entropy": -50.27621841430664,
588
+ "objective/kl": 23.557315826416016,
589
+ "objective/non_score_reward": -1.177865743637085,
590
+ "objective/rlhf_reward": 2.409815788269043,
591
+ "objective/scores": 3.587681531906128,
592
+ "policy/approxkl_avg": 0.002932226750999689,
593
+ "policy/clipfrac_avg": 0.03354738652706146,
594
+ "policy/entropy_avg": 2.4124414920806885,
595
+ "step": 28,
596
+ "val/clipfrac_avg": 0.01959545910358429,
597
+ "val/num_eos_tokens": 15394,
598
+ "val/ratio": 0.9994858503341675,
599
+ "val/ratio_var": 3.2119653496920364e-06
600
+ },
601
+ {
602
+ "episode": 14848,
603
+ "epoch": 0.12721041809458533,
604
+ "eps": 5,
605
+ "loss/policy_avg": 0.0033660349436104298,
606
+ "loss/value_avg": 0.2832929491996765,
607
+ "lr": 2.671875e-06,
608
+ "objective/entropy": -51.316322326660156,
609
+ "objective/kl": 22.52779769897461,
610
+ "objective/non_score_reward": -1.1263898611068726,
611
+ "objective/rlhf_reward": 2.447460174560547,
612
+ "objective/scores": 3.573850154876709,
613
+ "policy/approxkl_avg": 0.0027271718718111515,
614
+ "policy/clipfrac_avg": 0.03330669924616814,
615
+ "policy/entropy_avg": 2.3689398765563965,
616
+ "step": 29,
617
+ "val/clipfrac_avg": 0.0231044739484787,
618
+ "val/num_eos_tokens": 19411,
619
+ "val/ratio": 0.9997625350952148,
620
+ "val/ratio_var": 3.876300070260186e-06
621
+ },
622
+ {
623
+ "episode": 15360,
624
+ "epoch": 0.13159698423577793,
625
+ "eps": 5,
626
+ "loss/policy_avg": 0.001225413754582405,
627
+ "loss/value_avg": 0.3369906544685364,
628
+ "lr": 2.66015625e-06,
629
+ "objective/entropy": -50.127891540527344,
630
+ "objective/kl": 22.9091854095459,
631
+ "objective/non_score_reward": -1.1454592943191528,
632
+ "objective/rlhf_reward": 2.26356840133667,
633
+ "objective/scores": 3.409027576446533,
634
+ "policy/approxkl_avg": 0.002824030816555023,
635
+ "policy/clipfrac_avg": 0.03506358712911606,
636
+ "policy/entropy_avg": 2.329383134841919,
637
+ "step": 30,
638
+ "val/clipfrac_avg": 0.02555329166352749,
639
+ "val/num_eos_tokens": 21576,
640
+ "val/ratio": 1.0000300407409668,
641
+ "val/ratio_var": 7.115605512808543e-06
642
+ },
643
+ {
644
+ "episode": 15872,
645
+ "epoch": 0.13598355037697052,
646
+ "eps": 5,
647
+ "loss/policy_avg": -0.003297283314168453,
648
+ "loss/value_avg": 0.3402513265609741,
649
+ "lr": 2.6484375e-06,
650
+ "objective/entropy": -52.112213134765625,
651
+ "objective/kl": 22.56782341003418,
652
+ "objective/non_score_reward": -1.1283912658691406,
653
+ "objective/rlhf_reward": 2.382021427154541,
654
+ "objective/scores": 3.5104126930236816,
655
+ "policy/approxkl_avg": 0.0030852516647428274,
656
+ "policy/clipfrac_avg": 0.03482586890459061,
657
+ "policy/entropy_avg": 2.3571085929870605,
658
+ "step": 31,
659
+ "val/clipfrac_avg": 0.01955670304596424,
660
+ "val/num_eos_tokens": 20115,
661
+ "val/ratio": 1.0000096559524536,
662
+ "val/ratio_var": 5.159253760211868e-06
663
+ },
664
+ {
665
+ "episode": 16384,
666
+ "epoch": 0.1403701165181631,
667
+ "eps": 5,
668
+ "loss/policy_avg": 0.0004618735983967781,
669
+ "loss/value_avg": 0.30273759365081787,
670
+ "lr": 2.63671875e-06,
671
+ "objective/entropy": -52.41477584838867,
672
+ "objective/kl": 21.90145492553711,
673
+ "objective/non_score_reward": -1.0950727462768555,
674
+ "objective/rlhf_reward": 2.2722175121307373,
675
+ "objective/scores": 3.3672902584075928,
676
+ "policy/approxkl_avg": 0.003110222052782774,
677
+ "policy/clipfrac_avg": 0.03544200584292412,
678
+ "policy/entropy_avg": 2.381746768951416,
679
+ "step": 32,
680
+ "val/clipfrac_avg": 0.021619390696287155,
681
+ "val/num_eos_tokens": 16546,
682
+ "val/ratio": 0.9996892213821411,
683
+ "val/ratio_var": 5.161690751265269e-06
684
+ },
685
+ {
686
+ "episode": 16896,
687
+ "epoch": 0.14475668265935573,
688
+ "eps": 5,
689
+ "loss/policy_avg": 0.0029493197798728943,
690
+ "loss/value_avg": 0.2962348461151123,
691
+ "lr": 2.6250000000000003e-06,
692
+ "objective/entropy": -51.622154235839844,
693
+ "objective/kl": 21.047677993774414,
694
+ "objective/non_score_reward": -1.0523838996887207,
695
+ "objective/rlhf_reward": 2.3200583457946777,
696
+ "objective/scores": 3.3724422454833984,
697
+ "policy/approxkl_avg": 0.0023410473950207233,
698
+ "policy/clipfrac_avg": 0.03225007653236389,
699
+ "policy/entropy_avg": 2.3707313537597656,
700
+ "step": 33,
701
+ "val/clipfrac_avg": 0.018346378579735756,
702
+ "val/num_eos_tokens": 14465,
703
+ "val/ratio": 1.0001014471054077,
704
+ "val/ratio_var": 5.0973626457562204e-06
705
+ },
706
+ {
707
+ "episode": 17408,
708
+ "epoch": 0.14914324880054833,
709
+ "eps": 5,
710
+ "loss/policy_avg": -0.0006872201338410378,
711
+ "loss/value_avg": 0.27795565128326416,
712
+ "lr": 2.61328125e-06,
713
+ "objective/entropy": -51.323829650878906,
714
+ "objective/kl": 21.569639205932617,
715
+ "objective/non_score_reward": -1.078481912612915,
716
+ "objective/rlhf_reward": 2.390972375869751,
717
+ "objective/scores": 3.469454288482666,
718
+ "policy/approxkl_avg": 0.0028640958480536938,
719
+ "policy/clipfrac_avg": 0.03302103281021118,
720
+ "policy/entropy_avg": 2.3587939739227295,
721
+ "step": 34,
722
+ "val/clipfrac_avg": 0.025140725076198578,
723
+ "val/num_eos_tokens": 13605,
724
+ "val/ratio": 0.9998920559883118,
725
+ "val/ratio_var": 4.393466497276677e-06
726
+ },
727
+ {
728
+ "episode": 17920,
729
+ "epoch": 0.15352981494174092,
730
+ "eps": 5,
731
+ "loss/policy_avg": 0.0033647562377154827,
732
+ "loss/value_avg": 0.26732805371284485,
733
+ "lr": 2.6015625e-06,
734
+ "objective/entropy": -48.007991790771484,
735
+ "objective/kl": 22.043781280517578,
736
+ "objective/non_score_reward": -1.102189064025879,
737
+ "objective/rlhf_reward": 2.35861873626709,
738
+ "objective/scores": 3.4608078002929688,
739
+ "policy/approxkl_avg": 0.0027386369183659554,
740
+ "policy/clipfrac_avg": 0.03526460379362106,
741
+ "policy/entropy_avg": 2.2924861907958984,
742
+ "step": 35,
743
+ "val/clipfrac_avg": 0.02547530084848404,
744
+ "val/num_eos_tokens": 19570,
745
+ "val/ratio": 0.9999614953994751,
746
+ "val/ratio_var": 5.614921064989176e-06
747
+ },
748
+ {
749
+ "episode": 18432,
750
+ "epoch": 0.15791638108293352,
751
+ "eps": 5,
752
+ "loss/policy_avg": 0.0028674528002738953,
753
+ "loss/value_avg": 0.27996826171875,
754
+ "lr": 2.5898437500000003e-06,
755
+ "objective/entropy": -19.57564926147461,
756
+ "objective/kl": 21.769821166992188,
757
+ "objective/non_score_reward": -1.0884910821914673,
758
+ "objective/rlhf_reward": 2.424960136413574,
759
+ "objective/scores": 3.513451099395752,
760
+ "policy/approxkl_avg": 0.003606603480875492,
761
+ "policy/clipfrac_avg": 0.03163434565067291,
762
+ "policy/entropy_avg": 2.03568172454834,
763
+ "step": 36,
764
+ "val/clipfrac_avg": 0.027482207864522934,
765
+ "val/num_eos_tokens": 13789,
766
+ "val/ratio": 1.0000629425048828,
767
+ "val/ratio_var": 9.758717169461306e-06
768
+ },
769
+ {
770
+ "episode": 18944,
771
+ "epoch": 0.1623029472241261,
772
+ "eps": 5,
773
+ "loss/policy_avg": 0.00025663431733846664,
774
+ "loss/value_avg": 0.2603837251663208,
775
+ "lr": 2.578125e-06,
776
+ "objective/entropy": -48.72997283935547,
777
+ "objective/kl": 22.040477752685547,
778
+ "objective/non_score_reward": -1.102023959159851,
779
+ "objective/rlhf_reward": 2.416844367980957,
780
+ "objective/scores": 3.5188684463500977,
781
+ "policy/approxkl_avg": 0.002840830944478512,
782
+ "policy/clipfrac_avg": 0.032977405935525894,
783
+ "policy/entropy_avg": 2.2960309982299805,
784
+ "step": 37,
785
+ "val/clipfrac_avg": 0.026058048009872437,
786
+ "val/num_eos_tokens": 17630,
787
+ "val/ratio": 1.0000097751617432,
788
+ "val/ratio_var": 4.185341367701767e-06
789
+ },
790
+ {
791
+ "episode": 19456,
792
+ "epoch": 0.1666895133653187,
793
+ "eps": 5,
794
+ "loss/policy_avg": 0.0030144676566123962,
795
+ "loss/value_avg": 0.2525879740715027,
796
+ "lr": 2.56640625e-06,
797
+ "objective/entropy": 5.543879508972168,
798
+ "objective/kl": 21.94892120361328,
799
+ "objective/non_score_reward": -1.0974459648132324,
800
+ "objective/rlhf_reward": 2.3718080520629883,
801
+ "objective/scores": 3.4692540168762207,
802
+ "policy/approxkl_avg": 0.0053519075736403465,
803
+ "policy/clipfrac_avg": 0.03657833859324455,
804
+ "policy/entropy_avg": 1.6053663492202759,
805
+ "step": 38,
806
+ "val/clipfrac_avg": 0.02781175822019577,
807
+ "val/num_eos_tokens": 11537,
808
+ "val/ratio": 0.9999196529388428,
809
+ "val/ratio_var": 1.0832703082996886e-05
810
+ },
811
+ {
812
+ "episode": 19968,
813
+ "epoch": 0.1710760795065113,
814
+ "eps": 5,
815
+ "loss/policy_avg": 0.009016571566462517,
816
+ "loss/value_avg": 0.26280635595321655,
817
+ "lr": 2.5546875000000003e-06,
818
+ "objective/entropy": -50.39720153808594,
819
+ "objective/kl": 21.66283416748047,
820
+ "objective/non_score_reward": -1.0831416845321655,
821
+ "objective/rlhf_reward": 2.513495922088623,
822
+ "objective/scores": 3.596637725830078,
823
+ "policy/approxkl_avg": 0.002841135486960411,
824
+ "policy/clipfrac_avg": 0.03448709845542908,
825
+ "policy/entropy_avg": 2.252697467803955,
826
+ "step": 39,
827
+ "val/clipfrac_avg": 0.024258539080619812,
828
+ "val/num_eos_tokens": 12811,
829
+ "val/ratio": 0.9999498128890991,
830
+ "val/ratio_var": 4.850924142374424e-06
831
+ },
832
+ {
833
+ "episode": 20480,
834
+ "epoch": 0.17546264564770392,
835
+ "eps": 5,
836
+ "loss/policy_avg": 0.004262352362275124,
837
+ "loss/value_avg": 0.2522268295288086,
838
+ "lr": 2.54296875e-06,
839
+ "objective/entropy": 1.2119975090026855,
840
+ "objective/kl": 22.785415649414062,
841
+ "objective/non_score_reward": -1.1392709016799927,
842
+ "objective/rlhf_reward": 2.568258762359619,
843
+ "objective/scores": 3.7075295448303223,
844
+ "policy/approxkl_avg": 0.00519439484924078,
845
+ "policy/clipfrac_avg": 0.03151371330022812,
846
+ "policy/entropy_avg": 1.627762794494629,
847
+ "step": 40,
848
+ "val/clipfrac_avg": 0.019196556881070137,
849
+ "val/num_eos_tokens": 11974,
850
+ "val/ratio": 0.9988241791725159,
851
+ "val/ratio_var": 1.1656878086796496e-05
852
+ },
853
+ {
854
+ "episode": 20992,
855
+ "epoch": 0.1798492117888965,
856
+ "eps": 5,
857
+ "loss/policy_avg": 0.0003326535224914551,
858
+ "loss/value_avg": 0.274797260761261,
859
+ "lr": 2.53125e-06,
860
+ "objective/entropy": -47.69501495361328,
861
+ "objective/kl": 22.703166961669922,
862
+ "objective/non_score_reward": -1.1351585388183594,
863
+ "objective/rlhf_reward": 2.5548181533813477,
864
+ "objective/scores": 3.689976692199707,
865
+ "policy/approxkl_avg": 0.0031791897490620613,
866
+ "policy/clipfrac_avg": 0.035428449511528015,
867
+ "policy/entropy_avg": 2.220440626144409,
868
+ "step": 41,
869
+ "val/clipfrac_avg": 0.018629515543580055,
870
+ "val/num_eos_tokens": 17185,
871
+ "val/ratio": 1.0000611543655396,
872
+ "val/ratio_var": 6.102215593273286e-06
873
+ },
874
+ {
875
+ "episode": 21504,
876
+ "epoch": 0.1842357779300891,
877
+ "eps": 5,
878
+ "loss/policy_avg": 0.004429924301803112,
879
+ "loss/value_avg": 0.26965782046318054,
880
+ "lr": 2.5195312500000003e-06,
881
+ "objective/entropy": -45.22355651855469,
882
+ "objective/kl": 23.047122955322266,
883
+ "objective/non_score_reward": -1.1523562669754028,
884
+ "objective/rlhf_reward": 2.4558916091918945,
885
+ "objective/scores": 3.608247756958008,
886
+ "policy/approxkl_avg": 0.0032489311415702105,
887
+ "policy/clipfrac_avg": 0.03574991971254349,
888
+ "policy/entropy_avg": 2.1724371910095215,
889
+ "step": 42,
890
+ "val/clipfrac_avg": 0.021163295954465866,
891
+ "val/num_eos_tokens": 16152,
892
+ "val/ratio": 1.0000543594360352,
893
+ "val/ratio_var": 5.2228651838959195e-06
894
+ },
895
+ {
896
+ "episode": 22016,
897
+ "epoch": 0.1886223440712817,
898
+ "eps": 5,
899
+ "loss/policy_avg": 0.0034971339628100395,
900
+ "loss/value_avg": 0.24849510192871094,
901
+ "lr": 2.5078125e-06,
902
+ "objective/entropy": -47.84535217285156,
903
+ "objective/kl": 22.962020874023438,
904
+ "objective/non_score_reward": -1.1481010913848877,
905
+ "objective/rlhf_reward": 2.6502861976623535,
906
+ "objective/scores": 3.798387289047241,
907
+ "policy/approxkl_avg": 0.0026899862568825483,
908
+ "policy/clipfrac_avg": 0.03317224234342575,
909
+ "policy/entropy_avg": 2.1769700050354004,
910
+ "step": 43,
911
+ "val/clipfrac_avg": 0.02245986834168434,
912
+ "val/num_eos_tokens": 16865,
913
+ "val/ratio": 0.9999496936798096,
914
+ "val/ratio_var": 4.76645800517872e-06
915
+ },
916
+ {
917
+ "episode": 22528,
918
+ "epoch": 0.1930089102124743,
919
+ "eps": 5,
920
+ "loss/policy_avg": -0.0013549607247114182,
921
+ "loss/value_avg": 0.2586272954940796,
922
+ "lr": 2.49609375e-06,
923
+ "objective/entropy": -44.862213134765625,
924
+ "objective/kl": 24.869279861450195,
925
+ "objective/non_score_reward": -1.2434639930725098,
926
+ "objective/rlhf_reward": 2.643916606903076,
927
+ "objective/scores": 3.887380599975586,
928
+ "policy/approxkl_avg": 0.003170889802277088,
929
+ "policy/clipfrac_avg": 0.03518493473529816,
930
+ "policy/entropy_avg": 2.122673273086548,
931
+ "step": 44,
932
+ "val/clipfrac_avg": 0.023828640580177307,
933
+ "val/num_eos_tokens": 15571,
934
+ "val/ratio": 1.0002894401550293,
935
+ "val/ratio_var": 4.998578788217856e-06
936
+ },
937
+ {
938
+ "episode": 23040,
939
+ "epoch": 0.1973954763536669,
940
+ "eps": 5,
941
+ "loss/policy_avg": -0.00417162012308836,
942
+ "loss/value_avg": 0.24716231226921082,
943
+ "lr": 2.4843750000000002e-06,
944
+ "objective/entropy": -44.43798065185547,
945
+ "objective/kl": 25.453170776367188,
946
+ "objective/non_score_reward": -1.2726584672927856,
947
+ "objective/rlhf_reward": 2.675687789916992,
948
+ "objective/scores": 3.9483463764190674,
949
+ "policy/approxkl_avg": 0.0033037192188203335,
950
+ "policy/clipfrac_avg": 0.0344209223985672,
951
+ "policy/entropy_avg": 2.096439838409424,
952
+ "step": 45,
953
+ "val/clipfrac_avg": 0.03013107180595398,
954
+ "val/num_eos_tokens": 17842,
955
+ "val/ratio": 1.0000513792037964,
956
+ "val/ratio_var": 8.593594429839868e-06
957
+ },
958
+ {
959
+ "episode": 23552,
960
+ "epoch": 0.20178204249485948,
961
+ "eps": 5,
962
+ "loss/policy_avg": 0.006732858717441559,
963
+ "loss/value_avg": 0.2426295280456543,
964
+ "lr": 2.47265625e-06,
965
+ "objective/entropy": -5.966433048248291,
966
+ "objective/kl": 25.663631439208984,
967
+ "objective/non_score_reward": -1.2831816673278809,
968
+ "objective/rlhf_reward": 2.763052463531494,
969
+ "objective/scores": 4.046234130859375,
970
+ "policy/approxkl_avg": 0.004395271651446819,
971
+ "policy/clipfrac_avg": 0.035257115960121155,
972
+ "policy/entropy_avg": 1.677065134048462,
973
+ "step": 46,
974
+ "val/clipfrac_avg": 0.030731594190001488,
975
+ "val/num_eos_tokens": 15851,
976
+ "val/ratio": 0.9994746446609497,
977
+ "val/ratio_var": 9.96153448795667e-06
978
+ },
979
+ {
980
+ "episode": 24064,
981
+ "epoch": 0.2061686086360521,
982
+ "eps": 5,
983
+ "loss/policy_avg": 0.0013937074691057205,
984
+ "loss/value_avg": 0.2316712588071823,
985
+ "lr": 2.4609375e-06,
986
+ "objective/entropy": -43.65410614013672,
987
+ "objective/kl": 26.02899169921875,
988
+ "objective/non_score_reward": -1.3014496564865112,
989
+ "objective/rlhf_reward": 2.7257275581359863,
990
+ "objective/scores": 4.027177333831787,
991
+ "policy/approxkl_avg": 0.0038581183180212975,
992
+ "policy/clipfrac_avg": 0.0356968529522419,
993
+ "policy/entropy_avg": 2.0871620178222656,
994
+ "step": 47,
995
+ "val/clipfrac_avg": 0.02329973317682743,
996
+ "val/num_eos_tokens": 17669,
997
+ "val/ratio": 0.9996811151504517,
998
+ "val/ratio_var": 6.155441496957792e-06
999
+ },
1000
+ {
1001
+ "episode": 24576,
1002
+ "epoch": 0.2105551747772447,
1003
+ "eps": 5,
1004
+ "loss/policy_avg": 0.001014474779367447,
1005
+ "loss/value_avg": 0.24769575893878937,
1006
+ "lr": 2.4492187500000002e-06,
1007
+ "objective/entropy": -42.3983154296875,
1008
+ "objective/kl": 27.05030632019043,
1009
+ "objective/non_score_reward": -1.3525152206420898,
1010
+ "objective/rlhf_reward": 2.760283946990967,
1011
+ "objective/scores": 4.112799167633057,
1012
+ "policy/approxkl_avg": 0.003388539422303438,
1013
+ "policy/clipfrac_avg": 0.03586406260728836,
1014
+ "policy/entropy_avg": 2.031424045562744,
1015
+ "step": 48,
1016
+ "val/clipfrac_avg": 0.028523240238428116,
1017
+ "val/num_eos_tokens": 20346,
1018
+ "val/ratio": 1.0003292560577393,
1019
+ "val/ratio_var": 7.955940418469254e-06
1020
+ },
1021
+ {
1022
+ "episode": 25088,
1023
+ "epoch": 0.2149417409184373,
1024
+ "eps": 5,
1025
+ "loss/policy_avg": 0.005669362843036652,
1026
+ "loss/value_avg": 0.24370360374450684,
1027
+ "lr": 2.4375e-06,
1028
+ "objective/entropy": -6.888207912445068,
1029
+ "objective/kl": 27.578134536743164,
1030
+ "objective/non_score_reward": -1.3789067268371582,
1031
+ "objective/rlhf_reward": 2.679208278656006,
1032
+ "objective/scores": 4.058115005493164,
1033
+ "policy/approxkl_avg": 0.0043523721396923065,
1034
+ "policy/clipfrac_avg": 0.03430786356329918,
1035
+ "policy/entropy_avg": 1.65708589553833,
1036
+ "step": 49,
1037
+ "val/clipfrac_avg": 0.0249293502420187,
1038
+ "val/num_eos_tokens": 16730,
1039
+ "val/ratio": 0.9998753666877747,
1040
+ "val/ratio_var": 1.2587845048983581e-05
1041
+ },
1042
+ {
1043
+ "episode": 25600,
1044
+ "epoch": 0.21932830705962988,
1045
+ "eps": 5,
1046
+ "loss/policy_avg": 0.0024556703865528107,
1047
+ "loss/value_avg": 0.24001087248325348,
1048
+ "lr": 2.42578125e-06,
1049
+ "objective/entropy": -42.264007568359375,
1050
+ "objective/kl": 26.75133514404297,
1051
+ "objective/non_score_reward": -1.33756685256958,
1052
+ "objective/rlhf_reward": 2.798081874847412,
1053
+ "objective/scores": 4.135648727416992,
1054
+ "policy/approxkl_avg": 0.0036166657228022814,
1055
+ "policy/clipfrac_avg": 0.03266894817352295,
1056
+ "policy/entropy_avg": 2.02189302444458,
1057
+ "step": 50,
1058
+ "val/clipfrac_avg": 0.017730381339788437,
1059
+ "val/num_eos_tokens": 21750,
1060
+ "val/ratio": 1.000633716583252,
1061
+ "val/ratio_var": 8.395066288358066e-06
1062
+ },
1063
+ {
1064
+ "episode": 26112,
1065
+ "epoch": 0.22371487320082248,
1066
+ "eps": 5,
1067
+ "loss/policy_avg": 0.006465999409556389,
1068
+ "loss/value_avg": 0.22410720586776733,
1069
+ "lr": 2.4140625000000002e-06,
1070
+ "objective/entropy": -26.14785385131836,
1071
+ "objective/kl": 28.21823501586914,
1072
+ "objective/non_score_reward": -1.4109117984771729,
1073
+ "objective/rlhf_reward": 2.877530813217163,
1074
+ "objective/scores": 4.288442611694336,
1075
+ "policy/approxkl_avg": 0.0034957563038915396,
1076
+ "policy/clipfrac_avg": 0.033718906342983246,
1077
+ "policy/entropy_avg": 1.8430664539337158,
1078
+ "step": 51,
1079
+ "val/clipfrac_avg": 0.014801408164203167,
1080
+ "val/num_eos_tokens": 18836,
1081
+ "val/ratio": 0.9999322295188904,
1082
+ "val/ratio_var": 7.68390964367427e-06
1083
+ },
1084
+ {
1085
+ "episode": 26624,
1086
+ "epoch": 0.22810143934201507,
1087
+ "eps": 5,
1088
+ "loss/policy_avg": 0.004318266175687313,
1089
+ "loss/value_avg": 0.2414846122264862,
1090
+ "lr": 2.40234375e-06,
1091
+ "objective/entropy": -46.04804992675781,
1092
+ "objective/kl": 27.204673767089844,
1093
+ "objective/non_score_reward": -1.3602337837219238,
1094
+ "objective/rlhf_reward": 2.978355884552002,
1095
+ "objective/scores": 4.338589668273926,
1096
+ "policy/approxkl_avg": 0.002799399895593524,
1097
+ "policy/clipfrac_avg": 0.030691005289554596,
1098
+ "policy/entropy_avg": 1.9916101694107056,
1099
+ "step": 52,
1100
+ "val/clipfrac_avg": 0.014511026442050934,
1101
+ "val/num_eos_tokens": 20909,
1102
+ "val/ratio": 0.9999108910560608,
1103
+ "val/ratio_var": 4.210955921735149e-06
1104
+ },
1105
+ {
1106
+ "episode": 27136,
1107
+ "epoch": 0.23248800548320767,
1108
+ "eps": 5,
1109
+ "loss/policy_avg": 0.003912828862667084,
1110
+ "loss/value_avg": 0.24442175030708313,
1111
+ "lr": 2.390625e-06,
1112
+ "objective/entropy": -42.2852783203125,
1113
+ "objective/kl": 28.592121124267578,
1114
+ "objective/non_score_reward": -1.4296060800552368,
1115
+ "objective/rlhf_reward": 2.733473777770996,
1116
+ "objective/scores": 4.163079738616943,
1117
+ "policy/approxkl_avg": 0.0033199505414813757,
1118
+ "policy/clipfrac_avg": 0.03190267086029053,
1119
+ "policy/entropy_avg": 1.9791926145553589,
1120
+ "step": 53,
1121
+ "val/clipfrac_avg": 0.032533541321754456,
1122
+ "val/num_eos_tokens": 24410,
1123
+ "val/ratio": 0.9999375343322754,
1124
+ "val/ratio_var": 8.744671504246071e-06
1125
+ },
1126
+ {
1127
+ "episode": 27648,
1128
+ "epoch": 0.2368745716244003,
1129
+ "eps": 5,
1130
+ "loss/policy_avg": 0.007889004424214363,
1131
+ "loss/value_avg": 0.21356219053268433,
1132
+ "lr": 2.3789062500000002e-06,
1133
+ "objective/entropy": -18.047000885009766,
1134
+ "objective/kl": 27.99648094177246,
1135
+ "objective/non_score_reward": -1.3998240232467651,
1136
+ "objective/rlhf_reward": 2.8589439392089844,
1137
+ "objective/scores": 4.258768081665039,
1138
+ "policy/approxkl_avg": 0.003893472021445632,
1139
+ "policy/clipfrac_avg": 0.03211049363017082,
1140
+ "policy/entropy_avg": 1.7189104557037354,
1141
+ "step": 54,
1142
+ "val/clipfrac_avg": 0.03184382990002632,
1143
+ "val/num_eos_tokens": 18206,
1144
+ "val/ratio": 1.0007145404815674,
1145
+ "val/ratio_var": 1.4012421161169186e-05
1146
+ },
1147
+ {
1148
+ "episode": 28160,
1149
+ "epoch": 0.24126113776559288,
1150
+ "eps": 5,
1151
+ "loss/policy_avg": 0.0016269264742732048,
1152
+ "loss/value_avg": 0.21663957834243774,
1153
+ "lr": 2.3671875e-06,
1154
+ "objective/entropy": -28.89733123779297,
1155
+ "objective/kl": 27.212053298950195,
1156
+ "objective/non_score_reward": -1.3606027364730835,
1157
+ "objective/rlhf_reward": 2.8126540184020996,
1158
+ "objective/scores": 4.173256874084473,
1159
+ "policy/approxkl_avg": 0.0035315784625709057,
1160
+ "policy/clipfrac_avg": 0.03353913128376007,
1161
+ "policy/entropy_avg": 1.8504952192306519,
1162
+ "step": 55,
1163
+ "val/clipfrac_avg": 0.01966806687414646,
1164
+ "val/num_eos_tokens": 18017,
1165
+ "val/ratio": 1.0000567436218262,
1166
+ "val/ratio_var": 8.368031558347866e-06
1167
+ },
1168
+ {
1169
+ "episode": 28672,
1170
+ "epoch": 0.24564770390678548,
1171
+ "eps": 5,
1172
+ "loss/policy_avg": 0.004167681559920311,
1173
+ "loss/value_avg": 0.2348809689283371,
1174
+ "lr": 2.35546875e-06,
1175
+ "objective/entropy": -26.520727157592773,
1176
+ "objective/kl": 28.103900909423828,
1177
+ "objective/non_score_reward": -1.4051951169967651,
1178
+ "objective/rlhf_reward": 2.761594772338867,
1179
+ "objective/scores": 4.166790008544922,
1180
+ "policy/approxkl_avg": 0.0034154942259192467,
1181
+ "policy/clipfrac_avg": 0.03300219029188156,
1182
+ "policy/entropy_avg": 1.8018219470977783,
1183
+ "step": 56,
1184
+ "val/clipfrac_avg": 0.024162570014595985,
1185
+ "val/num_eos_tokens": 17735,
1186
+ "val/ratio": 1.0002646446228027,
1187
+ "val/ratio_var": 5.345529189071385e-06
1188
+ },
1189
+ {
1190
+ "episode": 29184,
1191
+ "epoch": 0.25003427004797807,
1192
+ "eps": 5,
1193
+ "loss/policy_avg": 0.003991998732089996,
1194
+ "loss/value_avg": 0.23025518655776978,
1195
+ "lr": 2.3437500000000002e-06,
1196
+ "objective/entropy": -43.95574188232422,
1197
+ "objective/kl": 28.330055236816406,
1198
+ "objective/non_score_reward": -1.4165027141571045,
1199
+ "objective/rlhf_reward": 2.8100287914276123,
1200
+ "objective/scores": 4.226531505584717,
1201
+ "policy/approxkl_avg": 0.0031039149034768343,
1202
+ "policy/clipfrac_avg": 0.0343279168009758,
1203
+ "policy/entropy_avg": 1.9948017597198486,
1204
+ "step": 57,
1205
+ "val/clipfrac_avg": 0.0235802810639143,
1206
+ "val/num_eos_tokens": 19582,
1207
+ "val/ratio": 0.9998975992202759,
1208
+ "val/ratio_var": 3.4535264603619e-06
1209
+ },
1210
+ {
1211
+ "episode": 29696,
1212
+ "epoch": 0.25442083618917066,
1213
+ "eps": 5,
1214
+ "loss/policy_avg": -0.005810210481286049,
1215
+ "loss/value_avg": 0.2340400516986847,
1216
+ "lr": 2.33203125e-06,
1217
+ "objective/entropy": -41.727108001708984,
1218
+ "objective/kl": 28.618722915649414,
1219
+ "objective/non_score_reward": -1.4309360980987549,
1220
+ "objective/rlhf_reward": 2.772367238998413,
1221
+ "objective/scores": 4.203303337097168,
1222
+ "policy/approxkl_avg": 0.0029092328622937202,
1223
+ "policy/clipfrac_avg": 0.03318607434630394,
1224
+ "policy/entropy_avg": 1.9372591972351074,
1225
+ "step": 58,
1226
+ "val/clipfrac_avg": 0.02145560272037983,
1227
+ "val/num_eos_tokens": 23466,
1228
+ "val/ratio": 0.9998555183410645,
1229
+ "val/ratio_var": 5.723162757931277e-06
1230
+ },
1231
+ {
1232
+ "episode": 30208,
1233
+ "epoch": 0.25880740233036326,
1234
+ "eps": 5,
1235
+ "loss/policy_avg": 0.0030759507790207863,
1236
+ "loss/value_avg": 0.2046152651309967,
1237
+ "lr": 2.3203125e-06,
1238
+ "objective/entropy": -23.432676315307617,
1239
+ "objective/kl": 27.8800048828125,
1240
+ "objective/non_score_reward": -1.3940002918243408,
1241
+ "objective/rlhf_reward": 2.849257230758667,
1242
+ "objective/scores": 4.243257522583008,
1243
+ "policy/approxkl_avg": 0.003216572804376483,
1244
+ "policy/clipfrac_avg": 0.03146814554929733,
1245
+ "policy/entropy_avg": 1.7635722160339355,
1246
+ "step": 59,
1247
+ "val/clipfrac_avg": 0.016629783436655998,
1248
+ "val/num_eos_tokens": 17909,
1249
+ "val/ratio": 1.000119686126709,
1250
+ "val/ratio_var": 7.115571406757226e-06
1251
+ },
1252
+ {
1253
+ "episode": 30720,
1254
+ "epoch": 0.26319396847155585,
1255
+ "eps": 5,
1256
+ "loss/policy_avg": 0.0051596928387880325,
1257
+ "loss/value_avg": 0.21278434991836548,
1258
+ "lr": 2.30859375e-06,
1259
+ "objective/entropy": -18.64295196533203,
1260
+ "objective/kl": 28.58637237548828,
1261
+ "objective/non_score_reward": -1.4293186664581299,
1262
+ "objective/rlhf_reward": 2.8722569942474365,
1263
+ "objective/scores": 4.301575660705566,
1264
+ "policy/approxkl_avg": 0.0036908402107656,
1265
+ "policy/clipfrac_avg": 0.03174355626106262,
1266
+ "policy/entropy_avg": 1.7053215503692627,
1267
+ "step": 60,
1268
+ "val/clipfrac_avg": 0.018147464841604233,
1269
+ "val/num_eos_tokens": 16563,
1270
+ "val/ratio": 0.9996346235275269,
1271
+ "val/ratio_var": 8.742338650336023e-06
1272
+ },
1273
+ {
1274
+ "episode": 31232,
1275
+ "epoch": 0.26758053461274844,
1276
+ "eps": 5,
1277
+ "loss/policy_avg": 0.0019552409648895264,
1278
+ "loss/value_avg": 0.21977171301841736,
1279
+ "lr": 2.296875e-06,
1280
+ "objective/entropy": -34.63594055175781,
1281
+ "objective/kl": 28.0120906829834,
1282
+ "objective/non_score_reward": -1.400604486465454,
1283
+ "objective/rlhf_reward": 2.865410566329956,
1284
+ "objective/scores": 4.26601505279541,
1285
+ "policy/approxkl_avg": 0.003109532408416271,
1286
+ "policy/clipfrac_avg": 0.031170206144452095,
1287
+ "policy/entropy_avg": 1.8843848705291748,
1288
+ "step": 61,
1289
+ "val/clipfrac_avg": 0.018924321979284286,
1290
+ "val/num_eos_tokens": 17492,
1291
+ "val/ratio": 1.000000238418579,
1292
+ "val/ratio_var": 6.473004305007635e-06
1293
+ },
1294
+ {
1295
+ "episode": 31744,
1296
+ "epoch": 0.27196710075394104,
1297
+ "eps": 5,
1298
+ "loss/policy_avg": 0.0027344757691025734,
1299
+ "loss/value_avg": 0.2190888375043869,
1300
+ "lr": 2.28515625e-06,
1301
+ "objective/entropy": -42.28523254394531,
1302
+ "objective/kl": 28.281116485595703,
1303
+ "objective/non_score_reward": -1.4140558242797852,
1304
+ "objective/rlhf_reward": 2.985428810119629,
1305
+ "objective/scores": 4.399484634399414,
1306
+ "policy/approxkl_avg": 0.003223988227546215,
1307
+ "policy/clipfrac_avg": 0.03254333883523941,
1308
+ "policy/entropy_avg": 1.9692389965057373,
1309
+ "step": 62,
1310
+ "val/clipfrac_avg": 0.024257011711597443,
1311
+ "val/num_eos_tokens": 19545,
1312
+ "val/ratio": 0.9998095035552979,
1313
+ "val/ratio_var": 5.689699264621595e-06
1314
+ },
1315
+ {
1316
+ "episode": 32256,
1317
+ "epoch": 0.27635366689513363,
1318
+ "eps": 5,
1319
+ "loss/policy_avg": 0.00750521756708622,
1320
+ "loss/value_avg": 0.21058180928230286,
1321
+ "lr": 2.2734375e-06,
1322
+ "objective/entropy": -41.52886962890625,
1323
+ "objective/kl": 28.18915367126465,
1324
+ "objective/non_score_reward": -1.4094576835632324,
1325
+ "objective/rlhf_reward": 2.7793426513671875,
1326
+ "objective/scores": 4.18880033493042,
1327
+ "policy/approxkl_avg": 0.00307050091214478,
1328
+ "policy/clipfrac_avg": 0.031591419130563736,
1329
+ "policy/entropy_avg": 1.9588576555252075,
1330
+ "step": 63,
1331
+ "val/clipfrac_avg": 0.02337929978966713,
1332
+ "val/num_eos_tokens": 16845,
1333
+ "val/ratio": 1.0004616975784302,
1334
+ "val/ratio_var": 9.101933756028302e-06
1335
+ },
1336
+ {
1337
+ "episode": 32768,
1338
+ "epoch": 0.2807402330363262,
1339
+ "eps": 5,
1340
+ "loss/policy_avg": 0.0035128919407725334,
1341
+ "loss/value_avg": 0.2123112976551056,
1342
+ "lr": 2.26171875e-06,
1343
+ "objective/entropy": -29.706064224243164,
1344
+ "objective/kl": 27.973098754882812,
1345
+ "objective/non_score_reward": -1.3986550569534302,
1346
+ "objective/rlhf_reward": 2.89738130569458,
1347
+ "objective/scores": 4.296036243438721,
1348
+ "policy/approxkl_avg": 0.0032222147565335035,
1349
+ "policy/clipfrac_avg": 0.03244758024811745,
1350
+ "policy/entropy_avg": 1.8412973880767822,
1351
+ "step": 64,
1352
+ "val/clipfrac_avg": 0.01701158657670021,
1353
+ "val/num_eos_tokens": 16471,
1354
+ "val/ratio": 0.999403715133667,
1355
+ "val/ratio_var": 7.529881258960813e-06
1356
+ },
1357
+ {
1358
+ "episode": 33280,
1359
+ "epoch": 0.2851267991775189,
1360
+ "eps": 5,
1361
+ "loss/policy_avg": 0.0004525454714894295,
1362
+ "loss/value_avg": 0.2151668667793274,
1363
+ "lr": 2.25e-06,
1364
+ "objective/entropy": -45.270347595214844,
1365
+ "objective/kl": 27.836875915527344,
1366
+ "objective/non_score_reward": -1.3918439149856567,
1367
+ "objective/rlhf_reward": 2.7277212142944336,
1368
+ "objective/scores": 4.119565010070801,
1369
+ "policy/approxkl_avg": 0.0030939963180571795,
1370
+ "policy/clipfrac_avg": 0.029837429523468018,
1371
+ "policy/entropy_avg": 1.9481849670410156,
1372
+ "step": 65,
1373
+ "val/clipfrac_avg": 0.026501141488552094,
1374
+ "val/num_eos_tokens": 17056,
1375
+ "val/ratio": 0.9998571276664734,
1376
+ "val/ratio_var": 6.92433559379424e-06
1377
+ },
1378
+ {
1379
+ "episode": 33792,
1380
+ "epoch": 0.28951336531871147,
1381
+ "eps": 5,
1382
+ "loss/policy_avg": 0.0063807349652051926,
1383
+ "loss/value_avg": 0.21251490712165833,
1384
+ "lr": 2.23828125e-06,
1385
+ "objective/entropy": -25.12890625,
1386
+ "objective/kl": 28.404438018798828,
1387
+ "objective/non_score_reward": -1.4202219247817993,
1388
+ "objective/rlhf_reward": 3.0330071449279785,
1389
+ "objective/scores": 4.453228950500488,
1390
+ "policy/approxkl_avg": 0.004077461082488298,
1391
+ "policy/clipfrac_avg": 0.03214584290981293,
1392
+ "policy/entropy_avg": 1.7610023021697998,
1393
+ "step": 66,
1394
+ "val/clipfrac_avg": 0.030560344457626343,
1395
+ "val/num_eos_tokens": 17430,
1396
+ "val/ratio": 1.000243067741394,
1397
+ "val/ratio_var": 1.1609051398409065e-05
1398
+ },
1399
+ {
1400
+ "episode": 34304,
1401
+ "epoch": 0.29389993145990406,
1402
+ "eps": 5,
1403
+ "loss/policy_avg": 0.005549637600779533,
1404
+ "loss/value_avg": 0.2186199426651001,
1405
+ "lr": 2.2265625e-06,
1406
+ "objective/entropy": -26.157814025878906,
1407
+ "objective/kl": 28.302576065063477,
1408
+ "objective/non_score_reward": -1.4151287078857422,
1409
+ "objective/rlhf_reward": 2.847787857055664,
1410
+ "objective/scores": 4.262916564941406,
1411
+ "policy/approxkl_avg": 0.003851176705211401,
1412
+ "policy/clipfrac_avg": 0.030647093430161476,
1413
+ "policy/entropy_avg": 1.7585735321044922,
1414
+ "step": 67,
1415
+ "val/clipfrac_avg": 0.016023114323616028,
1416
+ "val/num_eos_tokens": 15918,
1417
+ "val/ratio": 0.9992672204971313,
1418
+ "val/ratio_var": 6.194312845764216e-06
1419
+ },
1420
+ {
1421
+ "episode": 34816,
1422
+ "epoch": 0.29828649760109666,
1423
+ "eps": 5,
1424
+ "loss/policy_avg": -0.002579714171588421,
1425
+ "loss/value_avg": 0.22957545518875122,
1426
+ "lr": 2.21484375e-06,
1427
+ "objective/entropy": -46.926902770996094,
1428
+ "objective/kl": 28.35796356201172,
1429
+ "objective/non_score_reward": -1.417898178100586,
1430
+ "objective/rlhf_reward": 3.0081119537353516,
1431
+ "objective/scores": 4.4260101318359375,
1432
+ "policy/approxkl_avg": 0.0024162721820175648,
1433
+ "policy/clipfrac_avg": 0.030427245423197746,
1434
+ "policy/entropy_avg": 1.986201524734497,
1435
+ "step": 68,
1436
+ "val/clipfrac_avg": 0.02106384187936783,
1437
+ "val/num_eos_tokens": 16744,
1438
+ "val/ratio": 0.9999403953552246,
1439
+ "val/ratio_var": 4.121975962334545e-06
1440
+ },
1441
+ {
1442
+ "episode": 35328,
1443
+ "epoch": 0.30267306374228925,
1444
+ "eps": 5,
1445
+ "loss/policy_avg": 0.0004816874861717224,
1446
+ "loss/value_avg": 0.2244284301996231,
1447
+ "lr": 2.203125e-06,
1448
+ "objective/entropy": -45.291831970214844,
1449
+ "objective/kl": 28.912921905517578,
1450
+ "objective/non_score_reward": -1.4456461668014526,
1451
+ "objective/rlhf_reward": 2.9440298080444336,
1452
+ "objective/scores": 4.389676094055176,
1453
+ "policy/approxkl_avg": 0.002792609389871359,
1454
+ "policy/clipfrac_avg": 0.03056211769580841,
1455
+ "policy/entropy_avg": 1.9658164978027344,
1456
+ "step": 69,
1457
+ "val/clipfrac_avg": 0.013326474465429783,
1458
+ "val/num_eos_tokens": 21689,
1459
+ "val/ratio": 0.9998493790626526,
1460
+ "val/ratio_var": 7.042471224849578e-06
1461
+ },
1462
+ {
1463
+ "episode": 35840,
1464
+ "epoch": 0.30705962988348184,
1465
+ "eps": 5,
1466
+ "loss/policy_avg": 0.0018423181027173996,
1467
+ "loss/value_avg": 0.2250077724456787,
1468
+ "lr": 2.19140625e-06,
1469
+ "objective/entropy": -45.46595764160156,
1470
+ "objective/kl": 29.026504516601562,
1471
+ "objective/non_score_reward": -1.4513252973556519,
1472
+ "objective/rlhf_reward": 2.9487290382385254,
1473
+ "objective/scores": 4.400054454803467,
1474
+ "policy/approxkl_avg": 0.003939199261367321,
1475
+ "policy/clipfrac_avg": 0.031094079837203026,
1476
+ "policy/entropy_avg": 1.9610083103179932,
1477
+ "step": 70,
1478
+ "val/clipfrac_avg": 0.01343858614563942,
1479
+ "val/num_eos_tokens": 20499,
1480
+ "val/ratio": 0.999684751033783,
1481
+ "val/ratio_var": 6.334644240268972e-06
1482
+ },
1483
+ {
1484
+ "episode": 36352,
1485
+ "epoch": 0.31144619602467444,
1486
+ "eps": 5,
1487
+ "loss/policy_avg": -0.0021029403433203697,
1488
+ "loss/value_avg": 0.19206474721431732,
1489
+ "lr": 2.1796875e-06,
1490
+ "objective/entropy": -42.51447296142578,
1491
+ "objective/kl": 30.721092224121094,
1492
+ "objective/non_score_reward": -1.5360547304153442,
1493
+ "objective/rlhf_reward": 2.885496139526367,
1494
+ "objective/scores": 4.421550750732422,
1495
+ "policy/approxkl_avg": 0.002968056360259652,
1496
+ "policy/clipfrac_avg": 0.028208117932081223,
1497
+ "policy/entropy_avg": 1.9163283109664917,
1498
+ "step": 71,
1499
+ "val/clipfrac_avg": 0.01538503635674715,
1500
+ "val/num_eos_tokens": 21238,
1501
+ "val/ratio": 1.0007789134979248,
1502
+ "val/ratio_var": 4.891162916464964e-06
1503
+ },
1504
+ {
1505
+ "episode": 36864,
1506
+ "epoch": 0.31583276216586703,
1507
+ "eps": 5,
1508
+ "loss/policy_avg": -0.001980050466954708,
1509
+ "loss/value_avg": 0.2347830832004547,
1510
+ "lr": 2.16796875e-06,
1511
+ "objective/entropy": -44.55876541137695,
1512
+ "objective/kl": 31.202877044677734,
1513
+ "objective/non_score_reward": -1.5601439476013184,
1514
+ "objective/rlhf_reward": 3.0216851234436035,
1515
+ "objective/scores": 4.581829071044922,
1516
+ "policy/approxkl_avg": 0.003652730491012335,
1517
+ "policy/clipfrac_avg": 0.029209883883595467,
1518
+ "policy/entropy_avg": 1.894657015800476,
1519
+ "step": 72,
1520
+ "val/clipfrac_avg": 0.013316318392753601,
1521
+ "val/num_eos_tokens": 22115,
1522
+ "val/ratio": 1.0000977516174316,
1523
+ "val/ratio_var": 7.180529337347252e-06
1524
+ },
1525
+ {
1526
+ "episode": 37376,
1527
+ "epoch": 0.3202193283070596,
1528
+ "eps": 5,
1529
+ "loss/policy_avg": -0.0009753154590725899,
1530
+ "loss/value_avg": 0.21619904041290283,
1531
+ "lr": 2.15625e-06,
1532
+ "objective/entropy": -45.03413391113281,
1533
+ "objective/kl": 33.025238037109375,
1534
+ "objective/non_score_reward": -1.6512620449066162,
1535
+ "objective/rlhf_reward": 2.9435040950775146,
1536
+ "objective/scores": 4.594766139984131,
1537
+ "policy/approxkl_avg": 0.0027754739858210087,
1538
+ "policy/clipfrac_avg": 0.02909201756119728,
1539
+ "policy/entropy_avg": 1.883234977722168,
1540
+ "step": 73,
1541
+ "val/clipfrac_avg": 0.015321025624871254,
1542
+ "val/num_eos_tokens": 22473,
1543
+ "val/ratio": 0.9997678995132446,
1544
+ "val/ratio_var": 5.409801815403625e-06
1545
+ },
1546
+ {
1547
+ "episode": 37888,
1548
+ "epoch": 0.3246058944482522,
1549
+ "eps": 5,
1550
+ "loss/policy_avg": 0.0035352958366274834,
1551
+ "loss/value_avg": 0.22200468182563782,
1552
+ "lr": 2.14453125e-06,
1553
+ "objective/entropy": -43.88129425048828,
1554
+ "objective/kl": 31.707870483398438,
1555
+ "objective/non_score_reward": -1.5853935480117798,
1556
+ "objective/rlhf_reward": 3.0117926597595215,
1557
+ "objective/scores": 4.597186088562012,
1558
+ "policy/approxkl_avg": 0.003338968614116311,
1559
+ "policy/clipfrac_avg": 0.02952185645699501,
1560
+ "policy/entropy_avg": 1.8764941692352295,
1561
+ "step": 74,
1562
+ "val/clipfrac_avg": 0.01802271604537964,
1563
+ "val/num_eos_tokens": 23645,
1564
+ "val/ratio": 0.9994179010391235,
1565
+ "val/ratio_var": 4.616674232238438e-06
1566
+ },
1567
+ {
1568
+ "episode": 38400,
1569
+ "epoch": 0.3289924605894448,
1570
+ "eps": 5,
1571
+ "loss/policy_avg": -0.0047211856581270695,
1572
+ "loss/value_avg": 0.22870859503746033,
1573
+ "lr": 2.1328125e-06,
1574
+ "objective/entropy": -45.87744903564453,
1575
+ "objective/kl": 32.42133331298828,
1576
+ "objective/non_score_reward": -1.621066927909851,
1577
+ "objective/rlhf_reward": 3.0603857040405273,
1578
+ "objective/scores": 4.681452751159668,
1579
+ "policy/approxkl_avg": 0.00320438458584249,
1580
+ "policy/clipfrac_avg": 0.027557462453842163,
1581
+ "policy/entropy_avg": 1.89240562915802,
1582
+ "step": 75,
1583
+ "val/clipfrac_avg": 0.028391696512699127,
1584
+ "val/num_eos_tokens": 22337,
1585
+ "val/ratio": 1.0004334449768066,
1586
+ "val/ratio_var": 5.193583547224989e-06
1587
+ },
1588
+ {
1589
+ "episode": 38912,
1590
+ "epoch": 0.3333790267306374,
1591
+ "eps": 5,
1592
+ "loss/policy_avg": -0.0018232641741633415,
1593
+ "loss/value_avg": 0.22250117361545563,
1594
+ "lr": 2.12109375e-06,
1595
+ "objective/entropy": -44.9810905456543,
1596
+ "objective/kl": 32.4366569519043,
1597
+ "objective/non_score_reward": -1.6218329668045044,
1598
+ "objective/rlhf_reward": 2.8586840629577637,
1599
+ "objective/scores": 4.4805169105529785,
1600
+ "policy/approxkl_avg": 0.0027742942329496145,
1601
+ "policy/clipfrac_avg": 0.028607677668333054,
1602
+ "policy/entropy_avg": 1.8870813846588135,
1603
+ "step": 76,
1604
+ "val/clipfrac_avg": 0.01954970508813858,
1605
+ "val/num_eos_tokens": 23025,
1606
+ "val/ratio": 0.999954879283905,
1607
+ "val/ratio_var": 7.360893505392596e-06
1608
+ },
1609
+ {
1610
+ "episode": 39424,
1611
+ "epoch": 0.33776559287183,
1612
+ "eps": 5,
1613
+ "loss/policy_avg": 0.002040557563304901,
1614
+ "loss/value_avg": 0.20820286870002747,
1615
+ "lr": 2.109375e-06,
1616
+ "objective/entropy": -45.3465461730957,
1617
+ "objective/kl": 33.44585418701172,
1618
+ "objective/non_score_reward": -1.6722928285598755,
1619
+ "objective/rlhf_reward": 2.9919791221618652,
1620
+ "objective/scores": 4.664271831512451,
1621
+ "policy/approxkl_avg": 0.0030342168174684048,
1622
+ "policy/clipfrac_avg": 0.02885330654680729,
1623
+ "policy/entropy_avg": 1.8954873085021973,
1624
+ "step": 77,
1625
+ "val/clipfrac_avg": 0.016231678426265717,
1626
+ "val/num_eos_tokens": 20887,
1627
+ "val/ratio": 1.0000638961791992,
1628
+ "val/ratio_var": 9.632998626329936e-06
1629
+ },
1630
+ {
1631
+ "episode": 39936,
1632
+ "epoch": 0.3421521590130226,
1633
+ "eps": 5,
1634
+ "loss/policy_avg": -0.00032701529562473297,
1635
+ "loss/value_avg": 0.2173679769039154,
1636
+ "lr": 2.09765625e-06,
1637
+ "objective/entropy": -47.16279602050781,
1638
+ "objective/kl": 32.62769317626953,
1639
+ "objective/non_score_reward": -1.6313844919204712,
1640
+ "objective/rlhf_reward": 3.0506396293640137,
1641
+ "objective/scores": 4.682024002075195,
1642
+ "policy/approxkl_avg": 0.0030062790028750896,
1643
+ "policy/clipfrac_avg": 0.027039213106036186,
1644
+ "policy/entropy_avg": 1.8599188327789307,
1645
+ "step": 78,
1646
+ "val/clipfrac_avg": 0.015515465289354324,
1647
+ "val/num_eos_tokens": 23715,
1648
+ "val/ratio": 1.0002635717391968,
1649
+ "val/ratio_var": 6.824520369264064e-06
1650
+ },
1651
+ {
1652
+ "episode": 40448,
1653
+ "epoch": 0.34653872515421524,
1654
+ "eps": 5,
1655
+ "loss/policy_avg": 0.0014511756598949432,
1656
+ "loss/value_avg": 0.2262498438358307,
1657
+ "lr": 2.0859375e-06,
1658
+ "objective/entropy": -48.478755950927734,
1659
+ "objective/kl": 32.30998229980469,
1660
+ "objective/non_score_reward": -1.6154991388320923,
1661
+ "objective/rlhf_reward": 3.0539960861206055,
1662
+ "objective/scores": 4.669495105743408,
1663
+ "policy/approxkl_avg": 0.002757284790277481,
1664
+ "policy/clipfrac_avg": 0.0267037246376276,
1665
+ "policy/entropy_avg": 1.8800439834594727,
1666
+ "step": 79,
1667
+ "val/clipfrac_avg": 0.01783629134297371,
1668
+ "val/num_eos_tokens": 23726,
1669
+ "val/ratio": 1.0002996921539307,
1670
+ "val/ratio_var": 5.29440694663208e-06
1671
+ },
1672
+ {
1673
+ "episode": 40960,
1674
+ "epoch": 0.35092529129540784,
1675
+ "eps": 5,
1676
+ "loss/policy_avg": -0.00030362606048583984,
1677
+ "loss/value_avg": 0.23645125329494476,
1678
+ "lr": 2.07421875e-06,
1679
+ "objective/entropy": -48.08945846557617,
1680
+ "objective/kl": 32.43437194824219,
1681
+ "objective/non_score_reward": -1.6217186450958252,
1682
+ "objective/rlhf_reward": 3.0068461894989014,
1683
+ "objective/scores": 4.628564834594727,
1684
+ "policy/approxkl_avg": 0.002676879521459341,
1685
+ "policy/clipfrac_avg": 0.027121271938085556,
1686
+ "policy/entropy_avg": 1.8344483375549316,
1687
+ "step": 80,
1688
+ "val/clipfrac_avg": 0.017351917922496796,
1689
+ "val/num_eos_tokens": 26035,
1690
+ "val/ratio": 0.9995693564414978,
1691
+ "val/ratio_var": 4.564853497868171e-06
1692
+ },
1693
+ {
1694
+ "episode": 41472,
1695
+ "epoch": 0.35531185743660043,
1696
+ "eps": 5,
1697
+ "loss/policy_avg": -0.0004015360027551651,
1698
+ "loss/value_avg": 0.2039456069469452,
1699
+ "lr": 2.0625e-06,
1700
+ "objective/entropy": -50.942020416259766,
1701
+ "objective/kl": 31.24224090576172,
1702
+ "objective/non_score_reward": -1.5621120929718018,
1703
+ "objective/rlhf_reward": 2.9734346866607666,
1704
+ "objective/scores": 4.535546779632568,
1705
+ "policy/approxkl_avg": 0.0028801425360143185,
1706
+ "policy/clipfrac_avg": 0.028739815577864647,
1707
+ "policy/entropy_avg": 1.8809466361999512,
1708
+ "step": 81,
1709
+ "val/clipfrac_avg": 0.02381589636206627,
1710
+ "val/num_eos_tokens": 22486,
1711
+ "val/ratio": 1.000486135482788,
1712
+ "val/ratio_var": 5.7636229939816985e-06
1713
+ },
1714
+ {
1715
+ "episode": 41984,
1716
+ "epoch": 0.359698423577793,
1717
+ "eps": 5,
1718
+ "loss/policy_avg": 0.007100890390574932,
1719
+ "loss/value_avg": 0.19654756784439087,
1720
+ "lr": 2.05078125e-06,
1721
+ "objective/entropy": -48.43541717529297,
1722
+ "objective/kl": 33.017799377441406,
1723
+ "objective/non_score_reward": -1.6508899927139282,
1724
+ "objective/rlhf_reward": 3.181412696838379,
1725
+ "objective/scores": 4.832302570343018,
1726
+ "policy/approxkl_avg": 0.002802068367600441,
1727
+ "policy/clipfrac_avg": 0.025402868166565895,
1728
+ "policy/entropy_avg": 1.847480058670044,
1729
+ "step": 82,
1730
+ "val/clipfrac_avg": 0.020801864564418793,
1731
+ "val/num_eos_tokens": 23198,
1732
+ "val/ratio": 1.0002295970916748,
1733
+ "val/ratio_var": 4.7933585847204085e-06
1734
+ },
1735
+ {
1736
+ "episode": 42496,
1737
+ "epoch": 0.3640849897189856,
1738
+ "eps": 5,
1739
+ "loss/policy_avg": 0.0014801118522882462,
1740
+ "loss/value_avg": 0.20368510484695435,
1741
+ "lr": 2.0390625e-06,
1742
+ "objective/entropy": -51.05237579345703,
1743
+ "objective/kl": 32.151611328125,
1744
+ "objective/non_score_reward": -1.6075806617736816,
1745
+ "objective/rlhf_reward": 3.0617752075195312,
1746
+ "objective/scores": 4.669355869293213,
1747
+ "policy/approxkl_avg": 0.0025460361503064632,
1748
+ "policy/clipfrac_avg": 0.02481936663389206,
1749
+ "policy/entropy_avg": 1.8554236888885498,
1750
+ "step": 83,
1751
+ "val/clipfrac_avg": 0.024282867088913918,
1752
+ "val/num_eos_tokens": 23201,
1753
+ "val/ratio": 0.9998276829719543,
1754
+ "val/ratio_var": 3.309921112304437e-06
1755
+ },
1756
+ {
1757
+ "episode": 43008,
1758
+ "epoch": 0.3684715558601782,
1759
+ "eps": 5,
1760
+ "loss/policy_avg": 0.0065193381160497665,
1761
+ "loss/value_avg": 0.21418508887290955,
1762
+ "lr": 2.02734375e-06,
1763
+ "objective/entropy": -51.27957534790039,
1764
+ "objective/kl": 32.4244270324707,
1765
+ "objective/non_score_reward": -1.6212214231491089,
1766
+ "objective/rlhf_reward": 3.128838539123535,
1767
+ "objective/scores": 4.750060081481934,
1768
+ "policy/approxkl_avg": 0.002565734088420868,
1769
+ "policy/clipfrac_avg": 0.025620460510253906,
1770
+ "policy/entropy_avg": 1.815622329711914,
1771
+ "step": 84,
1772
+ "val/clipfrac_avg": 0.014234954491257668,
1773
+ "val/num_eos_tokens": 24471,
1774
+ "val/ratio": 1.0001951456069946,
1775
+ "val/ratio_var": 6.2828203226672485e-06
1776
+ },
1777
+ {
1778
+ "episode": 43520,
1779
+ "epoch": 0.3728581220013708,
1780
+ "eps": 5,
1781
+ "loss/policy_avg": 0.002269826363772154,
1782
+ "loss/value_avg": 0.22035224735736847,
1783
+ "lr": 2.015625e-06,
1784
+ "objective/entropy": -50.43995666503906,
1785
+ "objective/kl": 33.585693359375,
1786
+ "objective/non_score_reward": -1.679284691810608,
1787
+ "objective/rlhf_reward": 2.9680705070495605,
1788
+ "objective/scores": 4.647355079650879,
1789
+ "policy/approxkl_avg": 0.0028171499725431204,
1790
+ "policy/clipfrac_avg": 0.0254150852560997,
1791
+ "policy/entropy_avg": 1.8183186054229736,
1792
+ "step": 85,
1793
+ "val/clipfrac_avg": 0.016307460144162178,
1794
+ "val/num_eos_tokens": 25083,
1795
+ "val/ratio": 0.999747097492218,
1796
+ "val/ratio_var": 4.847966465604259e-06
1797
+ },
1798
+ {
1799
+ "episode": 44032,
1800
+ "epoch": 0.3772446881425634,
1801
+ "eps": 5,
1802
+ "loss/policy_avg": -0.0038543278351426125,
1803
+ "loss/value_avg": 0.21373863518238068,
1804
+ "lr": 2.00390625e-06,
1805
+ "objective/entropy": -51.48972702026367,
1806
+ "objective/kl": 34.424583435058594,
1807
+ "objective/non_score_reward": -1.7212293148040771,
1808
+ "objective/rlhf_reward": 3.021406412124634,
1809
+ "objective/scores": 4.742635726928711,
1810
+ "policy/approxkl_avg": 0.0027820314280688763,
1811
+ "policy/clipfrac_avg": 0.02570197358727455,
1812
+ "policy/entropy_avg": 1.849883794784546,
1813
+ "step": 86,
1814
+ "val/clipfrac_avg": 0.016377516090869904,
1815
+ "val/num_eos_tokens": 23250,
1816
+ "val/ratio": 0.9997796416282654,
1817
+ "val/ratio_var": 4.842361704504583e-06
1818
+ },
1819
+ {
1820
+ "episode": 44544,
1821
+ "epoch": 0.381631254283756,
1822
+ "eps": 5,
1823
+ "loss/policy_avg": -0.0015501310117542744,
1824
+ "loss/value_avg": 0.23201991617679596,
1825
+ "lr": 1.9921875e-06,
1826
+ "objective/entropy": -51.2038688659668,
1827
+ "objective/kl": 34.93539047241211,
1828
+ "objective/non_score_reward": -1.746769666671753,
1829
+ "objective/rlhf_reward": 3.0822598934173584,
1830
+ "objective/scores": 4.829029560089111,
1831
+ "policy/approxkl_avg": 0.0028443641494959593,
1832
+ "policy/clipfrac_avg": 0.025669323280453682,
1833
+ "policy/entropy_avg": 1.8496443033218384,
1834
+ "step": 87,
1835
+ "val/clipfrac_avg": 0.012260429561138153,
1836
+ "val/num_eos_tokens": 23147,
1837
+ "val/ratio": 0.999835729598999,
1838
+ "val/ratio_var": 6.624535672017373e-06
1839
+ },
1840
+ {
1841
+ "episode": 45056,
1842
+ "epoch": 0.3860178204249486,
1843
+ "eps": 5,
1844
+ "loss/policy_avg": 0.004490493796765804,
1845
+ "loss/value_avg": 0.21559040248394012,
1846
+ "lr": 1.98046875e-06,
1847
+ "objective/entropy": -49.819862365722656,
1848
+ "objective/kl": 35.83782196044922,
1849
+ "objective/non_score_reward": -1.7918909788131714,
1850
+ "objective/rlhf_reward": 3.0484533309936523,
1851
+ "objective/scores": 4.840344429016113,
1852
+ "policy/approxkl_avg": 0.0029672845266759396,
1853
+ "policy/clipfrac_avg": 0.025179177522659302,
1854
+ "policy/entropy_avg": 1.8241046667099,
1855
+ "step": 88,
1856
+ "val/clipfrac_avg": 0.011389493942260742,
1857
+ "val/num_eos_tokens": 22325,
1858
+ "val/ratio": 0.9995887279510498,
1859
+ "val/ratio_var": 3.6028022805112414e-06
1860
+ },
1861
+ {
1862
+ "episode": 45568,
1863
+ "epoch": 0.3904043865661412,
1864
+ "eps": 5,
1865
+ "loss/policy_avg": 0.003285501152276993,
1866
+ "loss/value_avg": 0.20261438190937042,
1867
+ "lr": 1.96875e-06,
1868
+ "objective/entropy": -51.72087478637695,
1869
+ "objective/kl": 34.95625305175781,
1870
+ "objective/non_score_reward": -1.7478127479553223,
1871
+ "objective/rlhf_reward": 3.141611099243164,
1872
+ "objective/scores": 4.889423847198486,
1873
+ "policy/approxkl_avg": 0.0026469011791050434,
1874
+ "policy/clipfrac_avg": 0.025560175999999046,
1875
+ "policy/entropy_avg": 1.8512508869171143,
1876
+ "step": 89,
1877
+ "val/clipfrac_avg": 0.018557574599981308,
1878
+ "val/num_eos_tokens": 20308,
1879
+ "val/ratio": 1.0001813173294067,
1880
+ "val/ratio_var": 4.480873940337915e-06
1881
+ },
1882
+ {
1883
+ "episode": 46080,
1884
+ "epoch": 0.3947909527073338,
1885
+ "eps": 5,
1886
+ "loss/policy_avg": -0.0009333370253443718,
1887
+ "loss/value_avg": 0.1969861537218094,
1888
+ "lr": 1.95703125e-06,
1889
+ "objective/entropy": -51.90027618408203,
1890
+ "objective/kl": 35.76751708984375,
1891
+ "objective/non_score_reward": -1.7883758544921875,
1892
+ "objective/rlhf_reward": 3.085066795349121,
1893
+ "objective/scores": 4.873442649841309,
1894
+ "policy/approxkl_avg": 0.002973160706460476,
1895
+ "policy/clipfrac_avg": 0.02556038275361061,
1896
+ "policy/entropy_avg": 1.844714879989624,
1897
+ "step": 90,
1898
+ "val/clipfrac_avg": 0.008994678035378456,
1899
+ "val/num_eos_tokens": 23411,
1900
+ "val/ratio": 1.0002460479736328,
1901
+ "val/ratio_var": 6.159632448543562e-06
1902
+ },
1903
+ {
1904
+ "episode": 46592,
1905
+ "epoch": 0.39917751884852637,
1906
+ "eps": 5,
1907
+ "loss/policy_avg": 0.0008165971376001835,
1908
+ "loss/value_avg": 0.21764297783374786,
1909
+ "lr": 1.9453125e-06,
1910
+ "objective/entropy": -51.29227066040039,
1911
+ "objective/kl": 35.93562316894531,
1912
+ "objective/non_score_reward": -1.796781301498413,
1913
+ "objective/rlhf_reward": 3.1337616443634033,
1914
+ "objective/scores": 4.930542945861816,
1915
+ "policy/approxkl_avg": 0.003850769717246294,
1916
+ "policy/clipfrac_avg": 0.025531645864248276,
1917
+ "policy/entropy_avg": 1.8263435363769531,
1918
+ "step": 91,
1919
+ "val/clipfrac_avg": 0.009705501608550549,
1920
+ "val/num_eos_tokens": 24653,
1921
+ "val/ratio": 1.0005769729614258,
1922
+ "val/ratio_var": 5.8719551816466264e-06
1923
+ },
1924
+ {
1925
+ "episode": 47104,
1926
+ "epoch": 0.40356408498971896,
1927
+ "eps": 5,
1928
+ "loss/policy_avg": -0.001339934766292572,
1929
+ "loss/value_avg": 0.22606465220451355,
1930
+ "lr": 1.93359375e-06,
1931
+ "objective/entropy": -52.69164276123047,
1932
+ "objective/kl": 34.70777130126953,
1933
+ "objective/non_score_reward": -1.7353887557983398,
1934
+ "objective/rlhf_reward": 3.1563901901245117,
1935
+ "objective/scores": 4.891778945922852,
1936
+ "policy/approxkl_avg": 0.0032106563448905945,
1937
+ "policy/clipfrac_avg": 0.025430142879486084,
1938
+ "policy/entropy_avg": 1.874713659286499,
1939
+ "step": 92,
1940
+ "val/clipfrac_avg": 0.01480791624635458,
1941
+ "val/num_eos_tokens": 24275,
1942
+ "val/ratio": 1.0002000331878662,
1943
+ "val/ratio_var": 5.490754574566381e-06
1944
+ },
1945
+ {
1946
+ "episode": 47616,
1947
+ "epoch": 0.40795065113091156,
1948
+ "eps": 5,
1949
+ "loss/policy_avg": 0.003710628254339099,
1950
+ "loss/value_avg": 0.17547845840454102,
1951
+ "lr": 1.921875e-06,
1952
+ "objective/entropy": -42.1109619140625,
1953
+ "objective/kl": 34.52031707763672,
1954
+ "objective/non_score_reward": -1.7260158061981201,
1955
+ "objective/rlhf_reward": 3.3212482929229736,
1956
+ "objective/scores": 5.047264099121094,
1957
+ "policy/approxkl_avg": 0.0023800067137926817,
1958
+ "policy/clipfrac_avg": 0.02241857722401619,
1959
+ "policy/entropy_avg": 1.6762866973876953,
1960
+ "step": 93,
1961
+ "val/clipfrac_avg": 0.013620151206851006,
1962
+ "val/num_eos_tokens": 21856,
1963
+ "val/ratio": 1.0000417232513428,
1964
+ "val/ratio_var": 6.8420908974076156e-06
1965
+ },
1966
+ {
1967
+ "episode": 48128,
1968
+ "epoch": 0.4123372172721042,
1969
+ "eps": 5,
1970
+ "loss/policy_avg": -0.003424527123570442,
1971
+ "loss/value_avg": 0.20876729488372803,
1972
+ "lr": 1.91015625e-06,
1973
+ "objective/entropy": -56.00605010986328,
1974
+ "objective/kl": 32.756813049316406,
1975
+ "objective/non_score_reward": -1.637840747833252,
1976
+ "objective/rlhf_reward": 3.344707489013672,
1977
+ "objective/scores": 4.982548236846924,
1978
+ "policy/approxkl_avg": 0.003393033053725958,
1979
+ "policy/clipfrac_avg": 0.024908877909183502,
1980
+ "policy/entropy_avg": 1.872148036956787,
1981
+ "step": 94,
1982
+ "val/clipfrac_avg": 0.014186862856149673,
1983
+ "val/num_eos_tokens": 22696,
1984
+ "val/ratio": 0.9998812079429626,
1985
+ "val/ratio_var": 6.546584245370468e-06
1986
+ },
1987
+ {
1988
+ "episode": 48640,
1989
+ "epoch": 0.4167237834132968,
1990
+ "eps": 5,
1991
+ "loss/policy_avg": 0.0005242200568318367,
1992
+ "loss/value_avg": 0.2137468159198761,
1993
+ "lr": 1.8984375e-06,
1994
+ "objective/entropy": -55.2701301574707,
1995
+ "objective/kl": 33.65769577026367,
1996
+ "objective/non_score_reward": -1.6828848123550415,
1997
+ "objective/rlhf_reward": 3.08296537399292,
1998
+ "objective/scores": 4.765850067138672,
1999
+ "policy/approxkl_avg": 0.002929868409410119,
2000
+ "policy/clipfrac_avg": 0.023102108389139175,
2001
+ "policy/entropy_avg": 1.8781063556671143,
2002
+ "step": 95,
2003
+ "val/clipfrac_avg": 0.013048425316810608,
2004
+ "val/num_eos_tokens": 23018,
2005
+ "val/ratio": 1.0001842975616455,
2006
+ "val/ratio_var": 4.114185685466509e-06
2007
+ },
2008
+ {
2009
+ "episode": 49152,
2010
+ "epoch": 0.4211103495544894,
2011
+ "eps": 5,
2012
+ "loss/policy_avg": 0.006132413633167744,
2013
+ "loss/value_avg": 0.21268463134765625,
2014
+ "lr": 1.8867187500000001e-06,
2015
+ "objective/entropy": -56.19593048095703,
2016
+ "objective/kl": 33.65611267089844,
2017
+ "objective/non_score_reward": -1.6828055381774902,
2018
+ "objective/rlhf_reward": 3.1711764335632324,
2019
+ "objective/scores": 4.853981971740723,
2020
+ "policy/approxkl_avg": 0.00293334829621017,
2021
+ "policy/clipfrac_avg": 0.024867385625839233,
2022
+ "policy/entropy_avg": 1.8930773735046387,
2023
+ "step": 96,
2024
+ "val/clipfrac_avg": 0.015137835405766964,
2025
+ "val/num_eos_tokens": 23050,
2026
+ "val/ratio": 1.000009536743164,
2027
+ "val/ratio_var": 4.815322427020874e-06
2028
+ },
2029
+ {
2030
+ "episode": 49664,
2031
+ "epoch": 0.425496915695682,
2032
+ "eps": 5,
2033
+ "loss/policy_avg": 0.002121499739587307,
2034
+ "loss/value_avg": 0.19983580708503723,
2035
+ "lr": 1.875e-06,
2036
+ "objective/entropy": -57.503963470458984,
2037
+ "objective/kl": 33.145103454589844,
2038
+ "objective/non_score_reward": -1.657254934310913,
2039
+ "objective/rlhf_reward": 3.171117067337036,
2040
+ "objective/scores": 4.828372001647949,
2041
+ "policy/approxkl_avg": 0.002225325908511877,
2042
+ "policy/clipfrac_avg": 0.022431831806898117,
2043
+ "policy/entropy_avg": 1.919158935546875,
2044
+ "step": 97,
2045
+ "val/clipfrac_avg": 0.008227003738284111,
2046
+ "val/num_eos_tokens": 23071,
2047
+ "val/ratio": 0.9998757839202881,
2048
+ "val/ratio_var": 3.6738720154971816e-06
2049
+ },
2050
+ {
2051
+ "episode": 50176,
2052
+ "epoch": 0.4298834818368746,
2053
+ "eps": 5,
2054
+ "loss/policy_avg": 0.0034734588116407394,
2055
+ "loss/value_avg": 0.17848807573318481,
2056
+ "lr": 1.86328125e-06,
2057
+ "objective/entropy": -35.84461212158203,
2058
+ "objective/kl": 32.006187438964844,
2059
+ "objective/non_score_reward": -1.6003093719482422,
2060
+ "objective/rlhf_reward": 3.182417869567871,
2061
+ "objective/scores": 4.782727241516113,
2062
+ "policy/approxkl_avg": 0.004119177348911762,
2063
+ "policy/clipfrac_avg": 0.024367572739720345,
2064
+ "policy/entropy_avg": 1.6725337505340576,
2065
+ "step": 98,
2066
+ "val/clipfrac_avg": 0.012022973969578743,
2067
+ "val/num_eos_tokens": 19432,
2068
+ "val/ratio": 0.9999524354934692,
2069
+ "val/ratio_var": 6.887255949550308e-06
2070
+ },
2071
+ {
2072
+ "episode": 50688,
2073
+ "epoch": 0.4342700479780672,
2074
+ "eps": 5,
2075
+ "loss/policy_avg": 0.001079482026398182,
2076
+ "loss/value_avg": 0.17016559839248657,
2077
+ "lr": 1.8515625000000001e-06,
2078
+ "objective/entropy": -38.25457000732422,
2079
+ "objective/kl": 32.945953369140625,
2080
+ "objective/non_score_reward": -1.6472976207733154,
2081
+ "objective/rlhf_reward": 3.248924970626831,
2082
+ "objective/scores": 4.8962225914001465,
2083
+ "policy/approxkl_avg": 0.0032758493907749653,
2084
+ "policy/clipfrac_avg": 0.023456059396266937,
2085
+ "policy/entropy_avg": 1.7023566961288452,
2086
+ "step": 99,
2087
+ "val/clipfrac_avg": 0.016986040398478508,
2088
+ "val/num_eos_tokens": 20758,
2089
+ "val/ratio": 1.0001511573791504,
2090
+ "val/ratio_var": 7.0640057856508065e-06
2091
+ },
2092
+ {
2093
+ "episode": 51200,
2094
+ "epoch": 0.43865661411925977,
2095
+ "eps": 5,
2096
+ "loss/policy_avg": 0.002577817067503929,
2097
+ "loss/value_avg": 0.18741527199745178,
2098
+ "lr": 1.83984375e-06,
2099
+ "objective/entropy": -56.75590896606445,
2100
+ "objective/kl": 33.12295913696289,
2101
+ "objective/non_score_reward": -1.656147837638855,
2102
+ "objective/rlhf_reward": 3.1702961921691895,
2103
+ "objective/scores": 4.826444149017334,
2104
+ "policy/approxkl_avg": 0.0025197030045092106,
2105
+ "policy/clipfrac_avg": 0.023469911888241768,
2106
+ "policy/entropy_avg": 1.935408592224121,
2107
+ "step": 100,
2108
+ "val/clipfrac_avg": 0.00803416594862938,
2109
+ "val/num_eos_tokens": 22328,
2110
+ "val/ratio": 1.0003705024719238,
2111
+ "val/ratio_var": 5.774788405688014e-06
2112
+ },
2113
+ {
2114
+ "episode": 51712,
2115
+ "epoch": 0.44304318026045236,
2116
+ "eps": 5,
2117
+ "loss/policy_avg": 0.005681059788912535,
2118
+ "loss/value_avg": 0.18603692948818207,
2119
+ "lr": 1.828125e-06,
2120
+ "objective/entropy": -56.08619689941406,
2121
+ "objective/kl": 31.718780517578125,
2122
+ "objective/non_score_reward": -1.5859390497207642,
2123
+ "objective/rlhf_reward": 3.3355307579040527,
2124
+ "objective/scores": 4.921469688415527,
2125
+ "policy/approxkl_avg": 0.0033193572890013456,
2126
+ "policy/clipfrac_avg": 0.022443672642111778,
2127
+ "policy/entropy_avg": 1.9039394855499268,
2128
+ "step": 101,
2129
+ "val/clipfrac_avg": 0.01020752266049385,
2130
+ "val/num_eos_tokens": 20720,
2131
+ "val/ratio": 0.9998416900634766,
2132
+ "val/ratio_var": 8.256325600086711e-06
2133
+ },
2134
+ {
2135
+ "episode": 52224,
2136
+ "epoch": 0.44742974640164496,
2137
+ "eps": 5,
2138
+ "loss/policy_avg": 0.0005138497799634933,
2139
+ "loss/value_avg": 0.19289636611938477,
2140
+ "lr": 1.81640625e-06,
2141
+ "objective/entropy": -55.88417053222656,
2142
+ "objective/kl": 33.3431510925293,
2143
+ "objective/non_score_reward": -1.6671576499938965,
2144
+ "objective/rlhf_reward": 3.1966233253479004,
2145
+ "objective/scores": 4.863780975341797,
2146
+ "policy/approxkl_avg": 0.002363540232181549,
2147
+ "policy/clipfrac_avg": 0.022522956132888794,
2148
+ "policy/entropy_avg": 1.914175271987915,
2149
+ "step": 102,
2150
+ "val/clipfrac_avg": 0.01081082969903946,
2151
+ "val/num_eos_tokens": 19440,
2152
+ "val/ratio": 0.9995956420898438,
2153
+ "val/ratio_var": 3.734937536137295e-06
2154
+ },
2155
+ {
2156
+ "episode": 52736,
2157
+ "epoch": 0.45181631254283755,
2158
+ "eps": 5,
2159
+ "loss/policy_avg": 0.004322280641645193,
2160
+ "loss/value_avg": 0.19218376278877258,
2161
+ "lr": 1.8046875e-06,
2162
+ "objective/entropy": -55.414825439453125,
2163
+ "objective/kl": 32.86573791503906,
2164
+ "objective/non_score_reward": -1.6432868242263794,
2165
+ "objective/rlhf_reward": 3.255722999572754,
2166
+ "objective/scores": 4.899009704589844,
2167
+ "policy/approxkl_avg": 0.0021291770972311497,
2168
+ "policy/clipfrac_avg": 0.023938871920108795,
2169
+ "policy/entropy_avg": 1.886290431022644,
2170
+ "step": 103,
2171
+ "val/clipfrac_avg": 0.012509731575846672,
2172
+ "val/num_eos_tokens": 22684,
2173
+ "val/ratio": 1.0003352165222168,
2174
+ "val/ratio_var": 5.712623078579782e-06
2175
+ },
2176
+ {
2177
+ "episode": 53248,
2178
+ "epoch": 0.45620287868403014,
2179
+ "eps": 5,
2180
+ "loss/policy_avg": 0.005551911890506744,
2181
+ "loss/value_avg": 0.17338958382606506,
2182
+ "lr": 1.79296875e-06,
2183
+ "objective/entropy": -40.899742126464844,
2184
+ "objective/kl": 33.91618347167969,
2185
+ "objective/non_score_reward": -1.6958091259002686,
2186
+ "objective/rlhf_reward": 3.210798501968384,
2187
+ "objective/scores": 4.906607627868652,
2188
+ "policy/approxkl_avg": 0.002731644082814455,
2189
+ "policy/clipfrac_avg": 0.022299444302916527,
2190
+ "policy/entropy_avg": 1.7232073545455933,
2191
+ "step": 104,
2192
+ "val/clipfrac_avg": 0.011527610011398792,
2193
+ "val/num_eos_tokens": 20397,
2194
+ "val/ratio": 0.9999278783798218,
2195
+ "val/ratio_var": 4.719611297332449e-06
2196
+ }
2197
+ ],
2198
+ "logging_steps": 100,
2199
+ "max_steps": 256,
2200
+ "num_input_tokens_seen": 0,
2201
+ "num_train_epochs": 1.122960932145305,
2202
+ "save_steps": 52,
2203
+ "stateful_callbacks": {
2204
+ "TrainerControl": {
2205
+ "args": {
2206
+ "should_epoch_stop": false,
2207
+ "should_evaluate": false,
2208
+ "should_log": false,
2209
+ "should_save": true,
2210
+ "should_training_stop": false
2211
+ },
2212
+ "attributes": {}
2213
+ }
2214
+ },
2215
+ "total_flos": 0,
2216
+ "train_batch_size": null,
2217
+ "trial_name": null,
2218
+ "trial_params": null
2219
+ }
checkpoint-104/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ebe64b3cb4195fbfd7e38621ff998e75d530687d86d07fc80f878142ab7f178
3
+ size 6328
checkpoint-156/config.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "GPTNeoXForCausalLM"
4
+ ],
5
+ "attention_bias": true,
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 0,
8
+ "classifier_dropout": 0.1,
9
+ "eos_token_id": 0,
10
+ "hidden_act": "gelu",
11
+ "hidden_dropout": 0.0,
12
+ "hidden_size": 1024,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 4096,
15
+ "layer_norm_eps": 1e-05,
16
+ "max_position_embeddings": 2048,
17
+ "model_type": "gpt_neox",
18
+ "num_attention_heads": 16,
19
+ "num_hidden_layers": 24,
20
+ "partial_rotary_factor": 0.25,
21
+ "rope_scaling": null,
22
+ "rope_theta": 10000,
23
+ "rotary_emb_base": 10000,
24
+ "rotary_pct": 0.25,
25
+ "tie_word_embeddings": false,
26
+ "torch_dtype": "float32",
27
+ "transformers_version": "4.50.3",
28
+ "use_cache": true,
29
+ "use_parallel_residual": true,
30
+ "vocab_size": 50304
31
+ }
checkpoint-156/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 0,
4
+ "eos_token_id": 0,
5
+ "transformers_version": "4.50.3"
6
+ }
checkpoint-156/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5a72b0fa0fdcaa42f4a7dfc1211da9f7b9169928ee309ca0c33c6eac21bf0875
3
+ size 1621370224
checkpoint-156/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c6ab2e4bcf164990e7b6b66819a4939eea7292054d4ea47931853282cfc4fd3
3
+ size 6073758958
checkpoint-156/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8bd5b442c30ba906a8070b3f29d1c7f32e39419d234e58e0ebb561e1890fda7
3
+ size 14308
checkpoint-156/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:77149524653640dd782e6e0575c6ebd8bd533b1341d34905209b604788ff862c
3
+ size 1064
checkpoint-156/special_tokens_map.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<|endoftext|>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "<|endoftext|>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": {
17
+ "content": "<|padding|>",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ },
23
+ "unk_token": {
24
+ "content": "<|endoftext|>",
25
+ "lstrip": false,
26
+ "normalized": false,
27
+ "rstrip": false,
28
+ "single_word": false
29
+ }
30
+ }
checkpoint-156/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-156/tokenizer_config.json ADDED
@@ -0,0 +1,220 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_eos_token": false,
4
+ "add_prefix_space": false,
5
+ "added_tokens_decoder": {
6
+ "0": {
7
+ "content": "<|endoftext|>",
8
+ "lstrip": false,
9
+ "normalized": false,
10
+ "rstrip": false,
11
+ "single_word": false,
12
+ "special": true
13
+ },
14
+ "1": {
15
+ "content": "<|padding|>",
16
+ "lstrip": false,
17
+ "normalized": false,
18
+ "rstrip": false,
19
+ "single_word": false,
20
+ "special": true
21
+ },
22
+ "50254": {
23
+ "content": " ",
24
+ "lstrip": false,
25
+ "normalized": true,
26
+ "rstrip": false,
27
+ "single_word": false,
28
+ "special": false
29
+ },
30
+ "50255": {
31
+ "content": " ",
32
+ "lstrip": false,
33
+ "normalized": true,
34
+ "rstrip": false,
35
+ "single_word": false,
36
+ "special": false
37
+ },
38
+ "50256": {
39
+ "content": " ",
40
+ "lstrip": false,
41
+ "normalized": true,
42
+ "rstrip": false,
43
+ "single_word": false,
44
+ "special": false
45
+ },
46
+ "50257": {
47
+ "content": " ",
48
+ "lstrip": false,
49
+ "normalized": true,
50
+ "rstrip": false,
51
+ "single_word": false,
52
+ "special": false
53
+ },
54
+ "50258": {
55
+ "content": " ",
56
+ "lstrip": false,
57
+ "normalized": true,
58
+ "rstrip": false,
59
+ "single_word": false,
60
+ "special": false
61
+ },
62
+ "50259": {
63
+ "content": " ",
64
+ "lstrip": false,
65
+ "normalized": true,
66
+ "rstrip": false,
67
+ "single_word": false,
68
+ "special": false
69
+ },
70
+ "50260": {
71
+ "content": " ",
72
+ "lstrip": false,
73
+ "normalized": true,
74
+ "rstrip": false,
75
+ "single_word": false,
76
+ "special": false
77
+ },
78
+ "50261": {
79
+ "content": " ",
80
+ "lstrip": false,
81
+ "normalized": true,
82
+ "rstrip": false,
83
+ "single_word": false,
84
+ "special": false
85
+ },
86
+ "50262": {
87
+ "content": " ",
88
+ "lstrip": false,
89
+ "normalized": true,
90
+ "rstrip": false,
91
+ "single_word": false,
92
+ "special": false
93
+ },
94
+ "50263": {
95
+ "content": " ",
96
+ "lstrip": false,
97
+ "normalized": true,
98
+ "rstrip": false,
99
+ "single_word": false,
100
+ "special": false
101
+ },
102
+ "50264": {
103
+ "content": " ",
104
+ "lstrip": false,
105
+ "normalized": true,
106
+ "rstrip": false,
107
+ "single_word": false,
108
+ "special": false
109
+ },
110
+ "50265": {
111
+ "content": " ",
112
+ "lstrip": false,
113
+ "normalized": true,
114
+ "rstrip": false,
115
+ "single_word": false,
116
+ "special": false
117
+ },
118
+ "50266": {
119
+ "content": " ",
120
+ "lstrip": false,
121
+ "normalized": true,
122
+ "rstrip": false,
123
+ "single_word": false,
124
+ "special": false
125
+ },
126
+ "50267": {
127
+ "content": " ",
128
+ "lstrip": false,
129
+ "normalized": true,
130
+ "rstrip": false,
131
+ "single_word": false,
132
+ "special": false
133
+ },
134
+ "50268": {
135
+ "content": " ",
136
+ "lstrip": false,
137
+ "normalized": true,
138
+ "rstrip": false,
139
+ "single_word": false,
140
+ "special": false
141
+ },
142
+ "50269": {
143
+ "content": " ",
144
+ "lstrip": false,
145
+ "normalized": true,
146
+ "rstrip": false,
147
+ "single_word": false,
148
+ "special": false
149
+ },
150
+ "50270": {
151
+ "content": " ",
152
+ "lstrip": false,
153
+ "normalized": true,
154
+ "rstrip": false,
155
+ "single_word": false,
156
+ "special": false
157
+ },
158
+ "50271": {
159
+ "content": " ",
160
+ "lstrip": false,
161
+ "normalized": true,
162
+ "rstrip": false,
163
+ "single_word": false,
164
+ "special": false
165
+ },
166
+ "50272": {
167
+ "content": " ",
168
+ "lstrip": false,
169
+ "normalized": true,
170
+ "rstrip": false,
171
+ "single_word": false,
172
+ "special": false
173
+ },
174
+ "50273": {
175
+ "content": " ",
176
+ "lstrip": false,
177
+ "normalized": true,
178
+ "rstrip": false,
179
+ "single_word": false,
180
+ "special": false
181
+ },
182
+ "50274": {
183
+ "content": " ",
184
+ "lstrip": false,
185
+ "normalized": true,
186
+ "rstrip": false,
187
+ "single_word": false,
188
+ "special": false
189
+ },
190
+ "50275": {
191
+ "content": " ",
192
+ "lstrip": false,
193
+ "normalized": true,
194
+ "rstrip": false,
195
+ "single_word": false,
196
+ "special": false
197
+ },
198
+ "50276": {
199
+ "content": " ",
200
+ "lstrip": false,
201
+ "normalized": true,
202
+ "rstrip": false,
203
+ "single_word": false,
204
+ "special": false
205
+ }
206
+ },
207
+ "bos_token": "<|endoftext|>",
208
+ "clean_up_tokenization_spaces": true,
209
+ "eos_token": "<|endoftext|>",
210
+ "extra_special_tokens": {},
211
+ "max_length": 580,
212
+ "model_max_length": 1000000000000000019884624838656,
213
+ "pad_token": "<|padding|>",
214
+ "padding_side": "left",
215
+ "stride": 0,
216
+ "tokenizer_class": "GPTNeoXTokenizer",
217
+ "truncation_side": "right",
218
+ "truncation_strategy": "longest_first",
219
+ "unk_token": "<|endoftext|>"
220
+ }
checkpoint-156/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-156/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ebe64b3cb4195fbfd7e38621ff998e75d530687d86d07fc80f878142ab7f178
3
+ size 6328
checkpoint-208/config.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "GPTNeoXForCausalLM"
4
+ ],
5
+ "attention_bias": true,
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 0,
8
+ "classifier_dropout": 0.1,
9
+ "eos_token_id": 0,
10
+ "hidden_act": "gelu",
11
+ "hidden_dropout": 0.0,
12
+ "hidden_size": 1024,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 4096,
15
+ "layer_norm_eps": 1e-05,
16
+ "max_position_embeddings": 2048,
17
+ "model_type": "gpt_neox",
18
+ "num_attention_heads": 16,
19
+ "num_hidden_layers": 24,
20
+ "partial_rotary_factor": 0.25,
21
+ "rope_scaling": null,
22
+ "rope_theta": 10000,
23
+ "rotary_emb_base": 10000,
24
+ "rotary_pct": 0.25,
25
+ "tie_word_embeddings": false,
26
+ "torch_dtype": "float32",
27
+ "transformers_version": "4.50.3",
28
+ "use_cache": true,
29
+ "use_parallel_residual": true,
30
+ "vocab_size": 50304
31
+ }
checkpoint-208/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 0,
4
+ "eos_token_id": 0,
5
+ "transformers_version": "4.50.3"
6
+ }
checkpoint-208/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ffcb920ea438d2067db0bf56c5a2b8765ce46e5fe6e0854395cc0d7f72530b47
3
+ size 1621370224
checkpoint-208/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe9427c3990d89300c835eda0e0e08cbedc52af907469838663b92f41131ae18
3
+ size 6073758958
checkpoint-208/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fafbd2caa47c3117f4754c693175f2147166136091d26331f79fabbeec9651ef
3
+ size 14308
checkpoint-208/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a732a23fbb571460c6a96bddae3100b1253bf8fc063fec70759eed2b43e8a60
3
+ size 1064
checkpoint-208/special_tokens_map.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<|endoftext|>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "<|endoftext|>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": {
17
+ "content": "<|padding|>",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ },
23
+ "unk_token": {
24
+ "content": "<|endoftext|>",
25
+ "lstrip": false,
26
+ "normalized": false,
27
+ "rstrip": false,
28
+ "single_word": false
29
+ }
30
+ }
checkpoint-208/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-208/tokenizer_config.json ADDED
@@ -0,0 +1,220 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_eos_token": false,
4
+ "add_prefix_space": false,
5
+ "added_tokens_decoder": {
6
+ "0": {
7
+ "content": "<|endoftext|>",
8
+ "lstrip": false,
9
+ "normalized": false,
10
+ "rstrip": false,
11
+ "single_word": false,
12
+ "special": true
13
+ },
14
+ "1": {
15
+ "content": "<|padding|>",
16
+ "lstrip": false,
17
+ "normalized": false,
18
+ "rstrip": false,
19
+ "single_word": false,
20
+ "special": true
21
+ },
22
+ "50254": {
23
+ "content": " ",
24
+ "lstrip": false,
25
+ "normalized": true,
26
+ "rstrip": false,
27
+ "single_word": false,
28
+ "special": false
29
+ },
30
+ "50255": {
31
+ "content": " ",
32
+ "lstrip": false,
33
+ "normalized": true,
34
+ "rstrip": false,
35
+ "single_word": false,
36
+ "special": false
37
+ },
38
+ "50256": {
39
+ "content": " ",
40
+ "lstrip": false,
41
+ "normalized": true,
42
+ "rstrip": false,
43
+ "single_word": false,
44
+ "special": false
45
+ },
46
+ "50257": {
47
+ "content": " ",
48
+ "lstrip": false,
49
+ "normalized": true,
50
+ "rstrip": false,
51
+ "single_word": false,
52
+ "special": false
53
+ },
54
+ "50258": {
55
+ "content": " ",
56
+ "lstrip": false,
57
+ "normalized": true,
58
+ "rstrip": false,
59
+ "single_word": false,
60
+ "special": false
61
+ },
62
+ "50259": {
63
+ "content": " ",
64
+ "lstrip": false,
65
+ "normalized": true,
66
+ "rstrip": false,
67
+ "single_word": false,
68
+ "special": false
69
+ },
70
+ "50260": {
71
+ "content": " ",
72
+ "lstrip": false,
73
+ "normalized": true,
74
+ "rstrip": false,
75
+ "single_word": false,
76
+ "special": false
77
+ },
78
+ "50261": {
79
+ "content": " ",
80
+ "lstrip": false,
81
+ "normalized": true,
82
+ "rstrip": false,
83
+ "single_word": false,
84
+ "special": false
85
+ },
86
+ "50262": {
87
+ "content": " ",
88
+ "lstrip": false,
89
+ "normalized": true,
90
+ "rstrip": false,
91
+ "single_word": false,
92
+ "special": false
93
+ },
94
+ "50263": {
95
+ "content": " ",
96
+ "lstrip": false,
97
+ "normalized": true,
98
+ "rstrip": false,
99
+ "single_word": false,
100
+ "special": false
101
+ },
102
+ "50264": {
103
+ "content": " ",
104
+ "lstrip": false,
105
+ "normalized": true,
106
+ "rstrip": false,
107
+ "single_word": false,
108
+ "special": false
109
+ },
110
+ "50265": {
111
+ "content": " ",
112
+ "lstrip": false,
113
+ "normalized": true,
114
+ "rstrip": false,
115
+ "single_word": false,
116
+ "special": false
117
+ },
118
+ "50266": {
119
+ "content": " ",
120
+ "lstrip": false,
121
+ "normalized": true,
122
+ "rstrip": false,
123
+ "single_word": false,
124
+ "special": false
125
+ },
126
+ "50267": {
127
+ "content": " ",
128
+ "lstrip": false,
129
+ "normalized": true,
130
+ "rstrip": false,
131
+ "single_word": false,
132
+ "special": false
133
+ },
134
+ "50268": {
135
+ "content": " ",
136
+ "lstrip": false,
137
+ "normalized": true,
138
+ "rstrip": false,
139
+ "single_word": false,
140
+ "special": false
141
+ },
142
+ "50269": {
143
+ "content": " ",
144
+ "lstrip": false,
145
+ "normalized": true,
146
+ "rstrip": false,
147
+ "single_word": false,
148
+ "special": false
149
+ },
150
+ "50270": {
151
+ "content": " ",
152
+ "lstrip": false,
153
+ "normalized": true,
154
+ "rstrip": false,
155
+ "single_word": false,
156
+ "special": false
157
+ },
158
+ "50271": {
159
+ "content": " ",
160
+ "lstrip": false,
161
+ "normalized": true,
162
+ "rstrip": false,
163
+ "single_word": false,
164
+ "special": false
165
+ },
166
+ "50272": {
167
+ "content": " ",
168
+ "lstrip": false,
169
+ "normalized": true,
170
+ "rstrip": false,
171
+ "single_word": false,
172
+ "special": false
173
+ },
174
+ "50273": {
175
+ "content": " ",
176
+ "lstrip": false,
177
+ "normalized": true,
178
+ "rstrip": false,
179
+ "single_word": false,
180
+ "special": false
181
+ },
182
+ "50274": {
183
+ "content": " ",
184
+ "lstrip": false,
185
+ "normalized": true,
186
+ "rstrip": false,
187
+ "single_word": false,
188
+ "special": false
189
+ },
190
+ "50275": {
191
+ "content": " ",
192
+ "lstrip": false,
193
+ "normalized": true,
194
+ "rstrip": false,
195
+ "single_word": false,
196
+ "special": false
197
+ },
198
+ "50276": {
199
+ "content": " ",
200
+ "lstrip": false,
201
+ "normalized": true,
202
+ "rstrip": false,
203
+ "single_word": false,
204
+ "special": false
205
+ }
206
+ },
207
+ "bos_token": "<|endoftext|>",
208
+ "clean_up_tokenization_spaces": true,
209
+ "eos_token": "<|endoftext|>",
210
+ "extra_special_tokens": {},
211
+ "max_length": 580,
212
+ "model_max_length": 1000000000000000019884624838656,
213
+ "pad_token": "<|padding|>",
214
+ "padding_side": "left",
215
+ "stride": 0,
216
+ "tokenizer_class": "GPTNeoXTokenizer",
217
+ "truncation_side": "right",
218
+ "truncation_strategy": "longest_first",
219
+ "unk_token": "<|endoftext|>"
220
+ }
checkpoint-208/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-208/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ebe64b3cb4195fbfd7e38621ff998e75d530687d86d07fc80f878142ab7f178
3
+ size 6328
checkpoint-256/config.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "GPTNeoXForCausalLM"
4
+ ],
5
+ "attention_bias": true,
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 0,
8
+ "classifier_dropout": 0.1,
9
+ "eos_token_id": 0,
10
+ "hidden_act": "gelu",
11
+ "hidden_dropout": 0.0,
12
+ "hidden_size": 1024,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 4096,
15
+ "layer_norm_eps": 1e-05,
16
+ "max_position_embeddings": 2048,
17
+ "model_type": "gpt_neox",
18
+ "num_attention_heads": 16,
19
+ "num_hidden_layers": 24,
20
+ "partial_rotary_factor": 0.25,
21
+ "rope_scaling": null,
22
+ "rope_theta": 10000,
23
+ "rotary_emb_base": 10000,
24
+ "rotary_pct": 0.25,
25
+ "tie_word_embeddings": false,
26
+ "torch_dtype": "float32",
27
+ "transformers_version": "4.50.3",
28
+ "use_cache": true,
29
+ "use_parallel_residual": true,
30
+ "vocab_size": 50304
31
+ }
checkpoint-256/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 0,
4
+ "eos_token_id": 0,
5
+ "transformers_version": "4.50.3"
6
+ }
checkpoint-256/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f493a89ee334f5883dfb087c13087b27e612c0702882afe91888b9957a41114e
3
+ size 1621370224
checkpoint-256/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28e23428bdf511d56c4f523628f09b7fc33b606103a18eb77fce3609b5b21e03
3
+ size 6073758958
checkpoint-256/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:920808f149d546fe6352dff53da4551371b5339c61664e7fb064bb6f91e1f18b
3
+ size 14308
checkpoint-256/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87ab05b39f525a9c92e5f8665d5a3111e407a81132e61964425d0d2854481513
3
+ size 1064
checkpoint-256/special_tokens_map.json ADDED
@@ -0,0 +1,30 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token": {
3
+ "content": "<|endoftext|>",
4
+ "lstrip": false,
5
+ "normalized": false,
6
+ "rstrip": false,
7
+ "single_word": false
8
+ },
9
+ "eos_token": {
10
+ "content": "<|endoftext|>",
11
+ "lstrip": false,
12
+ "normalized": false,
13
+ "rstrip": false,
14
+ "single_word": false
15
+ },
16
+ "pad_token": {
17
+ "content": "<|padding|>",
18
+ "lstrip": false,
19
+ "normalized": false,
20
+ "rstrip": false,
21
+ "single_word": false
22
+ },
23
+ "unk_token": {
24
+ "content": "<|endoftext|>",
25
+ "lstrip": false,
26
+ "normalized": false,
27
+ "rstrip": false,
28
+ "single_word": false
29
+ }
30
+ }
checkpoint-256/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-256/tokenizer_config.json ADDED
@@ -0,0 +1,220 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_eos_token": false,
4
+ "add_prefix_space": false,
5
+ "added_tokens_decoder": {
6
+ "0": {
7
+ "content": "<|endoftext|>",
8
+ "lstrip": false,
9
+ "normalized": false,
10
+ "rstrip": false,
11
+ "single_word": false,
12
+ "special": true
13
+ },
14
+ "1": {
15
+ "content": "<|padding|>",
16
+ "lstrip": false,
17
+ "normalized": false,
18
+ "rstrip": false,
19
+ "single_word": false,
20
+ "special": true
21
+ },
22
+ "50254": {
23
+ "content": " ",
24
+ "lstrip": false,
25
+ "normalized": true,
26
+ "rstrip": false,
27
+ "single_word": false,
28
+ "special": false
29
+ },
30
+ "50255": {
31
+ "content": " ",
32
+ "lstrip": false,
33
+ "normalized": true,
34
+ "rstrip": false,
35
+ "single_word": false,
36
+ "special": false
37
+ },
38
+ "50256": {
39
+ "content": " ",
40
+ "lstrip": false,
41
+ "normalized": true,
42
+ "rstrip": false,
43
+ "single_word": false,
44
+ "special": false
45
+ },
46
+ "50257": {
47
+ "content": " ",
48
+ "lstrip": false,
49
+ "normalized": true,
50
+ "rstrip": false,
51
+ "single_word": false,
52
+ "special": false
53
+ },
54
+ "50258": {
55
+ "content": " ",
56
+ "lstrip": false,
57
+ "normalized": true,
58
+ "rstrip": false,
59
+ "single_word": false,
60
+ "special": false
61
+ },
62
+ "50259": {
63
+ "content": " ",
64
+ "lstrip": false,
65
+ "normalized": true,
66
+ "rstrip": false,
67
+ "single_word": false,
68
+ "special": false
69
+ },
70
+ "50260": {
71
+ "content": " ",
72
+ "lstrip": false,
73
+ "normalized": true,
74
+ "rstrip": false,
75
+ "single_word": false,
76
+ "special": false
77
+ },
78
+ "50261": {
79
+ "content": " ",
80
+ "lstrip": false,
81
+ "normalized": true,
82
+ "rstrip": false,
83
+ "single_word": false,
84
+ "special": false
85
+ },
86
+ "50262": {
87
+ "content": " ",
88
+ "lstrip": false,
89
+ "normalized": true,
90
+ "rstrip": false,
91
+ "single_word": false,
92
+ "special": false
93
+ },
94
+ "50263": {
95
+ "content": " ",
96
+ "lstrip": false,
97
+ "normalized": true,
98
+ "rstrip": false,
99
+ "single_word": false,
100
+ "special": false
101
+ },
102
+ "50264": {
103
+ "content": " ",
104
+ "lstrip": false,
105
+ "normalized": true,
106
+ "rstrip": false,
107
+ "single_word": false,
108
+ "special": false
109
+ },
110
+ "50265": {
111
+ "content": " ",
112
+ "lstrip": false,
113
+ "normalized": true,
114
+ "rstrip": false,
115
+ "single_word": false,
116
+ "special": false
117
+ },
118
+ "50266": {
119
+ "content": " ",
120
+ "lstrip": false,
121
+ "normalized": true,
122
+ "rstrip": false,
123
+ "single_word": false,
124
+ "special": false
125
+ },
126
+ "50267": {
127
+ "content": " ",
128
+ "lstrip": false,
129
+ "normalized": true,
130
+ "rstrip": false,
131
+ "single_word": false,
132
+ "special": false
133
+ },
134
+ "50268": {
135
+ "content": " ",
136
+ "lstrip": false,
137
+ "normalized": true,
138
+ "rstrip": false,
139
+ "single_word": false,
140
+ "special": false
141
+ },
142
+ "50269": {
143
+ "content": " ",
144
+ "lstrip": false,
145
+ "normalized": true,
146
+ "rstrip": false,
147
+ "single_word": false,
148
+ "special": false
149
+ },
150
+ "50270": {
151
+ "content": " ",
152
+ "lstrip": false,
153
+ "normalized": true,
154
+ "rstrip": false,
155
+ "single_word": false,
156
+ "special": false
157
+ },
158
+ "50271": {
159
+ "content": " ",
160
+ "lstrip": false,
161
+ "normalized": true,
162
+ "rstrip": false,
163
+ "single_word": false,
164
+ "special": false
165
+ },
166
+ "50272": {
167
+ "content": " ",
168
+ "lstrip": false,
169
+ "normalized": true,
170
+ "rstrip": false,
171
+ "single_word": false,
172
+ "special": false
173
+ },
174
+ "50273": {
175
+ "content": " ",
176
+ "lstrip": false,
177
+ "normalized": true,
178
+ "rstrip": false,
179
+ "single_word": false,
180
+ "special": false
181
+ },
182
+ "50274": {
183
+ "content": " ",
184
+ "lstrip": false,
185
+ "normalized": true,
186
+ "rstrip": false,
187
+ "single_word": false,
188
+ "special": false
189
+ },
190
+ "50275": {
191
+ "content": " ",
192
+ "lstrip": false,
193
+ "normalized": true,
194
+ "rstrip": false,
195
+ "single_word": false,
196
+ "special": false
197
+ },
198
+ "50276": {
199
+ "content": " ",
200
+ "lstrip": false,
201
+ "normalized": true,
202
+ "rstrip": false,
203
+ "single_word": false,
204
+ "special": false
205
+ }
206
+ },
207
+ "bos_token": "<|endoftext|>",
208
+ "clean_up_tokenization_spaces": true,
209
+ "eos_token": "<|endoftext|>",
210
+ "extra_special_tokens": {},
211
+ "max_length": 580,
212
+ "model_max_length": 1000000000000000019884624838656,
213
+ "pad_token": "<|padding|>",
214
+ "padding_side": "left",
215
+ "stride": 0,
216
+ "tokenizer_class": "GPTNeoXTokenizer",
217
+ "truncation_side": "right",
218
+ "truncation_strategy": "longest_first",
219
+ "unk_token": "<|endoftext|>"
220
+ }
checkpoint-256/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-256/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ebe64b3cb4195fbfd7e38621ff998e75d530687d86d07fc80f878142ab7f178
3
+ size 6328
checkpoint-52/config.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "GPTNeoXForCausalLM"
4
+ ],
5
+ "attention_bias": true,
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 0,
8
+ "classifier_dropout": 0.1,
9
+ "eos_token_id": 0,
10
+ "hidden_act": "gelu",
11
+ "hidden_dropout": 0.0,
12
+ "hidden_size": 1024,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 4096,
15
+ "layer_norm_eps": 1e-05,
16
+ "max_position_embeddings": 2048,
17
+ "model_type": "gpt_neox",
18
+ "num_attention_heads": 16,
19
+ "num_hidden_layers": 24,
20
+ "partial_rotary_factor": 0.25,
21
+ "rope_scaling": null,
22
+ "rope_theta": 10000,
23
+ "rotary_emb_base": 10000,
24
+ "rotary_pct": 0.25,
25
+ "tie_word_embeddings": false,
26
+ "torch_dtype": "float32",
27
+ "transformers_version": "4.50.3",
28
+ "use_cache": true,
29
+ "use_parallel_residual": true,
30
+ "vocab_size": 50304
31
+ }
checkpoint-52/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 0,
4
+ "eos_token_id": 0,
5
+ "transformers_version": "4.50.3"
6
+ }
checkpoint-52/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f998ead29e2ea1f0fe3303ca421875e382b0a380da43192897d54c65076c5bed
3
+ size 1621370224
checkpoint-52/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a7b48e9746ebe8fdf8a057622f058eb0314dacb4e3baad239dabdd0fc0546550
3
+ size 6073758958
checkpoint-52/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c8ae0ea37d3b7d74786194b318f52f109f397e612f1568f8e74d6428ce7274d4
3
+ size 14244
checkpoint-52/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:15adc088a35512de96eaf76adc88eb47505d933e9ec793220480f6c500342492
3
+ size 1064