Training in progress, step 135, checkpoint
Browse files- checkpoint-135/config.json +31 -0
- checkpoint-135/generation_config.json +5 -0
- checkpoint-135/model.safetensors +3 -0
- checkpoint-135/optimizer.pt +3 -0
- checkpoint-135/rng_state.pth +3 -0
- checkpoint-135/scheduler.pt +3 -0
- checkpoint-135/special_tokens_map.json +30 -0
- checkpoint-135/tokenizer.json +0 -0
- checkpoint-135/tokenizer_config.json +220 -0
- checkpoint-135/trainer_state.json +2735 -0
- checkpoint-135/training_args.bin +3 -0
checkpoint-135/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-135/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-135/model.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:67f9be6557378a2505406de2b34cdf22da660422fe52142cb999b22a90e8d7f3
|
3 |
+
size 1621370224
|
checkpoint-135/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:25567b438e610fe36728b4e6589b6d7f93f4474a56d1547f5d1ed36a033894ec
|
3 |
+
size 3242916026
|
checkpoint-135/rng_state.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0eb2675ab041568329aaea3487a71211e6610c8c752e215e656df84a4d8d7763
|
3 |
+
size 14308
|
checkpoint-135/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c899d8e0c23be141a2d7164712adee661b83ee7fea13a1481cea3e2a8dd8a50a
|
3 |
+
size 1064
|
checkpoint-135/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-135/tokenizer.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-135/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-135/trainer_state.json
ADDED
@@ -0,0 +1,2735 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_global_step": null,
|
3 |
+
"best_metric": null,
|
4 |
+
"best_model_checkpoint": null,
|
5 |
+
"episode": 69120,
|
6 |
+
"epoch": 0.29609321453050036,
|
7 |
+
"eval_steps": 500,
|
8 |
+
"global_step": 135,
|
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": 0.14956334233283997,
|
18 |
+
"lr": 3e-06,
|
19 |
+
"objective/entropy": -45.407432556152344,
|
20 |
+
"objective/kl": 0.3935524821281433,
|
21 |
+
"objective/non_score_reward": -0.019677624106407166,
|
22 |
+
"objective/rlhf_reward": 4.659376621246338,
|
23 |
+
"objective/scores": 4.679054260253906,
|
24 |
+
"policy/approxkl_avg": 0.11690396815538406,
|
25 |
+
"policy/clipfrac_avg": 0.345703125,
|
26 |
+
"policy/entropy_avg": 0.8878304958343506,
|
27 |
+
"step": 0,
|
28 |
+
"val/clipfrac_avg": 0.0,
|
29 |
+
"val/num_eos_tokens": 2729,
|
30 |
+
"val/ratio": 1.000101089477539,
|
31 |
+
"val/ratio_var": 1.7632828530622646e-06
|
32 |
+
},
|
33 |
+
{
|
34 |
+
"episode": 1024,
|
35 |
+
"epoch": 0.004386566141192597,
|
36 |
+
"eps": 6,
|
37 |
+
"loss/policy_avg": 0.12376213073730469,
|
38 |
+
"lr": 2.9882812500000002e-06,
|
39 |
+
"objective/entropy": -49.89072036743164,
|
40 |
+
"objective/kl": 0.5856812000274658,
|
41 |
+
"objective/non_score_reward": -0.02928406000137329,
|
42 |
+
"objective/rlhf_reward": 4.568350315093994,
|
43 |
+
"objective/scores": 4.597634315490723,
|
44 |
+
"policy/approxkl_avg": 0.09876184910535812,
|
45 |
+
"policy/clipfrac_avg": 0.328125,
|
46 |
+
"policy/entropy_avg": 0.8194690346717834,
|
47 |
+
"step": 1,
|
48 |
+
"val/clipfrac_avg": 0.0,
|
49 |
+
"val/num_eos_tokens": 3452,
|
50 |
+
"val/ratio": 1.0000417232513428,
|
51 |
+
"val/ratio_var": 1.4841128859188757e-06
|
52 |
+
},
|
53 |
+
{
|
54 |
+
"episode": 1536,
|
55 |
+
"epoch": 0.006579849211788897,
|
56 |
+
"eps": 6,
|
57 |
+
"loss/policy_avg": 0.18096689879894257,
|
58 |
+
"lr": 2.9765625e-06,
|
59 |
+
"objective/entropy": -49.44681930541992,
|
60 |
+
"objective/kl": 1.1118967533111572,
|
61 |
+
"objective/non_score_reward": -0.055594831705093384,
|
62 |
+
"objective/rlhf_reward": 4.8199357986450195,
|
63 |
+
"objective/scores": 4.87553071975708,
|
64 |
+
"policy/approxkl_avg": 0.1327531635761261,
|
65 |
+
"policy/clipfrac_avg": 0.32421875,
|
66 |
+
"policy/entropy_avg": 0.7880593538284302,
|
67 |
+
"step": 2,
|
68 |
+
"val/clipfrac_avg": 0.0,
|
69 |
+
"val/num_eos_tokens": 3433,
|
70 |
+
"val/ratio": 1.0000628232955933,
|
71 |
+
"val/ratio_var": 1.982313960979809e-06
|
72 |
+
},
|
73 |
+
{
|
74 |
+
"episode": 2048,
|
75 |
+
"epoch": 0.008773132282385195,
|
76 |
+
"eps": 6,
|
77 |
+
"loss/policy_avg": 0.15055972337722778,
|
78 |
+
"lr": 2.96484375e-06,
|
79 |
+
"objective/entropy": -48.96665954589844,
|
80 |
+
"objective/kl": 1.6585638523101807,
|
81 |
+
"objective/non_score_reward": -0.08292818814516068,
|
82 |
+
"objective/rlhf_reward": 4.777359485626221,
|
83 |
+
"objective/scores": 4.860287666320801,
|
84 |
+
"policy/approxkl_avg": 0.17208586633205414,
|
85 |
+
"policy/clipfrac_avg": 0.3046875,
|
86 |
+
"policy/entropy_avg": 0.7760875225067139,
|
87 |
+
"step": 3,
|
88 |
+
"val/clipfrac_avg": 0.0,
|
89 |
+
"val/num_eos_tokens": 2870,
|
90 |
+
"val/ratio": 0.9998505711555481,
|
91 |
+
"val/ratio_var": 1.9419330783421174e-06
|
92 |
+
},
|
93 |
+
{
|
94 |
+
"episode": 2560,
|
95 |
+
"epoch": 0.010966415352981495,
|
96 |
+
"eps": 6,
|
97 |
+
"loss/policy_avg": 0.2313799411058426,
|
98 |
+
"lr": 2.953125e-06,
|
99 |
+
"objective/entropy": -50.714088439941406,
|
100 |
+
"objective/kl": 2.431265354156494,
|
101 |
+
"objective/non_score_reward": -0.12156327068805695,
|
102 |
+
"objective/rlhf_reward": 4.976639747619629,
|
103 |
+
"objective/scores": 5.098202705383301,
|
104 |
+
"policy/approxkl_avg": 0.12468202412128448,
|
105 |
+
"policy/clipfrac_avg": 0.30078125,
|
106 |
+
"policy/entropy_avg": 0.7334984540939331,
|
107 |
+
"step": 4,
|
108 |
+
"val/clipfrac_avg": 0.0,
|
109 |
+
"val/num_eos_tokens": 2548,
|
110 |
+
"val/ratio": 1.0002238750457764,
|
111 |
+
"val/ratio_var": 1.9082790458924137e-06
|
112 |
+
},
|
113 |
+
{
|
114 |
+
"episode": 3072,
|
115 |
+
"epoch": 0.013159698423577794,
|
116 |
+
"eps": 6,
|
117 |
+
"loss/policy_avg": 0.14224857091903687,
|
118 |
+
"lr": 2.94140625e-06,
|
119 |
+
"objective/entropy": -52.42451477050781,
|
120 |
+
"objective/kl": 2.8470888137817383,
|
121 |
+
"objective/non_score_reward": -0.14235445857048035,
|
122 |
+
"objective/rlhf_reward": 5.1500654220581055,
|
123 |
+
"objective/scores": 5.292419910430908,
|
124 |
+
"policy/approxkl_avg": 0.142442524433136,
|
125 |
+
"policy/clipfrac_avg": 0.330078125,
|
126 |
+
"policy/entropy_avg": 0.6877429485321045,
|
127 |
+
"step": 5,
|
128 |
+
"val/clipfrac_avg": 0.0,
|
129 |
+
"val/num_eos_tokens": 2296,
|
130 |
+
"val/ratio": 0.9998785853385925,
|
131 |
+
"val/ratio_var": 2.157924882340012e-06
|
132 |
+
},
|
133 |
+
{
|
134 |
+
"episode": 3584,
|
135 |
+
"epoch": 0.015352981494174092,
|
136 |
+
"eps": 7,
|
137 |
+
"loss/policy_avg": 0.21579426527023315,
|
138 |
+
"lr": 2.9296875e-06,
|
139 |
+
"objective/entropy": -52.407081604003906,
|
140 |
+
"objective/kl": 3.6154074668884277,
|
141 |
+
"objective/non_score_reward": -0.1807703673839569,
|
142 |
+
"objective/rlhf_reward": 5.39359712600708,
|
143 |
+
"objective/scores": 5.574367523193359,
|
144 |
+
"policy/approxkl_avg": 0.13034296035766602,
|
145 |
+
"policy/clipfrac_avg": 0.302734375,
|
146 |
+
"policy/entropy_avg": 0.6544360518455505,
|
147 |
+
"step": 6,
|
148 |
+
"val/clipfrac_avg": 0.0,
|
149 |
+
"val/num_eos_tokens": 2067,
|
150 |
+
"val/ratio": 1.000053882598877,
|
151 |
+
"val/ratio_var": 2.065472699541715e-06
|
152 |
+
},
|
153 |
+
{
|
154 |
+
"episode": 4096,
|
155 |
+
"epoch": 0.01754626456477039,
|
156 |
+
"eps": 7,
|
157 |
+
"loss/policy_avg": 0.10370946675539017,
|
158 |
+
"lr": 2.91796875e-06,
|
159 |
+
"objective/entropy": -52.04652786254883,
|
160 |
+
"objective/kl": 4.228328704833984,
|
161 |
+
"objective/non_score_reward": -0.21141643822193146,
|
162 |
+
"objective/rlhf_reward": 5.659776210784912,
|
163 |
+
"objective/scores": 5.87119197845459,
|
164 |
+
"policy/approxkl_avg": 0.14525723457336426,
|
165 |
+
"policy/clipfrac_avg": 0.3671875,
|
166 |
+
"policy/entropy_avg": 0.6791333556175232,
|
167 |
+
"step": 7,
|
168 |
+
"val/clipfrac_avg": 0.0,
|
169 |
+
"val/num_eos_tokens": 1779,
|
170 |
+
"val/ratio": 1.000075101852417,
|
171 |
+
"val/ratio_var": 1.6199858237087028e-06
|
172 |
+
},
|
173 |
+
{
|
174 |
+
"episode": 4608,
|
175 |
+
"epoch": 0.01973954763536669,
|
176 |
+
"eps": 7,
|
177 |
+
"loss/policy_avg": 0.1990879625082016,
|
178 |
+
"lr": 2.90625e-06,
|
179 |
+
"objective/entropy": -45.851478576660156,
|
180 |
+
"objective/kl": 6.0336480140686035,
|
181 |
+
"objective/non_score_reward": -0.30168241262435913,
|
182 |
+
"objective/rlhf_reward": 5.796751976013184,
|
183 |
+
"objective/scores": 6.098433971405029,
|
184 |
+
"policy/approxkl_avg": 0.17888996005058289,
|
185 |
+
"policy/clipfrac_avg": 0.333984375,
|
186 |
+
"policy/entropy_avg": 0.6368688344955444,
|
187 |
+
"step": 8,
|
188 |
+
"val/clipfrac_avg": 0.0,
|
189 |
+
"val/num_eos_tokens": 1556,
|
190 |
+
"val/ratio": 0.9997376203536987,
|
191 |
+
"val/ratio_var": 2.452116405038396e-06
|
192 |
+
},
|
193 |
+
{
|
194 |
+
"episode": 5120,
|
195 |
+
"epoch": 0.02193283070596299,
|
196 |
+
"eps": 7,
|
197 |
+
"loss/policy_avg": 0.17245890200138092,
|
198 |
+
"lr": 2.89453125e-06,
|
199 |
+
"objective/entropy": -46.33478546142578,
|
200 |
+
"objective/kl": 6.777096748352051,
|
201 |
+
"objective/non_score_reward": -0.3388548493385315,
|
202 |
+
"objective/rlhf_reward": 6.0051045417785645,
|
203 |
+
"objective/scores": 6.343959331512451,
|
204 |
+
"policy/approxkl_avg": 0.1448870599269867,
|
205 |
+
"policy/clipfrac_avg": 0.3203125,
|
206 |
+
"policy/entropy_avg": 0.6177160143852234,
|
207 |
+
"step": 9,
|
208 |
+
"val/clipfrac_avg": 0.0,
|
209 |
+
"val/num_eos_tokens": 1484,
|
210 |
+
"val/ratio": 0.999801754951477,
|
211 |
+
"val/ratio_var": 1.8629060605235281e-06
|
212 |
+
},
|
213 |
+
{
|
214 |
+
"episode": 5632,
|
215 |
+
"epoch": 0.024126113776559287,
|
216 |
+
"eps": 7,
|
217 |
+
"loss/policy_avg": 0.2342580407857895,
|
218 |
+
"lr": 2.8828125e-06,
|
219 |
+
"objective/entropy": -39.85594177246094,
|
220 |
+
"objective/kl": 8.770881652832031,
|
221 |
+
"objective/non_score_reward": -0.43854403495788574,
|
222 |
+
"objective/rlhf_reward": 6.183110237121582,
|
223 |
+
"objective/scores": 6.621654510498047,
|
224 |
+
"policy/approxkl_avg": 0.1644381582736969,
|
225 |
+
"policy/clipfrac_avg": 0.3515625,
|
226 |
+
"policy/entropy_avg": 0.569062352180481,
|
227 |
+
"step": 10,
|
228 |
+
"val/clipfrac_avg": 0.0,
|
229 |
+
"val/num_eos_tokens": 1314,
|
230 |
+
"val/ratio": 0.9999855756759644,
|
231 |
+
"val/ratio_var": 2.38121810980374e-06
|
232 |
+
},
|
233 |
+
{
|
234 |
+
"episode": 6144,
|
235 |
+
"epoch": 0.026319396847155587,
|
236 |
+
"eps": 7,
|
237 |
+
"loss/policy_avg": 0.2069457322359085,
|
238 |
+
"lr": 2.87109375e-06,
|
239 |
+
"objective/entropy": -41.22259521484375,
|
240 |
+
"objective/kl": 10.35439395904541,
|
241 |
+
"objective/non_score_reward": -0.5177196860313416,
|
242 |
+
"objective/rlhf_reward": 6.071005821228027,
|
243 |
+
"objective/scores": 6.588725566864014,
|
244 |
+
"policy/approxkl_avg": 0.16374149918556213,
|
245 |
+
"policy/clipfrac_avg": 0.35546875,
|
246 |
+
"policy/entropy_avg": 0.5040609836578369,
|
247 |
+
"step": 11,
|
248 |
+
"val/clipfrac_avg": 0.0,
|
249 |
+
"val/num_eos_tokens": 1575,
|
250 |
+
"val/ratio": 1.0002975463867188,
|
251 |
+
"val/ratio_var": 2.9677821657969616e-06
|
252 |
+
},
|
253 |
+
{
|
254 |
+
"episode": 6656,
|
255 |
+
"epoch": 0.028512679917751884,
|
256 |
+
"eps": 7,
|
257 |
+
"loss/policy_avg": 0.27875930070877075,
|
258 |
+
"lr": 2.859375e-06,
|
259 |
+
"objective/entropy": -44.60713195800781,
|
260 |
+
"objective/kl": 11.936761856079102,
|
261 |
+
"objective/non_score_reward": -0.596838116645813,
|
262 |
+
"objective/rlhf_reward": 6.585817337036133,
|
263 |
+
"objective/scores": 7.182655334472656,
|
264 |
+
"policy/approxkl_avg": 0.15854549407958984,
|
265 |
+
"policy/clipfrac_avg": 0.32421875,
|
266 |
+
"policy/entropy_avg": 0.48229658603668213,
|
267 |
+
"step": 12,
|
268 |
+
"val/clipfrac_avg": 0.0,
|
269 |
+
"val/num_eos_tokens": 1957,
|
270 |
+
"val/ratio": 0.999975323677063,
|
271 |
+
"val/ratio_var": 2.387672111581196e-06
|
272 |
+
},
|
273 |
+
{
|
274 |
+
"episode": 7168,
|
275 |
+
"epoch": 0.030705962988348184,
|
276 |
+
"eps": 7,
|
277 |
+
"loss/policy_avg": 0.19576790928840637,
|
278 |
+
"lr": 2.84765625e-06,
|
279 |
+
"objective/entropy": -47.06153106689453,
|
280 |
+
"objective/kl": 14.905025482177734,
|
281 |
+
"objective/non_score_reward": -0.7452512979507446,
|
282 |
+
"objective/rlhf_reward": 7.080592155456543,
|
283 |
+
"objective/scores": 7.82584285736084,
|
284 |
+
"policy/approxkl_avg": 0.15632450580596924,
|
285 |
+
"policy/clipfrac_avg": 0.322265625,
|
286 |
+
"policy/entropy_avg": 0.45476478338241577,
|
287 |
+
"step": 13,
|
288 |
+
"val/clipfrac_avg": 0.0,
|
289 |
+
"val/num_eos_tokens": 2210,
|
290 |
+
"val/ratio": 0.9998209476470947,
|
291 |
+
"val/ratio_var": 2.1720893528254237e-06
|
292 |
+
},
|
293 |
+
{
|
294 |
+
"episode": 7680,
|
295 |
+
"epoch": 0.03289924605894448,
|
296 |
+
"eps": 7,
|
297 |
+
"loss/policy_avg": 0.17670106887817383,
|
298 |
+
"lr": 2.8359375e-06,
|
299 |
+
"objective/entropy": -44.32414627075195,
|
300 |
+
"objective/kl": 17.04126739501953,
|
301 |
+
"objective/non_score_reward": -0.8520634770393372,
|
302 |
+
"objective/rlhf_reward": 6.854432582855225,
|
303 |
+
"objective/scores": 7.706496238708496,
|
304 |
+
"policy/approxkl_avg": 0.15034478902816772,
|
305 |
+
"policy/clipfrac_avg": 0.3359375,
|
306 |
+
"policy/entropy_avg": 0.41826024651527405,
|
307 |
+
"step": 14,
|
308 |
+
"val/clipfrac_avg": 0.0,
|
309 |
+
"val/num_eos_tokens": 2183,
|
310 |
+
"val/ratio": 1.0000519752502441,
|
311 |
+
"val/ratio_var": 2.0224842955940403e-06
|
312 |
+
},
|
313 |
+
{
|
314 |
+
"episode": 8192,
|
315 |
+
"epoch": 0.03509252912954078,
|
316 |
+
"eps": 7,
|
317 |
+
"loss/policy_avg": 0.44327589869499207,
|
318 |
+
"lr": 2.82421875e-06,
|
319 |
+
"objective/entropy": -48.27879333496094,
|
320 |
+
"objective/kl": 18.870018005371094,
|
321 |
+
"objective/non_score_reward": -0.9435009956359863,
|
322 |
+
"objective/rlhf_reward": 6.901827335357666,
|
323 |
+
"objective/scores": 7.845328330993652,
|
324 |
+
"policy/approxkl_avg": 0.14197328686714172,
|
325 |
+
"policy/clipfrac_avg": 0.279296875,
|
326 |
+
"policy/entropy_avg": 0.39645904302597046,
|
327 |
+
"step": 15,
|
328 |
+
"val/clipfrac_avg": 0.0,
|
329 |
+
"val/num_eos_tokens": 2105,
|
330 |
+
"val/ratio": 1.000123143196106,
|
331 |
+
"val/ratio_var": 2.0512616174528375e-06
|
332 |
+
},
|
333 |
+
{
|
334 |
+
"episode": 8704,
|
335 |
+
"epoch": 0.03728581220013708,
|
336 |
+
"eps": 7,
|
337 |
+
"loss/policy_avg": 0.14077123999595642,
|
338 |
+
"lr": 2.8125e-06,
|
339 |
+
"objective/entropy": -51.67315673828125,
|
340 |
+
"objective/kl": 19.975967407226562,
|
341 |
+
"objective/non_score_reward": -0.9987983703613281,
|
342 |
+
"objective/rlhf_reward": 6.921031951904297,
|
343 |
+
"objective/scores": 7.919830322265625,
|
344 |
+
"policy/approxkl_avg": 0.13363662362098694,
|
345 |
+
"policy/clipfrac_avg": 0.30078125,
|
346 |
+
"policy/entropy_avg": 0.35817191004753113,
|
347 |
+
"step": 16,
|
348 |
+
"val/clipfrac_avg": 0.0,
|
349 |
+
"val/num_eos_tokens": 1926,
|
350 |
+
"val/ratio": 1.0001368522644043,
|
351 |
+
"val/ratio_var": 2.0405464056238998e-06
|
352 |
+
},
|
353 |
+
{
|
354 |
+
"episode": 9216,
|
355 |
+
"epoch": 0.03947909527073338,
|
356 |
+
"eps": 7,
|
357 |
+
"loss/policy_avg": 0.11229261755943298,
|
358 |
+
"lr": 2.80078125e-06,
|
359 |
+
"objective/entropy": -56.30708312988281,
|
360 |
+
"objective/kl": 19.739307403564453,
|
361 |
+
"objective/non_score_reward": -0.9869654774665833,
|
362 |
+
"objective/rlhf_reward": 6.998082160949707,
|
363 |
+
"objective/scores": 7.985047340393066,
|
364 |
+
"policy/approxkl_avg": 0.11359818279743195,
|
365 |
+
"policy/clipfrac_avg": 0.294921875,
|
366 |
+
"policy/entropy_avg": 0.3619132936000824,
|
367 |
+
"step": 17,
|
368 |
+
"val/clipfrac_avg": 0.0,
|
369 |
+
"val/num_eos_tokens": 1909,
|
370 |
+
"val/ratio": 0.99982750415802,
|
371 |
+
"val/ratio_var": 1.578110754962836e-06
|
372 |
+
},
|
373 |
+
{
|
374 |
+
"episode": 9728,
|
375 |
+
"epoch": 0.041672378341329676,
|
376 |
+
"eps": 7,
|
377 |
+
"loss/policy_avg": 0.2098175585269928,
|
378 |
+
"lr": 2.7890625e-06,
|
379 |
+
"objective/entropy": -57.04094696044922,
|
380 |
+
"objective/kl": 21.025684356689453,
|
381 |
+
"objective/non_score_reward": -1.0512843132019043,
|
382 |
+
"objective/rlhf_reward": 6.816586494445801,
|
383 |
+
"objective/scores": 7.867871284484863,
|
384 |
+
"policy/approxkl_avg": 0.12195070087909698,
|
385 |
+
"policy/clipfrac_avg": 0.25390625,
|
386 |
+
"policy/entropy_avg": 0.34247398376464844,
|
387 |
+
"step": 18,
|
388 |
+
"val/clipfrac_avg": 0.0,
|
389 |
+
"val/num_eos_tokens": 1961,
|
390 |
+
"val/ratio": 1.0000369548797607,
|
391 |
+
"val/ratio_var": 2.1834478047821904e-06
|
392 |
+
},
|
393 |
+
{
|
394 |
+
"episode": 10240,
|
395 |
+
"epoch": 0.04386566141192598,
|
396 |
+
"eps": 7,
|
397 |
+
"loss/policy_avg": 0.1683914214372635,
|
398 |
+
"lr": 2.77734375e-06,
|
399 |
+
"objective/entropy": -58.71490478515625,
|
400 |
+
"objective/kl": 20.541297912597656,
|
401 |
+
"objective/non_score_reward": -1.0270648002624512,
|
402 |
+
"objective/rlhf_reward": 6.755092144012451,
|
403 |
+
"objective/scores": 7.782156944274902,
|
404 |
+
"policy/approxkl_avg": 0.17478591203689575,
|
405 |
+
"policy/clipfrac_avg": 0.294921875,
|
406 |
+
"policy/entropy_avg": 0.3125288188457489,
|
407 |
+
"step": 19,
|
408 |
+
"val/clipfrac_avg": 0.0,
|
409 |
+
"val/num_eos_tokens": 1872,
|
410 |
+
"val/ratio": 1.0001912117004395,
|
411 |
+
"val/ratio_var": 2.5365768578922143e-06
|
412 |
+
},
|
413 |
+
{
|
414 |
+
"episode": 10752,
|
415 |
+
"epoch": 0.04605894448252228,
|
416 |
+
"eps": 7,
|
417 |
+
"loss/policy_avg": 0.2009236365556717,
|
418 |
+
"lr": 2.765625e-06,
|
419 |
+
"objective/entropy": -58.684837341308594,
|
420 |
+
"objective/kl": 21.819164276123047,
|
421 |
+
"objective/non_score_reward": -1.0909582376480103,
|
422 |
+
"objective/rlhf_reward": 6.992592811584473,
|
423 |
+
"objective/scores": 8.083551406860352,
|
424 |
+
"policy/approxkl_avg": 0.13064998388290405,
|
425 |
+
"policy/clipfrac_avg": 0.267578125,
|
426 |
+
"policy/entropy_avg": 0.34474819898605347,
|
427 |
+
"step": 20,
|
428 |
+
"val/clipfrac_avg": 0.0,
|
429 |
+
"val/num_eos_tokens": 2117,
|
430 |
+
"val/ratio": 0.9997358322143555,
|
431 |
+
"val/ratio_var": 1.91668755178398e-06
|
432 |
+
},
|
433 |
+
{
|
434 |
+
"episode": 11264,
|
435 |
+
"epoch": 0.048252227553118573,
|
436 |
+
"eps": 7,
|
437 |
+
"loss/policy_avg": 0.15789391100406647,
|
438 |
+
"lr": 2.75390625e-06,
|
439 |
+
"objective/entropy": -55.13008117675781,
|
440 |
+
"objective/kl": 22.038318634033203,
|
441 |
+
"objective/non_score_reward": -1.1019160747528076,
|
442 |
+
"objective/rlhf_reward": 7.230824947357178,
|
443 |
+
"objective/scores": 8.332740783691406,
|
444 |
+
"policy/approxkl_avg": 0.13166062533855438,
|
445 |
+
"policy/clipfrac_avg": 0.279296875,
|
446 |
+
"policy/entropy_avg": 0.3379790484905243,
|
447 |
+
"step": 21,
|
448 |
+
"val/clipfrac_avg": 0.0,
|
449 |
+
"val/num_eos_tokens": 2328,
|
450 |
+
"val/ratio": 0.9997003674507141,
|
451 |
+
"val/ratio_var": 1.897903189274075e-06
|
452 |
+
},
|
453 |
+
{
|
454 |
+
"episode": 11776,
|
455 |
+
"epoch": 0.05044551062371487,
|
456 |
+
"eps": 7,
|
457 |
+
"loss/policy_avg": 0.2752581536769867,
|
458 |
+
"lr": 2.7421875e-06,
|
459 |
+
"objective/entropy": -54.6153450012207,
|
460 |
+
"objective/kl": 22.95562744140625,
|
461 |
+
"objective/non_score_reward": -1.1477813720703125,
|
462 |
+
"objective/rlhf_reward": 7.375998497009277,
|
463 |
+
"objective/scores": 8.52377986907959,
|
464 |
+
"policy/approxkl_avg": 0.1369776725769043,
|
465 |
+
"policy/clipfrac_avg": 0.30859375,
|
466 |
+
"policy/entropy_avg": 0.3782145380973816,
|
467 |
+
"step": 22,
|
468 |
+
"val/clipfrac_avg": 0.0,
|
469 |
+
"val/num_eos_tokens": 2505,
|
470 |
+
"val/ratio": 0.9999332427978516,
|
471 |
+
"val/ratio_var": 2.100945494021289e-06
|
472 |
+
},
|
473 |
+
{
|
474 |
+
"episode": 12288,
|
475 |
+
"epoch": 0.052638793694311174,
|
476 |
+
"eps": 7,
|
477 |
+
"loss/policy_avg": 0.17188745737075806,
|
478 |
+
"lr": 2.73046875e-06,
|
479 |
+
"objective/entropy": -52.373382568359375,
|
480 |
+
"objective/kl": 24.686687469482422,
|
481 |
+
"objective/non_score_reward": -1.2343344688415527,
|
482 |
+
"objective/rlhf_reward": 7.556896686553955,
|
483 |
+
"objective/scores": 8.791231155395508,
|
484 |
+
"policy/approxkl_avg": 0.1562095284461975,
|
485 |
+
"policy/clipfrac_avg": 0.3125,
|
486 |
+
"policy/entropy_avg": 0.3631889820098877,
|
487 |
+
"step": 23,
|
488 |
+
"val/clipfrac_avg": 0.0,
|
489 |
+
"val/num_eos_tokens": 2827,
|
490 |
+
"val/ratio": 0.9997256994247437,
|
491 |
+
"val/ratio_var": 1.3294010159370373e-06
|
492 |
+
},
|
493 |
+
{
|
494 |
+
"episode": 12800,
|
495 |
+
"epoch": 0.05483207676490747,
|
496 |
+
"eps": 7,
|
497 |
+
"loss/policy_avg": 0.14826999604701996,
|
498 |
+
"lr": 2.71875e-06,
|
499 |
+
"objective/entropy": -48.87049102783203,
|
500 |
+
"objective/kl": 26.420700073242188,
|
501 |
+
"objective/non_score_reward": -1.3210350275039673,
|
502 |
+
"objective/rlhf_reward": 7.30073356628418,
|
503 |
+
"objective/scores": 8.6217679977417,
|
504 |
+
"policy/approxkl_avg": 0.150124192237854,
|
505 |
+
"policy/clipfrac_avg": 0.30859375,
|
506 |
+
"policy/entropy_avg": 0.36433207988739014,
|
507 |
+
"step": 24,
|
508 |
+
"val/clipfrac_avg": 0.0,
|
509 |
+
"val/num_eos_tokens": 2618,
|
510 |
+
"val/ratio": 1.0000042915344238,
|
511 |
+
"val/ratio_var": 1.7361713844366022e-06
|
512 |
+
},
|
513 |
+
{
|
514 |
+
"episode": 13312,
|
515 |
+
"epoch": 0.05702535983550377,
|
516 |
+
"eps": 7,
|
517 |
+
"loss/policy_avg": 0.26754873991012573,
|
518 |
+
"lr": 2.70703125e-06,
|
519 |
+
"objective/entropy": -44.881736755371094,
|
520 |
+
"objective/kl": 28.44316291809082,
|
521 |
+
"objective/non_score_reward": -1.4221582412719727,
|
522 |
+
"objective/rlhf_reward": 7.6062798500061035,
|
523 |
+
"objective/scores": 9.028438568115234,
|
524 |
+
"policy/approxkl_avg": 0.1850195825099945,
|
525 |
+
"policy/clipfrac_avg": 0.3046875,
|
526 |
+
"policy/entropy_avg": 0.3754498362541199,
|
527 |
+
"step": 25,
|
528 |
+
"val/clipfrac_avg": 0.0,
|
529 |
+
"val/num_eos_tokens": 2679,
|
530 |
+
"val/ratio": 1.0001829862594604,
|
531 |
+
"val/ratio_var": 3.7937434171908535e-06
|
532 |
+
},
|
533 |
+
{
|
534 |
+
"episode": 13824,
|
535 |
+
"epoch": 0.05921864290610007,
|
536 |
+
"eps": 7,
|
537 |
+
"loss/policy_avg": 0.4031262993812561,
|
538 |
+
"lr": 2.6953125e-06,
|
539 |
+
"objective/entropy": -46.877281188964844,
|
540 |
+
"objective/kl": 28.513225555419922,
|
541 |
+
"objective/non_score_reward": -1.425661325454712,
|
542 |
+
"objective/rlhf_reward": 7.531891822814941,
|
543 |
+
"objective/scores": 8.957552909851074,
|
544 |
+
"policy/approxkl_avg": 0.17947888374328613,
|
545 |
+
"policy/clipfrac_avg": 0.310546875,
|
546 |
+
"policy/entropy_avg": 0.37926802039146423,
|
547 |
+
"step": 26,
|
548 |
+
"val/clipfrac_avg": 0.0,
|
549 |
+
"val/num_eos_tokens": 2800,
|
550 |
+
"val/ratio": 1.0000090599060059,
|
551 |
+
"val/ratio_var": 2.2898825591255445e-06
|
552 |
+
},
|
553 |
+
{
|
554 |
+
"episode": 14336,
|
555 |
+
"epoch": 0.06141192597669637,
|
556 |
+
"eps": 7,
|
557 |
+
"loss/policy_avg": 0.2814100682735443,
|
558 |
+
"lr": 2.68359375e-06,
|
559 |
+
"objective/entropy": -42.25868225097656,
|
560 |
+
"objective/kl": 30.40261459350586,
|
561 |
+
"objective/non_score_reward": -1.5201308727264404,
|
562 |
+
"objective/rlhf_reward": 7.6179656982421875,
|
563 |
+
"objective/scores": 9.13809585571289,
|
564 |
+
"policy/approxkl_avg": 0.20318853855133057,
|
565 |
+
"policy/clipfrac_avg": 0.306640625,
|
566 |
+
"policy/entropy_avg": 0.3404484987258911,
|
567 |
+
"step": 27,
|
568 |
+
"val/clipfrac_avg": 0.0,
|
569 |
+
"val/num_eos_tokens": 2518,
|
570 |
+
"val/ratio": 0.9999587535858154,
|
571 |
+
"val/ratio_var": 3.4963336474902462e-06
|
572 |
+
},
|
573 |
+
{
|
574 |
+
"episode": 14848,
|
575 |
+
"epoch": 0.06360520904729267,
|
576 |
+
"eps": 7,
|
577 |
+
"loss/policy_avg": 0.5485799312591553,
|
578 |
+
"lr": 2.671875e-06,
|
579 |
+
"objective/entropy": -44.9633903503418,
|
580 |
+
"objective/kl": 29.625776290893555,
|
581 |
+
"objective/non_score_reward": -1.4812889099121094,
|
582 |
+
"objective/rlhf_reward": 7.636438369750977,
|
583 |
+
"objective/scores": 9.117727279663086,
|
584 |
+
"policy/approxkl_avg": 0.19322876632213593,
|
585 |
+
"policy/clipfrac_avg": 0.314453125,
|
586 |
+
"policy/entropy_avg": 0.35978442430496216,
|
587 |
+
"step": 28,
|
588 |
+
"val/clipfrac_avg": 0.0,
|
589 |
+
"val/num_eos_tokens": 2586,
|
590 |
+
"val/ratio": 1.0003840923309326,
|
591 |
+
"val/ratio_var": 3.6096953408559784e-06
|
592 |
+
},
|
593 |
+
{
|
594 |
+
"episode": 15360,
|
595 |
+
"epoch": 0.06579849211788896,
|
596 |
+
"eps": 7,
|
597 |
+
"loss/policy_avg": 0.25037768483161926,
|
598 |
+
"lr": 2.66015625e-06,
|
599 |
+
"objective/entropy": -45.554847717285156,
|
600 |
+
"objective/kl": 27.680625915527344,
|
601 |
+
"objective/non_score_reward": -1.3840312957763672,
|
602 |
+
"objective/rlhf_reward": 7.663791656494141,
|
603 |
+
"objective/scores": 9.047822952270508,
|
604 |
+
"policy/approxkl_avg": 0.2359601855278015,
|
605 |
+
"policy/clipfrac_avg": 0.337890625,
|
606 |
+
"policy/entropy_avg": 0.35273873805999756,
|
607 |
+
"step": 29,
|
608 |
+
"val/clipfrac_avg": 0.0,
|
609 |
+
"val/num_eos_tokens": 2580,
|
610 |
+
"val/ratio": 1.0000591278076172,
|
611 |
+
"val/ratio_var": 3.4416327707731398e-06
|
612 |
+
},
|
613 |
+
{
|
614 |
+
"episode": 15872,
|
615 |
+
"epoch": 0.06799177518848526,
|
616 |
+
"eps": 7,
|
617 |
+
"loss/policy_avg": 0.180610790848732,
|
618 |
+
"lr": 2.6484375e-06,
|
619 |
+
"objective/entropy": -51.34313201904297,
|
620 |
+
"objective/kl": 25.798786163330078,
|
621 |
+
"objective/non_score_reward": -1.289939284324646,
|
622 |
+
"objective/rlhf_reward": 8.065506935119629,
|
623 |
+
"objective/scores": 9.355446815490723,
|
624 |
+
"policy/approxkl_avg": 0.14466476440429688,
|
625 |
+
"policy/clipfrac_avg": 0.291015625,
|
626 |
+
"policy/entropy_avg": 0.3412173092365265,
|
627 |
+
"step": 30,
|
628 |
+
"val/clipfrac_avg": 0.0,
|
629 |
+
"val/num_eos_tokens": 2674,
|
630 |
+
"val/ratio": 1.0001661777496338,
|
631 |
+
"val/ratio_var": 2.756241656243219e-06
|
632 |
+
},
|
633 |
+
{
|
634 |
+
"episode": 16384,
|
635 |
+
"epoch": 0.07018505825908156,
|
636 |
+
"eps": 7,
|
637 |
+
"loss/policy_avg": 0.17584098875522614,
|
638 |
+
"lr": 2.63671875e-06,
|
639 |
+
"objective/entropy": -54.59495544433594,
|
640 |
+
"objective/kl": 24.21860122680664,
|
641 |
+
"objective/non_score_reward": -1.2109301090240479,
|
642 |
+
"objective/rlhf_reward": 7.664724826812744,
|
643 |
+
"objective/scores": 8.875654220581055,
|
644 |
+
"policy/approxkl_avg": 0.14981456100940704,
|
645 |
+
"policy/clipfrac_avg": 0.34375,
|
646 |
+
"policy/entropy_avg": 0.35965415835380554,
|
647 |
+
"step": 31,
|
648 |
+
"val/clipfrac_avg": 0.0,
|
649 |
+
"val/num_eos_tokens": 2703,
|
650 |
+
"val/ratio": 1.0002694129943848,
|
651 |
+
"val/ratio_var": 2.6581130896374816e-06
|
652 |
+
},
|
653 |
+
{
|
654 |
+
"episode": 16896,
|
655 |
+
"epoch": 0.07237834132967787,
|
656 |
+
"eps": 7,
|
657 |
+
"loss/policy_avg": 0.13980838656425476,
|
658 |
+
"lr": 2.6250000000000003e-06,
|
659 |
+
"objective/entropy": -58.448150634765625,
|
660 |
+
"objective/kl": 21.735061645507812,
|
661 |
+
"objective/non_score_reward": -1.0867531299591064,
|
662 |
+
"objective/rlhf_reward": 7.95279598236084,
|
663 |
+
"objective/scores": 9.039548873901367,
|
664 |
+
"policy/approxkl_avg": 0.1316492259502411,
|
665 |
+
"policy/clipfrac_avg": 0.318359375,
|
666 |
+
"policy/entropy_avg": 0.3467163145542145,
|
667 |
+
"step": 32,
|
668 |
+
"val/clipfrac_avg": 0.0,
|
669 |
+
"val/num_eos_tokens": 2919,
|
670 |
+
"val/ratio": 0.9999335408210754,
|
671 |
+
"val/ratio_var": 1.5475716281798668e-06
|
672 |
+
},
|
673 |
+
{
|
674 |
+
"episode": 17408,
|
675 |
+
"epoch": 0.07457162440027416,
|
676 |
+
"eps": 7,
|
677 |
+
"loss/policy_avg": 0.16001015901565552,
|
678 |
+
"lr": 2.61328125e-06,
|
679 |
+
"objective/entropy": -58.598934173583984,
|
680 |
+
"objective/kl": 20.68917465209961,
|
681 |
+
"objective/non_score_reward": -1.0344586372375488,
|
682 |
+
"objective/rlhf_reward": 7.836881637573242,
|
683 |
+
"objective/scores": 8.87134075164795,
|
684 |
+
"policy/approxkl_avg": 0.12066276371479034,
|
685 |
+
"policy/clipfrac_avg": 0.259765625,
|
686 |
+
"policy/entropy_avg": 0.38158828020095825,
|
687 |
+
"step": 33,
|
688 |
+
"val/clipfrac_avg": 0.0,
|
689 |
+
"val/num_eos_tokens": 2861,
|
690 |
+
"val/ratio": 1.0005499124526978,
|
691 |
+
"val/ratio_var": 2.171855157939717e-06
|
692 |
+
},
|
693 |
+
{
|
694 |
+
"episode": 17920,
|
695 |
+
"epoch": 0.07676490747087046,
|
696 |
+
"eps": 7,
|
697 |
+
"loss/policy_avg": 0.15362152457237244,
|
698 |
+
"lr": 2.6015625e-06,
|
699 |
+
"objective/entropy": -59.510032653808594,
|
700 |
+
"objective/kl": 19.810598373413086,
|
701 |
+
"objective/non_score_reward": -0.9905299544334412,
|
702 |
+
"objective/rlhf_reward": 7.790919780731201,
|
703 |
+
"objective/scores": 8.781450271606445,
|
704 |
+
"policy/approxkl_avg": 0.13420400023460388,
|
705 |
+
"policy/clipfrac_avg": 0.310546875,
|
706 |
+
"policy/entropy_avg": 0.37936142086982727,
|
707 |
+
"step": 34,
|
708 |
+
"val/clipfrac_avg": 0.0,
|
709 |
+
"val/num_eos_tokens": 2984,
|
710 |
+
"val/ratio": 1.0000555515289307,
|
711 |
+
"val/ratio_var": 2.221043132522027e-06
|
712 |
+
},
|
713 |
+
{
|
714 |
+
"episode": 18432,
|
715 |
+
"epoch": 0.07895819054146676,
|
716 |
+
"eps": 7,
|
717 |
+
"loss/policy_avg": 0.14148639142513275,
|
718 |
+
"lr": 2.5898437500000003e-06,
|
719 |
+
"objective/entropy": -62.267539978027344,
|
720 |
+
"objective/kl": 19.003326416015625,
|
721 |
+
"objective/non_score_reward": -0.9501664638519287,
|
722 |
+
"objective/rlhf_reward": 7.759812355041504,
|
723 |
+
"objective/scores": 8.709979057312012,
|
724 |
+
"policy/approxkl_avg": 0.13157205283641815,
|
725 |
+
"policy/clipfrac_avg": 0.322265625,
|
726 |
+
"policy/entropy_avg": 0.3889726400375366,
|
727 |
+
"step": 35,
|
728 |
+
"val/clipfrac_avg": 0.0,
|
729 |
+
"val/num_eos_tokens": 3040,
|
730 |
+
"val/ratio": 1.0000544786453247,
|
731 |
+
"val/ratio_var": 1.8092218851961661e-06
|
732 |
+
},
|
733 |
+
{
|
734 |
+
"episode": 18944,
|
735 |
+
"epoch": 0.08115147361206305,
|
736 |
+
"eps": 7,
|
737 |
+
"loss/policy_avg": 0.13877639174461365,
|
738 |
+
"lr": 2.578125e-06,
|
739 |
+
"objective/entropy": -63.75535202026367,
|
740 |
+
"objective/kl": 17.362323760986328,
|
741 |
+
"objective/non_score_reward": -0.8681161403656006,
|
742 |
+
"objective/rlhf_reward": 7.706595420837402,
|
743 |
+
"objective/scores": 8.574711799621582,
|
744 |
+
"policy/approxkl_avg": 0.11438442766666412,
|
745 |
+
"policy/clipfrac_avg": 0.26171875,
|
746 |
+
"policy/entropy_avg": 0.3780348300933838,
|
747 |
+
"step": 36,
|
748 |
+
"val/clipfrac_avg": 0.0,
|
749 |
+
"val/num_eos_tokens": 3103,
|
750 |
+
"val/ratio": 0.9998428821563721,
|
751 |
+
"val/ratio_var": 1.6111931699924753e-06
|
752 |
+
},
|
753 |
+
{
|
754 |
+
"episode": 19456,
|
755 |
+
"epoch": 0.08334475668265935,
|
756 |
+
"eps": 7,
|
757 |
+
"loss/policy_avg": 0.1436106264591217,
|
758 |
+
"lr": 2.56640625e-06,
|
759 |
+
"objective/entropy": -63.54912567138672,
|
760 |
+
"objective/kl": 17.546058654785156,
|
761 |
+
"objective/non_score_reward": -0.877302885055542,
|
762 |
+
"objective/rlhf_reward": 7.767643928527832,
|
763 |
+
"objective/scores": 8.644947052001953,
|
764 |
+
"policy/approxkl_avg": 0.1229478269815445,
|
765 |
+
"policy/clipfrac_avg": 0.30078125,
|
766 |
+
"policy/entropy_avg": 0.3848574161529541,
|
767 |
+
"step": 37,
|
768 |
+
"val/clipfrac_avg": 0.0,
|
769 |
+
"val/num_eos_tokens": 3251,
|
770 |
+
"val/ratio": 0.9997145533561707,
|
771 |
+
"val/ratio_var": 1.7358969444103423e-06
|
772 |
+
},
|
773 |
+
{
|
774 |
+
"episode": 19968,
|
775 |
+
"epoch": 0.08553803975325565,
|
776 |
+
"eps": 7,
|
777 |
+
"loss/policy_avg": 0.0948140099644661,
|
778 |
+
"lr": 2.5546875000000003e-06,
|
779 |
+
"objective/entropy": -65.45063018798828,
|
780 |
+
"objective/kl": 17.355745315551758,
|
781 |
+
"objective/non_score_reward": -0.8677873611450195,
|
782 |
+
"objective/rlhf_reward": 7.8387346267700195,
|
783 |
+
"objective/scores": 8.706522941589355,
|
784 |
+
"policy/approxkl_avg": 0.0997094064950943,
|
785 |
+
"policy/clipfrac_avg": 0.291015625,
|
786 |
+
"policy/entropy_avg": 0.39116984605789185,
|
787 |
+
"step": 38,
|
788 |
+
"val/clipfrac_avg": 0.0,
|
789 |
+
"val/num_eos_tokens": 3321,
|
790 |
+
"val/ratio": 1.0000455379486084,
|
791 |
+
"val/ratio_var": 1.6502441440024995e-06
|
792 |
+
},
|
793 |
+
{
|
794 |
+
"episode": 20480,
|
795 |
+
"epoch": 0.08773132282385196,
|
796 |
+
"eps": 7,
|
797 |
+
"loss/policy_avg": 0.143824964761734,
|
798 |
+
"lr": 2.54296875e-06,
|
799 |
+
"objective/entropy": -63.26994323730469,
|
800 |
+
"objective/kl": 17.91488265991211,
|
801 |
+
"objective/non_score_reward": -0.8957440853118896,
|
802 |
+
"objective/rlhf_reward": 7.531732559204102,
|
803 |
+
"objective/scores": 8.42747688293457,
|
804 |
+
"policy/approxkl_avg": 0.12001372873783112,
|
805 |
+
"policy/clipfrac_avg": 0.296875,
|
806 |
+
"policy/entropy_avg": 0.3968222141265869,
|
807 |
+
"step": 39,
|
808 |
+
"val/clipfrac_avg": 0.0,
|
809 |
+
"val/num_eos_tokens": 3432,
|
810 |
+
"val/ratio": 0.9998824596405029,
|
811 |
+
"val/ratio_var": 1.8353850919083925e-06
|
812 |
+
},
|
813 |
+
{
|
814 |
+
"episode": 20992,
|
815 |
+
"epoch": 0.08992460589444826,
|
816 |
+
"eps": 7,
|
817 |
+
"loss/policy_avg": 0.08669011294841766,
|
818 |
+
"lr": 2.53125e-06,
|
819 |
+
"objective/entropy": -66.6177978515625,
|
820 |
+
"objective/kl": 16.663658142089844,
|
821 |
+
"objective/non_score_reward": -0.8331829905509949,
|
822 |
+
"objective/rlhf_reward": 7.59118127822876,
|
823 |
+
"objective/scores": 8.42436408996582,
|
824 |
+
"policy/approxkl_avg": 0.11151742935180664,
|
825 |
+
"policy/clipfrac_avg": 0.314453125,
|
826 |
+
"policy/entropy_avg": 0.39587923884391785,
|
827 |
+
"step": 40,
|
828 |
+
"val/clipfrac_avg": 0.0,
|
829 |
+
"val/num_eos_tokens": 3491,
|
830 |
+
"val/ratio": 0.9999145269393921,
|
831 |
+
"val/ratio_var": 1.2908723192595062e-06
|
832 |
+
},
|
833 |
+
{
|
834 |
+
"episode": 21504,
|
835 |
+
"epoch": 0.09211788896504455,
|
836 |
+
"eps": 7,
|
837 |
+
"loss/policy_avg": 0.09066055715084076,
|
838 |
+
"lr": 2.5195312500000003e-06,
|
839 |
+
"objective/entropy": -66.88969421386719,
|
840 |
+
"objective/kl": 16.79983901977539,
|
841 |
+
"objective/non_score_reward": -0.8399919271469116,
|
842 |
+
"objective/rlhf_reward": 7.6204423904418945,
|
843 |
+
"objective/scores": 8.460433959960938,
|
844 |
+
"policy/approxkl_avg": 0.10316135734319687,
|
845 |
+
"policy/clipfrac_avg": 0.2890625,
|
846 |
+
"policy/entropy_avg": 0.39546722173690796,
|
847 |
+
"step": 41,
|
848 |
+
"val/clipfrac_avg": 0.0,
|
849 |
+
"val/num_eos_tokens": 3365,
|
850 |
+
"val/ratio": 1.0000004768371582,
|
851 |
+
"val/ratio_var": 1.8537729147283244e-06
|
852 |
+
},
|
853 |
+
{
|
854 |
+
"episode": 22016,
|
855 |
+
"epoch": 0.09431117203564085,
|
856 |
+
"eps": 7,
|
857 |
+
"loss/policy_avg": 0.12256267666816711,
|
858 |
+
"lr": 2.5078125e-06,
|
859 |
+
"objective/entropy": -63.886512756347656,
|
860 |
+
"objective/kl": 17.69078826904297,
|
861 |
+
"objective/non_score_reward": -0.8845393657684326,
|
862 |
+
"objective/rlhf_reward": 7.7145915031433105,
|
863 |
+
"objective/scores": 8.599130630493164,
|
864 |
+
"policy/approxkl_avg": 0.1147078201174736,
|
865 |
+
"policy/clipfrac_avg": 0.287109375,
|
866 |
+
"policy/entropy_avg": 0.4132445454597473,
|
867 |
+
"step": 42,
|
868 |
+
"val/clipfrac_avg": 0.0,
|
869 |
+
"val/num_eos_tokens": 3497,
|
870 |
+
"val/ratio": 1.000230073928833,
|
871 |
+
"val/ratio_var": 1.4803732710788609e-06
|
872 |
+
},
|
873 |
+
{
|
874 |
+
"episode": 22528,
|
875 |
+
"epoch": 0.09650445510623715,
|
876 |
+
"eps": 7,
|
877 |
+
"loss/policy_avg": 0.10973351448774338,
|
878 |
+
"lr": 2.49609375e-06,
|
879 |
+
"objective/entropy": -63.165733337402344,
|
880 |
+
"objective/kl": 17.609962463378906,
|
881 |
+
"objective/non_score_reward": -0.8804981708526611,
|
882 |
+
"objective/rlhf_reward": 7.543717384338379,
|
883 |
+
"objective/scores": 8.424215316772461,
|
884 |
+
"policy/approxkl_avg": 0.09955161064863205,
|
885 |
+
"policy/clipfrac_avg": 0.310546875,
|
886 |
+
"policy/entropy_avg": 0.40705782175064087,
|
887 |
+
"step": 43,
|
888 |
+
"val/clipfrac_avg": 0.0,
|
889 |
+
"val/num_eos_tokens": 3581,
|
890 |
+
"val/ratio": 0.999884307384491,
|
891 |
+
"val/ratio_var": 1.5909022295090836e-06
|
892 |
+
},
|
893 |
+
{
|
894 |
+
"episode": 23040,
|
895 |
+
"epoch": 0.09869773817683344,
|
896 |
+
"eps": 7,
|
897 |
+
"loss/policy_avg": 0.11654724180698395,
|
898 |
+
"lr": 2.4843750000000002e-06,
|
899 |
+
"objective/entropy": -64.02261352539062,
|
900 |
+
"objective/kl": 17.51599884033203,
|
901 |
+
"objective/non_score_reward": -0.8758000135421753,
|
902 |
+
"objective/rlhf_reward": 7.722189903259277,
|
903 |
+
"objective/scores": 8.597990036010742,
|
904 |
+
"policy/approxkl_avg": 0.10517577826976776,
|
905 |
+
"policy/clipfrac_avg": 0.294921875,
|
906 |
+
"policy/entropy_avg": 0.39612969756126404,
|
907 |
+
"step": 44,
|
908 |
+
"val/clipfrac_avg": 0.0,
|
909 |
+
"val/num_eos_tokens": 3514,
|
910 |
+
"val/ratio": 0.9999499320983887,
|
911 |
+
"val/ratio_var": 1.633706119719136e-06
|
912 |
+
},
|
913 |
+
{
|
914 |
+
"episode": 23552,
|
915 |
+
"epoch": 0.10089102124742974,
|
916 |
+
"eps": 7,
|
917 |
+
"loss/policy_avg": 0.08247450739145279,
|
918 |
+
"lr": 2.47265625e-06,
|
919 |
+
"objective/entropy": -63.1307487487793,
|
920 |
+
"objective/kl": 18.76070785522461,
|
921 |
+
"objective/non_score_reward": -0.9380353689193726,
|
922 |
+
"objective/rlhf_reward": 7.654726982116699,
|
923 |
+
"objective/scores": 8.59276294708252,
|
924 |
+
"policy/approxkl_avg": 0.11724384129047394,
|
925 |
+
"policy/clipfrac_avg": 0.283203125,
|
926 |
+
"policy/entropy_avg": 0.41404789686203003,
|
927 |
+
"step": 45,
|
928 |
+
"val/clipfrac_avg": 0.0,
|
929 |
+
"val/num_eos_tokens": 3442,
|
930 |
+
"val/ratio": 0.999979555606842,
|
931 |
+
"val/ratio_var": 1.9009071365871932e-06
|
932 |
+
},
|
933 |
+
{
|
934 |
+
"episode": 24064,
|
935 |
+
"epoch": 0.10308430431802605,
|
936 |
+
"eps": 7,
|
937 |
+
"loss/policy_avg": 0.0794791728258133,
|
938 |
+
"lr": 2.4609375e-06,
|
939 |
+
"objective/entropy": -61.37458419799805,
|
940 |
+
"objective/kl": 19.370513916015625,
|
941 |
+
"objective/non_score_reward": -0.968525767326355,
|
942 |
+
"objective/rlhf_reward": 7.9266767501831055,
|
943 |
+
"objective/scores": 8.89520263671875,
|
944 |
+
"policy/approxkl_avg": 0.10954838991165161,
|
945 |
+
"policy/clipfrac_avg": 0.322265625,
|
946 |
+
"policy/entropy_avg": 0.3940267860889435,
|
947 |
+
"step": 46,
|
948 |
+
"val/clipfrac_avg": 0.0,
|
949 |
+
"val/num_eos_tokens": 3371,
|
950 |
+
"val/ratio": 0.9999462366104126,
|
951 |
+
"val/ratio_var": 1.5500449990213383e-06
|
952 |
+
},
|
953 |
+
{
|
954 |
+
"episode": 24576,
|
955 |
+
"epoch": 0.10527758738862235,
|
956 |
+
"eps": 7,
|
957 |
+
"loss/policy_avg": 0.14141422510147095,
|
958 |
+
"lr": 2.4492187500000002e-06,
|
959 |
+
"objective/entropy": -59.750732421875,
|
960 |
+
"objective/kl": 20.578956604003906,
|
961 |
+
"objective/non_score_reward": -1.0289478302001953,
|
962 |
+
"objective/rlhf_reward": 7.836535453796387,
|
963 |
+
"objective/scores": 8.865484237670898,
|
964 |
+
"policy/approxkl_avg": 0.1274111419916153,
|
965 |
+
"policy/clipfrac_avg": 0.3125,
|
966 |
+
"policy/entropy_avg": 0.4056673049926758,
|
967 |
+
"step": 47,
|
968 |
+
"val/clipfrac_avg": 0.0,
|
969 |
+
"val/num_eos_tokens": 3306,
|
970 |
+
"val/ratio": 0.9998725056648254,
|
971 |
+
"val/ratio_var": 1.8748906995824655e-06
|
972 |
+
},
|
973 |
+
{
|
974 |
+
"episode": 25088,
|
975 |
+
"epoch": 0.10747087045921865,
|
976 |
+
"eps": 7,
|
977 |
+
"loss/policy_avg": 0.140358105301857,
|
978 |
+
"lr": 2.4375e-06,
|
979 |
+
"objective/entropy": -57.218482971191406,
|
980 |
+
"objective/kl": 20.602365493774414,
|
981 |
+
"objective/non_score_reward": -1.0301182270050049,
|
982 |
+
"objective/rlhf_reward": 7.965327262878418,
|
983 |
+
"objective/scores": 8.995445251464844,
|
984 |
+
"policy/approxkl_avg": 0.1341305673122406,
|
985 |
+
"policy/clipfrac_avg": 0.353515625,
|
986 |
+
"policy/entropy_avg": 0.4019361138343811,
|
987 |
+
"step": 48,
|
988 |
+
"val/clipfrac_avg": 0.0,
|
989 |
+
"val/num_eos_tokens": 3124,
|
990 |
+
"val/ratio": 1.0001575946807861,
|
991 |
+
"val/ratio_var": 1.877664203675522e-06
|
992 |
+
},
|
993 |
+
{
|
994 |
+
"episode": 25600,
|
995 |
+
"epoch": 0.10966415352981494,
|
996 |
+
"eps": 7,
|
997 |
+
"loss/policy_avg": 0.10467305779457092,
|
998 |
+
"lr": 2.42578125e-06,
|
999 |
+
"objective/entropy": -56.535526275634766,
|
1000 |
+
"objective/kl": 20.753368377685547,
|
1001 |
+
"objective/non_score_reward": -1.0376684665679932,
|
1002 |
+
"objective/rlhf_reward": 7.955059051513672,
|
1003 |
+
"objective/scores": 8.992727279663086,
|
1004 |
+
"policy/approxkl_avg": 0.12367159128189087,
|
1005 |
+
"policy/clipfrac_avg": 0.306640625,
|
1006 |
+
"policy/entropy_avg": 0.40172457695007324,
|
1007 |
+
"step": 49,
|
1008 |
+
"val/clipfrac_avg": 0.0,
|
1009 |
+
"val/num_eos_tokens": 3155,
|
1010 |
+
"val/ratio": 0.9997079372406006,
|
1011 |
+
"val/ratio_var": 1.7377291214870638e-06
|
1012 |
+
},
|
1013 |
+
{
|
1014 |
+
"episode": 26112,
|
1015 |
+
"epoch": 0.11185743660041124,
|
1016 |
+
"eps": 7,
|
1017 |
+
"loss/policy_avg": 0.13892027735710144,
|
1018 |
+
"lr": 2.4140625000000002e-06,
|
1019 |
+
"objective/entropy": -57.01996994018555,
|
1020 |
+
"objective/kl": 21.23623275756836,
|
1021 |
+
"objective/non_score_reward": -1.0618116855621338,
|
1022 |
+
"objective/rlhf_reward": 7.532562255859375,
|
1023 |
+
"objective/scores": 8.59437370300293,
|
1024 |
+
"policy/approxkl_avg": 0.12606856226921082,
|
1025 |
+
"policy/clipfrac_avg": 0.337890625,
|
1026 |
+
"policy/entropy_avg": 0.39646661281585693,
|
1027 |
+
"step": 50,
|
1028 |
+
"val/clipfrac_avg": 0.0,
|
1029 |
+
"val/num_eos_tokens": 3053,
|
1030 |
+
"val/ratio": 1.0001752376556396,
|
1031 |
+
"val/ratio_var": 1.693503804744978e-06
|
1032 |
+
},
|
1033 |
+
{
|
1034 |
+
"episode": 26624,
|
1035 |
+
"epoch": 0.11405071967100754,
|
1036 |
+
"eps": 7,
|
1037 |
+
"loss/policy_avg": 0.15690860152244568,
|
1038 |
+
"lr": 2.40234375e-06,
|
1039 |
+
"objective/entropy": -53.88324737548828,
|
1040 |
+
"objective/kl": 22.00820541381836,
|
1041 |
+
"objective/non_score_reward": -1.1004102230072021,
|
1042 |
+
"objective/rlhf_reward": 7.813861846923828,
|
1043 |
+
"objective/scores": 8.91427230834961,
|
1044 |
+
"policy/approxkl_avg": 0.13617350161075592,
|
1045 |
+
"policy/clipfrac_avg": 0.30078125,
|
1046 |
+
"policy/entropy_avg": 0.40143150091171265,
|
1047 |
+
"step": 51,
|
1048 |
+
"val/clipfrac_avg": 0.0,
|
1049 |
+
"val/num_eos_tokens": 2971,
|
1050 |
+
"val/ratio": 1.0001614093780518,
|
1051 |
+
"val/ratio_var": 1.4389252100954764e-06
|
1052 |
+
},
|
1053 |
+
{
|
1054 |
+
"episode": 27136,
|
1055 |
+
"epoch": 0.11624400274160383,
|
1056 |
+
"eps": 7,
|
1057 |
+
"loss/policy_avg": 0.1458064317703247,
|
1058 |
+
"lr": 2.390625e-06,
|
1059 |
+
"objective/entropy": -55.348575592041016,
|
1060 |
+
"objective/kl": 21.728233337402344,
|
1061 |
+
"objective/non_score_reward": -1.086411714553833,
|
1062 |
+
"objective/rlhf_reward": 7.780529499053955,
|
1063 |
+
"objective/scores": 8.866941452026367,
|
1064 |
+
"policy/approxkl_avg": 0.13537180423736572,
|
1065 |
+
"policy/clipfrac_avg": 0.33203125,
|
1066 |
+
"policy/entropy_avg": 0.4011831283569336,
|
1067 |
+
"step": 52,
|
1068 |
+
"val/clipfrac_avg": 0.0,
|
1069 |
+
"val/num_eos_tokens": 2986,
|
1070 |
+
"val/ratio": 0.9999277591705322,
|
1071 |
+
"val/ratio_var": 1.8148826939068385e-06
|
1072 |
+
},
|
1073 |
+
{
|
1074 |
+
"episode": 27648,
|
1075 |
+
"epoch": 0.11843728581220014,
|
1076 |
+
"eps": 7,
|
1077 |
+
"loss/policy_avg": 0.14188739657402039,
|
1078 |
+
"lr": 2.3789062500000002e-06,
|
1079 |
+
"objective/entropy": -55.3588981628418,
|
1080 |
+
"objective/kl": 22.07166290283203,
|
1081 |
+
"objective/non_score_reward": -1.1035832166671753,
|
1082 |
+
"objective/rlhf_reward": 8.1143798828125,
|
1083 |
+
"objective/scores": 9.217963218688965,
|
1084 |
+
"policy/approxkl_avg": 0.14667567610740662,
|
1085 |
+
"policy/clipfrac_avg": 0.294921875,
|
1086 |
+
"policy/entropy_avg": 0.3902939260005951,
|
1087 |
+
"step": 53,
|
1088 |
+
"val/clipfrac_avg": 0.0,
|
1089 |
+
"val/num_eos_tokens": 2825,
|
1090 |
+
"val/ratio": 1.0000810623168945,
|
1091 |
+
"val/ratio_var": 1.813857466004265e-06
|
1092 |
+
},
|
1093 |
+
{
|
1094 |
+
"episode": 28160,
|
1095 |
+
"epoch": 0.12063056888279644,
|
1096 |
+
"eps": 7,
|
1097 |
+
"loss/policy_avg": 0.16617950797080994,
|
1098 |
+
"lr": 2.3671875e-06,
|
1099 |
+
"objective/entropy": -54.18295669555664,
|
1100 |
+
"objective/kl": 22.74285125732422,
|
1101 |
+
"objective/non_score_reward": -1.1371426582336426,
|
1102 |
+
"objective/rlhf_reward": 8.002310752868652,
|
1103 |
+
"objective/scores": 9.139452934265137,
|
1104 |
+
"policy/approxkl_avg": 0.15451765060424805,
|
1105 |
+
"policy/clipfrac_avg": 0.298828125,
|
1106 |
+
"policy/entropy_avg": 0.39271724224090576,
|
1107 |
+
"step": 54,
|
1108 |
+
"val/clipfrac_avg": 0.0,
|
1109 |
+
"val/num_eos_tokens": 2710,
|
1110 |
+
"val/ratio": 0.9999623894691467,
|
1111 |
+
"val/ratio_var": 3.920497420040192e-06
|
1112 |
+
},
|
1113 |
+
{
|
1114 |
+
"episode": 28672,
|
1115 |
+
"epoch": 0.12282385195339274,
|
1116 |
+
"eps": 7,
|
1117 |
+
"loss/policy_avg": 0.1481824815273285,
|
1118 |
+
"lr": 2.35546875e-06,
|
1119 |
+
"objective/entropy": -54.8997917175293,
|
1120 |
+
"objective/kl": 21.62200927734375,
|
1121 |
+
"objective/non_score_reward": -1.0811004638671875,
|
1122 |
+
"objective/rlhf_reward": 7.914689540863037,
|
1123 |
+
"objective/scores": 8.995789527893066,
|
1124 |
+
"policy/approxkl_avg": 0.12237914651632309,
|
1125 |
+
"policy/clipfrac_avg": 0.2890625,
|
1126 |
+
"policy/entropy_avg": 0.3880503177642822,
|
1127 |
+
"step": 55,
|
1128 |
+
"val/clipfrac_avg": 0.0,
|
1129 |
+
"val/num_eos_tokens": 2766,
|
1130 |
+
"val/ratio": 1.0002869367599487,
|
1131 |
+
"val/ratio_var": 1.1631994993877015e-06
|
1132 |
+
},
|
1133 |
+
{
|
1134 |
+
"episode": 29184,
|
1135 |
+
"epoch": 0.12501713502398903,
|
1136 |
+
"eps": 7,
|
1137 |
+
"loss/policy_avg": 0.14690028131008148,
|
1138 |
+
"lr": 2.3437500000000002e-06,
|
1139 |
+
"objective/entropy": -57.363319396972656,
|
1140 |
+
"objective/kl": 22.510290145874023,
|
1141 |
+
"objective/non_score_reward": -1.1255145072937012,
|
1142 |
+
"objective/rlhf_reward": 8.185041427612305,
|
1143 |
+
"objective/scores": 9.310556411743164,
|
1144 |
+
"policy/approxkl_avg": 0.11880584061145782,
|
1145 |
+
"policy/clipfrac_avg": 0.31640625,
|
1146 |
+
"policy/entropy_avg": 0.3666766285896301,
|
1147 |
+
"step": 56,
|
1148 |
+
"val/clipfrac_avg": 0.0,
|
1149 |
+
"val/num_eos_tokens": 2641,
|
1150 |
+
"val/ratio": 0.999775767326355,
|
1151 |
+
"val/ratio_var": 1.7523042288303259e-06
|
1152 |
+
},
|
1153 |
+
{
|
1154 |
+
"episode": 29696,
|
1155 |
+
"epoch": 0.12721041809458533,
|
1156 |
+
"eps": 7,
|
1157 |
+
"loss/policy_avg": 0.2148008942604065,
|
1158 |
+
"lr": 2.33203125e-06,
|
1159 |
+
"objective/entropy": -53.91007995605469,
|
1160 |
+
"objective/kl": 22.808076858520508,
|
1161 |
+
"objective/non_score_reward": -1.1404038667678833,
|
1162 |
+
"objective/rlhf_reward": 7.914100170135498,
|
1163 |
+
"objective/scores": 9.05450439453125,
|
1164 |
+
"policy/approxkl_avg": 0.1747734248638153,
|
1165 |
+
"policy/clipfrac_avg": 0.32421875,
|
1166 |
+
"policy/entropy_avg": 0.3906838893890381,
|
1167 |
+
"step": 57,
|
1168 |
+
"val/clipfrac_avg": 0.0,
|
1169 |
+
"val/num_eos_tokens": 2627,
|
1170 |
+
"val/ratio": 0.999924898147583,
|
1171 |
+
"val/ratio_var": 2.3690538455412025e-06
|
1172 |
+
},
|
1173 |
+
{
|
1174 |
+
"episode": 30208,
|
1175 |
+
"epoch": 0.12940370116518163,
|
1176 |
+
"eps": 7,
|
1177 |
+
"loss/policy_avg": 0.10798689723014832,
|
1178 |
+
"lr": 2.3203125e-06,
|
1179 |
+
"objective/entropy": -56.803367614746094,
|
1180 |
+
"objective/kl": 22.174915313720703,
|
1181 |
+
"objective/non_score_reward": -1.108745813369751,
|
1182 |
+
"objective/rlhf_reward": 7.929474830627441,
|
1183 |
+
"objective/scores": 9.038220405578613,
|
1184 |
+
"policy/approxkl_avg": 0.13498836755752563,
|
1185 |
+
"policy/clipfrac_avg": 0.3046875,
|
1186 |
+
"policy/entropy_avg": 0.3794073462486267,
|
1187 |
+
"step": 58,
|
1188 |
+
"val/clipfrac_avg": 0.0,
|
1189 |
+
"val/num_eos_tokens": 2560,
|
1190 |
+
"val/ratio": 0.9999980926513672,
|
1191 |
+
"val/ratio_var": 1.8393640175418113e-06
|
1192 |
+
},
|
1193 |
+
{
|
1194 |
+
"episode": 30720,
|
1195 |
+
"epoch": 0.13159698423577793,
|
1196 |
+
"eps": 7,
|
1197 |
+
"loss/policy_avg": 0.13295301795005798,
|
1198 |
+
"lr": 2.30859375e-06,
|
1199 |
+
"objective/entropy": -57.542625427246094,
|
1200 |
+
"objective/kl": 22.180622100830078,
|
1201 |
+
"objective/non_score_reward": -1.1090312004089355,
|
1202 |
+
"objective/rlhf_reward": 7.812742233276367,
|
1203 |
+
"objective/scores": 8.921772956848145,
|
1204 |
+
"policy/approxkl_avg": 0.11531847715377808,
|
1205 |
+
"policy/clipfrac_avg": 0.310546875,
|
1206 |
+
"policy/entropy_avg": 0.3836986720561981,
|
1207 |
+
"step": 59,
|
1208 |
+
"val/clipfrac_avg": 0.0,
|
1209 |
+
"val/num_eos_tokens": 2577,
|
1210 |
+
"val/ratio": 1.000056505203247,
|
1211 |
+
"val/ratio_var": 1.6733740721974755e-06
|
1212 |
+
},
|
1213 |
+
{
|
1214 |
+
"episode": 31232,
|
1215 |
+
"epoch": 0.13379026730637422,
|
1216 |
+
"eps": 7,
|
1217 |
+
"loss/policy_avg": 0.13682591915130615,
|
1218 |
+
"lr": 2.296875e-06,
|
1219 |
+
"objective/entropy": -56.855690002441406,
|
1220 |
+
"objective/kl": 22.521320343017578,
|
1221 |
+
"objective/non_score_reward": -1.126065969467163,
|
1222 |
+
"objective/rlhf_reward": 8.064724922180176,
|
1223 |
+
"objective/scores": 9.190791130065918,
|
1224 |
+
"policy/approxkl_avg": 0.12303945422172546,
|
1225 |
+
"policy/clipfrac_avg": 0.2890625,
|
1226 |
+
"policy/entropy_avg": 0.35441768169403076,
|
1227 |
+
"step": 60,
|
1228 |
+
"val/clipfrac_avg": 0.0,
|
1229 |
+
"val/num_eos_tokens": 2446,
|
1230 |
+
"val/ratio": 1.0001006126403809,
|
1231 |
+
"val/ratio_var": 2.1652347186318366e-06
|
1232 |
+
},
|
1233 |
+
{
|
1234 |
+
"episode": 31744,
|
1235 |
+
"epoch": 0.13598355037697052,
|
1236 |
+
"eps": 7,
|
1237 |
+
"loss/policy_avg": 0.12291203439235687,
|
1238 |
+
"lr": 2.28515625e-06,
|
1239 |
+
"objective/entropy": -58.003257751464844,
|
1240 |
+
"objective/kl": 22.295536041259766,
|
1241 |
+
"objective/non_score_reward": -1.114776849746704,
|
1242 |
+
"objective/rlhf_reward": 8.06563949584961,
|
1243 |
+
"objective/scores": 9.180416107177734,
|
1244 |
+
"policy/approxkl_avg": 0.1267692595720291,
|
1245 |
+
"policy/clipfrac_avg": 0.30859375,
|
1246 |
+
"policy/entropy_avg": 0.3748782277107239,
|
1247 |
+
"step": 61,
|
1248 |
+
"val/clipfrac_avg": 0.0,
|
1249 |
+
"val/num_eos_tokens": 2472,
|
1250 |
+
"val/ratio": 0.9999428391456604,
|
1251 |
+
"val/ratio_var": 1.7910625729200547e-06
|
1252 |
+
},
|
1253 |
+
{
|
1254 |
+
"episode": 32256,
|
1255 |
+
"epoch": 0.13817683344756682,
|
1256 |
+
"eps": 7,
|
1257 |
+
"loss/policy_avg": 0.10350387543439865,
|
1258 |
+
"lr": 2.2734375e-06,
|
1259 |
+
"objective/entropy": -58.27714920043945,
|
1260 |
+
"objective/kl": 21.72718048095703,
|
1261 |
+
"objective/non_score_reward": -1.0863591432571411,
|
1262 |
+
"objective/rlhf_reward": 7.940745830535889,
|
1263 |
+
"objective/scores": 9.027105331420898,
|
1264 |
+
"policy/approxkl_avg": 0.11410067230463028,
|
1265 |
+
"policy/clipfrac_avg": 0.3125,
|
1266 |
+
"policy/entropy_avg": 0.35974133014678955,
|
1267 |
+
"step": 62,
|
1268 |
+
"val/clipfrac_avg": 0.0,
|
1269 |
+
"val/num_eos_tokens": 2396,
|
1270 |
+
"val/ratio": 1.0000014305114746,
|
1271 |
+
"val/ratio_var": 1.7253937585337553e-06
|
1272 |
+
},
|
1273 |
+
{
|
1274 |
+
"episode": 32768,
|
1275 |
+
"epoch": 0.1403701165181631,
|
1276 |
+
"eps": 7,
|
1277 |
+
"loss/policy_avg": 0.14234605431556702,
|
1278 |
+
"lr": 2.26171875e-06,
|
1279 |
+
"objective/entropy": -58.319828033447266,
|
1280 |
+
"objective/kl": 21.27573013305664,
|
1281 |
+
"objective/non_score_reward": -1.063786506652832,
|
1282 |
+
"objective/rlhf_reward": 7.917078018188477,
|
1283 |
+
"objective/scores": 8.980864524841309,
|
1284 |
+
"policy/approxkl_avg": 0.09502886235713959,
|
1285 |
+
"policy/clipfrac_avg": 0.267578125,
|
1286 |
+
"policy/entropy_avg": 0.36790037155151367,
|
1287 |
+
"step": 63,
|
1288 |
+
"val/clipfrac_avg": 0.0,
|
1289 |
+
"val/num_eos_tokens": 2341,
|
1290 |
+
"val/ratio": 1.0000803470611572,
|
1291 |
+
"val/ratio_var": 1.6205699466809165e-06
|
1292 |
+
},
|
1293 |
+
{
|
1294 |
+
"episode": 33280,
|
1295 |
+
"epoch": 0.14256339958875944,
|
1296 |
+
"eps": 7,
|
1297 |
+
"loss/policy_avg": 0.18124544620513916,
|
1298 |
+
"lr": 2.25e-06,
|
1299 |
+
"objective/entropy": -56.98847961425781,
|
1300 |
+
"objective/kl": 22.343021392822266,
|
1301 |
+
"objective/non_score_reward": -1.1171510219573975,
|
1302 |
+
"objective/rlhf_reward": 8.013626098632812,
|
1303 |
+
"objective/scores": 9.130776405334473,
|
1304 |
+
"policy/approxkl_avg": 0.12282164394855499,
|
1305 |
+
"policy/clipfrac_avg": 0.279296875,
|
1306 |
+
"policy/entropy_avg": 0.3647758364677429,
|
1307 |
+
"step": 64,
|
1308 |
+
"val/clipfrac_avg": 0.0,
|
1309 |
+
"val/num_eos_tokens": 2392,
|
1310 |
+
"val/ratio": 1.000246524810791,
|
1311 |
+
"val/ratio_var": 1.054077802109532e-05
|
1312 |
+
},
|
1313 |
+
{
|
1314 |
+
"episode": 33792,
|
1315 |
+
"epoch": 0.14475668265935573,
|
1316 |
+
"eps": 7,
|
1317 |
+
"loss/policy_avg": 0.10490100830793381,
|
1318 |
+
"lr": 2.23828125e-06,
|
1319 |
+
"objective/entropy": -59.54853820800781,
|
1320 |
+
"objective/kl": 21.909761428833008,
|
1321 |
+
"objective/non_score_reward": -1.0954880714416504,
|
1322 |
+
"objective/rlhf_reward": 7.99943733215332,
|
1323 |
+
"objective/scores": 9.094924926757812,
|
1324 |
+
"policy/approxkl_avg": 0.1441516876220703,
|
1325 |
+
"policy/clipfrac_avg": 0.263671875,
|
1326 |
+
"policy/entropy_avg": 0.35612696409225464,
|
1327 |
+
"step": 65,
|
1328 |
+
"val/clipfrac_avg": 0.0,
|
1329 |
+
"val/num_eos_tokens": 2385,
|
1330 |
+
"val/ratio": 0.9997637867927551,
|
1331 |
+
"val/ratio_var": 1.6264643818431068e-06
|
1332 |
+
},
|
1333 |
+
{
|
1334 |
+
"episode": 34304,
|
1335 |
+
"epoch": 0.14694996572995203,
|
1336 |
+
"eps": 7,
|
1337 |
+
"loss/policy_avg": 0.11742566525936127,
|
1338 |
+
"lr": 2.2265625e-06,
|
1339 |
+
"objective/entropy": -56.3593864440918,
|
1340 |
+
"objective/kl": 22.55255699157715,
|
1341 |
+
"objective/non_score_reward": -1.1276278495788574,
|
1342 |
+
"objective/rlhf_reward": 7.8993988037109375,
|
1343 |
+
"objective/scores": 9.027027130126953,
|
1344 |
+
"policy/approxkl_avg": 0.1287810504436493,
|
1345 |
+
"policy/clipfrac_avg": 0.330078125,
|
1346 |
+
"policy/entropy_avg": 0.37438803911209106,
|
1347 |
+
"step": 66,
|
1348 |
+
"val/clipfrac_avg": 0.0,
|
1349 |
+
"val/num_eos_tokens": 2274,
|
1350 |
+
"val/ratio": 0.9996710419654846,
|
1351 |
+
"val/ratio_var": 1.698708615549549e-06
|
1352 |
+
},
|
1353 |
+
{
|
1354 |
+
"episode": 34816,
|
1355 |
+
"epoch": 0.14914324880054833,
|
1356 |
+
"eps": 7,
|
1357 |
+
"loss/policy_avg": 0.13423027098178864,
|
1358 |
+
"lr": 2.21484375e-06,
|
1359 |
+
"objective/entropy": -59.66276550292969,
|
1360 |
+
"objective/kl": 21.369693756103516,
|
1361 |
+
"objective/non_score_reward": -1.0684847831726074,
|
1362 |
+
"objective/rlhf_reward": 8.15538215637207,
|
1363 |
+
"objective/scores": 9.22386646270752,
|
1364 |
+
"policy/approxkl_avg": 0.10780099779367447,
|
1365 |
+
"policy/clipfrac_avg": 0.2578125,
|
1366 |
+
"policy/entropy_avg": 0.3673896789550781,
|
1367 |
+
"step": 67,
|
1368 |
+
"val/clipfrac_avg": 0.0,
|
1369 |
+
"val/num_eos_tokens": 2384,
|
1370 |
+
"val/ratio": 1.0000427961349487,
|
1371 |
+
"val/ratio_var": 1.8952780465042451e-06
|
1372 |
+
},
|
1373 |
+
{
|
1374 |
+
"episode": 35328,
|
1375 |
+
"epoch": 0.15133653187114463,
|
1376 |
+
"eps": 7,
|
1377 |
+
"loss/policy_avg": 0.1079690158367157,
|
1378 |
+
"lr": 2.203125e-06,
|
1379 |
+
"objective/entropy": -56.822410583496094,
|
1380 |
+
"objective/kl": 22.660343170166016,
|
1381 |
+
"objective/non_score_reward": -1.1330170631408691,
|
1382 |
+
"objective/rlhf_reward": 8.052312850952148,
|
1383 |
+
"objective/scores": 9.18532943725586,
|
1384 |
+
"policy/approxkl_avg": 0.13137517869472504,
|
1385 |
+
"policy/clipfrac_avg": 0.30859375,
|
1386 |
+
"policy/entropy_avg": 0.3610047399997711,
|
1387 |
+
"step": 68,
|
1388 |
+
"val/clipfrac_avg": 0.0,
|
1389 |
+
"val/num_eos_tokens": 2307,
|
1390 |
+
"val/ratio": 0.9998396635055542,
|
1391 |
+
"val/ratio_var": 2.1211410512478324e-06
|
1392 |
+
},
|
1393 |
+
{
|
1394 |
+
"episode": 35840,
|
1395 |
+
"epoch": 0.15352981494174092,
|
1396 |
+
"eps": 7,
|
1397 |
+
"loss/policy_avg": 0.13532832264900208,
|
1398 |
+
"lr": 2.19140625e-06,
|
1399 |
+
"objective/entropy": -57.17053985595703,
|
1400 |
+
"objective/kl": 22.723447799682617,
|
1401 |
+
"objective/non_score_reward": -1.1361724138259888,
|
1402 |
+
"objective/rlhf_reward": 8.25572395324707,
|
1403 |
+
"objective/scores": 9.39189624786377,
|
1404 |
+
"policy/approxkl_avg": 0.11353135108947754,
|
1405 |
+
"policy/clipfrac_avg": 0.29296875,
|
1406 |
+
"policy/entropy_avg": 0.3422977328300476,
|
1407 |
+
"step": 69,
|
1408 |
+
"val/clipfrac_avg": 0.0,
|
1409 |
+
"val/num_eos_tokens": 2249,
|
1410 |
+
"val/ratio": 1.000388264656067,
|
1411 |
+
"val/ratio_var": 1.609949322300963e-06
|
1412 |
+
},
|
1413 |
+
{
|
1414 |
+
"episode": 36352,
|
1415 |
+
"epoch": 0.15572309801233722,
|
1416 |
+
"eps": 7,
|
1417 |
+
"loss/policy_avg": 0.11315006762742996,
|
1418 |
+
"lr": 2.1796875e-06,
|
1419 |
+
"objective/entropy": -55.464996337890625,
|
1420 |
+
"objective/kl": 23.183774948120117,
|
1421 |
+
"objective/non_score_reward": -1.1591888666152954,
|
1422 |
+
"objective/rlhf_reward": 8.152729034423828,
|
1423 |
+
"objective/scores": 9.311917304992676,
|
1424 |
+
"policy/approxkl_avg": 0.13305282592773438,
|
1425 |
+
"policy/clipfrac_avg": 0.3203125,
|
1426 |
+
"policy/entropy_avg": 0.33513692021369934,
|
1427 |
+
"step": 70,
|
1428 |
+
"val/clipfrac_avg": 0.0,
|
1429 |
+
"val/num_eos_tokens": 2241,
|
1430 |
+
"val/ratio": 1.0002946853637695,
|
1431 |
+
"val/ratio_var": 2.0558102278300794e-06
|
1432 |
+
},
|
1433 |
+
{
|
1434 |
+
"episode": 36864,
|
1435 |
+
"epoch": 0.15791638108293352,
|
1436 |
+
"eps": 7,
|
1437 |
+
"loss/policy_avg": 0.12405405938625336,
|
1438 |
+
"lr": 2.16796875e-06,
|
1439 |
+
"objective/entropy": -55.366004943847656,
|
1440 |
+
"objective/kl": 23.366641998291016,
|
1441 |
+
"objective/non_score_reward": -1.1683320999145508,
|
1442 |
+
"objective/rlhf_reward": 8.099801063537598,
|
1443 |
+
"objective/scores": 9.268133163452148,
|
1444 |
+
"policy/approxkl_avg": 0.1177653819322586,
|
1445 |
+
"policy/clipfrac_avg": 0.30078125,
|
1446 |
+
"policy/entropy_avg": 0.36095091700553894,
|
1447 |
+
"step": 71,
|
1448 |
+
"val/clipfrac_avg": 0.0,
|
1449 |
+
"val/num_eos_tokens": 2242,
|
1450 |
+
"val/ratio": 1.0000619888305664,
|
1451 |
+
"val/ratio_var": 1.8105098433807143e-06
|
1452 |
+
},
|
1453 |
+
{
|
1454 |
+
"episode": 37376,
|
1455 |
+
"epoch": 0.1601096641535298,
|
1456 |
+
"eps": 7,
|
1457 |
+
"loss/policy_avg": 0.13172870874404907,
|
1458 |
+
"lr": 2.15625e-06,
|
1459 |
+
"objective/entropy": -55.20182800292969,
|
1460 |
+
"objective/kl": 25.142457962036133,
|
1461 |
+
"objective/non_score_reward": -1.2571228742599487,
|
1462 |
+
"objective/rlhf_reward": 8.464044570922852,
|
1463 |
+
"objective/scores": 9.72116756439209,
|
1464 |
+
"policy/approxkl_avg": 0.13379380106925964,
|
1465 |
+
"policy/clipfrac_avg": 0.3125,
|
1466 |
+
"policy/entropy_avg": 0.3525581955909729,
|
1467 |
+
"step": 72,
|
1468 |
+
"val/clipfrac_avg": 0.0,
|
1469 |
+
"val/num_eos_tokens": 2087,
|
1470 |
+
"val/ratio": 1.00001859664917,
|
1471 |
+
"val/ratio_var": 2.2096453449194087e-06
|
1472 |
+
},
|
1473 |
+
{
|
1474 |
+
"episode": 37888,
|
1475 |
+
"epoch": 0.1623029472241261,
|
1476 |
+
"eps": 7,
|
1477 |
+
"loss/policy_avg": 0.10756374150514603,
|
1478 |
+
"lr": 2.14453125e-06,
|
1479 |
+
"objective/entropy": -55.84001159667969,
|
1480 |
+
"objective/kl": 24.132648468017578,
|
1481 |
+
"objective/non_score_reward": -1.206632375717163,
|
1482 |
+
"objective/rlhf_reward": 8.113821029663086,
|
1483 |
+
"objective/scores": 9.320453643798828,
|
1484 |
+
"policy/approxkl_avg": 0.11456403136253357,
|
1485 |
+
"policy/clipfrac_avg": 0.28515625,
|
1486 |
+
"policy/entropy_avg": 0.3468587398529053,
|
1487 |
+
"step": 73,
|
1488 |
+
"val/clipfrac_avg": 0.0,
|
1489 |
+
"val/num_eos_tokens": 2222,
|
1490 |
+
"val/ratio": 0.9999667406082153,
|
1491 |
+
"val/ratio_var": 1.6667198678987916e-06
|
1492 |
+
},
|
1493 |
+
{
|
1494 |
+
"episode": 38400,
|
1495 |
+
"epoch": 0.1644962302947224,
|
1496 |
+
"eps": 7,
|
1497 |
+
"loss/policy_avg": 0.12748099863529205,
|
1498 |
+
"lr": 2.1328125e-06,
|
1499 |
+
"objective/entropy": -53.43772888183594,
|
1500 |
+
"objective/kl": 25.255722045898438,
|
1501 |
+
"objective/non_score_reward": -1.2627861499786377,
|
1502 |
+
"objective/rlhf_reward": 8.202866554260254,
|
1503 |
+
"objective/scores": 9.465652465820312,
|
1504 |
+
"policy/approxkl_avg": 0.12267530709505081,
|
1505 |
+
"policy/clipfrac_avg": 0.2734375,
|
1506 |
+
"policy/entropy_avg": 0.33846238255500793,
|
1507 |
+
"step": 74,
|
1508 |
+
"val/clipfrac_avg": 0.0,
|
1509 |
+
"val/num_eos_tokens": 2067,
|
1510 |
+
"val/ratio": 0.999931812286377,
|
1511 |
+
"val/ratio_var": 1.4451366041612346e-06
|
1512 |
+
},
|
1513 |
+
{
|
1514 |
+
"episode": 38912,
|
1515 |
+
"epoch": 0.1666895133653187,
|
1516 |
+
"eps": 7,
|
1517 |
+
"loss/policy_avg": 0.10486993193626404,
|
1518 |
+
"lr": 2.12109375e-06,
|
1519 |
+
"objective/entropy": -53.43898391723633,
|
1520 |
+
"objective/kl": 25.309619903564453,
|
1521 |
+
"objective/non_score_reward": -1.2654809951782227,
|
1522 |
+
"objective/rlhf_reward": 8.32456111907959,
|
1523 |
+
"objective/scores": 9.590042114257812,
|
1524 |
+
"policy/approxkl_avg": 0.13389162719249725,
|
1525 |
+
"policy/clipfrac_avg": 0.310546875,
|
1526 |
+
"policy/entropy_avg": 0.336054265499115,
|
1527 |
+
"step": 75,
|
1528 |
+
"val/clipfrac_avg": 0.0,
|
1529 |
+
"val/num_eos_tokens": 2128,
|
1530 |
+
"val/ratio": 0.9998234510421753,
|
1531 |
+
"val/ratio_var": 2.210314505646238e-06
|
1532 |
+
},
|
1533 |
+
{
|
1534 |
+
"episode": 39424,
|
1535 |
+
"epoch": 0.168882796435915,
|
1536 |
+
"eps": 7,
|
1537 |
+
"loss/policy_avg": 0.14061179757118225,
|
1538 |
+
"lr": 2.109375e-06,
|
1539 |
+
"objective/entropy": -54.23058319091797,
|
1540 |
+
"objective/kl": 24.61229705810547,
|
1541 |
+
"objective/non_score_reward": -1.2306149005889893,
|
1542 |
+
"objective/rlhf_reward": 8.29504680633545,
|
1543 |
+
"objective/scores": 9.52566146850586,
|
1544 |
+
"policy/approxkl_avg": 0.1283368021249771,
|
1545 |
+
"policy/clipfrac_avg": 0.30859375,
|
1546 |
+
"policy/entropy_avg": 0.3492497205734253,
|
1547 |
+
"step": 76,
|
1548 |
+
"val/clipfrac_avg": 0.0,
|
1549 |
+
"val/num_eos_tokens": 2146,
|
1550 |
+
"val/ratio": 1.0000791549682617,
|
1551 |
+
"val/ratio_var": 2.4275761916214833e-06
|
1552 |
+
},
|
1553 |
+
{
|
1554 |
+
"episode": 39936,
|
1555 |
+
"epoch": 0.1710760795065113,
|
1556 |
+
"eps": 7,
|
1557 |
+
"loss/policy_avg": 0.13834567368030548,
|
1558 |
+
"lr": 2.09765625e-06,
|
1559 |
+
"objective/entropy": -53.97233200073242,
|
1560 |
+
"objective/kl": 25.951492309570312,
|
1561 |
+
"objective/non_score_reward": -1.297574520111084,
|
1562 |
+
"objective/rlhf_reward": 8.272811889648438,
|
1563 |
+
"objective/scores": 9.57038688659668,
|
1564 |
+
"policy/approxkl_avg": 0.13967689871788025,
|
1565 |
+
"policy/clipfrac_avg": 0.3046875,
|
1566 |
+
"policy/entropy_avg": 0.3345507085323334,
|
1567 |
+
"step": 77,
|
1568 |
+
"val/clipfrac_avg": 0.0,
|
1569 |
+
"val/num_eos_tokens": 2126,
|
1570 |
+
"val/ratio": 0.9998956918716431,
|
1571 |
+
"val/ratio_var": 1.739919639476284e-06
|
1572 |
+
},
|
1573 |
+
{
|
1574 |
+
"episode": 40448,
|
1575 |
+
"epoch": 0.17326936257710762,
|
1576 |
+
"eps": 7,
|
1577 |
+
"loss/policy_avg": 0.1322835385799408,
|
1578 |
+
"lr": 2.0859375e-06,
|
1579 |
+
"objective/entropy": -52.13136672973633,
|
1580 |
+
"objective/kl": 26.608680725097656,
|
1581 |
+
"objective/non_score_reward": -1.3304340839385986,
|
1582 |
+
"objective/rlhf_reward": 8.47679328918457,
|
1583 |
+
"objective/scores": 9.80722713470459,
|
1584 |
+
"policy/approxkl_avg": 0.11805111914873123,
|
1585 |
+
"policy/clipfrac_avg": 0.29296875,
|
1586 |
+
"policy/entropy_avg": 0.32995736598968506,
|
1587 |
+
"step": 78,
|
1588 |
+
"val/clipfrac_avg": 0.0,
|
1589 |
+
"val/num_eos_tokens": 2193,
|
1590 |
+
"val/ratio": 1.000328779220581,
|
1591 |
+
"val/ratio_var": 1.6122173747135093e-06
|
1592 |
+
},
|
1593 |
+
{
|
1594 |
+
"episode": 40960,
|
1595 |
+
"epoch": 0.17546264564770392,
|
1596 |
+
"eps": 7,
|
1597 |
+
"loss/policy_avg": 0.12710238993167877,
|
1598 |
+
"lr": 2.07421875e-06,
|
1599 |
+
"objective/entropy": -49.995399475097656,
|
1600 |
+
"objective/kl": 27.390361785888672,
|
1601 |
+
"objective/non_score_reward": -1.3695181608200073,
|
1602 |
+
"objective/rlhf_reward": 8.022126197814941,
|
1603 |
+
"objective/scores": 9.391643524169922,
|
1604 |
+
"policy/approxkl_avg": 0.14070570468902588,
|
1605 |
+
"policy/clipfrac_avg": 0.3203125,
|
1606 |
+
"policy/entropy_avg": 0.3400488793849945,
|
1607 |
+
"step": 79,
|
1608 |
+
"val/clipfrac_avg": 0.0,
|
1609 |
+
"val/num_eos_tokens": 2093,
|
1610 |
+
"val/ratio": 0.9999239444732666,
|
1611 |
+
"val/ratio_var": 2.390835788901313e-06
|
1612 |
+
},
|
1613 |
+
{
|
1614 |
+
"episode": 41472,
|
1615 |
+
"epoch": 0.17765592871830022,
|
1616 |
+
"eps": 7,
|
1617 |
+
"loss/policy_avg": 0.1422366350889206,
|
1618 |
+
"lr": 2.0625e-06,
|
1619 |
+
"objective/entropy": -51.26823043823242,
|
1620 |
+
"objective/kl": 26.9412841796875,
|
1621 |
+
"objective/non_score_reward": -1.3470642566680908,
|
1622 |
+
"objective/rlhf_reward": 8.401371002197266,
|
1623 |
+
"objective/scores": 9.748435020446777,
|
1624 |
+
"policy/approxkl_avg": 0.12222130596637726,
|
1625 |
+
"policy/clipfrac_avg": 0.3125,
|
1626 |
+
"policy/entropy_avg": 0.34341585636138916,
|
1627 |
+
"step": 80,
|
1628 |
+
"val/clipfrac_avg": 0.0,
|
1629 |
+
"val/num_eos_tokens": 2095,
|
1630 |
+
"val/ratio": 1.0000666379928589,
|
1631 |
+
"val/ratio_var": 1.9948297449445818e-06
|
1632 |
+
},
|
1633 |
+
{
|
1634 |
+
"episode": 41984,
|
1635 |
+
"epoch": 0.1798492117888965,
|
1636 |
+
"eps": 7,
|
1637 |
+
"loss/policy_avg": 0.14413389563560486,
|
1638 |
+
"lr": 2.05078125e-06,
|
1639 |
+
"objective/entropy": -48.53511047363281,
|
1640 |
+
"objective/kl": 27.005571365356445,
|
1641 |
+
"objective/non_score_reward": -1.350278615951538,
|
1642 |
+
"objective/rlhf_reward": 8.146627426147461,
|
1643 |
+
"objective/scores": 9.496906280517578,
|
1644 |
+
"policy/approxkl_avg": 0.1366291642189026,
|
1645 |
+
"policy/clipfrac_avg": 0.31640625,
|
1646 |
+
"policy/entropy_avg": 0.3490913510322571,
|
1647 |
+
"step": 81,
|
1648 |
+
"val/clipfrac_avg": 0.0,
|
1649 |
+
"val/num_eos_tokens": 2138,
|
1650 |
+
"val/ratio": 1.0002083778381348,
|
1651 |
+
"val/ratio_var": 2.321082774869865e-06
|
1652 |
+
},
|
1653 |
+
{
|
1654 |
+
"episode": 42496,
|
1655 |
+
"epoch": 0.1820424948594928,
|
1656 |
+
"eps": 7,
|
1657 |
+
"loss/policy_avg": 0.2456514984369278,
|
1658 |
+
"lr": 2.0390625e-06,
|
1659 |
+
"objective/entropy": -51.47251892089844,
|
1660 |
+
"objective/kl": 26.860395431518555,
|
1661 |
+
"objective/non_score_reward": -1.343019723892212,
|
1662 |
+
"objective/rlhf_reward": 8.177400588989258,
|
1663 |
+
"objective/scores": 9.52042007446289,
|
1664 |
+
"policy/approxkl_avg": 0.14327874779701233,
|
1665 |
+
"policy/clipfrac_avg": 0.333984375,
|
1666 |
+
"policy/entropy_avg": 0.3569631576538086,
|
1667 |
+
"step": 82,
|
1668 |
+
"val/clipfrac_avg": 0.0,
|
1669 |
+
"val/num_eos_tokens": 2066,
|
1670 |
+
"val/ratio": 1.000014066696167,
|
1671 |
+
"val/ratio_var": 2.138266609108541e-06
|
1672 |
+
},
|
1673 |
+
{
|
1674 |
+
"episode": 43008,
|
1675 |
+
"epoch": 0.1842357779300891,
|
1676 |
+
"eps": 7,
|
1677 |
+
"loss/policy_avg": 0.13275276124477386,
|
1678 |
+
"lr": 2.02734375e-06,
|
1679 |
+
"objective/entropy": -49.27789306640625,
|
1680 |
+
"objective/kl": 27.951335906982422,
|
1681 |
+
"objective/non_score_reward": -1.397566795349121,
|
1682 |
+
"objective/rlhf_reward": 8.149410247802734,
|
1683 |
+
"objective/scores": 9.546977043151855,
|
1684 |
+
"policy/approxkl_avg": 0.16321811079978943,
|
1685 |
+
"policy/clipfrac_avg": 0.291015625,
|
1686 |
+
"policy/entropy_avg": 0.3553319275379181,
|
1687 |
+
"step": 83,
|
1688 |
+
"val/clipfrac_avg": 0.0,
|
1689 |
+
"val/num_eos_tokens": 2158,
|
1690 |
+
"val/ratio": 0.9998036623001099,
|
1691 |
+
"val/ratio_var": 1.9729568521142937e-06
|
1692 |
+
},
|
1693 |
+
{
|
1694 |
+
"episode": 43520,
|
1695 |
+
"epoch": 0.1864290610006854,
|
1696 |
+
"eps": 7,
|
1697 |
+
"loss/policy_avg": 0.19560450315475464,
|
1698 |
+
"lr": 2.015625e-06,
|
1699 |
+
"objective/entropy": -49.69116973876953,
|
1700 |
+
"objective/kl": 26.85687255859375,
|
1701 |
+
"objective/non_score_reward": -1.342843770980835,
|
1702 |
+
"objective/rlhf_reward": 8.36276626586914,
|
1703 |
+
"objective/scores": 9.705608367919922,
|
1704 |
+
"policy/approxkl_avg": 0.1448563039302826,
|
1705 |
+
"policy/clipfrac_avg": 0.302734375,
|
1706 |
+
"policy/entropy_avg": 0.3514622449874878,
|
1707 |
+
"step": 84,
|
1708 |
+
"val/clipfrac_avg": 0.0,
|
1709 |
+
"val/num_eos_tokens": 2213,
|
1710 |
+
"val/ratio": 0.9999929070472717,
|
1711 |
+
"val/ratio_var": 2.3642696760362014e-06
|
1712 |
+
},
|
1713 |
+
{
|
1714 |
+
"episode": 44032,
|
1715 |
+
"epoch": 0.1886223440712817,
|
1716 |
+
"eps": 7,
|
1717 |
+
"loss/policy_avg": 0.11201312392950058,
|
1718 |
+
"lr": 2.00390625e-06,
|
1719 |
+
"objective/entropy": -50.87125778198242,
|
1720 |
+
"objective/kl": 27.770545959472656,
|
1721 |
+
"objective/non_score_reward": -1.3885273933410645,
|
1722 |
+
"objective/rlhf_reward": 8.33647346496582,
|
1723 |
+
"objective/scores": 9.725000381469727,
|
1724 |
+
"policy/approxkl_avg": 0.1465965211391449,
|
1725 |
+
"policy/clipfrac_avg": 0.341796875,
|
1726 |
+
"policy/entropy_avg": 0.3566615581512451,
|
1727 |
+
"step": 85,
|
1728 |
+
"val/clipfrac_avg": 0.0,
|
1729 |
+
"val/num_eos_tokens": 2141,
|
1730 |
+
"val/ratio": 0.9996919631958008,
|
1731 |
+
"val/ratio_var": 1.868749677669257e-06
|
1732 |
+
},
|
1733 |
+
{
|
1734 |
+
"episode": 44544,
|
1735 |
+
"epoch": 0.190815627141878,
|
1736 |
+
"eps": 7,
|
1737 |
+
"loss/policy_avg": 0.14001718163490295,
|
1738 |
+
"lr": 1.9921875e-06,
|
1739 |
+
"objective/entropy": -50.17118453979492,
|
1740 |
+
"objective/kl": 27.461868286132812,
|
1741 |
+
"objective/non_score_reward": -1.3730934858322144,
|
1742 |
+
"objective/rlhf_reward": 8.599668502807617,
|
1743 |
+
"objective/scores": 9.972763061523438,
|
1744 |
+
"policy/approxkl_avg": 0.17532342672348022,
|
1745 |
+
"policy/clipfrac_avg": 0.306640625,
|
1746 |
+
"policy/entropy_avg": 0.3518564701080322,
|
1747 |
+
"step": 86,
|
1748 |
+
"val/clipfrac_avg": 0.0,
|
1749 |
+
"val/num_eos_tokens": 2145,
|
1750 |
+
"val/ratio": 0.9997180700302124,
|
1751 |
+
"val/ratio_var": 2.6727666408987716e-06
|
1752 |
+
},
|
1753 |
+
{
|
1754 |
+
"episode": 45056,
|
1755 |
+
"epoch": 0.1930089102124743,
|
1756 |
+
"eps": 7,
|
1757 |
+
"loss/policy_avg": 0.153324156999588,
|
1758 |
+
"lr": 1.98046875e-06,
|
1759 |
+
"objective/entropy": -49.21909713745117,
|
1760 |
+
"objective/kl": 28.38106918334961,
|
1761 |
+
"objective/non_score_reward": -1.419053554534912,
|
1762 |
+
"objective/rlhf_reward": 8.374338150024414,
|
1763 |
+
"objective/scores": 9.793391227722168,
|
1764 |
+
"policy/approxkl_avg": 0.1718733012676239,
|
1765 |
+
"policy/clipfrac_avg": 0.3203125,
|
1766 |
+
"policy/entropy_avg": 0.3515265882015228,
|
1767 |
+
"step": 87,
|
1768 |
+
"val/clipfrac_avg": 0.0,
|
1769 |
+
"val/num_eos_tokens": 2285,
|
1770 |
+
"val/ratio": 0.9999737739562988,
|
1771 |
+
"val/ratio_var": 1.7272968761972152e-06
|
1772 |
+
},
|
1773 |
+
{
|
1774 |
+
"episode": 45568,
|
1775 |
+
"epoch": 0.1952021932830706,
|
1776 |
+
"eps": 7,
|
1777 |
+
"loss/policy_avg": 0.12960462272167206,
|
1778 |
+
"lr": 1.96875e-06,
|
1779 |
+
"objective/entropy": -52.55916976928711,
|
1780 |
+
"objective/kl": 26.743499755859375,
|
1781 |
+
"objective/non_score_reward": -1.337174892425537,
|
1782 |
+
"objective/rlhf_reward": 8.524429321289062,
|
1783 |
+
"objective/scores": 9.861604690551758,
|
1784 |
+
"policy/approxkl_avg": 0.12384246289730072,
|
1785 |
+
"policy/clipfrac_avg": 0.2890625,
|
1786 |
+
"policy/entropy_avg": 0.34913384914398193,
|
1787 |
+
"step": 88,
|
1788 |
+
"val/clipfrac_avg": 0.0,
|
1789 |
+
"val/num_eos_tokens": 2166,
|
1790 |
+
"val/ratio": 1.0001153945922852,
|
1791 |
+
"val/ratio_var": 1.8935652406071313e-06
|
1792 |
+
},
|
1793 |
+
{
|
1794 |
+
"episode": 46080,
|
1795 |
+
"epoch": 0.1973954763536669,
|
1796 |
+
"eps": 7,
|
1797 |
+
"loss/policy_avg": 0.13431459665298462,
|
1798 |
+
"lr": 1.95703125e-06,
|
1799 |
+
"objective/entropy": -49.22521209716797,
|
1800 |
+
"objective/kl": 27.571868896484375,
|
1801 |
+
"objective/non_score_reward": -1.3785934448242188,
|
1802 |
+
"objective/rlhf_reward": 8.233023643493652,
|
1803 |
+
"objective/scores": 9.611616134643555,
|
1804 |
+
"policy/approxkl_avg": 0.13268576562404633,
|
1805 |
+
"policy/clipfrac_avg": 0.30859375,
|
1806 |
+
"policy/entropy_avg": 0.35870999097824097,
|
1807 |
+
"step": 89,
|
1808 |
+
"val/clipfrac_avg": 0.0,
|
1809 |
+
"val/num_eos_tokens": 2166,
|
1810 |
+
"val/ratio": 1.000309944152832,
|
1811 |
+
"val/ratio_var": 1.657385723774496e-06
|
1812 |
+
},
|
1813 |
+
{
|
1814 |
+
"episode": 46592,
|
1815 |
+
"epoch": 0.19958875942426318,
|
1816 |
+
"eps": 7,
|
1817 |
+
"loss/policy_avg": 0.14388418197631836,
|
1818 |
+
"lr": 1.9453125e-06,
|
1819 |
+
"objective/entropy": -50.920501708984375,
|
1820 |
+
"objective/kl": 28.16626739501953,
|
1821 |
+
"objective/non_score_reward": -1.408313512802124,
|
1822 |
+
"objective/rlhf_reward": 8.35401725769043,
|
1823 |
+
"objective/scores": 9.762331008911133,
|
1824 |
+
"policy/approxkl_avg": 0.1588677167892456,
|
1825 |
+
"policy/clipfrac_avg": 0.283203125,
|
1826 |
+
"policy/entropy_avg": 0.35217928886413574,
|
1827 |
+
"step": 90,
|
1828 |
+
"val/clipfrac_avg": 0.0,
|
1829 |
+
"val/num_eos_tokens": 2249,
|
1830 |
+
"val/ratio": 0.9999381303787231,
|
1831 |
+
"val/ratio_var": 1.946563543242519e-06
|
1832 |
+
},
|
1833 |
+
{
|
1834 |
+
"episode": 47104,
|
1835 |
+
"epoch": 0.20178204249485948,
|
1836 |
+
"eps": 7,
|
1837 |
+
"loss/policy_avg": 0.10846646875143051,
|
1838 |
+
"lr": 1.93359375e-06,
|
1839 |
+
"objective/entropy": -49.05603790283203,
|
1840 |
+
"objective/kl": 27.825576782226562,
|
1841 |
+
"objective/non_score_reward": -1.3912789821624756,
|
1842 |
+
"objective/rlhf_reward": 8.477624893188477,
|
1843 |
+
"objective/scores": 9.868904113769531,
|
1844 |
+
"policy/approxkl_avg": 0.16012346744537354,
|
1845 |
+
"policy/clipfrac_avg": 0.337890625,
|
1846 |
+
"policy/entropy_avg": 0.354667991399765,
|
1847 |
+
"step": 91,
|
1848 |
+
"val/clipfrac_avg": 0.0,
|
1849 |
+
"val/num_eos_tokens": 2234,
|
1850 |
+
"val/ratio": 0.9999682903289795,
|
1851 |
+
"val/ratio_var": 2.205156761192484e-06
|
1852 |
+
},
|
1853 |
+
{
|
1854 |
+
"episode": 47616,
|
1855 |
+
"epoch": 0.20397532556545578,
|
1856 |
+
"eps": 7,
|
1857 |
+
"loss/policy_avg": 0.1895519196987152,
|
1858 |
+
"lr": 1.921875e-06,
|
1859 |
+
"objective/entropy": -45.23288345336914,
|
1860 |
+
"objective/kl": 29.413528442382812,
|
1861 |
+
"objective/non_score_reward": -1.4706764221191406,
|
1862 |
+
"objective/rlhf_reward": 8.364998817443848,
|
1863 |
+
"objective/scores": 9.835676193237305,
|
1864 |
+
"policy/approxkl_avg": 0.17167441546916962,
|
1865 |
+
"policy/clipfrac_avg": 0.345703125,
|
1866 |
+
"policy/entropy_avg": 0.3293291926383972,
|
1867 |
+
"step": 92,
|
1868 |
+
"val/clipfrac_avg": 0.0,
|
1869 |
+
"val/num_eos_tokens": 2080,
|
1870 |
+
"val/ratio": 1.0002005100250244,
|
1871 |
+
"val/ratio_var": 1.5699610003139242e-06
|
1872 |
+
},
|
1873 |
+
{
|
1874 |
+
"episode": 48128,
|
1875 |
+
"epoch": 0.2061686086360521,
|
1876 |
+
"eps": 7,
|
1877 |
+
"loss/policy_avg": 0.17603826522827148,
|
1878 |
+
"lr": 1.91015625e-06,
|
1879 |
+
"objective/entropy": -46.78435516357422,
|
1880 |
+
"objective/kl": 29.79530906677246,
|
1881 |
+
"objective/non_score_reward": -1.4897654056549072,
|
1882 |
+
"objective/rlhf_reward": 8.564033508300781,
|
1883 |
+
"objective/scores": 10.05379867553711,
|
1884 |
+
"policy/approxkl_avg": 0.1884271502494812,
|
1885 |
+
"policy/clipfrac_avg": 0.32421875,
|
1886 |
+
"policy/entropy_avg": 0.33979788422584534,
|
1887 |
+
"step": 93,
|
1888 |
+
"val/clipfrac_avg": 0.0,
|
1889 |
+
"val/num_eos_tokens": 2086,
|
1890 |
+
"val/ratio": 0.9996660947799683,
|
1891 |
+
"val/ratio_var": 2.4188666429836303e-06
|
1892 |
+
},
|
1893 |
+
{
|
1894 |
+
"episode": 48640,
|
1895 |
+
"epoch": 0.2083618917066484,
|
1896 |
+
"eps": 7,
|
1897 |
+
"loss/policy_avg": 0.20978803932666779,
|
1898 |
+
"lr": 1.8984375e-06,
|
1899 |
+
"objective/entropy": -44.69247055053711,
|
1900 |
+
"objective/kl": 30.15808868408203,
|
1901 |
+
"objective/non_score_reward": -1.5079045295715332,
|
1902 |
+
"objective/rlhf_reward": 8.472569465637207,
|
1903 |
+
"objective/scores": 9.980473518371582,
|
1904 |
+
"policy/approxkl_avg": 0.17732420563697815,
|
1905 |
+
"policy/clipfrac_avg": 0.328125,
|
1906 |
+
"policy/entropy_avg": 0.3413885235786438,
|
1907 |
+
"step": 94,
|
1908 |
+
"val/clipfrac_avg": 0.0,
|
1909 |
+
"val/num_eos_tokens": 2000,
|
1910 |
+
"val/ratio": 1.000209927558899,
|
1911 |
+
"val/ratio_var": 2.5605929749872303e-06
|
1912 |
+
},
|
1913 |
+
{
|
1914 |
+
"episode": 49152,
|
1915 |
+
"epoch": 0.2105551747772447,
|
1916 |
+
"eps": 7,
|
1917 |
+
"loss/policy_avg": 0.18885597586631775,
|
1918 |
+
"lr": 1.8867187500000001e-06,
|
1919 |
+
"objective/entropy": -43.72532653808594,
|
1920 |
+
"objective/kl": 30.476011276245117,
|
1921 |
+
"objective/non_score_reward": -1.5238006114959717,
|
1922 |
+
"objective/rlhf_reward": 8.562983512878418,
|
1923 |
+
"objective/scores": 10.086785316467285,
|
1924 |
+
"policy/approxkl_avg": 0.17724457383155823,
|
1925 |
+
"policy/clipfrac_avg": 0.33984375,
|
1926 |
+
"policy/entropy_avg": 0.3357614576816559,
|
1927 |
+
"step": 95,
|
1928 |
+
"val/clipfrac_avg": 0.0,
|
1929 |
+
"val/num_eos_tokens": 2093,
|
1930 |
+
"val/ratio": 0.9998537302017212,
|
1931 |
+
"val/ratio_var": 3.411851821510936e-06
|
1932 |
+
},
|
1933 |
+
{
|
1934 |
+
"episode": 49664,
|
1935 |
+
"epoch": 0.212748457847841,
|
1936 |
+
"eps": 7,
|
1937 |
+
"loss/policy_avg": 0.12626244127750397,
|
1938 |
+
"lr": 1.875e-06,
|
1939 |
+
"objective/entropy": -42.57123565673828,
|
1940 |
+
"objective/kl": 30.709941864013672,
|
1941 |
+
"objective/non_score_reward": -1.5354971885681152,
|
1942 |
+
"objective/rlhf_reward": 8.317207336425781,
|
1943 |
+
"objective/scores": 9.852704048156738,
|
1944 |
+
"policy/approxkl_avg": 0.1764955222606659,
|
1945 |
+
"policy/clipfrac_avg": 0.3359375,
|
1946 |
+
"policy/entropy_avg": 0.3455975651741028,
|
1947 |
+
"step": 96,
|
1948 |
+
"val/clipfrac_avg": 0.0,
|
1949 |
+
"val/num_eos_tokens": 2008,
|
1950 |
+
"val/ratio": 0.9999105930328369,
|
1951 |
+
"val/ratio_var": 2.7429066449258244e-06
|
1952 |
+
},
|
1953 |
+
{
|
1954 |
+
"episode": 50176,
|
1955 |
+
"epoch": 0.2149417409184373,
|
1956 |
+
"eps": 7,
|
1957 |
+
"loss/policy_avg": 0.22914332151412964,
|
1958 |
+
"lr": 1.86328125e-06,
|
1959 |
+
"objective/entropy": -46.18797302246094,
|
1960 |
+
"objective/kl": 30.78895378112793,
|
1961 |
+
"objective/non_score_reward": -1.5394476652145386,
|
1962 |
+
"objective/rlhf_reward": 8.30815601348877,
|
1963 |
+
"objective/scores": 9.847604751586914,
|
1964 |
+
"policy/approxkl_avg": 0.17377912998199463,
|
1965 |
+
"policy/clipfrac_avg": 0.3125,
|
1966 |
+
"policy/entropy_avg": 0.34196025133132935,
|
1967 |
+
"step": 97,
|
1968 |
+
"val/clipfrac_avg": 0.0,
|
1969 |
+
"val/num_eos_tokens": 1946,
|
1970 |
+
"val/ratio": 1.0004637241363525,
|
1971 |
+
"val/ratio_var": 2.6703025923779933e-06
|
1972 |
+
},
|
1973 |
+
{
|
1974 |
+
"episode": 50688,
|
1975 |
+
"epoch": 0.2171350239890336,
|
1976 |
+
"eps": 7,
|
1977 |
+
"loss/policy_avg": 0.1760575771331787,
|
1978 |
+
"lr": 1.8515625000000001e-06,
|
1979 |
+
"objective/entropy": -45.94915008544922,
|
1980 |
+
"objective/kl": 29.990123748779297,
|
1981 |
+
"objective/non_score_reward": -1.4995063543319702,
|
1982 |
+
"objective/rlhf_reward": 8.321691513061523,
|
1983 |
+
"objective/scores": 9.821197509765625,
|
1984 |
+
"policy/approxkl_avg": 0.18175339698791504,
|
1985 |
+
"policy/clipfrac_avg": 0.3359375,
|
1986 |
+
"policy/entropy_avg": 0.32052457332611084,
|
1987 |
+
"step": 98,
|
1988 |
+
"val/clipfrac_avg": 0.0,
|
1989 |
+
"val/num_eos_tokens": 1913,
|
1990 |
+
"val/ratio": 1.000022053718567,
|
1991 |
+
"val/ratio_var": 1.9145775240758667e-06
|
1992 |
+
},
|
1993 |
+
{
|
1994 |
+
"episode": 51200,
|
1995 |
+
"epoch": 0.21932830705962988,
|
1996 |
+
"eps": 7,
|
1997 |
+
"loss/policy_avg": 0.22435761988162994,
|
1998 |
+
"lr": 1.83984375e-06,
|
1999 |
+
"objective/entropy": -45.11737823486328,
|
2000 |
+
"objective/kl": 29.877592086791992,
|
2001 |
+
"objective/non_score_reward": -1.4938795566558838,
|
2002 |
+
"objective/rlhf_reward": 8.598479270935059,
|
2003 |
+
"objective/scores": 10.09235954284668,
|
2004 |
+
"policy/approxkl_avg": 0.16921335458755493,
|
2005 |
+
"policy/clipfrac_avg": 0.35546875,
|
2006 |
+
"policy/entropy_avg": 0.31710439920425415,
|
2007 |
+
"step": 99,
|
2008 |
+
"val/clipfrac_avg": 0.0,
|
2009 |
+
"val/num_eos_tokens": 2014,
|
2010 |
+
"val/ratio": 0.9999295473098755,
|
2011 |
+
"val/ratio_var": 1.5191413922366337e-06
|
2012 |
+
},
|
2013 |
+
{
|
2014 |
+
"episode": 51712,
|
2015 |
+
"epoch": 0.22152159013022618,
|
2016 |
+
"eps": 7,
|
2017 |
+
"loss/policy_avg": 0.17742179334163666,
|
2018 |
+
"lr": 1.828125e-06,
|
2019 |
+
"objective/entropy": -42.36469650268555,
|
2020 |
+
"objective/kl": 30.714750289916992,
|
2021 |
+
"objective/non_score_reward": -1.5357375144958496,
|
2022 |
+
"objective/rlhf_reward": 8.25560188293457,
|
2023 |
+
"objective/scores": 9.791338920593262,
|
2024 |
+
"policy/approxkl_avg": 0.16882562637329102,
|
2025 |
+
"policy/clipfrac_avg": 0.28125,
|
2026 |
+
"policy/entropy_avg": 0.3051363229751587,
|
2027 |
+
"step": 100,
|
2028 |
+
"val/clipfrac_avg": 0.0,
|
2029 |
+
"val/num_eos_tokens": 1976,
|
2030 |
+
"val/ratio": 0.9998465776443481,
|
2031 |
+
"val/ratio_var": 2.1597036266030045e-06
|
2032 |
+
},
|
2033 |
+
{
|
2034 |
+
"episode": 52224,
|
2035 |
+
"epoch": 0.22371487320082248,
|
2036 |
+
"eps": 7,
|
2037 |
+
"loss/policy_avg": 0.1551419198513031,
|
2038 |
+
"lr": 1.81640625e-06,
|
2039 |
+
"objective/entropy": -45.24191665649414,
|
2040 |
+
"objective/kl": 29.786151885986328,
|
2041 |
+
"objective/non_score_reward": -1.4893076419830322,
|
2042 |
+
"objective/rlhf_reward": 8.873478889465332,
|
2043 |
+
"objective/scores": 10.362787246704102,
|
2044 |
+
"policy/approxkl_avg": 0.1471608281135559,
|
2045 |
+
"policy/clipfrac_avg": 0.3203125,
|
2046 |
+
"policy/entropy_avg": 0.31351402401924133,
|
2047 |
+
"step": 101,
|
2048 |
+
"val/clipfrac_avg": 0.0,
|
2049 |
+
"val/num_eos_tokens": 2097,
|
2050 |
+
"val/ratio": 1.0002268552780151,
|
2051 |
+
"val/ratio_var": 2.6088496269949246e-06
|
2052 |
+
},
|
2053 |
+
{
|
2054 |
+
"episode": 52736,
|
2055 |
+
"epoch": 0.22590815627141878,
|
2056 |
+
"eps": 7,
|
2057 |
+
"loss/policy_avg": 0.1386115550994873,
|
2058 |
+
"lr": 1.8046875e-06,
|
2059 |
+
"objective/entropy": -44.687103271484375,
|
2060 |
+
"objective/kl": 30.421260833740234,
|
2061 |
+
"objective/non_score_reward": -1.5210630893707275,
|
2062 |
+
"objective/rlhf_reward": 8.674066543579102,
|
2063 |
+
"objective/scores": 10.19512939453125,
|
2064 |
+
"policy/approxkl_avg": 0.17369933426380157,
|
2065 |
+
"policy/clipfrac_avg": 0.283203125,
|
2066 |
+
"policy/entropy_avg": 0.31028440594673157,
|
2067 |
+
"step": 102,
|
2068 |
+
"val/clipfrac_avg": 0.0,
|
2069 |
+
"val/num_eos_tokens": 2195,
|
2070 |
+
"val/ratio": 0.9999023675918579,
|
2071 |
+
"val/ratio_var": 1.8804470300892717e-06
|
2072 |
+
},
|
2073 |
+
{
|
2074 |
+
"episode": 53248,
|
2075 |
+
"epoch": 0.22810143934201507,
|
2076 |
+
"eps": 7,
|
2077 |
+
"loss/policy_avg": 0.1150750070810318,
|
2078 |
+
"lr": 1.79296875e-06,
|
2079 |
+
"objective/entropy": -42.66975402832031,
|
2080 |
+
"objective/kl": 30.731414794921875,
|
2081 |
+
"objective/non_score_reward": -1.5365707874298096,
|
2082 |
+
"objective/rlhf_reward": 8.601308822631836,
|
2083 |
+
"objective/scores": 10.137880325317383,
|
2084 |
+
"policy/approxkl_avg": 0.1787860095500946,
|
2085 |
+
"policy/clipfrac_avg": 0.322265625,
|
2086 |
+
"policy/entropy_avg": 0.31559547781944275,
|
2087 |
+
"step": 103,
|
2088 |
+
"val/clipfrac_avg": 0.0,
|
2089 |
+
"val/num_eos_tokens": 2208,
|
2090 |
+
"val/ratio": 0.9999778270721436,
|
2091 |
+
"val/ratio_var": 2.7223902634432307e-06
|
2092 |
+
},
|
2093 |
+
{
|
2094 |
+
"episode": 53760,
|
2095 |
+
"epoch": 0.23029472241261137,
|
2096 |
+
"eps": 7,
|
2097 |
+
"loss/policy_avg": 0.13593605160713196,
|
2098 |
+
"lr": 1.78125e-06,
|
2099 |
+
"objective/entropy": -43.46843719482422,
|
2100 |
+
"objective/kl": 29.380115509033203,
|
2101 |
+
"objective/non_score_reward": -1.4690057039260864,
|
2102 |
+
"objective/rlhf_reward": 8.295011520385742,
|
2103 |
+
"objective/scores": 9.764018058776855,
|
2104 |
+
"policy/approxkl_avg": 0.17159876227378845,
|
2105 |
+
"policy/clipfrac_avg": 0.318359375,
|
2106 |
+
"policy/entropy_avg": 0.31496453285217285,
|
2107 |
+
"step": 104,
|
2108 |
+
"val/clipfrac_avg": 0.0,
|
2109 |
+
"val/num_eos_tokens": 2184,
|
2110 |
+
"val/ratio": 1.0000436305999756,
|
2111 |
+
"val/ratio_var": 2.5784688659769017e-06
|
2112 |
+
},
|
2113 |
+
{
|
2114 |
+
"episode": 54272,
|
2115 |
+
"epoch": 0.23248800548320767,
|
2116 |
+
"eps": 7,
|
2117 |
+
"loss/policy_avg": 0.17001928389072418,
|
2118 |
+
"lr": 1.76953125e-06,
|
2119 |
+
"objective/entropy": -43.86747741699219,
|
2120 |
+
"objective/kl": 29.18187713623047,
|
2121 |
+
"objective/non_score_reward": -1.4590939283370972,
|
2122 |
+
"objective/rlhf_reward": 8.445150375366211,
|
2123 |
+
"objective/scores": 9.904245376586914,
|
2124 |
+
"policy/approxkl_avg": 0.1594773530960083,
|
2125 |
+
"policy/clipfrac_avg": 0.349609375,
|
2126 |
+
"policy/entropy_avg": 0.3245583772659302,
|
2127 |
+
"step": 105,
|
2128 |
+
"val/clipfrac_avg": 0.0,
|
2129 |
+
"val/num_eos_tokens": 2127,
|
2130 |
+
"val/ratio": 1.0002386569976807,
|
2131 |
+
"val/ratio_var": 2.4451144327031216e-06
|
2132 |
+
},
|
2133 |
+
{
|
2134 |
+
"episode": 54784,
|
2135 |
+
"epoch": 0.23468128855380396,
|
2136 |
+
"eps": 7,
|
2137 |
+
"loss/policy_avg": 0.16130343079566956,
|
2138 |
+
"lr": 1.7578125e-06,
|
2139 |
+
"objective/entropy": -41.72479248046875,
|
2140 |
+
"objective/kl": 29.970767974853516,
|
2141 |
+
"objective/non_score_reward": -1.4985383749008179,
|
2142 |
+
"objective/rlhf_reward": 8.395501136779785,
|
2143 |
+
"objective/scores": 9.894039154052734,
|
2144 |
+
"policy/approxkl_avg": 0.1860203593969345,
|
2145 |
+
"policy/clipfrac_avg": 0.328125,
|
2146 |
+
"policy/entropy_avg": 0.29689711332321167,
|
2147 |
+
"step": 106,
|
2148 |
+
"val/clipfrac_avg": 0.0,
|
2149 |
+
"val/num_eos_tokens": 2206,
|
2150 |
+
"val/ratio": 0.9999003410339355,
|
2151 |
+
"val/ratio_var": 1.9462190721242223e-06
|
2152 |
+
},
|
2153 |
+
{
|
2154 |
+
"episode": 55296,
|
2155 |
+
"epoch": 0.2368745716244003,
|
2156 |
+
"eps": 7,
|
2157 |
+
"loss/policy_avg": 0.09807079285383224,
|
2158 |
+
"lr": 1.74609375e-06,
|
2159 |
+
"objective/entropy": -45.3763427734375,
|
2160 |
+
"objective/kl": 29.436267852783203,
|
2161 |
+
"objective/non_score_reward": -1.4718133211135864,
|
2162 |
+
"objective/rlhf_reward": 8.733297348022461,
|
2163 |
+
"objective/scores": 10.205111503601074,
|
2164 |
+
"policy/approxkl_avg": 0.17290550470352173,
|
2165 |
+
"policy/clipfrac_avg": 0.326171875,
|
2166 |
+
"policy/entropy_avg": 0.31182169914245605,
|
2167 |
+
"step": 107,
|
2168 |
+
"val/clipfrac_avg": 0.0,
|
2169 |
+
"val/num_eos_tokens": 2273,
|
2170 |
+
"val/ratio": 0.9998378753662109,
|
2171 |
+
"val/ratio_var": 1.5131832924453192e-06
|
2172 |
+
},
|
2173 |
+
{
|
2174 |
+
"episode": 55808,
|
2175 |
+
"epoch": 0.23906785469499658,
|
2176 |
+
"eps": 7,
|
2177 |
+
"loss/policy_avg": 0.15108568966388702,
|
2178 |
+
"lr": 1.734375e-06,
|
2179 |
+
"objective/entropy": -43.72712707519531,
|
2180 |
+
"objective/kl": 28.699119567871094,
|
2181 |
+
"objective/non_score_reward": -1.4349560737609863,
|
2182 |
+
"objective/rlhf_reward": 8.475278854370117,
|
2183 |
+
"objective/scores": 9.910234451293945,
|
2184 |
+
"policy/approxkl_avg": 0.14781951904296875,
|
2185 |
+
"policy/clipfrac_avg": 0.302734375,
|
2186 |
+
"policy/entropy_avg": 0.3345240652561188,
|
2187 |
+
"step": 108,
|
2188 |
+
"val/clipfrac_avg": 0.0,
|
2189 |
+
"val/num_eos_tokens": 2432,
|
2190 |
+
"val/ratio": 0.9997607469558716,
|
2191 |
+
"val/ratio_var": 2.047207090072334e-06
|
2192 |
+
},
|
2193 |
+
{
|
2194 |
+
"episode": 56320,
|
2195 |
+
"epoch": 0.24126113776559288,
|
2196 |
+
"eps": 7,
|
2197 |
+
"loss/policy_avg": 0.15849685668945312,
|
2198 |
+
"lr": 1.72265625e-06,
|
2199 |
+
"objective/entropy": -46.727020263671875,
|
2200 |
+
"objective/kl": 28.58669662475586,
|
2201 |
+
"objective/non_score_reward": -1.4293348789215088,
|
2202 |
+
"objective/rlhf_reward": 8.42021656036377,
|
2203 |
+
"objective/scores": 9.849552154541016,
|
2204 |
+
"policy/approxkl_avg": 0.13590219616889954,
|
2205 |
+
"policy/clipfrac_avg": 0.3125,
|
2206 |
+
"policy/entropy_avg": 0.3165989816188812,
|
2207 |
+
"step": 109,
|
2208 |
+
"val/clipfrac_avg": 0.0,
|
2209 |
+
"val/num_eos_tokens": 2363,
|
2210 |
+
"val/ratio": 0.9999260306358337,
|
2211 |
+
"val/ratio_var": 2.04604066311731e-06
|
2212 |
+
},
|
2213 |
+
{
|
2214 |
+
"episode": 56832,
|
2215 |
+
"epoch": 0.24345442083618918,
|
2216 |
+
"eps": 7,
|
2217 |
+
"loss/policy_avg": 0.09374918043613434,
|
2218 |
+
"lr": 1.7109375e-06,
|
2219 |
+
"objective/entropy": -47.752201080322266,
|
2220 |
+
"objective/kl": 26.962312698364258,
|
2221 |
+
"objective/non_score_reward": -1.3481156826019287,
|
2222 |
+
"objective/rlhf_reward": 8.627408027648926,
|
2223 |
+
"objective/scores": 9.975523948669434,
|
2224 |
+
"policy/approxkl_avg": 0.15003812313079834,
|
2225 |
+
"policy/clipfrac_avg": 0.3515625,
|
2226 |
+
"policy/entropy_avg": 0.32964205741882324,
|
2227 |
+
"step": 110,
|
2228 |
+
"val/clipfrac_avg": 0.0,
|
2229 |
+
"val/num_eos_tokens": 2352,
|
2230 |
+
"val/ratio": 1.0004621744155884,
|
2231 |
+
"val/ratio_var": 3.1989563922252273e-06
|
2232 |
+
},
|
2233 |
+
{
|
2234 |
+
"episode": 57344,
|
2235 |
+
"epoch": 0.24564770390678548,
|
2236 |
+
"eps": 7,
|
2237 |
+
"loss/policy_avg": 0.25338220596313477,
|
2238 |
+
"lr": 1.69921875e-06,
|
2239 |
+
"objective/entropy": -47.662654876708984,
|
2240 |
+
"objective/kl": 27.328208923339844,
|
2241 |
+
"objective/non_score_reward": -1.366410493850708,
|
2242 |
+
"objective/rlhf_reward": 8.411155700683594,
|
2243 |
+
"objective/scores": 9.777565956115723,
|
2244 |
+
"policy/approxkl_avg": 0.11963716149330139,
|
2245 |
+
"policy/clipfrac_avg": 0.2890625,
|
2246 |
+
"policy/entropy_avg": 0.3197368085384369,
|
2247 |
+
"step": 111,
|
2248 |
+
"val/clipfrac_avg": 0.0,
|
2249 |
+
"val/num_eos_tokens": 2252,
|
2250 |
+
"val/ratio": 1.0002801418304443,
|
2251 |
+
"val/ratio_var": 1.0876318583541433e-06
|
2252 |
+
},
|
2253 |
+
{
|
2254 |
+
"episode": 57856,
|
2255 |
+
"epoch": 0.24784098697738177,
|
2256 |
+
"eps": 7,
|
2257 |
+
"loss/policy_avg": 0.16557908058166504,
|
2258 |
+
"lr": 1.6875e-06,
|
2259 |
+
"objective/entropy": -48.649131774902344,
|
2260 |
+
"objective/kl": 26.810922622680664,
|
2261 |
+
"objective/non_score_reward": -1.3405461311340332,
|
2262 |
+
"objective/rlhf_reward": 8.496389389038086,
|
2263 |
+
"objective/scores": 9.836935043334961,
|
2264 |
+
"policy/approxkl_avg": 0.14137035608291626,
|
2265 |
+
"policy/clipfrac_avg": 0.314453125,
|
2266 |
+
"policy/entropy_avg": 0.3187577724456787,
|
2267 |
+
"step": 112,
|
2268 |
+
"val/clipfrac_avg": 0.0,
|
2269 |
+
"val/num_eos_tokens": 2431,
|
2270 |
+
"val/ratio": 0.9999630451202393,
|
2271 |
+
"val/ratio_var": 2.2570845885638846e-06
|
2272 |
+
},
|
2273 |
+
{
|
2274 |
+
"episode": 58368,
|
2275 |
+
"epoch": 0.25003427004797807,
|
2276 |
+
"eps": 7,
|
2277 |
+
"loss/policy_avg": 0.06345733255147934,
|
2278 |
+
"lr": 1.67578125e-06,
|
2279 |
+
"objective/entropy": -49.035865783691406,
|
2280 |
+
"objective/kl": 26.79833221435547,
|
2281 |
+
"objective/non_score_reward": -1.339916706085205,
|
2282 |
+
"objective/rlhf_reward": 8.691230773925781,
|
2283 |
+
"objective/scores": 10.031147003173828,
|
2284 |
+
"policy/approxkl_avg": 0.11882881820201874,
|
2285 |
+
"policy/clipfrac_avg": 0.318359375,
|
2286 |
+
"policy/entropy_avg": 0.3099344074726105,
|
2287 |
+
"step": 113,
|
2288 |
+
"val/clipfrac_avg": 0.0,
|
2289 |
+
"val/num_eos_tokens": 2329,
|
2290 |
+
"val/ratio": 1.000173568725586,
|
2291 |
+
"val/ratio_var": 1.6486970935147838e-06
|
2292 |
+
},
|
2293 |
+
{
|
2294 |
+
"episode": 58880,
|
2295 |
+
"epoch": 0.25222755311857437,
|
2296 |
+
"eps": 7,
|
2297 |
+
"loss/policy_avg": 0.15583859384059906,
|
2298 |
+
"lr": 1.6640625e-06,
|
2299 |
+
"objective/entropy": -46.74812316894531,
|
2300 |
+
"objective/kl": 27.792320251464844,
|
2301 |
+
"objective/non_score_reward": -1.3896160125732422,
|
2302 |
+
"objective/rlhf_reward": 8.57948112487793,
|
2303 |
+
"objective/scores": 9.969097137451172,
|
2304 |
+
"policy/approxkl_avg": 0.17242445051670074,
|
2305 |
+
"policy/clipfrac_avg": 0.328125,
|
2306 |
+
"policy/entropy_avg": 0.33402085304260254,
|
2307 |
+
"step": 114,
|
2308 |
+
"val/clipfrac_avg": 0.0,
|
2309 |
+
"val/num_eos_tokens": 2459,
|
2310 |
+
"val/ratio": 1.0000650882720947,
|
2311 |
+
"val/ratio_var": 1.9053392179557704e-06
|
2312 |
+
},
|
2313 |
+
{
|
2314 |
+
"episode": 59392,
|
2315 |
+
"epoch": 0.25442083618917066,
|
2316 |
+
"eps": 7,
|
2317 |
+
"loss/policy_avg": 0.12703779339790344,
|
2318 |
+
"lr": 1.6523437500000001e-06,
|
2319 |
+
"objective/entropy": -48.9515495300293,
|
2320 |
+
"objective/kl": 27.333660125732422,
|
2321 |
+
"objective/non_score_reward": -1.366683006286621,
|
2322 |
+
"objective/rlhf_reward": 8.453899383544922,
|
2323 |
+
"objective/scores": 9.820582389831543,
|
2324 |
+
"policy/approxkl_avg": 0.15347963571548462,
|
2325 |
+
"policy/clipfrac_avg": 0.3671875,
|
2326 |
+
"policy/entropy_avg": 0.3327234983444214,
|
2327 |
+
"step": 115,
|
2328 |
+
"val/clipfrac_avg": 0.0,
|
2329 |
+
"val/num_eos_tokens": 2632,
|
2330 |
+
"val/ratio": 1.0000560283660889,
|
2331 |
+
"val/ratio_var": 2.397124944764073e-06
|
2332 |
+
},
|
2333 |
+
{
|
2334 |
+
"episode": 59904,
|
2335 |
+
"epoch": 0.25661411925976696,
|
2336 |
+
"eps": 7,
|
2337 |
+
"loss/policy_avg": 0.11752994358539581,
|
2338 |
+
"lr": 1.640625e-06,
|
2339 |
+
"objective/entropy": -48.20651626586914,
|
2340 |
+
"objective/kl": 27.3616943359375,
|
2341 |
+
"objective/non_score_reward": -1.3680846691131592,
|
2342 |
+
"objective/rlhf_reward": 8.65481948852539,
|
2343 |
+
"objective/scores": 10.022903442382812,
|
2344 |
+
"policy/approxkl_avg": 0.17291083931922913,
|
2345 |
+
"policy/clipfrac_avg": 0.359375,
|
2346 |
+
"policy/entropy_avg": 0.31415560841560364,
|
2347 |
+
"step": 116,
|
2348 |
+
"val/clipfrac_avg": 0.0,
|
2349 |
+
"val/num_eos_tokens": 2394,
|
2350 |
+
"val/ratio": 0.9997780323028564,
|
2351 |
+
"val/ratio_var": 2.4493372166034533e-06
|
2352 |
+
},
|
2353 |
+
{
|
2354 |
+
"episode": 60416,
|
2355 |
+
"epoch": 0.25880740233036326,
|
2356 |
+
"eps": 7,
|
2357 |
+
"loss/policy_avg": 0.11132411658763885,
|
2358 |
+
"lr": 1.62890625e-06,
|
2359 |
+
"objective/entropy": -48.8924560546875,
|
2360 |
+
"objective/kl": 27.538108825683594,
|
2361 |
+
"objective/non_score_reward": -1.3769053220748901,
|
2362 |
+
"objective/rlhf_reward": 8.483039855957031,
|
2363 |
+
"objective/scores": 9.859945297241211,
|
2364 |
+
"policy/approxkl_avg": 0.13603341579437256,
|
2365 |
+
"policy/clipfrac_avg": 0.291015625,
|
2366 |
+
"policy/entropy_avg": 0.31023937463760376,
|
2367 |
+
"step": 117,
|
2368 |
+
"val/clipfrac_avg": 0.0,
|
2369 |
+
"val/num_eos_tokens": 2411,
|
2370 |
+
"val/ratio": 1.000200629234314,
|
2371 |
+
"val/ratio_var": 1.9740023162739817e-06
|
2372 |
+
},
|
2373 |
+
{
|
2374 |
+
"episode": 60928,
|
2375 |
+
"epoch": 0.26100068540095955,
|
2376 |
+
"eps": 7,
|
2377 |
+
"loss/policy_avg": 0.11453983187675476,
|
2378 |
+
"lr": 1.6171875000000001e-06,
|
2379 |
+
"objective/entropy": -49.235565185546875,
|
2380 |
+
"objective/kl": 27.43050765991211,
|
2381 |
+
"objective/non_score_reward": -1.3715254068374634,
|
2382 |
+
"objective/rlhf_reward": 8.758109092712402,
|
2383 |
+
"objective/scores": 10.129634857177734,
|
2384 |
+
"policy/approxkl_avg": 0.14746010303497314,
|
2385 |
+
"policy/clipfrac_avg": 0.302734375,
|
2386 |
+
"policy/entropy_avg": 0.3134969174861908,
|
2387 |
+
"step": 118,
|
2388 |
+
"val/clipfrac_avg": 0.0,
|
2389 |
+
"val/num_eos_tokens": 2285,
|
2390 |
+
"val/ratio": 1.0001026391983032,
|
2391 |
+
"val/ratio_var": 2.20556512431358e-06
|
2392 |
+
},
|
2393 |
+
{
|
2394 |
+
"episode": 61440,
|
2395 |
+
"epoch": 0.26319396847155585,
|
2396 |
+
"eps": 7,
|
2397 |
+
"loss/policy_avg": 0.10771258175373077,
|
2398 |
+
"lr": 1.60546875e-06,
|
2399 |
+
"objective/entropy": -48.77255630493164,
|
2400 |
+
"objective/kl": 27.55831527709961,
|
2401 |
+
"objective/non_score_reward": -1.377915859222412,
|
2402 |
+
"objective/rlhf_reward": 8.477496147155762,
|
2403 |
+
"objective/scores": 9.855412483215332,
|
2404 |
+
"policy/approxkl_avg": 0.12517622113227844,
|
2405 |
+
"policy/clipfrac_avg": 0.361328125,
|
2406 |
+
"policy/entropy_avg": 0.328954815864563,
|
2407 |
+
"step": 119,
|
2408 |
+
"val/clipfrac_avg": 0.0,
|
2409 |
+
"val/num_eos_tokens": 2477,
|
2410 |
+
"val/ratio": 0.9999545812606812,
|
2411 |
+
"val/ratio_var": 1.424527908966411e-06
|
2412 |
+
},
|
2413 |
+
{
|
2414 |
+
"episode": 61952,
|
2415 |
+
"epoch": 0.26538725154215215,
|
2416 |
+
"eps": 7,
|
2417 |
+
"loss/policy_avg": 0.11551685631275177,
|
2418 |
+
"lr": 1.59375e-06,
|
2419 |
+
"objective/entropy": -48.61798095703125,
|
2420 |
+
"objective/kl": 27.41283416748047,
|
2421 |
+
"objective/non_score_reward": -1.3706417083740234,
|
2422 |
+
"objective/rlhf_reward": 8.553994178771973,
|
2423 |
+
"objective/scores": 9.924636840820312,
|
2424 |
+
"policy/approxkl_avg": 0.1658802330493927,
|
2425 |
+
"policy/clipfrac_avg": 0.322265625,
|
2426 |
+
"policy/entropy_avg": 0.3142884373664856,
|
2427 |
+
"step": 120,
|
2428 |
+
"val/clipfrac_avg": 0.0,
|
2429 |
+
"val/num_eos_tokens": 2462,
|
2430 |
+
"val/ratio": 1.000496506690979,
|
2431 |
+
"val/ratio_var": 3.072740128118312e-06
|
2432 |
+
},
|
2433 |
+
{
|
2434 |
+
"episode": 62464,
|
2435 |
+
"epoch": 0.26758053461274844,
|
2436 |
+
"eps": 7,
|
2437 |
+
"loss/policy_avg": 0.10897368937730789,
|
2438 |
+
"lr": 1.5820312500000001e-06,
|
2439 |
+
"objective/entropy": -50.19673538208008,
|
2440 |
+
"objective/kl": 27.115222930908203,
|
2441 |
+
"objective/non_score_reward": -1.355761170387268,
|
2442 |
+
"objective/rlhf_reward": 8.635272979736328,
|
2443 |
+
"objective/scores": 9.991033554077148,
|
2444 |
+
"policy/approxkl_avg": 0.14301660656929016,
|
2445 |
+
"policy/clipfrac_avg": 0.33203125,
|
2446 |
+
"policy/entropy_avg": 0.3095218241214752,
|
2447 |
+
"step": 121,
|
2448 |
+
"val/clipfrac_avg": 0.0,
|
2449 |
+
"val/num_eos_tokens": 2456,
|
2450 |
+
"val/ratio": 0.9999917149543762,
|
2451 |
+
"val/ratio_var": 2.4075486635410925e-06
|
2452 |
+
},
|
2453 |
+
{
|
2454 |
+
"episode": 62976,
|
2455 |
+
"epoch": 0.26977381768334474,
|
2456 |
+
"eps": 7,
|
2457 |
+
"loss/policy_avg": 0.14559023082256317,
|
2458 |
+
"lr": 1.5703125e-06,
|
2459 |
+
"objective/entropy": -48.252967834472656,
|
2460 |
+
"objective/kl": 27.653305053710938,
|
2461 |
+
"objective/non_score_reward": -1.3826653957366943,
|
2462 |
+
"objective/rlhf_reward": 8.649492263793945,
|
2463 |
+
"objective/scores": 10.032157897949219,
|
2464 |
+
"policy/approxkl_avg": 0.1397351324558258,
|
2465 |
+
"policy/clipfrac_avg": 0.298828125,
|
2466 |
+
"policy/entropy_avg": 0.3301985263824463,
|
2467 |
+
"step": 122,
|
2468 |
+
"val/clipfrac_avg": 0.0,
|
2469 |
+
"val/num_eos_tokens": 2471,
|
2470 |
+
"val/ratio": 1.0004826784133911,
|
2471 |
+
"val/ratio_var": 2.057088977380772e-06
|
2472 |
+
},
|
2473 |
+
{
|
2474 |
+
"episode": 63488,
|
2475 |
+
"epoch": 0.27196710075394104,
|
2476 |
+
"eps": 7,
|
2477 |
+
"loss/policy_avg": 0.10405009239912033,
|
2478 |
+
"lr": 1.55859375e-06,
|
2479 |
+
"objective/entropy": -46.9781494140625,
|
2480 |
+
"objective/kl": 28.512248992919922,
|
2481 |
+
"objective/non_score_reward": -1.4256125688552856,
|
2482 |
+
"objective/rlhf_reward": 8.570206642150879,
|
2483 |
+
"objective/scores": 9.995819091796875,
|
2484 |
+
"policy/approxkl_avg": 0.1661127209663391,
|
2485 |
+
"policy/clipfrac_avg": 0.365234375,
|
2486 |
+
"policy/entropy_avg": 0.30900561809539795,
|
2487 |
+
"step": 123,
|
2488 |
+
"val/clipfrac_avg": 0.0,
|
2489 |
+
"val/num_eos_tokens": 2323,
|
2490 |
+
"val/ratio": 1.000049352645874,
|
2491 |
+
"val/ratio_var": 2.1198638933128677e-06
|
2492 |
+
},
|
2493 |
+
{
|
2494 |
+
"episode": 64000,
|
2495 |
+
"epoch": 0.27416038382453733,
|
2496 |
+
"eps": 7,
|
2497 |
+
"loss/policy_avg": 0.11964317411184311,
|
2498 |
+
"lr": 1.5468750000000001e-06,
|
2499 |
+
"objective/entropy": -46.61250305175781,
|
2500 |
+
"objective/kl": 28.343521118164062,
|
2501 |
+
"objective/non_score_reward": -1.4171760082244873,
|
2502 |
+
"objective/rlhf_reward": 8.400045394897461,
|
2503 |
+
"objective/scores": 9.817220687866211,
|
2504 |
+
"policy/approxkl_avg": 0.16542762517929077,
|
2505 |
+
"policy/clipfrac_avg": 0.337890625,
|
2506 |
+
"policy/entropy_avg": 0.31432512402534485,
|
2507 |
+
"step": 124,
|
2508 |
+
"val/clipfrac_avg": 0.0,
|
2509 |
+
"val/num_eos_tokens": 2390,
|
2510 |
+
"val/ratio": 0.9999516606330872,
|
2511 |
+
"val/ratio_var": 2.199320988438558e-06
|
2512 |
+
},
|
2513 |
+
{
|
2514 |
+
"episode": 64512,
|
2515 |
+
"epoch": 0.27635366689513363,
|
2516 |
+
"eps": 7,
|
2517 |
+
"loss/policy_avg": 0.0921807587146759,
|
2518 |
+
"lr": 1.53515625e-06,
|
2519 |
+
"objective/entropy": -50.23047637939453,
|
2520 |
+
"objective/kl": 27.537601470947266,
|
2521 |
+
"objective/non_score_reward": -1.376880168914795,
|
2522 |
+
"objective/rlhf_reward": 8.707250595092773,
|
2523 |
+
"objective/scores": 10.084131240844727,
|
2524 |
+
"policy/approxkl_avg": 0.15336409211158752,
|
2525 |
+
"policy/clipfrac_avg": 0.259765625,
|
2526 |
+
"policy/entropy_avg": 0.30232733488082886,
|
2527 |
+
"step": 125,
|
2528 |
+
"val/clipfrac_avg": 0.0,
|
2529 |
+
"val/num_eos_tokens": 2417,
|
2530 |
+
"val/ratio": 0.999809205532074,
|
2531 |
+
"val/ratio_var": 1.704518240330799e-06
|
2532 |
+
},
|
2533 |
+
{
|
2534 |
+
"episode": 65024,
|
2535 |
+
"epoch": 0.27854694996572993,
|
2536 |
+
"eps": 7,
|
2537 |
+
"loss/policy_avg": 0.20316940546035767,
|
2538 |
+
"lr": 1.5234375e-06,
|
2539 |
+
"objective/entropy": -48.22269058227539,
|
2540 |
+
"objective/kl": 27.629222869873047,
|
2541 |
+
"objective/non_score_reward": -1.3814611434936523,
|
2542 |
+
"objective/rlhf_reward": 8.414966583251953,
|
2543 |
+
"objective/scores": 9.796428680419922,
|
2544 |
+
"policy/approxkl_avg": 0.13752874732017517,
|
2545 |
+
"policy/clipfrac_avg": 0.3046875,
|
2546 |
+
"policy/entropy_avg": 0.3049372434616089,
|
2547 |
+
"step": 126,
|
2548 |
+
"val/clipfrac_avg": 0.0,
|
2549 |
+
"val/num_eos_tokens": 2327,
|
2550 |
+
"val/ratio": 0.9998852014541626,
|
2551 |
+
"val/ratio_var": 1.6696317288733553e-06
|
2552 |
+
},
|
2553 |
+
{
|
2554 |
+
"episode": 65536,
|
2555 |
+
"epoch": 0.2807402330363262,
|
2556 |
+
"eps": 7,
|
2557 |
+
"loss/policy_avg": 0.13236911594867706,
|
2558 |
+
"lr": 1.5117187500000001e-06,
|
2559 |
+
"objective/entropy": -46.95719528198242,
|
2560 |
+
"objective/kl": 27.658584594726562,
|
2561 |
+
"objective/non_score_reward": -1.3829292058944702,
|
2562 |
+
"objective/rlhf_reward": 8.654300689697266,
|
2563 |
+
"objective/scores": 10.03722858428955,
|
2564 |
+
"policy/approxkl_avg": 0.15404567122459412,
|
2565 |
+
"policy/clipfrac_avg": 0.28515625,
|
2566 |
+
"policy/entropy_avg": 0.29800164699554443,
|
2567 |
+
"step": 127,
|
2568 |
+
"val/clipfrac_avg": 0.0,
|
2569 |
+
"val/num_eos_tokens": 2225,
|
2570 |
+
"val/ratio": 1.0000332593917847,
|
2571 |
+
"val/ratio_var": 1.8714601992542157e-06
|
2572 |
+
},
|
2573 |
+
{
|
2574 |
+
"episode": 66048,
|
2575 |
+
"epoch": 0.2829335161069225,
|
2576 |
+
"eps": 7,
|
2577 |
+
"loss/policy_avg": 0.10175906121730804,
|
2578 |
+
"lr": 1.5e-06,
|
2579 |
+
"objective/entropy": -48.04456329345703,
|
2580 |
+
"objective/kl": 27.31212615966797,
|
2581 |
+
"objective/non_score_reward": -1.365606427192688,
|
2582 |
+
"objective/rlhf_reward": 8.639174461364746,
|
2583 |
+
"objective/scores": 10.004779815673828,
|
2584 |
+
"policy/approxkl_avg": 0.1497514545917511,
|
2585 |
+
"policy/clipfrac_avg": 0.33984375,
|
2586 |
+
"policy/entropy_avg": 0.30890870094299316,
|
2587 |
+
"step": 128,
|
2588 |
+
"val/clipfrac_avg": 0.0,
|
2589 |
+
"val/num_eos_tokens": 2256,
|
2590 |
+
"val/ratio": 1.0001022815704346,
|
2591 |
+
"val/ratio_var": 2.219981297457707e-06
|
2592 |
+
},
|
2593 |
+
{
|
2594 |
+
"episode": 66560,
|
2595 |
+
"epoch": 0.2851267991775189,
|
2596 |
+
"eps": 7,
|
2597 |
+
"loss/policy_avg": 0.13455700874328613,
|
2598 |
+
"lr": 1.48828125e-06,
|
2599 |
+
"objective/entropy": -46.605934143066406,
|
2600 |
+
"objective/kl": 28.286909103393555,
|
2601 |
+
"objective/non_score_reward": -1.4143455028533936,
|
2602 |
+
"objective/rlhf_reward": 8.570450782775879,
|
2603 |
+
"objective/scores": 9.984796524047852,
|
2604 |
+
"policy/approxkl_avg": 0.1603183150291443,
|
2605 |
+
"policy/clipfrac_avg": 0.296875,
|
2606 |
+
"policy/entropy_avg": 0.3032073676586151,
|
2607 |
+
"step": 129,
|
2608 |
+
"val/clipfrac_avg": 0.0,
|
2609 |
+
"val/num_eos_tokens": 2338,
|
2610 |
+
"val/ratio": 0.9999357461929321,
|
2611 |
+
"val/ratio_var": 2.086506810883293e-06
|
2612 |
+
},
|
2613 |
+
{
|
2614 |
+
"episode": 67072,
|
2615 |
+
"epoch": 0.28732008224811517,
|
2616 |
+
"eps": 7,
|
2617 |
+
"loss/policy_avg": 0.15740051865577698,
|
2618 |
+
"lr": 1.4765625e-06,
|
2619 |
+
"objective/entropy": -48.83148956298828,
|
2620 |
+
"objective/kl": 27.573869705200195,
|
2621 |
+
"objective/non_score_reward": -1.3786934614181519,
|
2622 |
+
"objective/rlhf_reward": 8.459827423095703,
|
2623 |
+
"objective/scores": 9.838521957397461,
|
2624 |
+
"policy/approxkl_avg": 0.1607588231563568,
|
2625 |
+
"policy/clipfrac_avg": 0.314453125,
|
2626 |
+
"policy/entropy_avg": 0.3094344735145569,
|
2627 |
+
"step": 130,
|
2628 |
+
"val/clipfrac_avg": 0.0,
|
2629 |
+
"val/num_eos_tokens": 2403,
|
2630 |
+
"val/ratio": 0.9996756911277771,
|
2631 |
+
"val/ratio_var": 2.3813338430045405e-06
|
2632 |
+
},
|
2633 |
+
{
|
2634 |
+
"episode": 67584,
|
2635 |
+
"epoch": 0.28951336531871147,
|
2636 |
+
"eps": 7,
|
2637 |
+
"loss/policy_avg": 0.17430683970451355,
|
2638 |
+
"lr": 1.46484375e-06,
|
2639 |
+
"objective/entropy": -47.313819885253906,
|
2640 |
+
"objective/kl": 27.770536422729492,
|
2641 |
+
"objective/non_score_reward": -1.3885267972946167,
|
2642 |
+
"objective/rlhf_reward": 8.759931564331055,
|
2643 |
+
"objective/scores": 10.148458480834961,
|
2644 |
+
"policy/approxkl_avg": 0.18929040431976318,
|
2645 |
+
"policy/clipfrac_avg": 0.345703125,
|
2646 |
+
"policy/entropy_avg": 0.3068851828575134,
|
2647 |
+
"step": 131,
|
2648 |
+
"val/clipfrac_avg": 0.0,
|
2649 |
+
"val/num_eos_tokens": 2399,
|
2650 |
+
"val/ratio": 1.0000063180923462,
|
2651 |
+
"val/ratio_var": 3.177356802552822e-06
|
2652 |
+
},
|
2653 |
+
{
|
2654 |
+
"episode": 68096,
|
2655 |
+
"epoch": 0.29170664838930777,
|
2656 |
+
"eps": 7,
|
2657 |
+
"loss/policy_avg": 0.13198316097259521,
|
2658 |
+
"lr": 1.453125e-06,
|
2659 |
+
"objective/entropy": -45.62812042236328,
|
2660 |
+
"objective/kl": 28.616018295288086,
|
2661 |
+
"objective/non_score_reward": -1.4308009147644043,
|
2662 |
+
"objective/rlhf_reward": 8.519526481628418,
|
2663 |
+
"objective/scores": 9.950326919555664,
|
2664 |
+
"policy/approxkl_avg": 0.17275509238243103,
|
2665 |
+
"policy/clipfrac_avg": 0.31640625,
|
2666 |
+
"policy/entropy_avg": 0.3065592646598816,
|
2667 |
+
"step": 132,
|
2668 |
+
"val/clipfrac_avg": 0.0,
|
2669 |
+
"val/num_eos_tokens": 2372,
|
2670 |
+
"val/ratio": 0.9997407793998718,
|
2671 |
+
"val/ratio_var": 2.1874118374398677e-06
|
2672 |
+
},
|
2673 |
+
{
|
2674 |
+
"episode": 68608,
|
2675 |
+
"epoch": 0.29389993145990406,
|
2676 |
+
"eps": 6,
|
2677 |
+
"loss/policy_avg": 0.09811846166849136,
|
2678 |
+
"lr": 1.44140625e-06,
|
2679 |
+
"objective/entropy": -48.6571044921875,
|
2680 |
+
"objective/kl": 28.160383224487305,
|
2681 |
+
"objective/non_score_reward": -1.4080193042755127,
|
2682 |
+
"objective/rlhf_reward": 8.600898742675781,
|
2683 |
+
"objective/scores": 10.008918762207031,
|
2684 |
+
"policy/approxkl_avg": 0.1812044382095337,
|
2685 |
+
"policy/clipfrac_avg": 0.310546875,
|
2686 |
+
"policy/entropy_avg": 0.2906162738800049,
|
2687 |
+
"step": 133,
|
2688 |
+
"val/clipfrac_avg": 0.0,
|
2689 |
+
"val/num_eos_tokens": 2347,
|
2690 |
+
"val/ratio": 0.9997673630714417,
|
2691 |
+
"val/ratio_var": 3.1602842227584915e-06
|
2692 |
+
},
|
2693 |
+
{
|
2694 |
+
"episode": 69120,
|
2695 |
+
"epoch": 0.29609321453050036,
|
2696 |
+
"eps": 6,
|
2697 |
+
"loss/policy_avg": 0.08539767563343048,
|
2698 |
+
"lr": 1.4296875e-06,
|
2699 |
+
"objective/entropy": -48.618141174316406,
|
2700 |
+
"objective/kl": 28.692398071289062,
|
2701 |
+
"objective/non_score_reward": -1.4346199035644531,
|
2702 |
+
"objective/rlhf_reward": 8.672279357910156,
|
2703 |
+
"objective/scores": 10.10689926147461,
|
2704 |
+
"policy/approxkl_avg": 0.1650359034538269,
|
2705 |
+
"policy/clipfrac_avg": 0.35546875,
|
2706 |
+
"policy/entropy_avg": 0.293681263923645,
|
2707 |
+
"step": 134,
|
2708 |
+
"val/clipfrac_avg": 0.0,
|
2709 |
+
"val/num_eos_tokens": 2265,
|
2710 |
+
"val/ratio": 1.0002015829086304,
|
2711 |
+
"val/ratio_var": 2.1058724541944684e-06
|
2712 |
+
}
|
2713 |
+
],
|
2714 |
+
"logging_steps": 100,
|
2715 |
+
"max_steps": 128,
|
2716 |
+
"num_input_tokens_seen": 0,
|
2717 |
+
"num_train_epochs": 1.122960932145305,
|
2718 |
+
"save_steps": 26,
|
2719 |
+
"stateful_callbacks": {
|
2720 |
+
"TrainerControl": {
|
2721 |
+
"args": {
|
2722 |
+
"should_epoch_stop": false,
|
2723 |
+
"should_evaluate": false,
|
2724 |
+
"should_log": false,
|
2725 |
+
"should_save": true,
|
2726 |
+
"should_training_stop": true
|
2727 |
+
},
|
2728 |
+
"attributes": {}
|
2729 |
+
}
|
2730 |
+
},
|
2731 |
+
"total_flos": 0,
|
2732 |
+
"train_batch_size": null,
|
2733 |
+
"trial_name": null,
|
2734 |
+
"trial_params": null
|
2735 |
+
}
|
checkpoint-135/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:6177cda3990adf01bb87fb26103e192998a74ab964915c191455ea7217f4e472
|
3 |
+
size 6584
|