Hugging Face's logo

YYYYYYibo
/
nash_dpo_doff_real_no_golden_iter_3

PEFT
Safetensors
mistral
alignment-handbook
Generated from Trainer
trl
dpo
Model card Files Community
YYYYYYibo commited on
Commit
4a01f72
·
verified ·
1 Parent(s): 89ace55

Model save

Browse files
Files changed (4) hide show
  1. README.md +75 -0
  2. all_results.json +8 -0
  3. train_results.json +8 -0
  4. trainer_state.json +284 -0
README.md ADDED
@@ -0,0 +1,75 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ library_name: peft
4
+ tags:
5
+ - trl
6
+ - dpo
7
+ - generated_from_trainer
8
+ base_model: alignment-handbook/zephyr-7b-sft-full
9
+ model-index:
10
+ - name: nash_dpo_doff_real_no_golden_iter_3
11
+ results: []
12
+ ---
13
+
14
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
15
+ should probably proofread and complete it, then remove this comment. -->
16
+
17
+ # nash_dpo_doff_real_no_golden_iter_3
18
+
19
+ This model is a fine-tuned version of [alignment-handbook/zephyr-7b-sft-full](https://huggingface.co/alignment-handbook/zephyr-7b-sft-full) on the None dataset.
20
+ It achieves the following results on the evaluation set:
21
+ - Loss: 0.6635
22
+ - Rewards/chosen: 0.0266
23
+ - Rewards/rejected: -0.0491
24
+ - Rewards/accuracies: 0.5920
25
+ - Rewards/margins: 0.0757
26
+ - Logps/rejected: -276.3960
27
+ - Logps/chosen: -293.4469
28
+ - Logits/rejected: -2.5081
29
+ - Logits/chosen: -2.6072
30
+
31
+ ## Model description
32
+
33
+ More information needed
34
+
35
+ ## Intended uses & limitations
36
+
37
+ More information needed
38
+
39
+ ## Training and evaluation data
40
+
41
+ More information needed
42
+
43
+ ## Training procedure
44
+
45
+ ### Training hyperparameters
46
+
47
+ The following hyperparameters were used during training:
48
+ - learning_rate: 5e-06
49
+ - train_batch_size: 2
50
+ - eval_batch_size: 2
51
+ - seed: 42
52
+ - distributed_type: multi-GPU
53
+ - num_devices: 4
54
+ - gradient_accumulation_steps: 16
55
+ - total_train_batch_size: 128
56
+ - total_eval_batch_size: 8
57
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
58
+ - lr_scheduler_type: cosine
59
+ - lr_scheduler_warmup_ratio: 0.1
60
+ - num_epochs: 1
61
+
62
+ ### Training results
63
+
64
+ | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
65
+ |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
66
+ | 0.6619 | 0.62 | 100 | 0.6635 | 0.0266 | -0.0491 | 0.5920 | 0.0757 | -276.3960 | -293.4469 | -2.5081 | -2.6072 |
67
+
68
+
69
+ ### Framework versions
70
+
71
+ - PEFT 0.7.1
72
+ - Transformers 4.36.2
73
+ - Pytorch 2.1.2+cu121
74
+ - Datasets 2.14.6
75
+ - Tokenizers 0.15.2
all_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "train_loss": 0.6703717057008921,
4
+ "train_runtime": 7352.5467,
5
+ "train_samples": 20635,
6
+ "train_samples_per_second": 2.807,
7
+ "train_steps_per_second": 0.022
8
+ }
train_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "train_loss": 0.6703717057008921,
4
+ "train_runtime": 7352.5467,
5
+ "train_samples": 20635,
6
+ "train_samples_per_second": 2.807,
7
+ "train_steps_per_second": 0.022
8
+ }
trainer_state.json ADDED
@@ -0,0 +1,284 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.9984496124031008,
5
+ "eval_steps": 100,
6
+ "global_step": 161,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.01,
13
+ "learning_rate": 2.9411764705882356e-07,
14
+ "logits/chosen": -2.7490084171295166,
15
+ "logits/rejected": -2.6610748767852783,
16
+ "logps/chosen": -299.482421875,
17
+ "logps/rejected": -284.2821960449219,
18
+ "loss": 0.6931,
19
+ "rewards/accuracies": 0.0,
20
+ "rewards/chosen": 0.0,
21
+ "rewards/margins": 0.0,
22
+ "rewards/rejected": 0.0,
23
+ "step": 1
24
+ },
25
+ {
26
+ "epoch": 0.06,
27
+ "learning_rate": 2.9411764705882355e-06,
28
+ "logits/chosen": -2.71052885055542,
29
+ "logits/rejected": -2.7085328102111816,
30
+ "logps/chosen": -263.4867858886719,
31
+ "logps/rejected": -292.8466491699219,
32
+ "loss": 0.6931,
33
+ "rewards/accuracies": 0.4305555522441864,
34
+ "rewards/chosen": 0.0012557146837934852,
35
+ "rewards/margins": -7.775126141496003e-05,
36
+ "rewards/rejected": 0.0013334659161046147,
37
+ "step": 10
38
+ },
39
+ {
40
+ "epoch": 0.12,
41
+ "learning_rate": 4.994647308096509e-06,
42
+ "logits/chosen": -2.7212631702423096,
43
+ "logits/rejected": -2.6979565620422363,
44
+ "logps/chosen": -281.99365234375,
45
+ "logps/rejected": -305.00482177734375,
46
+ "loss": 0.6912,
47
+ "rewards/accuracies": 0.5218750238418579,
48
+ "rewards/chosen": 0.04867880418896675,
49
+ "rewards/margins": 0.002921257633715868,
50
+ "rewards/rejected": 0.04575754702091217,
51
+ "step": 20
52
+ },
53
+ {
54
+ "epoch": 0.19,
55
+ "learning_rate": 4.900124635964823e-06,
56
+ "logits/chosen": -2.673609495162964,
57
+ "logits/rejected": -2.6867575645446777,
58
+ "logps/chosen": -256.4033203125,
59
+ "logps/rejected": -289.47882080078125,
60
+ "loss": 0.6871,
61
+ "rewards/accuracies": 0.565625011920929,
62
+ "rewards/chosen": 0.1824948489665985,
63
+ "rewards/margins": 0.019166249781847,
64
+ "rewards/rejected": 0.1633286029100418,
65
+ "step": 30
66
+ },
67
+ {
68
+ "epoch": 0.25,
69
+ "learning_rate": 4.69181688926877e-06,
70
+ "logits/chosen": -2.6670594215393066,
71
+ "logits/rejected": -2.669649362564087,
72
+ "logps/chosen": -255.5750732421875,
73
+ "logps/rejected": -286.7828063964844,
74
+ "loss": 0.6806,
75
+ "rewards/accuracies": 0.625,
76
+ "rewards/chosen": 0.09983734786510468,
77
+ "rewards/margins": 0.02368326112627983,
78
+ "rewards/rejected": 0.07615408301353455,
79
+ "step": 40
80
+ },
81
+ {
82
+ "epoch": 0.31,
83
+ "learning_rate": 4.379599518697444e-06,
84
+ "logits/chosen": -2.6448562145233154,
85
+ "logits/rejected": -2.626420497894287,
86
+ "logps/chosen": -261.439208984375,
87
+ "logps/rejected": -301.18280029296875,
88
+ "loss": 0.6705,
89
+ "rewards/accuracies": 0.6656249761581421,
90
+ "rewards/chosen": 0.13415098190307617,
91
+ "rewards/margins": 0.053799472749233246,
92
+ "rewards/rejected": 0.08035150915384293,
93
+ "step": 50
94
+ },
95
+ {
96
+ "epoch": 0.37,
97
+ "learning_rate": 3.978274120908957e-06,
98
+ "logits/chosen": -2.6496951580047607,
99
+ "logits/rejected": -2.611518144607544,
100
+ "logps/chosen": -280.60791015625,
101
+ "logps/rejected": -300.1086730957031,
102
+ "loss": 0.6729,
103
+ "rewards/accuracies": 0.59375,
104
+ "rewards/chosen": 0.054178010672330856,
105
+ "rewards/margins": 0.04839291423559189,
106
+ "rewards/rejected": 0.005785099230706692,
107
+ "step": 60
108
+ },
109
+ {
110
+ "epoch": 0.43,
111
+ "learning_rate": 3.5068667246468437e-06,
112
+ "logits/chosen": -2.590028762817383,
113
+ "logits/rejected": -2.576371669769287,
114
+ "logps/chosen": -266.51116943359375,
115
+ "logps/rejected": -330.80792236328125,
116
+ "loss": 0.6691,
117
+ "rewards/accuracies": 0.643750011920929,
118
+ "rewards/chosen": -0.03388174623250961,
119
+ "rewards/margins": 0.06915116310119629,
120
+ "rewards/rejected": -0.1030329242348671,
121
+ "step": 70
122
+ },
123
+ {
124
+ "epoch": 0.5,
125
+ "learning_rate": 2.9877258050403214e-06,
126
+ "logits/chosen": -2.577664375305176,
127
+ "logits/rejected": -2.585880756378174,
128
+ "logps/chosen": -255.4573516845703,
129
+ "logps/rejected": -296.2555236816406,
130
+ "loss": 0.6672,
131
+ "rewards/accuracies": 0.628125011920929,
132
+ "rewards/chosen": 0.07994996011257172,
133
+ "rewards/margins": 0.06973399966955185,
134
+ "rewards/rejected": 0.01021595485508442,
135
+ "step": 80
136
+ },
137
+ {
138
+ "epoch": 0.56,
139
+ "learning_rate": 2.4454627874135976e-06,
140
+ "logits/chosen": -2.5908098220825195,
141
+ "logits/rejected": -2.602696657180786,
142
+ "logps/chosen": -274.9224853515625,
143
+ "logps/rejected": -314.12384033203125,
144
+ "loss": 0.6678,
145
+ "rewards/accuracies": 0.590624988079071,
146
+ "rewards/chosen": -0.0693972036242485,
147
+ "rewards/margins": 0.05152568221092224,
148
+ "rewards/rejected": -0.12092288583517075,
149
+ "step": 90
150
+ },
151
+ {
152
+ "epoch": 0.62,
153
+ "learning_rate": 1.9057852691845677e-06,
154
+ "logits/chosen": -2.565824031829834,
155
+ "logits/rejected": -2.594512462615967,
156
+ "logps/chosen": -260.5049743652344,
157
+ "logps/rejected": -310.77191162109375,
158
+ "loss": 0.6619,
159
+ "rewards/accuracies": 0.6499999761581421,
160
+ "rewards/chosen": 0.0450400672852993,
161
+ "rewards/margins": 0.08433112502098083,
162
+ "rewards/rejected": -0.03929106146097183,
163
+ "step": 100
164
+ },
165
+ {
166
+ "epoch": 0.62,
167
+ "eval_logits/chosen": -2.6071722507476807,
168
+ "eval_logits/rejected": -2.508082628250122,
169
+ "eval_logps/chosen": -293.4469299316406,
170
+ "eval_logps/rejected": -276.3959655761719,
171
+ "eval_loss": 0.663548469543457,
172
+ "eval_rewards/accuracies": 0.5920000076293945,
173
+ "eval_rewards/chosen": 0.02658846043050289,
174
+ "eval_rewards/margins": 0.07567868381738663,
175
+ "eval_rewards/rejected": -0.04909021407365799,
176
+ "eval_runtime": 397.106,
177
+ "eval_samples_per_second": 5.036,
178
+ "eval_steps_per_second": 0.63,
179
+ "step": 100
180
+ },
181
+ {
182
+ "epoch": 0.68,
183
+ "learning_rate": 1.3942782744524974e-06,
184
+ "logits/chosen": -2.6018948554992676,
185
+ "logits/rejected": -2.5732669830322266,
186
+ "logps/chosen": -280.30120849609375,
187
+ "logps/rejected": -298.91925048828125,
188
+ "loss": 0.6676,
189
+ "rewards/accuracies": 0.6499999761581421,
190
+ "rewards/chosen": 0.04486658424139023,
191
+ "rewards/margins": 0.06423817574977875,
192
+ "rewards/rejected": -0.019371582195162773,
193
+ "step": 110
194
+ },
195
+ {
196
+ "epoch": 0.74,
197
+ "learning_rate": 9.351913195398523e-07,
198
+ "logits/chosen": -2.6056082248687744,
199
+ "logits/rejected": -2.623044490814209,
200
+ "logps/chosen": -284.22393798828125,
201
+ "logps/rejected": -323.26153564453125,
202
+ "loss": 0.6574,
203
+ "rewards/accuracies": 0.653124988079071,
204
+ "rewards/chosen": -0.029515612870454788,
205
+ "rewards/margins": 0.09018988907337189,
206
+ "rewards/rejected": -0.11970548331737518,
207
+ "step": 120
208
+ },
209
+ {
210
+ "epoch": 0.81,
211
+ "learning_rate": 5.50288792267796e-07,
212
+ "logits/chosen": -2.5670642852783203,
213
+ "logits/rejected": -2.546692371368408,
214
+ "logps/chosen": -275.8234558105469,
215
+ "logps/rejected": -316.80291748046875,
216
+ "loss": 0.6611,
217
+ "rewards/accuracies": 0.6000000238418579,
218
+ "rewards/chosen": 0.015085836872458458,
219
+ "rewards/margins": 0.08981107175350189,
220
+ "rewards/rejected": -0.07472522556781769,
221
+ "step": 130
222
+ },
223
+ {
224
+ "epoch": 0.87,
225
+ "learning_rate": 2.5781814616827936e-07,
226
+ "logits/chosen": -2.602719783782959,
227
+ "logits/rejected": -2.62253999710083,
228
+ "logps/chosen": -277.62322998046875,
229
+ "logps/rejected": -338.15216064453125,
230
+ "loss": 0.664,
231
+ "rewards/accuracies": 0.59375,
232
+ "rewards/chosen": 0.039921779185533524,
233
+ "rewards/margins": 0.1004234328866005,
234
+ "rewards/rejected": -0.06050165742635727,
235
+ "step": 140
236
+ },
237
+ {
238
+ "epoch": 0.93,
239
+ "learning_rate": 7.164482546684642e-08,
240
+ "logits/chosen": -2.5598411560058594,
241
+ "logits/rejected": -2.5621132850646973,
242
+ "logps/chosen": -266.3095703125,
243
+ "logps/rejected": -310.7863464355469,
244
+ "loss": 0.6615,
245
+ "rewards/accuracies": 0.653124988079071,
246
+ "rewards/chosen": 0.02628152072429657,
247
+ "rewards/margins": 0.10357757657766342,
248
+ "rewards/rejected": -0.07729605585336685,
249
+ "step": 150
250
+ },
251
+ {
252
+ "epoch": 0.99,
253
+ "learning_rate": 5.94932300227169e-10,
254
+ "logits/chosen": -2.5754361152648926,
255
+ "logits/rejected": -2.580561876296997,
256
+ "logps/chosen": -266.0080871582031,
257
+ "logps/rejected": -315.78167724609375,
258
+ "loss": 0.6573,
259
+ "rewards/accuracies": 0.606249988079071,
260
+ "rewards/chosen": 0.02933870628476143,
261
+ "rewards/margins": 0.08014971017837524,
262
+ "rewards/rejected": -0.050811003893613815,
263
+ "step": 160
264
+ },
265
+ {
266
+ "epoch": 1.0,
267
+ "step": 161,
268
+ "total_flos": 0.0,
269
+ "train_loss": 0.6703717057008921,
270
+ "train_runtime": 7352.5467,
271
+ "train_samples_per_second": 2.807,
272
+ "train_steps_per_second": 0.022
273
+ }
274
+ ],
275
+ "logging_steps": 10,
276
+ "max_steps": 161,
277
+ "num_input_tokens_seen": 0,
278
+ "num_train_epochs": 1,
279
+ "save_steps": 100,
280
+ "total_flos": 0.0,
281
+ "train_batch_size": 2,
282
+ "trial_name": null,
283
+ "trial_params": null
284
+ }