yiran-wang3 commited on
Commit
c3652ca
1 Parent(s): 1139f1a

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_coder_adamw_iter4
5
+ tags:
6
+ - alignment-handbook
7
+ - generated_from_trainer
8
+ - trl
9
+ - dpo
10
+ datasets:
11
+ - self-generate/ds_coder_sppo_hard_new_cn_mining_oj_iter4-binarized
12
+ model-index:
13
+ - name: ds_coder_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_coder_adamw_iter5
21
+
22
+ This model is a fine-tuned version of [yiran-wang3/ds_coder_adamw_iter4](https://huggingface.co/yiran-wang3/ds_coder_adamw_iter4) on the self-generate/ds_coder_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": 0.4971384649926966,
5
+ "train_runtime": 409.9588,
6
+ "train_samples": 2077,
7
+ "train_samples_per_second": 5.066,
8
+ "train_steps_per_second": 0.08
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": 0.4971384649926966,
5
+ "train_runtime": 409.9588,
6
+ "train_samples": 2077,
7
+ "train_samples_per_second": 5.066,
8
+ "train_steps_per_second": 0.08
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": 29.093868255615234,
13
+ "debug/policy_chosen_logps": -394.1856689453125,
14
+ "debug/policy_rejected_logits": 25.88316535949707,
15
+ "debug/policy_rejected_logps": -393.7677307128906,
16
+ "debug/reference_chosen_logps": -394.1856689453125,
17
+ "debug/reference_rejected_logps": -393.7677307128906,
18
+ "epoch": 0.030303030303030304,
19
+ "grad_norm": 5.030868755913746,
20
+ "learning_rate": 1e-06,
21
+ "logits/chosen": 29.093868255615234,
22
+ "logits/rejected": 25.88316535949707,
23
+ "logps/chosen": -394.1856689453125,
24
+ "logps/rejected": -393.7677307128906,
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": 24.404436111450195,
34
+ "debug/policy_chosen_logps": -377.15850830078125,
35
+ "debug/policy_rejected_logits": 27.354230880737305,
36
+ "debug/policy_rejected_logps": -394.011962890625,
37
+ "debug/reference_chosen_logps": -376.9351806640625,
38
+ "debug/reference_rejected_logps": -393.5160827636719,
39
+ "epoch": 0.06060606060606061,
40
+ "grad_norm": 6.25518297669566,
41
+ "learning_rate": 1e-06,
42
+ "logits/chosen": 24.404436111450195,
43
+ "logits/rejected": 27.354230880737305,
44
+ "logps/chosen": -377.15850830078125,
45
+ "logps/rejected": -394.011962890625,
46
+ "loss": 0.4994,
47
+ "rewards/accuracies": 0.625,
48
+ "rewards/chosen": -0.0022330854553729296,
49
+ "rewards/margins": 0.0027256396133452654,
50
+ "rewards/rejected": -0.004958725534379482,
51
+ "step": 2
52
+ },
53
+ {
54
+ "debug/policy_chosen_logits": 28.0947208404541,
55
+ "debug/policy_chosen_logps": -412.77642822265625,
56
+ "debug/policy_rejected_logits": 26.817684173583984,
57
+ "debug/policy_rejected_logps": -395.5458679199219,
58
+ "debug/reference_chosen_logps": -412.11737060546875,
59
+ "debug/reference_rejected_logps": -395.4435729980469,
60
+ "epoch": 0.09090909090909091,
61
+ "grad_norm": 5.536591360832764,
62
+ "learning_rate": 1e-06,
63
+ "logits/chosen": 28.0947208404541,
64
+ "logits/rejected": 26.817684173583984,
65
+ "logps/chosen": -412.77642822265625,
66
+ "logps/rejected": -395.5458679199219,
67
+ "loss": 0.4984,
68
+ "rewards/accuracies": 0.375,
69
+ "rewards/chosen": -0.00659027136862278,
70
+ "rewards/margins": -0.005567245185375214,
71
+ "rewards/rejected": -0.001023025717586279,
72
+ "step": 3
73
+ },
74
+ {
75
+ "debug/policy_chosen_logits": 24.52150535583496,
76
+ "debug/policy_chosen_logps": -417.7979736328125,
77
+ "debug/policy_rejected_logits": 26.255943298339844,
78
+ "debug/policy_rejected_logps": -416.5947570800781,
79
+ "debug/reference_chosen_logps": -417.81341552734375,
80
+ "debug/reference_rejected_logps": -416.580322265625,
81
+ "epoch": 0.12121212121212122,
82
+ "grad_norm": 5.63081759397413,
83
+ "learning_rate": 1e-06,
84
+ "logits/chosen": 24.52150535583496,
85
+ "logits/rejected": 26.255943298339844,
86
+ "logps/chosen": -417.7979736328125,
87
+ "logps/rejected": -416.5947570800781,
88
+ "loss": 0.5028,
89
+ "rewards/accuracies": 0.625,
90
+ "rewards/chosen": 0.00015426566824316978,
91
+ "rewards/margins": 0.0002988434862345457,
92
+ "rewards/rejected": -0.0001445768866688013,
93
+ "step": 4
94
+ },
95
+ {
96
+ "debug/policy_chosen_logits": 26.004487991333008,
97
+ "debug/policy_chosen_logps": -388.8539733886719,
98
+ "debug/policy_rejected_logits": 28.043724060058594,
99
+ "debug/policy_rejected_logps": -416.6321105957031,
100
+ "debug/reference_chosen_logps": -388.64068603515625,
101
+ "debug/reference_rejected_logps": -416.25311279296875,
102
+ "epoch": 0.15151515151515152,
103
+ "grad_norm": 5.64911268621937,
104
+ "learning_rate": 1e-06,
105
+ "logits/chosen": 26.004487991333008,
106
+ "logits/rejected": 28.043724060058594,
107
+ "logps/chosen": -388.8539733886719,
108
+ "logps/rejected": -416.6321105957031,
109
+ "loss": 0.4994,
110
+ "rewards/accuracies": 0.625,
111
+ "rewards/chosen": -0.0021327207796275616,
112
+ "rewards/margins": 0.0016573716420680285,
113
+ "rewards/rejected": -0.00379009242169559,
114
+ "step": 5
115
+ },
116
+ {
117
+ "debug/policy_chosen_logits": 30.30371856689453,
118
+ "debug/policy_chosen_logps": -390.15362548828125,
119
+ "debug/policy_rejected_logits": 27.941354751586914,
120
+ "debug/policy_rejected_logps": -398.5522155761719,
121
+ "debug/reference_chosen_logps": -390.7499694824219,
122
+ "debug/reference_rejected_logps": -398.3262023925781,
123
+ "epoch": 0.18181818181818182,
124
+ "grad_norm": 7.405074953159111,
125
+ "learning_rate": 1e-06,
126
+ "logits/chosen": 30.30371856689453,
127
+ "logits/rejected": 27.941354751586914,
128
+ "logps/chosen": -390.15362548828125,
129
+ "logps/rejected": -398.5522155761719,
130
+ "loss": 0.4994,
131
+ "rewards/accuracies": 0.625,
132
+ "rewards/chosen": 0.005963439587503672,
133
+ "rewards/margins": 0.008223343640565872,
134
+ "rewards/rejected": -0.0022599028889089823,
135
+ "step": 6
136
+ },
137
+ {
138
+ "debug/policy_chosen_logits": 26.8443603515625,
139
+ "debug/policy_chosen_logps": -393.0838623046875,
140
+ "debug/policy_rejected_logits": 25.924436569213867,
141
+ "debug/policy_rejected_logps": -411.60015869140625,
142
+ "debug/reference_chosen_logps": -393.399169921875,
143
+ "debug/reference_rejected_logps": -410.80096435546875,
144
+ "epoch": 0.21212121212121213,
145
+ "grad_norm": 5.507645597677901,
146
+ "learning_rate": 1e-06,
147
+ "logits/chosen": 26.8443603515625,
148
+ "logits/rejected": 25.924436569213867,
149
+ "logps/chosen": -393.0838623046875,
150
+ "logps/rejected": -411.60015869140625,
151
+ "loss": 0.4981,
152
+ "rewards/accuracies": 0.625,
153
+ "rewards/chosen": 0.003153343452140689,
154
+ "rewards/margins": 0.011145325377583504,
155
+ "rewards/rejected": -0.007991980761289597,
156
+ "step": 7
157
+ },
158
+ {
159
+ "debug/policy_chosen_logits": 26.57024383544922,
160
+ "debug/policy_chosen_logps": -404.73974609375,
161
+ "debug/policy_rejected_logits": 23.179155349731445,
162
+ "debug/policy_rejected_logps": -381.5411376953125,
163
+ "debug/reference_chosen_logps": -405.3680419921875,
164
+ "debug/reference_rejected_logps": -382.30010986328125,
165
+ "epoch": 0.24242424242424243,
166
+ "grad_norm": 5.23273638015015,
167
+ "learning_rate": 1e-06,
168
+ "logits/chosen": 26.57024383544922,
169
+ "logits/rejected": 23.179155349731445,
170
+ "logps/chosen": -404.73974609375,
171
+ "logps/rejected": -381.5411376953125,
172
+ "loss": 0.5002,
173
+ "rewards/accuracies": 0.375,
174
+ "rewards/chosen": 0.00628303550183773,
175
+ "rewards/margins": -0.0013064193772152066,
176
+ "rewards/rejected": 0.007589454762637615,
177
+ "step": 8
178
+ },
179
+ {
180
+ "debug/policy_chosen_logits": 27.397151947021484,
181
+ "debug/policy_chosen_logps": -389.5999755859375,
182
+ "debug/policy_rejected_logits": 26.98357391357422,
183
+ "debug/policy_rejected_logps": -410.5415954589844,
184
+ "debug/reference_chosen_logps": -388.646728515625,
185
+ "debug/reference_rejected_logps": -410.95263671875,
186
+ "epoch": 0.2727272727272727,
187
+ "grad_norm": 4.833818621369551,
188
+ "learning_rate": 1e-06,
189
+ "logits/chosen": 27.397151947021484,
190
+ "logits/rejected": 26.98357391357422,
191
+ "logps/chosen": -389.5999755859375,
192
+ "logps/rejected": -410.5415954589844,
193
+ "loss": 0.5042,
194
+ "rewards/accuracies": 0.125,
195
+ "rewards/chosen": -0.009532546624541283,
196
+ "rewards/margins": -0.013642844744026661,
197
+ "rewards/rejected": 0.004110298119485378,
198
+ "step": 9
199
+ },
200
+ {
201
+ "debug/policy_chosen_logits": 26.80759620666504,
202
+ "debug/policy_chosen_logps": -399.30169677734375,
203
+ "debug/policy_rejected_logits": 28.478506088256836,
204
+ "debug/policy_rejected_logps": -394.08697509765625,
205
+ "debug/reference_chosen_logps": -398.6517333984375,
206
+ "debug/reference_rejected_logps": -394.05316162109375,
207
+ "epoch": 0.30303030303030304,
208
+ "grad_norm": 5.107308392124932,
209
+ "learning_rate": 1e-06,
210
+ "logits/chosen": 26.80759620666504,
211
+ "logits/rejected": 28.478506088256836,
212
+ "logps/chosen": -399.30169677734375,
213
+ "logps/rejected": -394.08697509765625,
214
+ "loss": 0.5001,
215
+ "rewards/accuracies": 0.375,
216
+ "rewards/chosen": -0.006500015035271645,
217
+ "rewards/margins": -0.0061615752056241035,
218
+ "rewards/rejected": -0.00033844029530882835,
219
+ "step": 10
220
+ },
221
+ {
222
+ "debug/policy_chosen_logits": 26.44737434387207,
223
+ "debug/policy_chosen_logps": -408.57733154296875,
224
+ "debug/policy_rejected_logits": 26.65883445739746,
225
+ "debug/policy_rejected_logps": -391.9090576171875,
226
+ "debug/reference_chosen_logps": -409.13458251953125,
227
+ "debug/reference_rejected_logps": -391.6588439941406,
228
+ "epoch": 0.3333333333333333,
229
+ "grad_norm": 4.974671977298996,
230
+ "learning_rate": 1e-06,
231
+ "logits/chosen": 26.44737434387207,
232
+ "logits/rejected": 26.65883445739746,
233
+ "logps/chosen": -408.57733154296875,
234
+ "logps/rejected": -391.9090576171875,
235
+ "loss": 0.4995,
236
+ "rewards/accuracies": 0.625,
237
+ "rewards/chosen": 0.005572776310145855,
238
+ "rewards/margins": 0.008074606768786907,
239
+ "rewards/rejected": -0.002501830691471696,
240
+ "step": 11
241
+ },
242
+ {
243
+ "debug/policy_chosen_logits": 27.20985221862793,
244
+ "debug/policy_chosen_logps": -377.276611328125,
245
+ "debug/policy_rejected_logits": 28.053346633911133,
246
+ "debug/policy_rejected_logps": -383.6709289550781,
247
+ "debug/reference_chosen_logps": -376.66357421875,
248
+ "debug/reference_rejected_logps": -383.989990234375,
249
+ "epoch": 0.36363636363636365,
250
+ "grad_norm": 5.327027762788645,
251
+ "learning_rate": 1e-06,
252
+ "logits/chosen": 27.20985221862793,
253
+ "logits/rejected": 28.053346633911133,
254
+ "logps/chosen": -377.276611328125,
255
+ "logps/rejected": -383.6709289550781,
256
+ "loss": 0.4994,
257
+ "rewards/accuracies": 0.25,
258
+ "rewards/chosen": -0.00613048579543829,
259
+ "rewards/margins": -0.00932083185762167,
260
+ "rewards/rejected": 0.003190345596522093,
261
+ "step": 12
262
+ },
263
+ {
264
+ "debug/policy_chosen_logits": 25.8612117767334,
265
+ "debug/policy_chosen_logps": -392.2992858886719,
266
+ "debug/policy_rejected_logits": 24.847339630126953,
267
+ "debug/policy_rejected_logps": -402.8356628417969,
268
+ "debug/reference_chosen_logps": -391.6371765136719,
269
+ "debug/reference_rejected_logps": -402.23583984375,
270
+ "epoch": 0.3939393939393939,
271
+ "grad_norm": 5.361687372093286,
272
+ "learning_rate": 1e-06,
273
+ "logits/chosen": 25.8612117767334,
274
+ "logits/rejected": 24.847339630126953,
275
+ "logps/chosen": -392.2992858886719,
276
+ "logps/rejected": -402.8356628417969,
277
+ "loss": 0.496,
278
+ "rewards/accuracies": 0.375,
279
+ "rewards/chosen": -0.0066207884810864925,
280
+ "rewards/margins": -0.0006225210381671786,
281
+ "rewards/rejected": -0.005998267792165279,
282
+ "step": 13
283
+ },
284
+ {
285
+ "debug/policy_chosen_logits": 29.508888244628906,
286
+ "debug/policy_chosen_logps": -387.1098327636719,
287
+ "debug/policy_rejected_logits": 29.381460189819336,
288
+ "debug/policy_rejected_logps": -405.422607421875,
289
+ "debug/reference_chosen_logps": -387.8492126464844,
290
+ "debug/reference_rejected_logps": -406.4776306152344,
291
+ "epoch": 0.42424242424242425,
292
+ "grad_norm": 5.461096503844373,
293
+ "learning_rate": 1e-06,
294
+ "logits/chosen": 29.508888244628906,
295
+ "logits/rejected": 29.381460189819336,
296
+ "logps/chosen": -387.1098327636719,
297
+ "logps/rejected": -405.422607421875,
298
+ "loss": 0.4966,
299
+ "rewards/accuracies": 0.25,
300
+ "rewards/chosen": 0.007393989246338606,
301
+ "rewards/margins": -0.0031561660580337048,
302
+ "rewards/rejected": 0.010550156235694885,
303
+ "step": 14
304
+ },
305
+ {
306
+ "debug/policy_chosen_logits": 30.304807662963867,
307
+ "debug/policy_chosen_logps": -388.3260498046875,
308
+ "debug/policy_rejected_logits": 29.830333709716797,
309
+ "debug/policy_rejected_logps": -399.8182373046875,
310
+ "debug/reference_chosen_logps": -387.77960205078125,
311
+ "debug/reference_rejected_logps": -398.8104553222656,
312
+ "epoch": 0.45454545454545453,
313
+ "grad_norm": 5.250363808071301,
314
+ "learning_rate": 1e-06,
315
+ "logits/chosen": 30.304807662963867,
316
+ "logits/rejected": 29.830333709716797,
317
+ "logps/chosen": -388.3260498046875,
318
+ "logps/rejected": -399.8182373046875,
319
+ "loss": 0.5039,
320
+ "rewards/accuracies": 0.625,
321
+ "rewards/chosen": -0.0054648201912641525,
322
+ "rewards/margins": 0.004612846300005913,
323
+ "rewards/rejected": -0.01007766742259264,
324
+ "step": 15
325
+ },
326
+ {
327
+ "debug/policy_chosen_logits": 28.09935188293457,
328
+ "debug/policy_chosen_logps": -397.7403259277344,
329
+ "debug/policy_rejected_logits": 31.767385482788086,
330
+ "debug/policy_rejected_logps": -376.7595520019531,
331
+ "debug/reference_chosen_logps": -397.562744140625,
332
+ "debug/reference_rejected_logps": -377.139892578125,
333
+ "epoch": 0.48484848484848486,
334
+ "grad_norm": 5.660459055452235,
335
+ "learning_rate": 1e-06,
336
+ "logits/chosen": 28.09935188293457,
337
+ "logits/rejected": 31.767385482788086,
338
+ "logps/chosen": -397.7403259277344,
339
+ "logps/rejected": -376.7595520019531,
340
+ "loss": 0.4983,
341
+ "rewards/accuracies": 0.125,
342
+ "rewards/chosen": -0.0017757411114871502,
343
+ "rewards/margins": -0.00557926157489419,
344
+ "rewards/rejected": 0.0038035199977457523,
345
+ "step": 16
346
+ },
347
+ {
348
+ "debug/policy_chosen_logits": 26.62754249572754,
349
+ "debug/policy_chosen_logps": -388.61407470703125,
350
+ "debug/policy_rejected_logits": 27.771495819091797,
351
+ "debug/policy_rejected_logps": -422.8292541503906,
352
+ "debug/reference_chosen_logps": -389.57568359375,
353
+ "debug/reference_rejected_logps": -422.9500732421875,
354
+ "epoch": 0.5151515151515151,
355
+ "grad_norm": 5.458787195966993,
356
+ "learning_rate": 1e-06,
357
+ "logits/chosen": 26.62754249572754,
358
+ "logits/rejected": 27.771495819091797,
359
+ "logps/chosen": -388.61407470703125,
360
+ "logps/rejected": -422.8292541503906,
361
+ "loss": 0.4997,
362
+ "rewards/accuracies": 0.625,
363
+ "rewards/chosen": 0.009616240859031677,
364
+ "rewards/margins": 0.008408088237047195,
365
+ "rewards/rejected": 0.0012081530876457691,
366
+ "step": 17
367
+ },
368
+ {
369
+ "debug/policy_chosen_logits": 26.25288200378418,
370
+ "debug/policy_chosen_logps": -395.8739318847656,
371
+ "debug/policy_rejected_logits": 26.34918975830078,
372
+ "debug/policy_rejected_logps": -418.2298889160156,
373
+ "debug/reference_chosen_logps": -396.3597412109375,
374
+ "debug/reference_rejected_logps": -419.4313049316406,
375
+ "epoch": 0.5454545454545454,
376
+ "grad_norm": 6.045438153278639,
377
+ "learning_rate": 1e-06,
378
+ "logits/chosen": 26.25288200378418,
379
+ "logits/rejected": 26.34918975830078,
380
+ "logps/chosen": -395.8739318847656,
381
+ "logps/rejected": -418.2298889160156,
382
+ "loss": 0.4971,
383
+ "rewards/accuracies": 0.375,
384
+ "rewards/chosen": 0.004857787862420082,
385
+ "rewards/margins": -0.007156638894230127,
386
+ "rewards/rejected": 0.012014427222311497,
387
+ "step": 18
388
+ },
389
+ {
390
+ "debug/policy_chosen_logits": 28.588306427001953,
391
+ "debug/policy_chosen_logps": -380.94671630859375,
392
+ "debug/policy_rejected_logits": 29.382522583007812,
393
+ "debug/policy_rejected_logps": -404.1590576171875,
394
+ "debug/reference_chosen_logps": -381.28717041015625,
395
+ "debug/reference_rejected_logps": -404.83154296875,
396
+ "epoch": 0.5757575757575758,
397
+ "grad_norm": 5.5743203305237,
398
+ "learning_rate": 1e-06,
399
+ "logits/chosen": 28.588306427001953,
400
+ "logits/rejected": 29.382522583007812,
401
+ "logps/chosen": -380.94671630859375,
402
+ "logps/rejected": -404.1590576171875,
403
+ "loss": 0.5007,
404
+ "rewards/accuracies": 0.5,
405
+ "rewards/chosen": 0.0034041597973555326,
406
+ "rewards/margins": -0.0033205407671630383,
407
+ "rewards/rejected": 0.00672469986602664,
408
+ "step": 19
409
+ },
410
+ {
411
+ "debug/policy_chosen_logits": 24.93841552734375,
412
+ "debug/policy_chosen_logps": -415.7802734375,
413
+ "debug/policy_rejected_logits": 27.447294235229492,
414
+ "debug/policy_rejected_logps": -404.31024169921875,
415
+ "debug/reference_chosen_logps": -416.76080322265625,
416
+ "debug/reference_rejected_logps": -404.63677978515625,
417
+ "epoch": 0.6060606060606061,
418
+ "grad_norm": 5.261485130685494,
419
+ "learning_rate": 1e-06,
420
+ "logits/chosen": 24.93841552734375,
421
+ "logits/rejected": 27.447294235229492,
422
+ "logps/chosen": -415.7802734375,
423
+ "logps/rejected": -404.31024169921875,
424
+ "loss": 0.4912,
425
+ "rewards/accuracies": 0.5,
426
+ "rewards/chosen": 0.009805488400161266,
427
+ "rewards/margins": 0.0065397643484175205,
428
+ "rewards/rejected": 0.003265724051743746,
429
+ "step": 20
430
+ },
431
+ {
432
+ "debug/policy_chosen_logits": 26.202421188354492,
433
+ "debug/policy_chosen_logps": -425.5368957519531,
434
+ "debug/policy_rejected_logits": 24.03887939453125,
435
+ "debug/policy_rejected_logps": -397.7702941894531,
436
+ "debug/reference_chosen_logps": -426.2981262207031,
437
+ "debug/reference_rejected_logps": -397.6723937988281,
438
+ "epoch": 0.6363636363636364,
439
+ "grad_norm": 5.590167982725568,
440
+ "learning_rate": 1e-06,
441
+ "logits/chosen": 26.202421188354492,
442
+ "logits/rejected": 24.03887939453125,
443
+ "logps/chosen": -425.5368957519531,
444
+ "logps/rejected": -397.7702941894531,
445
+ "loss": 0.4985,
446
+ "rewards/accuracies": 0.875,
447
+ "rewards/chosen": 0.007612571120262146,
448
+ "rewards/margins": 0.008591575548052788,
449
+ "rewards/rejected": -0.0009790037292987108,
450
+ "step": 21
451
+ },
452
+ {
453
+ "debug/policy_chosen_logits": 28.777624130249023,
454
+ "debug/policy_chosen_logps": -383.98968505859375,
455
+ "debug/policy_rejected_logits": 30.540597915649414,
456
+ "debug/policy_rejected_logps": -404.1961975097656,
457
+ "debug/reference_chosen_logps": -385.23529052734375,
458
+ "debug/reference_rejected_logps": -405.3203125,
459
+ "epoch": 0.6666666666666666,
460
+ "grad_norm": 6.204641362035411,
461
+ "learning_rate": 1e-06,
462
+ "logits/chosen": 28.777624130249023,
463
+ "logits/rejected": 30.540597915649414,
464
+ "logps/chosen": -383.98968505859375,
465
+ "logps/rejected": -404.1961975097656,
466
+ "loss": 0.5028,
467
+ "rewards/accuracies": 0.375,
468
+ "rewards/chosen": 0.012456131167709827,
469
+ "rewards/margins": 0.001215210184454918,
470
+ "rewards/rejected": 0.011240921914577484,
471
+ "step": 22
472
+ },
473
+ {
474
+ "debug/policy_chosen_logits": 29.740785598754883,
475
+ "debug/policy_chosen_logps": -386.64471435546875,
476
+ "debug/policy_rejected_logits": 28.254274368286133,
477
+ "debug/policy_rejected_logps": -388.5518493652344,
478
+ "debug/reference_chosen_logps": -386.5996398925781,
479
+ "debug/reference_rejected_logps": -387.55853271484375,
480
+ "epoch": 0.696969696969697,
481
+ "grad_norm": 5.534634639642078,
482
+ "learning_rate": 1e-06,
483
+ "logits/chosen": 29.740785598754883,
484
+ "logits/rejected": 28.254274368286133,
485
+ "logps/chosen": -386.64471435546875,
486
+ "logps/rejected": -388.5518493652344,
487
+ "loss": 0.4938,
488
+ "rewards/accuracies": 0.75,
489
+ "rewards/chosen": -0.000450668390840292,
490
+ "rewards/margins": 0.00948253646492958,
491
+ "rewards/rejected": -0.00993320532143116,
492
+ "step": 23
493
+ },
494
+ {
495
+ "debug/policy_chosen_logits": 27.11961555480957,
496
+ "debug/policy_chosen_logps": -401.6017761230469,
497
+ "debug/policy_rejected_logits": 27.432086944580078,
498
+ "debug/policy_rejected_logps": -412.71929931640625,
499
+ "debug/reference_chosen_logps": -400.8846130371094,
500
+ "debug/reference_rejected_logps": -412.95587158203125,
501
+ "epoch": 0.7272727272727273,
502
+ "grad_norm": 5.536134798973258,
503
+ "learning_rate": 1e-06,
504
+ "logits/chosen": 27.11961555480957,
505
+ "logits/rejected": 27.432086944580078,
506
+ "logps/chosen": -401.6017761230469,
507
+ "logps/rejected": -412.71929931640625,
508
+ "loss": 0.492,
509
+ "rewards/accuracies": 0.25,
510
+ "rewards/chosen": -0.007171629928052425,
511
+ "rewards/margins": -0.00953700952231884,
512
+ "rewards/rejected": 0.0023653795942664146,
513
+ "step": 24
514
+ },
515
+ {
516
+ "debug/policy_chosen_logits": 24.8132266998291,
517
+ "debug/policy_chosen_logps": -396.219482421875,
518
+ "debug/policy_rejected_logits": 27.473514556884766,
519
+ "debug/policy_rejected_logps": -407.45550537109375,
520
+ "debug/reference_chosen_logps": -395.4239501953125,
521
+ "debug/reference_rejected_logps": -405.3516540527344,
522
+ "epoch": 0.7575757575757576,
523
+ "grad_norm": 5.248733249901598,
524
+ "learning_rate": 1e-06,
525
+ "logits/chosen": 24.8132266998291,
526
+ "logits/rejected": 27.473514556884766,
527
+ "logps/chosen": -396.219482421875,
528
+ "logps/rejected": -407.45550537109375,
529
+ "loss": 0.4923,
530
+ "rewards/accuracies": 0.875,
531
+ "rewards/chosen": -0.007955474779009819,
532
+ "rewards/margins": 0.013082961551845074,
533
+ "rewards/rejected": -0.021038437262177467,
534
+ "step": 25
535
+ },
536
+ {
537
+ "debug/policy_chosen_logits": 26.792572021484375,
538
+ "debug/policy_chosen_logps": -388.32012939453125,
539
+ "debug/policy_rejected_logits": 26.965463638305664,
540
+ "debug/policy_rejected_logps": -401.08880615234375,
541
+ "debug/reference_chosen_logps": -388.4382629394531,
542
+ "debug/reference_rejected_logps": -399.5977783203125,
543
+ "epoch": 0.7878787878787878,
544
+ "grad_norm": 5.024632086381175,
545
+ "learning_rate": 1e-06,
546
+ "logits/chosen": 26.792572021484375,
547
+ "logits/rejected": 26.965463638305664,
548
+ "logps/chosen": -388.32012939453125,
549
+ "logps/rejected": -401.08880615234375,
550
+ "loss": 0.4907,
551
+ "rewards/accuracies": 0.875,
552
+ "rewards/chosen": 0.0011813733726739883,
553
+ "rewards/margins": 0.016091765835881233,
554
+ "rewards/rejected": -0.014910392463207245,
555
+ "step": 26
556
+ },
557
+ {
558
+ "debug/policy_chosen_logits": 21.125619888305664,
559
+ "debug/policy_chosen_logps": -363.61517333984375,
560
+ "debug/policy_rejected_logits": 26.54195213317871,
561
+ "debug/policy_rejected_logps": -408.9364013671875,
562
+ "debug/reference_chosen_logps": -363.9629821777344,
563
+ "debug/reference_rejected_logps": -408.07269287109375,
564
+ "epoch": 0.8181818181818182,
565
+ "grad_norm": 5.432070893156278,
566
+ "learning_rate": 1e-06,
567
+ "logits/chosen": 21.125619888305664,
568
+ "logits/rejected": 26.54195213317871,
569
+ "logps/chosen": -363.61517333984375,
570
+ "logps/rejected": -408.9364013671875,
571
+ "loss": 0.4934,
572
+ "rewards/accuracies": 0.625,
573
+ "rewards/chosen": 0.0034781266003847122,
574
+ "rewards/margins": 0.012114867568016052,
575
+ "rewards/rejected": -0.008636741898953915,
576
+ "step": 27
577
+ },
578
+ {
579
+ "debug/policy_chosen_logits": 24.872333526611328,
580
+ "debug/policy_chosen_logps": -372.0343933105469,
581
+ "debug/policy_rejected_logits": 26.66975975036621,
582
+ "debug/policy_rejected_logps": -398.46728515625,
583
+ "debug/reference_chosen_logps": -371.37255859375,
584
+ "debug/reference_rejected_logps": -398.9984130859375,
585
+ "epoch": 0.8484848484848485,
586
+ "grad_norm": 5.03229338667667,
587
+ "learning_rate": 1e-06,
588
+ "logits/chosen": 24.872333526611328,
589
+ "logits/rejected": 26.66975975036621,
590
+ "logps/chosen": -372.0343933105469,
591
+ "logps/rejected": -398.46728515625,
592
+ "loss": 0.4967,
593
+ "rewards/accuracies": 0.5,
594
+ "rewards/chosen": -0.006618117913603783,
595
+ "rewards/margins": -0.011929434724152088,
596
+ "rewards/rejected": 0.0053113182075321674,
597
+ "step": 28
598
+ },
599
+ {
600
+ "debug/policy_chosen_logits": 29.60181427001953,
601
+ "debug/policy_chosen_logps": -406.96124267578125,
602
+ "debug/policy_rejected_logits": 28.26951026916504,
603
+ "debug/policy_rejected_logps": -390.26763916015625,
604
+ "debug/reference_chosen_logps": -407.25262451171875,
605
+ "debug/reference_rejected_logps": -390.38494873046875,
606
+ "epoch": 0.8787878787878788,
607
+ "grad_norm": 5.040026373746743,
608
+ "learning_rate": 1e-06,
609
+ "logits/chosen": 29.60181427001953,
610
+ "logits/rejected": 28.26951026916504,
611
+ "logps/chosen": -406.96124267578125,
612
+ "logps/rejected": -390.26763916015625,
613
+ "loss": 0.494,
614
+ "rewards/accuracies": 0.25,
615
+ "rewards/chosen": 0.0029140852857381105,
616
+ "rewards/margins": 0.0017412565648555756,
617
+ "rewards/rejected": 0.0011728284880518913,
618
+ "step": 29
619
+ },
620
+ {
621
+ "debug/policy_chosen_logits": 27.035560607910156,
622
+ "debug/policy_chosen_logps": -392.25396728515625,
623
+ "debug/policy_rejected_logits": 27.442691802978516,
624
+ "debug/policy_rejected_logps": -391.0264892578125,
625
+ "debug/reference_chosen_logps": -392.8270263671875,
626
+ "debug/reference_rejected_logps": -390.5660400390625,
627
+ "epoch": 0.9090909090909091,
628
+ "grad_norm": 5.259600833580871,
629
+ "learning_rate": 1e-06,
630
+ "logits/chosen": 27.035560607910156,
631
+ "logits/rejected": 27.442691802978516,
632
+ "logps/chosen": -392.25396728515625,
633
+ "logps/rejected": -391.0264892578125,
634
+ "loss": 0.498,
635
+ "rewards/accuracies": 0.625,
636
+ "rewards/chosen": 0.005730743054300547,
637
+ "rewards/margins": 0.01033527310937643,
638
+ "rewards/rejected": -0.0046045295894145966,
639
+ "step": 30
640
+ },
641
+ {
642
+ "debug/policy_chosen_logits": 23.07676887512207,
643
+ "debug/policy_chosen_logps": -368.69171142578125,
644
+ "debug/policy_rejected_logits": 27.108469009399414,
645
+ "debug/policy_rejected_logps": -401.52239990234375,
646
+ "debug/reference_chosen_logps": -369.28192138671875,
647
+ "debug/reference_rejected_logps": -402.61138916015625,
648
+ "epoch": 0.9393939393939394,
649
+ "grad_norm": 5.210897913069892,
650
+ "learning_rate": 1e-06,
651
+ "logits/chosen": 23.07676887512207,
652
+ "logits/rejected": 27.108469009399414,
653
+ "logps/chosen": -368.69171142578125,
654
+ "logps/rejected": -401.52239990234375,
655
+ "loss": 0.5014,
656
+ "rewards/accuracies": 0.25,
657
+ "rewards/chosen": 0.005902099423110485,
658
+ "rewards/margins": -0.004987678490579128,
659
+ "rewards/rejected": 0.010889777913689613,
660
+ "step": 31
661
+ },
662
+ {
663
+ "debug/policy_chosen_logits": 26.6132869720459,
664
+ "debug/policy_chosen_logps": -397.4622802734375,
665
+ "debug/policy_rejected_logits": 27.003686904907227,
666
+ "debug/policy_rejected_logps": -413.6773681640625,
667
+ "debug/reference_chosen_logps": -397.5897216796875,
668
+ "debug/reference_rejected_logps": -411.7838134765625,
669
+ "epoch": 0.9696969696969697,
670
+ "grad_norm": 5.060163056079328,
671
+ "learning_rate": 1e-06,
672
+ "logits/chosen": 26.6132869720459,
673
+ "logits/rejected": 27.003686904907227,
674
+ "logps/chosen": -397.4622802734375,
675
+ "logps/rejected": -413.6773681640625,
676
+ "loss": 0.498,
677
+ "rewards/accuracies": 0.75,
678
+ "rewards/chosen": 0.0012746425345540047,
679
+ "rewards/margins": 0.020210150629281998,
680
+ "rewards/rejected": -0.018935509026050568,
681
+ "step": 32
682
+ },
683
+ {
684
+ "debug/policy_chosen_logits": 25.71558952331543,
685
+ "debug/policy_chosen_logps": -366.22833251953125,
686
+ "debug/policy_rejected_logits": 26.291645050048828,
687
+ "debug/policy_rejected_logps": -416.4896545410156,
688
+ "debug/reference_chosen_logps": -366.5025939941406,
689
+ "debug/reference_rejected_logps": -416.4383544921875,
690
+ "epoch": 1.0,
691
+ "grad_norm": 5.67721223467369,
692
+ "learning_rate": 1e-06,
693
+ "logits/chosen": 25.71558952331543,
694
+ "logits/rejected": 26.291645050048828,
695
+ "logps/chosen": -366.22833251953125,
696
+ "logps/rejected": -416.4896545410156,
697
+ "loss": 0.4695,
698
+ "rewards/accuracies": 0.625,
699
+ "rewards/chosen": 0.0027426911983639,
700
+ "rewards/margins": 0.003255615010857582,
701
+ "rewards/rejected": -0.0005129238124936819,
702
+ "step": 33
703
+ },
704
+ {
705
+ "epoch": 1.0,
706
+ "step": 33,
707
+ "total_flos": 0.0,
708
+ "train_loss": 0.4971384649926966,
709
+ "train_runtime": 409.9588,
710
+ "train_samples_per_second": 5.066,
711
+ "train_steps_per_second": 0.08
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
+ }