yiran-wang3 commited on
Commit
f799945
1 Parent(s): 313ac4d

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/qwen2_chat_reflct_adamw_iter5
5
+ tags:
6
+ - alignment-handbook
7
+ - generated_from_trainer
8
+ - trl
9
+ - dpo
10
+ datasets:
11
+ - self-generate/qw2_reflct_sppo_hard_new_cn_mining_oj_iter5-binarized-reflection-scored
12
+ model-index:
13
+ - name: qwen2_chat_reflct_adamw_iter6
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
+ # qwen2_chat_reflct_adamw_iter6
21
+
22
+ This model is a fine-tuned version of [yiran-wang3/qwen2_chat_reflct_adamw_iter5](https://huggingface.co/yiran-wang3/qwen2_chat_reflct_adamw_iter5) on the self-generate/qw2_reflct_sppo_hard_new_cn_mining_oj_iter5-binarized-reflection-scored 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 0.20.2
all_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.49317165862682255,
5
+ "train_runtime": 155.8617,
6
+ "train_samples": 2719,
7
+ "train_samples_per_second": 17.445,
8
+ "train_steps_per_second": 0.276
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": 0.49317165862682255,
5
+ "train_runtime": 155.8617,
6
+ "train_samples": 2719,
7
+ "train_samples_per_second": 17.445,
8
+ "train_steps_per_second": 0.276
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,945 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 100,
6
+ "global_step": 43,
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": -1.5398600101470947,
13
+ "debug/policy_chosen_logps": -263.839111328125,
14
+ "debug/policy_rejected_logits": -1.3813989162445068,
15
+ "debug/policy_rejected_logps": -320.2253112792969,
16
+ "debug/reference_chosen_logps": -263.839111328125,
17
+ "debug/reference_rejected_logps": -320.2253112792969,
18
+ "epoch": 0.023255813953488372,
19
+ "grad_norm": 17.674977894001827,
20
+ "learning_rate": 1e-06,
21
+ "logits/chosen": -1.5398600101470947,
22
+ "logits/rejected": -1.3813989162445068,
23
+ "logps/chosen": -263.839111328125,
24
+ "logps/rejected": -320.2253112792969,
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": -1.61044442653656,
34
+ "debug/policy_chosen_logps": -222.2188720703125,
35
+ "debug/policy_rejected_logits": -1.3342773914337158,
36
+ "debug/policy_rejected_logps": -305.922607421875,
37
+ "debug/reference_chosen_logps": -221.88748168945312,
38
+ "debug/reference_rejected_logps": -306.1925048828125,
39
+ "epoch": 0.046511627906976744,
40
+ "grad_norm": 40.70183376910324,
41
+ "learning_rate": 1e-06,
42
+ "logits/chosen": -1.61044442653656,
43
+ "logits/rejected": -1.3342773914337158,
44
+ "logps/chosen": -222.2188720703125,
45
+ "logps/rejected": -305.922607421875,
46
+ "loss": 0.5035,
47
+ "rewards/accuracies": 0.375,
48
+ "rewards/chosen": -0.003313712775707245,
49
+ "rewards/margins": -0.006012477912008762,
50
+ "rewards/rejected": 0.00269876467064023,
51
+ "step": 2
52
+ },
53
+ {
54
+ "debug/policy_chosen_logits": -1.4612830877304077,
55
+ "debug/policy_chosen_logps": -211.158935546875,
56
+ "debug/policy_rejected_logits": -1.4117305278778076,
57
+ "debug/policy_rejected_logps": -295.76220703125,
58
+ "debug/reference_chosen_logps": -210.71981811523438,
59
+ "debug/reference_rejected_logps": -295.57293701171875,
60
+ "epoch": 0.06976744186046512,
61
+ "grad_norm": 18.75915861796485,
62
+ "learning_rate": 1e-06,
63
+ "logits/chosen": -1.4612830877304077,
64
+ "logits/rejected": -1.4117305278778076,
65
+ "logps/chosen": -211.158935546875,
66
+ "logps/rejected": -295.76220703125,
67
+ "loss": 0.5071,
68
+ "rewards/accuracies": 0.5,
69
+ "rewards/chosen": -0.004391231574118137,
70
+ "rewards/margins": -0.0024983598850667477,
71
+ "rewards/rejected": -0.001892871456220746,
72
+ "step": 3
73
+ },
74
+ {
75
+ "debug/policy_chosen_logits": -1.613046646118164,
76
+ "debug/policy_chosen_logps": -302.75689697265625,
77
+ "debug/policy_rejected_logits": -1.5550684928894043,
78
+ "debug/policy_rejected_logps": -374.60919189453125,
79
+ "debug/reference_chosen_logps": -300.4857177734375,
80
+ "debug/reference_rejected_logps": -372.85357666015625,
81
+ "epoch": 0.09302325581395349,
82
+ "grad_norm": 16.079388960546417,
83
+ "learning_rate": 1e-06,
84
+ "logits/chosen": -1.613046646118164,
85
+ "logits/rejected": -1.5550684928894043,
86
+ "logps/chosen": -302.75689697265625,
87
+ "logps/rejected": -374.60919189453125,
88
+ "loss": 0.503,
89
+ "rewards/accuracies": 0.25,
90
+ "rewards/chosen": -0.02271179109811783,
91
+ "rewards/margins": -0.005155391059815884,
92
+ "rewards/rejected": -0.01755640096962452,
93
+ "step": 4
94
+ },
95
+ {
96
+ "debug/policy_chosen_logits": -1.453744649887085,
97
+ "debug/policy_chosen_logps": -324.67041015625,
98
+ "debug/policy_rejected_logits": -1.337669849395752,
99
+ "debug/policy_rejected_logps": -347.67181396484375,
100
+ "debug/reference_chosen_logps": -320.1375732421875,
101
+ "debug/reference_rejected_logps": -345.66351318359375,
102
+ "epoch": 0.11627906976744186,
103
+ "grad_norm": 41.72011905946601,
104
+ "learning_rate": 1e-06,
105
+ "logits/chosen": -1.453744649887085,
106
+ "logits/rejected": -1.337669849395752,
107
+ "logps/chosen": -324.67041015625,
108
+ "logps/rejected": -347.67181396484375,
109
+ "loss": 0.5138,
110
+ "rewards/accuracies": 0.5,
111
+ "rewards/chosen": -0.04532808065414429,
112
+ "rewards/margins": -0.02524520829319954,
113
+ "rewards/rejected": -0.020082874223589897,
114
+ "step": 5
115
+ },
116
+ {
117
+ "debug/policy_chosen_logits": -1.6281760931015015,
118
+ "debug/policy_chosen_logps": -235.80917358398438,
119
+ "debug/policy_rejected_logits": -1.4969021081924438,
120
+ "debug/policy_rejected_logps": -356.2055358886719,
121
+ "debug/reference_chosen_logps": -230.56964111328125,
122
+ "debug/reference_rejected_logps": -353.9713134765625,
123
+ "epoch": 0.13953488372093023,
124
+ "grad_norm": 51.19218432335038,
125
+ "learning_rate": 1e-06,
126
+ "logits/chosen": -1.6281760931015015,
127
+ "logits/rejected": -1.4969021081924438,
128
+ "logps/chosen": -235.80917358398438,
129
+ "logps/rejected": -356.2055358886719,
130
+ "loss": 0.5101,
131
+ "rewards/accuracies": 0.625,
132
+ "rewards/chosen": -0.0523955337703228,
133
+ "rewards/margins": -0.030053365975618362,
134
+ "rewards/rejected": -0.022342165932059288,
135
+ "step": 6
136
+ },
137
+ {
138
+ "debug/policy_chosen_logits": -1.3837248086929321,
139
+ "debug/policy_chosen_logps": -246.68148803710938,
140
+ "debug/policy_rejected_logits": -1.1755256652832031,
141
+ "debug/policy_rejected_logps": -349.03863525390625,
142
+ "debug/reference_chosen_logps": -243.39242553710938,
143
+ "debug/reference_rejected_logps": -341.8122863769531,
144
+ "epoch": 0.16279069767441862,
145
+ "grad_norm": 87.4581502140832,
146
+ "learning_rate": 1e-06,
147
+ "logits/chosen": -1.3837248086929321,
148
+ "logits/rejected": -1.1755256652832031,
149
+ "logps/chosen": -246.68148803710938,
150
+ "logps/rejected": -349.03863525390625,
151
+ "loss": 0.5205,
152
+ "rewards/accuracies": 0.75,
153
+ "rewards/chosen": -0.03289069980382919,
154
+ "rewards/margins": 0.03937286138534546,
155
+ "rewards/rejected": -0.07226356118917465,
156
+ "step": 7
157
+ },
158
+ {
159
+ "debug/policy_chosen_logits": -1.4155532121658325,
160
+ "debug/policy_chosen_logps": -323.04852294921875,
161
+ "debug/policy_rejected_logits": -1.4017454385757446,
162
+ "debug/policy_rejected_logps": -340.23590087890625,
163
+ "debug/reference_chosen_logps": -320.93182373046875,
164
+ "debug/reference_rejected_logps": -338.517578125,
165
+ "epoch": 0.18604651162790697,
166
+ "grad_norm": 16.872050626378687,
167
+ "learning_rate": 1e-06,
168
+ "logits/chosen": -1.4155532121658325,
169
+ "logits/rejected": -1.4017454385757446,
170
+ "logps/chosen": -323.04852294921875,
171
+ "logps/rejected": -340.23590087890625,
172
+ "loss": 0.4939,
173
+ "rewards/accuracies": 0.5,
174
+ "rewards/chosen": -0.021166954189538956,
175
+ "rewards/margins": -0.003983440343290567,
176
+ "rewards/rejected": -0.017183512449264526,
177
+ "step": 8
178
+ },
179
+ {
180
+ "debug/policy_chosen_logits": -1.5506556034088135,
181
+ "debug/policy_chosen_logps": -292.11676025390625,
182
+ "debug/policy_rejected_logits": -1.4309040307998657,
183
+ "debug/policy_rejected_logps": -307.30072021484375,
184
+ "debug/reference_chosen_logps": -288.0973205566406,
185
+ "debug/reference_rejected_logps": -306.0505676269531,
186
+ "epoch": 0.20930232558139536,
187
+ "grad_norm": 12.819676343754237,
188
+ "learning_rate": 1e-06,
189
+ "logits/chosen": -1.5506556034088135,
190
+ "logits/rejected": -1.4309040307998657,
191
+ "logps/chosen": -292.11676025390625,
192
+ "logps/rejected": -307.30072021484375,
193
+ "loss": 0.509,
194
+ "rewards/accuracies": 0.5,
195
+ "rewards/chosen": -0.04019462317228317,
196
+ "rewards/margins": -0.027693118900060654,
197
+ "rewards/rejected": -0.012501506134867668,
198
+ "step": 9
199
+ },
200
+ {
201
+ "debug/policy_chosen_logits": -1.5686500072479248,
202
+ "debug/policy_chosen_logps": -248.2123260498047,
203
+ "debug/policy_rejected_logits": -1.543015480041504,
204
+ "debug/policy_rejected_logps": -311.9925537109375,
205
+ "debug/reference_chosen_logps": -249.5205078125,
206
+ "debug/reference_rejected_logps": -309.8684997558594,
207
+ "epoch": 0.23255813953488372,
208
+ "grad_norm": 15.183229742103123,
209
+ "learning_rate": 1e-06,
210
+ "logits/chosen": -1.5686500072479248,
211
+ "logits/rejected": -1.543015480041504,
212
+ "logps/chosen": -248.2123260498047,
213
+ "logps/rejected": -311.9925537109375,
214
+ "loss": 0.479,
215
+ "rewards/accuracies": 0.75,
216
+ "rewards/chosen": 0.013081644661724567,
217
+ "rewards/margins": 0.0343221053481102,
218
+ "rewards/rejected": -0.021240463480353355,
219
+ "step": 10
220
+ },
221
+ {
222
+ "debug/policy_chosen_logits": -1.6759617328643799,
223
+ "debug/policy_chosen_logps": -241.01116943359375,
224
+ "debug/policy_rejected_logits": -1.577236533164978,
225
+ "debug/policy_rejected_logps": -277.329833984375,
226
+ "debug/reference_chosen_logps": -243.44976806640625,
227
+ "debug/reference_rejected_logps": -281.47589111328125,
228
+ "epoch": 0.2558139534883721,
229
+ "grad_norm": 31.90728084918248,
230
+ "learning_rate": 1e-06,
231
+ "logits/chosen": -1.6759617328643799,
232
+ "logits/rejected": -1.577236533164978,
233
+ "logps/chosen": -241.01116943359375,
234
+ "logps/rejected": -277.329833984375,
235
+ "loss": 0.4928,
236
+ "rewards/accuracies": 0.625,
237
+ "rewards/chosen": 0.024385996162891388,
238
+ "rewards/margins": -0.01707470789551735,
239
+ "rewards/rejected": 0.04146070405840874,
240
+ "step": 11
241
+ },
242
+ {
243
+ "debug/policy_chosen_logits": -1.5532420873641968,
244
+ "debug/policy_chosen_logps": -223.29437255859375,
245
+ "debug/policy_rejected_logits": -1.4593762159347534,
246
+ "debug/policy_rejected_logps": -401.5757141113281,
247
+ "debug/reference_chosen_logps": -226.59278869628906,
248
+ "debug/reference_rejected_logps": -400.6639099121094,
249
+ "epoch": 0.27906976744186046,
250
+ "grad_norm": 19.326544561851396,
251
+ "learning_rate": 1e-06,
252
+ "logits/chosen": -1.5532420873641968,
253
+ "logits/rejected": -1.4593762159347534,
254
+ "logps/chosen": -223.29437255859375,
255
+ "logps/rejected": -401.5757141113281,
256
+ "loss": 0.506,
257
+ "rewards/accuracies": 0.875,
258
+ "rewards/chosen": 0.03298404812812805,
259
+ "rewards/margins": 0.042102016508579254,
260
+ "rewards/rejected": -0.009117968380451202,
261
+ "step": 12
262
+ },
263
+ {
264
+ "debug/policy_chosen_logits": -1.4542545080184937,
265
+ "debug/policy_chosen_logps": -257.8163146972656,
266
+ "debug/policy_rejected_logits": -1.385650634765625,
267
+ "debug/policy_rejected_logps": -263.74664306640625,
268
+ "debug/reference_chosen_logps": -256.629638671875,
269
+ "debug/reference_rejected_logps": -271.56353759765625,
270
+ "epoch": 0.3023255813953488,
271
+ "grad_norm": 68.62278258007824,
272
+ "learning_rate": 1e-06,
273
+ "logits/chosen": -1.4542545080184937,
274
+ "logits/rejected": -1.385650634765625,
275
+ "logps/chosen": -257.8163146972656,
276
+ "logps/rejected": -263.74664306640625,
277
+ "loss": 0.5367,
278
+ "rewards/accuracies": 0.5,
279
+ "rewards/chosen": -0.011866549961268902,
280
+ "rewards/margins": -0.09003522992134094,
281
+ "rewards/rejected": 0.07816867530345917,
282
+ "step": 13
283
+ },
284
+ {
285
+ "debug/policy_chosen_logits": -1.6388806104660034,
286
+ "debug/policy_chosen_logps": -191.90548706054688,
287
+ "debug/policy_rejected_logits": -1.3299458026885986,
288
+ "debug/policy_rejected_logps": -283.2551574707031,
289
+ "debug/reference_chosen_logps": -193.77301025390625,
290
+ "debug/reference_rejected_logps": -284.05615234375,
291
+ "epoch": 0.32558139534883723,
292
+ "grad_norm": 54.79498866692366,
293
+ "learning_rate": 1e-06,
294
+ "logits/chosen": -1.6388806104660034,
295
+ "logits/rejected": -1.3299458026885986,
296
+ "logps/chosen": -191.90548706054688,
297
+ "logps/rejected": -283.2551574707031,
298
+ "loss": 0.5391,
299
+ "rewards/accuracies": 0.75,
300
+ "rewards/chosen": 0.018675191327929497,
301
+ "rewards/margins": 0.010665085166692734,
302
+ "rewards/rejected": 0.008010102435946465,
303
+ "step": 14
304
+ },
305
+ {
306
+ "debug/policy_chosen_logits": -1.4526491165161133,
307
+ "debug/policy_chosen_logps": -284.34832763671875,
308
+ "debug/policy_rejected_logits": -1.4107521772384644,
309
+ "debug/policy_rejected_logps": -363.9824523925781,
310
+ "debug/reference_chosen_logps": -282.1776123046875,
311
+ "debug/reference_rejected_logps": -362.1690368652344,
312
+ "epoch": 0.3488372093023256,
313
+ "grad_norm": 55.58252467843115,
314
+ "learning_rate": 1e-06,
315
+ "logits/chosen": -1.4526491165161133,
316
+ "logits/rejected": -1.4107521772384644,
317
+ "logps/chosen": -284.34832763671875,
318
+ "logps/rejected": -363.9824523925781,
319
+ "loss": 0.5002,
320
+ "rewards/accuracies": 0.5,
321
+ "rewards/chosen": -0.021707000210881233,
322
+ "rewards/margins": -0.0035727117210626602,
323
+ "rewards/rejected": -0.018134288489818573,
324
+ "step": 15
325
+ },
326
+ {
327
+ "debug/policy_chosen_logits": -1.5884922742843628,
328
+ "debug/policy_chosen_logps": -250.37115478515625,
329
+ "debug/policy_rejected_logits": -1.519769549369812,
330
+ "debug/policy_rejected_logps": -297.04339599609375,
331
+ "debug/reference_chosen_logps": -246.65272521972656,
332
+ "debug/reference_rejected_logps": -291.1093444824219,
333
+ "epoch": 0.37209302325581395,
334
+ "grad_norm": 20.97275013484062,
335
+ "learning_rate": 1e-06,
336
+ "logits/chosen": -1.5884922742843628,
337
+ "logits/rejected": -1.519769549369812,
338
+ "logps/chosen": -250.37115478515625,
339
+ "logps/rejected": -297.04339599609375,
340
+ "loss": 0.4843,
341
+ "rewards/accuracies": 0.625,
342
+ "rewards/chosen": -0.037184178829193115,
343
+ "rewards/margins": 0.022156618535518646,
344
+ "rewards/rejected": -0.05934080481529236,
345
+ "step": 16
346
+ },
347
+ {
348
+ "debug/policy_chosen_logits": -1.5344979763031006,
349
+ "debug/policy_chosen_logps": -214.46148681640625,
350
+ "debug/policy_rejected_logits": -1.4357935190200806,
351
+ "debug/policy_rejected_logps": -316.449951171875,
352
+ "debug/reference_chosen_logps": -210.27589416503906,
353
+ "debug/reference_rejected_logps": -310.38165283203125,
354
+ "epoch": 0.3953488372093023,
355
+ "grad_norm": 43.215377561087685,
356
+ "learning_rate": 1e-06,
357
+ "logits/chosen": -1.5344979763031006,
358
+ "logits/rejected": -1.4357935190200806,
359
+ "logps/chosen": -214.46148681640625,
360
+ "logps/rejected": -316.449951171875,
361
+ "loss": 0.5047,
362
+ "rewards/accuracies": 0.625,
363
+ "rewards/chosen": -0.04185575246810913,
364
+ "rewards/margins": 0.01882733218371868,
365
+ "rewards/rejected": -0.06068308651447296,
366
+ "step": 17
367
+ },
368
+ {
369
+ "debug/policy_chosen_logits": -1.5920026302337646,
370
+ "debug/policy_chosen_logps": -313.2801513671875,
371
+ "debug/policy_rejected_logits": -1.4743777513504028,
372
+ "debug/policy_rejected_logps": -344.209716796875,
373
+ "debug/reference_chosen_logps": -304.38165283203125,
374
+ "debug/reference_rejected_logps": -334.8297424316406,
375
+ "epoch": 0.4186046511627907,
376
+ "grad_norm": 46.01762275116522,
377
+ "learning_rate": 1e-06,
378
+ "logits/chosen": -1.5920026302337646,
379
+ "logits/rejected": -1.4743777513504028,
380
+ "logps/chosen": -313.2801513671875,
381
+ "logps/rejected": -344.209716796875,
382
+ "loss": 0.5061,
383
+ "rewards/accuracies": 0.625,
384
+ "rewards/chosen": -0.0889848843216896,
385
+ "rewards/margins": 0.004814721643924713,
386
+ "rewards/rejected": -0.09379960596561432,
387
+ "step": 18
388
+ },
389
+ {
390
+ "debug/policy_chosen_logits": -1.5471267700195312,
391
+ "debug/policy_chosen_logps": -275.2354736328125,
392
+ "debug/policy_rejected_logits": -1.5307285785675049,
393
+ "debug/policy_rejected_logps": -272.6752624511719,
394
+ "debug/reference_chosen_logps": -266.9699401855469,
395
+ "debug/reference_rejected_logps": -267.95025634765625,
396
+ "epoch": 0.4418604651162791,
397
+ "grad_norm": 78.66006276541071,
398
+ "learning_rate": 1e-06,
399
+ "logits/chosen": -1.5471267700195312,
400
+ "logits/rejected": -1.5307285785675049,
401
+ "logps/chosen": -275.2354736328125,
402
+ "logps/rejected": -272.6752624511719,
403
+ "loss": 0.5021,
404
+ "rewards/accuracies": 0.25,
405
+ "rewards/chosen": -0.08265500515699387,
406
+ "rewards/margins": -0.03540489077568054,
407
+ "rewards/rejected": -0.047250114381313324,
408
+ "step": 19
409
+ },
410
+ {
411
+ "debug/policy_chosen_logits": -1.5639888048171997,
412
+ "debug/policy_chosen_logps": -260.4048767089844,
413
+ "debug/policy_rejected_logits": -1.4268858432769775,
414
+ "debug/policy_rejected_logps": -320.64324951171875,
415
+ "debug/reference_chosen_logps": -257.83966064453125,
416
+ "debug/reference_rejected_logps": -314.41796875,
417
+ "epoch": 0.46511627906976744,
418
+ "grad_norm": 76.10262014579365,
419
+ "learning_rate": 1e-06,
420
+ "logits/chosen": -1.5639888048171997,
421
+ "logits/rejected": -1.4268858432769775,
422
+ "logps/chosen": -260.4048767089844,
423
+ "logps/rejected": -320.64324951171875,
424
+ "loss": 0.4966,
425
+ "rewards/accuracies": 0.5,
426
+ "rewards/chosen": -0.025652198120951653,
427
+ "rewards/margins": 0.036600261926651,
428
+ "rewards/rejected": -0.0622524619102478,
429
+ "step": 20
430
+ },
431
+ {
432
+ "debug/policy_chosen_logits": -1.5689480304718018,
433
+ "debug/policy_chosen_logps": -301.9723205566406,
434
+ "debug/policy_rejected_logits": -1.5015813112258911,
435
+ "debug/policy_rejected_logps": -363.80145263671875,
436
+ "debug/reference_chosen_logps": -296.4005432128906,
437
+ "debug/reference_rejected_logps": -355.6829833984375,
438
+ "epoch": 0.4883720930232558,
439
+ "grad_norm": 57.20135841653825,
440
+ "learning_rate": 1e-06,
441
+ "logits/chosen": -1.5689480304718018,
442
+ "logits/rejected": -1.5015813112258911,
443
+ "logps/chosen": -301.9723205566406,
444
+ "logps/rejected": -363.80145263671875,
445
+ "loss": 0.5027,
446
+ "rewards/accuracies": 0.625,
447
+ "rewards/chosen": -0.055717699229717255,
448
+ "rewards/margins": 0.025466954335570335,
449
+ "rewards/rejected": -0.08118465542793274,
450
+ "step": 21
451
+ },
452
+ {
453
+ "debug/policy_chosen_logits": -1.6258084774017334,
454
+ "debug/policy_chosen_logps": -271.4725341796875,
455
+ "debug/policy_rejected_logits": -1.5413790941238403,
456
+ "debug/policy_rejected_logps": -298.4366760253906,
457
+ "debug/reference_chosen_logps": -269.2203369140625,
458
+ "debug/reference_rejected_logps": -294.7615966796875,
459
+ "epoch": 0.5116279069767442,
460
+ "grad_norm": 24.247441456009533,
461
+ "learning_rate": 1e-06,
462
+ "logits/chosen": -1.6258084774017334,
463
+ "logits/rejected": -1.5413790941238403,
464
+ "logps/chosen": -271.4725341796875,
465
+ "logps/rejected": -298.4366760253906,
466
+ "loss": 0.4965,
467
+ "rewards/accuracies": 0.5,
468
+ "rewards/chosen": -0.022522106766700745,
469
+ "rewards/margins": 0.014228800311684608,
470
+ "rewards/rejected": -0.0367509089410305,
471
+ "step": 22
472
+ },
473
+ {
474
+ "debug/policy_chosen_logits": -1.4780645370483398,
475
+ "debug/policy_chosen_logps": -238.43905639648438,
476
+ "debug/policy_rejected_logits": -1.4756673574447632,
477
+ "debug/policy_rejected_logps": -300.5503845214844,
478
+ "debug/reference_chosen_logps": -234.42031860351562,
479
+ "debug/reference_rejected_logps": -297.0867919921875,
480
+ "epoch": 0.5348837209302325,
481
+ "grad_norm": 40.36957484697572,
482
+ "learning_rate": 1e-06,
483
+ "logits/chosen": -1.4780645370483398,
484
+ "logits/rejected": -1.4756673574447632,
485
+ "logps/chosen": -238.43905639648438,
486
+ "logps/rejected": -300.5503845214844,
487
+ "loss": 0.5159,
488
+ "rewards/accuracies": 0.625,
489
+ "rewards/chosen": -0.040187377482652664,
490
+ "rewards/margins": -0.005551624111831188,
491
+ "rewards/rejected": -0.0346357524394989,
492
+ "step": 23
493
+ },
494
+ {
495
+ "debug/policy_chosen_logits": -1.653166651725769,
496
+ "debug/policy_chosen_logps": -278.77362060546875,
497
+ "debug/policy_rejected_logits": -1.4196488857269287,
498
+ "debug/policy_rejected_logps": -267.0096435546875,
499
+ "debug/reference_chosen_logps": -280.239013671875,
500
+ "debug/reference_rejected_logps": -267.35540771484375,
501
+ "epoch": 0.5581395348837209,
502
+ "grad_norm": 43.31573267916226,
503
+ "learning_rate": 1e-06,
504
+ "logits/chosen": -1.653166651725769,
505
+ "logits/rejected": -1.4196488857269287,
506
+ "logps/chosen": -278.77362060546875,
507
+ "logps/rejected": -267.0096435546875,
508
+ "loss": 0.5133,
509
+ "rewards/accuracies": 0.625,
510
+ "rewards/chosen": 0.014654004946351051,
511
+ "rewards/margins": 0.011196250095963478,
512
+ "rewards/rejected": 0.0034577548503875732,
513
+ "step": 24
514
+ },
515
+ {
516
+ "debug/policy_chosen_logits": -1.5763293504714966,
517
+ "debug/policy_chosen_logps": -231.9792938232422,
518
+ "debug/policy_rejected_logits": -1.5205391645431519,
519
+ "debug/policy_rejected_logps": -304.4539794921875,
520
+ "debug/reference_chosen_logps": -230.9835662841797,
521
+ "debug/reference_rejected_logps": -301.9881896972656,
522
+ "epoch": 0.5813953488372093,
523
+ "grad_norm": 26.712778912083667,
524
+ "learning_rate": 1e-06,
525
+ "logits/chosen": -1.5763293504714966,
526
+ "logits/rejected": -1.5205391645431519,
527
+ "logps/chosen": -231.9792938232422,
528
+ "logps/rejected": -304.4539794921875,
529
+ "loss": 0.4975,
530
+ "rewards/accuracies": 0.625,
531
+ "rewards/chosen": -0.009957179427146912,
532
+ "rewards/margins": 0.014700757339596748,
533
+ "rewards/rejected": -0.02465793862938881,
534
+ "step": 25
535
+ },
536
+ {
537
+ "debug/policy_chosen_logits": -1.4429453611373901,
538
+ "debug/policy_chosen_logps": -223.239501953125,
539
+ "debug/policy_rejected_logits": -1.4664965867996216,
540
+ "debug/policy_rejected_logps": -303.63232421875,
541
+ "debug/reference_chosen_logps": -226.43624877929688,
542
+ "debug/reference_rejected_logps": -305.3199157714844,
543
+ "epoch": 0.6046511627906976,
544
+ "grad_norm": 19.046211934497915,
545
+ "learning_rate": 1e-06,
546
+ "logits/chosen": -1.4429453611373901,
547
+ "logits/rejected": -1.4664965867996216,
548
+ "logps/chosen": -223.239501953125,
549
+ "logps/rejected": -303.63232421875,
550
+ "loss": 0.4694,
551
+ "rewards/accuracies": 0.625,
552
+ "rewards/chosen": 0.03196748346090317,
553
+ "rewards/margins": 0.015091512352228165,
554
+ "rewards/rejected": 0.016875972971320152,
555
+ "step": 26
556
+ },
557
+ {
558
+ "debug/policy_chosen_logits": -1.565663456916809,
559
+ "debug/policy_chosen_logps": -309.9948425292969,
560
+ "debug/policy_rejected_logits": -1.5396883487701416,
561
+ "debug/policy_rejected_logps": -343.0301818847656,
562
+ "debug/reference_chosen_logps": -308.83489990234375,
563
+ "debug/reference_rejected_logps": -338.6658935546875,
564
+ "epoch": 0.627906976744186,
565
+ "grad_norm": 37.980256590319236,
566
+ "learning_rate": 1e-06,
567
+ "logits/chosen": -1.565663456916809,
568
+ "logits/rejected": -1.5396883487701416,
569
+ "logps/chosen": -309.9948425292969,
570
+ "logps/rejected": -343.0301818847656,
571
+ "loss": 0.4978,
572
+ "rewards/accuracies": 0.75,
573
+ "rewards/chosen": -0.011599387973546982,
574
+ "rewards/margins": 0.032043568789958954,
575
+ "rewards/rejected": -0.043642960488796234,
576
+ "step": 27
577
+ },
578
+ {
579
+ "debug/policy_chosen_logits": -1.5224864482879639,
580
+ "debug/policy_chosen_logps": -277.25262451171875,
581
+ "debug/policy_rejected_logits": -1.2660479545593262,
582
+ "debug/policy_rejected_logps": -272.9588623046875,
583
+ "debug/reference_chosen_logps": -273.28631591796875,
584
+ "debug/reference_rejected_logps": -273.3193359375,
585
+ "epoch": 0.6511627906976745,
586
+ "grad_norm": 26.186511827710103,
587
+ "learning_rate": 1e-06,
588
+ "logits/chosen": -1.5224864482879639,
589
+ "logits/rejected": -1.2660479545593262,
590
+ "logps/chosen": -277.25262451171875,
591
+ "logps/rejected": -272.9588623046875,
592
+ "loss": 0.4751,
593
+ "rewards/accuracies": 0.5,
594
+ "rewards/chosen": -0.0396631583571434,
595
+ "rewards/margins": -0.04326799511909485,
596
+ "rewards/rejected": 0.003604833036661148,
597
+ "step": 28
598
+ },
599
+ {
600
+ "debug/policy_chosen_logits": -1.366220235824585,
601
+ "debug/policy_chosen_logps": -253.40408325195312,
602
+ "debug/policy_rejected_logits": -1.3099398612976074,
603
+ "debug/policy_rejected_logps": -305.3914489746094,
604
+ "debug/reference_chosen_logps": -254.5312957763672,
605
+ "debug/reference_rejected_logps": -306.05023193359375,
606
+ "epoch": 0.6744186046511628,
607
+ "grad_norm": 17.36440597445389,
608
+ "learning_rate": 1e-06,
609
+ "logits/chosen": -1.366220235824585,
610
+ "logits/rejected": -1.3099398612976074,
611
+ "logps/chosen": -253.40408325195312,
612
+ "logps/rejected": -305.3914489746094,
613
+ "loss": 0.4956,
614
+ "rewards/accuracies": 0.75,
615
+ "rewards/chosen": 0.011272145435214043,
616
+ "rewards/margins": 0.004684487357735634,
617
+ "rewards/rejected": 0.00658765435218811,
618
+ "step": 29
619
+ },
620
+ {
621
+ "debug/policy_chosen_logits": -1.464882493019104,
622
+ "debug/policy_chosen_logps": -207.21820068359375,
623
+ "debug/policy_rejected_logits": -1.0828033685684204,
624
+ "debug/policy_rejected_logps": -319.6998291015625,
625
+ "debug/reference_chosen_logps": -202.86053466796875,
626
+ "debug/reference_rejected_logps": -312.14776611328125,
627
+ "epoch": 0.6976744186046512,
628
+ "grad_norm": 35.44932621138554,
629
+ "learning_rate": 1e-06,
630
+ "logits/chosen": -1.464882493019104,
631
+ "logits/rejected": -1.0828033685684204,
632
+ "logps/chosen": -207.21820068359375,
633
+ "logps/rejected": -319.6998291015625,
634
+ "loss": 0.5006,
635
+ "rewards/accuracies": 0.75,
636
+ "rewards/chosen": -0.043576523661613464,
637
+ "rewards/margins": 0.03194410353899002,
638
+ "rewards/rejected": -0.07552062720060349,
639
+ "step": 30
640
+ },
641
+ {
642
+ "debug/policy_chosen_logits": -1.4816325902938843,
643
+ "debug/policy_chosen_logps": -232.97183227539062,
644
+ "debug/policy_rejected_logits": -1.4852979183197021,
645
+ "debug/policy_rejected_logps": -330.938232421875,
646
+ "debug/reference_chosen_logps": -228.60247802734375,
647
+ "debug/reference_rejected_logps": -322.2069091796875,
648
+ "epoch": 0.7209302325581395,
649
+ "grad_norm": 69.51972685873595,
650
+ "learning_rate": 1e-06,
651
+ "logits/chosen": -1.4816325902938843,
652
+ "logits/rejected": -1.4852979183197021,
653
+ "logps/chosen": -232.97183227539062,
654
+ "logps/rejected": -330.938232421875,
655
+ "loss": 0.5002,
656
+ "rewards/accuracies": 0.5,
657
+ "rewards/chosen": -0.04369337856769562,
658
+ "rewards/margins": 0.04361977428197861,
659
+ "rewards/rejected": -0.08731315284967422,
660
+ "step": 31
661
+ },
662
+ {
663
+ "debug/policy_chosen_logits": -1.5637061595916748,
664
+ "debug/policy_chosen_logps": -243.05120849609375,
665
+ "debug/policy_rejected_logits": -1.4639126062393188,
666
+ "debug/policy_rejected_logps": -335.36761474609375,
667
+ "debug/reference_chosen_logps": -240.65737915039062,
668
+ "debug/reference_rejected_logps": -330.8468322753906,
669
+ "epoch": 0.7441860465116279,
670
+ "grad_norm": 27.689750506820502,
671
+ "learning_rate": 1e-06,
672
+ "logits/chosen": -1.5637061595916748,
673
+ "logits/rejected": -1.4639126062393188,
674
+ "logps/chosen": -243.05120849609375,
675
+ "logps/rejected": -335.36761474609375,
676
+ "loss": 0.4746,
677
+ "rewards/accuracies": 0.5,
678
+ "rewards/chosen": -0.023938149213790894,
679
+ "rewards/margins": 0.021269541233778,
680
+ "rewards/rejected": -0.04520769417285919,
681
+ "step": 32
682
+ },
683
+ {
684
+ "debug/policy_chosen_logits": -1.5378698110580444,
685
+ "debug/policy_chosen_logps": -269.176513671875,
686
+ "debug/policy_rejected_logits": -1.480587363243103,
687
+ "debug/policy_rejected_logps": -268.183349609375,
688
+ "debug/reference_chosen_logps": -267.6761474609375,
689
+ "debug/reference_rejected_logps": -264.76300048828125,
690
+ "epoch": 0.7674418604651163,
691
+ "grad_norm": 16.782807931183452,
692
+ "learning_rate": 1e-06,
693
+ "logits/chosen": -1.5378698110580444,
694
+ "logits/rejected": -1.480587363243103,
695
+ "logps/chosen": -269.176513671875,
696
+ "logps/rejected": -268.183349609375,
697
+ "loss": 0.491,
698
+ "rewards/accuracies": 0.5,
699
+ "rewards/chosen": -0.015003662556409836,
700
+ "rewards/margins": 0.01920013502240181,
701
+ "rewards/rejected": -0.034203797578811646,
702
+ "step": 33
703
+ },
704
+ {
705
+ "debug/policy_chosen_logits": -1.509446620941162,
706
+ "debug/policy_chosen_logps": -289.96466064453125,
707
+ "debug/policy_rejected_logits": -1.3191570043563843,
708
+ "debug/policy_rejected_logps": -325.3876953125,
709
+ "debug/reference_chosen_logps": -288.51171875,
710
+ "debug/reference_rejected_logps": -320.84259033203125,
711
+ "epoch": 0.7906976744186046,
712
+ "grad_norm": 43.72267267692569,
713
+ "learning_rate": 1e-06,
714
+ "logits/chosen": -1.509446620941162,
715
+ "logits/rejected": -1.3191570043563843,
716
+ "logps/chosen": -289.96466064453125,
717
+ "logps/rejected": -325.3876953125,
718
+ "loss": 0.4688,
719
+ "rewards/accuracies": 0.875,
720
+ "rewards/chosen": -0.014529475010931492,
721
+ "rewards/margins": 0.030921556055545807,
722
+ "rewards/rejected": -0.045451030135154724,
723
+ "step": 34
724
+ },
725
+ {
726
+ "debug/policy_chosen_logits": -1.5837349891662598,
727
+ "debug/policy_chosen_logps": -256.12457275390625,
728
+ "debug/policy_rejected_logits": -1.4979993104934692,
729
+ "debug/policy_rejected_logps": -346.24853515625,
730
+ "debug/reference_chosen_logps": -261.654052734375,
731
+ "debug/reference_rejected_logps": -343.11456298828125,
732
+ "epoch": 0.813953488372093,
733
+ "grad_norm": 20.72530257477532,
734
+ "learning_rate": 1e-06,
735
+ "logits/chosen": -1.5837349891662598,
736
+ "logits/rejected": -1.4979993104934692,
737
+ "logps/chosen": -256.12457275390625,
738
+ "logps/rejected": -346.24853515625,
739
+ "loss": 0.4555,
740
+ "rewards/accuracies": 0.75,
741
+ "rewards/chosen": 0.05529472604393959,
742
+ "rewards/margins": 0.08663440495729446,
743
+ "rewards/rejected": -0.03133968263864517,
744
+ "step": 35
745
+ },
746
+ {
747
+ "debug/policy_chosen_logits": -1.5132783651351929,
748
+ "debug/policy_chosen_logps": -203.21551513671875,
749
+ "debug/policy_rejected_logits": -1.5186952352523804,
750
+ "debug/policy_rejected_logps": -217.06405639648438,
751
+ "debug/reference_chosen_logps": -207.17909240722656,
752
+ "debug/reference_rejected_logps": -222.4234161376953,
753
+ "epoch": 0.8372093023255814,
754
+ "grad_norm": 21.776316541560245,
755
+ "learning_rate": 1e-06,
756
+ "logits/chosen": -1.5132783651351929,
757
+ "logits/rejected": -1.5186952352523804,
758
+ "logps/chosen": -203.21551513671875,
759
+ "logps/rejected": -217.06405639648438,
760
+ "loss": 0.4668,
761
+ "rewards/accuracies": 0.5,
762
+ "rewards/chosen": 0.03963588550686836,
763
+ "rewards/margins": -0.013957731425762177,
764
+ "rewards/rejected": 0.05359361693263054,
765
+ "step": 36
766
+ },
767
+ {
768
+ "debug/policy_chosen_logits": -1.588102102279663,
769
+ "debug/policy_chosen_logps": -200.3746795654297,
770
+ "debug/policy_rejected_logits": -1.5428119897842407,
771
+ "debug/policy_rejected_logps": -309.95166015625,
772
+ "debug/reference_chosen_logps": -209.82540893554688,
773
+ "debug/reference_rejected_logps": -305.77777099609375,
774
+ "epoch": 0.8604651162790697,
775
+ "grad_norm": 19.97759398330279,
776
+ "learning_rate": 1e-06,
777
+ "logits/chosen": -1.588102102279663,
778
+ "logits/rejected": -1.5428119897842407,
779
+ "logps/chosen": -200.3746795654297,
780
+ "logps/rejected": -309.95166015625,
781
+ "loss": 0.4543,
782
+ "rewards/accuracies": 1.0,
783
+ "rewards/chosen": 0.09450740367174149,
784
+ "rewards/margins": 0.13624611496925354,
785
+ "rewards/rejected": -0.04173869639635086,
786
+ "step": 37
787
+ },
788
+ {
789
+ "debug/policy_chosen_logits": -1.6548794507980347,
790
+ "debug/policy_chosen_logps": -257.4342956542969,
791
+ "debug/policy_rejected_logits": -1.496774673461914,
792
+ "debug/policy_rejected_logps": -306.6890563964844,
793
+ "debug/reference_chosen_logps": -263.0815734863281,
794
+ "debug/reference_rejected_logps": -307.71331787109375,
795
+ "epoch": 0.8837209302325582,
796
+ "grad_norm": 35.00520893739856,
797
+ "learning_rate": 1e-06,
798
+ "logits/chosen": -1.6548794507980347,
799
+ "logits/rejected": -1.496774673461914,
800
+ "logps/chosen": -257.4342956542969,
801
+ "logps/rejected": -306.6890563964844,
802
+ "loss": 0.476,
803
+ "rewards/accuracies": 0.625,
804
+ "rewards/chosen": 0.05647262558341026,
805
+ "rewards/margins": 0.046230219304561615,
806
+ "rewards/rejected": 0.01024240255355835,
807
+ "step": 38
808
+ },
809
+ {
810
+ "debug/policy_chosen_logits": -1.449007272720337,
811
+ "debug/policy_chosen_logps": -267.24298095703125,
812
+ "debug/policy_rejected_logits": -1.4096888303756714,
813
+ "debug/policy_rejected_logps": -289.27166748046875,
814
+ "debug/reference_chosen_logps": -269.7801818847656,
815
+ "debug/reference_rejected_logps": -288.45440673828125,
816
+ "epoch": 0.9069767441860465,
817
+ "grad_norm": 18.02813938430424,
818
+ "learning_rate": 1e-06,
819
+ "logits/chosen": -1.449007272720337,
820
+ "logits/rejected": -1.4096888303756714,
821
+ "logps/chosen": -267.24298095703125,
822
+ "logps/rejected": -289.27166748046875,
823
+ "loss": 0.4743,
824
+ "rewards/accuracies": 0.625,
825
+ "rewards/chosen": 0.025372076779603958,
826
+ "rewards/margins": 0.033544398844242096,
827
+ "rewards/rejected": -0.008172323927283287,
828
+ "step": 39
829
+ },
830
+ {
831
+ "debug/policy_chosen_logits": -1.5711109638214111,
832
+ "debug/policy_chosen_logps": -211.0213623046875,
833
+ "debug/policy_rejected_logits": -1.473678708076477,
834
+ "debug/policy_rejected_logps": -334.5294189453125,
835
+ "debug/reference_chosen_logps": -212.76715087890625,
836
+ "debug/reference_rejected_logps": -332.358154296875,
837
+ "epoch": 0.9302325581395349,
838
+ "grad_norm": 13.548944902515293,
839
+ "learning_rate": 1e-06,
840
+ "logits/chosen": -1.5711109638214111,
841
+ "logits/rejected": -1.473678708076477,
842
+ "logps/chosen": -211.0213623046875,
843
+ "logps/rejected": -334.5294189453125,
844
+ "loss": 0.4847,
845
+ "rewards/accuracies": 0.875,
846
+ "rewards/chosen": 0.01745777204632759,
847
+ "rewards/margins": 0.039170339703559875,
848
+ "rewards/rejected": -0.021712567657232285,
849
+ "step": 40
850
+ },
851
+ {
852
+ "debug/policy_chosen_logits": -1.5397162437438965,
853
+ "debug/policy_chosen_logps": -238.81680297851562,
854
+ "debug/policy_rejected_logits": -1.4631048440933228,
855
+ "debug/policy_rejected_logps": -299.21343994140625,
856
+ "debug/reference_chosen_logps": -237.25958251953125,
857
+ "debug/reference_rejected_logps": -294.679931640625,
858
+ "epoch": 0.9534883720930233,
859
+ "grad_norm": 13.60929652129651,
860
+ "learning_rate": 1e-06,
861
+ "logits/chosen": -1.5397162437438965,
862
+ "logits/rejected": -1.4631048440933228,
863
+ "logps/chosen": -238.81680297851562,
864
+ "logps/rejected": -299.21343994140625,
865
+ "loss": 0.4666,
866
+ "rewards/accuracies": 0.5,
867
+ "rewards/chosen": -0.01557239517569542,
868
+ "rewards/margins": 0.029763124883174896,
869
+ "rewards/rejected": -0.045335520058870316,
870
+ "step": 41
871
+ },
872
+ {
873
+ "debug/policy_chosen_logits": -1.4046356678009033,
874
+ "debug/policy_chosen_logps": -228.942138671875,
875
+ "debug/policy_rejected_logits": -1.301234245300293,
876
+ "debug/policy_rejected_logps": -319.590087890625,
877
+ "debug/reference_chosen_logps": -227.15533447265625,
878
+ "debug/reference_rejected_logps": -310.7023620605469,
879
+ "epoch": 0.9767441860465116,
880
+ "grad_norm": 20.029756661206523,
881
+ "learning_rate": 1e-06,
882
+ "logits/chosen": -1.4046356678009033,
883
+ "logits/rejected": -1.301234245300293,
884
+ "logps/chosen": -228.942138671875,
885
+ "logps/rejected": -319.590087890625,
886
+ "loss": 0.4742,
887
+ "rewards/accuracies": 0.875,
888
+ "rewards/chosen": -0.017868060618638992,
889
+ "rewards/margins": 0.07100919634103775,
890
+ "rewards/rejected": -0.08887725323438644,
891
+ "step": 42
892
+ },
893
+ {
894
+ "debug/policy_chosen_logits": -1.4242496490478516,
895
+ "debug/policy_chosen_logps": -248.90008544921875,
896
+ "debug/policy_rejected_logits": -1.344144582748413,
897
+ "debug/policy_rejected_logps": -312.7857360839844,
898
+ "debug/reference_chosen_logps": -248.36306762695312,
899
+ "debug/reference_rejected_logps": -306.93707275390625,
900
+ "epoch": 1.0,
901
+ "grad_norm": 16.354244880023245,
902
+ "learning_rate": 1e-06,
903
+ "logits/chosen": -1.4242496490478516,
904
+ "logits/rejected": -1.344144582748413,
905
+ "logps/chosen": -248.90008544921875,
906
+ "logps/rejected": -312.7857360839844,
907
+ "loss": 0.4466,
908
+ "rewards/accuracies": 0.75,
909
+ "rewards/chosen": -0.005370025523006916,
910
+ "rewards/margins": 0.053116798400878906,
911
+ "rewards/rejected": -0.05848682299256325,
912
+ "step": 43
913
+ },
914
+ {
915
+ "epoch": 1.0,
916
+ "step": 43,
917
+ "total_flos": 0.0,
918
+ "train_loss": 0.49317165862682255,
919
+ "train_runtime": 155.8617,
920
+ "train_samples_per_second": 17.445,
921
+ "train_steps_per_second": 0.276
922
+ }
923
+ ],
924
+ "logging_steps": 1,
925
+ "max_steps": 43,
926
+ "num_input_tokens_seen": 0,
927
+ "num_train_epochs": 1,
928
+ "save_steps": 500,
929
+ "stateful_callbacks": {
930
+ "TrainerControl": {
931
+ "args": {
932
+ "should_epoch_stop": false,
933
+ "should_evaluate": false,
934
+ "should_log": false,
935
+ "should_save": true,
936
+ "should_training_stop": true
937
+ },
938
+ "attributes": {}
939
+ }
940
+ },
941
+ "total_flos": 0.0,
942
+ "train_batch_size": 8,
943
+ "trial_name": null,
944
+ "trial_params": null
945
+ }