yiran-wang3 commited on
Commit
d5687bc
1 Parent(s): 0a7e7db

End of training

Browse files
README.md ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ license: apache-2.0
4
+ base_model: yiran-wang3/qwen1_chat_adamw_iter4
5
+ tags:
6
+ - alignment-handbook
7
+ - generated_from_trainer
8
+ - trl
9
+ - dpo
10
+ datasets:
11
+ - self-generate/qw1_sppo_hard_new_cn_mining_oj_iter4-binarized
12
+ model-index:
13
+ - name: qwen1_chat_adamw_iter5
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
+ # qwen1_chat_adamw_iter5
21
+
22
+ This model is a fine-tuned version of [yiran-wang3/qwen1_chat_adamw_iter4](https://huggingface.co/yiran-wang3/qwen1_chat_adamw_iter4) on the self-generate/qw1_sppo_hard_new_cn_mining_oj_iter4-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: 1e-06
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 0.20.1
all_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.49933575202779074,
5
+ "train_runtime": 145.5342,
6
+ "train_samples": 2618,
7
+ "train_samples_per_second": 17.989,
8
+ "train_steps_per_second": 0.282
9
+ }
config.json CHANGED
@@ -23,7 +23,7 @@
23
  "tie_word_embeddings": true,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.45.0",
26
- "use_cache": false,
27
  "use_sliding_window": false,
28
  "vocab_size": 151936
29
  }
 
23
  "tie_word_embeddings": true,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.45.0",
26
+ "use_cache": true,
27
  "use_sliding_window": false,
28
  "vocab_size": 151936
29
  }
generation_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 151643,
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 151645,
6
+ 151643
7
+ ],
8
+ "pad_token_id": 151643,
9
+ "repetition_penalty": 1.1,
10
+ "temperature": 0.7,
11
+ "top_k": 20,
12
+ "top_p": 0.8,
13
+ "transformers_version": "4.45.0"
14
+ }
train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.49933575202779074,
5
+ "train_runtime": 145.5342,
6
+ "train_samples": 2618,
7
+ "train_samples_per_second": 17.989,
8
+ "train_steps_per_second": 0.282
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,903 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 100,
6
+ "global_step": 41,
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": -1.0657888650894165,
13
+ "debug/policy_chosen_logps": -135.17166137695312,
14
+ "debug/policy_rejected_logits": -1.0082604885101318,
15
+ "debug/policy_rejected_logps": -164.67491149902344,
16
+ "debug/reference_chosen_logps": -135.17166137695312,
17
+ "debug/reference_rejected_logps": -164.67491149902344,
18
+ "epoch": 0.024390243902439025,
19
+ "grad_norm": 7.455705642028753,
20
+ "learning_rate": 1e-06,
21
+ "logits/chosen": -1.0657888650894165,
22
+ "logits/rejected": -1.0082604885101318,
23
+ "logps/chosen": -135.17166137695312,
24
+ "logps/rejected": -164.67491149902344,
25
+ "loss": 0.5,
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": -0.9316875338554382,
34
+ "debug/policy_chosen_logps": -168.0986785888672,
35
+ "debug/policy_rejected_logits": -0.9154044389724731,
36
+ "debug/policy_rejected_logps": -152.3076934814453,
37
+ "debug/reference_chosen_logps": -168.41038513183594,
38
+ "debug/reference_rejected_logps": -152.78643798828125,
39
+ "epoch": 0.04878048780487805,
40
+ "grad_norm": 7.29758593491183,
41
+ "learning_rate": 1e-06,
42
+ "logits/chosen": -0.9316875338554382,
43
+ "logits/rejected": -0.9154044389724731,
44
+ "logps/chosen": -168.0986785888672,
45
+ "logps/rejected": -152.3076934814453,
46
+ "loss": 0.4997,
47
+ "rewards/accuracies": 0.375,
48
+ "rewards/chosen": 0.003117017913609743,
49
+ "rewards/margins": -0.001670417725108564,
50
+ "rewards/rejected": 0.004787435755133629,
51
+ "step": 2
52
+ },
53
+ {
54
+ "debug/policy_chosen_logits": -1.1543760299682617,
55
+ "debug/policy_chosen_logps": -162.00730895996094,
56
+ "debug/policy_rejected_logits": -0.9926377534866333,
57
+ "debug/policy_rejected_logps": -148.77609252929688,
58
+ "debug/reference_chosen_logps": -161.8112030029297,
59
+ "debug/reference_rejected_logps": -148.79986572265625,
60
+ "epoch": 0.07317073170731707,
61
+ "grad_norm": 7.862425079439745,
62
+ "learning_rate": 1e-06,
63
+ "logits/chosen": -1.1543760299682617,
64
+ "logits/rejected": -0.9926377534866333,
65
+ "logps/chosen": -162.00730895996094,
66
+ "logps/rejected": -148.77609252929688,
67
+ "loss": 0.4992,
68
+ "rewards/accuracies": 0.5,
69
+ "rewards/chosen": -0.0019610594026744366,
70
+ "rewards/margins": -0.0021987915970385075,
71
+ "rewards/rejected": 0.00023773196153342724,
72
+ "step": 3
73
+ },
74
+ {
75
+ "debug/policy_chosen_logits": -0.9400157332420349,
76
+ "debug/policy_chosen_logps": -156.6750030517578,
77
+ "debug/policy_rejected_logits": -1.0246046781539917,
78
+ "debug/policy_rejected_logps": -140.80905151367188,
79
+ "debug/reference_chosen_logps": -156.9263153076172,
80
+ "debug/reference_rejected_logps": -141.17587280273438,
81
+ "epoch": 0.0975609756097561,
82
+ "grad_norm": 8.32764259608944,
83
+ "learning_rate": 1e-06,
84
+ "logits/chosen": -0.9400157332420349,
85
+ "logits/rejected": -1.0246046781539917,
86
+ "logps/chosen": -156.6750030517578,
87
+ "logps/rejected": -140.80905151367188,
88
+ "loss": 0.4991,
89
+ "rewards/accuracies": 0.5,
90
+ "rewards/chosen": 0.0025131413713097572,
91
+ "rewards/margins": -0.0011551285861060023,
92
+ "rewards/rejected": 0.0036682700738310814,
93
+ "step": 4
94
+ },
95
+ {
96
+ "debug/policy_chosen_logits": -0.8135665059089661,
97
+ "debug/policy_chosen_logps": -180.49978637695312,
98
+ "debug/policy_rejected_logits": -0.966259241104126,
99
+ "debug/policy_rejected_logps": -180.0769500732422,
100
+ "debug/reference_chosen_logps": -180.33763122558594,
101
+ "debug/reference_rejected_logps": -180.13702392578125,
102
+ "epoch": 0.12195121951219512,
103
+ "grad_norm": 9.5687290716109,
104
+ "learning_rate": 1e-06,
105
+ "logits/chosen": -0.8135665059089661,
106
+ "logits/rejected": -0.966259241104126,
107
+ "logps/chosen": -180.49978637695312,
108
+ "logps/rejected": -180.0769500732422,
109
+ "loss": 0.5012,
110
+ "rewards/accuracies": 0.5,
111
+ "rewards/chosen": -0.0016215991927310824,
112
+ "rewards/margins": -0.002222432754933834,
113
+ "rewards/rejected": 0.0006008340278640389,
114
+ "step": 5
115
+ },
116
+ {
117
+ "debug/policy_chosen_logits": -0.875511109828949,
118
+ "debug/policy_chosen_logps": -169.20863342285156,
119
+ "debug/policy_rejected_logits": -0.8739838600158691,
120
+ "debug/policy_rejected_logps": -161.75338745117188,
121
+ "debug/reference_chosen_logps": -168.50558471679688,
122
+ "debug/reference_rejected_logps": -160.78341674804688,
123
+ "epoch": 0.14634146341463414,
124
+ "grad_norm": 8.253308110293828,
125
+ "learning_rate": 1e-06,
126
+ "logits/chosen": -0.875511109828949,
127
+ "logits/rejected": -0.8739838600158691,
128
+ "logps/chosen": -169.20863342285156,
129
+ "logps/rejected": -161.75338745117188,
130
+ "loss": 0.5028,
131
+ "rewards/accuracies": 0.625,
132
+ "rewards/chosen": -0.007030477747321129,
133
+ "rewards/margins": 0.0026691434904932976,
134
+ "rewards/rejected": -0.009699621237814426,
135
+ "step": 6
136
+ },
137
+ {
138
+ "debug/policy_chosen_logits": -0.9064666628837585,
139
+ "debug/policy_chosen_logps": -163.56883239746094,
140
+ "debug/policy_rejected_logits": -0.88660728931427,
141
+ "debug/policy_rejected_logps": -167.95022583007812,
142
+ "debug/reference_chosen_logps": -162.90780639648438,
143
+ "debug/reference_rejected_logps": -166.8964080810547,
144
+ "epoch": 0.17073170731707318,
145
+ "grad_norm": 9.148150186683857,
146
+ "learning_rate": 1e-06,
147
+ "logits/chosen": -0.9064666628837585,
148
+ "logits/rejected": -0.88660728931427,
149
+ "logps/chosen": -163.56883239746094,
150
+ "logps/rejected": -167.95022583007812,
151
+ "loss": 0.5021,
152
+ "rewards/accuracies": 0.625,
153
+ "rewards/chosen": -0.0066101741977036,
154
+ "rewards/margins": 0.0039279647171497345,
155
+ "rewards/rejected": -0.010538139380514622,
156
+ "step": 7
157
+ },
158
+ {
159
+ "debug/policy_chosen_logits": -1.0406209230422974,
160
+ "debug/policy_chosen_logps": -145.71910095214844,
161
+ "debug/policy_rejected_logits": -0.9474231004714966,
162
+ "debug/policy_rejected_logps": -176.6788330078125,
163
+ "debug/reference_chosen_logps": -145.38890075683594,
164
+ "debug/reference_rejected_logps": -176.26510620117188,
165
+ "epoch": 0.1951219512195122,
166
+ "grad_norm": 7.873026337514701,
167
+ "learning_rate": 1e-06,
168
+ "logits/chosen": -1.0406209230422974,
169
+ "logits/rejected": -0.9474231004714966,
170
+ "logps/chosen": -145.71910095214844,
171
+ "logps/rejected": -176.6788330078125,
172
+ "loss": 0.4983,
173
+ "rewards/accuracies": 0.375,
174
+ "rewards/chosen": -0.003301925491541624,
175
+ "rewards/margins": 0.0008353232406079769,
176
+ "rewards/rejected": -0.004137249197810888,
177
+ "step": 8
178
+ },
179
+ {
180
+ "debug/policy_chosen_logits": -0.7725314497947693,
181
+ "debug/policy_chosen_logps": -191.3946533203125,
182
+ "debug/policy_rejected_logits": -0.9446115493774414,
183
+ "debug/policy_rejected_logps": -157.00643920898438,
184
+ "debug/reference_chosen_logps": -191.58938598632812,
185
+ "debug/reference_rejected_logps": -157.259521484375,
186
+ "epoch": 0.21951219512195122,
187
+ "grad_norm": 8.113917191223486,
188
+ "learning_rate": 1e-06,
189
+ "logits/chosen": -0.7725314497947693,
190
+ "logits/rejected": -0.9446115493774414,
191
+ "logps/chosen": -191.3946533203125,
192
+ "logps/rejected": -157.00643920898438,
193
+ "loss": 0.4984,
194
+ "rewards/accuracies": 0.625,
195
+ "rewards/chosen": 0.0019471454434096813,
196
+ "rewards/margins": -0.0005835914053022861,
197
+ "rewards/rejected": 0.00253073638305068,
198
+ "step": 9
199
+ },
200
+ {
201
+ "debug/policy_chosen_logits": -0.9539018273353577,
202
+ "debug/policy_chosen_logps": -156.33084106445312,
203
+ "debug/policy_rejected_logits": -0.8661950826644897,
204
+ "debug/policy_rejected_logps": -166.2339630126953,
205
+ "debug/reference_chosen_logps": -156.4985809326172,
206
+ "debug/reference_rejected_logps": -165.98721313476562,
207
+ "epoch": 0.24390243902439024,
208
+ "grad_norm": 8.579040246656625,
209
+ "learning_rate": 1e-06,
210
+ "logits/chosen": -0.9539018273353577,
211
+ "logits/rejected": -0.8661950826644897,
212
+ "logps/chosen": -156.33084106445312,
213
+ "logps/rejected": -166.2339630126953,
214
+ "loss": 0.502,
215
+ "rewards/accuracies": 0.75,
216
+ "rewards/chosen": 0.001677446300163865,
217
+ "rewards/margins": 0.0041448017582297325,
218
+ "rewards/rejected": -0.002467355690896511,
219
+ "step": 10
220
+ },
221
+ {
222
+ "debug/policy_chosen_logits": -0.9540179371833801,
223
+ "debug/policy_chosen_logps": -182.48159790039062,
224
+ "debug/policy_rejected_logits": -0.9883304834365845,
225
+ "debug/policy_rejected_logps": -147.31826782226562,
226
+ "debug/reference_chosen_logps": -184.1331024169922,
227
+ "debug/reference_rejected_logps": -149.22958374023438,
228
+ "epoch": 0.2682926829268293,
229
+ "grad_norm": 8.658564154949419,
230
+ "learning_rate": 1e-06,
231
+ "logits/chosen": -0.9540179371833801,
232
+ "logits/rejected": -0.9883304834365845,
233
+ "logps/chosen": -182.48159790039062,
234
+ "logps/rejected": -147.31826782226562,
235
+ "loss": 0.4978,
236
+ "rewards/accuracies": 0.375,
237
+ "rewards/chosen": 0.016515053808689117,
238
+ "rewards/margins": -0.0025980568025261164,
239
+ "rewards/rejected": 0.01911311037838459,
240
+ "step": 11
241
+ },
242
+ {
243
+ "debug/policy_chosen_logits": -0.9231241345405579,
244
+ "debug/policy_chosen_logps": -173.3875274658203,
245
+ "debug/policy_rejected_logits": -0.9746717810630798,
246
+ "debug/policy_rejected_logps": -186.23223876953125,
247
+ "debug/reference_chosen_logps": -173.79617309570312,
248
+ "debug/reference_rejected_logps": -185.99221801757812,
249
+ "epoch": 0.2926829268292683,
250
+ "grad_norm": 9.641106516800372,
251
+ "learning_rate": 1e-06,
252
+ "logits/chosen": -0.9231241345405579,
253
+ "logits/rejected": -0.9746717810630798,
254
+ "logps/chosen": -173.3875274658203,
255
+ "logps/rejected": -186.23223876953125,
256
+ "loss": 0.4994,
257
+ "rewards/accuracies": 0.75,
258
+ "rewards/chosen": 0.0040865708142519,
259
+ "rewards/margins": 0.006486768834292889,
260
+ "rewards/rejected": -0.0024001982528716326,
261
+ "step": 12
262
+ },
263
+ {
264
+ "debug/policy_chosen_logits": -0.8564068675041199,
265
+ "debug/policy_chosen_logps": -155.58351135253906,
266
+ "debug/policy_rejected_logits": -0.78647381067276,
267
+ "debug/policy_rejected_logps": -180.34735107421875,
268
+ "debug/reference_chosen_logps": -155.46743774414062,
269
+ "debug/reference_rejected_logps": -179.4715576171875,
270
+ "epoch": 0.3170731707317073,
271
+ "grad_norm": 7.778587819967809,
272
+ "learning_rate": 1e-06,
273
+ "logits/chosen": -0.8564068675041199,
274
+ "logits/rejected": -0.78647381067276,
275
+ "logps/chosen": -155.58351135253906,
276
+ "logps/rejected": -180.34735107421875,
277
+ "loss": 0.4997,
278
+ "rewards/accuracies": 0.5,
279
+ "rewards/chosen": -0.0011608502827584743,
280
+ "rewards/margins": 0.007597027346491814,
281
+ "rewards/rejected": -0.008757877163589,
282
+ "step": 13
283
+ },
284
+ {
285
+ "debug/policy_chosen_logits": -0.9291447997093201,
286
+ "debug/policy_chosen_logps": -137.7774658203125,
287
+ "debug/policy_rejected_logits": -0.8161742687225342,
288
+ "debug/policy_rejected_logps": -154.59149169921875,
289
+ "debug/reference_chosen_logps": -136.43026733398438,
290
+ "debug/reference_rejected_logps": -152.96876525878906,
291
+ "epoch": 0.34146341463414637,
292
+ "grad_norm": 9.452087496130199,
293
+ "learning_rate": 1e-06,
294
+ "logits/chosen": -0.9291447997093201,
295
+ "logits/rejected": -0.8161742687225342,
296
+ "logps/chosen": -137.7774658203125,
297
+ "logps/rejected": -154.59149169921875,
298
+ "loss": 0.4991,
299
+ "rewards/accuracies": 0.5,
300
+ "rewards/chosen": -0.013471984304487705,
301
+ "rewards/margins": 0.002755412831902504,
302
+ "rewards/rejected": -0.016227398067712784,
303
+ "step": 14
304
+ },
305
+ {
306
+ "debug/policy_chosen_logits": -0.8452920317649841,
307
+ "debug/policy_chosen_logps": -175.22425842285156,
308
+ "debug/policy_rejected_logits": -0.7951302528381348,
309
+ "debug/policy_rejected_logps": -175.16729736328125,
310
+ "debug/reference_chosen_logps": -174.3819122314453,
311
+ "debug/reference_rejected_logps": -172.59144592285156,
312
+ "epoch": 0.36585365853658536,
313
+ "grad_norm": 8.577281209186165,
314
+ "learning_rate": 1e-06,
315
+ "logits/chosen": -0.8452920317649841,
316
+ "logits/rejected": -0.7951302528381348,
317
+ "logps/chosen": -175.22425842285156,
318
+ "logps/rejected": -175.16729736328125,
319
+ "loss": 0.4999,
320
+ "rewards/accuracies": 0.75,
321
+ "rewards/chosen": -0.008423404768109322,
322
+ "rewards/margins": 0.017334945499897003,
323
+ "rewards/rejected": -0.025758352130651474,
324
+ "step": 15
325
+ },
326
+ {
327
+ "debug/policy_chosen_logits": -0.8255157470703125,
328
+ "debug/policy_chosen_logps": -196.6995391845703,
329
+ "debug/policy_rejected_logits": -1.0168753862380981,
330
+ "debug/policy_rejected_logps": -144.28012084960938,
331
+ "debug/reference_chosen_logps": -194.49815368652344,
332
+ "debug/reference_rejected_logps": -141.6895751953125,
333
+ "epoch": 0.3902439024390244,
334
+ "grad_norm": 10.569580209063629,
335
+ "learning_rate": 1e-06,
336
+ "logits/chosen": -0.8255157470703125,
337
+ "logits/rejected": -1.0168753862380981,
338
+ "logps/chosen": -196.6995391845703,
339
+ "logps/rejected": -144.28012084960938,
340
+ "loss": 0.5055,
341
+ "rewards/accuracies": 0.5,
342
+ "rewards/chosen": -0.02201385423541069,
343
+ "rewards/margins": 0.003891811240464449,
344
+ "rewards/rejected": -0.025905665010213852,
345
+ "step": 16
346
+ },
347
+ {
348
+ "debug/policy_chosen_logits": -0.8963515162467957,
349
+ "debug/policy_chosen_logps": -167.45254516601562,
350
+ "debug/policy_rejected_logits": -1.1090606451034546,
351
+ "debug/policy_rejected_logps": -132.12088012695312,
352
+ "debug/reference_chosen_logps": -165.4241943359375,
353
+ "debug/reference_rejected_logps": -131.6002655029297,
354
+ "epoch": 0.4146341463414634,
355
+ "grad_norm": 8.216685819750042,
356
+ "learning_rate": 1e-06,
357
+ "logits/chosen": -0.8963515162467957,
358
+ "logits/rejected": -1.1090606451034546,
359
+ "logps/chosen": -167.45254516601562,
360
+ "logps/rejected": -132.12088012695312,
361
+ "loss": 0.5023,
362
+ "rewards/accuracies": 0.375,
363
+ "rewards/chosen": -0.020283488556742668,
364
+ "rewards/margins": -0.015077323652803898,
365
+ "rewards/rejected": -0.00520616490393877,
366
+ "step": 17
367
+ },
368
+ {
369
+ "debug/policy_chosen_logits": -1.1278187036514282,
370
+ "debug/policy_chosen_logps": -163.56838989257812,
371
+ "debug/policy_rejected_logits": -0.7523326277732849,
372
+ "debug/policy_rejected_logps": -194.7676239013672,
373
+ "debug/reference_chosen_logps": -164.0474395751953,
374
+ "debug/reference_rejected_logps": -195.21795654296875,
375
+ "epoch": 0.43902439024390244,
376
+ "grad_norm": 8.12497169240311,
377
+ "learning_rate": 1e-06,
378
+ "logits/chosen": -1.1278187036514282,
379
+ "logits/rejected": -0.7523326277732849,
380
+ "logps/chosen": -163.56838989257812,
381
+ "logps/rejected": -194.7676239013672,
382
+ "loss": 0.5046,
383
+ "rewards/accuracies": 0.5,
384
+ "rewards/chosen": 0.004790563136339188,
385
+ "rewards/margins": 0.0002871984615921974,
386
+ "rewards/rejected": 0.00450336467474699,
387
+ "step": 18
388
+ },
389
+ {
390
+ "debug/policy_chosen_logits": -0.8814194202423096,
391
+ "debug/policy_chosen_logps": -187.4921417236328,
392
+ "debug/policy_rejected_logits": -0.9221204519271851,
393
+ "debug/policy_rejected_logps": -184.78768920898438,
394
+ "debug/reference_chosen_logps": -187.89466857910156,
395
+ "debug/reference_rejected_logps": -186.1348876953125,
396
+ "epoch": 0.4634146341463415,
397
+ "grad_norm": 8.289942810390798,
398
+ "learning_rate": 1e-06,
399
+ "logits/chosen": -0.8814194202423096,
400
+ "logits/rejected": -0.9221204519271851,
401
+ "logps/chosen": -187.4921417236328,
402
+ "logps/rejected": -184.78768920898438,
403
+ "loss": 0.4986,
404
+ "rewards/accuracies": 0.375,
405
+ "rewards/chosen": 0.004025382921099663,
406
+ "rewards/margins": -0.009446544572710991,
407
+ "rewards/rejected": 0.013471927493810654,
408
+ "step": 19
409
+ },
410
+ {
411
+ "debug/policy_chosen_logits": -0.9697524905204773,
412
+ "debug/policy_chosen_logps": -142.56431579589844,
413
+ "debug/policy_rejected_logits": -0.7087419033050537,
414
+ "debug/policy_rejected_logps": -181.8319091796875,
415
+ "debug/reference_chosen_logps": -145.44419860839844,
416
+ "debug/reference_rejected_logps": -184.153564453125,
417
+ "epoch": 0.4878048780487805,
418
+ "grad_norm": 9.078292331479483,
419
+ "learning_rate": 1e-06,
420
+ "logits/chosen": -0.9697524905204773,
421
+ "logits/rejected": -0.7087419033050537,
422
+ "logps/chosen": -142.56431579589844,
423
+ "logps/rejected": -181.8319091796875,
424
+ "loss": 0.4974,
425
+ "rewards/accuracies": 0.625,
426
+ "rewards/chosen": 0.028798799961805344,
427
+ "rewards/margins": 0.005582190118730068,
428
+ "rewards/rejected": 0.0232166089117527,
429
+ "step": 20
430
+ },
431
+ {
432
+ "debug/policy_chosen_logits": -0.8631553649902344,
433
+ "debug/policy_chosen_logps": -181.8375701904297,
434
+ "debug/policy_rejected_logits": -0.8101003170013428,
435
+ "debug/policy_rejected_logps": -204.69305419921875,
436
+ "debug/reference_chosen_logps": -181.6129150390625,
437
+ "debug/reference_rejected_logps": -206.70823669433594,
438
+ "epoch": 0.5121951219512195,
439
+ "grad_norm": 9.63889554364432,
440
+ "learning_rate": 1e-06,
441
+ "logits/chosen": -0.8631553649902344,
442
+ "logits/rejected": -0.8101003170013428,
443
+ "logps/chosen": -181.8375701904297,
444
+ "logps/rejected": -204.69305419921875,
445
+ "loss": 0.5031,
446
+ "rewards/accuracies": 0.125,
447
+ "rewards/chosen": -0.0022465987130999565,
448
+ "rewards/margins": -0.02239835634827614,
449
+ "rewards/rejected": 0.020151756703853607,
450
+ "step": 21
451
+ },
452
+ {
453
+ "debug/policy_chosen_logits": -0.900942862033844,
454
+ "debug/policy_chosen_logps": -156.7996826171875,
455
+ "debug/policy_rejected_logits": -0.7672395706176758,
456
+ "debug/policy_rejected_logps": -190.6102294921875,
457
+ "debug/reference_chosen_logps": -157.8141326904297,
458
+ "debug/reference_rejected_logps": -191.65597534179688,
459
+ "epoch": 0.5365853658536586,
460
+ "grad_norm": 8.86160952811184,
461
+ "learning_rate": 1e-06,
462
+ "logits/chosen": -0.900942862033844,
463
+ "logits/rejected": -0.7672395706176758,
464
+ "logps/chosen": -156.7996826171875,
465
+ "logps/rejected": -190.6102294921875,
466
+ "loss": 0.4989,
467
+ "rewards/accuracies": 0.375,
468
+ "rewards/chosen": 0.010144614614546299,
469
+ "rewards/margins": -0.0003129197284579277,
470
+ "rewards/rejected": 0.010457534343004227,
471
+ "step": 22
472
+ },
473
+ {
474
+ "debug/policy_chosen_logits": -0.8332923054695129,
475
+ "debug/policy_chosen_logps": -155.85830688476562,
476
+ "debug/policy_rejected_logits": -0.7771956324577332,
477
+ "debug/policy_rejected_logps": -165.27749633789062,
478
+ "debug/reference_chosen_logps": -156.92462158203125,
479
+ "debug/reference_rejected_logps": -165.16729736328125,
480
+ "epoch": 0.5609756097560976,
481
+ "grad_norm": 8.143779892532098,
482
+ "learning_rate": 1e-06,
483
+ "logits/chosen": -0.8332923054695129,
484
+ "logits/rejected": -0.7771956324577332,
485
+ "logps/chosen": -155.85830688476562,
486
+ "logps/rejected": -165.27749633789062,
487
+ "loss": 0.5003,
488
+ "rewards/accuracies": 0.625,
489
+ "rewards/chosen": 0.0106631089001894,
490
+ "rewards/margins": 0.011765326373279095,
491
+ "rewards/rejected": -0.001102218870073557,
492
+ "step": 23
493
+ },
494
+ {
495
+ "debug/policy_chosen_logits": -1.01643705368042,
496
+ "debug/policy_chosen_logps": -142.14419555664062,
497
+ "debug/policy_rejected_logits": -0.9811778664588928,
498
+ "debug/policy_rejected_logps": -161.79629516601562,
499
+ "debug/reference_chosen_logps": -141.1592559814453,
500
+ "debug/reference_rejected_logps": -159.3284149169922,
501
+ "epoch": 0.5853658536585366,
502
+ "grad_norm": 8.084865068688668,
503
+ "learning_rate": 1e-06,
504
+ "logits/chosen": -1.01643705368042,
505
+ "logits/rejected": -0.9811778664588928,
506
+ "logps/chosen": -142.14419555664062,
507
+ "logps/rejected": -161.79629516601562,
508
+ "loss": 0.5022,
509
+ "rewards/accuracies": 0.875,
510
+ "rewards/chosen": -0.009849337860941887,
511
+ "rewards/margins": 0.014829404652118683,
512
+ "rewards/rejected": -0.02467874437570572,
513
+ "step": 24
514
+ },
515
+ {
516
+ "debug/policy_chosen_logits": -0.7919837236404419,
517
+ "debug/policy_chosen_logps": -162.75352478027344,
518
+ "debug/policy_rejected_logits": -0.96061772108078,
519
+ "debug/policy_rejected_logps": -169.0644073486328,
520
+ "debug/reference_chosen_logps": -161.2084503173828,
521
+ "debug/reference_rejected_logps": -167.456787109375,
522
+ "epoch": 0.6097560975609756,
523
+ "grad_norm": 9.111682216370516,
524
+ "learning_rate": 1e-06,
525
+ "logits/chosen": -0.7919837236404419,
526
+ "logits/rejected": -0.96061772108078,
527
+ "logps/chosen": -162.75352478027344,
528
+ "logps/rejected": -169.0644073486328,
529
+ "loss": 0.5016,
530
+ "rewards/accuracies": 0.5,
531
+ "rewards/chosen": -0.015450754202902317,
532
+ "rewards/margins": 0.0006255432963371277,
533
+ "rewards/rejected": -0.01607629843056202,
534
+ "step": 25
535
+ },
536
+ {
537
+ "debug/policy_chosen_logits": -0.9698644876480103,
538
+ "debug/policy_chosen_logps": -165.7355194091797,
539
+ "debug/policy_rejected_logits": -0.7252393364906311,
540
+ "debug/policy_rejected_logps": -183.15887451171875,
541
+ "debug/reference_chosen_logps": -164.15481567382812,
542
+ "debug/reference_rejected_logps": -182.38124084472656,
543
+ "epoch": 0.6341463414634146,
544
+ "grad_norm": 10.115977003883836,
545
+ "learning_rate": 1e-06,
546
+ "logits/chosen": -0.9698644876480103,
547
+ "logits/rejected": -0.7252393364906311,
548
+ "logps/chosen": -165.7355194091797,
549
+ "logps/rejected": -183.15887451171875,
550
+ "loss": 0.502,
551
+ "rewards/accuracies": 0.5,
552
+ "rewards/chosen": -0.015806876122951508,
553
+ "rewards/margins": -0.008030500262975693,
554
+ "rewards/rejected": -0.00777637492865324,
555
+ "step": 26
556
+ },
557
+ {
558
+ "debug/policy_chosen_logits": -0.7707141637802124,
559
+ "debug/policy_chosen_logps": -205.21841430664062,
560
+ "debug/policy_rejected_logits": -0.9271693229675293,
561
+ "debug/policy_rejected_logps": -168.64834594726562,
562
+ "debug/reference_chosen_logps": -205.48428344726562,
563
+ "debug/reference_rejected_logps": -168.75538635253906,
564
+ "epoch": 0.6585365853658537,
565
+ "grad_norm": 9.132022098455337,
566
+ "learning_rate": 1e-06,
567
+ "logits/chosen": -0.7707141637802124,
568
+ "logits/rejected": -0.9271693229675293,
569
+ "logps/chosen": -205.21841430664062,
570
+ "logps/rejected": -168.64834594726562,
571
+ "loss": 0.4964,
572
+ "rewards/accuracies": 0.625,
573
+ "rewards/chosen": 0.002658710815012455,
574
+ "rewards/margins": 0.0015882402658462524,
575
+ "rewards/rejected": 0.0010704714804887772,
576
+ "step": 27
577
+ },
578
+ {
579
+ "debug/policy_chosen_logits": -0.9867036938667297,
580
+ "debug/policy_chosen_logps": -170.24920654296875,
581
+ "debug/policy_rejected_logits": -0.958678126335144,
582
+ "debug/policy_rejected_logps": -172.13462829589844,
583
+ "debug/reference_chosen_logps": -169.69461059570312,
584
+ "debug/reference_rejected_logps": -171.46791076660156,
585
+ "epoch": 0.6829268292682927,
586
+ "grad_norm": 8.165335264629883,
587
+ "learning_rate": 1e-06,
588
+ "logits/chosen": -0.9867036938667297,
589
+ "logits/rejected": -0.958678126335144,
590
+ "logps/chosen": -170.24920654296875,
591
+ "logps/rejected": -172.13462829589844,
592
+ "loss": 0.4934,
593
+ "rewards/accuracies": 0.375,
594
+ "rewards/chosen": -0.005545825697481632,
595
+ "rewards/margins": 0.0011212928220629692,
596
+ "rewards/rejected": -0.006667118053883314,
597
+ "step": 28
598
+ },
599
+ {
600
+ "debug/policy_chosen_logits": -0.6699000000953674,
601
+ "debug/policy_chosen_logps": -189.43618774414062,
602
+ "debug/policy_rejected_logits": -0.8863887786865234,
603
+ "debug/policy_rejected_logps": -179.74417114257812,
604
+ "debug/reference_chosen_logps": -188.31427001953125,
605
+ "debug/reference_rejected_logps": -179.5847625732422,
606
+ "epoch": 0.7073170731707317,
607
+ "grad_norm": 8.679731931376267,
608
+ "learning_rate": 1e-06,
609
+ "logits/chosen": -0.6699000000953674,
610
+ "logits/rejected": -0.8863887786865234,
611
+ "logps/chosen": -189.43618774414062,
612
+ "logps/rejected": -179.74417114257812,
613
+ "loss": 0.5078,
614
+ "rewards/accuracies": 0.375,
615
+ "rewards/chosen": -0.011219177395105362,
616
+ "rewards/margins": -0.009625071659684181,
617
+ "rewards/rejected": -0.0015941057354211807,
618
+ "step": 29
619
+ },
620
+ {
621
+ "debug/policy_chosen_logits": -0.9941346645355225,
622
+ "debug/policy_chosen_logps": -139.71783447265625,
623
+ "debug/policy_rejected_logits": -0.9757084846496582,
624
+ "debug/policy_rejected_logps": -147.81222534179688,
625
+ "debug/reference_chosen_logps": -140.21261596679688,
626
+ "debug/reference_rejected_logps": -150.31747436523438,
627
+ "epoch": 0.7317073170731707,
628
+ "grad_norm": 8.152161947975058,
629
+ "learning_rate": 1e-06,
630
+ "logits/chosen": -0.9941346645355225,
631
+ "logits/rejected": -0.9757084846496582,
632
+ "logps/chosen": -139.71783447265625,
633
+ "logps/rejected": -147.81222534179688,
634
+ "loss": 0.4989,
635
+ "rewards/accuracies": 0.375,
636
+ "rewards/chosen": 0.0049479007720947266,
637
+ "rewards/margins": -0.020104561001062393,
638
+ "rewards/rejected": 0.02505246177315712,
639
+ "step": 30
640
+ },
641
+ {
642
+ "debug/policy_chosen_logits": -0.9950529932975769,
643
+ "debug/policy_chosen_logps": -167.46485900878906,
644
+ "debug/policy_rejected_logits": -0.9336649179458618,
645
+ "debug/policy_rejected_logps": -161.46160888671875,
646
+ "debug/reference_chosen_logps": -170.01150512695312,
647
+ "debug/reference_rejected_logps": -164.1259765625,
648
+ "epoch": 0.7560975609756098,
649
+ "grad_norm": 8.534988911622145,
650
+ "learning_rate": 1e-06,
651
+ "logits/chosen": -0.9950529932975769,
652
+ "logits/rejected": -0.9336649179458618,
653
+ "logps/chosen": -167.46485900878906,
654
+ "logps/rejected": -161.46160888671875,
655
+ "loss": 0.4947,
656
+ "rewards/accuracies": 0.375,
657
+ "rewards/chosen": 0.02546636573970318,
658
+ "rewards/margins": -0.0011772820726037025,
659
+ "rewards/rejected": 0.026643646880984306,
660
+ "step": 31
661
+ },
662
+ {
663
+ "debug/policy_chosen_logits": -0.99274080991745,
664
+ "debug/policy_chosen_logps": -141.17715454101562,
665
+ "debug/policy_rejected_logits": -0.9643717408180237,
666
+ "debug/policy_rejected_logps": -163.40673828125,
667
+ "debug/reference_chosen_logps": -142.4697723388672,
668
+ "debug/reference_rejected_logps": -163.57907104492188,
669
+ "epoch": 0.7804878048780488,
670
+ "grad_norm": 8.868022634011657,
671
+ "learning_rate": 1e-06,
672
+ "logits/chosen": -0.99274080991745,
673
+ "logits/rejected": -0.9643717408180237,
674
+ "logps/chosen": -141.17715454101562,
675
+ "logps/rejected": -163.40673828125,
676
+ "loss": 0.4887,
677
+ "rewards/accuracies": 0.625,
678
+ "rewards/chosen": 0.012926094233989716,
679
+ "rewards/margins": 0.01120278425514698,
680
+ "rewards/rejected": 0.0017233085818588734,
681
+ "step": 32
682
+ },
683
+ {
684
+ "debug/policy_chosen_logits": -0.7678695321083069,
685
+ "debug/policy_chosen_logps": -179.30178833007812,
686
+ "debug/policy_rejected_logits": -0.9433090686798096,
687
+ "debug/policy_rejected_logps": -166.5570068359375,
688
+ "debug/reference_chosen_logps": -178.72622680664062,
689
+ "debug/reference_rejected_logps": -164.87799072265625,
690
+ "epoch": 0.8048780487804879,
691
+ "grad_norm": 8.118291328693896,
692
+ "learning_rate": 1e-06,
693
+ "logits/chosen": -0.7678695321083069,
694
+ "logits/rejected": -0.9433090686798096,
695
+ "logps/chosen": -179.30178833007812,
696
+ "logps/rejected": -166.5570068359375,
697
+ "loss": 0.4949,
698
+ "rewards/accuracies": 0.625,
699
+ "rewards/chosen": -0.005755499936640263,
700
+ "rewards/margins": 0.011034755036234856,
701
+ "rewards/rejected": -0.016790255904197693,
702
+ "step": 33
703
+ },
704
+ {
705
+ "debug/policy_chosen_logits": -0.9250268936157227,
706
+ "debug/policy_chosen_logps": -174.24758911132812,
707
+ "debug/policy_rejected_logits": -0.8358609676361084,
708
+ "debug/policy_rejected_logps": -207.80772399902344,
709
+ "debug/reference_chosen_logps": -174.22793579101562,
710
+ "debug/reference_rejected_logps": -206.686279296875,
711
+ "epoch": 0.8292682926829268,
712
+ "grad_norm": 9.100751206625086,
713
+ "learning_rate": 1e-06,
714
+ "logits/chosen": -0.9250268936157227,
715
+ "logits/rejected": -0.8358609676361084,
716
+ "logps/chosen": -174.24758911132812,
717
+ "logps/rejected": -207.80772399902344,
718
+ "loss": 0.5021,
719
+ "rewards/accuracies": 0.75,
720
+ "rewards/chosen": -0.00019640009850263596,
721
+ "rewards/margins": 0.01101800799369812,
722
+ "rewards/rejected": -0.01121440902352333,
723
+ "step": 34
724
+ },
725
+ {
726
+ "debug/policy_chosen_logits": -0.9135526418685913,
727
+ "debug/policy_chosen_logps": -159.21298217773438,
728
+ "debug/policy_rejected_logits": -1.1572299003601074,
729
+ "debug/policy_rejected_logps": -147.13705444335938,
730
+ "debug/reference_chosen_logps": -159.4290771484375,
731
+ "debug/reference_rejected_logps": -145.97109985351562,
732
+ "epoch": 0.8536585365853658,
733
+ "grad_norm": 8.661910409548431,
734
+ "learning_rate": 1e-06,
735
+ "logits/chosen": -0.9135526418685913,
736
+ "logits/rejected": -1.1572299003601074,
737
+ "logps/chosen": -159.21298217773438,
738
+ "logps/rejected": -147.13705444335938,
739
+ "loss": 0.4925,
740
+ "rewards/accuracies": 0.5,
741
+ "rewards/chosen": 0.0021607973612844944,
742
+ "rewards/margins": 0.013820314779877663,
743
+ "rewards/rejected": -0.011659517884254456,
744
+ "step": 35
745
+ },
746
+ {
747
+ "debug/policy_chosen_logits": -0.706568717956543,
748
+ "debug/policy_chosen_logps": -185.29225158691406,
749
+ "debug/policy_rejected_logits": -0.772286593914032,
750
+ "debug/policy_rejected_logps": -181.357177734375,
751
+ "debug/reference_chosen_logps": -186.060302734375,
752
+ "debug/reference_rejected_logps": -181.81985473632812,
753
+ "epoch": 0.8780487804878049,
754
+ "grad_norm": 8.794546491268,
755
+ "learning_rate": 1e-06,
756
+ "logits/chosen": -0.706568717956543,
757
+ "logits/rejected": -0.772286593914032,
758
+ "logps/chosen": -185.29225158691406,
759
+ "logps/rejected": -181.357177734375,
760
+ "loss": 0.4963,
761
+ "rewards/accuracies": 0.625,
762
+ "rewards/chosen": 0.007680453825742006,
763
+ "rewards/margins": 0.0030536362901329994,
764
+ "rewards/rejected": 0.004626817535609007,
765
+ "step": 36
766
+ },
767
+ {
768
+ "debug/policy_chosen_logits": -0.9353126883506775,
769
+ "debug/policy_chosen_logps": -148.95077514648438,
770
+ "debug/policy_rejected_logits": -0.8417949080467224,
771
+ "debug/policy_rejected_logps": -192.06149291992188,
772
+ "debug/reference_chosen_logps": -148.9204864501953,
773
+ "debug/reference_rejected_logps": -192.1736297607422,
774
+ "epoch": 0.9024390243902439,
775
+ "grad_norm": 9.147337941113607,
776
+ "learning_rate": 1e-06,
777
+ "logits/chosen": -0.9353126883506775,
778
+ "logits/rejected": -0.8417949080467224,
779
+ "logps/chosen": -148.95077514648438,
780
+ "logps/rejected": -192.06149291992188,
781
+ "loss": 0.4959,
782
+ "rewards/accuracies": 0.625,
783
+ "rewards/chosen": -0.00030298251658678055,
784
+ "rewards/margins": -0.001424331683665514,
785
+ "rewards/rejected": 0.0011213491670787334,
786
+ "step": 37
787
+ },
788
+ {
789
+ "debug/policy_chosen_logits": -0.9579962491989136,
790
+ "debug/policy_chosen_logps": -145.99325561523438,
791
+ "debug/policy_rejected_logits": -0.771964430809021,
792
+ "debug/policy_rejected_logps": -181.51898193359375,
793
+ "debug/reference_chosen_logps": -146.1923828125,
794
+ "debug/reference_rejected_logps": -181.7994384765625,
795
+ "epoch": 0.926829268292683,
796
+ "grad_norm": 8.770906218078656,
797
+ "learning_rate": 1e-06,
798
+ "logits/chosen": -0.9579962491989136,
799
+ "logits/rejected": -0.771964430809021,
800
+ "logps/chosen": -145.99325561523438,
801
+ "logps/rejected": -181.51898193359375,
802
+ "loss": 0.5028,
803
+ "rewards/accuracies": 0.375,
804
+ "rewards/chosen": 0.0019913101568818092,
805
+ "rewards/margins": -0.0008130739443004131,
806
+ "rewards/rejected": 0.0028043838683515787,
807
+ "step": 38
808
+ },
809
+ {
810
+ "debug/policy_chosen_logits": -0.7748340964317322,
811
+ "debug/policy_chosen_logps": -192.96409606933594,
812
+ "debug/policy_rejected_logits": -0.7723251581192017,
813
+ "debug/policy_rejected_logps": -193.10032653808594,
814
+ "debug/reference_chosen_logps": -194.00680541992188,
815
+ "debug/reference_rejected_logps": -195.18478393554688,
816
+ "epoch": 0.9512195121951219,
817
+ "grad_norm": 9.455831322622839,
818
+ "learning_rate": 1e-06,
819
+ "logits/chosen": -0.7748340964317322,
820
+ "logits/rejected": -0.7723251581192017,
821
+ "logps/chosen": -192.96409606933594,
822
+ "logps/rejected": -193.10032653808594,
823
+ "loss": 0.4989,
824
+ "rewards/accuracies": 0.375,
825
+ "rewards/chosen": 0.010427169501781464,
826
+ "rewards/margins": -0.010417431592941284,
827
+ "rewards/rejected": 0.020844601094722748,
828
+ "step": 39
829
+ },
830
+ {
831
+ "debug/policy_chosen_logits": -0.9706252813339233,
832
+ "debug/policy_chosen_logps": -156.927490234375,
833
+ "debug/policy_rejected_logits": -0.8014242649078369,
834
+ "debug/policy_rejected_logps": -197.02548217773438,
835
+ "debug/reference_chosen_logps": -155.9664764404297,
836
+ "debug/reference_rejected_logps": -197.78335571289062,
837
+ "epoch": 0.975609756097561,
838
+ "grad_norm": 9.78790232486535,
839
+ "learning_rate": 1e-06,
840
+ "logits/chosen": -0.9706252813339233,
841
+ "logits/rejected": -0.8014242649078369,
842
+ "logps/chosen": -156.927490234375,
843
+ "logps/rejected": -197.02548217773438,
844
+ "loss": 0.5018,
845
+ "rewards/accuracies": 0.25,
846
+ "rewards/chosen": -0.009610195644199848,
847
+ "rewards/margins": -0.017188798636198044,
848
+ "rewards/rejected": 0.007578602526336908,
849
+ "step": 40
850
+ },
851
+ {
852
+ "debug/policy_chosen_logits": -0.7953402400016785,
853
+ "debug/policy_chosen_logps": -170.52337646484375,
854
+ "debug/policy_rejected_logits": -0.8171042799949646,
855
+ "debug/policy_rejected_logps": -175.24819946289062,
856
+ "debug/reference_chosen_logps": -169.29534912109375,
857
+ "debug/reference_rejected_logps": -173.3860626220703,
858
+ "epoch": 1.0,
859
+ "grad_norm": 10.67828699749108,
860
+ "learning_rate": 1e-06,
861
+ "logits/chosen": -0.7953402400016785,
862
+ "logits/rejected": -0.8171042799949646,
863
+ "logps/chosen": -170.52337646484375,
864
+ "logps/rejected": -175.24819946289062,
865
+ "loss": 0.4925,
866
+ "rewards/accuracies": 0.625,
867
+ "rewards/chosen": -0.01228022575378418,
868
+ "rewards/margins": 0.006341142114251852,
869
+ "rewards/rejected": -0.01862136647105217,
870
+ "step": 41
871
+ },
872
+ {
873
+ "epoch": 1.0,
874
+ "step": 41,
875
+ "total_flos": 0.0,
876
+ "train_loss": 0.49933575202779074,
877
+ "train_runtime": 145.5342,
878
+ "train_samples_per_second": 17.989,
879
+ "train_steps_per_second": 0.282
880
+ }
881
+ ],
882
+ "logging_steps": 1,
883
+ "max_steps": 41,
884
+ "num_input_tokens_seen": 0,
885
+ "num_train_epochs": 1,
886
+ "save_steps": 500,
887
+ "stateful_callbacks": {
888
+ "TrainerControl": {
889
+ "args": {
890
+ "should_epoch_stop": false,
891
+ "should_evaluate": false,
892
+ "should_log": false,
893
+ "should_save": true,
894
+ "should_training_stop": true
895
+ },
896
+ "attributes": {}
897
+ }
898
+ },
899
+ "total_flos": 0.0,
900
+ "train_batch_size": 8,
901
+ "trial_name": null,
902
+ "trial_params": null
903
+ }