Hugging Face's logo

YYYYYYibo
/
gshf_lora_iter_2

PEFT
Safetensors
mistral
alignment-handbook
Generated from Trainer
trl
dpo
Model card Files Community
YYYYYYibo commited on
Commit
135bd56
•
1 Parent(s): 36014bb

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 +270 -0
README.md ADDED
@@ -0,0 +1,75 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: alignment-handbook/zephyr-7b-sft-full
3
+ library_name: peft
4
+ license: apache-2.0
5
+ tags:
6
+ - trl
7
+ - dpo
8
+ - generated_from_trainer
9
+ model-index:
10
+ - name: gshf_lora_iter_2
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
+ # gshf_lora_iter_2
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: 0.6731
22
+ - Rewards/chosen: -0.6240
23
+ - Rewards/rejected: -0.7259
24
+ - Rewards/accuracies: 0.5700
25
+ - Rewards/margins: 0.1019
26
+ - Logps/rejected: -362.4261
27
+ - Logps/chosen: -360.8068
28
+ - Logits/rejected: -2.3366
29
+ - Logits/chosen: -2.4455
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
+ | 0.6268 | 0.64 | 100 | 0.6731 | -0.6240 | -0.7259 | 0.5700 | 0.1019 | -362.4261 | -360.8068 | -2.3366 | -2.4455 |
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": 0.6301614909599988,
4
+ "train_runtime": 6946.965,
5
+ "train_samples": 20000,
6
+ "train_samples_per_second": 2.879,
7
+ "train_steps_per_second": 0.022
8
+ }
train_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "train_loss": 0.6301614909599988,
4
+ "train_runtime": 6946.965,
5
+ "train_samples": 20000,
6
+ "train_samples_per_second": 2.879,
7
+ "train_steps_per_second": 0.022
8
+ }
trainer_state.json ADDED
@@ -0,0 +1,270 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.9984,
5
+ "eval_steps": 100,
6
+ "global_step": 156,
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": 3.125e-07,
14
+ "logits/chosen": -2.6029441356658936,
15
+ "logits/rejected": -2.5599632263183594,
16
+ "logps/chosen": -189.41246032714844,
17
+ "logps/rejected": -190.13034057617188,
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": 3.125e-06,
28
+ "logits/chosen": -2.6171505451202393,
29
+ "logits/rejected": -2.537480592727661,
30
+ "logps/chosen": -160.8114776611328,
31
+ "logps/rejected": -149.8831024169922,
32
+ "loss": 0.6913,
33
+ "rewards/accuracies": 0.4618055522441864,
34
+ "rewards/chosen": -0.009910209104418755,
35
+ "rewards/margins": 0.004514796659350395,
36
+ "rewards/rejected": -0.01442500576376915,
37
+ "step": 10
38
+ },
39
+ {
40
+ "epoch": 0.13,
41
+ "learning_rate": 4.989935734988098e-06,
42
+ "logits/chosen": -2.623683452606201,
43
+ "logits/rejected": -2.5250072479248047,
44
+ "logps/chosen": -167.83267211914062,
45
+ "logps/rejected": -158.3927764892578,
46
+ "loss": 0.676,
47
+ "rewards/accuracies": 0.643750011920929,
48
+ "rewards/chosen": -0.0703786239027977,
49
+ "rewards/margins": 0.037703804671764374,
50
+ "rewards/rejected": -0.10808243602514267,
51
+ "step": 20
52
+ },
53
+ {
54
+ "epoch": 0.19,
55
+ "learning_rate": 4.8776412907378845e-06,
56
+ "logits/chosen": -2.6501214504241943,
57
+ "logits/rejected": -2.4790871143341064,
58
+ "logps/chosen": -172.5028839111328,
59
+ "logps/rejected": -160.69882202148438,
60
+ "loss": 0.6547,
61
+ "rewards/accuracies": 0.6875,
62
+ "rewards/chosen": -0.05348671227693558,
63
+ "rewards/margins": 0.1112779751420021,
64
+ "rewards/rejected": -0.1647646725177765,
65
+ "step": 30
66
+ },
67
+ {
68
+ "epoch": 0.26,
69
+ "learning_rate": 4.646121984004666e-06,
70
+ "logits/chosen": -2.573338031768799,
71
+ "logits/rejected": -2.4198756217956543,
72
+ "logps/chosen": -172.42422485351562,
73
+ "logps/rejected": -173.40711975097656,
74
+ "loss": 0.6345,
75
+ "rewards/accuracies": 0.6937500238418579,
76
+ "rewards/chosen": -0.13879454135894775,
77
+ "rewards/margins": 0.16979987919330597,
78
+ "rewards/rejected": -0.30859440565109253,
79
+ "step": 40
80
+ },
81
+ {
82
+ "epoch": 0.32,
83
+ "learning_rate": 4.3069871595684795e-06,
84
+ "logits/chosen": -2.5714492797851562,
85
+ "logits/rejected": -2.4478890895843506,
86
+ "logps/chosen": -178.8824005126953,
87
+ "logps/rejected": -183.39822387695312,
88
+ "loss": 0.6304,
89
+ "rewards/accuracies": 0.6812499761581421,
90
+ "rewards/chosen": -0.18915040791034698,
91
+ "rewards/margins": 0.17477598786354065,
92
+ "rewards/rejected": -0.36392638087272644,
93
+ "step": 50
94
+ },
95
+ {
96
+ "epoch": 0.38,
97
+ "learning_rate": 3.8772424536302565e-06,
98
+ "logits/chosen": -2.4946584701538086,
99
+ "logits/rejected": -2.3706579208374023,
100
+ "logps/chosen": -199.9522247314453,
101
+ "logps/rejected": -195.98255920410156,
102
+ "loss": 0.631,
103
+ "rewards/accuracies": 0.668749988079071,
104
+ "rewards/chosen": -0.360156387090683,
105
+ "rewards/margins": 0.17627720534801483,
106
+ "rewards/rejected": -0.5364335775375366,
107
+ "step": 60
108
+ },
109
+ {
110
+ "epoch": 0.45,
111
+ "learning_rate": 3.3784370602033572e-06,
112
+ "logits/chosen": -2.4831032752990723,
113
+ "logits/rejected": -2.368504285812378,
114
+ "logps/chosen": -189.5545654296875,
115
+ "logps/rejected": -190.16159057617188,
116
+ "loss": 0.6327,
117
+ "rewards/accuracies": 0.621874988079071,
118
+ "rewards/chosen": -0.35114872455596924,
119
+ "rewards/margins": 0.14523856341838837,
120
+ "rewards/rejected": -0.4963873028755188,
121
+ "step": 70
122
+ },
123
+ {
124
+ "epoch": 0.51,
125
+ "learning_rate": 2.835583164544139e-06,
126
+ "logits/chosen": -2.4916341304779053,
127
+ "logits/rejected": -2.3981051445007324,
128
+ "logps/chosen": -186.53948974609375,
129
+ "logps/rejected": -197.5457305908203,
130
+ "loss": 0.6266,
131
+ "rewards/accuracies": 0.6499999761581421,
132
+ "rewards/chosen": -0.2931613028049469,
133
+ "rewards/margins": 0.19320784509181976,
134
+ "rewards/rejected": -0.48636919260025024,
135
+ "step": 80
136
+ },
137
+ {
138
+ "epoch": 0.58,
139
+ "learning_rate": 2.2759017277414165e-06,
140
+ "logits/chosen": -2.549828290939331,
141
+ "logits/rejected": -2.365307092666626,
142
+ "logps/chosen": -222.1770477294922,
143
+ "logps/rejected": -225.6692657470703,
144
+ "loss": 0.6123,
145
+ "rewards/accuracies": 0.6968749761581421,
146
+ "rewards/chosen": -0.42118874192237854,
147
+ "rewards/margins": 0.27340102195739746,
148
+ "rewards/rejected": -0.6945897340774536,
149
+ "step": 90
150
+ },
151
+ {
152
+ "epoch": 0.64,
153
+ "learning_rate": 1.7274575140626318e-06,
154
+ "logits/chosen": -2.5293145179748535,
155
+ "logits/rejected": -2.3946516513824463,
156
+ "logps/chosen": -191.87879943847656,
157
+ "logps/rejected": -201.5653839111328,
158
+ "loss": 0.6268,
159
+ "rewards/accuracies": 0.668749988079071,
160
+ "rewards/chosen": -0.3195632994174957,
161
+ "rewards/margins": 0.2095298022031784,
162
+ "rewards/rejected": -0.5290930271148682,
163
+ "step": 100
164
+ },
165
+ {
166
+ "epoch": 0.64,
167
+ "eval_logits/chosen": -2.445504665374756,
168
+ "eval_logits/rejected": -2.33659029006958,
169
+ "eval_logps/chosen": -360.80682373046875,
170
+ "eval_logps/rejected": -362.4261169433594,
171
+ "eval_loss": 0.6731494069099426,
172
+ "eval_rewards/accuracies": 0.5699999928474426,
173
+ "eval_rewards/chosen": -0.6240409016609192,
174
+ "eval_rewards/margins": 0.10187112540006638,
175
+ "eval_rewards/rejected": -0.7259120941162109,
176
+ "eval_runtime": 381.8816,
177
+ "eval_samples_per_second": 5.237,
178
+ "eval_steps_per_second": 0.655,
179
+ "step": 100
180
+ },
181
+ {
182
+ "epoch": 0.7,
183
+ "learning_rate": 1.217751806485235e-06,
184
+ "logits/chosen": -2.5149567127227783,
185
+ "logits/rejected": -2.3778505325317383,
186
+ "logps/chosen": -199.1842803955078,
187
+ "logps/rejected": -212.69827270507812,
188
+ "loss": 0.6132,
189
+ "rewards/accuracies": 0.6968749761581421,
190
+ "rewards/chosen": -0.2891606390476227,
191
+ "rewards/margins": 0.2867799699306488,
192
+ "rewards/rejected": -0.5759405493736267,
193
+ "step": 110
194
+ },
195
+ {
196
+ "epoch": 0.77,
197
+ "learning_rate": 7.723433775328385e-07,
198
+ "logits/chosen": -2.4683775901794434,
199
+ "logits/rejected": -2.3485958576202393,
200
+ "logps/chosen": -196.60177612304688,
201
+ "logps/rejected": -211.6150360107422,
202
+ "loss": 0.6128,
203
+ "rewards/accuracies": 0.7093750238418579,
204
+ "rewards/chosen": -0.29994842410087585,
205
+ "rewards/margins": 0.27045875787734985,
206
+ "rewards/rejected": -0.5704071521759033,
207
+ "step": 120
208
+ },
209
+ {
210
+ "epoch": 0.83,
211
+ "learning_rate": 4.1356686569674344e-07,
212
+ "logits/chosen": -2.4233736991882324,
213
+ "logits/rejected": -2.3031530380249023,
214
+ "logps/chosen": -190.49278259277344,
215
+ "logps/rejected": -205.83090209960938,
216
+ "loss": 0.6139,
217
+ "rewards/accuracies": 0.6625000238418579,
218
+ "rewards/chosen": -0.3687855005264282,
219
+ "rewards/margins": 0.23157742619514465,
220
+ "rewards/rejected": -0.6003628969192505,
221
+ "step": 130
222
+ },
223
+ {
224
+ "epoch": 0.9,
225
+ "learning_rate": 1.59412823400657e-07,
226
+ "logits/chosen": -2.4012184143066406,
227
+ "logits/rejected": -2.203899621963501,
228
+ "logps/chosen": -198.17774963378906,
229
+ "logps/rejected": -214.18106079101562,
230
+ "loss": 0.604,
231
+ "rewards/accuracies": 0.6968749761581421,
232
+ "rewards/chosen": -0.3292800188064575,
233
+ "rewards/margins": 0.3036113381385803,
234
+ "rewards/rejected": -0.6328912973403931,
235
+ "step": 140
236
+ },
237
+ {
238
+ "epoch": 0.96,
239
+ "learning_rate": 2.262559558016325e-08,
240
+ "logits/chosen": -2.4568684101104736,
241
+ "logits/rejected": -2.2847976684570312,
242
+ "logps/chosen": -184.83816528320312,
243
+ "logps/rejected": -199.9399871826172,
244
+ "loss": 0.6024,
245
+ "rewards/accuracies": 0.734375,
246
+ "rewards/chosen": -0.33008551597595215,
247
+ "rewards/margins": 0.30454346537590027,
248
+ "rewards/rejected": -0.63462895154953,
249
+ "step": 150
250
+ },
251
+ {
252
+ "epoch": 1.0,
253
+ "step": 156,
254
+ "total_flos": 0.0,
255
+ "train_loss": 0.6301614909599988,
256
+ "train_runtime": 6946.965,
257
+ "train_samples_per_second": 2.879,
258
+ "train_steps_per_second": 0.022
259
+ }
260
+ ],
261
+ "logging_steps": 10,
262
+ "max_steps": 156,
263
+ "num_input_tokens_seen": 0,
264
+ "num_train_epochs": 1,
265
+ "save_steps": 100,
266
+ "total_flos": 0.0,
267
+ "train_batch_size": 2,
268
+ "trial_name": null,
269
+ "trial_params": null
270
+ }