Muqeeth commited on
Commit
8b51d95
·
verified ·
1 Parent(s): e6953b5

Training in progress, step 104, checkpoint

Browse files
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 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 0,
4
+ "transformers_version": "4.50.3"
5
+ }
checkpoint-104/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02038f98f1e3a0e36302b5121827f0d95b24a0a3b76374ccf68ae16634ea1d79
3
+ size 1621370224
checkpoint-104/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a6c488bb968be2be2d7df38d47f32215253202376a75fdcbc978b3feabb5b0d7
3
+ size 3242916026
checkpoint-104/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a7aee469c62ec38fa7239d0e96663bd2470dcbcb2e3244e310cb1bcb87ce53b2
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 +1,2115 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "episode": 53248,
6
+ "epoch": 0.22810143934201507,
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": 0.0021932830705962986,
16
+ "eps": 6,
17
+ "loss/policy_avg": 0.12750720977783203,
18
+ "lr": 3e-06,
19
+ "objective/entropy": -47.1529541015625,
20
+ "objective/kl": 0.3561969995498657,
21
+ "objective/non_score_reward": 0.0,
22
+ "objective/rlhf_reward": 4.533186912536621,
23
+ "objective/scores": 4.533186912536621,
24
+ "policy/approxkl_avg": 0.12109499424695969,
25
+ "policy/clipfrac_avg": 0.359375,
26
+ "policy/entropy_avg": 0.934253454208374,
27
+ "step": 0,
28
+ "val/clipfrac_avg": 0.0,
29
+ "val/num_eos_tokens": 2466,
30
+ "val/ratio": 1.000074863433838,
31
+ "val/ratio_var": 1.5309906302718446e-06
32
+ },
33
+ {
34
+ "episode": 1024,
35
+ "epoch": 0.004386566141192597,
36
+ "eps": 6,
37
+ "loss/policy_avg": 0.14252598583698273,
38
+ "lr": 2.9882812500000002e-06,
39
+ "objective/entropy": -45.39274597167969,
40
+ "objective/kl": 0.6608734130859375,
41
+ "objective/non_score_reward": 0.0,
42
+ "objective/rlhf_reward": 4.582604885101318,
43
+ "objective/scores": 4.582604885101318,
44
+ "policy/approxkl_avg": 0.12793150544166565,
45
+ "policy/clipfrac_avg": 0.34765625,
46
+ "policy/entropy_avg": 0.9144482612609863,
47
+ "step": 1,
48
+ "val/clipfrac_avg": 0.0,
49
+ "val/num_eos_tokens": 2017,
50
+ "val/ratio": 0.9999631643295288,
51
+ "val/ratio_var": 2.3689563022344373e-06
52
+ },
53
+ {
54
+ "episode": 1536,
55
+ "epoch": 0.006579849211788897,
56
+ "eps": 6,
57
+ "loss/policy_avg": 0.15036526322364807,
58
+ "lr": 2.9765625e-06,
59
+ "objective/entropy": -45.15424728393555,
60
+ "objective/kl": 1.1336889266967773,
61
+ "objective/non_score_reward": 0.0,
62
+ "objective/rlhf_reward": 5.002047538757324,
63
+ "objective/scores": 5.002047538757324,
64
+ "policy/approxkl_avg": 0.12406677007675171,
65
+ "policy/clipfrac_avg": 0.3125,
66
+ "policy/entropy_avg": 0.9146890640258789,
67
+ "step": 2,
68
+ "val/clipfrac_avg": 0.0,
69
+ "val/num_eos_tokens": 1670,
70
+ "val/ratio": 0.9997891783714294,
71
+ "val/ratio_var": 1.4586486258849618e-06
72
+ },
73
+ {
74
+ "episode": 2048,
75
+ "epoch": 0.008773132282385195,
76
+ "eps": 6,
77
+ "loss/policy_avg": 0.1377829760313034,
78
+ "lr": 2.96484375e-06,
79
+ "objective/entropy": -42.19343948364258,
80
+ "objective/kl": 1.5318495035171509,
81
+ "objective/non_score_reward": 0.0,
82
+ "objective/rlhf_reward": 4.917952537536621,
83
+ "objective/scores": 4.917952537536621,
84
+ "policy/approxkl_avg": 0.11953485012054443,
85
+ "policy/clipfrac_avg": 0.306640625,
86
+ "policy/entropy_avg": 0.8580580949783325,
87
+ "step": 3,
88
+ "val/clipfrac_avg": 0.0,
89
+ "val/num_eos_tokens": 1605,
90
+ "val/ratio": 0.9999769926071167,
91
+ "val/ratio_var": 1.8137678807761404e-06
92
+ },
93
+ {
94
+ "episode": 2560,
95
+ "epoch": 0.010966415352981495,
96
+ "eps": 6,
97
+ "loss/policy_avg": 0.16134658455848694,
98
+ "lr": 2.953125e-06,
99
+ "objective/entropy": -47.00198745727539,
100
+ "objective/kl": 2.190272808074951,
101
+ "objective/non_score_reward": 0.0,
102
+ "objective/rlhf_reward": 5.18130350112915,
103
+ "objective/scores": 5.18130350112915,
104
+ "policy/approxkl_avg": 0.12579435110092163,
105
+ "policy/clipfrac_avg": 0.310546875,
106
+ "policy/entropy_avg": 0.768105149269104,
107
+ "step": 4,
108
+ "val/clipfrac_avg": 0.0,
109
+ "val/num_eos_tokens": 1625,
110
+ "val/ratio": 0.9999212026596069,
111
+ "val/ratio_var": 1.7392967492924072e-06
112
+ },
113
+ {
114
+ "episode": 3072,
115
+ "epoch": 0.013159698423577794,
116
+ "eps": 6,
117
+ "loss/policy_avg": 0.13819235563278198,
118
+ "lr": 2.94140625e-06,
119
+ "objective/entropy": -49.404972076416016,
120
+ "objective/kl": 2.7731494903564453,
121
+ "objective/non_score_reward": 0.0,
122
+ "objective/rlhf_reward": 5.2876763343811035,
123
+ "objective/scores": 5.2876763343811035,
124
+ "policy/approxkl_avg": 0.1179528534412384,
125
+ "policy/clipfrac_avg": 0.3359375,
126
+ "policy/entropy_avg": 0.7313921451568604,
127
+ "step": 5,
128
+ "val/clipfrac_avg": 0.0,
129
+ "val/num_eos_tokens": 1740,
130
+ "val/ratio": 1.0000691413879395,
131
+ "val/ratio_var": 1.3418108437690535e-06
132
+ },
133
+ {
134
+ "episode": 3584,
135
+ "epoch": 0.015352981494174092,
136
+ "eps": 6,
137
+ "loss/policy_avg": 0.21166321635246277,
138
+ "lr": 2.9296875e-06,
139
+ "objective/entropy": -50.94463348388672,
140
+ "objective/kl": 3.567718982696533,
141
+ "objective/non_score_reward": 0.0,
142
+ "objective/rlhf_reward": 5.437840461730957,
143
+ "objective/scores": 5.437840461730957,
144
+ "policy/approxkl_avg": 0.1175212487578392,
145
+ "policy/clipfrac_avg": 0.3046875,
146
+ "policy/entropy_avg": 0.6900524497032166,
147
+ "step": 6,
148
+ "val/clipfrac_avg": 0.0,
149
+ "val/num_eos_tokens": 1894,
150
+ "val/ratio": 1.000088095664978,
151
+ "val/ratio_var": 1.8867278868128778e-06
152
+ },
153
+ {
154
+ "episode": 4096,
155
+ "epoch": 0.01754626456477039,
156
+ "eps": 6,
157
+ "loss/policy_avg": 0.23977062106132507,
158
+ "lr": 2.91796875e-06,
159
+ "objective/entropy": -55.779144287109375,
160
+ "objective/kl": 4.140018939971924,
161
+ "objective/non_score_reward": 0.0,
162
+ "objective/rlhf_reward": 5.744781494140625,
163
+ "objective/scores": 5.744781494140625,
164
+ "policy/approxkl_avg": 0.1309545934200287,
165
+ "policy/clipfrac_avg": 0.333984375,
166
+ "policy/entropy_avg": 0.633447527885437,
167
+ "step": 7,
168
+ "val/clipfrac_avg": 0.0,
169
+ "val/num_eos_tokens": 2119,
170
+ "val/ratio": 0.999936580657959,
171
+ "val/ratio_var": 1.55381121658138e-06
172
+ },
173
+ {
174
+ "episode": 4608,
175
+ "epoch": 0.01973954763536669,
176
+ "eps": 6,
177
+ "loss/policy_avg": 0.2528630793094635,
178
+ "lr": 2.90625e-06,
179
+ "objective/entropy": -61.04898452758789,
180
+ "objective/kl": 4.573873519897461,
181
+ "objective/non_score_reward": 0.0,
182
+ "objective/rlhf_reward": 5.735022068023682,
183
+ "objective/scores": 5.735022068023682,
184
+ "policy/approxkl_avg": 0.11911365389823914,
185
+ "policy/clipfrac_avg": 0.28515625,
186
+ "policy/entropy_avg": 0.5363410711288452,
187
+ "step": 8,
188
+ "val/clipfrac_avg": 0.0,
189
+ "val/num_eos_tokens": 2481,
190
+ "val/ratio": 1.000119686126709,
191
+ "val/ratio_var": 1.5850573618081398e-06
192
+ },
193
+ {
194
+ "episode": 5120,
195
+ "epoch": 0.02193283070596299,
196
+ "eps": 6,
197
+ "loss/policy_avg": 0.14096947014331818,
198
+ "lr": 2.89453125e-06,
199
+ "objective/entropy": -62.552093505859375,
200
+ "objective/kl": 5.485900402069092,
201
+ "objective/non_score_reward": 0.0,
202
+ "objective/rlhf_reward": 5.751631259918213,
203
+ "objective/scores": 5.751631259918213,
204
+ "policy/approxkl_avg": 0.11631271243095398,
205
+ "policy/clipfrac_avg": 0.31640625,
206
+ "policy/entropy_avg": 0.5053629875183105,
207
+ "step": 9,
208
+ "val/clipfrac_avg": 0.0,
209
+ "val/num_eos_tokens": 2779,
210
+ "val/ratio": 0.9999102354049683,
211
+ "val/ratio_var": 1.997669187403517e-06
212
+ },
213
+ {
214
+ "episode": 5632,
215
+ "epoch": 0.024126113776559287,
216
+ "eps": 6,
217
+ "loss/policy_avg": 0.1477496325969696,
218
+ "lr": 2.8828125e-06,
219
+ "objective/entropy": -66.58514404296875,
220
+ "objective/kl": 6.1831889152526855,
221
+ "objective/non_score_reward": 0.0,
222
+ "objective/rlhf_reward": 6.104897499084473,
223
+ "objective/scores": 6.104897499084473,
224
+ "policy/approxkl_avg": 0.11205172538757324,
225
+ "policy/clipfrac_avg": 0.2890625,
226
+ "policy/entropy_avg": 0.472697913646698,
227
+ "step": 10,
228
+ "val/clipfrac_avg": 0.0,
229
+ "val/num_eos_tokens": 2989,
230
+ "val/ratio": 0.9999105334281921,
231
+ "val/ratio_var": 1.112983909479226e-06
232
+ },
233
+ {
234
+ "episode": 6144,
235
+ "epoch": 0.026319396847155587,
236
+ "eps": 6,
237
+ "loss/policy_avg": 0.1740282028913498,
238
+ "lr": 2.87109375e-06,
239
+ "objective/entropy": -68.2206039428711,
240
+ "objective/kl": 6.888599395751953,
241
+ "objective/non_score_reward": 0.0,
242
+ "objective/rlhf_reward": 5.872122764587402,
243
+ "objective/scores": 5.872122764587402,
244
+ "policy/approxkl_avg": 0.1110803633928299,
245
+ "policy/clipfrac_avg": 0.279296875,
246
+ "policy/entropy_avg": 0.4416949152946472,
247
+ "step": 11,
248
+ "val/clipfrac_avg": 0.0,
249
+ "val/num_eos_tokens": 3219,
250
+ "val/ratio": 1.0001301765441895,
251
+ "val/ratio_var": 1.5931813095448888e-06
252
+ },
253
+ {
254
+ "episode": 6656,
255
+ "epoch": 0.028512679917751884,
256
+ "eps": 6,
257
+ "loss/policy_avg": 0.17489734292030334,
258
+ "lr": 2.859375e-06,
259
+ "objective/entropy": -65.95451354980469,
260
+ "objective/kl": 7.667189121246338,
261
+ "objective/non_score_reward": 0.0,
262
+ "objective/rlhf_reward": 6.356113433837891,
263
+ "objective/scores": 6.356113433837891,
264
+ "policy/approxkl_avg": 0.11979297548532486,
265
+ "policy/clipfrac_avg": 0.2890625,
266
+ "policy/entropy_avg": 0.4357627332210541,
267
+ "step": 12,
268
+ "val/clipfrac_avg": 0.0,
269
+ "val/num_eos_tokens": 3045,
270
+ "val/ratio": 0.9999862313270569,
271
+ "val/ratio_var": 2.047661610049545e-06
272
+ },
273
+ {
274
+ "episode": 7168,
275
+ "epoch": 0.030705962988348184,
276
+ "eps": 6,
277
+ "loss/policy_avg": 0.175493061542511,
278
+ "lr": 2.84765625e-06,
279
+ "objective/entropy": -67.00045776367188,
280
+ "objective/kl": 8.392281532287598,
281
+ "objective/non_score_reward": 0.0,
282
+ "objective/rlhf_reward": 6.625285625457764,
283
+ "objective/scores": 6.625285625457764,
284
+ "policy/approxkl_avg": 0.1163678765296936,
285
+ "policy/clipfrac_avg": 0.306640625,
286
+ "policy/entropy_avg": 0.4532189965248108,
287
+ "step": 13,
288
+ "val/clipfrac_avg": 0.0,
289
+ "val/num_eos_tokens": 3123,
290
+ "val/ratio": 0.9998478889465332,
291
+ "val/ratio_var": 1.6296273770421976e-06
292
+ },
293
+ {
294
+ "episode": 7680,
295
+ "epoch": 0.03289924605894448,
296
+ "eps": 6,
297
+ "loss/policy_avg": 0.12691491842269897,
298
+ "lr": 2.8359375e-06,
299
+ "objective/entropy": -62.55009841918945,
300
+ "objective/kl": 9.323993682861328,
301
+ "objective/non_score_reward": 0.0,
302
+ "objective/rlhf_reward": 6.616338729858398,
303
+ "objective/scores": 6.616338729858398,
304
+ "policy/approxkl_avg": 0.15570726990699768,
305
+ "policy/clipfrac_avg": 0.328125,
306
+ "policy/entropy_avg": 0.4586066007614136,
307
+ "step": 14,
308
+ "val/clipfrac_avg": 0.0,
309
+ "val/num_eos_tokens": 2864,
310
+ "val/ratio": 0.9998098611831665,
311
+ "val/ratio_var": 1.96517794393003e-06
312
+ },
313
+ {
314
+ "episode": 8192,
315
+ "epoch": 0.03509252912954078,
316
+ "eps": 6,
317
+ "loss/policy_avg": 0.23178045451641083,
318
+ "lr": 2.82421875e-06,
319
+ "objective/entropy": -57.576053619384766,
320
+ "objective/kl": 11.68783950805664,
321
+ "objective/non_score_reward": 0.0,
322
+ "objective/rlhf_reward": 6.858205795288086,
323
+ "objective/scores": 6.858205795288086,
324
+ "policy/approxkl_avg": 0.17012163996696472,
325
+ "policy/clipfrac_avg": 0.322265625,
326
+ "policy/entropy_avg": 0.49115192890167236,
327
+ "step": 15,
328
+ "val/clipfrac_avg": 0.0,
329
+ "val/num_eos_tokens": 2729,
330
+ "val/ratio": 1.0000524520874023,
331
+ "val/ratio_var": 2.174719611502951e-06
332
+ },
333
+ {
334
+ "episode": 8704,
335
+ "epoch": 0.03728581220013708,
336
+ "eps": 6,
337
+ "loss/policy_avg": 0.22957196831703186,
338
+ "lr": 2.8125e-06,
339
+ "objective/entropy": -50.27229690551758,
340
+ "objective/kl": 15.081552505493164,
341
+ "objective/non_score_reward": 0.0,
342
+ "objective/rlhf_reward": 7.474477767944336,
343
+ "objective/scores": 7.474477767944336,
344
+ "policy/approxkl_avg": 0.18791517615318298,
345
+ "policy/clipfrac_avg": 0.3671875,
346
+ "policy/entropy_avg": 0.4883398115634918,
347
+ "step": 16,
348
+ "val/clipfrac_avg": 0.0,
349
+ "val/num_eos_tokens": 2471,
350
+ "val/ratio": 1.00034499168396,
351
+ "val/ratio_var": 3.5397249575908063e-06
352
+ },
353
+ {
354
+ "episode": 9216,
355
+ "epoch": 0.03947909527073338,
356
+ "eps": 6,
357
+ "loss/policy_avg": 0.49715545773506165,
358
+ "lr": 2.80078125e-06,
359
+ "objective/entropy": -45.675880432128906,
360
+ "objective/kl": 18.226818084716797,
361
+ "objective/non_score_reward": 0.0,
362
+ "objective/rlhf_reward": 7.699345588684082,
363
+ "objective/scores": 7.699345588684082,
364
+ "policy/approxkl_avg": 0.25131434202194214,
365
+ "policy/clipfrac_avg": 0.341796875,
366
+ "policy/entropy_avg": 0.5121623277664185,
367
+ "step": 17,
368
+ "val/clipfrac_avg": 0.0,
369
+ "val/num_eos_tokens": 2467,
370
+ "val/ratio": 1.0001544952392578,
371
+ "val/ratio_var": 4.669571353588253e-06
372
+ },
373
+ {
374
+ "episode": 9728,
375
+ "epoch": 0.041672378341329676,
376
+ "eps": 6,
377
+ "loss/policy_avg": 0.346825510263443,
378
+ "lr": 2.7890625e-06,
379
+ "objective/entropy": -37.562652587890625,
380
+ "objective/kl": 23.889034271240234,
381
+ "objective/non_score_reward": 0.0,
382
+ "objective/rlhf_reward": 7.912500381469727,
383
+ "objective/scores": 7.912500381469727,
384
+ "policy/approxkl_avg": 0.2981986999511719,
385
+ "policy/clipfrac_avg": 0.38671875,
386
+ "policy/entropy_avg": 0.5075176954269409,
387
+ "step": 18,
388
+ "val/clipfrac_avg": 0.0,
389
+ "val/num_eos_tokens": 2109,
390
+ "val/ratio": 0.9998643398284912,
391
+ "val/ratio_var": 3.7719155443483032e-06
392
+ },
393
+ {
394
+ "episode": 10240,
395
+ "epoch": 0.04386566141192598,
396
+ "eps": 6,
397
+ "loss/policy_avg": 0.35793837904930115,
398
+ "lr": 2.77734375e-06,
399
+ "objective/entropy": -38.90000534057617,
400
+ "objective/kl": 24.49026870727539,
401
+ "objective/non_score_reward": 0.0,
402
+ "objective/rlhf_reward": 7.865503311157227,
403
+ "objective/scores": 7.865503311157227,
404
+ "policy/approxkl_avg": 0.28773218393325806,
405
+ "policy/clipfrac_avg": 0.373046875,
406
+ "policy/entropy_avg": 0.4393501877784729,
407
+ "step": 19,
408
+ "val/clipfrac_avg": 0.0,
409
+ "val/num_eos_tokens": 1740,
410
+ "val/ratio": 1.0000510215759277,
411
+ "val/ratio_var": 4.729288775706664e-06
412
+ },
413
+ {
414
+ "episode": 10752,
415
+ "epoch": 0.04605894448252228,
416
+ "eps": 6,
417
+ "loss/policy_avg": 0.39905059337615967,
418
+ "lr": 2.765625e-06,
419
+ "objective/entropy": -37.25298309326172,
420
+ "objective/kl": 26.252962112426758,
421
+ "objective/non_score_reward": 0.0,
422
+ "objective/rlhf_reward": 7.933384895324707,
423
+ "objective/scores": 7.933384895324707,
424
+ "policy/approxkl_avg": 0.33784598112106323,
425
+ "policy/clipfrac_avg": 0.365234375,
426
+ "policy/entropy_avg": 0.45024600625038147,
427
+ "step": 20,
428
+ "val/clipfrac_avg": 0.0,
429
+ "val/num_eos_tokens": 1676,
430
+ "val/ratio": 0.9996903538703918,
431
+ "val/ratio_var": 4.557784450298641e-06
432
+ },
433
+ {
434
+ "episode": 11264,
435
+ "epoch": 0.048252227553118573,
436
+ "eps": 6,
437
+ "loss/policy_avg": 0.47556352615356445,
438
+ "lr": 2.75390625e-06,
439
+ "objective/entropy": -35.37464141845703,
440
+ "objective/kl": 26.959781646728516,
441
+ "objective/non_score_reward": 0.0,
442
+ "objective/rlhf_reward": 8.056526184082031,
443
+ "objective/scores": 8.056526184082031,
444
+ "policy/approxkl_avg": 0.29800647497177124,
445
+ "policy/clipfrac_avg": 0.37109375,
446
+ "policy/entropy_avg": 0.4291992783546448,
447
+ "step": 21,
448
+ "val/clipfrac_avg": 0.0,
449
+ "val/num_eos_tokens": 1512,
450
+ "val/ratio": 0.9997580051422119,
451
+ "val/ratio_var": 4.297321993362857e-06
452
+ },
453
+ {
454
+ "episode": 11776,
455
+ "epoch": 0.05044551062371487,
456
+ "eps": 6,
457
+ "loss/policy_avg": 0.4392542243003845,
458
+ "lr": 2.7421875e-06,
459
+ "objective/entropy": -41.550724029541016,
460
+ "objective/kl": 24.043529510498047,
461
+ "objective/non_score_reward": 0.0,
462
+ "objective/rlhf_reward": 8.027580261230469,
463
+ "objective/scores": 8.027580261230469,
464
+ "policy/approxkl_avg": 0.26326847076416016,
465
+ "policy/clipfrac_avg": 0.36328125,
466
+ "policy/entropy_avg": 0.4284892678260803,
467
+ "step": 22,
468
+ "val/clipfrac_avg": 0.0,
469
+ "val/num_eos_tokens": 1592,
470
+ "val/ratio": 1.0000464916229248,
471
+ "val/ratio_var": 4.475009518500883e-06
472
+ },
473
+ {
474
+ "episode": 12288,
475
+ "epoch": 0.052638793694311174,
476
+ "eps": 6,
477
+ "loss/policy_avg": 0.3007674515247345,
478
+ "lr": 2.73046875e-06,
479
+ "objective/entropy": -52.200408935546875,
480
+ "objective/kl": 21.862667083740234,
481
+ "objective/non_score_reward": 0.0,
482
+ "objective/rlhf_reward": 8.28553581237793,
483
+ "objective/scores": 8.28553581237793,
484
+ "policy/approxkl_avg": 0.2857671082019806,
485
+ "policy/clipfrac_avg": 0.37109375,
486
+ "policy/entropy_avg": 0.420071542263031,
487
+ "step": 23,
488
+ "val/clipfrac_avg": 0.0,
489
+ "val/num_eos_tokens": 1880,
490
+ "val/ratio": 0.999809980392456,
491
+ "val/ratio_var": 4.868047199124703e-06
492
+ },
493
+ {
494
+ "episode": 12800,
495
+ "epoch": 0.05483207676490747,
496
+ "eps": 6,
497
+ "loss/policy_avg": 0.2275318056344986,
498
+ "lr": 2.71875e-06,
499
+ "objective/entropy": -52.59136962890625,
500
+ "objective/kl": 21.929889678955078,
501
+ "objective/non_score_reward": 0.0,
502
+ "objective/rlhf_reward": 8.46428108215332,
503
+ "objective/scores": 8.46428108215332,
504
+ "policy/approxkl_avg": 0.24716982245445251,
505
+ "policy/clipfrac_avg": 0.34765625,
506
+ "policy/entropy_avg": 0.4286070466041565,
507
+ "step": 24,
508
+ "val/clipfrac_avg": 0.0,
509
+ "val/num_eos_tokens": 1946,
510
+ "val/ratio": 1.0000059604644775,
511
+ "val/ratio_var": 3.2738601021264913e-06
512
+ },
513
+ {
514
+ "episode": 13312,
515
+ "epoch": 0.05702535983550377,
516
+ "eps": 6,
517
+ "loss/policy_avg": 0.21806836128234863,
518
+ "lr": 2.70703125e-06,
519
+ "objective/entropy": -52.577674865722656,
520
+ "objective/kl": 20.580398559570312,
521
+ "objective/non_score_reward": 0.0,
522
+ "objective/rlhf_reward": 8.337871551513672,
523
+ "objective/scores": 8.337871551513672,
524
+ "policy/approxkl_avg": 0.26169508695602417,
525
+ "policy/clipfrac_avg": 0.361328125,
526
+ "policy/entropy_avg": 0.47688111662864685,
527
+ "step": 25,
528
+ "val/clipfrac_avg": 0.0,
529
+ "val/num_eos_tokens": 2202,
530
+ "val/ratio": 0.9996628165245056,
531
+ "val/ratio_var": 3.0032485938136233e-06
532
+ },
533
+ {
534
+ "episode": 13824,
535
+ "epoch": 0.05921864290610007,
536
+ "eps": 6,
537
+ "loss/policy_avg": 0.2163274586200714,
538
+ "lr": 2.6953125e-06,
539
+ "objective/entropy": -56.968421936035156,
540
+ "objective/kl": 19.904338836669922,
541
+ "objective/non_score_reward": 0.0,
542
+ "objective/rlhf_reward": 8.275657653808594,
543
+ "objective/scores": 8.275657653808594,
544
+ "policy/approxkl_avg": 0.23852765560150146,
545
+ "policy/clipfrac_avg": 0.35546875,
546
+ "policy/entropy_avg": 0.4648514986038208,
547
+ "step": 26,
548
+ "val/clipfrac_avg": 0.0,
549
+ "val/num_eos_tokens": 2346,
550
+ "val/ratio": 0.9999961853027344,
551
+ "val/ratio_var": 3.942259809264215e-06
552
+ },
553
+ {
554
+ "episode": 14336,
555
+ "epoch": 0.06141192597669637,
556
+ "eps": 6,
557
+ "loss/policy_avg": 0.1764034628868103,
558
+ "lr": 2.68359375e-06,
559
+ "objective/entropy": -56.630027770996094,
560
+ "objective/kl": 20.158096313476562,
561
+ "objective/non_score_reward": 0.0,
562
+ "objective/rlhf_reward": 8.34037971496582,
563
+ "objective/scores": 8.34037971496582,
564
+ "policy/approxkl_avg": 0.30687031149864197,
565
+ "policy/clipfrac_avg": 0.3828125,
566
+ "policy/entropy_avg": 0.4918164312839508,
567
+ "step": 27,
568
+ "val/clipfrac_avg": 0.0,
569
+ "val/num_eos_tokens": 2430,
570
+ "val/ratio": 0.999588131904602,
571
+ "val/ratio_var": 3.445997435846948e-06
572
+ },
573
+ {
574
+ "episode": 14848,
575
+ "epoch": 0.06360520904729267,
576
+ "eps": 6,
577
+ "loss/policy_avg": 0.248679518699646,
578
+ "lr": 2.671875e-06,
579
+ "objective/entropy": -56.24310302734375,
580
+ "objective/kl": 19.94865608215332,
581
+ "objective/non_score_reward": 0.0,
582
+ "objective/rlhf_reward": 8.386137008666992,
583
+ "objective/scores": 8.386137008666992,
584
+ "policy/approxkl_avg": 0.2245277464389801,
585
+ "policy/clipfrac_avg": 0.361328125,
586
+ "policy/entropy_avg": 0.49217647314071655,
587
+ "step": 28,
588
+ "val/clipfrac_avg": 0.0,
589
+ "val/num_eos_tokens": 2555,
590
+ "val/ratio": 0.9999157786369324,
591
+ "val/ratio_var": 3.2236650895356433e-06
592
+ },
593
+ {
594
+ "episode": 15360,
595
+ "epoch": 0.06579849211788896,
596
+ "eps": 6,
597
+ "loss/policy_avg": 0.26906225085258484,
598
+ "lr": 2.66015625e-06,
599
+ "objective/entropy": -56.197872161865234,
600
+ "objective/kl": 19.09078025817871,
601
+ "objective/non_score_reward": 0.0,
602
+ "objective/rlhf_reward": 8.399325370788574,
603
+ "objective/scores": 8.399325370788574,
604
+ "policy/approxkl_avg": 0.2715945839881897,
605
+ "policy/clipfrac_avg": 0.375,
606
+ "policy/entropy_avg": 0.52540123462677,
607
+ "step": 29,
608
+ "val/clipfrac_avg": 0.0,
609
+ "val/num_eos_tokens": 2600,
610
+ "val/ratio": 1.0000545978546143,
611
+ "val/ratio_var": 4.418570824782364e-06
612
+ },
613
+ {
614
+ "episode": 15872,
615
+ "epoch": 0.06799177518848526,
616
+ "eps": 6,
617
+ "loss/policy_avg": 0.2973303198814392,
618
+ "lr": 2.6484375e-06,
619
+ "objective/entropy": -62.014461517333984,
620
+ "objective/kl": 19.058860778808594,
621
+ "objective/non_score_reward": 0.0,
622
+ "objective/rlhf_reward": 8.694437026977539,
623
+ "objective/scores": 8.694437026977539,
624
+ "policy/approxkl_avg": 0.2335938811302185,
625
+ "policy/clipfrac_avg": 0.400390625,
626
+ "policy/entropy_avg": 0.5033716559410095,
627
+ "step": 30,
628
+ "val/clipfrac_avg": 0.0,
629
+ "val/num_eos_tokens": 2786,
630
+ "val/ratio": 0.9996792674064636,
631
+ "val/ratio_var": 2.6598288513923762e-06
632
+ },
633
+ {
634
+ "episode": 16384,
635
+ "epoch": 0.07018505825908156,
636
+ "eps": 6,
637
+ "loss/policy_avg": 0.19679370522499084,
638
+ "lr": 2.63671875e-06,
639
+ "objective/entropy": -58.170814514160156,
640
+ "objective/kl": 18.921110153198242,
641
+ "objective/non_score_reward": 0.0,
642
+ "objective/rlhf_reward": 8.450736045837402,
643
+ "objective/scores": 8.450736045837402,
644
+ "policy/approxkl_avg": 0.21560396254062653,
645
+ "policy/clipfrac_avg": 0.359375,
646
+ "policy/entropy_avg": 0.5557649731636047,
647
+ "step": 31,
648
+ "val/clipfrac_avg": 0.0,
649
+ "val/num_eos_tokens": 2790,
650
+ "val/ratio": 1.0001705884933472,
651
+ "val/ratio_var": 2.6364186851424165e-06
652
+ },
653
+ {
654
+ "episode": 16896,
655
+ "epoch": 0.07237834132967787,
656
+ "eps": 6,
657
+ "loss/policy_avg": 0.1725379377603531,
658
+ "lr": 2.6250000000000003e-06,
659
+ "objective/entropy": -59.47856140136719,
660
+ "objective/kl": 19.420867919921875,
661
+ "objective/non_score_reward": 0.0,
662
+ "objective/rlhf_reward": 8.567611694335938,
663
+ "objective/scores": 8.567611694335938,
664
+ "policy/approxkl_avg": 0.28828394412994385,
665
+ "policy/clipfrac_avg": 0.345703125,
666
+ "policy/entropy_avg": 0.5723543167114258,
667
+ "step": 32,
668
+ "val/clipfrac_avg": 0.0,
669
+ "val/num_eos_tokens": 3227,
670
+ "val/ratio": 0.9996543526649475,
671
+ "val/ratio_var": 2.863696181520936e-06
672
+ },
673
+ {
674
+ "episode": 17408,
675
+ "epoch": 0.07457162440027416,
676
+ "eps": 6,
677
+ "loss/policy_avg": 0.3564334511756897,
678
+ "lr": 2.61328125e-06,
679
+ "objective/entropy": -59.503028869628906,
680
+ "objective/kl": 19.913440704345703,
681
+ "objective/non_score_reward": 0.0,
682
+ "objective/rlhf_reward": 8.610246658325195,
683
+ "objective/scores": 8.610246658325195,
684
+ "policy/approxkl_avg": 0.22897233068943024,
685
+ "policy/clipfrac_avg": 0.359375,
686
+ "policy/entropy_avg": 0.5609124898910522,
687
+ "step": 33,
688
+ "val/clipfrac_avg": 0.0,
689
+ "val/num_eos_tokens": 2937,
690
+ "val/ratio": 0.9998010396957397,
691
+ "val/ratio_var": 3.5082136946584797e-06
692
+ },
693
+ {
694
+ "episode": 17920,
695
+ "epoch": 0.07676490747087046,
696
+ "eps": 6,
697
+ "loss/policy_avg": 0.2554902136325836,
698
+ "lr": 2.6015625e-06,
699
+ "objective/entropy": -59.88847732543945,
700
+ "objective/kl": 18.962574005126953,
701
+ "objective/non_score_reward": 0.0,
702
+ "objective/rlhf_reward": 8.546546936035156,
703
+ "objective/scores": 8.546546936035156,
704
+ "policy/approxkl_avg": 0.26599401235580444,
705
+ "policy/clipfrac_avg": 0.328125,
706
+ "policy/entropy_avg": 0.5997524857521057,
707
+ "step": 34,
708
+ "val/clipfrac_avg": 0.0,
709
+ "val/num_eos_tokens": 3095,
710
+ "val/ratio": 1.00026535987854,
711
+ "val/ratio_var": 3.7096822325111134e-06
712
+ },
713
+ {
714
+ "episode": 18432,
715
+ "epoch": 0.07895819054146676,
716
+ "eps": 6,
717
+ "loss/policy_avg": 0.3109360337257385,
718
+ "lr": 2.5898437500000003e-06,
719
+ "objective/entropy": -62.252044677734375,
720
+ "objective/kl": 19.137222290039062,
721
+ "objective/non_score_reward": 0.0,
722
+ "objective/rlhf_reward": 8.649435043334961,
723
+ "objective/scores": 8.649435043334961,
724
+ "policy/approxkl_avg": 0.2142583727836609,
725
+ "policy/clipfrac_avg": 0.345703125,
726
+ "policy/entropy_avg": 0.607221245765686,
727
+ "step": 35,
728
+ "val/clipfrac_avg": 0.0,
729
+ "val/num_eos_tokens": 3229,
730
+ "val/ratio": 1.0000017881393433,
731
+ "val/ratio_var": 2.4627622678963235e-06
732
+ },
733
+ {
734
+ "episode": 18944,
735
+ "epoch": 0.08115147361206305,
736
+ "eps": 6,
737
+ "loss/policy_avg": 0.254541277885437,
738
+ "lr": 2.578125e-06,
739
+ "objective/entropy": -63.467262268066406,
740
+ "objective/kl": 18.373220443725586,
741
+ "objective/non_score_reward": 0.0,
742
+ "objective/rlhf_reward": 8.687257766723633,
743
+ "objective/scores": 8.687257766723633,
744
+ "policy/approxkl_avg": 0.19740049540996552,
745
+ "policy/clipfrac_avg": 0.326171875,
746
+ "policy/entropy_avg": 0.6166640520095825,
747
+ "step": 36,
748
+ "val/clipfrac_avg": 0.0,
749
+ "val/num_eos_tokens": 3321,
750
+ "val/ratio": 0.9999884366989136,
751
+ "val/ratio_var": 3.2546743113925913e-06
752
+ },
753
+ {
754
+ "episode": 19456,
755
+ "epoch": 0.08334475668265935,
756
+ "eps": 6,
757
+ "loss/policy_avg": 0.17846369743347168,
758
+ "lr": 2.56640625e-06,
759
+ "objective/entropy": -63.46952819824219,
760
+ "objective/kl": 19.115474700927734,
761
+ "objective/non_score_reward": 0.0,
762
+ "objective/rlhf_reward": 8.613758087158203,
763
+ "objective/scores": 8.613758087158203,
764
+ "policy/approxkl_avg": 0.23868654668331146,
765
+ "policy/clipfrac_avg": 0.375,
766
+ "policy/entropy_avg": 0.5862524509429932,
767
+ "step": 37,
768
+ "val/clipfrac_avg": 0.0,
769
+ "val/num_eos_tokens": 3248,
770
+ "val/ratio": 0.9998351335525513,
771
+ "val/ratio_var": 2.714186166485888e-06
772
+ },
773
+ {
774
+ "episode": 19968,
775
+ "epoch": 0.08553803975325565,
776
+ "eps": 6,
777
+ "loss/policy_avg": 0.19094648957252502,
778
+ "lr": 2.5546875000000003e-06,
779
+ "objective/entropy": -63.260986328125,
780
+ "objective/kl": 19.352394104003906,
781
+ "objective/non_score_reward": 0.0,
782
+ "objective/rlhf_reward": 8.733724594116211,
783
+ "objective/scores": 8.733724594116211,
784
+ "policy/approxkl_avg": 0.2169877141714096,
785
+ "policy/clipfrac_avg": 0.3203125,
786
+ "policy/entropy_avg": 0.5959694981575012,
787
+ "step": 38,
788
+ "val/clipfrac_avg": 0.0,
789
+ "val/num_eos_tokens": 3289,
790
+ "val/ratio": 1.0003280639648438,
791
+ "val/ratio_var": 3.3813628306234023e-06
792
+ },
793
+ {
794
+ "episode": 20480,
795
+ "epoch": 0.08773132282385196,
796
+ "eps": 6,
797
+ "loss/policy_avg": 0.2243296056985855,
798
+ "lr": 2.54296875e-06,
799
+ "objective/entropy": -59.47791290283203,
800
+ "objective/kl": 19.82537841796875,
801
+ "objective/non_score_reward": 0.0,
802
+ "objective/rlhf_reward": 8.559447288513184,
803
+ "objective/scores": 8.559447288513184,
804
+ "policy/approxkl_avg": 0.30144354701042175,
805
+ "policy/clipfrac_avg": 0.3515625,
806
+ "policy/entropy_avg": 0.6129642724990845,
807
+ "step": 39,
808
+ "val/clipfrac_avg": 0.0,
809
+ "val/num_eos_tokens": 3091,
810
+ "val/ratio": 0.9997188448905945,
811
+ "val/ratio_var": 4.063951564603485e-06
812
+ },
813
+ {
814
+ "episode": 20992,
815
+ "epoch": 0.08992460589444826,
816
+ "eps": 6,
817
+ "loss/policy_avg": 0.14945922791957855,
818
+ "lr": 2.53125e-06,
819
+ "objective/entropy": -63.837181091308594,
820
+ "objective/kl": 19.955276489257812,
821
+ "objective/non_score_reward": 0.0,
822
+ "objective/rlhf_reward": 8.650920867919922,
823
+ "objective/scores": 8.650920867919922,
824
+ "policy/approxkl_avg": 0.26656606793403625,
825
+ "policy/clipfrac_avg": 0.326171875,
826
+ "policy/entropy_avg": 0.6257542371749878,
827
+ "step": 40,
828
+ "val/clipfrac_avg": 0.0,
829
+ "val/num_eos_tokens": 2992,
830
+ "val/ratio": 1.0000975131988525,
831
+ "val/ratio_var": 3.132868187094573e-06
832
+ },
833
+ {
834
+ "episode": 21504,
835
+ "epoch": 0.09211788896504455,
836
+ "eps": 6,
837
+ "loss/policy_avg": 0.15452629327774048,
838
+ "lr": 2.5195312500000003e-06,
839
+ "objective/entropy": -63.7586784362793,
840
+ "objective/kl": 20.086402893066406,
841
+ "objective/non_score_reward": 0.0,
842
+ "objective/rlhf_reward": 8.547872543334961,
843
+ "objective/scores": 8.547872543334961,
844
+ "policy/approxkl_avg": 0.20186936855316162,
845
+ "policy/clipfrac_avg": 0.3515625,
846
+ "policy/entropy_avg": 0.5954059362411499,
847
+ "step": 41,
848
+ "val/clipfrac_avg": 0.0,
849
+ "val/num_eos_tokens": 3124,
850
+ "val/ratio": 1.0001468658447266,
851
+ "val/ratio_var": 3.0967376005719416e-06
852
+ },
853
+ {
854
+ "episode": 22016,
855
+ "epoch": 0.09431117203564085,
856
+ "eps": 6,
857
+ "loss/policy_avg": 0.23816156387329102,
858
+ "lr": 2.5078125e-06,
859
+ "objective/entropy": -61.8187141418457,
860
+ "objective/kl": 20.341209411621094,
861
+ "objective/non_score_reward": 0.0,
862
+ "objective/rlhf_reward": 8.72125244140625,
863
+ "objective/scores": 8.72125244140625,
864
+ "policy/approxkl_avg": 0.22096118330955505,
865
+ "policy/clipfrac_avg": 0.357421875,
866
+ "policy/entropy_avg": 0.6372725963592529,
867
+ "step": 42,
868
+ "val/clipfrac_avg": 0.0,
869
+ "val/num_eos_tokens": 2821,
870
+ "val/ratio": 1.0003211498260498,
871
+ "val/ratio_var": 6.170133019622881e-06
872
+ },
873
+ {
874
+ "episode": 22528,
875
+ "epoch": 0.09650445510623715,
876
+ "eps": 6,
877
+ "loss/policy_avg": 0.31142109632492065,
878
+ "lr": 2.49609375e-06,
879
+ "objective/entropy": -60.373191833496094,
880
+ "objective/kl": 21.558738708496094,
881
+ "objective/non_score_reward": 0.0,
882
+ "objective/rlhf_reward": 8.520058631896973,
883
+ "objective/scores": 8.520058631896973,
884
+ "policy/approxkl_avg": 0.3703784942626953,
885
+ "policy/clipfrac_avg": 0.333984375,
886
+ "policy/entropy_avg": 0.6228446960449219,
887
+ "step": 43,
888
+ "val/clipfrac_avg": 0.0,
889
+ "val/num_eos_tokens": 2950,
890
+ "val/ratio": 0.9998255372047424,
891
+ "val/ratio_var": 3.503203515720088e-06
892
+ },
893
+ {
894
+ "episode": 23040,
895
+ "epoch": 0.09869773817683344,
896
+ "eps": 6,
897
+ "loss/policy_avg": 0.20921407639980316,
898
+ "lr": 2.4843750000000002e-06,
899
+ "objective/entropy": -62.82386779785156,
900
+ "objective/kl": 21.283199310302734,
901
+ "objective/non_score_reward": 0.0,
902
+ "objective/rlhf_reward": 8.563481330871582,
903
+ "objective/scores": 8.563481330871582,
904
+ "policy/approxkl_avg": 0.245059072971344,
905
+ "policy/clipfrac_avg": 0.37890625,
906
+ "policy/entropy_avg": 0.6021170616149902,
907
+ "step": 44,
908
+ "val/clipfrac_avg": 0.0,
909
+ "val/num_eos_tokens": 3006,
910
+ "val/ratio": 1.000809907913208,
911
+ "val/ratio_var": 3.419398490223102e-05
912
+ },
913
+ {
914
+ "episode": 23552,
915
+ "epoch": 0.10089102124742974,
916
+ "eps": 6,
917
+ "loss/policy_avg": 0.22898651659488678,
918
+ "lr": 2.47265625e-06,
919
+ "objective/entropy": -61.859039306640625,
920
+ "objective/kl": 21.692474365234375,
921
+ "objective/non_score_reward": 0.0,
922
+ "objective/rlhf_reward": 8.69731330871582,
923
+ "objective/scores": 8.69731330871582,
924
+ "policy/approxkl_avg": 0.26175111532211304,
925
+ "policy/clipfrac_avg": 0.328125,
926
+ "policy/entropy_avg": 0.6003079414367676,
927
+ "step": 45,
928
+ "val/clipfrac_avg": 0.0,
929
+ "val/num_eos_tokens": 2893,
930
+ "val/ratio": 0.9998016357421875,
931
+ "val/ratio_var": 5.31077648702194e-06
932
+ },
933
+ {
934
+ "episode": 24064,
935
+ "epoch": 0.10308430431802605,
936
+ "eps": 6,
937
+ "loss/policy_avg": 0.280842125415802,
938
+ "lr": 2.4609375e-06,
939
+ "objective/entropy": -61.460044860839844,
940
+ "objective/kl": 21.635478973388672,
941
+ "objective/non_score_reward": 0.0,
942
+ "objective/rlhf_reward": 8.86025333404541,
943
+ "objective/scores": 8.86025333404541,
944
+ "policy/approxkl_avg": 0.26439762115478516,
945
+ "policy/clipfrac_avg": 0.3359375,
946
+ "policy/entropy_avg": 0.6330952644348145,
947
+ "step": 46,
948
+ "val/clipfrac_avg": 0.0,
949
+ "val/num_eos_tokens": 2932,
950
+ "val/ratio": 0.9998016357421875,
951
+ "val/ratio_var": 2.7805131139757577e-06
952
+ },
953
+ {
954
+ "episode": 24576,
955
+ "epoch": 0.10527758738862235,
956
+ "eps": 6,
957
+ "loss/policy_avg": 0.33989423513412476,
958
+ "lr": 2.4492187500000002e-06,
959
+ "objective/entropy": -62.55036163330078,
960
+ "objective/kl": 21.94875717163086,
961
+ "objective/non_score_reward": 0.0,
962
+ "objective/rlhf_reward": 8.946436882019043,
963
+ "objective/scores": 8.946436882019043,
964
+ "policy/approxkl_avg": 0.21244999766349792,
965
+ "policy/clipfrac_avg": 0.359375,
966
+ "policy/entropy_avg": 0.6167851686477661,
967
+ "step": 47,
968
+ "val/clipfrac_avg": 0.0,
969
+ "val/num_eos_tokens": 2966,
970
+ "val/ratio": 1.000269889831543,
971
+ "val/ratio_var": 3.7640895698132226e-06
972
+ },
973
+ {
974
+ "episode": 25088,
975
+ "epoch": 0.10747087045921865,
976
+ "eps": 6,
977
+ "loss/policy_avg": 0.18155589699745178,
978
+ "lr": 2.4375e-06,
979
+ "objective/entropy": -62.10002136230469,
980
+ "objective/kl": 22.01198959350586,
981
+ "objective/non_score_reward": 0.0,
982
+ "objective/rlhf_reward": 8.775335311889648,
983
+ "objective/scores": 8.775335311889648,
984
+ "policy/approxkl_avg": 0.2533162236213684,
985
+ "policy/clipfrac_avg": 0.36328125,
986
+ "policy/entropy_avg": 0.6513728499412537,
987
+ "step": 48,
988
+ "val/clipfrac_avg": 0.0,
989
+ "val/num_eos_tokens": 2958,
990
+ "val/ratio": 0.9997414350509644,
991
+ "val/ratio_var": 2.8600227324204752e-06
992
+ },
993
+ {
994
+ "episode": 25600,
995
+ "epoch": 0.10966415352981494,
996
+ "eps": 6,
997
+ "loss/policy_avg": 0.22356639802455902,
998
+ "lr": 2.42578125e-06,
999
+ "objective/entropy": -60.96599578857422,
1000
+ "objective/kl": 22.15947723388672,
1001
+ "objective/non_score_reward": 0.0,
1002
+ "objective/rlhf_reward": 8.907562255859375,
1003
+ "objective/scores": 8.907562255859375,
1004
+ "policy/approxkl_avg": 0.21106398105621338,
1005
+ "policy/clipfrac_avg": 0.369140625,
1006
+ "policy/entropy_avg": 0.6532235145568848,
1007
+ "step": 49,
1008
+ "val/clipfrac_avg": 0.0,
1009
+ "val/num_eos_tokens": 3107,
1010
+ "val/ratio": 0.9998568296432495,
1011
+ "val/ratio_var": 2.8039498829457443e-06
1012
+ },
1013
+ {
1014
+ "episode": 26112,
1015
+ "epoch": 0.11185743660041124,
1016
+ "eps": 6,
1017
+ "loss/policy_avg": 0.2772133946418762,
1018
+ "lr": 2.4140625000000002e-06,
1019
+ "objective/entropy": -60.50039291381836,
1020
+ "objective/kl": 22.602027893066406,
1021
+ "objective/non_score_reward": 0.0,
1022
+ "objective/rlhf_reward": 8.589601516723633,
1023
+ "objective/scores": 8.589601516723633,
1024
+ "policy/approxkl_avg": 0.2172616720199585,
1025
+ "policy/clipfrac_avg": 0.349609375,
1026
+ "policy/entropy_avg": 0.6324688792228699,
1027
+ "step": 50,
1028
+ "val/clipfrac_avg": 0.0,
1029
+ "val/num_eos_tokens": 3201,
1030
+ "val/ratio": 1.0002442598342896,
1031
+ "val/ratio_var": 2.36768028116785e-06
1032
+ },
1033
+ {
1034
+ "episode": 26624,
1035
+ "epoch": 0.11405071967100754,
1036
+ "eps": 6,
1037
+ "loss/policy_avg": 0.17067407071590424,
1038
+ "lr": 2.40234375e-06,
1039
+ "objective/entropy": -58.01458740234375,
1040
+ "objective/kl": 23.73537826538086,
1041
+ "objective/non_score_reward": 0.0,
1042
+ "objective/rlhf_reward": 8.827688217163086,
1043
+ "objective/scores": 8.827688217163086,
1044
+ "policy/approxkl_avg": 0.23201116919517517,
1045
+ "policy/clipfrac_avg": 0.37109375,
1046
+ "policy/entropy_avg": 0.6225080490112305,
1047
+ "step": 51,
1048
+ "val/clipfrac_avg": 0.0,
1049
+ "val/num_eos_tokens": 3656,
1050
+ "val/ratio": 1.000308871269226,
1051
+ "val/ratio_var": 2.6556886041362304e-06
1052
+ },
1053
+ {
1054
+ "episode": 27136,
1055
+ "epoch": 0.11624400274160383,
1056
+ "eps": 6,
1057
+ "loss/policy_avg": 0.18175722658634186,
1058
+ "lr": 2.390625e-06,
1059
+ "objective/entropy": -58.348419189453125,
1060
+ "objective/kl": 23.386844635009766,
1061
+ "objective/non_score_reward": 0.0,
1062
+ "objective/rlhf_reward": 8.72817611694336,
1063
+ "objective/scores": 8.72817611694336,
1064
+ "policy/approxkl_avg": 0.2951381504535675,
1065
+ "policy/clipfrac_avg": 0.37890625,
1066
+ "policy/entropy_avg": 0.6238501667976379,
1067
+ "step": 52,
1068
+ "val/clipfrac_avg": 0.0,
1069
+ "val/num_eos_tokens": 3423,
1070
+ "val/ratio": 1.000365972518921,
1071
+ "val/ratio_var": 2.8732440569001483e-06
1072
+ },
1073
+ {
1074
+ "episode": 27648,
1075
+ "epoch": 0.11843728581220014,
1076
+ "eps": 6,
1077
+ "loss/policy_avg": 0.3416135609149933,
1078
+ "lr": 2.3789062500000002e-06,
1079
+ "objective/entropy": -57.76121520996094,
1080
+ "objective/kl": 23.697275161743164,
1081
+ "objective/non_score_reward": 0.0,
1082
+ "objective/rlhf_reward": 9.052407264709473,
1083
+ "objective/scores": 9.052407264709473,
1084
+ "policy/approxkl_avg": 0.2662898600101471,
1085
+ "policy/clipfrac_avg": 0.380859375,
1086
+ "policy/entropy_avg": 0.6075088977813721,
1087
+ "step": 53,
1088
+ "val/clipfrac_avg": 0.0,
1089
+ "val/num_eos_tokens": 3651,
1090
+ "val/ratio": 1.0000418424606323,
1091
+ "val/ratio_var": 4.378874109534081e-06
1092
+ },
1093
+ {
1094
+ "episode": 28160,
1095
+ "epoch": 0.12063056888279644,
1096
+ "eps": 6,
1097
+ "loss/policy_avg": 0.265788733959198,
1098
+ "lr": 2.3671875e-06,
1099
+ "objective/entropy": -59.14204788208008,
1100
+ "objective/kl": 23.922271728515625,
1101
+ "objective/non_score_reward": 0.0,
1102
+ "objective/rlhf_reward": 8.96934700012207,
1103
+ "objective/scores": 8.96934700012207,
1104
+ "policy/approxkl_avg": 0.29015815258026123,
1105
+ "policy/clipfrac_avg": 0.365234375,
1106
+ "policy/entropy_avg": 0.6201518177986145,
1107
+ "step": 54,
1108
+ "val/clipfrac_avg": 0.0,
1109
+ "val/num_eos_tokens": 3441,
1110
+ "val/ratio": 0.9999364614486694,
1111
+ "val/ratio_var": 2.862538394765579e-06
1112
+ },
1113
+ {
1114
+ "episode": 28672,
1115
+ "epoch": 0.12282385195339274,
1116
+ "eps": 6,
1117
+ "loss/policy_avg": 0.2591712474822998,
1118
+ "lr": 2.35546875e-06,
1119
+ "objective/entropy": -58.195640563964844,
1120
+ "objective/kl": 24.767526626586914,
1121
+ "objective/non_score_reward": 0.0,
1122
+ "objective/rlhf_reward": 8.780952453613281,
1123
+ "objective/scores": 8.780952453613281,
1124
+ "policy/approxkl_avg": 0.3050951361656189,
1125
+ "policy/clipfrac_avg": 0.3203125,
1126
+ "policy/entropy_avg": 0.6067063808441162,
1127
+ "step": 55,
1128
+ "val/clipfrac_avg": 0.0,
1129
+ "val/num_eos_tokens": 3560,
1130
+ "val/ratio": 0.9996646642684937,
1131
+ "val/ratio_var": 3.838819793600123e-06
1132
+ },
1133
+ {
1134
+ "episode": 29184,
1135
+ "epoch": 0.12501713502398903,
1136
+ "eps": 6,
1137
+ "loss/policy_avg": 0.1893739104270935,
1138
+ "lr": 2.3437500000000002e-06,
1139
+ "objective/entropy": -61.51493835449219,
1140
+ "objective/kl": 24.346567153930664,
1141
+ "objective/non_score_reward": 0.0,
1142
+ "objective/rlhf_reward": 8.864131927490234,
1143
+ "objective/scores": 8.864131927490234,
1144
+ "policy/approxkl_avg": 0.2597726583480835,
1145
+ "policy/clipfrac_avg": 0.357421875,
1146
+ "policy/entropy_avg": 0.5995157957077026,
1147
+ "step": 56,
1148
+ "val/clipfrac_avg": 0.0,
1149
+ "val/num_eos_tokens": 3598,
1150
+ "val/ratio": 0.9998759031295776,
1151
+ "val/ratio_var": 3.2004745662561618e-06
1152
+ },
1153
+ {
1154
+ "episode": 29696,
1155
+ "epoch": 0.12721041809458533,
1156
+ "eps": 6,
1157
+ "loss/policy_avg": 0.17370668053627014,
1158
+ "lr": 2.33203125e-06,
1159
+ "objective/entropy": -60.52989196777344,
1160
+ "objective/kl": 25.029788970947266,
1161
+ "objective/non_score_reward": 0.0,
1162
+ "objective/rlhf_reward": 8.666007041931152,
1163
+ "objective/scores": 8.666007041931152,
1164
+ "policy/approxkl_avg": 0.2449224293231964,
1165
+ "policy/clipfrac_avg": 0.361328125,
1166
+ "policy/entropy_avg": 0.587178111076355,
1167
+ "step": 57,
1168
+ "val/clipfrac_avg": 0.0,
1169
+ "val/num_eos_tokens": 3571,
1170
+ "val/ratio": 1.0002458095550537,
1171
+ "val/ratio_var": 3.114056880804128e-06
1172
+ },
1173
+ {
1174
+ "episode": 30208,
1175
+ "epoch": 0.12940370116518163,
1176
+ "eps": 6,
1177
+ "loss/policy_avg": 0.22951845824718475,
1178
+ "lr": 2.3203125e-06,
1179
+ "objective/entropy": -61.93848419189453,
1180
+ "objective/kl": 25.332937240600586,
1181
+ "objective/non_score_reward": 0.0,
1182
+ "objective/rlhf_reward": 8.495318412780762,
1183
+ "objective/scores": 8.495318412780762,
1184
+ "policy/approxkl_avg": 0.27128294110298157,
1185
+ "policy/clipfrac_avg": 0.349609375,
1186
+ "policy/entropy_avg": 0.5786547064781189,
1187
+ "step": 58,
1188
+ "val/clipfrac_avg": 0.0,
1189
+ "val/num_eos_tokens": 3478,
1190
+ "val/ratio": 1.0003738403320312,
1191
+ "val/ratio_var": 1.3942043551651295e-05
1192
+ },
1193
+ {
1194
+ "episode": 30720,
1195
+ "epoch": 0.13159698423577793,
1196
+ "eps": 6,
1197
+ "loss/policy_avg": 0.22642306983470917,
1198
+ "lr": 2.30859375e-06,
1199
+ "objective/entropy": -65.1754150390625,
1200
+ "objective/kl": 24.58494758605957,
1201
+ "objective/non_score_reward": 0.0,
1202
+ "objective/rlhf_reward": 8.604547500610352,
1203
+ "objective/scores": 8.604547500610352,
1204
+ "policy/approxkl_avg": 0.28292161226272583,
1205
+ "policy/clipfrac_avg": 0.310546875,
1206
+ "policy/entropy_avg": 0.5747346878051758,
1207
+ "step": 59,
1208
+ "val/clipfrac_avg": 0.0,
1209
+ "val/num_eos_tokens": 3528,
1210
+ "val/ratio": 0.9998360276222229,
1211
+ "val/ratio_var": 3.136041641482734e-06
1212
+ },
1213
+ {
1214
+ "episode": 31232,
1215
+ "epoch": 0.13379026730637422,
1216
+ "eps": 6,
1217
+ "loss/policy_avg": 0.1339200735092163,
1218
+ "lr": 2.296875e-06,
1219
+ "objective/entropy": -65.92912292480469,
1220
+ "objective/kl": 24.22686004638672,
1221
+ "objective/non_score_reward": 0.0,
1222
+ "objective/rlhf_reward": 8.52584457397461,
1223
+ "objective/scores": 8.52584457397461,
1224
+ "policy/approxkl_avg": 0.21744689345359802,
1225
+ "policy/clipfrac_avg": 0.32421875,
1226
+ "policy/entropy_avg": 0.5810856223106384,
1227
+ "step": 60,
1228
+ "val/clipfrac_avg": 0.0,
1229
+ "val/num_eos_tokens": 3584,
1230
+ "val/ratio": 0.9998077154159546,
1231
+ "val/ratio_var": 2.1918256152275717e-06
1232
+ },
1233
+ {
1234
+ "episode": 31744,
1235
+ "epoch": 0.13598355037697052,
1236
+ "eps": 6,
1237
+ "loss/policy_avg": 0.19992566108703613,
1238
+ "lr": 2.28515625e-06,
1239
+ "objective/entropy": -65.07493591308594,
1240
+ "objective/kl": 25.419109344482422,
1241
+ "objective/non_score_reward": 0.0,
1242
+ "objective/rlhf_reward": 8.409069061279297,
1243
+ "objective/scores": 8.409069061279297,
1244
+ "policy/approxkl_avg": 0.1962146908044815,
1245
+ "policy/clipfrac_avg": 0.310546875,
1246
+ "policy/entropy_avg": 0.5588007569313049,
1247
+ "step": 61,
1248
+ "val/clipfrac_avg": 0.0,
1249
+ "val/num_eos_tokens": 3595,
1250
+ "val/ratio": 1.0000882148742676,
1251
+ "val/ratio_var": 2.9485702270903857e-06
1252
+ },
1253
+ {
1254
+ "episode": 32256,
1255
+ "epoch": 0.13817683344756682,
1256
+ "eps": 6,
1257
+ "loss/policy_avg": 0.19996881484985352,
1258
+ "lr": 2.2734375e-06,
1259
+ "objective/entropy": -67.38516235351562,
1260
+ "objective/kl": 23.758987426757812,
1261
+ "objective/non_score_reward": 0.0,
1262
+ "objective/rlhf_reward": 8.108330726623535,
1263
+ "objective/scores": 8.108330726623535,
1264
+ "policy/approxkl_avg": 0.1896720677614212,
1265
+ "policy/clipfrac_avg": 0.26953125,
1266
+ "policy/entropy_avg": 0.5402642488479614,
1267
+ "step": 62,
1268
+ "val/clipfrac_avg": 0.0,
1269
+ "val/num_eos_tokens": 3944,
1270
+ "val/ratio": 1.0001147985458374,
1271
+ "val/ratio_var": 3.3895339583978057e-06
1272
+ },
1273
+ {
1274
+ "episode": 32768,
1275
+ "epoch": 0.1403701165181631,
1276
+ "eps": 6,
1277
+ "loss/policy_avg": 0.13677772879600525,
1278
+ "lr": 2.26171875e-06,
1279
+ "objective/entropy": -71.50131225585938,
1280
+ "objective/kl": 23.17635726928711,
1281
+ "objective/non_score_reward": 0.0,
1282
+ "objective/rlhf_reward": 8.140105247497559,
1283
+ "objective/scores": 8.140105247497559,
1284
+ "policy/approxkl_avg": 0.20787930488586426,
1285
+ "policy/clipfrac_avg": 0.32421875,
1286
+ "policy/entropy_avg": 0.5493172407150269,
1287
+ "step": 63,
1288
+ "val/clipfrac_avg": 0.0,
1289
+ "val/num_eos_tokens": 3830,
1290
+ "val/ratio": 0.9997697472572327,
1291
+ "val/ratio_var": 2.91177570943546e-06
1292
+ },
1293
+ {
1294
+ "episode": 33280,
1295
+ "epoch": 0.14256339958875944,
1296
+ "eps": 6,
1297
+ "loss/policy_avg": 0.20616966485977173,
1298
+ "lr": 2.25e-06,
1299
+ "objective/entropy": -70.77236938476562,
1300
+ "objective/kl": 24.19706153869629,
1301
+ "objective/non_score_reward": 0.0,
1302
+ "objective/rlhf_reward": 8.366374015808105,
1303
+ "objective/scores": 8.366374015808105,
1304
+ "policy/approxkl_avg": 0.18721507489681244,
1305
+ "policy/clipfrac_avg": 0.26171875,
1306
+ "policy/entropy_avg": 0.555661678314209,
1307
+ "step": 64,
1308
+ "val/clipfrac_avg": 0.0,
1309
+ "val/num_eos_tokens": 3883,
1310
+ "val/ratio": 0.9998935461044312,
1311
+ "val/ratio_var": 2.8920271688548382e-06
1312
+ },
1313
+ {
1314
+ "episode": 33792,
1315
+ "epoch": 0.14475668265935573,
1316
+ "eps": 6,
1317
+ "loss/policy_avg": 0.14455527067184448,
1318
+ "lr": 2.23828125e-06,
1319
+ "objective/entropy": -71.4616928100586,
1320
+ "objective/kl": 24.86420440673828,
1321
+ "objective/non_score_reward": 0.0,
1322
+ "objective/rlhf_reward": 8.183595657348633,
1323
+ "objective/scores": 8.183595657348633,
1324
+ "policy/approxkl_avg": 0.16570612788200378,
1325
+ "policy/clipfrac_avg": 0.267578125,
1326
+ "policy/entropy_avg": 0.527056872844696,
1327
+ "step": 65,
1328
+ "val/clipfrac_avg": 0.0,
1329
+ "val/num_eos_tokens": 4113,
1330
+ "val/ratio": 1.0001296997070312,
1331
+ "val/ratio_var": 2.1795549400849268e-06
1332
+ },
1333
+ {
1334
+ "episode": 34304,
1335
+ "epoch": 0.14694996572995203,
1336
+ "eps": 6,
1337
+ "loss/policy_avg": 0.1541716307401657,
1338
+ "lr": 2.2265625e-06,
1339
+ "objective/entropy": -73.83854675292969,
1340
+ "objective/kl": 23.82467269897461,
1341
+ "objective/non_score_reward": 0.0,
1342
+ "objective/rlhf_reward": 8.430439949035645,
1343
+ "objective/scores": 8.430439949035645,
1344
+ "policy/approxkl_avg": 0.17509257793426514,
1345
+ "policy/clipfrac_avg": 0.30078125,
1346
+ "policy/entropy_avg": 0.5003349781036377,
1347
+ "step": 66,
1348
+ "val/clipfrac_avg": 0.0,
1349
+ "val/num_eos_tokens": 4152,
1350
+ "val/ratio": 0.999921441078186,
1351
+ "val/ratio_var": 2.0571305867633782e-06
1352
+ },
1353
+ {
1354
+ "episode": 34816,
1355
+ "epoch": 0.14914324880054833,
1356
+ "eps": 6,
1357
+ "loss/policy_avg": 0.2405724823474884,
1358
+ "lr": 2.21484375e-06,
1359
+ "objective/entropy": -73.2585220336914,
1360
+ "objective/kl": 24.3407039642334,
1361
+ "objective/non_score_reward": 0.0,
1362
+ "objective/rlhf_reward": 8.195615768432617,
1363
+ "objective/scores": 8.195615768432617,
1364
+ "policy/approxkl_avg": 0.17023462057113647,
1365
+ "policy/clipfrac_avg": 0.2890625,
1366
+ "policy/entropy_avg": 0.46991419792175293,
1367
+ "step": 67,
1368
+ "val/clipfrac_avg": 0.0,
1369
+ "val/num_eos_tokens": 4058,
1370
+ "val/ratio": 0.9999005198478699,
1371
+ "val/ratio_var": 1.952504135260824e-06
1372
+ },
1373
+ {
1374
+ "episode": 35328,
1375
+ "epoch": 0.15133653187114463,
1376
+ "eps": 6,
1377
+ "loss/policy_avg": 0.17481882870197296,
1378
+ "lr": 2.203125e-06,
1379
+ "objective/entropy": -74.03469848632812,
1380
+ "objective/kl": 24.17955207824707,
1381
+ "objective/non_score_reward": 0.0,
1382
+ "objective/rlhf_reward": 8.267098426818848,
1383
+ "objective/scores": 8.267098426818848,
1384
+ "policy/approxkl_avg": 0.1741240918636322,
1385
+ "policy/clipfrac_avg": 0.26171875,
1386
+ "policy/entropy_avg": 0.4679723083972931,
1387
+ "step": 68,
1388
+ "val/clipfrac_avg": 0.0,
1389
+ "val/num_eos_tokens": 3994,
1390
+ "val/ratio": 1.000124216079712,
1391
+ "val/ratio_var": 2.194252829212928e-06
1392
+ },
1393
+ {
1394
+ "episode": 35840,
1395
+ "epoch": 0.15352981494174092,
1396
+ "eps": 6,
1397
+ "loss/policy_avg": 0.2236326038837433,
1398
+ "lr": 2.19140625e-06,
1399
+ "objective/entropy": -74.11293029785156,
1400
+ "objective/kl": 25.54970932006836,
1401
+ "objective/non_score_reward": 0.0,
1402
+ "objective/rlhf_reward": 8.242114067077637,
1403
+ "objective/scores": 8.242114067077637,
1404
+ "policy/approxkl_avg": 0.16678233444690704,
1405
+ "policy/clipfrac_avg": 0.275390625,
1406
+ "policy/entropy_avg": 0.459768146276474,
1407
+ "step": 69,
1408
+ "val/clipfrac_avg": 0.0,
1409
+ "val/num_eos_tokens": 3875,
1410
+ "val/ratio": 0.9999948740005493,
1411
+ "val/ratio_var": 2.9980392355355434e-06
1412
+ },
1413
+ {
1414
+ "episode": 36352,
1415
+ "epoch": 0.15572309801233722,
1416
+ "eps": 6,
1417
+ "loss/policy_avg": 0.2104223519563675,
1418
+ "lr": 2.1796875e-06,
1419
+ "objective/entropy": -70.23101806640625,
1420
+ "objective/kl": 26.61813735961914,
1421
+ "objective/non_score_reward": 0.0,
1422
+ "objective/rlhf_reward": 8.240058898925781,
1423
+ "objective/scores": 8.240058898925781,
1424
+ "policy/approxkl_avg": 0.17932359874248505,
1425
+ "policy/clipfrac_avg": 0.26171875,
1426
+ "policy/entropy_avg": 0.44719618558883667,
1427
+ "step": 70,
1428
+ "val/clipfrac_avg": 0.0,
1429
+ "val/num_eos_tokens": 3672,
1430
+ "val/ratio": 1.0002201795578003,
1431
+ "val/ratio_var": 2.288593123012106e-06
1432
+ },
1433
+ {
1434
+ "episode": 36864,
1435
+ "epoch": 0.15791638108293352,
1436
+ "eps": 6,
1437
+ "loss/policy_avg": 0.18882869184017181,
1438
+ "lr": 2.16796875e-06,
1439
+ "objective/entropy": -72.48193359375,
1440
+ "objective/kl": 26.345766067504883,
1441
+ "objective/non_score_reward": 0.0,
1442
+ "objective/rlhf_reward": 8.317778587341309,
1443
+ "objective/scores": 8.317778587341309,
1444
+ "policy/approxkl_avg": 0.2453937828540802,
1445
+ "policy/clipfrac_avg": 0.2578125,
1446
+ "policy/entropy_avg": 0.43665581941604614,
1447
+ "step": 71,
1448
+ "val/clipfrac_avg": 0.0,
1449
+ "val/num_eos_tokens": 3908,
1450
+ "val/ratio": 1.0000020265579224,
1451
+ "val/ratio_var": 4.2595497689035255e-06
1452
+ },
1453
+ {
1454
+ "episode": 37376,
1455
+ "epoch": 0.1601096641535298,
1456
+ "eps": 6,
1457
+ "loss/policy_avg": 0.19868019223213196,
1458
+ "lr": 2.15625e-06,
1459
+ "objective/entropy": -70.25636291503906,
1460
+ "objective/kl": 27.215253829956055,
1461
+ "objective/non_score_reward": 0.0,
1462
+ "objective/rlhf_reward": 8.491509437561035,
1463
+ "objective/scores": 8.491509437561035,
1464
+ "policy/approxkl_avg": 0.18058954179286957,
1465
+ "policy/clipfrac_avg": 0.265625,
1466
+ "policy/entropy_avg": 0.4683910608291626,
1467
+ "step": 72,
1468
+ "val/clipfrac_avg": 0.0,
1469
+ "val/num_eos_tokens": 3603,
1470
+ "val/ratio": 0.999990701675415,
1471
+ "val/ratio_var": 2.4616608698124764e-06
1472
+ },
1473
+ {
1474
+ "episode": 37888,
1475
+ "epoch": 0.1623029472241261,
1476
+ "eps": 6,
1477
+ "loss/policy_avg": 0.1968342512845993,
1478
+ "lr": 2.14453125e-06,
1479
+ "objective/entropy": -68.2578353881836,
1480
+ "objective/kl": 28.091264724731445,
1481
+ "objective/non_score_reward": 0.0,
1482
+ "objective/rlhf_reward": 8.216062545776367,
1483
+ "objective/scores": 8.216062545776367,
1484
+ "policy/approxkl_avg": 0.19425219297409058,
1485
+ "policy/clipfrac_avg": 0.298828125,
1486
+ "policy/entropy_avg": 0.4262210726737976,
1487
+ "step": 73,
1488
+ "val/clipfrac_avg": 0.0,
1489
+ "val/num_eos_tokens": 3467,
1490
+ "val/ratio": 1.000177264213562,
1491
+ "val/ratio_var": 2.9407167403405765e-06
1492
+ },
1493
+ {
1494
+ "episode": 38400,
1495
+ "epoch": 0.1644962302947224,
1496
+ "eps": 6,
1497
+ "loss/policy_avg": 0.21408286690711975,
1498
+ "lr": 2.1328125e-06,
1499
+ "objective/entropy": -70.67361450195312,
1500
+ "objective/kl": 27.946510314941406,
1501
+ "objective/non_score_reward": 0.0,
1502
+ "objective/rlhf_reward": 8.111261367797852,
1503
+ "objective/scores": 8.111261367797852,
1504
+ "policy/approxkl_avg": 0.21669934689998627,
1505
+ "policy/clipfrac_avg": 0.232421875,
1506
+ "policy/entropy_avg": 0.4281734228134155,
1507
+ "step": 74,
1508
+ "val/clipfrac_avg": 0.0,
1509
+ "val/num_eos_tokens": 3512,
1510
+ "val/ratio": 1.0006786584854126,
1511
+ "val/ratio_var": 2.5324472517240793e-05
1512
+ },
1513
+ {
1514
+ "episode": 38912,
1515
+ "epoch": 0.1666895133653187,
1516
+ "eps": 6,
1517
+ "loss/policy_avg": 0.1531296968460083,
1518
+ "lr": 2.12109375e-06,
1519
+ "objective/entropy": -69.2296142578125,
1520
+ "objective/kl": 28.850563049316406,
1521
+ "objective/non_score_reward": 0.0,
1522
+ "objective/rlhf_reward": 8.479706764221191,
1523
+ "objective/scores": 8.479706764221191,
1524
+ "policy/approxkl_avg": 0.20146965980529785,
1525
+ "policy/clipfrac_avg": 0.306640625,
1526
+ "policy/entropy_avg": 0.4345015287399292,
1527
+ "step": 75,
1528
+ "val/clipfrac_avg": 0.0,
1529
+ "val/num_eos_tokens": 3428,
1530
+ "val/ratio": 1.0000364780426025,
1531
+ "val/ratio_var": 2.9701880066568265e-06
1532
+ },
1533
+ {
1534
+ "episode": 39424,
1535
+ "epoch": 0.168882796435915,
1536
+ "eps": 6,
1537
+ "loss/policy_avg": 0.17767590284347534,
1538
+ "lr": 2.109375e-06,
1539
+ "objective/entropy": -70.01402282714844,
1540
+ "objective/kl": 28.191417694091797,
1541
+ "objective/non_score_reward": 0.0,
1542
+ "objective/rlhf_reward": 8.29566764831543,
1543
+ "objective/scores": 8.29566764831543,
1544
+ "policy/approxkl_avg": 0.1757415235042572,
1545
+ "policy/clipfrac_avg": 0.26953125,
1546
+ "policy/entropy_avg": 0.4495161175727844,
1547
+ "step": 76,
1548
+ "val/clipfrac_avg": 0.0,
1549
+ "val/num_eos_tokens": 3349,
1550
+ "val/ratio": 0.9998478889465332,
1551
+ "val/ratio_var": 2.1308803752617678e-06
1552
+ },
1553
+ {
1554
+ "episode": 39936,
1555
+ "epoch": 0.1710760795065113,
1556
+ "eps": 6,
1557
+ "loss/policy_avg": 0.33594539761543274,
1558
+ "lr": 2.09765625e-06,
1559
+ "objective/entropy": -69.42768859863281,
1560
+ "objective/kl": 28.3437442779541,
1561
+ "objective/non_score_reward": 0.0,
1562
+ "objective/rlhf_reward": 8.23046875,
1563
+ "objective/scores": 8.23046875,
1564
+ "policy/approxkl_avg": 0.20253780484199524,
1565
+ "policy/clipfrac_avg": 0.2734375,
1566
+ "policy/entropy_avg": 0.433658242225647,
1567
+ "step": 77,
1568
+ "val/clipfrac_avg": 0.0,
1569
+ "val/num_eos_tokens": 3187,
1570
+ "val/ratio": 0.9998530149459839,
1571
+ "val/ratio_var": 1.98079601432255e-06
1572
+ },
1573
+ {
1574
+ "episode": 40448,
1575
+ "epoch": 0.17326936257710762,
1576
+ "eps": 6,
1577
+ "loss/policy_avg": 0.19012102484703064,
1578
+ "lr": 2.0859375e-06,
1579
+ "objective/entropy": -66.31437683105469,
1580
+ "objective/kl": 29.664644241333008,
1581
+ "objective/non_score_reward": 0.0,
1582
+ "objective/rlhf_reward": 8.34729290008545,
1583
+ "objective/scores": 8.34729290008545,
1584
+ "policy/approxkl_avg": 0.22112396359443665,
1585
+ "policy/clipfrac_avg": 0.310546875,
1586
+ "policy/entropy_avg": 0.42181527614593506,
1587
+ "step": 78,
1588
+ "val/clipfrac_avg": 0.0,
1589
+ "val/num_eos_tokens": 3218,
1590
+ "val/ratio": 0.9998316168785095,
1591
+ "val/ratio_var": 3.1733356991026085e-06
1592
+ },
1593
+ {
1594
+ "episode": 40960,
1595
+ "epoch": 0.17546264564770392,
1596
+ "eps": 6,
1597
+ "loss/policy_avg": 0.11748410761356354,
1598
+ "lr": 2.07421875e-06,
1599
+ "objective/entropy": -67.87670135498047,
1600
+ "objective/kl": 30.676856994628906,
1601
+ "objective/non_score_reward": 0.0,
1602
+ "objective/rlhf_reward": 8.11823844909668,
1603
+ "objective/scores": 8.11823844909668,
1604
+ "policy/approxkl_avg": 0.27489668130874634,
1605
+ "policy/clipfrac_avg": 0.322265625,
1606
+ "policy/entropy_avg": 0.4536714553833008,
1607
+ "step": 79,
1608
+ "val/clipfrac_avg": 0.0,
1609
+ "val/num_eos_tokens": 3259,
1610
+ "val/ratio": 0.9998610615730286,
1611
+ "val/ratio_var": 2.2992560388956917e-06
1612
+ },
1613
+ {
1614
+ "episode": 41472,
1615
+ "epoch": 0.17765592871830022,
1616
+ "eps": 6,
1617
+ "loss/policy_avg": 0.2511185109615326,
1618
+ "lr": 2.0625e-06,
1619
+ "objective/entropy": -65.38703155517578,
1620
+ "objective/kl": 30.458276748657227,
1621
+ "objective/non_score_reward": 0.0,
1622
+ "objective/rlhf_reward": 8.453354835510254,
1623
+ "objective/scores": 8.453354835510254,
1624
+ "policy/approxkl_avg": 0.37790045142173767,
1625
+ "policy/clipfrac_avg": 0.34765625,
1626
+ "policy/entropy_avg": 0.4551919102668762,
1627
+ "step": 80,
1628
+ "val/clipfrac_avg": 0.0,
1629
+ "val/num_eos_tokens": 3147,
1630
+ "val/ratio": 1.0001468658447266,
1631
+ "val/ratio_var": 1.2064292604918592e-05
1632
+ },
1633
+ {
1634
+ "episode": 41984,
1635
+ "epoch": 0.1798492117888965,
1636
+ "eps": 6,
1637
+ "loss/policy_avg": 0.2766464948654175,
1638
+ "lr": 2.05078125e-06,
1639
+ "objective/entropy": -64.6202163696289,
1640
+ "objective/kl": 31.10712432861328,
1641
+ "objective/non_score_reward": 0.0,
1642
+ "objective/rlhf_reward": 8.16506290435791,
1643
+ "objective/scores": 8.16506290435791,
1644
+ "policy/approxkl_avg": 0.2353859394788742,
1645
+ "policy/clipfrac_avg": 0.330078125,
1646
+ "policy/entropy_avg": 0.44192785024642944,
1647
+ "step": 81,
1648
+ "val/clipfrac_avg": 0.0,
1649
+ "val/num_eos_tokens": 3189,
1650
+ "val/ratio": 0.9998602867126465,
1651
+ "val/ratio_var": 3.1267311442206847e-06
1652
+ },
1653
+ {
1654
+ "episode": 42496,
1655
+ "epoch": 0.1820424948594928,
1656
+ "eps": 6,
1657
+ "loss/policy_avg": 0.19304512441158295,
1658
+ "lr": 2.0390625e-06,
1659
+ "objective/entropy": -64.17574310302734,
1660
+ "objective/kl": 30.861011505126953,
1661
+ "objective/non_score_reward": 0.0,
1662
+ "objective/rlhf_reward": 8.137439727783203,
1663
+ "objective/scores": 8.137439727783203,
1664
+ "policy/approxkl_avg": 0.24140365421772003,
1665
+ "policy/clipfrac_avg": 0.30859375,
1666
+ "policy/entropy_avg": 0.43526899814605713,
1667
+ "step": 82,
1668
+ "val/clipfrac_avg": 0.0,
1669
+ "val/num_eos_tokens": 3233,
1670
+ "val/ratio": 0.9996709823608398,
1671
+ "val/ratio_var": 3.042540356545942e-06
1672
+ },
1673
+ {
1674
+ "episode": 43008,
1675
+ "epoch": 0.1842357779300891,
1676
+ "eps": 6,
1677
+ "loss/policy_avg": 0.21941827237606049,
1678
+ "lr": 2.02734375e-06,
1679
+ "objective/entropy": -63.085872650146484,
1680
+ "objective/kl": 31.999509811401367,
1681
+ "objective/non_score_reward": 0.0,
1682
+ "objective/rlhf_reward": 8.363846778869629,
1683
+ "objective/scores": 8.363846778869629,
1684
+ "policy/approxkl_avg": 0.23224441707134247,
1685
+ "policy/clipfrac_avg": 0.3359375,
1686
+ "policy/entropy_avg": 0.45573890209198,
1687
+ "step": 83,
1688
+ "val/clipfrac_avg": 0.0,
1689
+ "val/num_eos_tokens": 3165,
1690
+ "val/ratio": 0.9997715950012207,
1691
+ "val/ratio_var": 3.3492219699837733e-06
1692
+ },
1693
+ {
1694
+ "episode": 43520,
1695
+ "epoch": 0.1864290610006854,
1696
+ "eps": 6,
1697
+ "loss/policy_avg": 0.17782308161258698,
1698
+ "lr": 2.015625e-06,
1699
+ "objective/entropy": -63.105010986328125,
1700
+ "objective/kl": 30.78789520263672,
1701
+ "objective/non_score_reward": 0.0,
1702
+ "objective/rlhf_reward": 8.67146110534668,
1703
+ "objective/scores": 8.67146110534668,
1704
+ "policy/approxkl_avg": 0.24423187971115112,
1705
+ "policy/clipfrac_avg": 0.3046875,
1706
+ "policy/entropy_avg": 0.4529142379760742,
1707
+ "step": 84,
1708
+ "val/clipfrac_avg": 0.0,
1709
+ "val/num_eos_tokens": 3087,
1710
+ "val/ratio": 0.9999881386756897,
1711
+ "val/ratio_var": 2.5693764200696023e-06
1712
+ },
1713
+ {
1714
+ "episode": 44032,
1715
+ "epoch": 0.1886223440712817,
1716
+ "eps": 6,
1717
+ "loss/policy_avg": 0.16561922430992126,
1718
+ "lr": 2.00390625e-06,
1719
+ "objective/entropy": -60.95121765136719,
1720
+ "objective/kl": 31.70094108581543,
1721
+ "objective/non_score_reward": 0.0,
1722
+ "objective/rlhf_reward": 8.50253677368164,
1723
+ "objective/scores": 8.50253677368164,
1724
+ "policy/approxkl_avg": 0.24480891227722168,
1725
+ "policy/clipfrac_avg": 0.31640625,
1726
+ "policy/entropy_avg": 0.45602551102638245,
1727
+ "step": 85,
1728
+ "val/clipfrac_avg": 0.0,
1729
+ "val/num_eos_tokens": 2839,
1730
+ "val/ratio": 1.0002214908599854,
1731
+ "val/ratio_var": 4.56425777883851e-06
1732
+ },
1733
+ {
1734
+ "episode": 44544,
1735
+ "epoch": 0.190815627141878,
1736
+ "eps": 6,
1737
+ "loss/policy_avg": 0.26409026980400085,
1738
+ "lr": 1.9921875e-06,
1739
+ "objective/entropy": -62.31785583496094,
1740
+ "objective/kl": 31.997968673706055,
1741
+ "objective/non_score_reward": 0.0,
1742
+ "objective/rlhf_reward": 8.572611808776855,
1743
+ "objective/scores": 8.572611808776855,
1744
+ "policy/approxkl_avg": 0.2332761436700821,
1745
+ "policy/clipfrac_avg": 0.30859375,
1746
+ "policy/entropy_avg": 0.48270928859710693,
1747
+ "step": 86,
1748
+ "val/clipfrac_avg": 0.0,
1749
+ "val/num_eos_tokens": 3049,
1750
+ "val/ratio": 0.9998488426208496,
1751
+ "val/ratio_var": 2.2608548988500843e-06
1752
+ },
1753
+ {
1754
+ "episode": 45056,
1755
+ "epoch": 0.1930089102124743,
1756
+ "eps": 6,
1757
+ "loss/policy_avg": 0.2466927170753479,
1758
+ "lr": 1.98046875e-06,
1759
+ "objective/entropy": -63.5084114074707,
1760
+ "objective/kl": 32.784034729003906,
1761
+ "objective/non_score_reward": 0.0,
1762
+ "objective/rlhf_reward": 8.442138671875,
1763
+ "objective/scores": 8.442138671875,
1764
+ "policy/approxkl_avg": 0.19753462076187134,
1765
+ "policy/clipfrac_avg": 0.330078125,
1766
+ "policy/entropy_avg": 0.46065136790275574,
1767
+ "step": 87,
1768
+ "val/clipfrac_avg": 0.0,
1769
+ "val/num_eos_tokens": 3216,
1770
+ "val/ratio": 0.9997737407684326,
1771
+ "val/ratio_var": 1.8330310922465287e-06
1772
+ },
1773
+ {
1774
+ "episode": 45568,
1775
+ "epoch": 0.1952021932830706,
1776
+ "eps": 6,
1777
+ "loss/policy_avg": 0.21488100290298462,
1778
+ "lr": 1.96875e-06,
1779
+ "objective/entropy": -62.979454040527344,
1780
+ "objective/kl": 31.44158363342285,
1781
+ "objective/non_score_reward": 0.0,
1782
+ "objective/rlhf_reward": 8.53417682647705,
1783
+ "objective/scores": 8.53417682647705,
1784
+ "policy/approxkl_avg": 0.2771834433078766,
1785
+ "policy/clipfrac_avg": 0.33203125,
1786
+ "policy/entropy_avg": 0.468097984790802,
1787
+ "step": 88,
1788
+ "val/clipfrac_avg": 0.0,
1789
+ "val/num_eos_tokens": 2921,
1790
+ "val/ratio": 1.0001518726348877,
1791
+ "val/ratio_var": 3.3578060083527816e-06
1792
+ },
1793
+ {
1794
+ "episode": 46080,
1795
+ "epoch": 0.1973954763536669,
1796
+ "eps": 6,
1797
+ "loss/policy_avg": 0.3296290934085846,
1798
+ "lr": 1.95703125e-06,
1799
+ "objective/entropy": -58.682701110839844,
1800
+ "objective/kl": 34.13310623168945,
1801
+ "objective/non_score_reward": 0.0,
1802
+ "objective/rlhf_reward": 8.303296089172363,
1803
+ "objective/scores": 8.303296089172363,
1804
+ "policy/approxkl_avg": 0.37636691331863403,
1805
+ "policy/clipfrac_avg": 0.345703125,
1806
+ "policy/entropy_avg": 0.49518346786499023,
1807
+ "step": 89,
1808
+ "val/clipfrac_avg": 0.0,
1809
+ "val/num_eos_tokens": 2719,
1810
+ "val/ratio": 0.9997811317443848,
1811
+ "val/ratio_var": 3.208933321730001e-06
1812
+ },
1813
+ {
1814
+ "episode": 46592,
1815
+ "epoch": 0.19958875942426318,
1816
+ "eps": 6,
1817
+ "loss/policy_avg": 0.257150799036026,
1818
+ "lr": 1.9453125e-06,
1819
+ "objective/entropy": -62.256492614746094,
1820
+ "objective/kl": 31.841196060180664,
1821
+ "objective/non_score_reward": 0.0,
1822
+ "objective/rlhf_reward": 8.305122375488281,
1823
+ "objective/scores": 8.305122375488281,
1824
+ "policy/approxkl_avg": 0.24115335941314697,
1825
+ "policy/clipfrac_avg": 0.341796875,
1826
+ "policy/entropy_avg": 0.47345805168151855,
1827
+ "step": 90,
1828
+ "val/clipfrac_avg": 0.0,
1829
+ "val/num_eos_tokens": 3045,
1830
+ "val/ratio": 0.9996979236602783,
1831
+ "val/ratio_var": 2.477581347193336e-06
1832
+ },
1833
+ {
1834
+ "episode": 47104,
1835
+ "epoch": 0.20178204249485948,
1836
+ "eps": 6,
1837
+ "loss/policy_avg": 0.16495227813720703,
1838
+ "lr": 1.93359375e-06,
1839
+ "objective/entropy": -60.697792053222656,
1840
+ "objective/kl": 33.659996032714844,
1841
+ "objective/non_score_reward": 0.0,
1842
+ "objective/rlhf_reward": 8.492660522460938,
1843
+ "objective/scores": 8.492660522460938,
1844
+ "policy/approxkl_avg": 0.2478049099445343,
1845
+ "policy/clipfrac_avg": 0.365234375,
1846
+ "policy/entropy_avg": 0.47592592239379883,
1847
+ "step": 91,
1848
+ "val/clipfrac_avg": 0.0,
1849
+ "val/num_eos_tokens": 2758,
1850
+ "val/ratio": 0.9998412132263184,
1851
+ "val/ratio_var": 4.479879407881526e-06
1852
+ },
1853
+ {
1854
+ "episode": 47616,
1855
+ "epoch": 0.20397532556545578,
1856
+ "eps": 6,
1857
+ "loss/policy_avg": 0.24863702058792114,
1858
+ "lr": 1.921875e-06,
1859
+ "objective/entropy": -57.656898498535156,
1860
+ "objective/kl": 36.29644012451172,
1861
+ "objective/non_score_reward": 0.0,
1862
+ "objective/rlhf_reward": 8.122814178466797,
1863
+ "objective/scores": 8.122814178466797,
1864
+ "policy/approxkl_avg": 0.23853012919425964,
1865
+ "policy/clipfrac_avg": 0.314453125,
1866
+ "policy/entropy_avg": 0.4676399230957031,
1867
+ "step": 92,
1868
+ "val/clipfrac_avg": 0.0,
1869
+ "val/num_eos_tokens": 2766,
1870
+ "val/ratio": 1.0002670288085938,
1871
+ "val/ratio_var": 4.800449460162781e-06
1872
+ },
1873
+ {
1874
+ "episode": 48128,
1875
+ "epoch": 0.2061686086360521,
1876
+ "eps": 6,
1877
+ "loss/policy_avg": 0.33681780099868774,
1878
+ "lr": 1.91015625e-06,
1879
+ "objective/entropy": -57.09372329711914,
1880
+ "objective/kl": 36.896705627441406,
1881
+ "objective/non_score_reward": 0.0,
1882
+ "objective/rlhf_reward": 8.56679916381836,
1883
+ "objective/scores": 8.56679916381836,
1884
+ "policy/approxkl_avg": 0.23770955204963684,
1885
+ "policy/clipfrac_avg": 0.31640625,
1886
+ "policy/entropy_avg": 0.4676530957221985,
1887
+ "step": 93,
1888
+ "val/clipfrac_avg": 0.0,
1889
+ "val/num_eos_tokens": 2802,
1890
+ "val/ratio": 1.000120759010315,
1891
+ "val/ratio_var": 3.0969872568675783e-06
1892
+ },
1893
+ {
1894
+ "episode": 48640,
1895
+ "epoch": 0.2083618917066484,
1896
+ "eps": 6,
1897
+ "loss/policy_avg": 0.21784643828868866,
1898
+ "lr": 1.8984375e-06,
1899
+ "objective/entropy": -57.76728820800781,
1900
+ "objective/kl": 36.6060905456543,
1901
+ "objective/non_score_reward": 0.0,
1902
+ "objective/rlhf_reward": 8.36257266998291,
1903
+ "objective/scores": 8.36257266998291,
1904
+ "policy/approxkl_avg": 0.24328580498695374,
1905
+ "policy/clipfrac_avg": 0.3046875,
1906
+ "policy/entropy_avg": 0.486191987991333,
1907
+ "step": 94,
1908
+ "val/clipfrac_avg": 0.0,
1909
+ "val/num_eos_tokens": 2915,
1910
+ "val/ratio": 1.0001773834228516,
1911
+ "val/ratio_var": 3.4632869301276514e-06
1912
+ },
1913
+ {
1914
+ "episode": 49152,
1915
+ "epoch": 0.2105551747772447,
1916
+ "eps": 6,
1917
+ "loss/policy_avg": 0.19738544523715973,
1918
+ "lr": 1.8867187500000001e-06,
1919
+ "objective/entropy": -57.4933967590332,
1920
+ "objective/kl": 36.51211166381836,
1921
+ "objective/non_score_reward": 0.0,
1922
+ "objective/rlhf_reward": 8.441587448120117,
1923
+ "objective/scores": 8.441587448120117,
1924
+ "policy/approxkl_avg": 0.256747305393219,
1925
+ "policy/clipfrac_avg": 0.345703125,
1926
+ "policy/entropy_avg": 0.47086185216903687,
1927
+ "step": 95,
1928
+ "val/clipfrac_avg": 0.0,
1929
+ "val/num_eos_tokens": 2966,
1930
+ "val/ratio": 0.9996964931488037,
1931
+ "val/ratio_var": 2.7482572022563545e-06
1932
+ },
1933
+ {
1934
+ "episode": 49664,
1935
+ "epoch": 0.212748457847841,
1936
+ "eps": 7,
1937
+ "loss/policy_avg": 0.20914721488952637,
1938
+ "lr": 1.875e-06,
1939
+ "objective/entropy": -56.415225982666016,
1940
+ "objective/kl": 35.90780258178711,
1941
+ "objective/non_score_reward": 0.0,
1942
+ "objective/rlhf_reward": 8.193758964538574,
1943
+ "objective/scores": 8.193758964538574,
1944
+ "policy/approxkl_avg": 0.23544546961784363,
1945
+ "policy/clipfrac_avg": 0.333984375,
1946
+ "policy/entropy_avg": 0.4839785099029541,
1947
+ "step": 96,
1948
+ "val/clipfrac_avg": 0.0,
1949
+ "val/num_eos_tokens": 3114,
1950
+ "val/ratio": 0.9998562335968018,
1951
+ "val/ratio_var": 2.716004701142083e-06
1952
+ },
1953
+ {
1954
+ "episode": 50176,
1955
+ "epoch": 0.2149417409184373,
1956
+ "eps": 7,
1957
+ "loss/policy_avg": 0.1869872510433197,
1958
+ "lr": 1.86328125e-06,
1959
+ "objective/entropy": -56.17817687988281,
1960
+ "objective/kl": 38.056846618652344,
1961
+ "objective/non_score_reward": 0.0,
1962
+ "objective/rlhf_reward": 8.39655590057373,
1963
+ "objective/scores": 8.39655590057373,
1964
+ "policy/approxkl_avg": 0.2423265278339386,
1965
+ "policy/clipfrac_avg": 0.34375,
1966
+ "policy/entropy_avg": 0.4714318513870239,
1967
+ "step": 97,
1968
+ "val/clipfrac_avg": 0.0,
1969
+ "val/num_eos_tokens": 2883,
1970
+ "val/ratio": 0.9999042749404907,
1971
+ "val/ratio_var": 2.3066882022249047e-06
1972
+ },
1973
+ {
1974
+ "episode": 50688,
1975
+ "epoch": 0.2171350239890336,
1976
+ "eps": 7,
1977
+ "loss/policy_avg": 0.22085243463516235,
1978
+ "lr": 1.8515625000000001e-06,
1979
+ "objective/entropy": -56.651580810546875,
1980
+ "objective/kl": 36.86515808105469,
1981
+ "objective/non_score_reward": 0.0,
1982
+ "objective/rlhf_reward": 8.52478313446045,
1983
+ "objective/scores": 8.52478313446045,
1984
+ "policy/approxkl_avg": 0.21143081784248352,
1985
+ "policy/clipfrac_avg": 0.357421875,
1986
+ "policy/entropy_avg": 0.4802488088607788,
1987
+ "step": 98,
1988
+ "val/clipfrac_avg": 0.0,
1989
+ "val/num_eos_tokens": 3000,
1990
+ "val/ratio": 1.0000693798065186,
1991
+ "val/ratio_var": 2.6737234293250367e-06
1992
+ },
1993
+ {
1994
+ "episode": 51200,
1995
+ "epoch": 0.21932830705962988,
1996
+ "eps": 7,
1997
+ "loss/policy_avg": 0.21842840313911438,
1998
+ "lr": 1.83984375e-06,
1999
+ "objective/entropy": -56.499183654785156,
2000
+ "objective/kl": 37.15102005004883,
2001
+ "objective/non_score_reward": 0.0,
2002
+ "objective/rlhf_reward": 8.545164108276367,
2003
+ "objective/scores": 8.545164108276367,
2004
+ "policy/approxkl_avg": 0.23521457612514496,
2005
+ "policy/clipfrac_avg": 0.34765625,
2006
+ "policy/entropy_avg": 0.46721622347831726,
2007
+ "step": 99,
2008
+ "val/clipfrac_avg": 0.0,
2009
+ "val/num_eos_tokens": 3114,
2010
+ "val/ratio": 0.9996678233146667,
2011
+ "val/ratio_var": 3.113004140686826e-06
2012
+ },
2013
+ {
2014
+ "episode": 51712,
2015
+ "epoch": 0.22152159013022618,
2016
+ "eps": 7,
2017
+ "loss/policy_avg": 0.20084796845912933,
2018
+ "lr": 1.828125e-06,
2019
+ "objective/entropy": -53.680606842041016,
2020
+ "objective/kl": 40.248260498046875,
2021
+ "objective/non_score_reward": 0.0,
2022
+ "objective/rlhf_reward": 8.473481178283691,
2023
+ "objective/scores": 8.473481178283691,
2024
+ "policy/approxkl_avg": 0.22911617159843445,
2025
+ "policy/clipfrac_avg": 0.349609375,
2026
+ "policy/entropy_avg": 0.5091525316238403,
2027
+ "step": 100,
2028
+ "val/clipfrac_avg": 0.0,
2029
+ "val/num_eos_tokens": 3219,
2030
+ "val/ratio": 0.9998315572738647,
2031
+ "val/ratio_var": 3.1328345357906073e-06
2032
+ },
2033
+ {
2034
+ "episode": 52224,
2035
+ "epoch": 0.22371487320082248,
2036
+ "eps": 7,
2037
+ "loss/policy_avg": 0.21213960647583008,
2038
+ "lr": 1.81640625e-06,
2039
+ "objective/entropy": -55.78679656982422,
2040
+ "objective/kl": 37.61595916748047,
2041
+ "objective/non_score_reward": 0.0,
2042
+ "objective/rlhf_reward": 8.810025215148926,
2043
+ "objective/scores": 8.810025215148926,
2044
+ "policy/approxkl_avg": 0.19881097972393036,
2045
+ "policy/clipfrac_avg": 0.275390625,
2046
+ "policy/entropy_avg": 0.47496291995048523,
2047
+ "step": 101,
2048
+ "val/clipfrac_avg": 0.0,
2049
+ "val/num_eos_tokens": 3167,
2050
+ "val/ratio": 0.9999278783798218,
2051
+ "val/ratio_var": 2.2831025034975028e-06
2052
+ },
2053
+ {
2054
+ "episode": 52736,
2055
+ "epoch": 0.22590815627141878,
2056
+ "eps": 7,
2057
+ "loss/policy_avg": 0.2268027514219284,
2058
+ "lr": 1.8046875e-06,
2059
+ "objective/entropy": -56.765926361083984,
2060
+ "objective/kl": 36.390296936035156,
2061
+ "objective/non_score_reward": 0.0,
2062
+ "objective/rlhf_reward": 8.64309310913086,
2063
+ "objective/scores": 8.64309310913086,
2064
+ "policy/approxkl_avg": 0.3020716905593872,
2065
+ "policy/clipfrac_avg": 0.29296875,
2066
+ "policy/entropy_avg": 0.48782238364219666,
2067
+ "step": 102,
2068
+ "val/clipfrac_avg": 0.0,
2069
+ "val/num_eos_tokens": 3250,
2070
+ "val/ratio": 0.9999886155128479,
2071
+ "val/ratio_var": 9.797128768695984e-06
2072
+ },
2073
+ {
2074
+ "episode": 53248,
2075
+ "epoch": 0.22810143934201507,
2076
+ "eps": 7,
2077
+ "loss/policy_avg": 0.1790272742509842,
2078
+ "lr": 1.79296875e-06,
2079
+ "objective/entropy": -52.02347183227539,
2080
+ "objective/kl": 39.089656829833984,
2081
+ "objective/non_score_reward": 0.0,
2082
+ "objective/rlhf_reward": 8.757326126098633,
2083
+ "objective/scores": 8.757326126098633,
2084
+ "policy/approxkl_avg": 0.1798405945301056,
2085
+ "policy/clipfrac_avg": 0.341796875,
2086
+ "policy/entropy_avg": 0.47533801198005676,
2087
+ "step": 103,
2088
+ "val/clipfrac_avg": 0.0,
2089
+ "val/num_eos_tokens": 2851,
2090
+ "val/ratio": 1.0001875162124634,
2091
+ "val/ratio_var": 3.090276777584222e-06
2092
+ }
2093
+ ],
2094
+ "logging_steps": 1,
2095
+ "max_steps": 128,
2096
+ "num_input_tokens_seen": 0,
2097
+ "num_train_epochs": 1.122960932145305,
2098
+ "save_steps": 26,
2099
+ "stateful_callbacks": {
2100
+ "TrainerControl": {
2101
+ "args": {
2102
+ "should_epoch_stop": false,
2103
+ "should_evaluate": false,
2104
+ "should_log": true,
2105
+ "should_save": true,
2106
+ "should_training_stop": false
2107
+ },
2108
+ "attributes": {}
2109
+ }
2110
+ },
2111
+ "total_flos": 0,
2112
+ "train_batch_size": null,
2113
+ "trial_name": null,
2114
+ "trial_params": null
2115
+ }
checkpoint-104/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5e69611bee011a56c9eda92fc9354b22f6efc4357ae21e957602e49335de635
3
+ size 6584