yiran-wang3 commited on
Commit
cd17a82
1 Parent(s): 0b3449a

End of training

Browse files
README.md ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ license: other
4
+ base_model: yiran-wang3/ds_coder6.7b_adamw_iter4
5
+ tags:
6
+ - alignment-handbook
7
+ - generated_from_trainer
8
+ - trl
9
+ - dpo
10
+ datasets:
11
+ - self-generate/ds_coder6.7b_sppo_hard_new_cn_mining_oj_iter4-binarized
12
+ model-index:
13
+ - name: ds_coder6.7b_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
+ # ds_coder6.7b_adamw_iter5
21
+
22
+ This model is a fine-tuned version of [yiran-wang3/ds_coder6.7b_adamw_iter4](https://huggingface.co/yiran-wang3/ds_coder6.7b_adamw_iter4) on the self-generate/ds_coder6.7b_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 0.20.3
all_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.4962217591025613,
5
+ "train_runtime": 388.3481,
6
+ "train_samples": 2059,
7
+ "train_samples_per_second": 5.302,
8
+ "train_steps_per_second": 0.085
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 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 100000,
4
+ "eos_token_id": 100015,
5
+ "transformers_version": "4.45.0"
6
+ }
train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.4962217591025613,
5
+ "train_runtime": 388.3481,
6
+ "train_samples": 2059,
7
+ "train_samples_per_second": 5.302,
8
+ "train_steps_per_second": 0.085
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,735 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 100,
6
+ "global_step": 33,
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": 26.247488021850586,
13
+ "debug/policy_chosen_logps": -381.44134521484375,
14
+ "debug/policy_rejected_logits": 26.089710235595703,
15
+ "debug/policy_rejected_logps": -404.22979736328125,
16
+ "debug/reference_chosen_logps": -381.44134521484375,
17
+ "debug/reference_rejected_logps": -404.22979736328125,
18
+ "epoch": 0.030303030303030304,
19
+ "grad_norm": 5.442552264842066,
20
+ "learning_rate": 1e-06,
21
+ "logits/chosen": 26.247488021850586,
22
+ "logits/rejected": 26.089710235595703,
23
+ "logps/chosen": -381.44134521484375,
24
+ "logps/rejected": -404.22979736328125,
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": 29.335710525512695,
34
+ "debug/policy_chosen_logps": -403.6748352050781,
35
+ "debug/policy_rejected_logits": 29.501441955566406,
36
+ "debug/policy_rejected_logps": -394.9062194824219,
37
+ "debug/reference_chosen_logps": -403.30743408203125,
38
+ "debug/reference_rejected_logps": -394.7815856933594,
39
+ "epoch": 0.06060606060606061,
40
+ "grad_norm": 5.814697031201472,
41
+ "learning_rate": 1e-06,
42
+ "logits/chosen": 29.335710525512695,
43
+ "logits/rejected": 29.501441955566406,
44
+ "logps/chosen": -403.6748352050781,
45
+ "logps/rejected": -394.9062194824219,
46
+ "loss": 0.5004,
47
+ "rewards/accuracies": 0.5,
48
+ "rewards/chosen": -0.0036742021329700947,
49
+ "rewards/margins": -0.0024280548095703125,
50
+ "rewards/rejected": -0.0012461473233997822,
51
+ "step": 2
52
+ },
53
+ {
54
+ "debug/policy_chosen_logits": 26.286334991455078,
55
+ "debug/policy_chosen_logps": -385.650634765625,
56
+ "debug/policy_rejected_logits": 26.62445640563965,
57
+ "debug/policy_rejected_logps": -411.38519287109375,
58
+ "debug/reference_chosen_logps": -385.63262939453125,
59
+ "debug/reference_rejected_logps": -411.11822509765625,
60
+ "epoch": 0.09090909090909091,
61
+ "grad_norm": 5.064121289594221,
62
+ "learning_rate": 1e-06,
63
+ "logits/chosen": 26.286334991455078,
64
+ "logits/rejected": 26.62445640563965,
65
+ "logps/chosen": -385.650634765625,
66
+ "logps/rejected": -411.38519287109375,
67
+ "loss": 0.5017,
68
+ "rewards/accuracies": 0.625,
69
+ "rewards/chosen": -0.00018005399033427238,
70
+ "rewards/margins": 0.0024895858950912952,
71
+ "rewards/rejected": -0.0026696398854255676,
72
+ "step": 3
73
+ },
74
+ {
75
+ "debug/policy_chosen_logits": 30.732887268066406,
76
+ "debug/policy_chosen_logps": -401.61358642578125,
77
+ "debug/policy_rejected_logits": 26.973230361938477,
78
+ "debug/policy_rejected_logps": -379.72772216796875,
79
+ "debug/reference_chosen_logps": -401.27716064453125,
80
+ "debug/reference_rejected_logps": -380.01727294921875,
81
+ "epoch": 0.12121212121212122,
82
+ "grad_norm": 4.93218202903919,
83
+ "learning_rate": 1e-06,
84
+ "logits/chosen": 30.732887268066406,
85
+ "logits/rejected": 26.973230361938477,
86
+ "logps/chosen": -401.61358642578125,
87
+ "logps/rejected": -379.72772216796875,
88
+ "loss": 0.5001,
89
+ "rewards/accuracies": 0.25,
90
+ "rewards/chosen": -0.0033642961643636227,
91
+ "rewards/margins": -0.00626018550246954,
92
+ "rewards/rejected": 0.002895889338105917,
93
+ "step": 4
94
+ },
95
+ {
96
+ "debug/policy_chosen_logits": 26.53725814819336,
97
+ "debug/policy_chosen_logps": -364.45782470703125,
98
+ "debug/policy_rejected_logits": 29.11540412902832,
99
+ "debug/policy_rejected_logps": -386.5697326660156,
100
+ "debug/reference_chosen_logps": -364.01983642578125,
101
+ "debug/reference_rejected_logps": -386.7102966308594,
102
+ "epoch": 0.15151515151515152,
103
+ "grad_norm": 4.732508848857073,
104
+ "learning_rate": 1e-06,
105
+ "logits/chosen": 26.53725814819336,
106
+ "logits/rejected": 29.11540412902832,
107
+ "logps/chosen": -364.45782470703125,
108
+ "logps/rejected": -386.5697326660156,
109
+ "loss": 0.5019,
110
+ "rewards/accuracies": 0.375,
111
+ "rewards/chosen": -0.0043797302059829235,
112
+ "rewards/margins": -0.005785560701042414,
113
+ "rewards/rejected": 0.0014058304950594902,
114
+ "step": 5
115
+ },
116
+ {
117
+ "debug/policy_chosen_logits": 27.333480834960938,
118
+ "debug/policy_chosen_logps": -371.4402770996094,
119
+ "debug/policy_rejected_logits": 28.14923858642578,
120
+ "debug/policy_rejected_logps": -380.748291015625,
121
+ "debug/reference_chosen_logps": -370.5325927734375,
122
+ "debug/reference_rejected_logps": -379.92169189453125,
123
+ "epoch": 0.18181818181818182,
124
+ "grad_norm": 5.12780395559187,
125
+ "learning_rate": 1e-06,
126
+ "logits/chosen": 27.333480834960938,
127
+ "logits/rejected": 28.14923858642578,
128
+ "logps/chosen": -371.4402770996094,
129
+ "logps/rejected": -380.748291015625,
130
+ "loss": 0.5005,
131
+ "rewards/accuracies": 0.375,
132
+ "rewards/chosen": -0.009076766669750214,
133
+ "rewards/margins": -0.0008102413266897202,
134
+ "rewards/rejected": -0.008266524411737919,
135
+ "step": 6
136
+ },
137
+ {
138
+ "debug/policy_chosen_logits": 27.809894561767578,
139
+ "debug/policy_chosen_logps": -409.5631408691406,
140
+ "debug/policy_rejected_logits": 25.189990997314453,
141
+ "debug/policy_rejected_logps": -424.32916259765625,
142
+ "debug/reference_chosen_logps": -409.25799560546875,
143
+ "debug/reference_rejected_logps": -424.49981689453125,
144
+ "epoch": 0.21212121212121213,
145
+ "grad_norm": 6.280279145214255,
146
+ "learning_rate": 1e-06,
147
+ "logits/chosen": 27.809894561767578,
148
+ "logits/rejected": 25.189990997314453,
149
+ "logps/chosen": -409.5631408691406,
150
+ "logps/rejected": -424.32916259765625,
151
+ "loss": 0.4977,
152
+ "rewards/accuracies": 0.5,
153
+ "rewards/chosen": -0.0030513377860188484,
154
+ "rewards/margins": -0.0047579575330019,
155
+ "rewards/rejected": 0.0017066190484911203,
156
+ "step": 7
157
+ },
158
+ {
159
+ "debug/policy_chosen_logits": 28.680850982666016,
160
+ "debug/policy_chosen_logps": -367.8634033203125,
161
+ "debug/policy_rejected_logits": 26.834354400634766,
162
+ "debug/policy_rejected_logps": -373.39215087890625,
163
+ "debug/reference_chosen_logps": -367.77484130859375,
164
+ "debug/reference_rejected_logps": -373.0135192871094,
165
+ "epoch": 0.24242424242424243,
166
+ "grad_norm": 6.296849117602656,
167
+ "learning_rate": 1e-06,
168
+ "logits/chosen": 28.680850982666016,
169
+ "logits/rejected": 26.834354400634766,
170
+ "logps/chosen": -367.8634033203125,
171
+ "logps/rejected": -373.39215087890625,
172
+ "loss": 0.5004,
173
+ "rewards/accuracies": 0.5,
174
+ "rewards/chosen": -0.000885658198967576,
175
+ "rewards/margins": 0.002900313585996628,
176
+ "rewards/rejected": -0.003785972483456135,
177
+ "step": 8
178
+ },
179
+ {
180
+ "debug/policy_chosen_logits": 26.558706283569336,
181
+ "debug/policy_chosen_logps": -407.4287414550781,
182
+ "debug/policy_rejected_logits": 27.462989807128906,
183
+ "debug/policy_rejected_logps": -404.10052490234375,
184
+ "debug/reference_chosen_logps": -407.3221130371094,
185
+ "debug/reference_rejected_logps": -403.45672607421875,
186
+ "epoch": 0.2727272727272727,
187
+ "grad_norm": 5.217727377560115,
188
+ "learning_rate": 1e-06,
189
+ "logits/chosen": 26.558706283569336,
190
+ "logits/rejected": 27.462989807128906,
191
+ "logps/chosen": -407.4287414550781,
192
+ "logps/rejected": -404.10052490234375,
193
+ "loss": 0.4967,
194
+ "rewards/accuracies": 0.375,
195
+ "rewards/chosen": -0.0010662460699677467,
196
+ "rewards/margins": 0.0053720856085419655,
197
+ "rewards/rejected": -0.006438331678509712,
198
+ "step": 9
199
+ },
200
+ {
201
+ "debug/policy_chosen_logits": 27.330812454223633,
202
+ "debug/policy_chosen_logps": -397.4725646972656,
203
+ "debug/policy_rejected_logits": 27.048776626586914,
204
+ "debug/policy_rejected_logps": -385.05328369140625,
205
+ "debug/reference_chosen_logps": -397.6583251953125,
206
+ "debug/reference_rejected_logps": -385.533203125,
207
+ "epoch": 0.30303030303030304,
208
+ "grad_norm": 5.264064218965023,
209
+ "learning_rate": 1e-06,
210
+ "logits/chosen": 27.330812454223633,
211
+ "logits/rejected": 27.048776626586914,
212
+ "logps/chosen": -397.4725646972656,
213
+ "logps/rejected": -385.05328369140625,
214
+ "loss": 0.4993,
215
+ "rewards/accuracies": 0.375,
216
+ "rewards/chosen": 0.0018573375418782234,
217
+ "rewards/margins": -0.002941780723631382,
218
+ "rewards/rejected": 0.004799118265509605,
219
+ "step": 10
220
+ },
221
+ {
222
+ "debug/policy_chosen_logits": 26.38730812072754,
223
+ "debug/policy_chosen_logps": -394.4657287597656,
224
+ "debug/policy_rejected_logits": 25.263452529907227,
225
+ "debug/policy_rejected_logps": -401.41778564453125,
226
+ "debug/reference_chosen_logps": -394.7672119140625,
227
+ "debug/reference_rejected_logps": -401.0439453125,
228
+ "epoch": 0.3333333333333333,
229
+ "grad_norm": 5.091598075709861,
230
+ "learning_rate": 1e-06,
231
+ "logits/chosen": 26.38730812072754,
232
+ "logits/rejected": 25.263452529907227,
233
+ "logps/chosen": -394.4657287597656,
234
+ "logps/rejected": -401.41778564453125,
235
+ "loss": 0.4959,
236
+ "rewards/accuracies": 0.75,
237
+ "rewards/chosen": 0.0030147552024573088,
238
+ "rewards/margins": 0.006753234658390284,
239
+ "rewards/rejected": -0.003738479223102331,
240
+ "step": 11
241
+ },
242
+ {
243
+ "debug/policy_chosen_logits": 29.810720443725586,
244
+ "debug/policy_chosen_logps": -365.6493225097656,
245
+ "debug/policy_rejected_logits": 29.309846878051758,
246
+ "debug/policy_rejected_logps": -382.3474426269531,
247
+ "debug/reference_chosen_logps": -366.44732666015625,
248
+ "debug/reference_rejected_logps": -382.22650146484375,
249
+ "epoch": 0.36363636363636365,
250
+ "grad_norm": 4.68129717907089,
251
+ "learning_rate": 1e-06,
252
+ "logits/chosen": 29.810720443725586,
253
+ "logits/rejected": 29.309846878051758,
254
+ "logps/chosen": -365.6493225097656,
255
+ "logps/rejected": -382.3474426269531,
256
+ "loss": 0.4974,
257
+ "rewards/accuracies": 0.75,
258
+ "rewards/chosen": 0.00797988846898079,
259
+ "rewards/margins": 0.009189032949507236,
260
+ "rewards/rejected": -0.0012091444805264473,
261
+ "step": 12
262
+ },
263
+ {
264
+ "debug/policy_chosen_logits": 25.008668899536133,
265
+ "debug/policy_chosen_logps": -400.12109375,
266
+ "debug/policy_rejected_logits": 23.562170028686523,
267
+ "debug/policy_rejected_logps": -360.582275390625,
268
+ "debug/reference_chosen_logps": -400.1289978027344,
269
+ "debug/reference_rejected_logps": -360.8138427734375,
270
+ "epoch": 0.3939393939393939,
271
+ "grad_norm": 4.901760910772593,
272
+ "learning_rate": 1e-06,
273
+ "logits/chosen": 25.008668899536133,
274
+ "logits/rejected": 23.562170028686523,
275
+ "logps/chosen": -400.12109375,
276
+ "logps/rejected": -360.582275390625,
277
+ "loss": 0.5,
278
+ "rewards/accuracies": 0.5,
279
+ "rewards/chosen": 7.907883264124393e-05,
280
+ "rewards/margins": -0.0022365953773260117,
281
+ "rewards/rejected": 0.002315673977136612,
282
+ "step": 13
283
+ },
284
+ {
285
+ "debug/policy_chosen_logits": 28.53821563720703,
286
+ "debug/policy_chosen_logps": -389.4108581542969,
287
+ "debug/policy_rejected_logits": 26.43337631225586,
288
+ "debug/policy_rejected_logps": -399.672607421875,
289
+ "debug/reference_chosen_logps": -389.0559387207031,
290
+ "debug/reference_rejected_logps": -400.2510986328125,
291
+ "epoch": 0.42424242424242425,
292
+ "grad_norm": 5.316323576734348,
293
+ "learning_rate": 1e-06,
294
+ "logits/chosen": 28.53821563720703,
295
+ "logits/rejected": 26.43337631225586,
296
+ "logps/chosen": -389.4108581542969,
297
+ "logps/rejected": -399.672607421875,
298
+ "loss": 0.4975,
299
+ "rewards/accuracies": 0.375,
300
+ "rewards/chosen": -0.0035491182934492826,
301
+ "rewards/margins": -0.009334182366728783,
302
+ "rewards/rejected": 0.005785064306110144,
303
+ "step": 14
304
+ },
305
+ {
306
+ "debug/policy_chosen_logits": 29.288440704345703,
307
+ "debug/policy_chosen_logps": -396.9311218261719,
308
+ "debug/policy_rejected_logits": 25.288602828979492,
309
+ "debug/policy_rejected_logps": -369.3037109375,
310
+ "debug/reference_chosen_logps": -397.51153564453125,
311
+ "debug/reference_rejected_logps": -369.73687744140625,
312
+ "epoch": 0.45454545454545453,
313
+ "grad_norm": 4.921365130482872,
314
+ "learning_rate": 1e-06,
315
+ "logits/chosen": 29.288440704345703,
316
+ "logits/rejected": 25.288602828979492,
317
+ "logps/chosen": -396.9311218261719,
318
+ "logps/rejected": -369.3037109375,
319
+ "loss": 0.4949,
320
+ "rewards/accuracies": 0.375,
321
+ "rewards/chosen": 0.005803985986858606,
322
+ "rewards/margins": 0.0014723972417414188,
323
+ "rewards/rejected": 0.0043315887451171875,
324
+ "step": 15
325
+ },
326
+ {
327
+ "debug/policy_chosen_logits": 25.585439682006836,
328
+ "debug/policy_chosen_logps": -410.58551025390625,
329
+ "debug/policy_rejected_logits": 24.364900588989258,
330
+ "debug/policy_rejected_logps": -394.73638916015625,
331
+ "debug/reference_chosen_logps": -411.3353271484375,
332
+ "debug/reference_rejected_logps": -394.3494873046875,
333
+ "epoch": 0.48484848484848486,
334
+ "grad_norm": 5.084152192534468,
335
+ "learning_rate": 1e-06,
336
+ "logits/chosen": 25.585439682006836,
337
+ "logits/rejected": 24.364900588989258,
338
+ "logps/chosen": -410.58551025390625,
339
+ "logps/rejected": -394.73638916015625,
340
+ "loss": 0.4953,
341
+ "rewards/accuracies": 0.75,
342
+ "rewards/chosen": 0.0074979402124881744,
343
+ "rewards/margins": 0.01136699691414833,
344
+ "rewards/rejected": -0.0038690564688295126,
345
+ "step": 16
346
+ },
347
+ {
348
+ "debug/policy_chosen_logits": 25.545652389526367,
349
+ "debug/policy_chosen_logps": -390.6317138671875,
350
+ "debug/policy_rejected_logits": 24.665712356567383,
351
+ "debug/policy_rejected_logps": -388.66748046875,
352
+ "debug/reference_chosen_logps": -392.0401611328125,
353
+ "debug/reference_rejected_logps": -388.7672119140625,
354
+ "epoch": 0.5151515151515151,
355
+ "grad_norm": 6.040733481743162,
356
+ "learning_rate": 1e-06,
357
+ "logits/chosen": 25.545652389526367,
358
+ "logits/rejected": 24.665712356567383,
359
+ "logps/chosen": -390.6317138671875,
360
+ "logps/rejected": -388.66748046875,
361
+ "loss": 0.4979,
362
+ "rewards/accuracies": 0.5,
363
+ "rewards/chosen": 0.01408412866294384,
364
+ "rewards/margins": 0.013087119907140732,
365
+ "rewards/rejected": 0.0009970092214643955,
366
+ "step": 17
367
+ },
368
+ {
369
+ "debug/policy_chosen_logits": 21.49858283996582,
370
+ "debug/policy_chosen_logps": -373.1883850097656,
371
+ "debug/policy_rejected_logits": 20.456966400146484,
372
+ "debug/policy_rejected_logps": -389.541015625,
373
+ "debug/reference_chosen_logps": -374.1577453613281,
374
+ "debug/reference_rejected_logps": -389.02264404296875,
375
+ "epoch": 0.5454545454545454,
376
+ "grad_norm": 4.954322411747276,
377
+ "learning_rate": 1e-06,
378
+ "logits/chosen": 21.49858283996582,
379
+ "logits/rejected": 20.456966400146484,
380
+ "logps/chosen": -373.1883850097656,
381
+ "logps/rejected": -389.541015625,
382
+ "loss": 0.4967,
383
+ "rewards/accuracies": 0.75,
384
+ "rewards/chosen": 0.009693756699562073,
385
+ "rewards/margins": 0.01487777754664421,
386
+ "rewards/rejected": -0.005184021312743425,
387
+ "step": 18
388
+ },
389
+ {
390
+ "debug/policy_chosen_logits": 28.040218353271484,
391
+ "debug/policy_chosen_logps": -395.2466125488281,
392
+ "debug/policy_rejected_logits": 30.119606018066406,
393
+ "debug/policy_rejected_logps": -410.98419189453125,
394
+ "debug/reference_chosen_logps": -394.3302001953125,
395
+ "debug/reference_rejected_logps": -410.0677490234375,
396
+ "epoch": 0.5757575757575758,
397
+ "grad_norm": 4.970008551225456,
398
+ "learning_rate": 1e-06,
399
+ "logits/chosen": 28.040218353271484,
400
+ "logits/rejected": 30.119606018066406,
401
+ "logps/chosen": -395.2466125488281,
402
+ "logps/rejected": -410.98419189453125,
403
+ "loss": 0.5008,
404
+ "rewards/accuracies": 0.5,
405
+ "rewards/chosen": -0.009163931012153625,
406
+ "rewards/margins": 5.345791578292847e-07,
407
+ "rewards/rejected": -0.009164467453956604,
408
+ "step": 19
409
+ },
410
+ {
411
+ "debug/policy_chosen_logits": 25.36637306213379,
412
+ "debug/policy_chosen_logps": -359.6551208496094,
413
+ "debug/policy_rejected_logits": 28.496570587158203,
414
+ "debug/policy_rejected_logps": -396.62646484375,
415
+ "debug/reference_chosen_logps": -359.1936950683594,
416
+ "debug/reference_rejected_logps": -394.8607177734375,
417
+ "epoch": 0.6060606060606061,
418
+ "grad_norm": 5.36738350658573,
419
+ "learning_rate": 1e-06,
420
+ "logits/chosen": 25.36637306213379,
421
+ "logits/rejected": 28.496570587158203,
422
+ "logps/chosen": -359.6551208496094,
423
+ "logps/rejected": -396.62646484375,
424
+ "loss": 0.493,
425
+ "rewards/accuracies": 0.5,
426
+ "rewards/chosen": -0.00461437227204442,
427
+ "rewards/margins": 0.013043059967458248,
428
+ "rewards/rejected": -0.017657432705163956,
429
+ "step": 20
430
+ },
431
+ {
432
+ "debug/policy_chosen_logits": 27.611148834228516,
433
+ "debug/policy_chosen_logps": -388.72259521484375,
434
+ "debug/policy_rejected_logits": 27.41425323486328,
435
+ "debug/policy_rejected_logps": -376.91302490234375,
436
+ "debug/reference_chosen_logps": -388.17486572265625,
437
+ "debug/reference_rejected_logps": -376.16900634765625,
438
+ "epoch": 0.6363636363636364,
439
+ "grad_norm": 4.90749395193054,
440
+ "learning_rate": 1e-06,
441
+ "logits/chosen": 27.611148834228516,
442
+ "logits/rejected": 27.41425323486328,
443
+ "logps/chosen": -388.72259521484375,
444
+ "logps/rejected": -376.91302490234375,
445
+ "loss": 0.4944,
446
+ "rewards/accuracies": 0.5,
447
+ "rewards/chosen": -0.00547714252024889,
448
+ "rewards/margins": 0.0019630808383226395,
449
+ "rewards/rejected": -0.007440223358571529,
450
+ "step": 21
451
+ },
452
+ {
453
+ "debug/policy_chosen_logits": 31.425703048706055,
454
+ "debug/policy_chosen_logps": -415.8355712890625,
455
+ "debug/policy_rejected_logits": 30.123226165771484,
456
+ "debug/policy_rejected_logps": -390.62847900390625,
457
+ "debug/reference_chosen_logps": -416.2061767578125,
458
+ "debug/reference_rejected_logps": -391.3880310058594,
459
+ "epoch": 0.6666666666666666,
460
+ "grad_norm": 5.726245430763464,
461
+ "learning_rate": 1e-06,
462
+ "logits/chosen": 31.425703048706055,
463
+ "logits/rejected": 30.123226165771484,
464
+ "logps/chosen": -415.8355712890625,
465
+ "logps/rejected": -390.62847900390625,
466
+ "loss": 0.491,
467
+ "rewards/accuracies": 0.375,
468
+ "rewards/chosen": 0.003706092946231365,
469
+ "rewards/margins": -0.003889350686222315,
470
+ "rewards/rejected": 0.00759544363245368,
471
+ "step": 22
472
+ },
473
+ {
474
+ "debug/policy_chosen_logits": 28.34208869934082,
475
+ "debug/policy_chosen_logps": -383.0400085449219,
476
+ "debug/policy_rejected_logits": 27.740306854248047,
477
+ "debug/policy_rejected_logps": -396.38739013671875,
478
+ "debug/reference_chosen_logps": -382.1289367675781,
479
+ "debug/reference_rejected_logps": -395.863525390625,
480
+ "epoch": 0.696969696969697,
481
+ "grad_norm": 6.5017943561734475,
482
+ "learning_rate": 1e-06,
483
+ "logits/chosen": 28.34208869934082,
484
+ "logits/rejected": 27.740306854248047,
485
+ "logps/chosen": -383.0400085449219,
486
+ "logps/rejected": -396.38739013671875,
487
+ "loss": 0.4974,
488
+ "rewards/accuracies": 0.5,
489
+ "rewards/chosen": -0.009110527113080025,
490
+ "rewards/margins": -0.003872184082865715,
491
+ "rewards/rejected": -0.005238342098891735,
492
+ "step": 23
493
+ },
494
+ {
495
+ "debug/policy_chosen_logits": 28.894771575927734,
496
+ "debug/policy_chosen_logps": -379.154296875,
497
+ "debug/policy_rejected_logits": 29.083221435546875,
498
+ "debug/policy_rejected_logps": -398.38275146484375,
499
+ "debug/reference_chosen_logps": -379.416015625,
500
+ "debug/reference_rejected_logps": -397.15728759765625,
501
+ "epoch": 0.7272727272727273,
502
+ "grad_norm": 5.2990691392738665,
503
+ "learning_rate": 1e-06,
504
+ "logits/chosen": 28.894771575927734,
505
+ "logits/rejected": 29.083221435546875,
506
+ "logps/chosen": -379.154296875,
507
+ "logps/rejected": -398.38275146484375,
508
+ "loss": 0.4945,
509
+ "rewards/accuracies": 0.75,
510
+ "rewards/chosen": 0.0026170345954596996,
511
+ "rewards/margins": 0.014871787279844284,
512
+ "rewards/rejected": -0.012254752218723297,
513
+ "step": 24
514
+ },
515
+ {
516
+ "debug/policy_chosen_logits": 27.720739364624023,
517
+ "debug/policy_chosen_logps": -386.32635498046875,
518
+ "debug/policy_rejected_logits": 24.095592498779297,
519
+ "debug/policy_rejected_logps": -369.91998291015625,
520
+ "debug/reference_chosen_logps": -385.1112365722656,
521
+ "debug/reference_rejected_logps": -369.97222900390625,
522
+ "epoch": 0.7575757575757576,
523
+ "grad_norm": 5.341763911189841,
524
+ "learning_rate": 1e-06,
525
+ "logits/chosen": 27.720739364624023,
526
+ "logits/rejected": 24.095592498779297,
527
+ "logps/chosen": -386.32635498046875,
528
+ "logps/rejected": -369.91998291015625,
529
+ "loss": 0.4978,
530
+ "rewards/accuracies": 0.125,
531
+ "rewards/chosen": -0.0121515654027462,
532
+ "rewards/margins": -0.012673910707235336,
533
+ "rewards/rejected": 0.0005223462358117104,
534
+ "step": 25
535
+ },
536
+ {
537
+ "debug/policy_chosen_logits": 26.473604202270508,
538
+ "debug/policy_chosen_logps": -406.0578308105469,
539
+ "debug/policy_rejected_logits": 25.949445724487305,
540
+ "debug/policy_rejected_logps": -396.72174072265625,
541
+ "debug/reference_chosen_logps": -406.4796447753906,
542
+ "debug/reference_rejected_logps": -396.7256774902344,
543
+ "epoch": 0.7878787878787878,
544
+ "grad_norm": 5.2984045842092815,
545
+ "learning_rate": 1e-06,
546
+ "logits/chosen": 26.473604202270508,
547
+ "logits/rejected": 25.949445724487305,
548
+ "logps/chosen": -406.0578308105469,
549
+ "logps/rejected": -396.72174072265625,
550
+ "loss": 0.5009,
551
+ "rewards/accuracies": 0.5,
552
+ "rewards/chosen": 0.004218368325382471,
553
+ "rewards/margins": 0.0041790008544921875,
554
+ "rewards/rejected": 3.936793655157089e-05,
555
+ "step": 26
556
+ },
557
+ {
558
+ "debug/policy_chosen_logits": 25.699703216552734,
559
+ "debug/policy_chosen_logps": -412.0224609375,
560
+ "debug/policy_rejected_logits": 22.656808853149414,
561
+ "debug/policy_rejected_logps": -371.5995178222656,
562
+ "debug/reference_chosen_logps": -411.7939453125,
563
+ "debug/reference_rejected_logps": -374.21875,
564
+ "epoch": 0.8181818181818182,
565
+ "grad_norm": 5.565761201808661,
566
+ "learning_rate": 1e-06,
567
+ "logits/chosen": 25.699703216552734,
568
+ "logits/rejected": 22.656808853149414,
569
+ "logps/chosen": -412.0224609375,
570
+ "logps/rejected": -371.5995178222656,
571
+ "loss": 0.5049,
572
+ "rewards/accuracies": 0.0,
573
+ "rewards/chosen": -0.0022848891094326973,
574
+ "rewards/margins": -0.028477095067501068,
575
+ "rewards/rejected": 0.026192206889390945,
576
+ "step": 27
577
+ },
578
+ {
579
+ "debug/policy_chosen_logits": 28.45078468322754,
580
+ "debug/policy_chosen_logps": -388.2000732421875,
581
+ "debug/policy_rejected_logits": 24.951025009155273,
582
+ "debug/policy_rejected_logps": -383.76446533203125,
583
+ "debug/reference_chosen_logps": -389.6691589355469,
584
+ "debug/reference_rejected_logps": -383.29156494140625,
585
+ "epoch": 0.8484848484848485,
586
+ "grad_norm": 5.763656144736158,
587
+ "learning_rate": 1e-06,
588
+ "logits/chosen": 28.45078468322754,
589
+ "logits/rejected": 24.951025009155273,
590
+ "logps/chosen": -388.2000732421875,
591
+ "logps/rejected": -383.76446533203125,
592
+ "loss": 0.4931,
593
+ "rewards/accuracies": 0.625,
594
+ "rewards/chosen": 0.014691008254885674,
595
+ "rewards/margins": 0.019420089200139046,
596
+ "rewards/rejected": -0.004729080013930798,
597
+ "step": 28
598
+ },
599
+ {
600
+ "debug/policy_chosen_logits": 27.9750919342041,
601
+ "debug/policy_chosen_logps": -392.9916076660156,
602
+ "debug/policy_rejected_logits": 30.114521026611328,
603
+ "debug/policy_rejected_logps": -411.13897705078125,
604
+ "debug/reference_chosen_logps": -394.3369140625,
605
+ "debug/reference_rejected_logps": -412.72320556640625,
606
+ "epoch": 0.8787878787878788,
607
+ "grad_norm": 5.254216092353236,
608
+ "learning_rate": 1e-06,
609
+ "logits/chosen": 27.9750919342041,
610
+ "logits/rejected": 30.114521026611328,
611
+ "logps/chosen": -392.9916076660156,
612
+ "logps/rejected": -411.13897705078125,
613
+ "loss": 0.4989,
614
+ "rewards/accuracies": 0.375,
615
+ "rewards/chosen": 0.013453138992190361,
616
+ "rewards/margins": -0.0023886507842689753,
617
+ "rewards/rejected": 0.015841789543628693,
618
+ "step": 29
619
+ },
620
+ {
621
+ "debug/policy_chosen_logits": 29.86224937438965,
622
+ "debug/policy_chosen_logps": -391.58087158203125,
623
+ "debug/policy_rejected_logits": 30.319177627563477,
624
+ "debug/policy_rejected_logps": -398.785400390625,
625
+ "debug/reference_chosen_logps": -391.1322326660156,
626
+ "debug/reference_rejected_logps": -397.64776611328125,
627
+ "epoch": 0.9090909090909091,
628
+ "grad_norm": 5.01142693933464,
629
+ "learning_rate": 1e-06,
630
+ "logits/chosen": 29.86224937438965,
631
+ "logits/rejected": 30.319177627563477,
632
+ "logps/chosen": -391.58087158203125,
633
+ "logps/rejected": -398.785400390625,
634
+ "loss": 0.4928,
635
+ "rewards/accuracies": 0.375,
636
+ "rewards/chosen": -0.0044861603528261185,
637
+ "rewards/margins": 0.006889915093779564,
638
+ "rewards/rejected": -0.011376074515283108,
639
+ "step": 30
640
+ },
641
+ {
642
+ "debug/policy_chosen_logits": 24.41091537475586,
643
+ "debug/policy_chosen_logps": -374.98590087890625,
644
+ "debug/policy_rejected_logits": 25.041379928588867,
645
+ "debug/policy_rejected_logps": -392.67889404296875,
646
+ "debug/reference_chosen_logps": -376.893798828125,
647
+ "debug/reference_rejected_logps": -391.7608337402344,
648
+ "epoch": 0.9393939393939394,
649
+ "grad_norm": 5.68017363572054,
650
+ "learning_rate": 1e-06,
651
+ "logits/chosen": 24.41091537475586,
652
+ "logits/rejected": 25.041379928588867,
653
+ "logps/chosen": -374.98590087890625,
654
+ "logps/rejected": -392.67889404296875,
655
+ "loss": 0.4959,
656
+ "rewards/accuracies": 0.75,
657
+ "rewards/chosen": 0.019078826531767845,
658
+ "rewards/margins": 0.028259430080652237,
659
+ "rewards/rejected": -0.009180602617561817,
660
+ "step": 31
661
+ },
662
+ {
663
+ "debug/policy_chosen_logits": 31.064552307128906,
664
+ "debug/policy_chosen_logps": -416.615966796875,
665
+ "debug/policy_rejected_logits": 27.05549430847168,
666
+ "debug/policy_rejected_logps": -407.7713623046875,
667
+ "debug/reference_chosen_logps": -416.4413757324219,
668
+ "debug/reference_rejected_logps": -407.89178466796875,
669
+ "epoch": 0.9696969696969697,
670
+ "grad_norm": 5.3155545548133,
671
+ "learning_rate": 1e-06,
672
+ "logits/chosen": 31.064552307128906,
673
+ "logits/rejected": 27.05549430847168,
674
+ "logps/chosen": -416.615966796875,
675
+ "logps/rejected": -407.7713623046875,
676
+ "loss": 0.4929,
677
+ "rewards/accuracies": 0.625,
678
+ "rewards/chosen": -0.001745910383760929,
679
+ "rewards/margins": -0.002950286725535989,
680
+ "rewards/rejected": 0.0012043765746057034,
681
+ "step": 32
682
+ },
683
+ {
684
+ "debug/policy_chosen_logits": 28.52979850769043,
685
+ "debug/policy_chosen_logps": -404.65618896484375,
686
+ "debug/policy_rejected_logits": 23.76580047607422,
687
+ "debug/policy_rejected_logps": -377.091796875,
688
+ "debug/reference_chosen_logps": -406.3551940917969,
689
+ "debug/reference_rejected_logps": -375.24481201171875,
690
+ "epoch": 1.0,
691
+ "grad_norm": 4.941707849448349,
692
+ "learning_rate": 1e-06,
693
+ "logits/chosen": 28.52979850769043,
694
+ "logits/rejected": 23.76580047607422,
695
+ "logps/chosen": -404.65618896484375,
696
+ "logps/rejected": -377.091796875,
697
+ "loss": 0.4529,
698
+ "rewards/accuracies": 0.75,
699
+ "rewards/chosen": 0.016990013420581818,
700
+ "rewards/margins": 0.03545970842242241,
701
+ "rewards/rejected": -0.01846969500184059,
702
+ "step": 33
703
+ },
704
+ {
705
+ "epoch": 1.0,
706
+ "step": 33,
707
+ "total_flos": 0.0,
708
+ "train_loss": 0.4962217591025613,
709
+ "train_runtime": 388.3481,
710
+ "train_samples_per_second": 5.302,
711
+ "train_steps_per_second": 0.085
712
+ }
713
+ ],
714
+ "logging_steps": 1,
715
+ "max_steps": 33,
716
+ "num_input_tokens_seen": 0,
717
+ "num_train_epochs": 1,
718
+ "save_steps": 500,
719
+ "stateful_callbacks": {
720
+ "TrainerControl": {
721
+ "args": {
722
+ "should_epoch_stop": false,
723
+ "should_evaluate": false,
724
+ "should_log": false,
725
+ "should_save": true,
726
+ "should_training_stop": true
727
+ },
728
+ "attributes": {}
729
+ }
730
+ },
731
+ "total_flos": 0.0,
732
+ "train_batch_size": 8,
733
+ "trial_name": null,
734
+ "trial_params": null
735
+ }