Muqeeth commited on
Commit
35d6da8
·
verified ·
1 Parent(s): 541612a

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 sha256:40dae372f5d833e80db65a19652d324b98bb714972dbb658722aa8065f9477ca
3
+ size 1621370224
checkpoint-104/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34092a148a916aee76e51774cc0c6d4db242fb0957ca8c345d3af21db2f57a93
3
+ size 6073758958
checkpoint-104/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:882a892b6a2fef9c9c150f6cf2873252a1c71ae49cdc593d70d019b550d79b18
3
+ size 14244
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": -0.0013769641518592834,
18
+ "loss/value_avg": 1.9423742294311523,
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": 4.956933975219727,
24
+ "objective/scores": 4.956933975219727,
25
+ "policy/approxkl_avg": 0.001525376457720995,
26
+ "policy/clipfrac_avg": 0.022334059700369835,
27
+ "policy/entropy_avg": 1.2976276874542236,
28
+ "step": 1,
29
+ "val/clipfrac_avg": 0.006576753221452236,
30
+ "val/num_eos_tokens": 13845,
31
+ "val/ratio": 0.9997893571853638,
32
+ "val/ratio_var": 4.300739419704769e-06
33
+ },
34
+ {
35
+ "episode": 1024,
36
+ "epoch": 0.008773132282385195,
37
+ "eps": 6,
38
+ "loss/policy_avg": 0.0031125713139772415,
39
+ "loss/value_avg": 1.3287988901138306,
40
+ "lr": 2.9882812500000002e-06,
41
+ "objective/entropy": 7.951052665710449,
42
+ "objective/kl": 0.7057143449783325,
43
+ "objective/non_score_reward": -0.035285718739032745,
44
+ "objective/rlhf_reward": 5.132782936096191,
45
+ "objective/scores": 5.168068885803223,
46
+ "policy/approxkl_avg": 0.0028971810825169086,
47
+ "policy/clipfrac_avg": 0.02630750462412834,
48
+ "policy/entropy_avg": 0.9777142405509949,
49
+ "step": 2,
50
+ "val/clipfrac_avg": 0.017064230516552925,
51
+ "val/num_eos_tokens": 8612,
52
+ "val/ratio": 1.000025749206543,
53
+ "val/ratio_var": 1.2729425179713871e-05
54
+ },
55
+ {
56
+ "episode": 1536,
57
+ "epoch": 0.013159698423577794,
58
+ "eps": 6,
59
+ "loss/policy_avg": 0.004569530487060547,
60
+ "loss/value_avg": 1.8551630973815918,
61
+ "lr": 2.9765625e-06,
62
+ "objective/entropy": 6.086846351623535,
63
+ "objective/kl": 1.7537505626678467,
64
+ "objective/non_score_reward": -0.08768752217292786,
65
+ "objective/rlhf_reward": 5.270642280578613,
66
+ "objective/scores": 5.358329772949219,
67
+ "policy/approxkl_avg": 0.0035523450933396816,
68
+ "policy/clipfrac_avg": 0.02881922945380211,
69
+ "policy/entropy_avg": 0.99476557970047,
70
+ "step": 3,
71
+ "val/clipfrac_avg": 0.023893361911177635,
72
+ "val/num_eos_tokens": 8606,
73
+ "val/ratio": 0.999504804611206,
74
+ "val/ratio_var": 2.043731728917919e-05
75
+ },
76
+ {
77
+ "episode": 2048,
78
+ "epoch": 0.01754626456477039,
79
+ "eps": 6,
80
+ "loss/policy_avg": 0.00686279870569706,
81
+ "loss/value_avg": 2.0762438774108887,
82
+ "lr": 2.96484375e-06,
83
+ "objective/entropy": -72.34222412109375,
84
+ "objective/kl": 2.7035207748413086,
85
+ "objective/non_score_reward": -0.13517603278160095,
86
+ "objective/rlhf_reward": 5.987885475158691,
87
+ "objective/scores": 6.123061656951904,
88
+ "policy/approxkl_avg": 0.0016497070901095867,
89
+ "policy/clipfrac_avg": 0.02908516302704811,
90
+ "policy/entropy_avg": 1.7031397819519043,
91
+ "step": 4,
92
+ "val/clipfrac_avg": 0.027677347883582115,
93
+ "val/num_eos_tokens": 14870,
94
+ "val/ratio": 1.000091552734375,
95
+ "val/ratio_var": 2.1516761989914812e-06
96
+ },
97
+ {
98
+ "episode": 2560,
99
+ "epoch": 0.02193283070596299,
100
+ "eps": 6,
101
+ "loss/policy_avg": 0.00263179000467062,
102
+ "loss/value_avg": 1.606541633605957,
103
+ "lr": 2.953125e-06,
104
+ "objective/entropy": -65.28475952148438,
105
+ "objective/kl": 4.556520462036133,
106
+ "objective/non_score_reward": -0.22782602906227112,
107
+ "objective/rlhf_reward": 6.3742218017578125,
108
+ "objective/scores": 6.602047920227051,
109
+ "policy/approxkl_avg": 0.0017431321321055293,
110
+ "policy/clipfrac_avg": 0.02620251104235649,
111
+ "policy/entropy_avg": 2.0425772666931152,
112
+ "step": 5,
113
+ "val/clipfrac_avg": 0.023111186921596527,
114
+ "val/num_eos_tokens": 13325,
115
+ "val/ratio": 0.9996423721313477,
116
+ "val/ratio_var": 2.722375029406976e-06
117
+ },
118
+ {
119
+ "episode": 3072,
120
+ "epoch": 0.026319396847155587,
121
+ "eps": 6,
122
+ "loss/policy_avg": 0.0040515996515750885,
123
+ "loss/value_avg": 1.1282833814620972,
124
+ "lr": 2.94140625e-06,
125
+ "objective/entropy": -56.7620964050293,
126
+ "objective/kl": 10.2549467086792,
127
+ "objective/non_score_reward": -0.5127473473548889,
128
+ "objective/rlhf_reward": 7.096510887145996,
129
+ "objective/scores": 7.60925817489624,
130
+ "policy/approxkl_avg": 0.002651135204359889,
131
+ "policy/clipfrac_avg": 0.02610427513718605,
132
+ "policy/entropy_avg": 2.47774600982666,
133
+ "step": 6,
134
+ "val/clipfrac_avg": 0.01816035807132721,
135
+ "val/num_eos_tokens": 20916,
136
+ "val/ratio": 0.9997857809066772,
137
+ "val/ratio_var": 3.7805486954312073e-06
138
+ },
139
+ {
140
+ "episode": 3584,
141
+ "epoch": 0.030705962988348184,
142
+ "eps": 6,
143
+ "loss/policy_avg": -0.0022932137362658978,
144
+ "loss/value_avg": 0.9223657846450806,
145
+ "lr": 2.9296875e-06,
146
+ "objective/entropy": -51.8396110534668,
147
+ "objective/kl": 15.692605972290039,
148
+ "objective/non_score_reward": -0.784630298614502,
149
+ "objective/rlhf_reward": 7.749014377593994,
150
+ "objective/scores": 8.533644676208496,
151
+ "policy/approxkl_avg": 0.002274099038913846,
152
+ "policy/clipfrac_avg": 0.025012236088514328,
153
+ "policy/entropy_avg": 2.5470399856567383,
154
+ "step": 7,
155
+ "val/clipfrac_avg": 0.018111344426870346,
156
+ "val/num_eos_tokens": 27788,
157
+ "val/ratio": 1.0004068613052368,
158
+ "val/ratio_var": 3.3165326840389753e-06
159
+ },
160
+ {
161
+ "episode": 4096,
162
+ "epoch": 0.03509252912954078,
163
+ "eps": 5,
164
+ "loss/policy_avg": -0.02864566072821617,
165
+ "loss/value_avg": 0.8586666584014893,
166
+ "lr": 2.91796875e-06,
167
+ "objective/entropy": -44.25556564331055,
168
+ "objective/kl": 18.9859619140625,
169
+ "objective/non_score_reward": -0.9492980241775513,
170
+ "objective/rlhf_reward": 7.566014289855957,
171
+ "objective/scores": 8.515312194824219,
172
+ "policy/approxkl_avg": 0.0024859202094376087,
173
+ "policy/clipfrac_avg": 0.02374047227203846,
174
+ "policy/entropy_avg": 2.4501914978027344,
175
+ "step": 8,
176
+ "val/clipfrac_avg": 0.012094689533114433,
177
+ "val/num_eos_tokens": 39697,
178
+ "val/ratio": 0.9997825622558594,
179
+ "val/ratio_var": 2.589023324617301e-06
180
+ },
181
+ {
182
+ "episode": 4608,
183
+ "epoch": 0.03947909527073338,
184
+ "eps": 5,
185
+ "loss/policy_avg": -0.03993895649909973,
186
+ "loss/value_avg": 0.8039427995681763,
187
+ "lr": 2.90625e-06,
188
+ "objective/entropy": -41.25428771972656,
189
+ "objective/kl": 21.300134658813477,
190
+ "objective/non_score_reward": -1.0650067329406738,
191
+ "objective/rlhf_reward": 7.658048152923584,
192
+ "objective/scores": 8.723054885864258,
193
+ "policy/approxkl_avg": 0.002434398978948593,
194
+ "policy/clipfrac_avg": 0.020647358149290085,
195
+ "policy/entropy_avg": 2.318600654602051,
196
+ "step": 9,
197
+ "val/clipfrac_avg": 0.012456808239221573,
198
+ "val/num_eos_tokens": 38279,
199
+ "val/ratio": 0.9994508028030396,
200
+ "val/ratio_var": 2.497590912753367e-06
201
+ },
202
+ {
203
+ "episode": 5120,
204
+ "epoch": 0.04386566141192598,
205
+ "eps": 5,
206
+ "loss/policy_avg": -0.038270700722932816,
207
+ "loss/value_avg": 0.7831208109855652,
208
+ "lr": 2.89453125e-06,
209
+ "objective/entropy": -40.375099182128906,
210
+ "objective/kl": 23.66585922241211,
211
+ "objective/non_score_reward": -1.1832929849624634,
212
+ "objective/rlhf_reward": 7.1149516105651855,
213
+ "objective/scores": 8.29824447631836,
214
+ "policy/approxkl_avg": 0.0023879953660070896,
215
+ "policy/clipfrac_avg": 0.02045682817697525,
216
+ "policy/entropy_avg": 2.171109199523926,
217
+ "step": 10,
218
+ "val/clipfrac_avg": 0.01572496071457863,
219
+ "val/num_eos_tokens": 36087,
220
+ "val/ratio": 0.9995651245117188,
221
+ "val/ratio_var": 4.97030487167649e-06
222
+ },
223
+ {
224
+ "episode": 5632,
225
+ "epoch": 0.048252227553118573,
226
+ "eps": 5,
227
+ "loss/policy_avg": -0.03474128991365433,
228
+ "loss/value_avg": 0.7656794786453247,
229
+ "lr": 2.8828125e-06,
230
+ "objective/entropy": -43.84745788574219,
231
+ "objective/kl": 24.729175567626953,
232
+ "objective/non_score_reward": -1.2364587783813477,
233
+ "objective/rlhf_reward": 7.341504096984863,
234
+ "objective/scores": 8.577962875366211,
235
+ "policy/approxkl_avg": 0.0027273856103420258,
236
+ "policy/clipfrac_avg": 0.022427301853895187,
237
+ "policy/entropy_avg": 2.208101272583008,
238
+ "step": 11,
239
+ "val/clipfrac_avg": 0.016871843487024307,
240
+ "val/num_eos_tokens": 36666,
241
+ "val/ratio": 0.9991950988769531,
242
+ "val/ratio_var": 4.898628048977116e-06
243
+ },
244
+ {
245
+ "episode": 6144,
246
+ "epoch": 0.052638793694311174,
247
+ "eps": 5,
248
+ "loss/policy_avg": -0.0249569620937109,
249
+ "loss/value_avg": 0.7158781290054321,
250
+ "lr": 2.87109375e-06,
251
+ "objective/entropy": -48.97076416015625,
252
+ "objective/kl": 25.502344131469727,
253
+ "objective/non_score_reward": -1.27511727809906,
254
+ "objective/rlhf_reward": 7.928722858428955,
255
+ "objective/scores": 9.203840255737305,
256
+ "policy/approxkl_avg": 0.0023700755555182695,
257
+ "policy/clipfrac_avg": 0.02416691742837429,
258
+ "policy/entropy_avg": 2.2255916595458984,
259
+ "step": 12,
260
+ "val/clipfrac_avg": 0.014078151434659958,
261
+ "val/num_eos_tokens": 37746,
262
+ "val/ratio": 0.9998907446861267,
263
+ "val/ratio_var": 5.1026213441218715e-06
264
+ },
265
+ {
266
+ "episode": 6656,
267
+ "epoch": 0.05702535983550377,
268
+ "eps": 5,
269
+ "loss/policy_avg": -0.018949007615447044,
270
+ "loss/value_avg": 0.7686400413513184,
271
+ "lr": 2.859375e-06,
272
+ "objective/entropy": -47.11943817138672,
273
+ "objective/kl": 27.999988555908203,
274
+ "objective/non_score_reward": -1.3999994993209839,
275
+ "objective/rlhf_reward": 7.812624931335449,
276
+ "objective/scores": 9.212624549865723,
277
+ "policy/approxkl_avg": 0.002762872725725174,
278
+ "policy/clipfrac_avg": 0.024748779833316803,
279
+ "policy/entropy_avg": 2.155306339263916,
280
+ "step": 13,
281
+ "val/clipfrac_avg": 0.015288694761693478,
282
+ "val/num_eos_tokens": 36580,
283
+ "val/ratio": 1.000169277191162,
284
+ "val/ratio_var": 5.652353593177395e-06
285
+ },
286
+ {
287
+ "episode": 7168,
288
+ "epoch": 0.06141192597669637,
289
+ "eps": 5,
290
+ "loss/policy_avg": -0.009994697757065296,
291
+ "loss/value_avg": 0.7283180952072144,
292
+ "lr": 2.84765625e-06,
293
+ "objective/entropy": -53.18806457519531,
294
+ "objective/kl": 27.922462463378906,
295
+ "objective/non_score_reward": -1.3961231708526611,
296
+ "objective/rlhf_reward": 8.238632202148438,
297
+ "objective/scores": 9.63475513458252,
298
+ "policy/approxkl_avg": 0.002478944603353739,
299
+ "policy/clipfrac_avg": 0.021354785189032555,
300
+ "policy/entropy_avg": 2.150880813598633,
301
+ "step": 14,
302
+ "val/clipfrac_avg": 0.012931845150887966,
303
+ "val/num_eos_tokens": 36004,
304
+ "val/ratio": 1.0006251335144043,
305
+ "val/ratio_var": 7.0911496550252195e-06
306
+ },
307
+ {
308
+ "episode": 7680,
309
+ "epoch": 0.06579849211788896,
310
+ "eps": 5,
311
+ "loss/policy_avg": -0.006545604206621647,
312
+ "loss/value_avg": 0.6837283968925476,
313
+ "lr": 2.8359375e-06,
314
+ "objective/entropy": -51.048362731933594,
315
+ "objective/kl": 29.153331756591797,
316
+ "objective/non_score_reward": -1.4576666355133057,
317
+ "objective/rlhf_reward": 8.508513450622559,
318
+ "objective/scores": 9.966179847717285,
319
+ "policy/approxkl_avg": 0.003793728072196245,
320
+ "policy/clipfrac_avg": 0.023298773914575577,
321
+ "policy/entropy_avg": 2.1069443225860596,
322
+ "step": 15,
323
+ "val/clipfrac_avg": 0.016515308991074562,
324
+ "val/num_eos_tokens": 33784,
325
+ "val/ratio": 1.0005180835723877,
326
+ "val/ratio_var": 1.2400751074892469e-05
327
+ },
328
+ {
329
+ "episode": 8192,
330
+ "epoch": 0.07018505825908156,
331
+ "eps": 5,
332
+ "loss/policy_avg": -0.010740559548139572,
333
+ "loss/value_avg": 0.6928367614746094,
334
+ "lr": 2.82421875e-06,
335
+ "objective/entropy": -50.633888244628906,
336
+ "objective/kl": 29.69721794128418,
337
+ "objective/non_score_reward": -1.484860897064209,
338
+ "objective/rlhf_reward": 8.630350112915039,
339
+ "objective/scores": 10.115211486816406,
340
+ "policy/approxkl_avg": 0.002925678389146924,
341
+ "policy/clipfrac_avg": 0.022546332329511642,
342
+ "policy/entropy_avg": 2.0426688194274902,
343
+ "step": 16,
344
+ "val/clipfrac_avg": 0.01313821505755186,
345
+ "val/num_eos_tokens": 34515,
346
+ "val/ratio": 0.9999200105667114,
347
+ "val/ratio_var": 3.5139350984536577e-06
348
+ },
349
+ {
350
+ "episode": 8704,
351
+ "epoch": 0.07457162440027416,
352
+ "eps": 5,
353
+ "loss/policy_avg": -0.0003042006865143776,
354
+ "loss/value_avg": 0.630737841129303,
355
+ "lr": 2.8125e-06,
356
+ "objective/entropy": -54.703208923339844,
357
+ "objective/kl": 29.418441772460938,
358
+ "objective/non_score_reward": -1.4709219932556152,
359
+ "objective/rlhf_reward": 8.782114028930664,
360
+ "objective/scores": 10.253036499023438,
361
+ "policy/approxkl_avg": 0.0034269755706191063,
362
+ "policy/clipfrac_avg": 0.021969493478536606,
363
+ "policy/entropy_avg": 2.0263655185699463,
364
+ "step": 17,
365
+ "val/clipfrac_avg": 0.013666579499840736,
366
+ "val/num_eos_tokens": 32440,
367
+ "val/ratio": 0.9990756511688232,
368
+ "val/ratio_var": 5.783027063444024e-06
369
+ },
370
+ {
371
+ "episode": 9216,
372
+ "epoch": 0.07895819054146676,
373
+ "eps": 5,
374
+ "loss/policy_avg": -0.0013809381052851677,
375
+ "loss/value_avg": 0.625673234462738,
376
+ "lr": 2.80078125e-06,
377
+ "objective/entropy": -50.05335235595703,
378
+ "objective/kl": 32.24420166015625,
379
+ "objective/non_score_reward": -1.6122100353240967,
380
+ "objective/rlhf_reward": 8.869917869567871,
381
+ "objective/scores": 10.482128143310547,
382
+ "policy/approxkl_avg": 0.0037200015503913164,
383
+ "policy/clipfrac_avg": 0.023221854120492935,
384
+ "policy/entropy_avg": 1.9334304332733154,
385
+ "step": 18,
386
+ "val/clipfrac_avg": 0.012253889814019203,
387
+ "val/num_eos_tokens": 32755,
388
+ "val/ratio": 1.0010600090026855,
389
+ "val/ratio_var": 1.8864358935388736e-05
390
+ },
391
+ {
392
+ "episode": 9728,
393
+ "epoch": 0.08334475668265935,
394
+ "eps": 5,
395
+ "loss/policy_avg": -0.001600255724042654,
396
+ "loss/value_avg": 0.5548046231269836,
397
+ "lr": 2.7890625e-06,
398
+ "objective/entropy": -52.448543548583984,
399
+ "objective/kl": 31.141429901123047,
400
+ "objective/non_score_reward": -1.5570714473724365,
401
+ "objective/rlhf_reward": 9.025203704833984,
402
+ "objective/scores": 10.582275390625,
403
+ "policy/approxkl_avg": 0.00342507753521204,
404
+ "policy/clipfrac_avg": 0.022851863875985146,
405
+ "policy/entropy_avg": 1.9168764352798462,
406
+ "step": 19,
407
+ "val/clipfrac_avg": 0.011491503566503525,
408
+ "val/num_eos_tokens": 33170,
409
+ "val/ratio": 0.999735951423645,
410
+ "val/ratio_var": 5.808557489217492e-06
411
+ },
412
+ {
413
+ "episode": 10240,
414
+ "epoch": 0.08773132282385196,
415
+ "eps": 5,
416
+ "loss/policy_avg": 0.00035352353006601334,
417
+ "loss/value_avg": 0.5776740312576294,
418
+ "lr": 2.77734375e-06,
419
+ "objective/entropy": -49.64569854736328,
420
+ "objective/kl": 32.26124572753906,
421
+ "objective/non_score_reward": -1.6130623817443848,
422
+ "objective/rlhf_reward": 9.020620346069336,
423
+ "objective/scores": 10.633682250976562,
424
+ "policy/approxkl_avg": 0.003732402576133609,
425
+ "policy/clipfrac_avg": 0.02131691202521324,
426
+ "policy/entropy_avg": 1.8495761156082153,
427
+ "step": 20,
428
+ "val/clipfrac_avg": 0.014484411105513573,
429
+ "val/num_eos_tokens": 32124,
430
+ "val/ratio": 1.000256061553955,
431
+ "val/ratio_var": 1.3850245522917248e-05
432
+ },
433
+ {
434
+ "episode": 10752,
435
+ "epoch": 0.09211788896504455,
436
+ "eps": 5,
437
+ "loss/policy_avg": 0.004326590336859226,
438
+ "loss/value_avg": 0.5091855525970459,
439
+ "lr": 2.765625e-06,
440
+ "objective/entropy": -52.526283264160156,
441
+ "objective/kl": 32.356727600097656,
442
+ "objective/non_score_reward": -1.6178364753723145,
443
+ "objective/rlhf_reward": 8.98788070678711,
444
+ "objective/scores": 10.605716705322266,
445
+ "policy/approxkl_avg": 0.002898063976317644,
446
+ "policy/clipfrac_avg": 0.02005818486213684,
447
+ "policy/entropy_avg": 1.8272948265075684,
448
+ "step": 21,
449
+ "val/clipfrac_avg": 0.010025454685091972,
450
+ "val/num_eos_tokens": 31636,
451
+ "val/ratio": 0.9998326301574707,
452
+ "val/ratio_var": 4.743439149024198e-06
453
+ },
454
+ {
455
+ "episode": 11264,
456
+ "epoch": 0.09650445510623715,
457
+ "eps": 5,
458
+ "loss/policy_avg": 0.001326502999290824,
459
+ "loss/value_avg": 0.5260515213012695,
460
+ "lr": 2.75390625e-06,
461
+ "objective/entropy": -50.08943557739258,
462
+ "objective/kl": 32.325721740722656,
463
+ "objective/non_score_reward": -1.6162861585617065,
464
+ "objective/rlhf_reward": 9.08689022064209,
465
+ "objective/scores": 10.703176498413086,
466
+ "policy/approxkl_avg": 0.004229161888360977,
467
+ "policy/clipfrac_avg": 0.019012872129678726,
468
+ "policy/entropy_avg": 1.7281861305236816,
469
+ "step": 22,
470
+ "val/clipfrac_avg": 0.012183602899312973,
471
+ "val/num_eos_tokens": 30727,
472
+ "val/ratio": 0.9997547268867493,
473
+ "val/ratio_var": 5.818676982016768e-06
474
+ },
475
+ {
476
+ "episode": 11776,
477
+ "epoch": 0.10089102124742974,
478
+ "eps": 5,
479
+ "loss/policy_avg": 0.0023662205785512924,
480
+ "loss/value_avg": 0.4883805811405182,
481
+ "lr": 2.7421875e-06,
482
+ "objective/entropy": -53.46156692504883,
483
+ "objective/kl": 31.849475860595703,
484
+ "objective/non_score_reward": -1.5924739837646484,
485
+ "objective/rlhf_reward": 9.090110778808594,
486
+ "objective/scores": 10.682584762573242,
487
+ "policy/approxkl_avg": 0.003692931029945612,
488
+ "policy/clipfrac_avg": 0.01948227360844612,
489
+ "policy/entropy_avg": 1.7976675033569336,
490
+ "step": 23,
491
+ "val/clipfrac_avg": 0.012527575716376305,
492
+ "val/num_eos_tokens": 31978,
493
+ "val/ratio": 1.0005892515182495,
494
+ "val/ratio_var": 5.0292437663301826e-05
495
+ },
496
+ {
497
+ "episode": 12288,
498
+ "epoch": 0.10527758738862235,
499
+ "eps": 5,
500
+ "loss/policy_avg": -0.002815414220094681,
501
+ "loss/value_avg": 0.47801172733306885,
502
+ "lr": 2.73046875e-06,
503
+ "objective/entropy": -54.20009231567383,
504
+ "objective/kl": 30.75657081604004,
505
+ "objective/non_score_reward": -1.5378286838531494,
506
+ "objective/rlhf_reward": 9.155134201049805,
507
+ "objective/scores": 10.692962646484375,
508
+ "policy/approxkl_avg": 0.0031356574036180973,
509
+ "policy/clipfrac_avg": 0.01814776286482811,
510
+ "policy/entropy_avg": 1.7910568714141846,
511
+ "step": 24,
512
+ "val/clipfrac_avg": 0.0124925896525383,
513
+ "val/num_eos_tokens": 30355,
514
+ "val/ratio": 0.9999936819076538,
515
+ "val/ratio_var": 3.8795255932200234e-06
516
+ },
517
+ {
518
+ "episode": 12800,
519
+ "epoch": 0.10966415352981494,
520
+ "eps": 5,
521
+ "loss/policy_avg": 0.004107598215341568,
522
+ "loss/value_avg": 0.42799514532089233,
523
+ "lr": 2.71875e-06,
524
+ "objective/entropy": -56.50166320800781,
525
+ "objective/kl": 29.329524993896484,
526
+ "objective/non_score_reward": -1.466476321220398,
527
+ "objective/rlhf_reward": 9.282110214233398,
528
+ "objective/scores": 10.748586654663086,
529
+ "policy/approxkl_avg": 0.0033666701056063175,
530
+ "policy/clipfrac_avg": 0.018708810210227966,
531
+ "policy/entropy_avg": 1.836845874786377,
532
+ "step": 25,
533
+ "val/clipfrac_avg": 0.007121690083295107,
534
+ "val/num_eos_tokens": 29066,
535
+ "val/ratio": 0.9996716976165771,
536
+ "val/ratio_var": 5.8772338888957165e-06
537
+ },
538
+ {
539
+ "episode": 13312,
540
+ "epoch": 0.11405071967100754,
541
+ "eps": 5,
542
+ "loss/policy_avg": 0.00938927847892046,
543
+ "loss/value_avg": 0.3949218690395355,
544
+ "lr": 2.70703125e-06,
545
+ "objective/entropy": -60.591705322265625,
546
+ "objective/kl": 28.337303161621094,
547
+ "objective/non_score_reward": -1.4168651103973389,
548
+ "objective/rlhf_reward": 9.06982707977295,
549
+ "objective/scores": 10.486692428588867,
550
+ "policy/approxkl_avg": 0.0028319661505520344,
551
+ "policy/clipfrac_avg": 0.021056555211544037,
552
+ "policy/entropy_avg": 1.9272569417953491,
553
+ "step": 26,
554
+ "val/clipfrac_avg": 0.00876462459564209,
555
+ "val/num_eos_tokens": 28183,
556
+ "val/ratio": 0.9999039173126221,
557
+ "val/ratio_var": 4.642893145501148e-06
558
+ },
559
+ {
560
+ "episode": 13824,
561
+ "epoch": 0.11843728581220014,
562
+ "eps": 5,
563
+ "loss/policy_avg": 0.006120752543210983,
564
+ "loss/value_avg": 0.3891026973724365,
565
+ "lr": 2.6953125e-06,
566
+ "objective/entropy": -60.93333435058594,
567
+ "objective/kl": 28.110050201416016,
568
+ "objective/non_score_reward": -1.4055025577545166,
569
+ "objective/rlhf_reward": 9.255789756774902,
570
+ "objective/scores": 10.66129207611084,
571
+ "policy/approxkl_avg": 0.0032745348289608955,
572
+ "policy/clipfrac_avg": 0.0199459008872509,
573
+ "policy/entropy_avg": 1.9059574604034424,
574
+ "step": 27,
575
+ "val/clipfrac_avg": 0.007393369916826487,
576
+ "val/num_eos_tokens": 29010,
577
+ "val/ratio": 0.999842643737793,
578
+ "val/ratio_var": 8.019253073143773e-06
579
+ },
580
+ {
581
+ "episode": 14336,
582
+ "epoch": 0.12282385195339274,
583
+ "eps": 5,
584
+ "loss/policy_avg": 0.008166640996932983,
585
+ "loss/value_avg": 0.33365607261657715,
586
+ "lr": 2.68359375e-06,
587
+ "objective/entropy": -61.89513397216797,
588
+ "objective/kl": 27.650426864624023,
589
+ "objective/non_score_reward": -1.382521390914917,
590
+ "objective/rlhf_reward": 9.297162055969238,
591
+ "objective/scores": 10.679683685302734,
592
+ "policy/approxkl_avg": 0.0037113018333911896,
593
+ "policy/clipfrac_avg": 0.020008713006973267,
594
+ "policy/entropy_avg": 1.9356969594955444,
595
+ "step": 28,
596
+ "val/clipfrac_avg": 0.008162135258316994,
597
+ "val/num_eos_tokens": 25073,
598
+ "val/ratio": 1.0000048875808716,
599
+ "val/ratio_var": 6.840070909674978e-06
600
+ },
601
+ {
602
+ "episode": 14848,
603
+ "epoch": 0.12721041809458533,
604
+ "eps": 5,
605
+ "loss/policy_avg": 0.01543242298066616,
606
+ "loss/value_avg": 0.3317902088165283,
607
+ "lr": 2.671875e-06,
608
+ "objective/entropy": -53.287864685058594,
609
+ "objective/kl": 28.098434448242188,
610
+ "objective/non_score_reward": -1.4049216508865356,
611
+ "objective/rlhf_reward": 9.339518547058105,
612
+ "objective/scores": 10.744440078735352,
613
+ "policy/approxkl_avg": 0.00311127002350986,
614
+ "policy/clipfrac_avg": 0.02064812183380127,
615
+ "policy/entropy_avg": 1.7959502935409546,
616
+ "step": 29,
617
+ "val/clipfrac_avg": 0.00857294537127018,
618
+ "val/num_eos_tokens": 24235,
619
+ "val/ratio": 1.000222086906433,
620
+ "val/ratio_var": 5.494624929269776e-06
621
+ },
622
+ {
623
+ "episode": 15360,
624
+ "epoch": 0.13159698423577793,
625
+ "eps": 5,
626
+ "loss/policy_avg": 0.010417070239782333,
627
+ "loss/value_avg": 0.34546566009521484,
628
+ "lr": 2.66015625e-06,
629
+ "objective/entropy": -58.94676971435547,
630
+ "objective/kl": 28.113197326660156,
631
+ "objective/non_score_reward": -1.4056599140167236,
632
+ "objective/rlhf_reward": 9.112338066101074,
633
+ "objective/scores": 10.517997741699219,
634
+ "policy/approxkl_avg": 0.003508294001221657,
635
+ "policy/clipfrac_avg": 0.021732624620199203,
636
+ "policy/entropy_avg": 1.8900206089019775,
637
+ "step": 30,
638
+ "val/clipfrac_avg": 0.006745144259184599,
639
+ "val/num_eos_tokens": 26990,
640
+ "val/ratio": 0.9994415044784546,
641
+ "val/ratio_var": 4.539244855550351e-06
642
+ },
643
+ {
644
+ "episode": 15872,
645
+ "epoch": 0.13598355037697052,
646
+ "eps": 5,
647
+ "loss/policy_avg": 0.011747404932975769,
648
+ "loss/value_avg": 0.32529520988464355,
649
+ "lr": 2.6484375e-06,
650
+ "objective/entropy": -50.62657165527344,
651
+ "objective/kl": 28.064453125,
652
+ "objective/non_score_reward": -1.403222680091858,
653
+ "objective/rlhf_reward": 9.3327054977417,
654
+ "objective/scores": 10.735928535461426,
655
+ "policy/approxkl_avg": 0.0032155239023268223,
656
+ "policy/clipfrac_avg": 0.02317969501018524,
657
+ "policy/entropy_avg": 1.7806997299194336,
658
+ "step": 31,
659
+ "val/clipfrac_avg": 0.006999637931585312,
660
+ "val/num_eos_tokens": 23939,
661
+ "val/ratio": 1.0002076625823975,
662
+ "val/ratio_var": 2.5230063329217955e-05
663
+ },
664
+ {
665
+ "episode": 16384,
666
+ "epoch": 0.1403701165181631,
667
+ "eps": 5,
668
+ "loss/policy_avg": 0.005137978587299585,
669
+ "loss/value_avg": 0.32520678639411926,
670
+ "lr": 2.63671875e-06,
671
+ "objective/entropy": -57.196502685546875,
672
+ "objective/kl": 28.533008575439453,
673
+ "objective/non_score_reward": -1.4266504049301147,
674
+ "objective/rlhf_reward": 9.323891639709473,
675
+ "objective/scores": 10.750541687011719,
676
+ "policy/approxkl_avg": 0.003107226686552167,
677
+ "policy/clipfrac_avg": 0.02019701898097992,
678
+ "policy/entropy_avg": 1.8475944995880127,
679
+ "step": 32,
680
+ "val/clipfrac_avg": 0.006430475041270256,
681
+ "val/num_eos_tokens": 25239,
682
+ "val/ratio": 1.0004234313964844,
683
+ "val/ratio_var": 4.5365977712208405e-06
684
+ },
685
+ {
686
+ "episode": 16896,
687
+ "epoch": 0.14475668265935573,
688
+ "eps": 5,
689
+ "loss/policy_avg": 0.01028348132967949,
690
+ "loss/value_avg": 0.35146623849868774,
691
+ "lr": 2.6250000000000003e-06,
692
+ "objective/entropy": -54.473472595214844,
693
+ "objective/kl": 30.472434997558594,
694
+ "objective/non_score_reward": -1.523621678352356,
695
+ "objective/rlhf_reward": 9.17609977722168,
696
+ "objective/scores": 10.699721336364746,
697
+ "policy/approxkl_avg": 0.003178014885634184,
698
+ "policy/clipfrac_avg": 0.0210000891238451,
699
+ "policy/entropy_avg": 1.8384833335876465,
700
+ "step": 33,
701
+ "val/clipfrac_avg": 0.005109312012791634,
702
+ "val/num_eos_tokens": 24597,
703
+ "val/ratio": 1.000359058380127,
704
+ "val/ratio_var": 6.385233064065687e-06
705
+ },
706
+ {
707
+ "episode": 17408,
708
+ "epoch": 0.14914324880054833,
709
+ "eps": 5,
710
+ "loss/policy_avg": 0.004060306120663881,
711
+ "loss/value_avg": 0.3234812617301941,
712
+ "lr": 2.61328125e-06,
713
+ "objective/entropy": -51.67649841308594,
714
+ "objective/kl": 31.204999923706055,
715
+ "objective/non_score_reward": -1.5602500438690186,
716
+ "objective/rlhf_reward": 9.3241605758667,
717
+ "objective/scores": 10.884410858154297,
718
+ "policy/approxkl_avg": 0.0030823112465441227,
719
+ "policy/clipfrac_avg": 0.02090715430676937,
720
+ "policy/entropy_avg": 1.7866475582122803,
721
+ "step": 34,
722
+ "val/clipfrac_avg": 0.006479810923337936,
723
+ "val/num_eos_tokens": 26150,
724
+ "val/ratio": 0.9998917579650879,
725
+ "val/ratio_var": 4.709030235972023e-06
726
+ },
727
+ {
728
+ "episode": 17920,
729
+ "epoch": 0.15352981494174092,
730
+ "eps": 5,
731
+ "loss/policy_avg": -0.0002043084241449833,
732
+ "loss/value_avg": 0.331778347492218,
733
+ "lr": 2.6015625e-06,
734
+ "objective/entropy": -50.791839599609375,
735
+ "objective/kl": 32.52722930908203,
736
+ "objective/non_score_reward": -1.62636137008667,
737
+ "objective/rlhf_reward": 9.229442596435547,
738
+ "objective/scores": 10.855804443359375,
739
+ "policy/approxkl_avg": 0.0038988732267171144,
740
+ "policy/clipfrac_avg": 0.02275794744491577,
741
+ "policy/entropy_avg": 1.7861764430999756,
742
+ "step": 35,
743
+ "val/clipfrac_avg": 0.0063492972403764725,
744
+ "val/num_eos_tokens": 24757,
745
+ "val/ratio": 0.999879002571106,
746
+ "val/ratio_var": 6.5008221099560615e-06
747
+ },
748
+ {
749
+ "episode": 18432,
750
+ "epoch": 0.15791638108293352,
751
+ "eps": 5,
752
+ "loss/policy_avg": 0.0009975926950573921,
753
+ "loss/value_avg": 0.3189573884010315,
754
+ "lr": 2.5898437500000003e-06,
755
+ "objective/entropy": -48.66301727294922,
756
+ "objective/kl": 33.829776763916016,
757
+ "objective/non_score_reward": -1.6914888620376587,
758
+ "objective/rlhf_reward": 9.39903736114502,
759
+ "objective/scores": 11.090526580810547,
760
+ "policy/approxkl_avg": 0.0024533928371965885,
761
+ "policy/clipfrac_avg": 0.020511234179139137,
762
+ "policy/entropy_avg": 1.7290668487548828,
763
+ "step": 36,
764
+ "val/clipfrac_avg": 0.007418747525662184,
765
+ "val/num_eos_tokens": 25660,
766
+ "val/ratio": 1.0001412630081177,
767
+ "val/ratio_var": 4.908704795525409e-06
768
+ },
769
+ {
770
+ "episode": 18944,
771
+ "epoch": 0.1623029472241261,
772
+ "eps": 5,
773
+ "loss/policy_avg": -0.00018343282863497734,
774
+ "loss/value_avg": 0.3374910354614258,
775
+ "lr": 2.578125e-06,
776
+ "objective/entropy": -48.08941650390625,
777
+ "objective/kl": 34.32440948486328,
778
+ "objective/non_score_reward": -1.716220498085022,
779
+ "objective/rlhf_reward": 9.393345832824707,
780
+ "objective/scores": 11.109566688537598,
781
+ "policy/approxkl_avg": 0.002999143209308386,
782
+ "policy/clipfrac_avg": 0.02006850577890873,
783
+ "policy/entropy_avg": 1.6796194314956665,
784
+ "step": 37,
785
+ "val/clipfrac_avg": 0.006178020033985376,
786
+ "val/num_eos_tokens": 23563,
787
+ "val/ratio": 1.0002793073654175,
788
+ "val/ratio_var": 4.9771865633374546e-06
789
+ },
790
+ {
791
+ "episode": 19456,
792
+ "epoch": 0.1666895133653187,
793
+ "eps": 5,
794
+ "loss/policy_avg": 0.006254453212022781,
795
+ "loss/value_avg": 0.35177081823349,
796
+ "lr": 2.56640625e-06,
797
+ "objective/entropy": -48.374507904052734,
798
+ "objective/kl": 34.51807403564453,
799
+ "objective/non_score_reward": -1.725903868675232,
800
+ "objective/rlhf_reward": 9.309269905090332,
801
+ "objective/scores": 11.035173416137695,
802
+ "policy/approxkl_avg": 0.0031045477371662855,
803
+ "policy/clipfrac_avg": 0.020648740231990814,
804
+ "policy/entropy_avg": 1.7128504514694214,
805
+ "step": 38,
806
+ "val/clipfrac_avg": 0.008920802734792233,
807
+ "val/num_eos_tokens": 25658,
808
+ "val/ratio": 0.9995635747909546,
809
+ "val/ratio_var": 6.544411462527933e-06
810
+ },
811
+ {
812
+ "episode": 19968,
813
+ "epoch": 0.1710760795065113,
814
+ "eps": 5,
815
+ "loss/policy_avg": -0.0010195476934313774,
816
+ "loss/value_avg": 0.3460981845855713,
817
+ "lr": 2.5546875000000003e-06,
818
+ "objective/entropy": -51.683372497558594,
819
+ "objective/kl": 32.773372650146484,
820
+ "objective/non_score_reward": -1.6386685371398926,
821
+ "objective/rlhf_reward": 9.41845703125,
822
+ "objective/scores": 11.057125091552734,
823
+ "policy/approxkl_avg": 0.0031053770799189806,
824
+ "policy/clipfrac_avg": 0.02040484920144081,
825
+ "policy/entropy_avg": 1.729933261871338,
826
+ "step": 39,
827
+ "val/clipfrac_avg": 0.006355272606015205,
828
+ "val/num_eos_tokens": 24834,
829
+ "val/ratio": 0.9998283982276917,
830
+ "val/ratio_var": 4.5907336243544705e-06
831
+ },
832
+ {
833
+ "episode": 20480,
834
+ "epoch": 0.17546264564770392,
835
+ "eps": 5,
836
+ "loss/policy_avg": 0.0011649285443127155,
837
+ "loss/value_avg": 0.32062214612960815,
838
+ "lr": 2.54296875e-06,
839
+ "objective/entropy": -53.09362030029297,
840
+ "objective/kl": 33.102630615234375,
841
+ "objective/non_score_reward": -1.6551315784454346,
842
+ "objective/rlhf_reward": 9.409041404724121,
843
+ "objective/scores": 11.064172744750977,
844
+ "policy/approxkl_avg": 0.0030981849413365126,
845
+ "policy/clipfrac_avg": 0.01863136701285839,
846
+ "policy/entropy_avg": 1.7126240730285645,
847
+ "step": 40,
848
+ "val/clipfrac_avg": 0.007702820934355259,
849
+ "val/num_eos_tokens": 24091,
850
+ "val/ratio": 0.9998593330383301,
851
+ "val/ratio_var": 5.052448614151217e-06
852
+ },
853
+ {
854
+ "episode": 20992,
855
+ "epoch": 0.1798492117888965,
856
+ "eps": 5,
857
+ "loss/policy_avg": 0.0019053546711802483,
858
+ "loss/value_avg": 0.3188174068927765,
859
+ "lr": 2.53125e-06,
860
+ "objective/entropy": -54.4024543762207,
861
+ "objective/kl": 32.57151794433594,
862
+ "objective/non_score_reward": -1.6285758018493652,
863
+ "objective/rlhf_reward": 9.431205749511719,
864
+ "objective/scores": 11.059782028198242,
865
+ "policy/approxkl_avg": 0.0028727450408041477,
866
+ "policy/clipfrac_avg": 0.01854875311255455,
867
+ "policy/entropy_avg": 1.7159607410430908,
868
+ "step": 41,
869
+ "val/clipfrac_avg": 0.008048100396990776,
870
+ "val/num_eos_tokens": 24981,
871
+ "val/ratio": 0.9997231960296631,
872
+ "val/ratio_var": 4.052901658724295e-06
873
+ },
874
+ {
875
+ "episode": 21504,
876
+ "epoch": 0.1842357779300891,
877
+ "eps": 5,
878
+ "loss/policy_avg": 0.001745171844959259,
879
+ "loss/value_avg": 0.35853347182273865,
880
+ "lr": 2.5195312500000003e-06,
881
+ "objective/entropy": -54.93397521972656,
882
+ "objective/kl": 32.59296417236328,
883
+ "objective/non_score_reward": -1.6296483278274536,
884
+ "objective/rlhf_reward": 9.271963119506836,
885
+ "objective/scores": 10.901611328125,
886
+ "policy/approxkl_avg": 0.002658254001289606,
887
+ "policy/clipfrac_avg": 0.01734079420566559,
888
+ "policy/entropy_avg": 1.7013740539550781,
889
+ "step": 42,
890
+ "val/clipfrac_avg": 0.005504803732037544,
891
+ "val/num_eos_tokens": 27864,
892
+ "val/ratio": 1.0002515316009521,
893
+ "val/ratio_var": 4.718350737675792e-06
894
+ },
895
+ {
896
+ "episode": 22016,
897
+ "epoch": 0.1886223440712817,
898
+ "eps": 5,
899
+ "loss/policy_avg": 0.0031983088701963425,
900
+ "loss/value_avg": 0.32714200019836426,
901
+ "lr": 2.5078125e-06,
902
+ "objective/entropy": -56.087486267089844,
903
+ "objective/kl": 31.602561950683594,
904
+ "objective/non_score_reward": -1.5801280736923218,
905
+ "objective/rlhf_reward": 9.48602294921875,
906
+ "objective/scores": 11.066150665283203,
907
+ "policy/approxkl_avg": 0.0033945119939744473,
908
+ "policy/clipfrac_avg": 0.017319316044449806,
909
+ "policy/entropy_avg": 1.7106884717941284,
910
+ "step": 43,
911
+ "val/clipfrac_avg": 0.00773418415337801,
912
+ "val/num_eos_tokens": 27303,
913
+ "val/ratio": 0.9996844530105591,
914
+ "val/ratio_var": 4.000376975454856e-06
915
+ },
916
+ {
917
+ "episode": 22528,
918
+ "epoch": 0.1930089102124743,
919
+ "eps": 5,
920
+ "loss/policy_avg": 0.0024800747632980347,
921
+ "loss/value_avg": 0.3157750368118286,
922
+ "lr": 2.49609375e-06,
923
+ "objective/entropy": -58.142826080322266,
924
+ "objective/kl": 31.376550674438477,
925
+ "objective/non_score_reward": -1.5688276290893555,
926
+ "objective/rlhf_reward": 9.588839530944824,
927
+ "objective/scores": 11.15766716003418,
928
+ "policy/approxkl_avg": 0.0025360295549035072,
929
+ "policy/clipfrac_avg": 0.015292399562895298,
930
+ "policy/entropy_avg": 1.6972548961639404,
931
+ "step": 44,
932
+ "val/clipfrac_avg": 0.007428675889968872,
933
+ "val/num_eos_tokens": 26608,
934
+ "val/ratio": 1.0000505447387695,
935
+ "val/ratio_var": 4.718193395092385e-06
936
+ },
937
+ {
938
+ "episode": 23040,
939
+ "epoch": 0.1973954763536669,
940
+ "eps": 5,
941
+ "loss/policy_avg": 0.001205185428261757,
942
+ "loss/value_avg": 0.3156132996082306,
943
+ "lr": 2.4843750000000002e-06,
944
+ "objective/entropy": -58.6962890625,
945
+ "objective/kl": 30.90795135498047,
946
+ "objective/non_score_reward": -1.5453976392745972,
947
+ "objective/rlhf_reward": 9.614949226379395,
948
+ "objective/scores": 11.160346984863281,
949
+ "policy/approxkl_avg": 0.0026144087314605713,
950
+ "policy/clipfrac_avg": 0.016292206943035126,
951
+ "policy/entropy_avg": 1.7031623125076294,
952
+ "step": 45,
953
+ "val/clipfrac_avg": 0.0070708440616726875,
954
+ "val/num_eos_tokens": 26980,
955
+ "val/ratio": 0.9996762275695801,
956
+ "val/ratio_var": 4.505399829213275e-06
957
+ },
958
+ {
959
+ "episode": 23552,
960
+ "epoch": 0.20178204249485948,
961
+ "eps": 5,
962
+ "loss/policy_avg": 0.004359962418675423,
963
+ "loss/value_avg": 0.32259702682495117,
964
+ "lr": 2.47265625e-06,
965
+ "objective/entropy": -61.484619140625,
966
+ "objective/kl": 29.10392189025879,
967
+ "objective/non_score_reward": -1.4551961421966553,
968
+ "objective/rlhf_reward": 9.753060340881348,
969
+ "objective/scores": 11.208256721496582,
970
+ "policy/approxkl_avg": 0.002397476462647319,
971
+ "policy/clipfrac_avg": 0.013912687078118324,
972
+ "policy/entropy_avg": 1.727178931236267,
973
+ "step": 46,
974
+ "val/clipfrac_avg": 0.00816885195672512,
975
+ "val/num_eos_tokens": 28750,
976
+ "val/ratio": 0.999366283416748,
977
+ "val/ratio_var": 4.691919002652867e-06
978
+ },
979
+ {
980
+ "episode": 24064,
981
+ "epoch": 0.2061686086360521,
982
+ "eps": 5,
983
+ "loss/policy_avg": 0.001963222399353981,
984
+ "loss/value_avg": 0.31864023208618164,
985
+ "lr": 2.4609375e-06,
986
+ "objective/entropy": -61.265846252441406,
987
+ "objective/kl": 29.692306518554688,
988
+ "objective/non_score_reward": -1.4846153259277344,
989
+ "objective/rlhf_reward": 9.666860580444336,
990
+ "objective/scores": 11.15147590637207,
991
+ "policy/approxkl_avg": 0.002565302886068821,
992
+ "policy/clipfrac_avg": 0.013390684500336647,
993
+ "policy/entropy_avg": 1.7250237464904785,
994
+ "step": 47,
995
+ "val/clipfrac_avg": 0.007603655569255352,
996
+ "val/num_eos_tokens": 28052,
997
+ "val/ratio": 1.0003018379211426,
998
+ "val/ratio_var": 5.1418255679891445e-06
999
+ },
1000
+ {
1001
+ "episode": 24576,
1002
+ "epoch": 0.2105551747772447,
1003
+ "eps": 5,
1004
+ "loss/policy_avg": 0.007894441485404968,
1005
+ "loss/value_avg": 0.32393255829811096,
1006
+ "lr": 2.4492187500000002e-06,
1007
+ "objective/entropy": -61.45293426513672,
1008
+ "objective/kl": 29.473636627197266,
1009
+ "objective/non_score_reward": -1.4736818075180054,
1010
+ "objective/rlhf_reward": 9.70655345916748,
1011
+ "objective/scores": 11.180234909057617,
1012
+ "policy/approxkl_avg": 0.0018561662873253226,
1013
+ "policy/clipfrac_avg": 0.012259826064109802,
1014
+ "policy/entropy_avg": 1.7109990119934082,
1015
+ "step": 48,
1016
+ "val/clipfrac_avg": 0.007823488675057888,
1017
+ "val/num_eos_tokens": 30985,
1018
+ "val/ratio": 1.0001144409179688,
1019
+ "val/ratio_var": 3.26874760503415e-06
1020
+ },
1021
+ {
1022
+ "episode": 25088,
1023
+ "epoch": 0.2149417409184373,
1024
+ "eps": 5,
1025
+ "loss/policy_avg": 0.00292217917740345,
1026
+ "loss/value_avg": 0.3294008672237396,
1027
+ "lr": 2.4375e-06,
1028
+ "objective/entropy": -60.33291244506836,
1029
+ "objective/kl": 30.310314178466797,
1030
+ "objective/non_score_reward": -1.5155158042907715,
1031
+ "objective/rlhf_reward": 9.636428833007812,
1032
+ "objective/scores": 11.151945114135742,
1033
+ "policy/approxkl_avg": 0.0021009996999055147,
1034
+ "policy/clipfrac_avg": 0.012688988819718361,
1035
+ "policy/entropy_avg": 1.6571977138519287,
1036
+ "step": 49,
1037
+ "val/clipfrac_avg": 0.0057515716180205345,
1038
+ "val/num_eos_tokens": 27460,
1039
+ "val/ratio": 1.0001912117004395,
1040
+ "val/ratio_var": 3.742007493201527e-06
1041
+ },
1042
+ {
1043
+ "episode": 25600,
1044
+ "epoch": 0.21932830705962988,
1045
+ "eps": 5,
1046
+ "loss/policy_avg": 0.003821900114417076,
1047
+ "loss/value_avg": 0.30453431606292725,
1048
+ "lr": 2.42578125e-06,
1049
+ "objective/entropy": -59.68878173828125,
1050
+ "objective/kl": 30.681987762451172,
1051
+ "objective/non_score_reward": -1.5340994596481323,
1052
+ "objective/rlhf_reward": 9.784141540527344,
1053
+ "objective/scores": 11.318241119384766,
1054
+ "policy/approxkl_avg": 0.0016853193519636989,
1055
+ "policy/clipfrac_avg": 0.011813260614871979,
1056
+ "policy/entropy_avg": 1.642488956451416,
1057
+ "step": 50,
1058
+ "val/clipfrac_avg": 0.00796700082719326,
1059
+ "val/num_eos_tokens": 28255,
1060
+ "val/ratio": 1.0001150369644165,
1061
+ "val/ratio_var": 3.880189069604967e-06
1062
+ },
1063
+ {
1064
+ "episode": 26112,
1065
+ "epoch": 0.22371487320082248,
1066
+ "eps": 5,
1067
+ "loss/policy_avg": 0.0011352086439728737,
1068
+ "loss/value_avg": 0.30040693283081055,
1069
+ "lr": 2.4140625000000002e-06,
1070
+ "objective/entropy": -55.53452682495117,
1071
+ "objective/kl": 33.050025939941406,
1072
+ "objective/non_score_reward": -1.6525013446807861,
1073
+ "objective/rlhf_reward": 9.85799503326416,
1074
+ "objective/scores": 11.510496139526367,
1075
+ "policy/approxkl_avg": 0.0019366666674613953,
1076
+ "policy/clipfrac_avg": 0.011372741311788559,
1077
+ "policy/entropy_avg": 1.5299839973449707,
1078
+ "step": 51,
1079
+ "val/clipfrac_avg": 0.008255399763584137,
1080
+ "val/num_eos_tokens": 27884,
1081
+ "val/ratio": 0.9998023509979248,
1082
+ "val/ratio_var": 4.162790446571307e-06
1083
+ },
1084
+ {
1085
+ "episode": 26624,
1086
+ "epoch": 0.22810143934201507,
1087
+ "eps": 5,
1088
+ "loss/policy_avg": 0.008498594164848328,
1089
+ "loss/value_avg": 0.2648542523384094,
1090
+ "lr": 2.40234375e-06,
1091
+ "objective/entropy": -56.34510803222656,
1092
+ "objective/kl": 32.5233154296875,
1093
+ "objective/non_score_reward": -1.626165747642517,
1094
+ "objective/rlhf_reward": 9.983355522155762,
1095
+ "objective/scores": 11.60952091217041,
1096
+ "policy/approxkl_avg": 0.0027015511877834797,
1097
+ "policy/clipfrac_avg": 0.011898016557097435,
1098
+ "policy/entropy_avg": 1.5591082572937012,
1099
+ "step": 52,
1100
+ "val/clipfrac_avg": 0.006908833980560303,
1101
+ "val/num_eos_tokens": 27153,
1102
+ "val/ratio": 0.9998370409011841,
1103
+ "val/ratio_var": 2.8140475478721783e-06
1104
+ },
1105
+ {
1106
+ "episode": 27136,
1107
+ "epoch": 0.23248800548320767,
1108
+ "eps": 5,
1109
+ "loss/policy_avg": -0.0007923748344182968,
1110
+ "loss/value_avg": 0.2664443850517273,
1111
+ "lr": 2.390625e-06,
1112
+ "objective/entropy": -52.48090362548828,
1113
+ "objective/kl": 35.079097747802734,
1114
+ "objective/non_score_reward": -1.7539548873901367,
1115
+ "objective/rlhf_reward": 9.7537202835083,
1116
+ "objective/scores": 11.507675170898438,
1117
+ "policy/approxkl_avg": 0.002516430802643299,
1118
+ "policy/clipfrac_avg": 0.012202695943415165,
1119
+ "policy/entropy_avg": 1.4821527004241943,
1120
+ "step": 53,
1121
+ "val/clipfrac_avg": 0.005511893425136805,
1122
+ "val/num_eos_tokens": 27268,
1123
+ "val/ratio": 1.0001299381256104,
1124
+ "val/ratio_var": 2.9662376164196758e-06
1125
+ },
1126
+ {
1127
+ "episode": 27648,
1128
+ "epoch": 0.2368745716244003,
1129
+ "eps": 5,
1130
+ "loss/policy_avg": 0.00248061865568161,
1131
+ "loss/value_avg": 0.29047083854675293,
1132
+ "lr": 2.3789062500000002e-06,
1133
+ "objective/entropy": -50.3958740234375,
1134
+ "objective/kl": 35.103553771972656,
1135
+ "objective/non_score_reward": -1.755177617073059,
1136
+ "objective/rlhf_reward": 9.844841003417969,
1137
+ "objective/scores": 11.600018501281738,
1138
+ "policy/approxkl_avg": 0.002680136589333415,
1139
+ "policy/clipfrac_avg": 0.010980302467942238,
1140
+ "policy/entropy_avg": 1.4363842010498047,
1141
+ "step": 54,
1142
+ "val/clipfrac_avg": 0.006265181582421064,
1143
+ "val/num_eos_tokens": 26313,
1144
+ "val/ratio": 1.0002726316452026,
1145
+ "val/ratio_var": 3.265275154262781e-06
1146
+ },
1147
+ {
1148
+ "episode": 28160,
1149
+ "epoch": 0.24126113776559288,
1150
+ "eps": 5,
1151
+ "loss/policy_avg": 0.00040969252586364746,
1152
+ "loss/value_avg": 0.27804386615753174,
1153
+ "lr": 2.3671875e-06,
1154
+ "objective/entropy": -50.88187789916992,
1155
+ "objective/kl": 34.17361068725586,
1156
+ "objective/non_score_reward": -1.7086806297302246,
1157
+ "objective/rlhf_reward": 9.788152694702148,
1158
+ "objective/scores": 11.496833801269531,
1159
+ "policy/approxkl_avg": 0.00247185374610126,
1160
+ "policy/clipfrac_avg": 0.011061925441026688,
1161
+ "policy/entropy_avg": 1.4272186756134033,
1162
+ "step": 55,
1163
+ "val/clipfrac_avg": 0.004378362558782101,
1164
+ "val/num_eos_tokens": 28037,
1165
+ "val/ratio": 0.9997042417526245,
1166
+ "val/ratio_var": 3.098047955063521e-06
1167
+ },
1168
+ {
1169
+ "episode": 28672,
1170
+ "epoch": 0.24564770390678548,
1171
+ "eps": 5,
1172
+ "loss/policy_avg": 0.001188849564641714,
1173
+ "loss/value_avg": 0.2825721502304077,
1174
+ "lr": 2.35546875e-06,
1175
+ "objective/entropy": -52.136688232421875,
1176
+ "objective/kl": 33.818946838378906,
1177
+ "objective/non_score_reward": -1.690947413444519,
1178
+ "objective/rlhf_reward": 9.844747543334961,
1179
+ "objective/scores": 11.53569507598877,
1180
+ "policy/approxkl_avg": 0.0022125309333205223,
1181
+ "policy/clipfrac_avg": 0.011263608932495117,
1182
+ "policy/entropy_avg": 1.4427647590637207,
1183
+ "step": 56,
1184
+ "val/clipfrac_avg": 0.004626412410289049,
1185
+ "val/num_eos_tokens": 27546,
1186
+ "val/ratio": 0.999881386756897,
1187
+ "val/ratio_var": 4.78684432891896e-06
1188
+ },
1189
+ {
1190
+ "episode": 29184,
1191
+ "epoch": 0.25003427004797807,
1192
+ "eps": 4,
1193
+ "loss/policy_avg": 0.009166279807686806,
1194
+ "loss/value_avg": 0.2479139119386673,
1195
+ "lr": 2.3437500000000002e-06,
1196
+ "objective/entropy": -54.860206604003906,
1197
+ "objective/kl": 32.90901184082031,
1198
+ "objective/non_score_reward": -1.6454508304595947,
1199
+ "objective/rlhf_reward": 9.943949699401855,
1200
+ "objective/scores": 11.589400291442871,
1201
+ "policy/approxkl_avg": 0.0023442034143954515,
1202
+ "policy/clipfrac_avg": 0.011214188300073147,
1203
+ "policy/entropy_avg": 1.4958158731460571,
1204
+ "step": 57,
1205
+ "val/clipfrac_avg": 0.005244936794042587,
1206
+ "val/num_eos_tokens": 26152,
1207
+ "val/ratio": 1.0008132457733154,
1208
+ "val/ratio_var": 4.664201696868986e-06
1209
+ },
1210
+ {
1211
+ "episode": 29696,
1212
+ "epoch": 0.25442083618917066,
1213
+ "eps": 4,
1214
+ "loss/policy_avg": 0.004646984860301018,
1215
+ "loss/value_avg": 0.2771589457988739,
1216
+ "lr": 2.33203125e-06,
1217
+ "objective/entropy": -53.552146911621094,
1218
+ "objective/kl": 33.505615234375,
1219
+ "objective/non_score_reward": -1.6752808094024658,
1220
+ "objective/rlhf_reward": 9.639711380004883,
1221
+ "objective/scores": 11.31499195098877,
1222
+ "policy/approxkl_avg": 0.0028129604179412127,
1223
+ "policy/clipfrac_avg": 0.013663064688444138,
1224
+ "policy/entropy_avg": 1.483880639076233,
1225
+ "step": 58,
1226
+ "val/clipfrac_avg": 0.004003824666142464,
1227
+ "val/num_eos_tokens": 26445,
1228
+ "val/ratio": 0.9997650980949402,
1229
+ "val/ratio_var": 2.8647132239711937e-06
1230
+ },
1231
+ {
1232
+ "episode": 30208,
1233
+ "epoch": 0.25880740233036326,
1234
+ "eps": 4,
1235
+ "loss/policy_avg": 0.006868647411465645,
1236
+ "loss/value_avg": 0.25104641914367676,
1237
+ "lr": 2.3203125e-06,
1238
+ "objective/entropy": -57.64221954345703,
1239
+ "objective/kl": 30.967973709106445,
1240
+ "objective/non_score_reward": -1.548398733139038,
1241
+ "objective/rlhf_reward": 9.892714500427246,
1242
+ "objective/scores": 11.441113471984863,
1243
+ "policy/approxkl_avg": 0.001667265547439456,
1244
+ "policy/clipfrac_avg": 0.010815152898430824,
1245
+ "policy/entropy_avg": 1.515237808227539,
1246
+ "step": 59,
1247
+ "val/clipfrac_avg": 0.004705238156020641,
1248
+ "val/num_eos_tokens": 25173,
1249
+ "val/ratio": 0.999830424785614,
1250
+ "val/ratio_var": 3.4093393423972884e-06
1251
+ },
1252
+ {
1253
+ "episode": 30720,
1254
+ "epoch": 0.26319396847155585,
1255
+ "eps": 4,
1256
+ "loss/policy_avg": 0.011243993416428566,
1257
+ "loss/value_avg": 0.2495395541191101,
1258
+ "lr": 2.30859375e-06,
1259
+ "objective/entropy": -60.84779357910156,
1260
+ "objective/kl": 31.15279769897461,
1261
+ "objective/non_score_reward": -1.5576398372650146,
1262
+ "objective/rlhf_reward": 9.788341522216797,
1263
+ "objective/scores": 11.34598159790039,
1264
+ "policy/approxkl_avg": 0.002522544004023075,
1265
+ "policy/clipfrac_avg": 0.01168464682996273,
1266
+ "policy/entropy_avg": 1.5892558097839355,
1267
+ "step": 60,
1268
+ "val/clipfrac_avg": 0.0036916325334459543,
1269
+ "val/num_eos_tokens": 25030,
1270
+ "val/ratio": 1.000048041343689,
1271
+ "val/ratio_var": 5.555901225307025e-06
1272
+ },
1273
+ {
1274
+ "episode": 31232,
1275
+ "epoch": 0.26758053461274844,
1276
+ "eps": 4,
1277
+ "loss/policy_avg": 0.005039035342633724,
1278
+ "loss/value_avg": 0.23521627485752106,
1279
+ "lr": 2.296875e-06,
1280
+ "objective/entropy": -52.48558807373047,
1281
+ "objective/kl": 29.97555923461914,
1282
+ "objective/non_score_reward": -1.498777985572815,
1283
+ "objective/rlhf_reward": 9.8745756149292,
1284
+ "objective/scores": 11.373353958129883,
1285
+ "policy/approxkl_avg": 0.0022459528408944607,
1286
+ "policy/clipfrac_avg": 0.011031204834580421,
1287
+ "policy/entropy_avg": 1.452695608139038,
1288
+ "step": 61,
1289
+ "val/clipfrac_avg": 0.0035397186875343323,
1290
+ "val/num_eos_tokens": 22719,
1291
+ "val/ratio": 1.0000181198120117,
1292
+ "val/ratio_var": 7.56223380449228e-06
1293
+ },
1294
+ {
1295
+ "episode": 31744,
1296
+ "epoch": 0.27196710075394104,
1297
+ "eps": 5,
1298
+ "loss/policy_avg": 0.003956328146159649,
1299
+ "loss/value_avg": 0.2286529839038849,
1300
+ "lr": 2.28515625e-06,
1301
+ "objective/entropy": -62.272193908691406,
1302
+ "objective/kl": 30.68024444580078,
1303
+ "objective/non_score_reward": -1.5340123176574707,
1304
+ "objective/rlhf_reward": 9.862707138061523,
1305
+ "objective/scores": 11.396718978881836,
1306
+ "policy/approxkl_avg": 0.0018089789664372802,
1307
+ "policy/clipfrac_avg": 0.01095383707433939,
1308
+ "policy/entropy_avg": 1.5233193635940552,
1309
+ "step": 62,
1310
+ "val/clipfrac_avg": 0.0031363165471702814,
1311
+ "val/num_eos_tokens": 21682,
1312
+ "val/ratio": 1.0003973245620728,
1313
+ "val/ratio_var": 3.887281764036743e-06
1314
+ },
1315
+ {
1316
+ "episode": 32256,
1317
+ "epoch": 0.27635366689513363,
1318
+ "eps": 5,
1319
+ "loss/policy_avg": 0.006541299633681774,
1320
+ "loss/value_avg": 0.2510289251804352,
1321
+ "lr": 2.2734375e-06,
1322
+ "objective/entropy": -63.541847229003906,
1323
+ "objective/kl": 29.277267456054688,
1324
+ "objective/non_score_reward": -1.4638633728027344,
1325
+ "objective/rlhf_reward": 9.708379745483398,
1326
+ "objective/scores": 11.172243118286133,
1327
+ "policy/approxkl_avg": 0.0018989848904311657,
1328
+ "policy/clipfrac_avg": 0.01218925230205059,
1329
+ "policy/entropy_avg": 1.6170835494995117,
1330
+ "step": 63,
1331
+ "val/clipfrac_avg": 0.0036716386675834656,
1332
+ "val/num_eos_tokens": 23597,
1333
+ "val/ratio": 0.999724268913269,
1334
+ "val/ratio_var": 3.221219003535225e-06
1335
+ },
1336
+ {
1337
+ "episode": 32768,
1338
+ "epoch": 0.2807402330363262,
1339
+ "eps": 5,
1340
+ "loss/policy_avg": 0.00918935239315033,
1341
+ "loss/value_avg": 0.2206374704837799,
1342
+ "lr": 2.26171875e-06,
1343
+ "objective/entropy": -62.05379867553711,
1344
+ "objective/kl": 29.984851837158203,
1345
+ "objective/non_score_reward": -1.4992427825927734,
1346
+ "objective/rlhf_reward": 9.81045913696289,
1347
+ "objective/scores": 11.309701919555664,
1348
+ "policy/approxkl_avg": 0.0021062542218714952,
1349
+ "policy/clipfrac_avg": 0.010909339413046837,
1350
+ "policy/entropy_avg": 1.581649661064148,
1351
+ "step": 64,
1352
+ "val/clipfrac_avg": 0.0033828443847596645,
1353
+ "val/num_eos_tokens": 25070,
1354
+ "val/ratio": 0.9998538494110107,
1355
+ "val/ratio_var": 2.1343398657336365e-06
1356
+ },
1357
+ {
1358
+ "episode": 33280,
1359
+ "epoch": 0.2851267991775189,
1360
+ "eps": 5,
1361
+ "loss/policy_avg": 0.0031117182224988937,
1362
+ "loss/value_avg": 0.2214275449514389,
1363
+ "lr": 2.25e-06,
1364
+ "objective/entropy": -61.279544830322266,
1365
+ "objective/kl": 30.740419387817383,
1366
+ "objective/non_score_reward": -1.5370210409164429,
1367
+ "objective/rlhf_reward": 9.933812141418457,
1368
+ "objective/scores": 11.470832824707031,
1369
+ "policy/approxkl_avg": 0.003304574405774474,
1370
+ "policy/clipfrac_avg": 0.011563178151845932,
1371
+ "policy/entropy_avg": 1.5370697975158691,
1372
+ "step": 65,
1373
+ "val/clipfrac_avg": 0.0039400579407811165,
1374
+ "val/num_eos_tokens": 23425,
1375
+ "val/ratio": 1.0000369548797607,
1376
+ "val/ratio_var": 2.214629830632475e-06
1377
+ },
1378
+ {
1379
+ "episode": 33792,
1380
+ "epoch": 0.28951336531871147,
1381
+ "eps": 5,
1382
+ "loss/policy_avg": 0.006491166073828936,
1383
+ "loss/value_avg": 0.22057856619358063,
1384
+ "lr": 2.23828125e-06,
1385
+ "objective/entropy": -61.981842041015625,
1386
+ "objective/kl": 30.143644332885742,
1387
+ "objective/non_score_reward": -1.5071823596954346,
1388
+ "objective/rlhf_reward": 9.982759475708008,
1389
+ "objective/scores": 11.489941596984863,
1390
+ "policy/approxkl_avg": 0.0016894883010536432,
1391
+ "policy/clipfrac_avg": 0.012134966440498829,
1392
+ "policy/entropy_avg": 1.5720046758651733,
1393
+ "step": 66,
1394
+ "val/clipfrac_avg": 0.0033682635985314846,
1395
+ "val/num_eos_tokens": 22234,
1396
+ "val/ratio": 1.0001835823059082,
1397
+ "val/ratio_var": 5.471536951517919e-06
1398
+ },
1399
+ {
1400
+ "episode": 34304,
1401
+ "epoch": 0.29389993145990406,
1402
+ "eps": 5,
1403
+ "loss/policy_avg": 0.00841301865875721,
1404
+ "loss/value_avg": 0.2318386435508728,
1405
+ "lr": 2.2265625e-06,
1406
+ "objective/entropy": -60.53614044189453,
1407
+ "objective/kl": 31.55874252319336,
1408
+ "objective/non_score_reward": -1.577937126159668,
1409
+ "objective/rlhf_reward": 9.732769012451172,
1410
+ "objective/scores": 11.31070613861084,
1411
+ "policy/approxkl_avg": 0.0018887519836425781,
1412
+ "policy/clipfrac_avg": 0.011964268051087856,
1413
+ "policy/entropy_avg": 1.5244885683059692,
1414
+ "step": 67,
1415
+ "val/clipfrac_avg": 0.0025918553583323956,
1416
+ "val/num_eos_tokens": 22458,
1417
+ "val/ratio": 1.0001020431518555,
1418
+ "val/ratio_var": 1.6006388250389136e-06
1419
+ },
1420
+ {
1421
+ "episode": 34816,
1422
+ "epoch": 0.29828649760109666,
1423
+ "eps": 5,
1424
+ "loss/policy_avg": 0.004491077736020088,
1425
+ "loss/value_avg": 0.21774470806121826,
1426
+ "lr": 2.21484375e-06,
1427
+ "objective/entropy": -61.31328582763672,
1428
+ "objective/kl": 30.977596282958984,
1429
+ "objective/non_score_reward": -1.548879861831665,
1430
+ "objective/rlhf_reward": 9.94593620300293,
1431
+ "objective/scores": 11.494815826416016,
1432
+ "policy/approxkl_avg": 0.002278459956869483,
1433
+ "policy/clipfrac_avg": 0.01156328059732914,
1434
+ "policy/entropy_avg": 1.5633031129837036,
1435
+ "step": 68,
1436
+ "val/clipfrac_avg": 0.004319990985095501,
1437
+ "val/num_eos_tokens": 22420,
1438
+ "val/ratio": 0.9998349547386169,
1439
+ "val/ratio_var": 2.804338009809726e-06
1440
+ },
1441
+ {
1442
+ "episode": 35328,
1443
+ "epoch": 0.30267306374228925,
1444
+ "eps": 5,
1445
+ "loss/policy_avg": 0.012953916564583778,
1446
+ "loss/value_avg": 0.22159670293331146,
1447
+ "lr": 2.203125e-06,
1448
+ "objective/entropy": -60.40982437133789,
1449
+ "objective/kl": 31.161331176757812,
1450
+ "objective/non_score_reward": -1.5580666065216064,
1451
+ "objective/rlhf_reward": 9.979460716247559,
1452
+ "objective/scores": 11.537527084350586,
1453
+ "policy/approxkl_avg": 0.002316855126991868,
1454
+ "policy/clipfrac_avg": 0.012466225773096085,
1455
+ "policy/entropy_avg": 1.5732831954956055,
1456
+ "step": 69,
1457
+ "val/clipfrac_avg": 0.004431965295225382,
1458
+ "val/num_eos_tokens": 23475,
1459
+ "val/ratio": 0.9999319314956665,
1460
+ "val/ratio_var": 2.884213245124556e-06
1461
+ },
1462
+ {
1463
+ "episode": 35840,
1464
+ "epoch": 0.30705962988348184,
1465
+ "eps": 5,
1466
+ "loss/policy_avg": 0.0026891864836215973,
1467
+ "loss/value_avg": 0.22345781326293945,
1468
+ "lr": 2.19140625e-06,
1469
+ "objective/entropy": -59.21501541137695,
1470
+ "objective/kl": 31.3795166015625,
1471
+ "objective/non_score_reward": -1.568975806236267,
1472
+ "objective/rlhf_reward": 10.109546661376953,
1473
+ "objective/scores": 11.678522109985352,
1474
+ "policy/approxkl_avg": 0.0033681951463222504,
1475
+ "policy/clipfrac_avg": 0.01226731576025486,
1476
+ "policy/entropy_avg": 1.559905767440796,
1477
+ "step": 70,
1478
+ "val/clipfrac_avg": 0.004147649277001619,
1479
+ "val/num_eos_tokens": 26871,
1480
+ "val/ratio": 0.9998610615730286,
1481
+ "val/ratio_var": 3.8059629332565237e-06
1482
+ },
1483
+ {
1484
+ "episode": 36352,
1485
+ "epoch": 0.31144619602467444,
1486
+ "eps": 5,
1487
+ "loss/policy_avg": 0.010949812829494476,
1488
+ "loss/value_avg": 0.21867918968200684,
1489
+ "lr": 2.1796875e-06,
1490
+ "objective/entropy": -55.472938537597656,
1491
+ "objective/kl": 33.11960220336914,
1492
+ "objective/non_score_reward": -1.655980110168457,
1493
+ "objective/rlhf_reward": 9.923175811767578,
1494
+ "objective/scores": 11.579155921936035,
1495
+ "policy/approxkl_avg": 0.0021974374540150166,
1496
+ "policy/clipfrac_avg": 0.013162683695554733,
1497
+ "policy/entropy_avg": 1.446916103363037,
1498
+ "step": 71,
1499
+ "val/clipfrac_avg": 0.004135113209486008,
1500
+ "val/num_eos_tokens": 25201,
1501
+ "val/ratio": 1.0002843141555786,
1502
+ "val/ratio_var": 5.781992513220757e-06
1503
+ },
1504
+ {
1505
+ "episode": 36864,
1506
+ "epoch": 0.31583276216586703,
1507
+ "eps": 5,
1508
+ "loss/policy_avg": 0.008956819772720337,
1509
+ "loss/value_avg": 0.22117221355438232,
1510
+ "lr": 2.16796875e-06,
1511
+ "objective/entropy": -57.94511413574219,
1512
+ "objective/kl": 32.885406494140625,
1513
+ "objective/non_score_reward": -1.6442703008651733,
1514
+ "objective/rlhf_reward": 9.936177253723145,
1515
+ "objective/scores": 11.58044719696045,
1516
+ "policy/approxkl_avg": 0.0023873518221080303,
1517
+ "policy/clipfrac_avg": 0.011646779254078865,
1518
+ "policy/entropy_avg": 1.5099616050720215,
1519
+ "step": 72,
1520
+ "val/clipfrac_avg": 0.003031653817743063,
1521
+ "val/num_eos_tokens": 23619,
1522
+ "val/ratio": 0.9997458457946777,
1523
+ "val/ratio_var": 3.188845766999293e-06
1524
+ },
1525
+ {
1526
+ "episode": 37376,
1527
+ "epoch": 0.3202193283070596,
1528
+ "eps": 5,
1529
+ "loss/policy_avg": 0.006841774098575115,
1530
+ "loss/value_avg": 0.22015002369880676,
1531
+ "lr": 2.15625e-06,
1532
+ "objective/entropy": -53.92641830444336,
1533
+ "objective/kl": 34.366966247558594,
1534
+ "objective/non_score_reward": -1.7183483839035034,
1535
+ "objective/rlhf_reward": 10.033120155334473,
1536
+ "objective/scores": 11.751468658447266,
1537
+ "policy/approxkl_avg": 0.003308027284219861,
1538
+ "policy/clipfrac_avg": 0.011116940528154373,
1539
+ "policy/entropy_avg": 1.41293466091156,
1540
+ "step": 73,
1541
+ "val/clipfrac_avg": 0.004825121723115444,
1542
+ "val/num_eos_tokens": 23706,
1543
+ "val/ratio": 1.0000604391098022,
1544
+ "val/ratio_var": 3.1772717647982063e-06
1545
+ },
1546
+ {
1547
+ "episode": 37888,
1548
+ "epoch": 0.3246058944482522,
1549
+ "eps": 5,
1550
+ "loss/policy_avg": 8.680112659931183e-05,
1551
+ "loss/value_avg": 0.22518639266490936,
1552
+ "lr": 2.14453125e-06,
1553
+ "objective/entropy": -51.67547607421875,
1554
+ "objective/kl": 33.763038635253906,
1555
+ "objective/non_score_reward": -1.6881520748138428,
1556
+ "objective/rlhf_reward": 10.125733375549316,
1557
+ "objective/scores": 11.813885688781738,
1558
+ "policy/approxkl_avg": 0.002383920131251216,
1559
+ "policy/clipfrac_avg": 0.012023954652249813,
1560
+ "policy/entropy_avg": 1.3859004974365234,
1561
+ "step": 74,
1562
+ "val/clipfrac_avg": 0.0035750826355069876,
1563
+ "val/num_eos_tokens": 27803,
1564
+ "val/ratio": 0.9996781945228577,
1565
+ "val/ratio_var": 4.62918205812457e-06
1566
+ },
1567
+ {
1568
+ "episode": 38400,
1569
+ "epoch": 0.3289924605894448,
1570
+ "eps": 5,
1571
+ "loss/policy_avg": 0.00683591328561306,
1572
+ "loss/value_avg": 0.19435633718967438,
1573
+ "lr": 2.1328125e-06,
1574
+ "objective/entropy": -55.22687911987305,
1575
+ "objective/kl": 33.985172271728516,
1576
+ "objective/non_score_reward": -1.69925856590271,
1577
+ "objective/rlhf_reward": 10.095266342163086,
1578
+ "objective/scores": 11.794525146484375,
1579
+ "policy/approxkl_avg": 0.0020253900438547134,
1580
+ "policy/clipfrac_avg": 0.011549219489097595,
1581
+ "policy/entropy_avg": 1.4170671701431274,
1582
+ "step": 75,
1583
+ "val/clipfrac_avg": 0.005189975723624229,
1584
+ "val/num_eos_tokens": 23555,
1585
+ "val/ratio": 1.0000282526016235,
1586
+ "val/ratio_var": 2.6403824904264184e-06
1587
+ },
1588
+ {
1589
+ "episode": 38912,
1590
+ "epoch": 0.3333790267306374,
1591
+ "eps": 5,
1592
+ "loss/policy_avg": 0.0018177703022956848,
1593
+ "loss/value_avg": 0.2192225456237793,
1594
+ "lr": 2.12109375e-06,
1595
+ "objective/entropy": -52.719966888427734,
1596
+ "objective/kl": 34.939414978027344,
1597
+ "objective/non_score_reward": -1.746970772743225,
1598
+ "objective/rlhf_reward": 9.940811157226562,
1599
+ "objective/scores": 11.687782287597656,
1600
+ "policy/approxkl_avg": 0.003144835354760289,
1601
+ "policy/clipfrac_avg": 0.011414816603064537,
1602
+ "policy/entropy_avg": 1.3688035011291504,
1603
+ "step": 76,
1604
+ "val/clipfrac_avg": 0.00428430363535881,
1605
+ "val/num_eos_tokens": 25195,
1606
+ "val/ratio": 0.9994131326675415,
1607
+ "val/ratio_var": 3.1442646104551386e-06
1608
+ },
1609
+ {
1610
+ "episode": 39424,
1611
+ "epoch": 0.33776559287183,
1612
+ "eps": 5,
1613
+ "loss/policy_avg": 0.002414613962173462,
1614
+ "loss/value_avg": 0.22705943882465363,
1615
+ "lr": 2.109375e-06,
1616
+ "objective/entropy": -52.9849853515625,
1617
+ "objective/kl": 35.76004409790039,
1618
+ "objective/non_score_reward": -1.7880022525787354,
1619
+ "objective/rlhf_reward": 10.003686904907227,
1620
+ "objective/scores": 11.791688919067383,
1621
+ "policy/approxkl_avg": 0.0024730274453759193,
1622
+ "policy/clipfrac_avg": 0.012745586223900318,
1623
+ "policy/entropy_avg": 1.3729814291000366,
1624
+ "step": 77,
1625
+ "val/clipfrac_avg": 0.004229060374200344,
1626
+ "val/num_eos_tokens": 26826,
1627
+ "val/ratio": 0.9999415278434753,
1628
+ "val/ratio_var": 5.440687800728483e-06
1629
+ },
1630
+ {
1631
+ "episode": 39936,
1632
+ "epoch": 0.3421521590130226,
1633
+ "eps": 4,
1634
+ "loss/policy_avg": 0.007603425532579422,
1635
+ "loss/value_avg": 0.20695388317108154,
1636
+ "lr": 2.09765625e-06,
1637
+ "objective/entropy": -52.369571685791016,
1638
+ "objective/kl": 35.140933990478516,
1639
+ "objective/non_score_reward": -1.7570466995239258,
1640
+ "objective/rlhf_reward": 10.074074745178223,
1641
+ "objective/scores": 11.831121444702148,
1642
+ "policy/approxkl_avg": 0.002142505254596472,
1643
+ "policy/clipfrac_avg": 0.010965963825583458,
1644
+ "policy/entropy_avg": 1.3412797451019287,
1645
+ "step": 78,
1646
+ "val/clipfrac_avg": 0.0038626650348305702,
1647
+ "val/num_eos_tokens": 23611,
1648
+ "val/ratio": 0.9996312856674194,
1649
+ "val/ratio_var": 4.994772552890936e-06
1650
+ },
1651
+ {
1652
+ "episode": 40448,
1653
+ "epoch": 0.34653872515421524,
1654
+ "eps": 4,
1655
+ "loss/policy_avg": 0.0018220320343971252,
1656
+ "loss/value_avg": 0.21862611174583435,
1657
+ "lr": 2.0859375e-06,
1658
+ "objective/entropy": -51.15296936035156,
1659
+ "objective/kl": 35.11440658569336,
1660
+ "objective/non_score_reward": -1.7557203769683838,
1661
+ "objective/rlhf_reward": 10.102788925170898,
1662
+ "objective/scores": 11.858509063720703,
1663
+ "policy/approxkl_avg": 0.0024529770016670227,
1664
+ "policy/clipfrac_avg": 0.012614256702363491,
1665
+ "policy/entropy_avg": 1.3262717723846436,
1666
+ "step": 79,
1667
+ "val/clipfrac_avg": 0.003596197348088026,
1668
+ "val/num_eos_tokens": 24645,
1669
+ "val/ratio": 1.000266194343567,
1670
+ "val/ratio_var": 6.255781045183539e-06
1671
+ },
1672
+ {
1673
+ "episode": 40960,
1674
+ "epoch": 0.35092529129540784,
1675
+ "eps": 4,
1676
+ "loss/policy_avg": 0.0023819338530302048,
1677
+ "loss/value_avg": 0.22472721338272095,
1678
+ "lr": 2.07421875e-06,
1679
+ "objective/entropy": -50.92372512817383,
1680
+ "objective/kl": 36.36195373535156,
1681
+ "objective/non_score_reward": -1.8180978298187256,
1682
+ "objective/rlhf_reward": 10.099588394165039,
1683
+ "objective/scores": 11.917686462402344,
1684
+ "policy/approxkl_avg": 0.002738222246989608,
1685
+ "policy/clipfrac_avg": 0.01162832509726286,
1686
+ "policy/entropy_avg": 1.3125125169754028,
1687
+ "step": 80,
1688
+ "val/clipfrac_avg": 0.005824130028486252,
1689
+ "val/num_eos_tokens": 22979,
1690
+ "val/ratio": 0.9995706677436829,
1691
+ "val/ratio_var": 3.720711902133189e-06
1692
+ },
1693
+ {
1694
+ "episode": 41472,
1695
+ "epoch": 0.35531185743660043,
1696
+ "eps": 4,
1697
+ "loss/policy_avg": 0.004819877445697784,
1698
+ "loss/value_avg": 0.22295227646827698,
1699
+ "lr": 2.0625e-06,
1700
+ "objective/entropy": -50.77427291870117,
1701
+ "objective/kl": 35.64800262451172,
1702
+ "objective/non_score_reward": -1.782400131225586,
1703
+ "objective/rlhf_reward": 9.891586303710938,
1704
+ "objective/scores": 11.673986434936523,
1705
+ "policy/approxkl_avg": 0.0026750012766569853,
1706
+ "policy/clipfrac_avg": 0.012281844392418861,
1707
+ "policy/entropy_avg": 1.3098926544189453,
1708
+ "step": 81,
1709
+ "val/clipfrac_avg": 0.003562133526429534,
1710
+ "val/num_eos_tokens": 22481,
1711
+ "val/ratio": 1.0002925395965576,
1712
+ "val/ratio_var": 3.08896505885059e-06
1713
+ },
1714
+ {
1715
+ "episode": 41984,
1716
+ "epoch": 0.359698423577793,
1717
+ "eps": 4,
1718
+ "loss/policy_avg": 0.003094850108027458,
1719
+ "loss/value_avg": 0.20467893779277802,
1720
+ "lr": 2.05078125e-06,
1721
+ "objective/entropy": -49.50064468383789,
1722
+ "objective/kl": 35.082889556884766,
1723
+ "objective/non_score_reward": -1.7541444301605225,
1724
+ "objective/rlhf_reward": 10.12310791015625,
1725
+ "objective/scores": 11.877252578735352,
1726
+ "policy/approxkl_avg": 0.0031247385777533054,
1727
+ "policy/clipfrac_avg": 0.010663332417607307,
1728
+ "policy/entropy_avg": 1.2899810075759888,
1729
+ "step": 82,
1730
+ "val/clipfrac_avg": 0.003626835998147726,
1731
+ "val/num_eos_tokens": 23688,
1732
+ "val/ratio": 0.9997555613517761,
1733
+ "val/ratio_var": 4.246959178999532e-06
1734
+ },
1735
+ {
1736
+ "episode": 42496,
1737
+ "epoch": 0.3640849897189856,
1738
+ "eps": 4,
1739
+ "loss/policy_avg": 0.006415246054530144,
1740
+ "loss/value_avg": 0.2166299670934677,
1741
+ "lr": 2.0390625e-06,
1742
+ "objective/entropy": -49.80467987060547,
1743
+ "objective/kl": 35.149269104003906,
1744
+ "objective/non_score_reward": -1.7574634552001953,
1745
+ "objective/rlhf_reward": 9.97119140625,
1746
+ "objective/scores": 11.728654861450195,
1747
+ "policy/approxkl_avg": 0.002845556242391467,
1748
+ "policy/clipfrac_avg": 0.013346903957426548,
1749
+ "policy/entropy_avg": 1.293116569519043,
1750
+ "step": 83,
1751
+ "val/clipfrac_avg": 0.004545033443719149,
1752
+ "val/num_eos_tokens": 24437,
1753
+ "val/ratio": 0.9999039173126221,
1754
+ "val/ratio_var": 6.112253231549403e-06
1755
+ },
1756
+ {
1757
+ "episode": 43008,
1758
+ "epoch": 0.3684715558601782,
1759
+ "eps": 4,
1760
+ "loss/policy_avg": 0.004797011613845825,
1761
+ "loss/value_avg": 0.2133270502090454,
1762
+ "lr": 2.02734375e-06,
1763
+ "objective/entropy": -51.009857177734375,
1764
+ "objective/kl": 35.441497802734375,
1765
+ "objective/non_score_reward": -1.7720749378204346,
1766
+ "objective/rlhf_reward": 10.048232078552246,
1767
+ "objective/scores": 11.820306777954102,
1768
+ "policy/approxkl_avg": 0.002602731343358755,
1769
+ "policy/clipfrac_avg": 0.012205180712044239,
1770
+ "policy/entropy_avg": 1.3177757263183594,
1771
+ "step": 84,
1772
+ "val/clipfrac_avg": 0.004008126445114613,
1773
+ "val/num_eos_tokens": 24990,
1774
+ "val/ratio": 1.0002726316452026,
1775
+ "val/ratio_var": 5.879127456864808e-06
1776
+ },
1777
+ {
1778
+ "episode": 43520,
1779
+ "epoch": 0.3728581220013708,
1780
+ "eps": 4,
1781
+ "loss/policy_avg": 0.004787225276231766,
1782
+ "loss/value_avg": 0.213937908411026,
1783
+ "lr": 2.015625e-06,
1784
+ "objective/entropy": -48.71440124511719,
1785
+ "objective/kl": 35.934600830078125,
1786
+ "objective/non_score_reward": -1.7967300415039062,
1787
+ "objective/rlhf_reward": 10.027629852294922,
1788
+ "objective/scores": 11.824359893798828,
1789
+ "policy/approxkl_avg": 0.0032878173515200615,
1790
+ "policy/clipfrac_avg": 0.01084982417523861,
1791
+ "policy/entropy_avg": 1.26298189163208,
1792
+ "step": 85,
1793
+ "val/clipfrac_avg": 0.004145544022321701,
1794
+ "val/num_eos_tokens": 25423,
1795
+ "val/ratio": 0.9995351433753967,
1796
+ "val/ratio_var": 4.676774096878944e-06
1797
+ },
1798
+ {
1799
+ "episode": 44032,
1800
+ "epoch": 0.3772446881425634,
1801
+ "eps": 4,
1802
+ "loss/policy_avg": 0.009194480255246162,
1803
+ "loss/value_avg": 0.19612735509872437,
1804
+ "lr": 2.00390625e-06,
1805
+ "objective/entropy": -50.50733947753906,
1806
+ "objective/kl": 35.42319869995117,
1807
+ "objective/non_score_reward": -1.7711601257324219,
1808
+ "objective/rlhf_reward": 10.026796340942383,
1809
+ "objective/scores": 11.797956466674805,
1810
+ "policy/approxkl_avg": 0.0025596285704523325,
1811
+ "policy/clipfrac_avg": 0.011088266968727112,
1812
+ "policy/entropy_avg": 1.2955846786499023,
1813
+ "step": 86,
1814
+ "val/clipfrac_avg": 0.002618049271404743,
1815
+ "val/num_eos_tokens": 24385,
1816
+ "val/ratio": 1.0004734992980957,
1817
+ "val/ratio_var": 7.4206191129633226e-06
1818
+ },
1819
+ {
1820
+ "episode": 44544,
1821
+ "epoch": 0.381631254283756,
1822
+ "eps": 4,
1823
+ "loss/policy_avg": 0.0011375932954251766,
1824
+ "loss/value_avg": 0.19349417090415955,
1825
+ "lr": 1.9921875e-06,
1826
+ "objective/entropy": -48.215614318847656,
1827
+ "objective/kl": 36.09561538696289,
1828
+ "objective/non_score_reward": -1.8047807216644287,
1829
+ "objective/rlhf_reward": 10.043390274047852,
1830
+ "objective/scores": 11.84817123413086,
1831
+ "policy/approxkl_avg": 0.002257507061585784,
1832
+ "policy/clipfrac_avg": 0.011135649867355824,
1833
+ "policy/entropy_avg": 1.2575373649597168,
1834
+ "step": 87,
1835
+ "val/clipfrac_avg": 0.0035062048118561506,
1836
+ "val/num_eos_tokens": 26536,
1837
+ "val/ratio": 1.0005232095718384,
1838
+ "val/ratio_var": 7.836673830752261e-06
1839
+ },
1840
+ {
1841
+ "episode": 45056,
1842
+ "epoch": 0.3860178204249486,
1843
+ "eps": 4,
1844
+ "loss/policy_avg": 0.004499746486544609,
1845
+ "loss/value_avg": 0.20375367999076843,
1846
+ "lr": 1.98046875e-06,
1847
+ "objective/entropy": -48.469757080078125,
1848
+ "objective/kl": 35.357276916503906,
1849
+ "objective/non_score_reward": -1.7678639888763428,
1850
+ "objective/rlhf_reward": 10.101089477539062,
1851
+ "objective/scores": 11.868953704833984,
1852
+ "policy/approxkl_avg": 0.0027951907832175493,
1853
+ "policy/clipfrac_avg": 0.011439654044806957,
1854
+ "policy/entropy_avg": 1.2559094429016113,
1855
+ "step": 88,
1856
+ "val/clipfrac_avg": 0.0032751811668276787,
1857
+ "val/num_eos_tokens": 24904,
1858
+ "val/ratio": 0.999579668045044,
1859
+ "val/ratio_var": 3.818234290520195e-06
1860
+ },
1861
+ {
1862
+ "episode": 45568,
1863
+ "epoch": 0.3904043865661412,
1864
+ "eps": 4,
1865
+ "loss/policy_avg": 0.004391288384795189,
1866
+ "loss/value_avg": 0.19215956330299377,
1867
+ "lr": 1.96875e-06,
1868
+ "objective/entropy": -47.49878692626953,
1869
+ "objective/kl": 35.51283645629883,
1870
+ "objective/non_score_reward": -1.775641918182373,
1871
+ "objective/rlhf_reward": 10.171991348266602,
1872
+ "objective/scores": 11.947633743286133,
1873
+ "policy/approxkl_avg": 0.003075700718909502,
1874
+ "policy/clipfrac_avg": 0.011072011664509773,
1875
+ "policy/entropy_avg": 1.2150685787200928,
1876
+ "step": 89,
1877
+ "val/clipfrac_avg": 0.0027455922681838274,
1878
+ "val/num_eos_tokens": 23593,
1879
+ "val/ratio": 0.9998563528060913,
1880
+ "val/ratio_var": 3.37912251779926e-06
1881
+ },
1882
+ {
1883
+ "episode": 46080,
1884
+ "epoch": 0.3947909527073338,
1885
+ "eps": 4,
1886
+ "loss/policy_avg": 0.0004821214824914932,
1887
+ "loss/value_avg": 0.21164780855178833,
1888
+ "lr": 1.95703125e-06,
1889
+ "objective/entropy": -49.131282806396484,
1890
+ "objective/kl": 35.567405700683594,
1891
+ "objective/non_score_reward": -1.7783703804016113,
1892
+ "objective/rlhf_reward": 10.056915283203125,
1893
+ "objective/scores": 11.835285186767578,
1894
+ "policy/approxkl_avg": 0.0025346523616462946,
1895
+ "policy/clipfrac_avg": 0.01132851280272007,
1896
+ "policy/entropy_avg": 1.2548737525939941,
1897
+ "step": 90,
1898
+ "val/clipfrac_avg": 0.0035444346722215414,
1899
+ "val/num_eos_tokens": 23956,
1900
+ "val/ratio": 1.0001569986343384,
1901
+ "val/ratio_var": 4.048593382321997e-06
1902
+ },
1903
+ {
1904
+ "episode": 46592,
1905
+ "epoch": 0.39917751884852637,
1906
+ "eps": 4,
1907
+ "loss/policy_avg": 0.002457218011841178,
1908
+ "loss/value_avg": 0.20488248765468597,
1909
+ "lr": 1.9453125e-06,
1910
+ "objective/entropy": -50.223350524902344,
1911
+ "objective/kl": 34.380897521972656,
1912
+ "objective/non_score_reward": -1.719044804573059,
1913
+ "objective/rlhf_reward": 10.15352725982666,
1914
+ "objective/scores": 11.87257194519043,
1915
+ "policy/approxkl_avg": 0.0025883247144520283,
1916
+ "policy/clipfrac_avg": 0.01095154695212841,
1917
+ "policy/entropy_avg": 1.2806235551834106,
1918
+ "step": 91,
1919
+ "val/clipfrac_avg": 0.0030915343668311834,
1920
+ "val/num_eos_tokens": 24619,
1921
+ "val/ratio": 1.000389814376831,
1922
+ "val/ratio_var": 5.21131823916221e-06
1923
+ },
1924
+ {
1925
+ "episode": 47104,
1926
+ "epoch": 0.40356408498971896,
1927
+ "eps": 4,
1928
+ "loss/policy_avg": 0.009746008552610874,
1929
+ "loss/value_avg": 0.19489187002182007,
1930
+ "lr": 1.93359375e-06,
1931
+ "objective/entropy": -49.78590774536133,
1932
+ "objective/kl": 34.79002380371094,
1933
+ "objective/non_score_reward": -1.7395012378692627,
1934
+ "objective/rlhf_reward": 10.149117469787598,
1935
+ "objective/scores": 11.888618469238281,
1936
+ "policy/approxkl_avg": 0.0028350851498544216,
1937
+ "policy/clipfrac_avg": 0.012067590840160847,
1938
+ "policy/entropy_avg": 1.321395993232727,
1939
+ "step": 92,
1940
+ "val/clipfrac_avg": 0.0034278968814760447,
1941
+ "val/num_eos_tokens": 24707,
1942
+ "val/ratio": 1.000132441520691,
1943
+ "val/ratio_var": 3.817275683104526e-06
1944
+ },
1945
+ {
1946
+ "episode": 47616,
1947
+ "epoch": 0.40795065113091156,
1948
+ "eps": 4,
1949
+ "loss/policy_avg": 0.007891923189163208,
1950
+ "loss/value_avg": 0.18601296842098236,
1951
+ "lr": 1.921875e-06,
1952
+ "objective/entropy": -51.31802749633789,
1953
+ "objective/kl": 33.889923095703125,
1954
+ "objective/non_score_reward": -1.6944962739944458,
1955
+ "objective/rlhf_reward": 10.027242660522461,
1956
+ "objective/scores": 11.721738815307617,
1957
+ "policy/approxkl_avg": 0.0023833350278437138,
1958
+ "policy/clipfrac_avg": 0.011782050132751465,
1959
+ "policy/entropy_avg": 1.3522666692733765,
1960
+ "step": 93,
1961
+ "val/clipfrac_avg": 0.002397140022367239,
1962
+ "val/num_eos_tokens": 23485,
1963
+ "val/ratio": 0.9999991655349731,
1964
+ "val/ratio_var": 4.472914497455349e-06
1965
+ },
1966
+ {
1967
+ "episode": 48128,
1968
+ "epoch": 0.4123372172721042,
1969
+ "eps": 4,
1970
+ "loss/policy_avg": 0.008123669773340225,
1971
+ "loss/value_avg": 0.19486477971076965,
1972
+ "lr": 1.91015625e-06,
1973
+ "objective/entropy": -51.119022369384766,
1974
+ "objective/kl": 33.58214569091797,
1975
+ "objective/non_score_reward": -1.679107427597046,
1976
+ "objective/rlhf_reward": 10.173272132873535,
1977
+ "objective/scores": 11.85237979888916,
1978
+ "policy/approxkl_avg": 0.002184953773394227,
1979
+ "policy/clipfrac_avg": 0.010806472972035408,
1980
+ "policy/entropy_avg": 1.3613462448120117,
1981
+ "step": 94,
1982
+ "val/clipfrac_avg": 0.0037152436561882496,
1983
+ "val/num_eos_tokens": 24130,
1984
+ "val/ratio": 1.0001533031463623,
1985
+ "val/ratio_var": 7.5301450124243274e-06
1986
+ },
1987
+ {
1988
+ "episode": 48640,
1989
+ "epoch": 0.4167237834132968,
1990
+ "eps": 4,
1991
+ "loss/policy_avg": 0.006111526861786842,
1992
+ "loss/value_avg": 0.18791520595550537,
1993
+ "lr": 1.8984375e-06,
1994
+ "objective/entropy": -50.709373474121094,
1995
+ "objective/kl": 34.28173065185547,
1996
+ "objective/non_score_reward": -1.7140867710113525,
1997
+ "objective/rlhf_reward": 10.056640625,
1998
+ "objective/scores": 11.770727157592773,
1999
+ "policy/approxkl_avg": 0.0028409743681550026,
2000
+ "policy/clipfrac_avg": 0.010896073654294014,
2001
+ "policy/entropy_avg": 1.3718466758728027,
2002
+ "step": 95,
2003
+ "val/clipfrac_avg": 0.0028323421720415354,
2004
+ "val/num_eos_tokens": 24401,
2005
+ "val/ratio": 1.0004030466079712,
2006
+ "val/ratio_var": 5.722152764064958e-06
2007
+ },
2008
+ {
2009
+ "episode": 49152,
2010
+ "epoch": 0.4211103495544894,
2011
+ "eps": 4,
2012
+ "loss/policy_avg": 0.006566773168742657,
2013
+ "loss/value_avg": 0.21990007162094116,
2014
+ "lr": 1.8867187500000001e-06,
2015
+ "objective/entropy": -51.272518157958984,
2016
+ "objective/kl": 33.29633712768555,
2017
+ "objective/non_score_reward": -1.6648168563842773,
2018
+ "objective/rlhf_reward": 10.015409469604492,
2019
+ "objective/scores": 11.68022632598877,
2020
+ "policy/approxkl_avg": 0.002582971705123782,
2021
+ "policy/clipfrac_avg": 0.012000022456049919,
2022
+ "policy/entropy_avg": 1.3877381086349487,
2023
+ "step": 96,
2024
+ "val/clipfrac_avg": 0.00376589922234416,
2025
+ "val/num_eos_tokens": 26887,
2026
+ "val/ratio": 0.9997319579124451,
2027
+ "val/ratio_var": 3.3458989037171705e-06
2028
+ },
2029
+ {
2030
+ "episode": 49664,
2031
+ "epoch": 0.425496915695682,
2032
+ "eps": 4,
2033
+ "loss/policy_avg": 0.00574074499309063,
2034
+ "loss/value_avg": 0.20088031888008118,
2035
+ "lr": 1.875e-06,
2036
+ "objective/entropy": -50.57700729370117,
2037
+ "objective/kl": 33.54800033569336,
2038
+ "objective/non_score_reward": -1.67739999294281,
2039
+ "objective/rlhf_reward": 10.143403053283691,
2040
+ "objective/scores": 11.820802688598633,
2041
+ "policy/approxkl_avg": 0.00242623221129179,
2042
+ "policy/clipfrac_avg": 0.011794717982411385,
2043
+ "policy/entropy_avg": 1.3870760202407837,
2044
+ "step": 97,
2045
+ "val/clipfrac_avg": 0.002550810342654586,
2046
+ "val/num_eos_tokens": 26478,
2047
+ "val/ratio": 1.0001178979873657,
2048
+ "val/ratio_var": 3.899105195159791e-06
2049
+ },
2050
+ {
2051
+ "episode": 50176,
2052
+ "epoch": 0.4298834818368746,
2053
+ "eps": 4,
2054
+ "loss/policy_avg": 0.002134094014763832,
2055
+ "loss/value_avg": 0.2105971872806549,
2056
+ "lr": 1.86328125e-06,
2057
+ "objective/entropy": -50.969947814941406,
2058
+ "objective/kl": 32.82494354248047,
2059
+ "objective/non_score_reward": -1.6412471532821655,
2060
+ "objective/rlhf_reward": 10.093246459960938,
2061
+ "objective/scores": 11.734493255615234,
2062
+ "policy/approxkl_avg": 0.002924936590716243,
2063
+ "policy/clipfrac_avg": 0.012574190273880959,
2064
+ "policy/entropy_avg": 1.4206920862197876,
2065
+ "step": 98,
2066
+ "val/clipfrac_avg": 0.0022323690354824066,
2067
+ "val/num_eos_tokens": 25788,
2068
+ "val/ratio": 1.0000821352005005,
2069
+ "val/ratio_var": 4.282260761101497e-06
2070
+ },
2071
+ {
2072
+ "episode": 50688,
2073
+ "epoch": 0.4342700479780672,
2074
+ "eps": 4,
2075
+ "loss/policy_avg": 0.002779986709356308,
2076
+ "loss/value_avg": 0.21304886043071747,
2077
+ "lr": 1.8515625000000001e-06,
2078
+ "objective/entropy": -48.35157012939453,
2079
+ "objective/kl": 34.29575729370117,
2080
+ "objective/non_score_reward": -1.7147879600524902,
2081
+ "objective/rlhf_reward": 10.013525009155273,
2082
+ "objective/scores": 11.728313446044922,
2083
+ "policy/approxkl_avg": 0.0031486451625823975,
2084
+ "policy/clipfrac_avg": 0.011947648599743843,
2085
+ "policy/entropy_avg": 1.3698291778564453,
2086
+ "step": 99,
2087
+ "val/clipfrac_avg": 0.003086227923631668,
2088
+ "val/num_eos_tokens": 27226,
2089
+ "val/ratio": 1.0002012252807617,
2090
+ "val/ratio_var": 3.848301275866106e-06
2091
+ },
2092
+ {
2093
+ "episode": 51200,
2094
+ "epoch": 0.43865661411925977,
2095
+ "eps": 4,
2096
+ "loss/policy_avg": 0.002698383294045925,
2097
+ "loss/value_avg": 0.1958826333284378,
2098
+ "lr": 1.83984375e-06,
2099
+ "objective/entropy": -51.114051818847656,
2100
+ "objective/kl": 33.804664611816406,
2101
+ "objective/non_score_reward": -1.6902332305908203,
2102
+ "objective/rlhf_reward": 10.000195503234863,
2103
+ "objective/scores": 11.690428733825684,
2104
+ "policy/approxkl_avg": 0.002505134791135788,
2105
+ "policy/clipfrac_avg": 0.013553831726312637,
2106
+ "policy/entropy_avg": 1.435497522354126,
2107
+ "step": 100,
2108
+ "val/clipfrac_avg": 0.0029125860892236233,
2109
+ "val/num_eos_tokens": 27134,
2110
+ "val/ratio": 0.9996867179870605,
2111
+ "val/ratio_var": 4.967731001670472e-06
2112
+ },
2113
+ {
2114
+ "episode": 51712,
2115
+ "epoch": 0.44304318026045236,
2116
+ "eps": 4,
2117
+ "loss/policy_avg": 0.00523423682898283,
2118
+ "loss/value_avg": 0.20203697681427002,
2119
+ "lr": 1.828125e-06,
2120
+ "objective/entropy": -53.56639862060547,
2121
+ "objective/kl": 32.214115142822266,
2122
+ "objective/non_score_reward": -1.610705852508545,
2123
+ "objective/rlhf_reward": 10.233478546142578,
2124
+ "objective/scores": 11.844184875488281,
2125
+ "policy/approxkl_avg": 0.003430293407291174,
2126
+ "policy/clipfrac_avg": 0.013302096165716648,
2127
+ "policy/entropy_avg": 1.4647541046142578,
2128
+ "step": 101,
2129
+ "val/clipfrac_avg": 0.0026878612115979195,
2130
+ "val/num_eos_tokens": 24773,
2131
+ "val/ratio": 1.0002996921539307,
2132
+ "val/ratio_var": 1.0531030056881718e-05
2133
+ },
2134
+ {
2135
+ "episode": 52224,
2136
+ "epoch": 0.44742974640164496,
2137
+ "eps": 4,
2138
+ "loss/policy_avg": 0.007096525281667709,
2139
+ "loss/value_avg": 0.2158614844083786,
2140
+ "lr": 1.81640625e-06,
2141
+ "objective/entropy": -51.106422424316406,
2142
+ "objective/kl": 33.817195892333984,
2143
+ "objective/non_score_reward": -1.6908597946166992,
2144
+ "objective/rlhf_reward": 10.014439582824707,
2145
+ "objective/scores": 11.705299377441406,
2146
+ "policy/approxkl_avg": 0.0026541282422840595,
2147
+ "policy/clipfrac_avg": 0.012738144025206566,
2148
+ "policy/entropy_avg": 1.4463238716125488,
2149
+ "step": 102,
2150
+ "val/clipfrac_avg": 0.0025819321162998676,
2151
+ "val/num_eos_tokens": 26794,
2152
+ "val/ratio": 1.0001683235168457,
2153
+ "val/ratio_var": 4.307699327910086e-06
2154
+ },
2155
+ {
2156
+ "episode": 52736,
2157
+ "epoch": 0.45181631254283755,
2158
+ "eps": 4,
2159
+ "loss/policy_avg": 0.0076528494246304035,
2160
+ "loss/value_avg": 0.22683024406433105,
2161
+ "lr": 1.8046875e-06,
2162
+ "objective/entropy": -51.95180130004883,
2163
+ "objective/kl": 32.943721771240234,
2164
+ "objective/non_score_reward": -1.647186040878296,
2165
+ "objective/rlhf_reward": 9.92637825012207,
2166
+ "objective/scores": 11.573564529418945,
2167
+ "policy/approxkl_avg": 0.002516075037419796,
2168
+ "policy/clipfrac_avg": 0.012894165702164173,
2169
+ "policy/entropy_avg": 1.4716830253601074,
2170
+ "step": 103,
2171
+ "val/clipfrac_avg": 0.003462804015725851,
2172
+ "val/num_eos_tokens": 27258,
2173
+ "val/ratio": 0.9998781681060791,
2174
+ "val/ratio_var": 3.6184294458507793e-06
2175
+ },
2176
+ {
2177
+ "episode": 53248,
2178
+ "epoch": 0.45620287868403014,
2179
+ "eps": 4,
2180
+ "loss/policy_avg": 0.0059026554226875305,
2181
+ "loss/value_avg": 0.18866774439811707,
2182
+ "lr": 1.79296875e-06,
2183
+ "objective/entropy": -51.065303802490234,
2184
+ "objective/kl": 33.99561309814453,
2185
+ "objective/non_score_reward": -1.699780821800232,
2186
+ "objective/rlhf_reward": 10.005899429321289,
2187
+ "objective/scores": 11.705679893493652,
2188
+ "policy/approxkl_avg": 0.003799548838287592,
2189
+ "policy/clipfrac_avg": 0.011900994926691055,
2190
+ "policy/entropy_avg": 1.453720211982727,
2191
+ "step": 104,
2192
+ "val/clipfrac_avg": 0.002593266312032938,
2193
+ "val/num_eos_tokens": 25703,
2194
+ "val/ratio": 1.0001416206359863,
2195
+ "val/ratio_var": 5.38453923581983e-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 sha256:44d16ae9aae200635358494ccef464ffa9c9751a418a133a3449130c00833108
3
+ size 6392
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 sha256:8c27e7dd6cfaef2fb26e08b0aa982498eada5d1b699caafaf6e8abf371d13dd2
3
+ size 1621370224
checkpoint-156/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b9635a4f3a9fc4e7054223ac9f7846d8a85e753f696c90b5dcc3a7e095e6315
3
+ size 6073758958
checkpoint-156/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:645da38c17f04d14d7af6ddabb6011cd438862f9469dee12386401af710cc078
3
+ size 14244
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 sha256:44d16ae9aae200635358494ccef464ffa9c9751a418a133a3449130c00833108
3
+ size 6392
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 sha256:0b04ee1587f3e206dc3c82e4eaadfe6f85ff6da7486a5fe7f99e1c92ee4dba31
3
+ size 1621370224
checkpoint-208/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:896cde5bf67b921cea844a05fb065d4493bfb199db732b2a5b561a7b1dc77aaa
3
+ size 6073758958
checkpoint-208/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c16fe4708dd9f189934c63f0d9680397beedb29bfe6a8887f92b7b41b56419e
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 sha256:44d16ae9aae200635358494ccef464ffa9c9751a418a133a3449130c00833108
3
+ size 6392
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 sha256:2ea91d2a3ef45ab61ec81a23be6c82fe7a0ef6e716c10a2a10bebe80b938d3d0
3
+ size 1621370224
checkpoint-256/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2858120f05e255fd66f7867efd1538ba51c400a1dfcddfbaeaa95e04f34cbc5a
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:acb0d392a4380f3c943fd3f0f02ff2157a6c3a26f26cdf857012dc74d9daca2e
3
+ size 14244
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 sha256:44d16ae9aae200635358494ccef464ffa9c9751a418a133a3449130c00833108
3
+ size 6392
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:f1f8d4278bb773b46d73935b6a7ab4651c323a52dd26e04c951ef5ac816df564
3
+ size 1621370224
checkpoint-52/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7dd1be996d78aea757034ea1242b9cbab10a26e7eab540429ee0e631397317df
3
+ size 6073758958
checkpoint-52/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9009c55c0f0a06fc27c5527322bbb09377b6cc6f3d192808674eb6435d8cd493
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