yiran-wang3 commited on
Commit
eeef5a8
1 Parent(s): 7770691

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_coder_reflct_adamw_iter7
5
+ tags:
6
+ - alignment-handbook
7
+ - generated_from_trainer
8
+ - trl
9
+ - dpo
10
+ datasets:
11
+ - self-generate/qwcoder2_reflct_sppo_hard_new_cn_mining_oj_iter7-binarized-reflection-scored
12
+ model-index:
13
+ - name: qwen2_coder_reflct_adamw_iter8
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_coder_reflct_adamw_iter8
21
+
22
+ This model is a fine-tuned version of [yiran-wang3/qwen2_coder_reflct_adamw_iter7](https://huggingface.co/yiran-wang3/qwen2_coder_reflct_adamw_iter7) on the self-generate/qwcoder2_reflct_sppo_hard_new_cn_mining_oj_iter7-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": 0.4911675299367597,
5
+ "train_runtime": 120.215,
6
+ "train_samples": 1968,
7
+ "train_samples_per_second": 16.371,
8
+ "train_steps_per_second": 0.258
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": 0.4911675299367597,
5
+ "train_runtime": 120.215,
6
+ "train_samples": 1968,
7
+ "train_samples_per_second": 16.371,
8
+ "train_steps_per_second": 0.258
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,693 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 100,
6
+ "global_step": 31,
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": -3.0837113857269287,
13
+ "debug/policy_chosen_logps": -156.1900634765625,
14
+ "debug/policy_rejected_logits": -3.0168697834014893,
15
+ "debug/policy_rejected_logps": -213.46035766601562,
16
+ "debug/reference_chosen_logps": -156.1900634765625,
17
+ "debug/reference_rejected_logps": -213.46035766601562,
18
+ "epoch": 0.03225806451612903,
19
+ "grad_norm": 33.635131555501395,
20
+ "learning_rate": 1e-06,
21
+ "logits/chosen": -3.0837113857269287,
22
+ "logits/rejected": -3.0168697834014893,
23
+ "logps/chosen": -156.1900634765625,
24
+ "logps/rejected": -213.46035766601562,
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": -3.188256025314331,
34
+ "debug/policy_chosen_logps": -173.60809326171875,
35
+ "debug/policy_rejected_logits": -3.2439301013946533,
36
+ "debug/policy_rejected_logps": -204.31166076660156,
37
+ "debug/reference_chosen_logps": -172.73226928710938,
38
+ "debug/reference_rejected_logps": -198.9984588623047,
39
+ "epoch": 0.06451612903225806,
40
+ "grad_norm": 12.890042733243831,
41
+ "learning_rate": 1e-06,
42
+ "logits/chosen": -3.188256025314331,
43
+ "logits/rejected": -3.2439301013946533,
44
+ "logps/chosen": -173.60809326171875,
45
+ "logps/rejected": -204.31166076660156,
46
+ "loss": 0.498,
47
+ "rewards/accuracies": 0.625,
48
+ "rewards/chosen": -0.008758220821619034,
49
+ "rewards/margins": 0.04437387362122536,
50
+ "rewards/rejected": -0.05313209444284439,
51
+ "step": 2
52
+ },
53
+ {
54
+ "debug/policy_chosen_logits": -3.208176374435425,
55
+ "debug/policy_chosen_logps": -172.39854431152344,
56
+ "debug/policy_rejected_logits": -3.1775765419006348,
57
+ "debug/policy_rejected_logps": -211.1417236328125,
58
+ "debug/reference_chosen_logps": -171.08482360839844,
59
+ "debug/reference_rejected_logps": -208.74172973632812,
60
+ "epoch": 0.0967741935483871,
61
+ "grad_norm": 19.598718664572683,
62
+ "learning_rate": 1e-06,
63
+ "logits/chosen": -3.208176374435425,
64
+ "logits/rejected": -3.1775765419006348,
65
+ "logps/chosen": -172.39854431152344,
66
+ "logps/rejected": -211.1417236328125,
67
+ "loss": 0.4905,
68
+ "rewards/accuracies": 0.5,
69
+ "rewards/chosen": -0.013137312605977058,
70
+ "rewards/margins": 0.010862598195672035,
71
+ "rewards/rejected": -0.023999910801649094,
72
+ "step": 3
73
+ },
74
+ {
75
+ "debug/policy_chosen_logits": -3.0628340244293213,
76
+ "debug/policy_chosen_logps": -170.09300231933594,
77
+ "debug/policy_rejected_logits": -2.9403834342956543,
78
+ "debug/policy_rejected_logps": -206.1832275390625,
79
+ "debug/reference_chosen_logps": -162.90957641601562,
80
+ "debug/reference_rejected_logps": -198.26235961914062,
81
+ "epoch": 0.12903225806451613,
82
+ "grad_norm": 19.89756602440321,
83
+ "learning_rate": 1e-06,
84
+ "logits/chosen": -3.0628340244293213,
85
+ "logits/rejected": -2.9403834342956543,
86
+ "logps/chosen": -170.09300231933594,
87
+ "logps/rejected": -206.1832275390625,
88
+ "loss": 0.4921,
89
+ "rewards/accuracies": 0.375,
90
+ "rewards/chosen": -0.07183413207530975,
91
+ "rewards/margins": 0.0073744915425777435,
92
+ "rewards/rejected": -0.0792086198925972,
93
+ "step": 4
94
+ },
95
+ {
96
+ "debug/policy_chosen_logits": -3.1047394275665283,
97
+ "debug/policy_chosen_logps": -169.22952270507812,
98
+ "debug/policy_rejected_logits": -3.178271532058716,
99
+ "debug/policy_rejected_logps": -193.72286987304688,
100
+ "debug/reference_chosen_logps": -168.54092407226562,
101
+ "debug/reference_rejected_logps": -186.80690002441406,
102
+ "epoch": 0.16129032258064516,
103
+ "grad_norm": 13.088334676288056,
104
+ "learning_rate": 1e-06,
105
+ "logits/chosen": -3.1047394275665283,
106
+ "logits/rejected": -3.178271532058716,
107
+ "logps/chosen": -169.22952270507812,
108
+ "logps/rejected": -193.72286987304688,
109
+ "loss": 0.4858,
110
+ "rewards/accuracies": 0.75,
111
+ "rewards/chosen": -0.006886100862175226,
112
+ "rewards/margins": 0.062273673713207245,
113
+ "rewards/rejected": -0.06915977597236633,
114
+ "step": 5
115
+ },
116
+ {
117
+ "debug/policy_chosen_logits": -3.1648635864257812,
118
+ "debug/policy_chosen_logps": -161.87496948242188,
119
+ "debug/policy_rejected_logits": -3.1185457706451416,
120
+ "debug/policy_rejected_logps": -213.39263916015625,
121
+ "debug/reference_chosen_logps": -154.76771545410156,
122
+ "debug/reference_rejected_logps": -211.57601928710938,
123
+ "epoch": 0.1935483870967742,
124
+ "grad_norm": 67.05357176610438,
125
+ "learning_rate": 1e-06,
126
+ "logits/chosen": -3.1648635864257812,
127
+ "logits/rejected": -3.1185457706451416,
128
+ "logps/chosen": -161.87496948242188,
129
+ "logps/rejected": -213.39263916015625,
130
+ "loss": 0.5157,
131
+ "rewards/accuracies": 0.375,
132
+ "rewards/chosen": -0.07107262313365936,
133
+ "rewards/margins": -0.052906349301338196,
134
+ "rewards/rejected": -0.018166273832321167,
135
+ "step": 6
136
+ },
137
+ {
138
+ "debug/policy_chosen_logits": -3.2189788818359375,
139
+ "debug/policy_chosen_logps": -177.56878662109375,
140
+ "debug/policy_rejected_logits": -3.11997389793396,
141
+ "debug/policy_rejected_logps": -189.31292724609375,
142
+ "debug/reference_chosen_logps": -179.44760131835938,
143
+ "debug/reference_rejected_logps": -187.13697814941406,
144
+ "epoch": 0.22580645161290322,
145
+ "grad_norm": 11.843697734815683,
146
+ "learning_rate": 1e-06,
147
+ "logits/chosen": -3.2189788818359375,
148
+ "logits/rejected": -3.11997389793396,
149
+ "logps/chosen": -177.56878662109375,
150
+ "logps/rejected": -189.31292724609375,
151
+ "loss": 0.4878,
152
+ "rewards/accuracies": 0.75,
153
+ "rewards/chosen": 0.01878824271261692,
154
+ "rewards/margins": 0.04054774343967438,
155
+ "rewards/rejected": -0.021759498864412308,
156
+ "step": 7
157
+ },
158
+ {
159
+ "debug/policy_chosen_logits": -3.2227280139923096,
160
+ "debug/policy_chosen_logps": -163.78762817382812,
161
+ "debug/policy_rejected_logits": -3.290872573852539,
162
+ "debug/policy_rejected_logps": -193.32077026367188,
163
+ "debug/reference_chosen_logps": -165.69235229492188,
164
+ "debug/reference_rejected_logps": -189.6918182373047,
165
+ "epoch": 0.25806451612903225,
166
+ "grad_norm": 15.145376881649522,
167
+ "learning_rate": 1e-06,
168
+ "logits/chosen": -3.2227280139923096,
169
+ "logits/rejected": -3.290872573852539,
170
+ "logps/chosen": -163.78762817382812,
171
+ "logps/rejected": -193.32077026367188,
172
+ "loss": 0.4912,
173
+ "rewards/accuracies": 0.625,
174
+ "rewards/chosen": 0.019047202542424202,
175
+ "rewards/margins": 0.05533674359321594,
176
+ "rewards/rejected": -0.03628953918814659,
177
+ "step": 8
178
+ },
179
+ {
180
+ "debug/policy_chosen_logits": -3.159303903579712,
181
+ "debug/policy_chosen_logps": -167.47830200195312,
182
+ "debug/policy_rejected_logits": -3.1892285346984863,
183
+ "debug/policy_rejected_logps": -166.064208984375,
184
+ "debug/reference_chosen_logps": -168.6324462890625,
185
+ "debug/reference_rejected_logps": -168.05596923828125,
186
+ "epoch": 0.2903225806451613,
187
+ "grad_norm": 14.37792159617078,
188
+ "learning_rate": 1e-06,
189
+ "logits/chosen": -3.159303903579712,
190
+ "logits/rejected": -3.1892285346984863,
191
+ "logps/chosen": -167.47830200195312,
192
+ "logps/rejected": -166.064208984375,
193
+ "loss": 0.4962,
194
+ "rewards/accuracies": 0.75,
195
+ "rewards/chosen": 0.011541502550244331,
196
+ "rewards/margins": -0.0083759855479002,
197
+ "rewards/rejected": 0.01991748809814453,
198
+ "step": 9
199
+ },
200
+ {
201
+ "debug/policy_chosen_logits": -2.971123218536377,
202
+ "debug/policy_chosen_logps": -211.6798858642578,
203
+ "debug/policy_rejected_logits": -3.1354713439941406,
204
+ "debug/policy_rejected_logps": -175.7381134033203,
205
+ "debug/reference_chosen_logps": -216.06980895996094,
206
+ "debug/reference_rejected_logps": -179.71377563476562,
207
+ "epoch": 0.3225806451612903,
208
+ "grad_norm": 12.173478354163274,
209
+ "learning_rate": 1e-06,
210
+ "logits/chosen": -2.971123218536377,
211
+ "logits/rejected": -3.1354713439941406,
212
+ "logps/chosen": -211.6798858642578,
213
+ "logps/rejected": -175.7381134033203,
214
+ "loss": 0.4989,
215
+ "rewards/accuracies": 0.5,
216
+ "rewards/chosen": 0.04389926791191101,
217
+ "rewards/margins": 0.004142570309340954,
218
+ "rewards/rejected": 0.03975669667124748,
219
+ "step": 10
220
+ },
221
+ {
222
+ "debug/policy_chosen_logits": -3.1046411991119385,
223
+ "debug/policy_chosen_logps": -179.2141571044922,
224
+ "debug/policy_rejected_logits": -3.006396770477295,
225
+ "debug/policy_rejected_logps": -231.02593994140625,
226
+ "debug/reference_chosen_logps": -183.09750366210938,
227
+ "debug/reference_rejected_logps": -234.02491760253906,
228
+ "epoch": 0.3548387096774194,
229
+ "grad_norm": 19.69641784212433,
230
+ "learning_rate": 1e-06,
231
+ "logits/chosen": -3.1046411991119385,
232
+ "logits/rejected": -3.006396770477295,
233
+ "logps/chosen": -179.2141571044922,
234
+ "logps/rejected": -231.02593994140625,
235
+ "loss": 0.4927,
236
+ "rewards/accuracies": 0.625,
237
+ "rewards/chosen": 0.038833435624837875,
238
+ "rewards/margins": 0.008843660354614258,
239
+ "rewards/rejected": 0.029989775270223618,
240
+ "step": 11
241
+ },
242
+ {
243
+ "debug/policy_chosen_logits": -3.148892402648926,
244
+ "debug/policy_chosen_logps": -164.12550354003906,
245
+ "debug/policy_rejected_logits": -3.2069592475891113,
246
+ "debug/policy_rejected_logps": -196.435791015625,
247
+ "debug/reference_chosen_logps": -173.92369079589844,
248
+ "debug/reference_rejected_logps": -200.88906860351562,
249
+ "epoch": 0.3870967741935484,
250
+ "grad_norm": 39.94934023839017,
251
+ "learning_rate": 1e-06,
252
+ "logits/chosen": -3.148892402648926,
253
+ "logits/rejected": -3.2069592475891113,
254
+ "logps/chosen": -164.12550354003906,
255
+ "logps/rejected": -196.435791015625,
256
+ "loss": 0.5015,
257
+ "rewards/accuracies": 0.875,
258
+ "rewards/chosen": 0.09798187762498856,
259
+ "rewards/margins": 0.05344902724027634,
260
+ "rewards/rejected": 0.04453285038471222,
261
+ "step": 12
262
+ },
263
+ {
264
+ "debug/policy_chosen_logits": -3.147637367248535,
265
+ "debug/policy_chosen_logps": -142.30979919433594,
266
+ "debug/policy_rejected_logits": -3.1137590408325195,
267
+ "debug/policy_rejected_logps": -179.00892639160156,
268
+ "debug/reference_chosen_logps": -149.29544067382812,
269
+ "debug/reference_rejected_logps": -185.37518310546875,
270
+ "epoch": 0.41935483870967744,
271
+ "grad_norm": 65.71234154752543,
272
+ "learning_rate": 1e-06,
273
+ "logits/chosen": -3.147637367248535,
274
+ "logits/rejected": -3.1137590408325195,
275
+ "logps/chosen": -142.30979919433594,
276
+ "logps/rejected": -179.00892639160156,
277
+ "loss": 0.5062,
278
+ "rewards/accuracies": 0.625,
279
+ "rewards/chosen": 0.06985651701688766,
280
+ "rewards/margins": 0.006193867418915033,
281
+ "rewards/rejected": 0.06366264820098877,
282
+ "step": 13
283
+ },
284
+ {
285
+ "debug/policy_chosen_logits": -3.2451846599578857,
286
+ "debug/policy_chosen_logps": -183.0009765625,
287
+ "debug/policy_rejected_logits": -3.2852346897125244,
288
+ "debug/policy_rejected_logps": -197.824951171875,
289
+ "debug/reference_chosen_logps": -183.507568359375,
290
+ "debug/reference_rejected_logps": -201.4381103515625,
291
+ "epoch": 0.45161290322580644,
292
+ "grad_norm": 26.124571295930675,
293
+ "learning_rate": 1e-06,
294
+ "logits/chosen": -3.2451846599578857,
295
+ "logits/rejected": -3.2852346897125244,
296
+ "logps/chosen": -183.0009765625,
297
+ "logps/rejected": -197.824951171875,
298
+ "loss": 0.5122,
299
+ "rewards/accuracies": 0.375,
300
+ "rewards/chosen": 0.005065831355750561,
301
+ "rewards/margins": -0.031065817922353745,
302
+ "rewards/rejected": 0.03613165393471718,
303
+ "step": 14
304
+ },
305
+ {
306
+ "debug/policy_chosen_logits": -3.228227376937866,
307
+ "debug/policy_chosen_logps": -150.36203002929688,
308
+ "debug/policy_rejected_logits": -3.2135226726531982,
309
+ "debug/policy_rejected_logps": -236.50494384765625,
310
+ "debug/reference_chosen_logps": -158.17320251464844,
311
+ "debug/reference_rejected_logps": -239.26577758789062,
312
+ "epoch": 0.4838709677419355,
313
+ "grad_norm": 15.171711401478728,
314
+ "learning_rate": 1e-06,
315
+ "logits/chosen": -3.228227376937866,
316
+ "logits/rejected": -3.2135226726531982,
317
+ "logps/chosen": -150.36203002929688,
318
+ "logps/rejected": -236.50494384765625,
319
+ "loss": 0.5003,
320
+ "rewards/accuracies": 0.75,
321
+ "rewards/chosen": 0.07811184972524643,
322
+ "rewards/margins": 0.05050337687134743,
323
+ "rewards/rejected": 0.027608470991253853,
324
+ "step": 15
325
+ },
326
+ {
327
+ "debug/policy_chosen_logits": -3.164198398590088,
328
+ "debug/policy_chosen_logps": -178.2091064453125,
329
+ "debug/policy_rejected_logits": -3.2197299003601074,
330
+ "debug/policy_rejected_logps": -188.035888671875,
331
+ "debug/reference_chosen_logps": -180.6068115234375,
332
+ "debug/reference_rejected_logps": -190.4276123046875,
333
+ "epoch": 0.5161290322580645,
334
+ "grad_norm": 21.853695731678606,
335
+ "learning_rate": 1e-06,
336
+ "logits/chosen": -3.164198398590088,
337
+ "logits/rejected": -3.2197299003601074,
338
+ "logps/chosen": -178.2091064453125,
339
+ "logps/rejected": -188.035888671875,
340
+ "loss": 0.5007,
341
+ "rewards/accuracies": 0.375,
342
+ "rewards/chosen": 0.023976972326636314,
343
+ "rewards/margins": 5.9641897678375244e-05,
344
+ "rewards/rejected": 0.02391733229160309,
345
+ "step": 16
346
+ },
347
+ {
348
+ "debug/policy_chosen_logits": -3.1247196197509766,
349
+ "debug/policy_chosen_logps": -138.4497528076172,
350
+ "debug/policy_rejected_logits": -3.148193359375,
351
+ "debug/policy_rejected_logps": -177.52459716796875,
352
+ "debug/reference_chosen_logps": -141.9833526611328,
353
+ "debug/reference_rejected_logps": -176.2555694580078,
354
+ "epoch": 0.5483870967741935,
355
+ "grad_norm": 16.184529966630762,
356
+ "learning_rate": 1e-06,
357
+ "logits/chosen": -3.1247196197509766,
358
+ "logits/rejected": -3.148193359375,
359
+ "logps/chosen": -138.4497528076172,
360
+ "logps/rejected": -177.52459716796875,
361
+ "loss": 0.5052,
362
+ "rewards/accuracies": 0.5,
363
+ "rewards/chosen": 0.035335883498191833,
364
+ "rewards/margins": 0.04802631214261055,
365
+ "rewards/rejected": -0.012690428644418716,
366
+ "step": 17
367
+ },
368
+ {
369
+ "debug/policy_chosen_logits": -3.064986228942871,
370
+ "debug/policy_chosen_logps": -200.385009765625,
371
+ "debug/policy_rejected_logits": -3.0621964931488037,
372
+ "debug/policy_rejected_logps": -207.35183715820312,
373
+ "debug/reference_chosen_logps": -202.05747985839844,
374
+ "debug/reference_rejected_logps": -206.5642547607422,
375
+ "epoch": 0.5806451612903226,
376
+ "grad_norm": 12.667063930754026,
377
+ "learning_rate": 1e-06,
378
+ "logits/chosen": -3.064986228942871,
379
+ "logits/rejected": -3.0621964931488037,
380
+ "logps/chosen": -200.385009765625,
381
+ "logps/rejected": -207.35183715820312,
382
+ "loss": 0.4804,
383
+ "rewards/accuracies": 0.75,
384
+ "rewards/chosen": 0.016724586486816406,
385
+ "rewards/margins": 0.024600541219115257,
386
+ "rewards/rejected": -0.007875956594944,
387
+ "step": 18
388
+ },
389
+ {
390
+ "debug/policy_chosen_logits": -3.2486557960510254,
391
+ "debug/policy_chosen_logps": -152.92401123046875,
392
+ "debug/policy_rejected_logits": -3.239227771759033,
393
+ "debug/policy_rejected_logps": -164.85879516601562,
394
+ "debug/reference_chosen_logps": -152.27206420898438,
395
+ "debug/reference_rejected_logps": -161.17071533203125,
396
+ "epoch": 0.6129032258064516,
397
+ "grad_norm": 17.915281455929325,
398
+ "learning_rate": 1e-06,
399
+ "logits/chosen": -3.2486557960510254,
400
+ "logits/rejected": -3.239227771759033,
401
+ "logps/chosen": -152.92401123046875,
402
+ "logps/rejected": -164.85879516601562,
403
+ "loss": 0.4612,
404
+ "rewards/accuracies": 0.875,
405
+ "rewards/chosen": -0.00651948805898428,
406
+ "rewards/margins": 0.03036128729581833,
407
+ "rewards/rejected": -0.03688077628612518,
408
+ "step": 19
409
+ },
410
+ {
411
+ "debug/policy_chosen_logits": -3.223217010498047,
412
+ "debug/policy_chosen_logps": -216.85635375976562,
413
+ "debug/policy_rejected_logits": -3.2773444652557373,
414
+ "debug/policy_rejected_logps": -170.92140197753906,
415
+ "debug/reference_chosen_logps": -215.15817260742188,
416
+ "debug/reference_rejected_logps": -165.83981323242188,
417
+ "epoch": 0.6451612903225806,
418
+ "grad_norm": 27.794458161611285,
419
+ "learning_rate": 1e-06,
420
+ "logits/chosen": -3.223217010498047,
421
+ "logits/rejected": -3.2773444652557373,
422
+ "logps/chosen": -216.85635375976562,
423
+ "logps/rejected": -170.92140197753906,
424
+ "loss": 0.4863,
425
+ "rewards/accuracies": 0.375,
426
+ "rewards/chosen": -0.01698194444179535,
427
+ "rewards/margins": 0.03383391723036766,
428
+ "rewards/rejected": -0.05081585794687271,
429
+ "step": 20
430
+ },
431
+ {
432
+ "debug/policy_chosen_logits": -3.0951452255249023,
433
+ "debug/policy_chosen_logps": -184.65277099609375,
434
+ "debug/policy_rejected_logits": -3.077794313430786,
435
+ "debug/policy_rejected_logps": -177.56741333007812,
436
+ "debug/reference_chosen_logps": -183.5254669189453,
437
+ "debug/reference_rejected_logps": -175.02011108398438,
438
+ "epoch": 0.6774193548387096,
439
+ "grad_norm": 20.0849838488403,
440
+ "learning_rate": 1e-06,
441
+ "logits/chosen": -3.0951452255249023,
442
+ "logits/rejected": -3.077794313430786,
443
+ "logps/chosen": -184.65277099609375,
444
+ "logps/rejected": -177.56741333007812,
445
+ "loss": 0.4889,
446
+ "rewards/accuracies": 0.625,
447
+ "rewards/chosen": -0.011273040436208248,
448
+ "rewards/margins": 0.014199809171259403,
449
+ "rewards/rejected": -0.02547284960746765,
450
+ "step": 21
451
+ },
452
+ {
453
+ "debug/policy_chosen_logits": -3.1871633529663086,
454
+ "debug/policy_chosen_logps": -147.81625366210938,
455
+ "debug/policy_rejected_logits": -3.2192680835723877,
456
+ "debug/policy_rejected_logps": -198.98519897460938,
457
+ "debug/reference_chosen_logps": -145.34873962402344,
458
+ "debug/reference_rejected_logps": -196.41769409179688,
459
+ "epoch": 0.7096774193548387,
460
+ "grad_norm": 22.495490487928247,
461
+ "learning_rate": 1e-06,
462
+ "logits/chosen": -3.1871633529663086,
463
+ "logits/rejected": -3.2192680835723877,
464
+ "logps/chosen": -147.81625366210938,
465
+ "logps/rejected": -198.98519897460938,
466
+ "loss": 0.4847,
467
+ "rewards/accuracies": 0.625,
468
+ "rewards/chosen": -0.024675264954566956,
469
+ "rewards/margins": 0.0009998120367527008,
470
+ "rewards/rejected": -0.025675076991319656,
471
+ "step": 22
472
+ },
473
+ {
474
+ "debug/policy_chosen_logits": -3.077518939971924,
475
+ "debug/policy_chosen_logps": -193.34442138671875,
476
+ "debug/policy_rejected_logits": -3.218148708343506,
477
+ "debug/policy_rejected_logps": -198.03993225097656,
478
+ "debug/reference_chosen_logps": -191.85537719726562,
479
+ "debug/reference_rejected_logps": -191.48312377929688,
480
+ "epoch": 0.7419354838709677,
481
+ "grad_norm": 25.201442820556927,
482
+ "learning_rate": 1e-06,
483
+ "logits/chosen": -3.077518939971924,
484
+ "logits/rejected": -3.218148708343506,
485
+ "logps/chosen": -193.34442138671875,
486
+ "logps/rejected": -198.03993225097656,
487
+ "loss": 0.4618,
488
+ "rewards/accuracies": 0.75,
489
+ "rewards/chosen": -0.014890595339238644,
490
+ "rewards/margins": 0.05067773163318634,
491
+ "rewards/rejected": -0.06556832045316696,
492
+ "step": 23
493
+ },
494
+ {
495
+ "debug/policy_chosen_logits": -3.2286179065704346,
496
+ "debug/policy_chosen_logps": -161.75,
497
+ "debug/policy_rejected_logits": -3.1389119625091553,
498
+ "debug/policy_rejected_logps": -189.98956298828125,
499
+ "debug/reference_chosen_logps": -161.94760131835938,
500
+ "debug/reference_rejected_logps": -190.900634765625,
501
+ "epoch": 0.7741935483870968,
502
+ "grad_norm": 12.967586243161568,
503
+ "learning_rate": 1e-06,
504
+ "logits/chosen": -3.2286179065704346,
505
+ "logits/rejected": -3.1389119625091553,
506
+ "logps/chosen": -161.75,
507
+ "logps/rejected": -189.98956298828125,
508
+ "loss": 0.4749,
509
+ "rewards/accuracies": 0.375,
510
+ "rewards/chosen": 0.0019761761650443077,
511
+ "rewards/margins": -0.007134380750358105,
512
+ "rewards/rejected": 0.009110554121434689,
513
+ "step": 24
514
+ },
515
+ {
516
+ "debug/policy_chosen_logits": -3.272780418395996,
517
+ "debug/policy_chosen_logps": -159.73602294921875,
518
+ "debug/policy_rejected_logits": -3.2736470699310303,
519
+ "debug/policy_rejected_logps": -184.5146484375,
520
+ "debug/reference_chosen_logps": -164.45745849609375,
521
+ "debug/reference_rejected_logps": -187.22787475585938,
522
+ "epoch": 0.8064516129032258,
523
+ "grad_norm": 30.149125998144832,
524
+ "learning_rate": 1e-06,
525
+ "logits/chosen": -3.272780418395996,
526
+ "logits/rejected": -3.2736470699310303,
527
+ "logps/chosen": -159.73602294921875,
528
+ "logps/rejected": -184.5146484375,
529
+ "loss": 0.4964,
530
+ "rewards/accuracies": 0.5,
531
+ "rewards/chosen": 0.0472143180668354,
532
+ "rewards/margins": 0.02008194290101528,
533
+ "rewards/rejected": 0.027132373303174973,
534
+ "step": 25
535
+ },
536
+ {
537
+ "debug/policy_chosen_logits": -3.0852928161621094,
538
+ "debug/policy_chosen_logps": -206.41693115234375,
539
+ "debug/policy_rejected_logits": -3.1496071815490723,
540
+ "debug/policy_rejected_logps": -199.11636352539062,
541
+ "debug/reference_chosen_logps": -207.51861572265625,
542
+ "debug/reference_rejected_logps": -200.44711303710938,
543
+ "epoch": 0.8387096774193549,
544
+ "grad_norm": 53.54284328779407,
545
+ "learning_rate": 1e-06,
546
+ "logits/chosen": -3.0852928161621094,
547
+ "logits/rejected": -3.1496071815490723,
548
+ "logps/chosen": -206.41693115234375,
549
+ "logps/rejected": -199.11636352539062,
550
+ "loss": 0.5242,
551
+ "rewards/accuracies": 0.5,
552
+ "rewards/chosen": 0.01101672276854515,
553
+ "rewards/margins": -0.0022906875237822533,
554
+ "rewards/rejected": 0.013307408429682255,
555
+ "step": 26
556
+ },
557
+ {
558
+ "debug/policy_chosen_logits": -3.176145315170288,
559
+ "debug/policy_chosen_logps": -136.85189819335938,
560
+ "debug/policy_rejected_logits": -3.165332555770874,
561
+ "debug/policy_rejected_logps": -172.10556030273438,
562
+ "debug/reference_chosen_logps": -137.76473999023438,
563
+ "debug/reference_rejected_logps": -177.917236328125,
564
+ "epoch": 0.8709677419354839,
565
+ "grad_norm": 48.38261399039204,
566
+ "learning_rate": 1e-06,
567
+ "logits/chosen": -3.176145315170288,
568
+ "logits/rejected": -3.165332555770874,
569
+ "logps/chosen": -136.85189819335938,
570
+ "logps/rejected": -172.10556030273438,
571
+ "loss": 0.5204,
572
+ "rewards/accuracies": 0.375,
573
+ "rewards/chosen": 0.009128322824835777,
574
+ "rewards/margins": -0.048988424241542816,
575
+ "rewards/rejected": 0.05811674892902374,
576
+ "step": 27
577
+ },
578
+ {
579
+ "debug/policy_chosen_logits": -3.070044755935669,
580
+ "debug/policy_chosen_logps": -188.4156036376953,
581
+ "debug/policy_rejected_logits": -3.1209747791290283,
582
+ "debug/policy_rejected_logps": -268.2892150878906,
583
+ "debug/reference_chosen_logps": -187.7012939453125,
584
+ "debug/reference_rejected_logps": -255.53848266601562,
585
+ "epoch": 0.9032258064516129,
586
+ "grad_norm": 12.887123147417002,
587
+ "learning_rate": 1e-06,
588
+ "logits/chosen": -3.070044755935669,
589
+ "logits/rejected": -3.1209747791290283,
590
+ "logps/chosen": -188.4156036376953,
591
+ "logps/rejected": -268.2892150878906,
592
+ "loss": 0.4722,
593
+ "rewards/accuracies": 1.0,
594
+ "rewards/chosen": -0.007143107242882252,
595
+ "rewards/margins": 0.12036430835723877,
596
+ "rewards/rejected": -0.12750741839408875,
597
+ "step": 28
598
+ },
599
+ {
600
+ "debug/policy_chosen_logits": -3.107494831085205,
601
+ "debug/policy_chosen_logps": -160.94583129882812,
602
+ "debug/policy_rejected_logits": -3.0909152030944824,
603
+ "debug/policy_rejected_logps": -195.18310546875,
604
+ "debug/reference_chosen_logps": -164.90542602539062,
605
+ "debug/reference_rejected_logps": -192.39305114746094,
606
+ "epoch": 0.9354838709677419,
607
+ "grad_norm": 43.32899603798361,
608
+ "learning_rate": 1e-06,
609
+ "logits/chosen": -3.107494831085205,
610
+ "logits/rejected": -3.0909152030944824,
611
+ "logps/chosen": -160.94583129882812,
612
+ "logps/rejected": -195.18310546875,
613
+ "loss": 0.4825,
614
+ "rewards/accuracies": 0.625,
615
+ "rewards/chosen": 0.03959597647190094,
616
+ "rewards/margins": 0.06749667972326279,
617
+ "rewards/rejected": -0.027900703251361847,
618
+ "step": 29
619
+ },
620
+ {
621
+ "debug/policy_chosen_logits": -3.106745958328247,
622
+ "debug/policy_chosen_logps": -197.70811462402344,
623
+ "debug/policy_rejected_logits": -3.103618860244751,
624
+ "debug/policy_rejected_logps": -207.11260986328125,
625
+ "debug/reference_chosen_logps": -194.18881225585938,
626
+ "debug/reference_rejected_logps": -205.00173950195312,
627
+ "epoch": 0.967741935483871,
628
+ "grad_norm": 9.647676539368454,
629
+ "learning_rate": 1e-06,
630
+ "logits/chosen": -3.106745958328247,
631
+ "logits/rejected": -3.103618860244751,
632
+ "logps/chosen": -197.70811462402344,
633
+ "logps/rejected": -207.11260986328125,
634
+ "loss": 0.4565,
635
+ "rewards/accuracies": 0.5,
636
+ "rewards/chosen": -0.035193100571632385,
637
+ "rewards/margins": -0.01408432051539421,
638
+ "rewards/rejected": -0.021108778193593025,
639
+ "step": 30
640
+ },
641
+ {
642
+ "debug/policy_chosen_logits": -3.026829957962036,
643
+ "debug/policy_chosen_logps": -198.42453002929688,
644
+ "debug/policy_rejected_logits": -2.9937751293182373,
645
+ "debug/policy_rejected_logps": -243.72132873535156,
646
+ "debug/reference_chosen_logps": -197.02398681640625,
647
+ "debug/reference_rejected_logps": -237.82403564453125,
648
+ "epoch": 1.0,
649
+ "grad_norm": 10.307557167599814,
650
+ "learning_rate": 1e-06,
651
+ "logits/chosen": -3.026829957962036,
652
+ "logits/rejected": -2.9937751293182373,
653
+ "logps/chosen": -198.42453002929688,
654
+ "logps/rejected": -243.72132873535156,
655
+ "loss": 0.461,
656
+ "rewards/accuracies": 0.625,
657
+ "rewards/chosen": -0.014005661010742188,
658
+ "rewards/margins": 0.04496738314628601,
659
+ "rewards/rejected": -0.0589730404317379,
660
+ "step": 31
661
+ },
662
+ {
663
+ "epoch": 1.0,
664
+ "step": 31,
665
+ "total_flos": 0.0,
666
+ "train_loss": 0.4911675299367597,
667
+ "train_runtime": 120.215,
668
+ "train_samples_per_second": 16.371,
669
+ "train_steps_per_second": 0.258
670
+ }
671
+ ],
672
+ "logging_steps": 1,
673
+ "max_steps": 31,
674
+ "num_input_tokens_seen": 0,
675
+ "num_train_epochs": 1,
676
+ "save_steps": 500,
677
+ "stateful_callbacks": {
678
+ "TrainerControl": {
679
+ "args": {
680
+ "should_epoch_stop": false,
681
+ "should_evaluate": false,
682
+ "should_log": false,
683
+ "should_save": true,
684
+ "should_training_stop": true
685
+ },
686
+ "attributes": {}
687
+ }
688
+ },
689
+ "total_flos": 0.0,
690
+ "train_batch_size": 8,
691
+ "trial_name": null,
692
+ "trial_params": null
693
+ }