yiran-wang3 commited on
Commit
d9f44d5
1 Parent(s): 809c89f

End of training

Browse files
README.md ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ license: other
4
+ base_model: deepseek-ai/deepseek-llm-7b-chat
5
+ tags:
6
+ - alignment-handbook
7
+ - generated_from_trainer
8
+ - trl
9
+ - dpo
10
+ datasets:
11
+ - reflection-gen/ds_chat_original_cn_mining_oj_iter0-binarized
12
+ model-index:
13
+ - name: sigmoidds_chat_rmsprop_iter1
14
+ results: []
15
+ ---
16
+
17
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
18
+ should probably proofread and complete it, then remove this comment. -->
19
+
20
+ # sigmoidds_chat_rmsprop_iter1
21
+
22
+ This model is a fine-tuned version of [deepseek-ai/deepseek-llm-7b-chat](https://huggingface.co/deepseek-ai/deepseek-llm-7b-chat) on the reflection-gen/ds_chat_original_cn_mining_oj_iter0-binarized dataset.
23
+
24
+ ## Model description
25
+
26
+ More information needed
27
+
28
+ ## Intended uses & limitations
29
+
30
+ More information needed
31
+
32
+ ## Training and evaluation data
33
+
34
+ More information needed
35
+
36
+ ## Training procedure
37
+
38
+ ### Training hyperparameters
39
+
40
+ The following hyperparameters were used during training:
41
+ - learning_rate: 5e-07
42
+ - train_batch_size: 8
43
+ - eval_batch_size: 4
44
+ - seed: 42
45
+ - distributed_type: multi-GPU
46
+ - num_devices: 8
47
+ - total_train_batch_size: 64
48
+ - total_eval_batch_size: 32
49
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
50
+ - lr_scheduler_type: constant
51
+ - lr_scheduler_warmup_ratio: 0.1
52
+ - lr_scheduler_warmup_steps: 100
53
+ - num_epochs: 1.0
54
+
55
+ ### Training results
56
+
57
+
58
+
59
+ ### Framework versions
60
+
61
+ - Transformers 4.45.0
62
+ - Pytorch 2.4.0+cu121
63
+ - Datasets 2.14.6
64
+ - Tokenizers 0.20.3
all_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.6775240507992831,
5
+ "train_runtime": 440.7264,
6
+ "train_samples": 3485,
7
+ "train_samples_per_second": 7.907,
8
+ "train_steps_per_second": 0.125
9
+ }
config.json CHANGED
@@ -25,6 +25,6 @@
25
  "tie_word_embeddings": false,
26
  "torch_dtype": "bfloat16",
27
  "transformers_version": "4.45.0",
28
- "use_cache": false,
29
  "vocab_size": 102400
30
  }
 
25
  "tie_word_embeddings": false,
26
  "torch_dtype": "bfloat16",
27
  "transformers_version": "4.45.0",
28
+ "use_cache": true,
29
  "vocab_size": 102400
30
  }
generation_config.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 100000,
4
+ "do_sample": true,
5
+ "eos_token_id": 100001,
6
+ "temperature": 0.7,
7
+ "top_p": 0.95,
8
+ "transformers_version": "4.45.0"
9
+ }
train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.6775240507992831,
5
+ "train_runtime": 440.7264,
6
+ "train_samples": 3485,
7
+ "train_samples_per_second": 7.907,
8
+ "train_steps_per_second": 0.125
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,1197 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 100,
6
+ "global_step": 55,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "debug/policy_chosen_logits": 1.3768084049224854,
13
+ "debug/policy_chosen_logps": -161.25889587402344,
14
+ "debug/policy_rejected_logits": 1.3094204664230347,
15
+ "debug/policy_rejected_logps": -151.26834106445312,
16
+ "debug/reference_chosen_logps": -161.25889587402344,
17
+ "debug/reference_rejected_logps": -151.26834106445312,
18
+ "epoch": 0.01818181818181818,
19
+ "grad_norm": 2.5769607314165515,
20
+ "learning_rate": 5e-07,
21
+ "logits/chosen": 1.3768084049224854,
22
+ "logits/rejected": 1.3094204664230347,
23
+ "logps/chosen": -161.25889587402344,
24
+ "logps/rejected": -151.26834106445312,
25
+ "loss": 0.6931,
26
+ "rewards/accuracies": 0.0,
27
+ "rewards/chosen": 0.0,
28
+ "rewards/margins": 0.0,
29
+ "rewards/rejected": 0.0,
30
+ "step": 1
31
+ },
32
+ {
33
+ "debug/policy_chosen_logits": 0.889292299747467,
34
+ "debug/policy_chosen_logps": -130.50076293945312,
35
+ "debug/policy_rejected_logits": 1.1603094339370728,
36
+ "debug/policy_rejected_logps": -134.81027221679688,
37
+ "debug/reference_chosen_logps": -130.64517211914062,
38
+ "debug/reference_rejected_logps": -135.13914489746094,
39
+ "epoch": 0.03636363636363636,
40
+ "grad_norm": 2.472634188308473,
41
+ "learning_rate": 5e-07,
42
+ "logits/chosen": 0.889292299747467,
43
+ "logits/rejected": 1.1603094339370728,
44
+ "logps/chosen": -130.50076293945312,
45
+ "logps/rejected": -134.81027221679688,
46
+ "loss": 0.6925,
47
+ "rewards/accuracies": 0.5,
48
+ "rewards/chosen": 0.0014441299717873335,
49
+ "rewards/margins": -0.0018445965833961964,
50
+ "rewards/rejected": 0.0032887267880141735,
51
+ "step": 2
52
+ },
53
+ {
54
+ "debug/policy_chosen_logits": 0.6971374154090881,
55
+ "debug/policy_chosen_logps": -137.5977783203125,
56
+ "debug/policy_rejected_logits": 1.2003397941589355,
57
+ "debug/policy_rejected_logps": -122.20698547363281,
58
+ "debug/reference_chosen_logps": -138.00997924804688,
59
+ "debug/reference_rejected_logps": -122.32334899902344,
60
+ "epoch": 0.05454545454545454,
61
+ "grad_norm": 2.960555045135284,
62
+ "learning_rate": 5e-07,
63
+ "logits/chosen": 0.6971374154090881,
64
+ "logits/rejected": 1.2003397941589355,
65
+ "logps/chosen": -137.5977783203125,
66
+ "logps/rejected": -122.20698547363281,
67
+ "loss": 0.6908,
68
+ "rewards/accuracies": 0.5,
69
+ "rewards/chosen": 0.004121971316635609,
70
+ "rewards/margins": 0.0029583165887743235,
71
+ "rewards/rejected": 0.001163654262199998,
72
+ "step": 3
73
+ },
74
+ {
75
+ "debug/policy_chosen_logits": 0.4667843282222748,
76
+ "debug/policy_chosen_logps": -124.55257415771484,
77
+ "debug/policy_rejected_logits": 1.1448464393615723,
78
+ "debug/policy_rejected_logps": -149.21067810058594,
79
+ "debug/reference_chosen_logps": -125.31111907958984,
80
+ "debug/reference_rejected_logps": -148.5081787109375,
81
+ "epoch": 0.07272727272727272,
82
+ "grad_norm": 2.6975903263061793,
83
+ "learning_rate": 5e-07,
84
+ "logits/chosen": 0.4667843282222748,
85
+ "logits/rejected": 1.1448464393615723,
86
+ "logps/chosen": -124.55257415771484,
87
+ "logps/rejected": -149.21067810058594,
88
+ "loss": 0.6893,
89
+ "rewards/accuracies": 0.625,
90
+ "rewards/chosen": 0.007585420273244381,
91
+ "rewards/margins": 0.014610327780246735,
92
+ "rewards/rejected": -0.007024907972663641,
93
+ "step": 4
94
+ },
95
+ {
96
+ "debug/policy_chosen_logits": 0.7746908664703369,
97
+ "debug/policy_chosen_logps": -121.77815246582031,
98
+ "debug/policy_rejected_logits": 1.0104609727859497,
99
+ "debug/policy_rejected_logps": -127.0907211303711,
100
+ "debug/reference_chosen_logps": -122.27123260498047,
101
+ "debug/reference_rejected_logps": -126.88908386230469,
102
+ "epoch": 0.09090909090909091,
103
+ "grad_norm": 2.3195072990326535,
104
+ "learning_rate": 5e-07,
105
+ "logits/chosen": 0.7746908664703369,
106
+ "logits/rejected": 1.0104609727859497,
107
+ "logps/chosen": -121.77815246582031,
108
+ "logps/rejected": -127.0907211303711,
109
+ "loss": 0.6902,
110
+ "rewards/accuracies": 0.5,
111
+ "rewards/chosen": 0.004930791910737753,
112
+ "rewards/margins": 0.006947212386876345,
113
+ "rewards/rejected": -0.0020164204761385918,
114
+ "step": 5
115
+ },
116
+ {
117
+ "debug/policy_chosen_logits": 0.4436347484588623,
118
+ "debug/policy_chosen_logps": -126.17998504638672,
119
+ "debug/policy_rejected_logits": 0.6884921789169312,
120
+ "debug/policy_rejected_logps": -126.66116333007812,
121
+ "debug/reference_chosen_logps": -126.64784240722656,
122
+ "debug/reference_rejected_logps": -126.05181884765625,
123
+ "epoch": 0.10909090909090909,
124
+ "grad_norm": 2.7552991032792282,
125
+ "learning_rate": 5e-07,
126
+ "logits/chosen": 0.4436347484588623,
127
+ "logits/rejected": 0.6884921789169312,
128
+ "logps/chosen": -126.17998504638672,
129
+ "logps/rejected": -126.66116333007812,
130
+ "loss": 0.6891,
131
+ "rewards/accuracies": 0.875,
132
+ "rewards/chosen": 0.004678516648709774,
133
+ "rewards/margins": 0.01077202707529068,
134
+ "rewards/rejected": -0.006093511823564768,
135
+ "step": 6
136
+ },
137
+ {
138
+ "debug/policy_chosen_logits": 0.3949795365333557,
139
+ "debug/policy_chosen_logps": -120.72560119628906,
140
+ "debug/policy_rejected_logits": 1.16127610206604,
141
+ "debug/policy_rejected_logps": -166.54122924804688,
142
+ "debug/reference_chosen_logps": -122.05653381347656,
143
+ "debug/reference_rejected_logps": -166.2891387939453,
144
+ "epoch": 0.12727272727272726,
145
+ "grad_norm": 2.618269906477997,
146
+ "learning_rate": 5e-07,
147
+ "logits/chosen": 0.3949795365333557,
148
+ "logits/rejected": 1.16127610206604,
149
+ "logps/chosen": -120.72560119628906,
150
+ "logps/rejected": -166.54122924804688,
151
+ "loss": 0.6896,
152
+ "rewards/accuracies": 0.75,
153
+ "rewards/chosen": 0.013309288769960403,
154
+ "rewards/margins": 0.015830213204026222,
155
+ "rewards/rejected": -0.002520923502743244,
156
+ "step": 7
157
+ },
158
+ {
159
+ "debug/policy_chosen_logits": 0.7735803127288818,
160
+ "debug/policy_chosen_logps": -135.4464569091797,
161
+ "debug/policy_rejected_logits": 0.9391681551933289,
162
+ "debug/policy_rejected_logps": -139.32675170898438,
163
+ "debug/reference_chosen_logps": -136.41879272460938,
164
+ "debug/reference_rejected_logps": -139.55023193359375,
165
+ "epoch": 0.14545454545454545,
166
+ "grad_norm": 2.5281133640812117,
167
+ "learning_rate": 5e-07,
168
+ "logits/chosen": 0.7735803127288818,
169
+ "logits/rejected": 0.9391681551933289,
170
+ "logps/chosen": -135.4464569091797,
171
+ "logps/rejected": -139.32675170898438,
172
+ "loss": 0.6896,
173
+ "rewards/accuracies": 0.75,
174
+ "rewards/chosen": 0.00972352921962738,
175
+ "rewards/margins": 0.007488718256354332,
176
+ "rewards/rejected": 0.002234811894595623,
177
+ "step": 8
178
+ },
179
+ {
180
+ "debug/policy_chosen_logits": 0.5073169469833374,
181
+ "debug/policy_chosen_logps": -135.45155334472656,
182
+ "debug/policy_rejected_logits": 1.3236225843429565,
183
+ "debug/policy_rejected_logps": -130.271728515625,
184
+ "debug/reference_chosen_logps": -135.94186401367188,
185
+ "debug/reference_rejected_logps": -129.65411376953125,
186
+ "epoch": 0.16363636363636364,
187
+ "grad_norm": 2.5354121562315766,
188
+ "learning_rate": 5e-07,
189
+ "logits/chosen": 0.5073169469833374,
190
+ "logits/rejected": 1.3236225843429565,
191
+ "logps/chosen": -135.45155334472656,
192
+ "logps/rejected": -130.271728515625,
193
+ "loss": 0.6887,
194
+ "rewards/accuracies": 0.75,
195
+ "rewards/chosen": 0.004903106484562159,
196
+ "rewards/margins": 0.011079207062721252,
197
+ "rewards/rejected": -0.006176099181175232,
198
+ "step": 9
199
+ },
200
+ {
201
+ "debug/policy_chosen_logits": 0.9053974151611328,
202
+ "debug/policy_chosen_logps": -129.9651641845703,
203
+ "debug/policy_rejected_logits": 1.6513068675994873,
204
+ "debug/policy_rejected_logps": -157.7940673828125,
205
+ "debug/reference_chosen_logps": -130.6460418701172,
206
+ "debug/reference_rejected_logps": -158.63949584960938,
207
+ "epoch": 0.18181818181818182,
208
+ "grad_norm": 2.810542042119807,
209
+ "learning_rate": 5e-07,
210
+ "logits/chosen": 0.9053974151611328,
211
+ "logits/rejected": 1.6513068675994873,
212
+ "logps/chosen": -129.9651641845703,
213
+ "logps/rejected": -157.7940673828125,
214
+ "loss": 0.6851,
215
+ "rewards/accuracies": 0.375,
216
+ "rewards/chosen": 0.006808795966207981,
217
+ "rewards/margins": -0.0016453838907182217,
218
+ "rewards/rejected": 0.008454179391264915,
219
+ "step": 10
220
+ },
221
+ {
222
+ "debug/policy_chosen_logits": 0.792403519153595,
223
+ "debug/policy_chosen_logps": -141.16513061523438,
224
+ "debug/policy_rejected_logits": 1.2816684246063232,
225
+ "debug/policy_rejected_logps": -152.61949157714844,
226
+ "debug/reference_chosen_logps": -141.37376403808594,
227
+ "debug/reference_rejected_logps": -153.21987915039062,
228
+ "epoch": 0.2,
229
+ "grad_norm": 2.488423861858769,
230
+ "learning_rate": 5e-07,
231
+ "logits/chosen": 0.792403519153595,
232
+ "logits/rejected": 1.2816684246063232,
233
+ "logps/chosen": -141.16513061523438,
234
+ "logps/rejected": -152.61949157714844,
235
+ "loss": 0.6883,
236
+ "rewards/accuracies": 0.625,
237
+ "rewards/chosen": 0.0020863343961536884,
238
+ "rewards/margins": -0.00391742680221796,
239
+ "rewards/rejected": 0.0060037607327103615,
240
+ "step": 11
241
+ },
242
+ {
243
+ "debug/policy_chosen_logits": 1.0979342460632324,
244
+ "debug/policy_chosen_logps": -127.43941497802734,
245
+ "debug/policy_rejected_logits": 0.8406213521957397,
246
+ "debug/policy_rejected_logps": -133.73110961914062,
247
+ "debug/reference_chosen_logps": -128.33518981933594,
248
+ "debug/reference_rejected_logps": -133.88735961914062,
249
+ "epoch": 0.21818181818181817,
250
+ "grad_norm": 2.5700027163746975,
251
+ "learning_rate": 5e-07,
252
+ "logits/chosen": 1.0979342460632324,
253
+ "logits/rejected": 0.8406213521957397,
254
+ "logps/chosen": -127.43941497802734,
255
+ "logps/rejected": -133.73110961914062,
256
+ "loss": 0.6829,
257
+ "rewards/accuracies": 0.75,
258
+ "rewards/chosen": 0.008957691490650177,
259
+ "rewards/margins": 0.007395162247121334,
260
+ "rewards/rejected": 0.0015625286614522338,
261
+ "step": 12
262
+ },
263
+ {
264
+ "debug/policy_chosen_logits": 0.8384308218955994,
265
+ "debug/policy_chosen_logps": -129.3570556640625,
266
+ "debug/policy_rejected_logits": 1.2891650199890137,
267
+ "debug/policy_rejected_logps": -140.60781860351562,
268
+ "debug/reference_chosen_logps": -130.777099609375,
269
+ "debug/reference_rejected_logps": -137.655029296875,
270
+ "epoch": 0.23636363636363636,
271
+ "grad_norm": 2.5021787713492976,
272
+ "learning_rate": 5e-07,
273
+ "logits/chosen": 0.8384308218955994,
274
+ "logits/rejected": 1.2891650199890137,
275
+ "logps/chosen": -129.3570556640625,
276
+ "logps/rejected": -140.60781860351562,
277
+ "loss": 0.6844,
278
+ "rewards/accuracies": 1.0,
279
+ "rewards/chosen": 0.014200429432094097,
280
+ "rewards/margins": 0.043728265911340714,
281
+ "rewards/rejected": -0.029527835547924042,
282
+ "step": 13
283
+ },
284
+ {
285
+ "debug/policy_chosen_logits": 0.7953339219093323,
286
+ "debug/policy_chosen_logps": -123.55655670166016,
287
+ "debug/policy_rejected_logits": 1.151228427886963,
288
+ "debug/policy_rejected_logps": -121.413330078125,
289
+ "debug/reference_chosen_logps": -122.65122985839844,
290
+ "debug/reference_rejected_logps": -119.04695892333984,
291
+ "epoch": 0.2545454545454545,
292
+ "grad_norm": 2.645883535930893,
293
+ "learning_rate": 5e-07,
294
+ "logits/chosen": 0.7953339219093323,
295
+ "logits/rejected": 1.151228427886963,
296
+ "logps/chosen": -123.55655670166016,
297
+ "logps/rejected": -121.413330078125,
298
+ "loss": 0.69,
299
+ "rewards/accuracies": 0.625,
300
+ "rewards/chosen": -0.00905330665409565,
301
+ "rewards/margins": 0.014610443264245987,
302
+ "rewards/rejected": -0.023663748055696487,
303
+ "step": 14
304
+ },
305
+ {
306
+ "debug/policy_chosen_logits": 0.7620865106582642,
307
+ "debug/policy_chosen_logps": -142.4312286376953,
308
+ "debug/policy_rejected_logits": 0.9396656155586243,
309
+ "debug/policy_rejected_logps": -148.87515258789062,
310
+ "debug/reference_chosen_logps": -139.45565795898438,
311
+ "debug/reference_rejected_logps": -145.11453247070312,
312
+ "epoch": 0.2727272727272727,
313
+ "grad_norm": 2.9294875671025875,
314
+ "learning_rate": 5e-07,
315
+ "logits/chosen": 0.7620865106582642,
316
+ "logits/rejected": 0.9396656155586243,
317
+ "logps/chosen": -142.4312286376953,
318
+ "logps/rejected": -148.87515258789062,
319
+ "loss": 0.6889,
320
+ "rewards/accuracies": 0.75,
321
+ "rewards/chosen": -0.02975570783019066,
322
+ "rewards/margins": 0.007850446738302708,
323
+ "rewards/rejected": -0.03760614991188049,
324
+ "step": 15
325
+ },
326
+ {
327
+ "debug/policy_chosen_logits": 1.0136598348617554,
328
+ "debug/policy_chosen_logps": -130.98876953125,
329
+ "debug/policy_rejected_logits": 0.7458176612854004,
330
+ "debug/policy_rejected_logps": -134.5099639892578,
331
+ "debug/reference_chosen_logps": -131.6763153076172,
332
+ "debug/reference_rejected_logps": -133.0446014404297,
333
+ "epoch": 0.2909090909090909,
334
+ "grad_norm": 2.2028450675710602,
335
+ "learning_rate": 5e-07,
336
+ "logits/chosen": 1.0136598348617554,
337
+ "logits/rejected": 0.7458176612854004,
338
+ "logps/chosen": -130.98876953125,
339
+ "logps/rejected": -134.5099639892578,
340
+ "loss": 0.6842,
341
+ "rewards/accuracies": 0.625,
342
+ "rewards/chosen": 0.006875476334244013,
343
+ "rewards/margins": 0.021529102697968483,
344
+ "rewards/rejected": -0.014653624035418034,
345
+ "step": 16
346
+ },
347
+ {
348
+ "debug/policy_chosen_logits": 0.8773481249809265,
349
+ "debug/policy_chosen_logps": -134.43736267089844,
350
+ "debug/policy_rejected_logits": 0.8306017518043518,
351
+ "debug/policy_rejected_logps": -131.41659545898438,
352
+ "debug/reference_chosen_logps": -134.44261169433594,
353
+ "debug/reference_rejected_logps": -130.02728271484375,
354
+ "epoch": 0.3090909090909091,
355
+ "grad_norm": 2.8510973680896896,
356
+ "learning_rate": 5e-07,
357
+ "logits/chosen": 0.8773481249809265,
358
+ "logits/rejected": 0.8306017518043518,
359
+ "logps/chosen": -134.43736267089844,
360
+ "logps/rejected": -131.41659545898438,
361
+ "loss": 0.6817,
362
+ "rewards/accuracies": 0.625,
363
+ "rewards/chosen": 5.249073728919029e-05,
364
+ "rewards/margins": 0.01394571177661419,
365
+ "rewards/rejected": -0.013893223367631435,
366
+ "step": 17
367
+ },
368
+ {
369
+ "debug/policy_chosen_logits": 0.32443925738334656,
370
+ "debug/policy_chosen_logps": -117.34490966796875,
371
+ "debug/policy_rejected_logits": 0.6942044496536255,
372
+ "debug/policy_rejected_logps": -133.7316436767578,
373
+ "debug/reference_chosen_logps": -118.01707458496094,
374
+ "debug/reference_rejected_logps": -132.0999755859375,
375
+ "epoch": 0.32727272727272727,
376
+ "grad_norm": 2.9624915134206953,
377
+ "learning_rate": 5e-07,
378
+ "logits/chosen": 0.32443925738334656,
379
+ "logits/rejected": 0.6942044496536255,
380
+ "logps/chosen": -117.34490966796875,
381
+ "logps/rejected": -133.7316436767578,
382
+ "loss": 0.6789,
383
+ "rewards/accuracies": 0.5,
384
+ "rewards/chosen": 0.006721640005707741,
385
+ "rewards/margins": 0.023038387298583984,
386
+ "rewards/rejected": -0.016316747292876244,
387
+ "step": 18
388
+ },
389
+ {
390
+ "debug/policy_chosen_logits": 0.6036683320999146,
391
+ "debug/policy_chosen_logps": -139.1854248046875,
392
+ "debug/policy_rejected_logits": 1.0925521850585938,
393
+ "debug/policy_rejected_logps": -129.66262817382812,
394
+ "debug/reference_chosen_logps": -139.90301513671875,
395
+ "debug/reference_rejected_logps": -128.35787963867188,
396
+ "epoch": 0.34545454545454546,
397
+ "grad_norm": 3.0283747066503546,
398
+ "learning_rate": 5e-07,
399
+ "logits/chosen": 0.6036683320999146,
400
+ "logits/rejected": 1.0925521850585938,
401
+ "logps/chosen": -139.1854248046875,
402
+ "logps/rejected": -129.66262817382812,
403
+ "loss": 0.6771,
404
+ "rewards/accuracies": 0.875,
405
+ "rewards/chosen": 0.007175750564783812,
406
+ "rewards/margins": 0.02022336982190609,
407
+ "rewards/rejected": -0.013047618791460991,
408
+ "step": 19
409
+ },
410
+ {
411
+ "debug/policy_chosen_logits": 0.19202691316604614,
412
+ "debug/policy_chosen_logps": -125.41830444335938,
413
+ "debug/policy_rejected_logits": 0.6729949116706848,
414
+ "debug/policy_rejected_logps": -136.26675415039062,
415
+ "debug/reference_chosen_logps": -129.2891845703125,
416
+ "debug/reference_rejected_logps": -131.7703857421875,
417
+ "epoch": 0.36363636363636365,
418
+ "grad_norm": 2.5690310299484906,
419
+ "learning_rate": 5e-07,
420
+ "logits/chosen": 0.19202691316604614,
421
+ "logits/rejected": 0.6729949116706848,
422
+ "logps/chosen": -125.41830444335938,
423
+ "logps/rejected": -136.26675415039062,
424
+ "loss": 0.6813,
425
+ "rewards/accuracies": 0.875,
426
+ "rewards/chosen": 0.03870879113674164,
427
+ "rewards/margins": 0.08367244899272919,
428
+ "rewards/rejected": -0.04496365413069725,
429
+ "step": 20
430
+ },
431
+ {
432
+ "debug/policy_chosen_logits": 0.7662903070449829,
433
+ "debug/policy_chosen_logps": -124.8802261352539,
434
+ "debug/policy_rejected_logits": 1.350595474243164,
435
+ "debug/policy_rejected_logps": -131.6766357421875,
436
+ "debug/reference_chosen_logps": -125.80413818359375,
437
+ "debug/reference_rejected_logps": -129.85723876953125,
438
+ "epoch": 0.38181818181818183,
439
+ "grad_norm": 2.9390962336045523,
440
+ "learning_rate": 5e-07,
441
+ "logits/chosen": 0.7662903070449829,
442
+ "logits/rejected": 1.350595474243164,
443
+ "logps/chosen": -124.8802261352539,
444
+ "logps/rejected": -131.6766357421875,
445
+ "loss": 0.6739,
446
+ "rewards/accuracies": 0.75,
447
+ "rewards/chosen": 0.009239129722118378,
448
+ "rewards/margins": 0.027433060109615326,
449
+ "rewards/rejected": -0.018193930387496948,
450
+ "step": 21
451
+ },
452
+ {
453
+ "debug/policy_chosen_logits": 1.21824312210083,
454
+ "debug/policy_chosen_logps": -118.64605712890625,
455
+ "debug/policy_rejected_logits": 1.3743467330932617,
456
+ "debug/policy_rejected_logps": -130.24822998046875,
457
+ "debug/reference_chosen_logps": -119.62400817871094,
458
+ "debug/reference_rejected_logps": -129.10247802734375,
459
+ "epoch": 0.4,
460
+ "grad_norm": 2.4995079289310267,
461
+ "learning_rate": 5e-07,
462
+ "logits/chosen": 1.21824312210083,
463
+ "logits/rejected": 1.3743467330932617,
464
+ "logps/chosen": -118.64605712890625,
465
+ "logps/rejected": -130.24822998046875,
466
+ "loss": 0.6711,
467
+ "rewards/accuracies": 0.5,
468
+ "rewards/chosen": 0.009779386222362518,
469
+ "rewards/margins": 0.021236935630440712,
470
+ "rewards/rejected": -0.011457548476755619,
471
+ "step": 22
472
+ },
473
+ {
474
+ "debug/policy_chosen_logits": 0.32036519050598145,
475
+ "debug/policy_chosen_logps": -137.20065307617188,
476
+ "debug/policy_rejected_logits": 0.4707115590572357,
477
+ "debug/policy_rejected_logps": -141.1231231689453,
478
+ "debug/reference_chosen_logps": -137.7662353515625,
479
+ "debug/reference_rejected_logps": -140.11264038085938,
480
+ "epoch": 0.41818181818181815,
481
+ "grad_norm": 2.726999800833551,
482
+ "learning_rate": 5e-07,
483
+ "logits/chosen": 0.32036519050598145,
484
+ "logits/rejected": 0.4707115590572357,
485
+ "logps/chosen": -137.20065307617188,
486
+ "logps/rejected": -141.1231231689453,
487
+ "loss": 0.6804,
488
+ "rewards/accuracies": 0.875,
489
+ "rewards/chosen": 0.00565571803599596,
490
+ "rewards/margins": 0.01576041243970394,
491
+ "rewards/rejected": -0.010104694403707981,
492
+ "step": 23
493
+ },
494
+ {
495
+ "debug/policy_chosen_logits": 1.174080491065979,
496
+ "debug/policy_chosen_logps": -130.7235870361328,
497
+ "debug/policy_rejected_logits": 1.6448463201522827,
498
+ "debug/policy_rejected_logps": -149.97439575195312,
499
+ "debug/reference_chosen_logps": -131.06661987304688,
500
+ "debug/reference_rejected_logps": -147.88232421875,
501
+ "epoch": 0.43636363636363634,
502
+ "grad_norm": 2.652353634767768,
503
+ "learning_rate": 5e-07,
504
+ "logits/chosen": 1.174080491065979,
505
+ "logits/rejected": 1.6448463201522827,
506
+ "logps/chosen": -130.7235870361328,
507
+ "logps/rejected": -149.97439575195312,
508
+ "loss": 0.6831,
509
+ "rewards/accuracies": 0.75,
510
+ "rewards/chosen": 0.0034304135479032993,
511
+ "rewards/margins": 0.02435128018260002,
512
+ "rewards/rejected": -0.02092086710035801,
513
+ "step": 24
514
+ },
515
+ {
516
+ "debug/policy_chosen_logits": 0.5194191336631775,
517
+ "debug/policy_chosen_logps": -125.35334014892578,
518
+ "debug/policy_rejected_logits": 0.8994636535644531,
519
+ "debug/policy_rejected_logps": -129.91224670410156,
520
+ "debug/reference_chosen_logps": -127.37261962890625,
521
+ "debug/reference_rejected_logps": -127.31116485595703,
522
+ "epoch": 0.45454545454545453,
523
+ "grad_norm": 2.970456164285477,
524
+ "learning_rate": 5e-07,
525
+ "logits/chosen": 0.5194191336631775,
526
+ "logits/rejected": 0.8994636535644531,
527
+ "logps/chosen": -125.35334014892578,
528
+ "logps/rejected": -129.91224670410156,
529
+ "loss": 0.682,
530
+ "rewards/accuracies": 0.5,
531
+ "rewards/chosen": 0.02019280195236206,
532
+ "rewards/margins": 0.04620352387428284,
533
+ "rewards/rejected": -0.026010721921920776,
534
+ "step": 25
535
+ },
536
+ {
537
+ "debug/policy_chosen_logits": 0.6115962862968445,
538
+ "debug/policy_chosen_logps": -150.66409301757812,
539
+ "debug/policy_rejected_logits": 0.6578989028930664,
540
+ "debug/policy_rejected_logps": -153.54840087890625,
541
+ "debug/reference_chosen_logps": -151.63339233398438,
542
+ "debug/reference_rejected_logps": -151.82666015625,
543
+ "epoch": 0.4727272727272727,
544
+ "grad_norm": 2.580941309384753,
545
+ "learning_rate": 5e-07,
546
+ "logits/chosen": 0.6115962862968445,
547
+ "logits/rejected": 0.6578989028930664,
548
+ "logps/chosen": -150.66409301757812,
549
+ "logps/rejected": -153.54840087890625,
550
+ "loss": 0.6722,
551
+ "rewards/accuracies": 0.875,
552
+ "rewards/chosen": 0.009693136438727379,
553
+ "rewards/margins": 0.026910629123449326,
554
+ "rewards/rejected": -0.017217492684721947,
555
+ "step": 26
556
+ },
557
+ {
558
+ "debug/policy_chosen_logits": 0.41199997067451477,
559
+ "debug/policy_chosen_logps": -148.89768981933594,
560
+ "debug/policy_rejected_logits": 0.782092273235321,
561
+ "debug/policy_rejected_logps": -147.73719787597656,
562
+ "debug/reference_chosen_logps": -149.99037170410156,
563
+ "debug/reference_rejected_logps": -146.23422241210938,
564
+ "epoch": 0.4909090909090909,
565
+ "grad_norm": 2.7525802515180575,
566
+ "learning_rate": 5e-07,
567
+ "logits/chosen": 0.41199997067451477,
568
+ "logits/rejected": 0.782092273235321,
569
+ "logps/chosen": -148.89768981933594,
570
+ "logps/rejected": -147.73719787597656,
571
+ "loss": 0.6753,
572
+ "rewards/accuracies": 0.75,
573
+ "rewards/chosen": 0.010926798917353153,
574
+ "rewards/margins": 0.02595655433833599,
575
+ "rewards/rejected": -0.015029754489660263,
576
+ "step": 27
577
+ },
578
+ {
579
+ "debug/policy_chosen_logits": 0.34080997109413147,
580
+ "debug/policy_chosen_logps": -126.34262084960938,
581
+ "debug/policy_rejected_logits": 0.36349666118621826,
582
+ "debug/policy_rejected_logps": -134.22015380859375,
583
+ "debug/reference_chosen_logps": -125.65828704833984,
584
+ "debug/reference_rejected_logps": -130.0347442626953,
585
+ "epoch": 0.509090909090909,
586
+ "grad_norm": 2.680704105439303,
587
+ "learning_rate": 5e-07,
588
+ "logits/chosen": 0.34080997109413147,
589
+ "logits/rejected": 0.36349666118621826,
590
+ "logps/chosen": -126.34262084960938,
591
+ "logps/rejected": -134.22015380859375,
592
+ "loss": 0.6773,
593
+ "rewards/accuracies": 0.625,
594
+ "rewards/chosen": -0.006843318697065115,
595
+ "rewards/margins": 0.03501085191965103,
596
+ "rewards/rejected": -0.04185417294502258,
597
+ "step": 28
598
+ },
599
+ {
600
+ "debug/policy_chosen_logits": 0.7369721531867981,
601
+ "debug/policy_chosen_logps": -137.26649475097656,
602
+ "debug/policy_rejected_logits": 0.7872332334518433,
603
+ "debug/policy_rejected_logps": -143.5257568359375,
604
+ "debug/reference_chosen_logps": -137.09490966796875,
605
+ "debug/reference_rejected_logps": -140.14822387695312,
606
+ "epoch": 0.5272727272727272,
607
+ "grad_norm": 2.8512448335063185,
608
+ "learning_rate": 5e-07,
609
+ "logits/chosen": 0.7369721531867981,
610
+ "logits/rejected": 0.7872332334518433,
611
+ "logps/chosen": -137.26649475097656,
612
+ "logps/rejected": -143.5257568359375,
613
+ "loss": 0.6764,
614
+ "rewards/accuracies": 0.625,
615
+ "rewards/chosen": -0.0017157741822302341,
616
+ "rewards/margins": 0.03205941990017891,
617
+ "rewards/rejected": -0.033775195479393005,
618
+ "step": 29
619
+ },
620
+ {
621
+ "debug/policy_chosen_logits": 0.7467297911643982,
622
+ "debug/policy_chosen_logps": -142.2055206298828,
623
+ "debug/policy_rejected_logits": 0.8480007648468018,
624
+ "debug/policy_rejected_logps": -145.74615478515625,
625
+ "debug/reference_chosen_logps": -141.9608154296875,
626
+ "debug/reference_rejected_logps": -141.98846435546875,
627
+ "epoch": 0.5454545454545454,
628
+ "grad_norm": 3.2448790034806456,
629
+ "learning_rate": 5e-07,
630
+ "logits/chosen": 0.7467297911643982,
631
+ "logits/rejected": 0.8480007648468018,
632
+ "logps/chosen": -142.2055206298828,
633
+ "logps/rejected": -145.74615478515625,
634
+ "loss": 0.6655,
635
+ "rewards/accuracies": 0.625,
636
+ "rewards/chosen": -0.0024469951167702675,
637
+ "rewards/margins": 0.03513010963797569,
638
+ "rewards/rejected": -0.03757710009813309,
639
+ "step": 30
640
+ },
641
+ {
642
+ "debug/policy_chosen_logits": 0.45751431584358215,
643
+ "debug/policy_chosen_logps": -119.06462097167969,
644
+ "debug/policy_rejected_logits": 1.1633354425430298,
645
+ "debug/policy_rejected_logps": -134.27395629882812,
646
+ "debug/reference_chosen_logps": -118.83071899414062,
647
+ "debug/reference_rejected_logps": -132.58956909179688,
648
+ "epoch": 0.5636363636363636,
649
+ "grad_norm": 2.604825578516903,
650
+ "learning_rate": 5e-07,
651
+ "logits/chosen": 0.45751431584358215,
652
+ "logits/rejected": 1.1633354425430298,
653
+ "logps/chosen": -119.06462097167969,
654
+ "logps/rejected": -134.27395629882812,
655
+ "loss": 0.6774,
656
+ "rewards/accuracies": 0.875,
657
+ "rewards/chosen": -0.002338953083381057,
658
+ "rewards/margins": 0.01450504269450903,
659
+ "rewards/rejected": -0.016843995079398155,
660
+ "step": 31
661
+ },
662
+ {
663
+ "debug/policy_chosen_logits": 0.47755831480026245,
664
+ "debug/policy_chosen_logps": -113.85858154296875,
665
+ "debug/policy_rejected_logits": 0.9192254543304443,
666
+ "debug/policy_rejected_logps": -142.87522888183594,
667
+ "debug/reference_chosen_logps": -114.60741424560547,
668
+ "debug/reference_rejected_logps": -141.06182861328125,
669
+ "epoch": 0.5818181818181818,
670
+ "grad_norm": 2.4614920126129927,
671
+ "learning_rate": 5e-07,
672
+ "logits/chosen": 0.47755831480026245,
673
+ "logits/rejected": 0.9192254543304443,
674
+ "logps/chosen": -113.85858154296875,
675
+ "logps/rejected": -142.87522888183594,
676
+ "loss": 0.6833,
677
+ "rewards/accuracies": 1.0,
678
+ "rewards/chosen": 0.007488393224775791,
679
+ "rewards/margins": 0.02562214806675911,
680
+ "rewards/rejected": -0.018133753910660744,
681
+ "step": 32
682
+ },
683
+ {
684
+ "debug/policy_chosen_logits": 0.7646514773368835,
685
+ "debug/policy_chosen_logps": -140.6458740234375,
686
+ "debug/policy_rejected_logits": 1.0664713382720947,
687
+ "debug/policy_rejected_logps": -154.0317840576172,
688
+ "debug/reference_chosen_logps": -141.03515625,
689
+ "debug/reference_rejected_logps": -152.28282165527344,
690
+ "epoch": 0.6,
691
+ "grad_norm": 2.892367660321636,
692
+ "learning_rate": 5e-07,
693
+ "logits/chosen": 0.7646514773368835,
694
+ "logits/rejected": 1.0664713382720947,
695
+ "logps/chosen": -140.6458740234375,
696
+ "logps/rejected": -154.0317840576172,
697
+ "loss": 0.6726,
698
+ "rewards/accuracies": 0.75,
699
+ "rewards/chosen": 0.0038927835412323475,
700
+ "rewards/margins": 0.021382424980401993,
701
+ "rewards/rejected": -0.017489641904830933,
702
+ "step": 33
703
+ },
704
+ {
705
+ "debug/policy_chosen_logits": 1.0055601596832275,
706
+ "debug/policy_chosen_logps": -130.6605224609375,
707
+ "debug/policy_rejected_logits": 0.745191216468811,
708
+ "debug/policy_rejected_logps": -126.47428894042969,
709
+ "debug/reference_chosen_logps": -129.04339599609375,
710
+ "debug/reference_rejected_logps": -119.90361022949219,
711
+ "epoch": 0.6181818181818182,
712
+ "grad_norm": 2.9093181223713334,
713
+ "learning_rate": 5e-07,
714
+ "logits/chosen": 1.0055601596832275,
715
+ "logits/rejected": 0.745191216468811,
716
+ "logps/chosen": -130.6605224609375,
717
+ "logps/rejected": -126.47428894042969,
718
+ "loss": 0.6777,
719
+ "rewards/accuracies": 0.625,
720
+ "rewards/chosen": -0.01617136038839817,
721
+ "rewards/margins": 0.049535419791936874,
722
+ "rewards/rejected": -0.0657067745923996,
723
+ "step": 34
724
+ },
725
+ {
726
+ "debug/policy_chosen_logits": 0.665949285030365,
727
+ "debug/policy_chosen_logps": -167.20567321777344,
728
+ "debug/policy_rejected_logits": 0.8378525972366333,
729
+ "debug/policy_rejected_logps": -176.95294189453125,
730
+ "debug/reference_chosen_logps": -166.0945281982422,
731
+ "debug/reference_rejected_logps": -174.29322814941406,
732
+ "epoch": 0.6363636363636364,
733
+ "grad_norm": 3.164591716357413,
734
+ "learning_rate": 5e-07,
735
+ "logits/chosen": 0.665949285030365,
736
+ "logits/rejected": 0.8378525972366333,
737
+ "logps/chosen": -167.20567321777344,
738
+ "logps/rejected": -176.95294189453125,
739
+ "loss": 0.6686,
740
+ "rewards/accuracies": 0.875,
741
+ "rewards/chosen": -0.011111450381577015,
742
+ "rewards/margins": 0.015485725365579128,
743
+ "rewards/rejected": -0.026597173884510994,
744
+ "step": 35
745
+ },
746
+ {
747
+ "debug/policy_chosen_logits": 0.9554933905601501,
748
+ "debug/policy_chosen_logps": -136.39425659179688,
749
+ "debug/policy_rejected_logits": 0.49876803159713745,
750
+ "debug/policy_rejected_logps": -153.6390380859375,
751
+ "debug/reference_chosen_logps": -136.20895385742188,
752
+ "debug/reference_rejected_logps": -146.877197265625,
753
+ "epoch": 0.6545454545454545,
754
+ "grad_norm": 3.098324286117154,
755
+ "learning_rate": 5e-07,
756
+ "logits/chosen": 0.9554933905601501,
757
+ "logits/rejected": 0.49876803159713745,
758
+ "logps/chosen": -136.39425659179688,
759
+ "logps/rejected": -153.6390380859375,
760
+ "loss": 0.6738,
761
+ "rewards/accuracies": 0.5,
762
+ "rewards/chosen": -0.0018529421649873257,
763
+ "rewards/margins": 0.06576552242040634,
764
+ "rewards/rejected": -0.0676184669137001,
765
+ "step": 36
766
+ },
767
+ {
768
+ "debug/policy_chosen_logits": 0.2520717680454254,
769
+ "debug/policy_chosen_logps": -121.43704223632812,
770
+ "debug/policy_rejected_logits": 0.5053534507751465,
771
+ "debug/policy_rejected_logps": -158.01513671875,
772
+ "debug/reference_chosen_logps": -120.23490142822266,
773
+ "debug/reference_rejected_logps": -155.57327270507812,
774
+ "epoch": 0.6727272727272727,
775
+ "grad_norm": 3.1244720060016986,
776
+ "learning_rate": 5e-07,
777
+ "logits/chosen": 0.2520717680454254,
778
+ "logits/rejected": 0.5053534507751465,
779
+ "logps/chosen": -121.43704223632812,
780
+ "logps/rejected": -158.01513671875,
781
+ "loss": 0.6738,
782
+ "rewards/accuracies": 0.625,
783
+ "rewards/chosen": -0.012021454982459545,
784
+ "rewards/margins": 0.012397251091897488,
785
+ "rewards/rejected": -0.024418707937002182,
786
+ "step": 37
787
+ },
788
+ {
789
+ "debug/policy_chosen_logits": 0.35868680477142334,
790
+ "debug/policy_chosen_logps": -139.0185546875,
791
+ "debug/policy_rejected_logits": 0.39101022481918335,
792
+ "debug/policy_rejected_logps": -153.45266723632812,
793
+ "debug/reference_chosen_logps": -138.5296630859375,
794
+ "debug/reference_rejected_logps": -147.90438842773438,
795
+ "epoch": 0.6909090909090909,
796
+ "grad_norm": 5.757112422371425,
797
+ "learning_rate": 5e-07,
798
+ "logits/chosen": 0.35868680477142334,
799
+ "logits/rejected": 0.39101022481918335,
800
+ "logps/chosen": -139.0185546875,
801
+ "logps/rejected": -153.45266723632812,
802
+ "loss": 0.6783,
803
+ "rewards/accuracies": 0.75,
804
+ "rewards/chosen": -0.0048888688907027245,
805
+ "rewards/margins": 0.050593726336956024,
806
+ "rewards/rejected": -0.05548259615898132,
807
+ "step": 38
808
+ },
809
+ {
810
+ "debug/policy_chosen_logits": 0.6587602496147156,
811
+ "debug/policy_chosen_logps": -140.99093627929688,
812
+ "debug/policy_rejected_logits": 0.5164349675178528,
813
+ "debug/policy_rejected_logps": -140.2601318359375,
814
+ "debug/reference_chosen_logps": -136.22140502929688,
815
+ "debug/reference_rejected_logps": -129.1555633544922,
816
+ "epoch": 0.7090909090909091,
817
+ "grad_norm": 3.224110720714856,
818
+ "learning_rate": 5e-07,
819
+ "logits/chosen": 0.6587602496147156,
820
+ "logits/rejected": 0.5164349675178528,
821
+ "logps/chosen": -140.99093627929688,
822
+ "logps/rejected": -140.2601318359375,
823
+ "loss": 0.6744,
824
+ "rewards/accuracies": 0.625,
825
+ "rewards/chosen": -0.04769532382488251,
826
+ "rewards/margins": 0.0633503645658493,
827
+ "rewards/rejected": -0.11104568839073181,
828
+ "step": 39
829
+ },
830
+ {
831
+ "debug/policy_chosen_logits": 0.36872461438179016,
832
+ "debug/policy_chosen_logps": -153.08731079101562,
833
+ "debug/policy_rejected_logits": 0.7442142367362976,
834
+ "debug/policy_rejected_logps": -148.73162841796875,
835
+ "debug/reference_chosen_logps": -152.52410888671875,
836
+ "debug/reference_rejected_logps": -143.73696899414062,
837
+ "epoch": 0.7272727272727273,
838
+ "grad_norm": 3.030981160478689,
839
+ "learning_rate": 5e-07,
840
+ "logits/chosen": 0.36872461438179016,
841
+ "logits/rejected": 0.7442142367362976,
842
+ "logps/chosen": -153.08731079101562,
843
+ "logps/rejected": -148.73162841796875,
844
+ "loss": 0.6692,
845
+ "rewards/accuracies": 0.875,
846
+ "rewards/chosen": -0.0056318375281989574,
847
+ "rewards/margins": 0.044314805418252945,
848
+ "rewards/rejected": -0.04994663968682289,
849
+ "step": 40
850
+ },
851
+ {
852
+ "debug/policy_chosen_logits": 0.9497181177139282,
853
+ "debug/policy_chosen_logps": -147.80657958984375,
854
+ "debug/policy_rejected_logits": 1.0767797231674194,
855
+ "debug/policy_rejected_logps": -147.0999755859375,
856
+ "debug/reference_chosen_logps": -144.19863891601562,
857
+ "debug/reference_rejected_logps": -141.43292236328125,
858
+ "epoch": 0.7454545454545455,
859
+ "grad_norm": 2.594544171794356,
860
+ "learning_rate": 5e-07,
861
+ "logits/chosen": 0.9497181177139282,
862
+ "logits/rejected": 1.0767797231674194,
863
+ "logps/chosen": -147.80657958984375,
864
+ "logps/rejected": -147.0999755859375,
865
+ "loss": 0.6729,
866
+ "rewards/accuracies": 0.5,
867
+ "rewards/chosen": -0.0360795296728611,
868
+ "rewards/margins": 0.020590826869010925,
869
+ "rewards/rejected": -0.05667036026716232,
870
+ "step": 41
871
+ },
872
+ {
873
+ "debug/policy_chosen_logits": 0.2258227914571762,
874
+ "debug/policy_chosen_logps": -125.74639129638672,
875
+ "debug/policy_rejected_logits": 0.5640020966529846,
876
+ "debug/policy_rejected_logps": -163.89407348632812,
877
+ "debug/reference_chosen_logps": -124.42830657958984,
878
+ "debug/reference_rejected_logps": -156.05613708496094,
879
+ "epoch": 0.7636363636363637,
880
+ "grad_norm": 2.6702311112671437,
881
+ "learning_rate": 5e-07,
882
+ "logits/chosen": 0.2258227914571762,
883
+ "logits/rejected": 0.5640020966529846,
884
+ "logps/chosen": -125.74639129638672,
885
+ "logps/rejected": -163.89407348632812,
886
+ "loss": 0.6707,
887
+ "rewards/accuracies": 0.75,
888
+ "rewards/chosen": -0.013180827721953392,
889
+ "rewards/margins": 0.06519865989685059,
890
+ "rewards/rejected": -0.07837948203086853,
891
+ "step": 42
892
+ },
893
+ {
894
+ "debug/policy_chosen_logits": 0.5549467206001282,
895
+ "debug/policy_chosen_logps": -146.0997314453125,
896
+ "debug/policy_rejected_logits": 0.725266695022583,
897
+ "debug/policy_rejected_logps": -147.44781494140625,
898
+ "debug/reference_chosen_logps": -145.01519775390625,
899
+ "debug/reference_rejected_logps": -139.47506713867188,
900
+ "epoch": 0.7818181818181819,
901
+ "grad_norm": 3.292126654382054,
902
+ "learning_rate": 5e-07,
903
+ "logits/chosen": 0.5549467206001282,
904
+ "logits/rejected": 0.725266695022583,
905
+ "logps/chosen": -146.0997314453125,
906
+ "logps/rejected": -147.44781494140625,
907
+ "loss": 0.6715,
908
+ "rewards/accuracies": 0.625,
909
+ "rewards/chosen": -0.010845517739653587,
910
+ "rewards/margins": 0.06888195872306824,
911
+ "rewards/rejected": -0.07972747832536697,
912
+ "step": 43
913
+ },
914
+ {
915
+ "debug/policy_chosen_logits": -0.019000336527824402,
916
+ "debug/policy_chosen_logps": -190.69863891601562,
917
+ "debug/policy_rejected_logits": 0.052988436073064804,
918
+ "debug/policy_rejected_logps": -200.59532165527344,
919
+ "debug/reference_chosen_logps": -116.88156127929688,
920
+ "debug/reference_rejected_logps": -129.81407165527344,
921
+ "epoch": 0.8,
922
+ "grad_norm": 6.5675247662578915,
923
+ "learning_rate": 5e-07,
924
+ "logits/chosen": -0.019000336527824402,
925
+ "logits/rejected": 0.052988436073064804,
926
+ "logps/chosen": -190.69863891601562,
927
+ "logps/rejected": -200.59532165527344,
928
+ "loss": 0.6624,
929
+ "rewards/accuracies": 0.625,
930
+ "rewards/chosen": -0.7381708025932312,
931
+ "rewards/margins": -0.030358243733644485,
932
+ "rewards/rejected": -0.7078125476837158,
933
+ "step": 44
934
+ },
935
+ {
936
+ "debug/policy_chosen_logits": -0.10405106842517853,
937
+ "debug/policy_chosen_logps": -157.44500732421875,
938
+ "debug/policy_rejected_logits": 0.3257129192352295,
939
+ "debug/policy_rejected_logps": -170.44888305664062,
940
+ "debug/reference_chosen_logps": -139.64646911621094,
941
+ "debug/reference_rejected_logps": -142.53656005859375,
942
+ "epoch": 0.8181818181818182,
943
+ "grad_norm": 4.692245763612819,
944
+ "learning_rate": 5e-07,
945
+ "logits/chosen": -0.10405106842517853,
946
+ "logits/rejected": 0.3257129192352295,
947
+ "logps/chosen": -157.44500732421875,
948
+ "logps/rejected": -170.44888305664062,
949
+ "loss": 0.6487,
950
+ "rewards/accuracies": 0.625,
951
+ "rewards/chosen": -0.17798534035682678,
952
+ "rewards/margins": 0.10113789886236191,
953
+ "rewards/rejected": -0.2791232466697693,
954
+ "step": 45
955
+ },
956
+ {
957
+ "debug/policy_chosen_logits": 0.4848513901233673,
958
+ "debug/policy_chosen_logps": -132.71551513671875,
959
+ "debug/policy_rejected_logits": 0.5532675981521606,
960
+ "debug/policy_rejected_logps": -146.56289672851562,
961
+ "debug/reference_chosen_logps": -128.2412109375,
962
+ "debug/reference_rejected_logps": -136.38479614257812,
963
+ "epoch": 0.8363636363636363,
964
+ "grad_norm": 3.7539173764574945,
965
+ "learning_rate": 5e-07,
966
+ "logits/chosen": 0.4848513901233673,
967
+ "logits/rejected": 0.5532675981521606,
968
+ "logps/chosen": -132.71551513671875,
969
+ "logps/rejected": -146.56289672851562,
970
+ "loss": 0.6673,
971
+ "rewards/accuracies": 0.375,
972
+ "rewards/chosen": -0.04474309831857681,
973
+ "rewards/margins": 0.05703797936439514,
974
+ "rewards/rejected": -0.10178108513355255,
975
+ "step": 46
976
+ },
977
+ {
978
+ "debug/policy_chosen_logits": 1.0264931917190552,
979
+ "debug/policy_chosen_logps": -127.01493072509766,
980
+ "debug/policy_rejected_logits": 1.2245111465454102,
981
+ "debug/policy_rejected_logps": -139.18240356445312,
982
+ "debug/reference_chosen_logps": -120.42625427246094,
983
+ "debug/reference_rejected_logps": -133.286865234375,
984
+ "epoch": 0.8545454545454545,
985
+ "grad_norm": 3.509455580959997,
986
+ "learning_rate": 5e-07,
987
+ "logits/chosen": 1.0264931917190552,
988
+ "logits/rejected": 1.2245111465454102,
989
+ "logps/chosen": -127.01493072509766,
990
+ "logps/rejected": -139.18240356445312,
991
+ "loss": 0.6752,
992
+ "rewards/accuracies": 0.5,
993
+ "rewards/chosen": -0.06588685512542725,
994
+ "rewards/margins": -0.0069314176216721535,
995
+ "rewards/rejected": -0.05895543843507767,
996
+ "step": 47
997
+ },
998
+ {
999
+ "debug/policy_chosen_logits": 0.8054248094558716,
1000
+ "debug/policy_chosen_logps": -141.4749755859375,
1001
+ "debug/policy_rejected_logits": 0.9782359004020691,
1002
+ "debug/policy_rejected_logps": -170.6127471923828,
1003
+ "debug/reference_chosen_logps": -137.53990173339844,
1004
+ "debug/reference_rejected_logps": -161.91265869140625,
1005
+ "epoch": 0.8727272727272727,
1006
+ "grad_norm": 2.906297231648741,
1007
+ "learning_rate": 5e-07,
1008
+ "logits/chosen": 0.8054248094558716,
1009
+ "logits/rejected": 0.9782359004020691,
1010
+ "logps/chosen": -141.4749755859375,
1011
+ "logps/rejected": -170.6127471923828,
1012
+ "loss": 0.6715,
1013
+ "rewards/accuracies": 0.875,
1014
+ "rewards/chosen": -0.039350755512714386,
1015
+ "rewards/margins": 0.04765000939369202,
1016
+ "rewards/rejected": -0.087000772356987,
1017
+ "step": 48
1018
+ },
1019
+ {
1020
+ "debug/policy_chosen_logits": 0.7787018418312073,
1021
+ "debug/policy_chosen_logps": -144.430419921875,
1022
+ "debug/policy_rejected_logits": 1.1615475416183472,
1023
+ "debug/policy_rejected_logps": -154.2880859375,
1024
+ "debug/reference_chosen_logps": -143.35760498046875,
1025
+ "debug/reference_rejected_logps": -149.20867919921875,
1026
+ "epoch": 0.8909090909090909,
1027
+ "grad_norm": 3.916602695749529,
1028
+ "learning_rate": 5e-07,
1029
+ "logits/chosen": 0.7787018418312073,
1030
+ "logits/rejected": 1.1615475416183472,
1031
+ "logps/chosen": -144.430419921875,
1032
+ "logps/rejected": -154.2880859375,
1033
+ "loss": 0.6714,
1034
+ "rewards/accuracies": 0.75,
1035
+ "rewards/chosen": -0.010728321969509125,
1036
+ "rewards/margins": 0.04006559029221535,
1037
+ "rewards/rejected": -0.05079391598701477,
1038
+ "step": 49
1039
+ },
1040
+ {
1041
+ "debug/policy_chosen_logits": 0.4613797962665558,
1042
+ "debug/policy_chosen_logps": -136.51910400390625,
1043
+ "debug/policy_rejected_logits": 0.8358896970748901,
1044
+ "debug/policy_rejected_logps": -144.94639587402344,
1045
+ "debug/reference_chosen_logps": -134.311279296875,
1046
+ "debug/reference_rejected_logps": -135.14649963378906,
1047
+ "epoch": 0.9090909090909091,
1048
+ "grad_norm": 3.261755038151918,
1049
+ "learning_rate": 5e-07,
1050
+ "logits/chosen": 0.4613797962665558,
1051
+ "logits/rejected": 0.8358896970748901,
1052
+ "logps/chosen": -136.51910400390625,
1053
+ "logps/rejected": -144.94639587402344,
1054
+ "loss": 0.6625,
1055
+ "rewards/accuracies": 0.75,
1056
+ "rewards/chosen": -0.022078264504671097,
1057
+ "rewards/margins": 0.07592067122459412,
1058
+ "rewards/rejected": -0.09799893200397491,
1059
+ "step": 50
1060
+ },
1061
+ {
1062
+ "debug/policy_chosen_logits": 0.2970171570777893,
1063
+ "debug/policy_chosen_logps": -131.04600524902344,
1064
+ "debug/policy_rejected_logits": 0.780690610408783,
1065
+ "debug/policy_rejected_logps": -150.49591064453125,
1066
+ "debug/reference_chosen_logps": -130.48240661621094,
1067
+ "debug/reference_rejected_logps": -142.41004943847656,
1068
+ "epoch": 0.9272727272727272,
1069
+ "grad_norm": 2.7673993762098075,
1070
+ "learning_rate": 5e-07,
1071
+ "logits/chosen": 0.2970171570777893,
1072
+ "logits/rejected": 0.780690610408783,
1073
+ "logps/chosen": -131.04600524902344,
1074
+ "logps/rejected": -150.49591064453125,
1075
+ "loss": 0.6769,
1076
+ "rewards/accuracies": 0.875,
1077
+ "rewards/chosen": -0.005636042915284634,
1078
+ "rewards/margins": 0.07522260397672653,
1079
+ "rewards/rejected": -0.08085864782333374,
1080
+ "step": 51
1081
+ },
1082
+ {
1083
+ "debug/policy_chosen_logits": 0.5221680998802185,
1084
+ "debug/policy_chosen_logps": -143.44683837890625,
1085
+ "debug/policy_rejected_logits": 1.1260162591934204,
1086
+ "debug/policy_rejected_logps": -153.56045532226562,
1087
+ "debug/reference_chosen_logps": -140.38705444335938,
1088
+ "debug/reference_rejected_logps": -145.02029418945312,
1089
+ "epoch": 0.9454545454545454,
1090
+ "grad_norm": 2.977638074764286,
1091
+ "learning_rate": 5e-07,
1092
+ "logits/chosen": 0.5221680998802185,
1093
+ "logits/rejected": 1.1260162591934204,
1094
+ "logps/chosen": -143.44683837890625,
1095
+ "logps/rejected": -153.56045532226562,
1096
+ "loss": 0.6616,
1097
+ "rewards/accuracies": 0.75,
1098
+ "rewards/chosen": -0.0305978674441576,
1099
+ "rewards/margins": 0.05480368435382843,
1100
+ "rewards/rejected": -0.08540154993534088,
1101
+ "step": 52
1102
+ },
1103
+ {
1104
+ "debug/policy_chosen_logits": 0.917246401309967,
1105
+ "debug/policy_chosen_logps": -144.25767517089844,
1106
+ "debug/policy_rejected_logits": 1.1883795261383057,
1107
+ "debug/policy_rejected_logps": -146.60035705566406,
1108
+ "debug/reference_chosen_logps": -142.2523193359375,
1109
+ "debug/reference_rejected_logps": -136.43551635742188,
1110
+ "epoch": 0.9636363636363636,
1111
+ "grad_norm": 2.933035711917657,
1112
+ "learning_rate": 5e-07,
1113
+ "logits/chosen": 0.917246401309967,
1114
+ "logits/rejected": 1.1883795261383057,
1115
+ "logps/chosen": -144.25767517089844,
1116
+ "logps/rejected": -146.60035705566406,
1117
+ "loss": 0.6687,
1118
+ "rewards/accuracies": 0.625,
1119
+ "rewards/chosen": -0.02005346305668354,
1120
+ "rewards/margins": 0.08159493654966354,
1121
+ "rewards/rejected": -0.10164839029312134,
1122
+ "step": 53
1123
+ },
1124
+ {
1125
+ "debug/policy_chosen_logits": 0.44729718565940857,
1126
+ "debug/policy_chosen_logps": -159.4932098388672,
1127
+ "debug/policy_rejected_logits": 0.5373156666755676,
1128
+ "debug/policy_rejected_logps": -145.55308532714844,
1129
+ "debug/reference_chosen_logps": -154.90708923339844,
1130
+ "debug/reference_rejected_logps": -141.65737915039062,
1131
+ "epoch": 0.9818181818181818,
1132
+ "grad_norm": 3.1066391658700105,
1133
+ "learning_rate": 5e-07,
1134
+ "logits/chosen": 0.44729718565940857,
1135
+ "logits/rejected": 0.5373156666755676,
1136
+ "logps/chosen": -159.4932098388672,
1137
+ "logps/rejected": -145.55308532714844,
1138
+ "loss": 0.6641,
1139
+ "rewards/accuracies": 0.375,
1140
+ "rewards/chosen": -0.04586116969585419,
1141
+ "rewards/margins": -0.006904135458171368,
1142
+ "rewards/rejected": -0.038957033306360245,
1143
+ "step": 54
1144
+ },
1145
+ {
1146
+ "debug/policy_chosen_logits": 0.4478115737438202,
1147
+ "debug/policy_chosen_logps": -131.2347412109375,
1148
+ "debug/policy_rejected_logits": 1.390880823135376,
1149
+ "debug/policy_rejected_logps": -158.60877990722656,
1150
+ "debug/reference_chosen_logps": -128.52542114257812,
1151
+ "debug/reference_rejected_logps": -150.23410034179688,
1152
+ "epoch": 1.0,
1153
+ "grad_norm": 3.235468961904597,
1154
+ "learning_rate": 5e-07,
1155
+ "logits/chosen": 0.4478115737438202,
1156
+ "logits/rejected": 1.390880823135376,
1157
+ "logps/chosen": -131.2347412109375,
1158
+ "logps/rejected": -158.60877990722656,
1159
+ "loss": 0.6763,
1160
+ "rewards/accuracies": 0.875,
1161
+ "rewards/chosen": -0.02709309384226799,
1162
+ "rewards/margins": 0.056653641164302826,
1163
+ "rewards/rejected": -0.08374673128128052,
1164
+ "step": 55
1165
+ },
1166
+ {
1167
+ "epoch": 1.0,
1168
+ "step": 55,
1169
+ "total_flos": 0.0,
1170
+ "train_loss": 0.6775240507992831,
1171
+ "train_runtime": 440.7264,
1172
+ "train_samples_per_second": 7.907,
1173
+ "train_steps_per_second": 0.125
1174
+ }
1175
+ ],
1176
+ "logging_steps": 1,
1177
+ "max_steps": 55,
1178
+ "num_input_tokens_seen": 0,
1179
+ "num_train_epochs": 1,
1180
+ "save_steps": 500,
1181
+ "stateful_callbacks": {
1182
+ "TrainerControl": {
1183
+ "args": {
1184
+ "should_epoch_stop": false,
1185
+ "should_evaluate": false,
1186
+ "should_log": false,
1187
+ "should_save": true,
1188
+ "should_training_stop": true
1189
+ },
1190
+ "attributes": {}
1191
+ }
1192
+ },
1193
+ "total_flos": 0.0,
1194
+ "train_batch_size": 8,
1195
+ "trial_name": null,
1196
+ "trial_params": null
1197
+ }