yiran-wang3 commited on
Commit
f9bf5c1
1 Parent(s): 20ad6e8

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_coder7b_reflct_adamw_iter1
5
+ tags:
6
+ - alignment-handbook
7
+ - generated_from_trainer
8
+ - trl
9
+ - dpo
10
+ datasets:
11
+ - self-generate/qw2coder_reflct7b_sppo_hard_new_cn_mining_oj_iter1-binarized-reflection-scored
12
+ model-index:
13
+ - name: qwen2_coder7b_reflct_adamw_iter2
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_coder7b_reflct_adamw_iter2
21
+
22
+ This model is a fine-tuned version of [yiran-wang3/qwen2_coder7b_reflct_adamw_iter1](https://huggingface.co/yiran-wang3/qwen2_coder7b_reflct_adamw_iter1) on the self-generate/qw2coder_reflct7b_sppo_hard_new_cn_mining_oj_iter1-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 0.20.3
all_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "total_flos": 0.0,
4
+ "train_loss": 0.48355864485104877,
5
+ "train_runtime": 407.656,
6
+ "train_samples": 2070,
7
+ "train_samples_per_second": 5.078,
8
+ "train_steps_per_second": 0.081
9
+ }
config.json CHANGED
@@ -23,7 +23,7 @@
23
  "tie_word_embeddings": false,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.45.0",
26
- "use_cache": false,
27
  "use_sliding_window": false,
28
  "vocab_size": 152064
29
  }
 
23
  "tie_word_embeddings": false,
24
  "torch_dtype": "bfloat16",
25
  "transformers_version": "4.45.0",
26
+ "use_cache": true,
27
  "use_sliding_window": false,
28
  "vocab_size": 152064
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": 0.48355864485104877,
5
+ "train_runtime": 407.656,
6
+ "train_samples": 2070,
7
+ "train_samples_per_second": 5.078,
8
+ "train_steps_per_second": 0.081
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": -2.4703779220581055,
13
+ "debug/policy_chosen_logps": -179.5443115234375,
14
+ "debug/policy_rejected_logits": -2.4724299907684326,
15
+ "debug/policy_rejected_logps": -182.86181640625,
16
+ "debug/reference_chosen_logps": -179.5443115234375,
17
+ "debug/reference_rejected_logps": -182.86181640625,
18
+ "epoch": 0.030303030303030304,
19
+ "grad_norm": 2.750340847346066,
20
+ "learning_rate": 1e-06,
21
+ "logits/chosen": -2.4703779220581055,
22
+ "logits/rejected": -2.4724299907684326,
23
+ "logps/chosen": -179.5443115234375,
24
+ "logps/rejected": -182.86181640625,
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": -2.612560987472534,
34
+ "debug/policy_chosen_logps": -187.98855590820312,
35
+ "debug/policy_rejected_logits": -2.5251338481903076,
36
+ "debug/policy_rejected_logps": -192.80223083496094,
37
+ "debug/reference_chosen_logps": -187.92822265625,
38
+ "debug/reference_rejected_logps": -192.76466369628906,
39
+ "epoch": 0.06060606060606061,
40
+ "grad_norm": 3.2222630610295133,
41
+ "learning_rate": 1e-06,
42
+ "logits/chosen": -2.612560987472534,
43
+ "logits/rejected": -2.5251338481903076,
44
+ "logps/chosen": -187.98855590820312,
45
+ "logps/rejected": -192.80223083496094,
46
+ "loss": 0.5,
47
+ "rewards/accuracies": 0.625,
48
+ "rewards/chosen": -0.0006031226948834956,
49
+ "rewards/margins": -0.00022733688820153475,
50
+ "rewards/rejected": -0.0003757858066819608,
51
+ "step": 2
52
+ },
53
+ {
54
+ "debug/policy_chosen_logits": -2.5202717781066895,
55
+ "debug/policy_chosen_logps": -172.94644165039062,
56
+ "debug/policy_rejected_logits": -2.446408748626709,
57
+ "debug/policy_rejected_logps": -182.64926147460938,
58
+ "debug/reference_chosen_logps": -172.74356079101562,
59
+ "debug/reference_rejected_logps": -182.5146484375,
60
+ "epoch": 0.09090909090909091,
61
+ "grad_norm": 3.296231138476821,
62
+ "learning_rate": 1e-06,
63
+ "logits/chosen": -2.5202717781066895,
64
+ "logits/rejected": -2.446408748626709,
65
+ "logps/chosen": -172.94644165039062,
66
+ "logps/rejected": -182.64926147460938,
67
+ "loss": 0.4995,
68
+ "rewards/accuracies": 0.5,
69
+ "rewards/chosen": -0.002028884831815958,
70
+ "rewards/margins": -0.0006827162578701973,
71
+ "rewards/rejected": -0.001346168457530439,
72
+ "step": 3
73
+ },
74
+ {
75
+ "debug/policy_chosen_logits": -2.5300447940826416,
76
+ "debug/policy_chosen_logps": -180.87469482421875,
77
+ "debug/policy_rejected_logits": -2.5812830924987793,
78
+ "debug/policy_rejected_logps": -196.6584014892578,
79
+ "debug/reference_chosen_logps": -180.65478515625,
80
+ "debug/reference_rejected_logps": -196.28567504882812,
81
+ "epoch": 0.12121212121212122,
82
+ "grad_norm": 2.654852833868078,
83
+ "learning_rate": 1e-06,
84
+ "logits/chosen": -2.5300447940826416,
85
+ "logits/rejected": -2.5812830924987793,
86
+ "logps/chosen": -180.87469482421875,
87
+ "logps/rejected": -196.6584014892578,
88
+ "loss": 0.4987,
89
+ "rewards/accuracies": 0.625,
90
+ "rewards/chosen": -0.0021991347894072533,
91
+ "rewards/margins": 0.0015281867235898972,
92
+ "rewards/rejected": -0.0037273215129971504,
93
+ "step": 4
94
+ },
95
+ {
96
+ "debug/policy_chosen_logits": -2.443774700164795,
97
+ "debug/policy_chosen_logps": -189.18003845214844,
98
+ "debug/policy_rejected_logits": -2.4312195777893066,
99
+ "debug/policy_rejected_logps": -172.596923828125,
100
+ "debug/reference_chosen_logps": -188.87039184570312,
101
+ "debug/reference_rejected_logps": -172.26124572753906,
102
+ "epoch": 0.15151515151515152,
103
+ "grad_norm": 2.765216242619178,
104
+ "learning_rate": 1e-06,
105
+ "logits/chosen": -2.443774700164795,
106
+ "logits/rejected": -2.4312195777893066,
107
+ "logps/chosen": -189.18003845214844,
108
+ "logps/rejected": -172.596923828125,
109
+ "loss": 0.4978,
110
+ "rewards/accuracies": 0.375,
111
+ "rewards/chosen": -0.0030963707249611616,
112
+ "rewards/margins": 0.00026035308837890625,
113
+ "rewards/rejected": -0.003356723580509424,
114
+ "step": 5
115
+ },
116
+ {
117
+ "debug/policy_chosen_logits": -2.5867741107940674,
118
+ "debug/policy_chosen_logps": -191.50296020507812,
119
+ "debug/policy_rejected_logits": -2.5049338340759277,
120
+ "debug/policy_rejected_logps": -172.64727783203125,
121
+ "debug/reference_chosen_logps": -191.35916137695312,
122
+ "debug/reference_rejected_logps": -171.7898406982422,
123
+ "epoch": 0.18181818181818182,
124
+ "grad_norm": 2.98493127681766,
125
+ "learning_rate": 1e-06,
126
+ "logits/chosen": -2.5867741107940674,
127
+ "logits/rejected": -2.5049338340759277,
128
+ "logps/chosen": -191.50296020507812,
129
+ "logps/rejected": -172.64727783203125,
130
+ "loss": 0.4949,
131
+ "rewards/accuracies": 0.75,
132
+ "rewards/chosen": -0.0014381217770278454,
133
+ "rewards/margins": 0.007136192638427019,
134
+ "rewards/rejected": -0.008574314415454865,
135
+ "step": 6
136
+ },
137
+ {
138
+ "debug/policy_chosen_logits": -2.4890458583831787,
139
+ "debug/policy_chosen_logps": -188.74664306640625,
140
+ "debug/policy_rejected_logits": -2.503746747970581,
141
+ "debug/policy_rejected_logps": -182.01272583007812,
142
+ "debug/reference_chosen_logps": -188.08114624023438,
143
+ "debug/reference_rejected_logps": -181.43170166015625,
144
+ "epoch": 0.21212121212121213,
145
+ "grad_norm": 2.565299522305327,
146
+ "learning_rate": 1e-06,
147
+ "logits/chosen": -2.4890458583831787,
148
+ "logits/rejected": -2.503746747970581,
149
+ "logps/chosen": -188.74664306640625,
150
+ "logps/rejected": -182.01272583007812,
151
+ "loss": 0.4962,
152
+ "rewards/accuracies": 0.375,
153
+ "rewards/chosen": -0.006655101664364338,
154
+ "rewards/margins": -0.0008448789012618363,
155
+ "rewards/rejected": -0.005810222588479519,
156
+ "step": 7
157
+ },
158
+ {
159
+ "debug/policy_chosen_logits": -2.5022552013397217,
160
+ "debug/policy_chosen_logps": -172.44332885742188,
161
+ "debug/policy_rejected_logits": -2.38063907623291,
162
+ "debug/policy_rejected_logps": -179.5426025390625,
163
+ "debug/reference_chosen_logps": -171.3497772216797,
164
+ "debug/reference_rejected_logps": -178.29132080078125,
165
+ "epoch": 0.24242424242424243,
166
+ "grad_norm": 2.683548775109707,
167
+ "learning_rate": 1e-06,
168
+ "logits/chosen": -2.5022552013397217,
169
+ "logits/rejected": -2.38063907623291,
170
+ "logps/chosen": -172.44332885742188,
171
+ "logps/rejected": -179.5426025390625,
172
+ "loss": 0.4946,
173
+ "rewards/accuracies": 0.75,
174
+ "rewards/chosen": -0.010935591533780098,
175
+ "rewards/margins": 0.001577300950884819,
176
+ "rewards/rejected": -0.012512893415987492,
177
+ "step": 8
178
+ },
179
+ {
180
+ "debug/policy_chosen_logits": -2.495532989501953,
181
+ "debug/policy_chosen_logps": -188.021728515625,
182
+ "debug/policy_rejected_logits": -2.334559202194214,
183
+ "debug/policy_rejected_logps": -184.0438690185547,
184
+ "debug/reference_chosen_logps": -187.92709350585938,
185
+ "debug/reference_rejected_logps": -183.24673461914062,
186
+ "epoch": 0.2727272727272727,
187
+ "grad_norm": 2.872420779341629,
188
+ "learning_rate": 1e-06,
189
+ "logits/chosen": -2.495532989501953,
190
+ "logits/rejected": -2.334559202194214,
191
+ "logps/chosen": -188.021728515625,
192
+ "logps/rejected": -184.0438690185547,
193
+ "loss": 0.4937,
194
+ "rewards/accuracies": 0.625,
195
+ "rewards/chosen": -0.0009464073227718472,
196
+ "rewards/margins": 0.007024994120001793,
197
+ "rewards/rejected": -0.007971401326358318,
198
+ "step": 9
199
+ },
200
+ {
201
+ "debug/policy_chosen_logits": -2.4814565181732178,
202
+ "debug/policy_chosen_logps": -198.445556640625,
203
+ "debug/policy_rejected_logits": -2.496356248855591,
204
+ "debug/policy_rejected_logps": -177.85418701171875,
205
+ "debug/reference_chosen_logps": -198.29083251953125,
206
+ "debug/reference_rejected_logps": -175.91622924804688,
207
+ "epoch": 0.30303030303030304,
208
+ "grad_norm": 3.0807114531930426,
209
+ "learning_rate": 1e-06,
210
+ "logits/chosen": -2.4814565181732178,
211
+ "logits/rejected": -2.496356248855591,
212
+ "logps/chosen": -198.445556640625,
213
+ "logps/rejected": -177.85418701171875,
214
+ "loss": 0.4914,
215
+ "rewards/accuracies": 0.625,
216
+ "rewards/chosen": -0.0015472793020308018,
217
+ "rewards/margins": 0.01783212646842003,
218
+ "rewards/rejected": -0.019379405304789543,
219
+ "step": 10
220
+ },
221
+ {
222
+ "debug/policy_chosen_logits": -2.5934786796569824,
223
+ "debug/policy_chosen_logps": -204.08248901367188,
224
+ "debug/policy_rejected_logits": -2.4578843116760254,
225
+ "debug/policy_rejected_logps": -186.8817138671875,
226
+ "debug/reference_chosen_logps": -203.44688415527344,
227
+ "debug/reference_rejected_logps": -187.01136779785156,
228
+ "epoch": 0.3333333333333333,
229
+ "grad_norm": 2.6552171041535537,
230
+ "learning_rate": 1e-06,
231
+ "logits/chosen": -2.5934786796569824,
232
+ "logits/rejected": -2.4578843116760254,
233
+ "logps/chosen": -204.08248901367188,
234
+ "logps/rejected": -186.8817138671875,
235
+ "loss": 0.4876,
236
+ "rewards/accuracies": 0.5,
237
+ "rewards/chosen": -0.006356086581945419,
238
+ "rewards/margins": -0.0076524922624230385,
239
+ "rewards/rejected": 0.001296405796892941,
240
+ "step": 11
241
+ },
242
+ {
243
+ "debug/policy_chosen_logits": -2.5135629177093506,
244
+ "debug/policy_chosen_logps": -177.83209228515625,
245
+ "debug/policy_rejected_logits": -2.414395332336426,
246
+ "debug/policy_rejected_logps": -174.76602172851562,
247
+ "debug/reference_chosen_logps": -179.32247924804688,
248
+ "debug/reference_rejected_logps": -171.77493286132812,
249
+ "epoch": 0.36363636363636365,
250
+ "grad_norm": 2.61493183338989,
251
+ "learning_rate": 1e-06,
252
+ "logits/chosen": -2.5135629177093506,
253
+ "logits/rejected": -2.414395332336426,
254
+ "logps/chosen": -177.83209228515625,
255
+ "logps/rejected": -174.76602172851562,
256
+ "loss": 0.4916,
257
+ "rewards/accuracies": 0.875,
258
+ "rewards/chosen": 0.014903735369443893,
259
+ "rewards/margins": 0.04481462389230728,
260
+ "rewards/rejected": -0.029910888522863388,
261
+ "step": 12
262
+ },
263
+ {
264
+ "debug/policy_chosen_logits": -2.533503532409668,
265
+ "debug/policy_chosen_logps": -183.77186584472656,
266
+ "debug/policy_rejected_logits": -2.3716490268707275,
267
+ "debug/policy_rejected_logps": -182.7346954345703,
268
+ "debug/reference_chosen_logps": -186.3679656982422,
269
+ "debug/reference_rejected_logps": -182.13394165039062,
270
+ "epoch": 0.3939393939393939,
271
+ "grad_norm": 2.664094734451926,
272
+ "learning_rate": 1e-06,
273
+ "logits/chosen": -2.533503532409668,
274
+ "logits/rejected": -2.3716490268707275,
275
+ "logps/chosen": -183.77186584472656,
276
+ "logps/rejected": -182.7346954345703,
277
+ "loss": 0.4855,
278
+ "rewards/accuracies": 0.75,
279
+ "rewards/chosen": 0.025960901752114296,
280
+ "rewards/margins": 0.03196847438812256,
281
+ "rewards/rejected": -0.006007576361298561,
282
+ "step": 13
283
+ },
284
+ {
285
+ "debug/policy_chosen_logits": -2.594430923461914,
286
+ "debug/policy_chosen_logps": -180.72747802734375,
287
+ "debug/policy_rejected_logits": -2.5284759998321533,
288
+ "debug/policy_rejected_logps": -187.50599670410156,
289
+ "debug/reference_chosen_logps": -182.18939208984375,
290
+ "debug/reference_rejected_logps": -188.8681182861328,
291
+ "epoch": 0.42424242424242425,
292
+ "grad_norm": 2.5123607464425324,
293
+ "learning_rate": 1e-06,
294
+ "logits/chosen": -2.594430923461914,
295
+ "logits/rejected": -2.5284759998321533,
296
+ "logps/chosen": -180.72747802734375,
297
+ "logps/rejected": -187.50599670410156,
298
+ "loss": 0.4808,
299
+ "rewards/accuracies": 0.75,
300
+ "rewards/chosen": 0.014619007706642151,
301
+ "rewards/margins": 0.0009978096932172775,
302
+ "rewards/rejected": 0.013621198013424873,
303
+ "step": 14
304
+ },
305
+ {
306
+ "debug/policy_chosen_logits": -2.5420515537261963,
307
+ "debug/policy_chosen_logps": -177.52317810058594,
308
+ "debug/policy_rejected_logits": -2.487260580062866,
309
+ "debug/policy_rejected_logps": -197.84640502929688,
310
+ "debug/reference_chosen_logps": -178.9560089111328,
311
+ "debug/reference_rejected_logps": -197.4163818359375,
312
+ "epoch": 0.45454545454545453,
313
+ "grad_norm": 2.680073567478649,
314
+ "learning_rate": 1e-06,
315
+ "logits/chosen": -2.5420515537261963,
316
+ "logits/rejected": -2.487260580062866,
317
+ "logps/chosen": -177.52317810058594,
318
+ "logps/rejected": -197.84640502929688,
319
+ "loss": 0.4794,
320
+ "rewards/accuracies": 0.75,
321
+ "rewards/chosen": 0.014328251592814922,
322
+ "rewards/margins": 0.018628500401973724,
323
+ "rewards/rejected": -0.004300250671803951,
324
+ "step": 15
325
+ },
326
+ {
327
+ "debug/policy_chosen_logits": -2.5443243980407715,
328
+ "debug/policy_chosen_logps": -175.14971923828125,
329
+ "debug/policy_rejected_logits": -2.615178108215332,
330
+ "debug/policy_rejected_logps": -195.92684936523438,
331
+ "debug/reference_chosen_logps": -177.9427490234375,
332
+ "debug/reference_rejected_logps": -192.58816528320312,
333
+ "epoch": 0.48484848484848486,
334
+ "grad_norm": 2.7557266784436454,
335
+ "learning_rate": 1e-06,
336
+ "logits/chosen": -2.5443243980407715,
337
+ "logits/rejected": -2.615178108215332,
338
+ "logps/chosen": -175.14971923828125,
339
+ "logps/rejected": -195.92684936523438,
340
+ "loss": 0.4813,
341
+ "rewards/accuracies": 0.875,
342
+ "rewards/chosen": 0.027930183336138725,
343
+ "rewards/margins": 0.061316829174757004,
344
+ "rewards/rejected": -0.03338664770126343,
345
+ "step": 16
346
+ },
347
+ {
348
+ "debug/policy_chosen_logits": -2.4633026123046875,
349
+ "debug/policy_chosen_logps": -189.36892700195312,
350
+ "debug/policy_rejected_logits": -2.456451654434204,
351
+ "debug/policy_rejected_logps": -171.43634033203125,
352
+ "debug/reference_chosen_logps": -192.05772399902344,
353
+ "debug/reference_rejected_logps": -170.10711669921875,
354
+ "epoch": 0.5151515151515151,
355
+ "grad_norm": 2.7126032387088785,
356
+ "learning_rate": 1e-06,
357
+ "logits/chosen": -2.4633026123046875,
358
+ "logits/rejected": -2.456451654434204,
359
+ "logps/chosen": -189.36892700195312,
360
+ "logps/rejected": -171.43634033203125,
361
+ "loss": 0.5013,
362
+ "rewards/accuracies": 0.75,
363
+ "rewards/chosen": 0.02688794955611229,
364
+ "rewards/margins": 0.04018016904592514,
365
+ "rewards/rejected": -0.013292218558490276,
366
+ "step": 17
367
+ },
368
+ {
369
+ "debug/policy_chosen_logits": -2.542309284210205,
370
+ "debug/policy_chosen_logps": -178.5828857421875,
371
+ "debug/policy_rejected_logits": -2.5415501594543457,
372
+ "debug/policy_rejected_logps": -202.34744262695312,
373
+ "debug/reference_chosen_logps": -180.57891845703125,
374
+ "debug/reference_rejected_logps": -201.40545654296875,
375
+ "epoch": 0.5454545454545454,
376
+ "grad_norm": 2.604213651718002,
377
+ "learning_rate": 1e-06,
378
+ "logits/chosen": -2.542309284210205,
379
+ "logits/rejected": -2.5415501594543457,
380
+ "logps/chosen": -178.5828857421875,
381
+ "logps/rejected": -202.34744262695312,
382
+ "loss": 0.4857,
383
+ "rewards/accuracies": 0.5,
384
+ "rewards/chosen": 0.01996036432683468,
385
+ "rewards/margins": 0.0293801911175251,
386
+ "rewards/rejected": -0.009419821202754974,
387
+ "step": 18
388
+ },
389
+ {
390
+ "debug/policy_chosen_logits": -2.51920223236084,
391
+ "debug/policy_chosen_logps": -173.57725524902344,
392
+ "debug/policy_rejected_logits": -2.4803247451782227,
393
+ "debug/policy_rejected_logps": -176.664306640625,
394
+ "debug/reference_chosen_logps": -171.49952697753906,
395
+ "debug/reference_rejected_logps": -174.46640014648438,
396
+ "epoch": 0.5757575757575758,
397
+ "grad_norm": 3.001719578710194,
398
+ "learning_rate": 1e-06,
399
+ "logits/chosen": -2.51920223236084,
400
+ "logits/rejected": -2.4803247451782227,
401
+ "logps/chosen": -173.57725524902344,
402
+ "logps/rejected": -176.664306640625,
403
+ "loss": 0.4888,
404
+ "rewards/accuracies": 0.625,
405
+ "rewards/chosen": -0.020777244120836258,
406
+ "rewards/margins": 0.0012017078697681427,
407
+ "rewards/rejected": -0.02197895012795925,
408
+ "step": 19
409
+ },
410
+ {
411
+ "debug/policy_chosen_logits": -2.60622501373291,
412
+ "debug/policy_chosen_logps": -195.92251586914062,
413
+ "debug/policy_rejected_logits": -2.4916000366210938,
414
+ "debug/policy_rejected_logps": -182.05645751953125,
415
+ "debug/reference_chosen_logps": -196.1642608642578,
416
+ "debug/reference_rejected_logps": -176.87411499023438,
417
+ "epoch": 0.6060606060606061,
418
+ "grad_norm": 2.8616257419732705,
419
+ "learning_rate": 1e-06,
420
+ "logits/chosen": -2.60622501373291,
421
+ "logits/rejected": -2.4916000366210938,
422
+ "logps/chosen": -195.92251586914062,
423
+ "logps/rejected": -182.05645751953125,
424
+ "loss": 0.4718,
425
+ "rewards/accuracies": 0.625,
426
+ "rewards/chosen": 0.002417393261566758,
427
+ "rewards/margins": 0.05424093082547188,
428
+ "rewards/rejected": -0.05182354152202606,
429
+ "step": 20
430
+ },
431
+ {
432
+ "debug/policy_chosen_logits": -2.449052572250366,
433
+ "debug/policy_chosen_logps": -171.43643188476562,
434
+ "debug/policy_rejected_logits": -2.4381000995635986,
435
+ "debug/policy_rejected_logps": -179.66990661621094,
436
+ "debug/reference_chosen_logps": -172.74703979492188,
437
+ "debug/reference_rejected_logps": -178.99798583984375,
438
+ "epoch": 0.6363636363636364,
439
+ "grad_norm": 2.7027165426616646,
440
+ "learning_rate": 1e-06,
441
+ "logits/chosen": -2.449052572250366,
442
+ "logits/rejected": -2.4381000995635986,
443
+ "logps/chosen": -171.43643188476562,
444
+ "logps/rejected": -179.66990661621094,
445
+ "loss": 0.4773,
446
+ "rewards/accuracies": 0.375,
447
+ "rewards/chosen": 0.013106021098792553,
448
+ "rewards/margins": 0.019825361669063568,
449
+ "rewards/rejected": -0.006719340104609728,
450
+ "step": 21
451
+ },
452
+ {
453
+ "debug/policy_chosen_logits": -2.5421152114868164,
454
+ "debug/policy_chosen_logps": -205.93255615234375,
455
+ "debug/policy_rejected_logits": -2.630420446395874,
456
+ "debug/policy_rejected_logps": -230.3187255859375,
457
+ "debug/reference_chosen_logps": -208.52468872070312,
458
+ "debug/reference_rejected_logps": -228.58242797851562,
459
+ "epoch": 0.6666666666666666,
460
+ "grad_norm": 2.7272923033258256,
461
+ "learning_rate": 1e-06,
462
+ "logits/chosen": -2.5421152114868164,
463
+ "logits/rejected": -2.630420446395874,
464
+ "logps/chosen": -205.93255615234375,
465
+ "logps/rejected": -230.3187255859375,
466
+ "loss": 0.4876,
467
+ "rewards/accuracies": 0.625,
468
+ "rewards/chosen": 0.02592136338353157,
469
+ "rewards/margins": 0.0432843416929245,
470
+ "rewards/rejected": -0.01736297458410263,
471
+ "step": 22
472
+ },
473
+ {
474
+ "debug/policy_chosen_logits": -2.6116299629211426,
475
+ "debug/policy_chosen_logps": -183.0703125,
476
+ "debug/policy_rejected_logits": -2.611067295074463,
477
+ "debug/policy_rejected_logps": -210.6973876953125,
478
+ "debug/reference_chosen_logps": -183.8590087890625,
479
+ "debug/reference_rejected_logps": -211.12478637695312,
480
+ "epoch": 0.696969696969697,
481
+ "grad_norm": 2.5040109839505282,
482
+ "learning_rate": 1e-06,
483
+ "logits/chosen": -2.6116299629211426,
484
+ "logits/rejected": -2.611067295074463,
485
+ "logps/chosen": -183.0703125,
486
+ "logps/rejected": -210.6973876953125,
487
+ "loss": 0.4779,
488
+ "rewards/accuracies": 0.5,
489
+ "rewards/chosen": 0.007887020707130432,
490
+ "rewards/margins": 0.003613131120800972,
491
+ "rewards/rejected": 0.004273890517652035,
492
+ "step": 23
493
+ },
494
+ {
495
+ "debug/policy_chosen_logits": -2.5641233921051025,
496
+ "debug/policy_chosen_logps": -160.5648193359375,
497
+ "debug/policy_rejected_logits": -2.555621862411499,
498
+ "debug/policy_rejected_logps": -203.09788513183594,
499
+ "debug/reference_chosen_logps": -165.14849853515625,
500
+ "debug/reference_rejected_logps": -203.03428649902344,
501
+ "epoch": 0.7272727272727273,
502
+ "grad_norm": 2.8849926512874946,
503
+ "learning_rate": 1e-06,
504
+ "logits/chosen": -2.5641233921051025,
505
+ "logits/rejected": -2.555621862411499,
506
+ "logps/chosen": -160.5648193359375,
507
+ "logps/rejected": -203.09788513183594,
508
+ "loss": 0.4788,
509
+ "rewards/accuracies": 0.75,
510
+ "rewards/chosen": 0.04583675414323807,
511
+ "rewards/margins": 0.046472661197185516,
512
+ "rewards/rejected": -0.000635908916592598,
513
+ "step": 24
514
+ },
515
+ {
516
+ "debug/policy_chosen_logits": -2.4982876777648926,
517
+ "debug/policy_chosen_logps": -182.13987731933594,
518
+ "debug/policy_rejected_logits": -2.5158534049987793,
519
+ "debug/policy_rejected_logps": -197.07666015625,
520
+ "debug/reference_chosen_logps": -180.93638610839844,
521
+ "debug/reference_rejected_logps": -195.06378173828125,
522
+ "epoch": 0.7575757575757576,
523
+ "grad_norm": 2.573419249626687,
524
+ "learning_rate": 1e-06,
525
+ "logits/chosen": -2.4982876777648926,
526
+ "logits/rejected": -2.5158534049987793,
527
+ "logps/chosen": -182.13987731933594,
528
+ "logps/rejected": -197.07666015625,
529
+ "loss": 0.4723,
530
+ "rewards/accuracies": 0.625,
531
+ "rewards/chosen": -0.012034913524985313,
532
+ "rewards/margins": 0.00809389166533947,
533
+ "rewards/rejected": -0.020128803327679634,
534
+ "step": 25
535
+ },
536
+ {
537
+ "debug/policy_chosen_logits": -2.5644371509552,
538
+ "debug/policy_chosen_logps": -190.55043029785156,
539
+ "debug/policy_rejected_logits": -2.563399314880371,
540
+ "debug/policy_rejected_logps": -203.17471313476562,
541
+ "debug/reference_chosen_logps": -189.57899475097656,
542
+ "debug/reference_rejected_logps": -200.677490234375,
543
+ "epoch": 0.7878787878787878,
544
+ "grad_norm": 2.6451303932135497,
545
+ "learning_rate": 1e-06,
546
+ "logits/chosen": -2.5644371509552,
547
+ "logits/rejected": -2.563399314880371,
548
+ "logps/chosen": -190.55043029785156,
549
+ "logps/rejected": -203.17471313476562,
550
+ "loss": 0.4921,
551
+ "rewards/accuracies": 0.625,
552
+ "rewards/chosen": -0.009714392013847828,
553
+ "rewards/margins": 0.015257682651281357,
554
+ "rewards/rejected": -0.02497207559645176,
555
+ "step": 26
556
+ },
557
+ {
558
+ "debug/policy_chosen_logits": -2.5107343196868896,
559
+ "debug/policy_chosen_logps": -157.84153747558594,
560
+ "debug/policy_rejected_logits": -2.465113401412964,
561
+ "debug/policy_rejected_logps": -193.6126708984375,
562
+ "debug/reference_chosen_logps": -162.47134399414062,
563
+ "debug/reference_rejected_logps": -187.2801971435547,
564
+ "epoch": 0.8181818181818182,
565
+ "grad_norm": 2.6477503407437504,
566
+ "learning_rate": 1e-06,
567
+ "logits/chosen": -2.5107343196868896,
568
+ "logits/rejected": -2.465113401412964,
569
+ "logps/chosen": -157.84153747558594,
570
+ "logps/rejected": -193.6126708984375,
571
+ "loss": 0.4575,
572
+ "rewards/accuracies": 1.0,
573
+ "rewards/chosen": 0.04629802703857422,
574
+ "rewards/margins": 0.10962266474962234,
575
+ "rewards/rejected": -0.06332464516162872,
576
+ "step": 27
577
+ },
578
+ {
579
+ "debug/policy_chosen_logits": -2.6364948749542236,
580
+ "debug/policy_chosen_logps": -170.59153747558594,
581
+ "debug/policy_rejected_logits": -2.5919628143310547,
582
+ "debug/policy_rejected_logps": -202.32334899902344,
583
+ "debug/reference_chosen_logps": -173.39129638671875,
584
+ "debug/reference_rejected_logps": -198.57765197753906,
585
+ "epoch": 0.8484848484848485,
586
+ "grad_norm": 3.545236381819129,
587
+ "learning_rate": 1e-06,
588
+ "logits/chosen": -2.6364948749542236,
589
+ "logits/rejected": -2.5919628143310547,
590
+ "logps/chosen": -170.59153747558594,
591
+ "logps/rejected": -202.32334899902344,
592
+ "loss": 0.4874,
593
+ "rewards/accuracies": 0.75,
594
+ "rewards/chosen": 0.02799762599170208,
595
+ "rewards/margins": 0.06545457243919373,
596
+ "rewards/rejected": -0.037456952035427094,
597
+ "step": 28
598
+ },
599
+ {
600
+ "debug/policy_chosen_logits": -2.543713331222534,
601
+ "debug/policy_chosen_logps": -171.46937561035156,
602
+ "debug/policy_rejected_logits": -2.5579335689544678,
603
+ "debug/policy_rejected_logps": -209.11444091796875,
604
+ "debug/reference_chosen_logps": -172.6361083984375,
605
+ "debug/reference_rejected_logps": -201.7142333984375,
606
+ "epoch": 0.8787878787878788,
607
+ "grad_norm": 3.1668304273549825,
608
+ "learning_rate": 1e-06,
609
+ "logits/chosen": -2.543713331222534,
610
+ "logits/rejected": -2.5579335689544678,
611
+ "logps/chosen": -171.46937561035156,
612
+ "logps/rejected": -209.11444091796875,
613
+ "loss": 0.4725,
614
+ "rewards/accuracies": 0.75,
615
+ "rewards/chosen": 0.011667365208268166,
616
+ "rewards/margins": 0.08566955476999283,
617
+ "rewards/rejected": -0.07400219142436981,
618
+ "step": 29
619
+ },
620
+ {
621
+ "debug/policy_chosen_logits": -2.5963022708892822,
622
+ "debug/policy_chosen_logps": -185.55584716796875,
623
+ "debug/policy_rejected_logits": -2.591168165206909,
624
+ "debug/policy_rejected_logps": -198.93161010742188,
625
+ "debug/reference_chosen_logps": -189.76705932617188,
626
+ "debug/reference_rejected_logps": -198.31280517578125,
627
+ "epoch": 0.9090909090909091,
628
+ "grad_norm": 2.9274538153529783,
629
+ "learning_rate": 1e-06,
630
+ "logits/chosen": -2.5963022708892822,
631
+ "logits/rejected": -2.591168165206909,
632
+ "logps/chosen": -185.55584716796875,
633
+ "logps/rejected": -198.93161010742188,
634
+ "loss": 0.4708,
635
+ "rewards/accuracies": 0.625,
636
+ "rewards/chosen": 0.0421120822429657,
637
+ "rewards/margins": 0.04830005392432213,
638
+ "rewards/rejected": -0.0061879754066467285,
639
+ "step": 30
640
+ },
641
+ {
642
+ "debug/policy_chosen_logits": -2.540618419647217,
643
+ "debug/policy_chosen_logps": -188.67941284179688,
644
+ "debug/policy_rejected_logits": -2.491089105606079,
645
+ "debug/policy_rejected_logps": -194.14166259765625,
646
+ "debug/reference_chosen_logps": -196.98406982421875,
647
+ "debug/reference_rejected_logps": -196.64881896972656,
648
+ "epoch": 0.9393939393939394,
649
+ "grad_norm": 2.76888166586592,
650
+ "learning_rate": 1e-06,
651
+ "logits/chosen": -2.540618419647217,
652
+ "logits/rejected": -2.491089105606079,
653
+ "logps/chosen": -188.67941284179688,
654
+ "logps/rejected": -194.14166259765625,
655
+ "loss": 0.474,
656
+ "rewards/accuracies": 0.75,
657
+ "rewards/chosen": 0.08304648846387863,
658
+ "rewards/margins": 0.05797498673200607,
659
+ "rewards/rejected": 0.025071505457162857,
660
+ "step": 31
661
+ },
662
+ {
663
+ "debug/policy_chosen_logits": -2.5790419578552246,
664
+ "debug/policy_chosen_logps": -173.6497344970703,
665
+ "debug/policy_rejected_logits": -2.4907901287078857,
666
+ "debug/policy_rejected_logps": -215.8475799560547,
667
+ "debug/reference_chosen_logps": -178.68959045410156,
668
+ "debug/reference_rejected_logps": -208.4277801513672,
669
+ "epoch": 0.9696969696969697,
670
+ "grad_norm": 2.612964150864111,
671
+ "learning_rate": 1e-06,
672
+ "logits/chosen": -2.5790419578552246,
673
+ "logits/rejected": -2.4907901287078857,
674
+ "logps/chosen": -173.6497344970703,
675
+ "logps/rejected": -215.8475799560547,
676
+ "loss": 0.4595,
677
+ "rewards/accuracies": 1.0,
678
+ "rewards/chosen": 0.0503985770046711,
679
+ "rewards/margins": 0.12459664791822433,
680
+ "rewards/rejected": -0.07419807463884354,
681
+ "step": 32
682
+ },
683
+ {
684
+ "debug/policy_chosen_logits": -2.553539991378784,
685
+ "debug/policy_chosen_logps": -181.01400756835938,
686
+ "debug/policy_rejected_logits": -2.5499818325042725,
687
+ "debug/policy_rejected_logps": -200.9937286376953,
688
+ "debug/reference_chosen_logps": -185.49740600585938,
689
+ "debug/reference_rejected_logps": -197.2001953125,
690
+ "epoch": 1.0,
691
+ "grad_norm": 2.5944736669447344,
692
+ "learning_rate": 1e-06,
693
+ "logits/chosen": -2.553539991378784,
694
+ "logits/rejected": -2.5499818325042725,
695
+ "logps/chosen": -181.01400756835938,
696
+ "logps/rejected": -200.9937286376953,
697
+ "loss": 0.429,
698
+ "rewards/accuracies": 0.75,
699
+ "rewards/chosen": 0.04483398422598839,
700
+ "rewards/margins": 0.08276933431625366,
701
+ "rewards/rejected": -0.03793535381555557,
702
+ "step": 33
703
+ },
704
+ {
705
+ "epoch": 1.0,
706
+ "step": 33,
707
+ "total_flos": 0.0,
708
+ "train_loss": 0.48355864485104877,
709
+ "train_runtime": 407.656,
710
+ "train_samples_per_second": 5.078,
711
+ "train_steps_per_second": 0.081
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
+ }