yiran-wang3 commited on
Commit
2d04b70
1 Parent(s): 41eccba

End of training

Browse files
README.md ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ license: other
4
+ base_model: yiran-wang3/ds_coder6.7b_adamw_iter5
5
+ tags:
6
+ - alignment-handbook
7
+ - generated_from_trainer
8
+ - trl
9
+ - dpo
10
+ datasets:
11
+ - self-generate/ds_coder6.7b_sppo_hard_new_cn_mining_oj_iter5-binarized
12
+ model-index:
13
+ - name: ds_coder6.7b_adamw_iter6
14
+ results: []
15
+ ---
16
+
17
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
18
+ should probably proofread and complete it, then remove this comment. -->
19
+
20
+ # ds_coder6.7b_adamw_iter6
21
+
22
+ This model is a fine-tuned version of [yiran-wang3/ds_coder6.7b_adamw_iter5](https://huggingface.co/yiran-wang3/ds_coder6.7b_adamw_iter5) on the self-generate/ds_coder6.7b_sppo_hard_new_cn_mining_oj_iter5-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": 0.49622934125363827,
5
+ "train_runtime": 397.2698,
6
+ "train_samples": 1995,
7
+ "train_samples_per_second": 5.022,
8
+ "train_steps_per_second": 0.081
9
+ }
config.json CHANGED
@@ -29,6 +29,6 @@
29
  "tie_word_embeddings": false,
30
  "torch_dtype": "bfloat16",
31
  "transformers_version": "4.45.0",
32
- "use_cache": false,
33
  "vocab_size": 32256
34
  }
 
29
  "tie_word_embeddings": false,
30
  "torch_dtype": "bfloat16",
31
  "transformers_version": "4.45.0",
32
+ "use_cache": true,
33
  "vocab_size": 32256
34
  }
generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 32013,
4
+ "eos_token_id": 32021,
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": 0.49622934125363827,
5
+ "train_runtime": 397.2698,
6
+ "train_samples": 1995,
7
+ "train_samples_per_second": 5.022,
8
+ "train_steps_per_second": 0.081
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,714 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0,
5
+ "eval_steps": 100,
6
+ "global_step": 32,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "debug/policy_chosen_logits": 1.5414493083953857,
13
+ "debug/policy_chosen_logps": -465.5361328125,
14
+ "debug/policy_rejected_logits": 1.6527317762374878,
15
+ "debug/policy_rejected_logps": -427.4836120605469,
16
+ "debug/reference_chosen_logps": -465.5361328125,
17
+ "debug/reference_rejected_logps": -427.4836120605469,
18
+ "epoch": 0.03125,
19
+ "grad_norm": 12.323765072380743,
20
+ "learning_rate": 1e-06,
21
+ "logits/chosen": 1.5414493083953857,
22
+ "logits/rejected": 1.6527317762374878,
23
+ "logps/chosen": -465.5361328125,
24
+ "logps/rejected": -427.4836120605469,
25
+ "loss": 0.5,
26
+ "rewards/accuracies": 0.0,
27
+ "rewards/chosen": 0.0,
28
+ "rewards/margins": 0.0,
29
+ "rewards/rejected": 0.0,
30
+ "step": 1
31
+ },
32
+ {
33
+ "debug/policy_chosen_logits": 1.8114635944366455,
34
+ "debug/policy_chosen_logps": -395.85809326171875,
35
+ "debug/policy_rejected_logits": 2.0628278255462646,
36
+ "debug/policy_rejected_logps": -405.449462890625,
37
+ "debug/reference_chosen_logps": -395.2831726074219,
38
+ "debug/reference_rejected_logps": -405.4757080078125,
39
+ "epoch": 0.0625,
40
+ "grad_norm": 5.688989109980878,
41
+ "learning_rate": 1e-06,
42
+ "logits/chosen": 1.8114635944366455,
43
+ "logits/rejected": 2.0628278255462646,
44
+ "logps/chosen": -395.85809326171875,
45
+ "logps/rejected": -405.449462890625,
46
+ "loss": 0.5021,
47
+ "rewards/accuracies": 0.375,
48
+ "rewards/chosen": -0.005749321077018976,
49
+ "rewards/margins": -0.006011696066707373,
50
+ "rewards/rejected": 0.0002623748732730746,
51
+ "step": 2
52
+ },
53
+ {
54
+ "debug/policy_chosen_logits": 1.734413504600525,
55
+ "debug/policy_chosen_logps": -416.40167236328125,
56
+ "debug/policy_rejected_logits": 1.5799931287765503,
57
+ "debug/policy_rejected_logps": -398.26263427734375,
58
+ "debug/reference_chosen_logps": -416.07275390625,
59
+ "debug/reference_rejected_logps": -398.1192321777344,
60
+ "epoch": 0.09375,
61
+ "grad_norm": 3.905789416952891,
62
+ "learning_rate": 1e-06,
63
+ "logits/chosen": 1.734413504600525,
64
+ "logits/rejected": 1.5799931287765503,
65
+ "logps/chosen": -416.40167236328125,
66
+ "logps/rejected": -398.26263427734375,
67
+ "loss": 0.5007,
68
+ "rewards/accuracies": 0.5,
69
+ "rewards/chosen": -0.00328903179615736,
70
+ "rewards/margins": -0.0018549349624663591,
71
+ "rewards/rejected": -0.0014340970665216446,
72
+ "step": 3
73
+ },
74
+ {
75
+ "debug/policy_chosen_logits": 1.7133558988571167,
76
+ "debug/policy_chosen_logps": -429.8788757324219,
77
+ "debug/policy_rejected_logits": 1.561805009841919,
78
+ "debug/policy_rejected_logps": -398.11846923828125,
79
+ "debug/reference_chosen_logps": -429.8458557128906,
80
+ "debug/reference_rejected_logps": -398.2086486816406,
81
+ "epoch": 0.125,
82
+ "grad_norm": 4.592860856362045,
83
+ "learning_rate": 1e-06,
84
+ "logits/chosen": 1.7133558988571167,
85
+ "logits/rejected": 1.561805009841919,
86
+ "logps/chosen": -429.8788757324219,
87
+ "logps/rejected": -398.11846923828125,
88
+ "loss": 0.5006,
89
+ "rewards/accuracies": 0.375,
90
+ "rewards/chosen": -0.0003304291167296469,
91
+ "rewards/margins": -0.0012322617694735527,
92
+ "rewards/rejected": 0.0009018326527439058,
93
+ "step": 4
94
+ },
95
+ {
96
+ "debug/policy_chosen_logits": 1.6188408136367798,
97
+ "debug/policy_chosen_logps": -405.05450439453125,
98
+ "debug/policy_rejected_logits": 1.7664055824279785,
99
+ "debug/policy_rejected_logps": -389.2750244140625,
100
+ "debug/reference_chosen_logps": -404.82635498046875,
101
+ "debug/reference_rejected_logps": -389.15972900390625,
102
+ "epoch": 0.15625,
103
+ "grad_norm": 4.105078966015891,
104
+ "learning_rate": 1e-06,
105
+ "logits/chosen": 1.6188408136367798,
106
+ "logits/rejected": 1.7664055824279785,
107
+ "logps/chosen": -405.05450439453125,
108
+ "logps/rejected": -389.2750244140625,
109
+ "loss": 0.4992,
110
+ "rewards/accuracies": 0.375,
111
+ "rewards/chosen": -0.002281513065099716,
112
+ "rewards/margins": -0.001129131531342864,
113
+ "rewards/rejected": -0.0011523817665874958,
114
+ "step": 5
115
+ },
116
+ {
117
+ "debug/policy_chosen_logits": 1.5052284002304077,
118
+ "debug/policy_chosen_logps": -417.0735778808594,
119
+ "debug/policy_rejected_logits": 1.241634488105774,
120
+ "debug/policy_rejected_logps": -425.4691467285156,
121
+ "debug/reference_chosen_logps": -417.2569580078125,
122
+ "debug/reference_rejected_logps": -425.63446044921875,
123
+ "epoch": 0.1875,
124
+ "grad_norm": 3.9533420030932884,
125
+ "learning_rate": 1e-06,
126
+ "logits/chosen": 1.5052284002304077,
127
+ "logits/rejected": 1.241634488105774,
128
+ "logps/chosen": -417.0735778808594,
129
+ "logps/rejected": -425.4691467285156,
130
+ "loss": 0.5005,
131
+ "rewards/accuracies": 0.5,
132
+ "rewards/chosen": 0.0018336105858907104,
133
+ "rewards/margins": 0.0001804353087209165,
134
+ "rewards/rejected": 0.0016531753353774548,
135
+ "step": 6
136
+ },
137
+ {
138
+ "debug/policy_chosen_logits": 1.5048482418060303,
139
+ "debug/policy_chosen_logps": -410.84844970703125,
140
+ "debug/policy_rejected_logits": 1.9221699237823486,
141
+ "debug/policy_rejected_logps": -365.1915283203125,
142
+ "debug/reference_chosen_logps": -411.6490478515625,
143
+ "debug/reference_rejected_logps": -365.34991455078125,
144
+ "epoch": 0.21875,
145
+ "grad_norm": 4.1754933609078195,
146
+ "learning_rate": 1e-06,
147
+ "logits/chosen": 1.5048482418060303,
148
+ "logits/rejected": 1.9221699237823486,
149
+ "logps/chosen": -410.84844970703125,
150
+ "logps/rejected": -365.1915283203125,
151
+ "loss": 0.4994,
152
+ "rewards/accuracies": 0.5,
153
+ "rewards/chosen": 0.008005637675523758,
154
+ "rewards/margins": 0.006421546917408705,
155
+ "rewards/rejected": 0.0015840912237763405,
156
+ "step": 7
157
+ },
158
+ {
159
+ "debug/policy_chosen_logits": 1.3764795064926147,
160
+ "debug/policy_chosen_logps": -410.59136962890625,
161
+ "debug/policy_rejected_logits": 1.3361552953720093,
162
+ "debug/policy_rejected_logps": -446.6079406738281,
163
+ "debug/reference_chosen_logps": -411.0892028808594,
164
+ "debug/reference_rejected_logps": -446.5020446777344,
165
+ "epoch": 0.25,
166
+ "grad_norm": 28.29512163500699,
167
+ "learning_rate": 1e-06,
168
+ "logits/chosen": 1.3764795064926147,
169
+ "logits/rejected": 1.3361552953720093,
170
+ "logps/chosen": -410.59136962890625,
171
+ "logps/rejected": -446.6079406738281,
172
+ "loss": 0.4985,
173
+ "rewards/accuracies": 0.75,
174
+ "rewards/chosen": 0.004978599026799202,
175
+ "rewards/margins": 0.006037559360265732,
176
+ "rewards/rejected": -0.0010589603334665298,
177
+ "step": 8
178
+ },
179
+ {
180
+ "debug/policy_chosen_logits": 1.610563039779663,
181
+ "debug/policy_chosen_logps": -406.68865966796875,
182
+ "debug/policy_rejected_logits": 1.793952226638794,
183
+ "debug/policy_rejected_logps": -407.90484619140625,
184
+ "debug/reference_chosen_logps": -406.7731018066406,
185
+ "debug/reference_rejected_logps": -408.06695556640625,
186
+ "epoch": 0.28125,
187
+ "grad_norm": 8.342311302611428,
188
+ "learning_rate": 1e-06,
189
+ "logits/chosen": 1.610563039779663,
190
+ "logits/rejected": 1.793952226638794,
191
+ "logps/chosen": -406.68865966796875,
192
+ "logps/rejected": -407.90484619140625,
193
+ "loss": 0.4981,
194
+ "rewards/accuracies": 0.375,
195
+ "rewards/chosen": 0.0008446313440799713,
196
+ "rewards/margins": -0.0007765005575492978,
197
+ "rewards/rejected": 0.0016211319016292691,
198
+ "step": 9
199
+ },
200
+ {
201
+ "debug/policy_chosen_logits": 1.8192218542099,
202
+ "debug/policy_chosen_logps": -399.52435302734375,
203
+ "debug/policy_rejected_logits": 1.5653293132781982,
204
+ "debug/policy_rejected_logps": -385.7630615234375,
205
+ "debug/reference_chosen_logps": -399.35186767578125,
206
+ "debug/reference_rejected_logps": -385.9487609863281,
207
+ "epoch": 0.3125,
208
+ "grad_norm": 4.2502799962912725,
209
+ "learning_rate": 1e-06,
210
+ "logits/chosen": 1.8192218542099,
211
+ "logits/rejected": 1.5653293132781982,
212
+ "logps/chosen": -399.52435302734375,
213
+ "logps/rejected": -385.7630615234375,
214
+ "loss": 0.5017,
215
+ "rewards/accuracies": 0.5,
216
+ "rewards/chosen": -0.0017250444507226348,
217
+ "rewards/margins": -0.003581695957109332,
218
+ "rewards/rejected": 0.0018566511571407318,
219
+ "step": 10
220
+ },
221
+ {
222
+ "debug/policy_chosen_logits": 1.8145008087158203,
223
+ "debug/policy_chosen_logps": -408.01934814453125,
224
+ "debug/policy_rejected_logits": 1.7603477239608765,
225
+ "debug/policy_rejected_logps": -435.0226135253906,
226
+ "debug/reference_chosen_logps": -408.4494323730469,
227
+ "debug/reference_rejected_logps": -435.68634033203125,
228
+ "epoch": 0.34375,
229
+ "grad_norm": 4.1787448496513075,
230
+ "learning_rate": 1e-06,
231
+ "logits/chosen": 1.8145008087158203,
232
+ "logits/rejected": 1.7603477239608765,
233
+ "logps/chosen": -408.01934814453125,
234
+ "logps/rejected": -435.0226135253906,
235
+ "loss": 0.4975,
236
+ "rewards/accuracies": 0.5,
237
+ "rewards/chosen": 0.004300727508962154,
238
+ "rewards/margins": -0.0023362734355032444,
239
+ "rewards/rejected": 0.006637001410126686,
240
+ "step": 11
241
+ },
242
+ {
243
+ "debug/policy_chosen_logits": 1.6767597198486328,
244
+ "debug/policy_chosen_logps": -384.9023742675781,
245
+ "debug/policy_rejected_logits": 1.4636709690093994,
246
+ "debug/policy_rejected_logps": -410.29254150390625,
247
+ "debug/reference_chosen_logps": -384.634033203125,
248
+ "debug/reference_rejected_logps": -410.1116943359375,
249
+ "epoch": 0.375,
250
+ "grad_norm": 4.319336871459386,
251
+ "learning_rate": 1e-06,
252
+ "logits/chosen": 1.6767597198486328,
253
+ "logits/rejected": 1.4636709690093994,
254
+ "logps/chosen": -384.9023742675781,
255
+ "logps/rejected": -410.29254150390625,
256
+ "loss": 0.497,
257
+ "rewards/accuracies": 0.625,
258
+ "rewards/chosen": -0.0026834101881831884,
259
+ "rewards/margins": -0.0008746716193854809,
260
+ "rewards/rejected": -0.0018087388016283512,
261
+ "step": 12
262
+ },
263
+ {
264
+ "debug/policy_chosen_logits": 1.508570909500122,
265
+ "debug/policy_chosen_logps": -404.47314453125,
266
+ "debug/policy_rejected_logits": 1.6251033544540405,
267
+ "debug/policy_rejected_logps": -459.6101379394531,
268
+ "debug/reference_chosen_logps": -404.60076904296875,
269
+ "debug/reference_rejected_logps": -459.2474365234375,
270
+ "epoch": 0.40625,
271
+ "grad_norm": 4.398229659396608,
272
+ "learning_rate": 1e-06,
273
+ "logits/chosen": 1.508570909500122,
274
+ "logits/rejected": 1.6251033544540405,
275
+ "logps/chosen": -404.47314453125,
276
+ "logps/rejected": -459.6101379394531,
277
+ "loss": 0.4981,
278
+ "rewards/accuracies": 0.625,
279
+ "rewards/chosen": 0.001276321243494749,
280
+ "rewards/margins": 0.004903449676930904,
281
+ "rewards/rejected": -0.0036271288990974426,
282
+ "step": 13
283
+ },
284
+ {
285
+ "debug/policy_chosen_logits": 1.7688783407211304,
286
+ "debug/policy_chosen_logps": -424.601318359375,
287
+ "debug/policy_rejected_logits": 1.3446720838546753,
288
+ "debug/policy_rejected_logps": -452.760009765625,
289
+ "debug/reference_chosen_logps": -423.881103515625,
290
+ "debug/reference_rejected_logps": -452.08984375,
291
+ "epoch": 0.4375,
292
+ "grad_norm": 4.8812354492275505,
293
+ "learning_rate": 1e-06,
294
+ "logits/chosen": 1.7688783407211304,
295
+ "logits/rejected": 1.3446720838546753,
296
+ "logps/chosen": -424.601318359375,
297
+ "logps/rejected": -452.760009765625,
298
+ "loss": 0.5007,
299
+ "rewards/accuracies": 0.25,
300
+ "rewards/chosen": -0.007202301174402237,
301
+ "rewards/margins": -0.0005010606837458909,
302
+ "rewards/rejected": -0.006701240316033363,
303
+ "step": 14
304
+ },
305
+ {
306
+ "debug/policy_chosen_logits": 1.4991490840911865,
307
+ "debug/policy_chosen_logps": -414.95416259765625,
308
+ "debug/policy_rejected_logits": 1.7444727420806885,
309
+ "debug/policy_rejected_logps": -397.67449951171875,
310
+ "debug/reference_chosen_logps": -414.56903076171875,
311
+ "debug/reference_rejected_logps": -398.7281494140625,
312
+ "epoch": 0.46875,
313
+ "grad_norm": 4.614738630368307,
314
+ "learning_rate": 1e-06,
315
+ "logits/chosen": 1.4991490840911865,
316
+ "logits/rejected": 1.7444727420806885,
317
+ "logps/chosen": -414.95416259765625,
318
+ "logps/rejected": -397.67449951171875,
319
+ "loss": 0.5,
320
+ "rewards/accuracies": 0.25,
321
+ "rewards/chosen": -0.0038512798491865396,
322
+ "rewards/margins": -0.014388121664524078,
323
+ "rewards/rejected": 0.010536842048168182,
324
+ "step": 15
325
+ },
326
+ {
327
+ "debug/policy_chosen_logits": 1.8573402166366577,
328
+ "debug/policy_chosen_logps": -411.7935485839844,
329
+ "debug/policy_rejected_logits": 1.655826449394226,
330
+ "debug/policy_rejected_logps": -428.7922668457031,
331
+ "debug/reference_chosen_logps": -411.98040771484375,
332
+ "debug/reference_rejected_logps": -428.9548645019531,
333
+ "epoch": 0.5,
334
+ "grad_norm": 4.37108878989768,
335
+ "learning_rate": 1e-06,
336
+ "logits/chosen": 1.8573402166366577,
337
+ "logits/rejected": 1.655826449394226,
338
+ "logps/chosen": -411.7935485839844,
339
+ "logps/rejected": -428.7922668457031,
340
+ "loss": 0.496,
341
+ "rewards/accuracies": 0.5,
342
+ "rewards/chosen": 0.0018686293624341488,
343
+ "rewards/margins": 0.0002428237348794937,
344
+ "rewards/rejected": 0.0016258049290627241,
345
+ "step": 16
346
+ },
347
+ {
348
+ "debug/policy_chosen_logits": 1.4710519313812256,
349
+ "debug/policy_chosen_logps": -422.3893737792969,
350
+ "debug/policy_rejected_logits": 1.6017435789108276,
351
+ "debug/policy_rejected_logps": -444.07470703125,
352
+ "debug/reference_chosen_logps": -421.87664794921875,
353
+ "debug/reference_rejected_logps": -443.59698486328125,
354
+ "epoch": 0.53125,
355
+ "grad_norm": 3.9347513368146982,
356
+ "learning_rate": 1e-06,
357
+ "logits/chosen": 1.4710519313812256,
358
+ "logits/rejected": 1.6017435789108276,
359
+ "logps/chosen": -422.3893737792969,
360
+ "logps/rejected": -444.07470703125,
361
+ "loss": 0.4916,
362
+ "rewards/accuracies": 0.625,
363
+ "rewards/chosen": -0.00512706721201539,
364
+ "rewards/margins": -0.0003494260599836707,
365
+ "rewards/rejected": -0.004777641035616398,
366
+ "step": 17
367
+ },
368
+ {
369
+ "debug/policy_chosen_logits": 1.6341142654418945,
370
+ "debug/policy_chosen_logps": -445.0538330078125,
371
+ "debug/policy_rejected_logits": 1.6374801397323608,
372
+ "debug/policy_rejected_logps": -459.182373046875,
373
+ "debug/reference_chosen_logps": -444.14422607421875,
374
+ "debug/reference_rejected_logps": -458.5662841796875,
375
+ "epoch": 0.5625,
376
+ "grad_norm": 4.343729471925386,
377
+ "learning_rate": 1e-06,
378
+ "logits/chosen": 1.6341142654418945,
379
+ "logits/rejected": 1.6374801397323608,
380
+ "logps/chosen": -445.0538330078125,
381
+ "logps/rejected": -459.182373046875,
382
+ "loss": 0.4974,
383
+ "rewards/accuracies": 0.375,
384
+ "rewards/chosen": -0.00909606833010912,
385
+ "rewards/margins": -0.0029354477301239967,
386
+ "rewards/rejected": -0.006160621531307697,
387
+ "step": 18
388
+ },
389
+ {
390
+ "debug/policy_chosen_logits": 1.8724883794784546,
391
+ "debug/policy_chosen_logps": -373.3547668457031,
392
+ "debug/policy_rejected_logits": 1.7754302024841309,
393
+ "debug/policy_rejected_logps": -470.7467956542969,
394
+ "debug/reference_chosen_logps": -374.51361083984375,
395
+ "debug/reference_rejected_logps": -470.7672119140625,
396
+ "epoch": 0.59375,
397
+ "grad_norm": 4.542983297185189,
398
+ "learning_rate": 1e-06,
399
+ "logits/chosen": 1.8724883794784546,
400
+ "logits/rejected": 1.7754302024841309,
401
+ "logps/chosen": -373.3547668457031,
402
+ "logps/rejected": -470.7467956542969,
403
+ "loss": 0.4987,
404
+ "rewards/accuracies": 0.625,
405
+ "rewards/chosen": 0.011588573455810547,
406
+ "rewards/margins": 0.011384601704776287,
407
+ "rewards/rejected": 0.0002039717510342598,
408
+ "step": 19
409
+ },
410
+ {
411
+ "debug/policy_chosen_logits": 1.6738826036453247,
412
+ "debug/policy_chosen_logps": -421.78070068359375,
413
+ "debug/policy_rejected_logits": 1.6627225875854492,
414
+ "debug/policy_rejected_logps": -457.22625732421875,
415
+ "debug/reference_chosen_logps": -422.5181884765625,
416
+ "debug/reference_rejected_logps": -456.66961669921875,
417
+ "epoch": 0.625,
418
+ "grad_norm": 4.212961782177093,
419
+ "learning_rate": 1e-06,
420
+ "logits/chosen": 1.6738826036453247,
421
+ "logits/rejected": 1.6627225875854492,
422
+ "logps/chosen": -421.78070068359375,
423
+ "logps/rejected": -457.22625732421875,
424
+ "loss": 0.4966,
425
+ "rewards/accuracies": 0.875,
426
+ "rewards/chosen": 0.007374763488769531,
427
+ "rewards/margins": 0.012941397726535797,
428
+ "rewards/rejected": -0.0055666351690888405,
429
+ "step": 20
430
+ },
431
+ {
432
+ "debug/policy_chosen_logits": 1.6218189001083374,
433
+ "debug/policy_chosen_logps": -396.4649658203125,
434
+ "debug/policy_rejected_logits": 1.6794122457504272,
435
+ "debug/policy_rejected_logps": -393.181640625,
436
+ "debug/reference_chosen_logps": -397.00018310546875,
437
+ "debug/reference_rejected_logps": -393.26422119140625,
438
+ "epoch": 0.65625,
439
+ "grad_norm": 4.415514740564342,
440
+ "learning_rate": 1e-06,
441
+ "logits/chosen": 1.6218189001083374,
442
+ "logits/rejected": 1.6794122457504272,
443
+ "logps/chosen": -396.4649658203125,
444
+ "logps/rejected": -393.181640625,
445
+ "loss": 0.4973,
446
+ "rewards/accuracies": 0.5,
447
+ "rewards/chosen": 0.0053519438952207565,
448
+ "rewards/margins": 0.004526557866483927,
449
+ "rewards/rejected": 0.0008253857959061861,
450
+ "step": 21
451
+ },
452
+ {
453
+ "debug/policy_chosen_logits": 1.8384886980056763,
454
+ "debug/policy_chosen_logps": -445.74896240234375,
455
+ "debug/policy_rejected_logits": 1.820858359336853,
456
+ "debug/policy_rejected_logps": -427.93316650390625,
457
+ "debug/reference_chosen_logps": -445.52197265625,
458
+ "debug/reference_rejected_logps": -427.1968688964844,
459
+ "epoch": 0.6875,
460
+ "grad_norm": 4.279507111277066,
461
+ "learning_rate": 1e-06,
462
+ "logits/chosen": 1.8384886980056763,
463
+ "logits/rejected": 1.820858359336853,
464
+ "logps/chosen": -445.74896240234375,
465
+ "logps/rejected": -427.93316650390625,
466
+ "loss": 0.4976,
467
+ "rewards/accuracies": 0.375,
468
+ "rewards/chosen": -0.002269973512738943,
469
+ "rewards/margins": 0.005093079060316086,
470
+ "rewards/rejected": -0.007363053038716316,
471
+ "step": 22
472
+ },
473
+ {
474
+ "debug/policy_chosen_logits": 1.7056151628494263,
475
+ "debug/policy_chosen_logps": -412.3712463378906,
476
+ "debug/policy_rejected_logits": 1.6649104356765747,
477
+ "debug/policy_rejected_logps": -451.2648620605469,
478
+ "debug/reference_chosen_logps": -412.66668701171875,
479
+ "debug/reference_rejected_logps": -449.33660888671875,
480
+ "epoch": 0.71875,
481
+ "grad_norm": 5.168198816512977,
482
+ "learning_rate": 1e-06,
483
+ "logits/chosen": 1.7056151628494263,
484
+ "logits/rejected": 1.6649104356765747,
485
+ "logps/chosen": -412.3712463378906,
486
+ "logps/rejected": -451.2648620605469,
487
+ "loss": 0.4922,
488
+ "rewards/accuracies": 0.75,
489
+ "rewards/chosen": 0.00295448349788785,
490
+ "rewards/margins": 0.022237051278352737,
491
+ "rewards/rejected": -0.019282570108771324,
492
+ "step": 23
493
+ },
494
+ {
495
+ "debug/policy_chosen_logits": 1.3328857421875,
496
+ "debug/policy_chosen_logps": -398.9266357421875,
497
+ "debug/policy_rejected_logits": 1.627695083618164,
498
+ "debug/policy_rejected_logps": -393.88433837890625,
499
+ "debug/reference_chosen_logps": -399.486328125,
500
+ "debug/reference_rejected_logps": -395.30322265625,
501
+ "epoch": 0.75,
502
+ "grad_norm": 5.635074846323111,
503
+ "learning_rate": 1e-06,
504
+ "logits/chosen": 1.3328857421875,
505
+ "logits/rejected": 1.627695083618164,
506
+ "logps/chosen": -398.9266357421875,
507
+ "logps/rejected": -393.88433837890625,
508
+ "loss": 0.5024,
509
+ "rewards/accuracies": 0.25,
510
+ "rewards/chosen": 0.005597000941634178,
511
+ "rewards/margins": -0.00859165284782648,
512
+ "rewards/rejected": 0.014188650995492935,
513
+ "step": 24
514
+ },
515
+ {
516
+ "debug/policy_chosen_logits": 1.1641021966934204,
517
+ "debug/policy_chosen_logps": -426.21893310546875,
518
+ "debug/policy_rejected_logits": 1.7478457689285278,
519
+ "debug/policy_rejected_logps": -509.15692138671875,
520
+ "debug/reference_chosen_logps": -426.0503845214844,
521
+ "debug/reference_rejected_logps": -508.83636474609375,
522
+ "epoch": 0.78125,
523
+ "grad_norm": 4.992982311072277,
524
+ "learning_rate": 1e-06,
525
+ "logits/chosen": 1.1641021966934204,
526
+ "logits/rejected": 1.7478457689285278,
527
+ "logps/chosen": -426.21893310546875,
528
+ "logps/rejected": -509.15692138671875,
529
+ "loss": 0.4925,
530
+ "rewards/accuracies": 0.5,
531
+ "rewards/chosen": -0.00168533343821764,
532
+ "rewards/margins": 0.0015203863149508834,
533
+ "rewards/rejected": -0.003205719171091914,
534
+ "step": 25
535
+ },
536
+ {
537
+ "debug/policy_chosen_logits": 1.3662775754928589,
538
+ "debug/policy_chosen_logps": -370.339111328125,
539
+ "debug/policy_rejected_logits": 1.361480951309204,
540
+ "debug/policy_rejected_logps": -431.66033935546875,
541
+ "debug/reference_chosen_logps": -370.40283203125,
542
+ "debug/reference_rejected_logps": -430.5782775878906,
543
+ "epoch": 0.8125,
544
+ "grad_norm": 5.426783468499563,
545
+ "learning_rate": 1e-06,
546
+ "logits/chosen": 1.3662775754928589,
547
+ "logits/rejected": 1.361480951309204,
548
+ "logps/chosen": -370.339111328125,
549
+ "logps/rejected": -431.66033935546875,
550
+ "loss": 0.4951,
551
+ "rewards/accuracies": 0.75,
552
+ "rewards/chosen": 0.0006370928604155779,
553
+ "rewards/margins": 0.01145759504288435,
554
+ "rewards/rejected": -0.01082050334662199,
555
+ "step": 26
556
+ },
557
+ {
558
+ "debug/policy_chosen_logits": 1.8342163562774658,
559
+ "debug/policy_chosen_logps": -379.77557373046875,
560
+ "debug/policy_rejected_logits": 1.710644245147705,
561
+ "debug/policy_rejected_logps": -458.7286682128906,
562
+ "debug/reference_chosen_logps": -381.13714599609375,
563
+ "debug/reference_rejected_logps": -457.7099304199219,
564
+ "epoch": 0.84375,
565
+ "grad_norm": 4.18822192571096,
566
+ "learning_rate": 1e-06,
567
+ "logits/chosen": 1.8342163562774658,
568
+ "logits/rejected": 1.710644245147705,
569
+ "logps/chosen": -379.77557373046875,
570
+ "logps/rejected": -458.7286682128906,
571
+ "loss": 0.4942,
572
+ "rewards/accuracies": 0.875,
573
+ "rewards/chosen": 0.013616141863167286,
574
+ "rewards/margins": 0.02380317449569702,
575
+ "rewards/rejected": -0.010187034495174885,
576
+ "step": 27
577
+ },
578
+ {
579
+ "debug/policy_chosen_logits": 1.6289498805999756,
580
+ "debug/policy_chosen_logps": -342.238037109375,
581
+ "debug/policy_rejected_logits": 1.474778413772583,
582
+ "debug/policy_rejected_logps": -447.5248718261719,
583
+ "debug/reference_chosen_logps": -343.45904541015625,
584
+ "debug/reference_rejected_logps": -446.7528076171875,
585
+ "epoch": 0.875,
586
+ "grad_norm": 4.120783195357869,
587
+ "learning_rate": 1e-06,
588
+ "logits/chosen": 1.6289498805999756,
589
+ "logits/rejected": 1.474778413772583,
590
+ "logps/chosen": -342.238037109375,
591
+ "logps/rejected": -447.5248718261719,
592
+ "loss": 0.4991,
593
+ "rewards/accuracies": 0.625,
594
+ "rewards/chosen": 0.012210273183882236,
595
+ "rewards/margins": 0.019930992275476456,
596
+ "rewards/rejected": -0.007720718160271645,
597
+ "step": 28
598
+ },
599
+ {
600
+ "debug/policy_chosen_logits": 1.6164311170578003,
601
+ "debug/policy_chosen_logps": -397.0234680175781,
602
+ "debug/policy_rejected_logits": 1.5657469034194946,
603
+ "debug/policy_rejected_logps": -414.18756103515625,
604
+ "debug/reference_chosen_logps": -397.58160400390625,
605
+ "debug/reference_rejected_logps": -414.6282958984375,
606
+ "epoch": 0.90625,
607
+ "grad_norm": 4.352593213110272,
608
+ "learning_rate": 1e-06,
609
+ "logits/chosen": 1.6164311170578003,
610
+ "logits/rejected": 1.5657469034194946,
611
+ "logps/chosen": -397.0234680175781,
612
+ "logps/rejected": -414.18756103515625,
613
+ "loss": 0.4981,
614
+ "rewards/accuracies": 0.5,
615
+ "rewards/chosen": 0.005581206642091274,
616
+ "rewards/margins": 0.0011737439781427383,
617
+ "rewards/rejected": 0.004407462663948536,
618
+ "step": 29
619
+ },
620
+ {
621
+ "debug/policy_chosen_logits": 1.3324956893920898,
622
+ "debug/policy_chosen_logps": -384.20556640625,
623
+ "debug/policy_rejected_logits": 1.4190467596054077,
624
+ "debug/policy_rejected_logps": -439.9374694824219,
625
+ "debug/reference_chosen_logps": -384.3076171875,
626
+ "debug/reference_rejected_logps": -439.364990234375,
627
+ "epoch": 0.9375,
628
+ "grad_norm": 4.24197252932316,
629
+ "learning_rate": 1e-06,
630
+ "logits/chosen": 1.3324956893920898,
631
+ "logits/rejected": 1.4190467596054077,
632
+ "logps/chosen": -384.20556640625,
633
+ "logps/rejected": -439.9374694824219,
634
+ "loss": 0.4902,
635
+ "rewards/accuracies": 0.625,
636
+ "rewards/chosen": 0.0010205453727394342,
637
+ "rewards/margins": 0.006745452992618084,
638
+ "rewards/rejected": -0.005724907387048006,
639
+ "step": 30
640
+ },
641
+ {
642
+ "debug/policy_chosen_logits": 1.6599881649017334,
643
+ "debug/policy_chosen_logps": -391.8392333984375,
644
+ "debug/policy_rejected_logits": 1.440269947052002,
645
+ "debug/policy_rejected_logps": -410.36962890625,
646
+ "debug/reference_chosen_logps": -391.9766845703125,
647
+ "debug/reference_rejected_logps": -410.27911376953125,
648
+ "epoch": 0.96875,
649
+ "grad_norm": 4.2333369790689614,
650
+ "learning_rate": 1e-06,
651
+ "logits/chosen": 1.6599881649017334,
652
+ "logits/rejected": 1.440269947052002,
653
+ "logps/chosen": -391.8392333984375,
654
+ "logps/rejected": -410.36962890625,
655
+ "loss": 0.4915,
656
+ "rewards/accuracies": 0.5,
657
+ "rewards/chosen": 0.0013742828741669655,
658
+ "rewards/margins": 0.002279548905789852,
659
+ "rewards/rejected": -0.0009052654495462775,
660
+ "step": 31
661
+ },
662
+ {
663
+ "debug/policy_chosen_logits": 1.3013134002685547,
664
+ "debug/policy_chosen_logps": -468.1767883300781,
665
+ "debug/policy_rejected_logits": 1.5036485195159912,
666
+ "debug/policy_rejected_logps": -528.7972412109375,
667
+ "debug/reference_chosen_logps": -468.2105712890625,
668
+ "debug/reference_rejected_logps": -525.7525634765625,
669
+ "epoch": 1.0,
670
+ "grad_norm": 4.709201002942774,
671
+ "learning_rate": 1e-06,
672
+ "logits/chosen": 1.3013134002685547,
673
+ "logits/rejected": 1.5036485195159912,
674
+ "logps/chosen": -468.1767883300781,
675
+ "logps/rejected": -528.7972412109375,
676
+ "loss": 0.4547,
677
+ "rewards/accuracies": 1.0,
678
+ "rewards/chosen": 0.00033790571615099907,
679
+ "rewards/margins": 0.03078407421708107,
680
+ "rewards/rejected": -0.030446168035268784,
681
+ "step": 32
682
+ },
683
+ {
684
+ "epoch": 1.0,
685
+ "step": 32,
686
+ "total_flos": 0.0,
687
+ "train_loss": 0.49622934125363827,
688
+ "train_runtime": 397.2698,
689
+ "train_samples_per_second": 5.022,
690
+ "train_steps_per_second": 0.081
691
+ }
692
+ ],
693
+ "logging_steps": 1,
694
+ "max_steps": 32,
695
+ "num_input_tokens_seen": 0,
696
+ "num_train_epochs": 1,
697
+ "save_steps": 500,
698
+ "stateful_callbacks": {
699
+ "TrainerControl": {
700
+ "args": {
701
+ "should_epoch_stop": false,
702
+ "should_evaluate": false,
703
+ "should_log": false,
704
+ "should_save": true,
705
+ "should_training_stop": true
706
+ },
707
+ "attributes": {}
708
+ }
709
+ },
710
+ "total_flos": 0.0,
711
+ "train_batch_size": 8,
712
+ "trial_name": null,
713
+ "trial_params": null
714
+ }