Hugging Face's logo

YYYYYYibo
/
gshf_lora_iter_3

PEFT
Safetensors
mistral
alignment-handbook
Generated from Trainer
trl
dpo
Model card Files Community
YYYYYYibo commited on
Commit
9359141
·
verified ·
1 Parent(s): f91474a

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
+ 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_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
+ # gshf_lora_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: 0.6864
22
+ - Rewards/chosen: -1.0331
23
+ - Rewards/rejected: -1.1393
24
+ - Rewards/accuracies: 0.5780
25
+ - Rewards/margins: 0.1061
26
+ - Logps/rejected: -401.6631
27
+ - Logps/chosen: -412.7206
28
+ - Logits/rejected: -2.2917
29
+ - Logits/chosen: -2.4065
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
+ | 0.5977 | 0.61 | 100 | 0.6864 | -1.0331 | -1.1393 | 0.5780 | 0.1061 | -401.6631 | -412.7206 | -2.2917 | -2.4065 |
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": 0.6085586136037653,
4
+ "train_runtime": 7087.3831,
5
+ "train_samples": 21135,
6
+ "train_samples_per_second": 2.982,
7
+ "train_steps_per_second": 0.023
8
+ }
train_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "train_loss": 0.6085586136037653,
4
+ "train_runtime": 7087.3831,
5
+ "train_samples": 21135,
6
+ "train_samples_per_second": 2.982,
7
+ "train_steps_per_second": 0.023
8
+ }
trainer_state.json ADDED
@@ -0,0 +1,284 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.9992429977289932,
5
+ "eval_steps": 100,
6
+ "global_step": 165,
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": -2.707052707672119,
15
+ "logits/rejected": -2.5629069805145264,
16
+ "logps/chosen": -140.5702667236328,
17
+ "logps/rejected": -149.23953247070312,
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": -2.6915149688720703,
29
+ "logits/rejected": -2.6174051761627197,
30
+ "logps/chosen": -127.6676025390625,
31
+ "logps/rejected": -134.06060791015625,
32
+ "loss": 0.6906,
33
+ "rewards/accuracies": 0.5277777910232544,
34
+ "rewards/chosen": 0.011390692554414272,
35
+ "rewards/margins": 0.004895765800029039,
36
+ "rewards/rejected": 0.00649492722004652,
37
+ "step": 10
38
+ },
39
+ {
40
+ "epoch": 0.12,
41
+ "learning_rate": 4.994932636402032e-06,
42
+ "logits/chosen": -2.697421073913574,
43
+ "logits/rejected": -2.6101489067077637,
44
+ "logps/chosen": -119.96502685546875,
45
+ "logps/rejected": -128.2425537109375,
46
+ "loss": 0.6673,
47
+ "rewards/accuracies": 0.668749988079071,
48
+ "rewards/chosen": 0.09520833939313889,
49
+ "rewards/margins": 0.04998597502708435,
50
+ "rewards/rejected": 0.04522235691547394,
51
+ "step": 20
52
+ },
53
+ {
54
+ "epoch": 0.18,
55
+ "learning_rate": 4.905416503522124e-06,
56
+ "logits/chosen": -2.696227550506592,
57
+ "logits/rejected": -2.5642526149749756,
58
+ "logps/chosen": -150.34628295898438,
59
+ "logps/rejected": -169.03726196289062,
60
+ "loss": 0.634,
61
+ "rewards/accuracies": 0.6812499761581421,
62
+ "rewards/chosen": -0.12244757264852524,
63
+ "rewards/margins": 0.17084087431430817,
64
+ "rewards/rejected": -0.2932884693145752,
65
+ "step": 30
66
+ },
67
+ {
68
+ "epoch": 0.24,
69
+ "learning_rate": 4.707922373336524e-06,
70
+ "logits/chosen": -2.538926601409912,
71
+ "logits/rejected": -2.455259323120117,
72
+ "logps/chosen": -162.61892700195312,
73
+ "logps/rejected": -183.95986938476562,
74
+ "loss": 0.6185,
75
+ "rewards/accuracies": 0.6468750238418579,
76
+ "rewards/chosen": -0.24682824313640594,
77
+ "rewards/margins": 0.21059636771678925,
78
+ "rewards/rejected": -0.4574246406555176,
79
+ "step": 40
80
+ },
81
+ {
82
+ "epoch": 0.3,
83
+ "learning_rate": 4.411315662967732e-06,
84
+ "logits/chosen": -2.5605010986328125,
85
+ "logits/rejected": -2.4333040714263916,
86
+ "logps/chosen": -166.7479248046875,
87
+ "logps/rejected": -208.0742645263672,
88
+ "loss": 0.59,
89
+ "rewards/accuracies": 0.7093750238418579,
90
+ "rewards/chosen": -0.3265220522880554,
91
+ "rewards/margins": 0.3464210033416748,
92
+ "rewards/rejected": -0.6729430556297302,
93
+ "step": 50
94
+ },
95
+ {
96
+ "epoch": 0.36,
97
+ "learning_rate": 4.028910905897229e-06,
98
+ "logits/chosen": -2.540019989013672,
99
+ "logits/rejected": -2.388812780380249,
100
+ "logps/chosen": -173.4254913330078,
101
+ "logps/rejected": -209.1867218017578,
102
+ "loss": 0.6063,
103
+ "rewards/accuracies": 0.715624988079071,
104
+ "rewards/chosen": -0.3530834913253784,
105
+ "rewards/margins": 0.3775056302547455,
106
+ "rewards/rejected": -0.7305890917778015,
107
+ "step": 60
108
+ },
109
+ {
110
+ "epoch": 0.42,
111
+ "learning_rate": 3.577874068920446e-06,
112
+ "logits/chosen": -2.474632978439331,
113
+ "logits/rejected": -2.355051040649414,
114
+ "logps/chosen": -178.6104278564453,
115
+ "logps/rejected": -210.9853057861328,
116
+ "loss": 0.5931,
117
+ "rewards/accuracies": 0.675000011920929,
118
+ "rewards/chosen": -0.4069199562072754,
119
+ "rewards/margins": 0.3201315402984619,
120
+ "rewards/rejected": -0.7270514369010925,
121
+ "step": 70
122
+ },
123
+ {
124
+ "epoch": 0.48,
125
+ "learning_rate": 3.0784519801008546e-06,
126
+ "logits/chosen": -2.416491985321045,
127
+ "logits/rejected": -2.2601351737976074,
128
+ "logps/chosen": -188.75547790527344,
129
+ "logps/rejected": -218.71762084960938,
130
+ "loss": 0.6059,
131
+ "rewards/accuracies": 0.703125,
132
+ "rewards/chosen": -0.48314571380615234,
133
+ "rewards/margins": 0.3314744830131531,
134
+ "rewards/rejected": -0.8146201372146606,
135
+ "step": 80
136
+ },
137
+ {
138
+ "epoch": 0.55,
139
+ "learning_rate": 2.553063458334059e-06,
140
+ "logits/chosen": -2.461280107498169,
141
+ "logits/rejected": -2.3336400985717773,
142
+ "logps/chosen": -193.59939575195312,
143
+ "logps/rejected": -233.89682006835938,
144
+ "loss": 0.5949,
145
+ "rewards/accuracies": 0.6875,
146
+ "rewards/chosen": -0.5311333537101746,
147
+ "rewards/margins": 0.3859925866127014,
148
+ "rewards/rejected": -0.9171259999275208,
149
+ "step": 90
150
+ },
151
+ {
152
+ "epoch": 0.61,
153
+ "learning_rate": 2.025292943281429e-06,
154
+ "logits/chosen": -2.468966245651245,
155
+ "logits/rejected": -2.3661489486694336,
156
+ "logps/chosen": -179.36941528320312,
157
+ "logps/rejected": -225.336669921875,
158
+ "loss": 0.5977,
159
+ "rewards/accuracies": 0.715624988079071,
160
+ "rewards/chosen": -0.4190208911895752,
161
+ "rewards/margins": 0.38450565934181213,
162
+ "rewards/rejected": -0.8035264015197754,
163
+ "step": 100
164
+ },
165
+ {
166
+ "epoch": 0.61,
167
+ "eval_logits/chosen": -2.406489849090576,
168
+ "eval_logits/rejected": -2.2917134761810303,
169
+ "eval_logps/chosen": -412.7205810546875,
170
+ "eval_logps/rejected": -401.6630859375,
171
+ "eval_loss": 0.6863529086112976,
172
+ "eval_rewards/accuracies": 0.578000009059906,
173
+ "eval_rewards/chosen": -1.033116102218628,
174
+ "eval_rewards/margins": 0.10613465309143066,
175
+ "eval_rewards/rejected": -1.1392508745193481,
176
+ "eval_runtime": 381.8968,
177
+ "eval_samples_per_second": 5.237,
178
+ "eval_steps_per_second": 0.655,
179
+ "step": 100
180
+ },
181
+ {
182
+ "epoch": 0.67,
183
+ "learning_rate": 1.5188318011445907e-06,
184
+ "logits/chosen": -2.464691638946533,
185
+ "logits/rejected": -2.3220224380493164,
186
+ "logps/chosen": -182.71714782714844,
187
+ "logps/rejected": -227.18356323242188,
188
+ "loss": 0.5785,
189
+ "rewards/accuracies": 0.721875011920929,
190
+ "rewards/chosen": -0.4125642776489258,
191
+ "rewards/margins": 0.46859368681907654,
192
+ "rewards/rejected": -0.8811580538749695,
193
+ "step": 110
194
+ },
195
+ {
196
+ "epoch": 0.73,
197
+ "learning_rate": 1.0564148305586296e-06,
198
+ "logits/chosen": -2.4078032970428467,
199
+ "logits/rejected": -2.303321361541748,
200
+ "logps/chosen": -178.89767456054688,
201
+ "logps/rejected": -210.9845733642578,
202
+ "loss": 0.6082,
203
+ "rewards/accuracies": 0.6781250238418579,
204
+ "rewards/chosen": -0.4273001551628113,
205
+ "rewards/margins": 0.29686012864112854,
206
+ "rewards/rejected": -0.7241603136062622,
207
+ "step": 120
208
+ },
209
+ {
210
+ "epoch": 0.79,
211
+ "learning_rate": 6.587997083462197e-07,
212
+ "logits/chosen": -2.425297498703003,
213
+ "logits/rejected": -2.2869625091552734,
214
+ "logps/chosen": -173.35562133789062,
215
+ "logps/rejected": -225.9993133544922,
216
+ "loss": 0.5929,
217
+ "rewards/accuracies": 0.734375,
218
+ "rewards/chosen": -0.3401731550693512,
219
+ "rewards/margins": 0.4647199511528015,
220
+ "rewards/rejected": -0.8048931360244751,
221
+ "step": 130
222
+ },
223
+ {
224
+ "epoch": 0.85,
225
+ "learning_rate": 3.438351873250492e-07,
226
+ "logits/chosen": -2.324622631072998,
227
+ "logits/rejected": -2.229893922805786,
228
+ "logps/chosen": -169.2745819091797,
229
+ "logps/rejected": -208.8419647216797,
230
+ "loss": 0.5979,
231
+ "rewards/accuracies": 0.6937500238418579,
232
+ "rewards/chosen": -0.3650784194469452,
233
+ "rewards/margins": 0.3623533844947815,
234
+ "rewards/rejected": -0.7274317145347595,
235
+ "step": 140
236
+ },
237
+ {
238
+ "epoch": 0.91,
239
+ "learning_rate": 1.2565987432367032e-07,
240
+ "logits/chosen": -2.3950939178466797,
241
+ "logits/rejected": -2.2864699363708496,
242
+ "logps/chosen": -173.4356689453125,
243
+ "logps/rejected": -218.30789184570312,
244
+ "loss": 0.5972,
245
+ "rewards/accuracies": 0.7093750238418579,
246
+ "rewards/chosen": -0.38028424978256226,
247
+ "rewards/margins": 0.3941970765590668,
248
+ "rewards/rejected": -0.7744813561439514,
249
+ "step": 150
250
+ },
251
+ {
252
+ "epoch": 0.97,
253
+ "learning_rate": 1.4067554877743861e-08,
254
+ "logits/chosen": -2.344088554382324,
255
+ "logits/rejected": -2.264390468597412,
256
+ "logps/chosen": -165.1439208984375,
257
+ "logps/rejected": -209.009765625,
258
+ "loss": 0.592,
259
+ "rewards/accuracies": 0.7437499761581421,
260
+ "rewards/chosen": -0.360766738653183,
261
+ "rewards/margins": 0.41205891966819763,
262
+ "rewards/rejected": -0.7728255987167358,
263
+ "step": 160
264
+ },
265
+ {
266
+ "epoch": 1.0,
267
+ "step": 165,
268
+ "total_flos": 0.0,
269
+ "train_loss": 0.6085586136037653,
270
+ "train_runtime": 7087.3831,
271
+ "train_samples_per_second": 2.982,
272
+ "train_steps_per_second": 0.023
273
+ }
274
+ ],
275
+ "logging_steps": 10,
276
+ "max_steps": 165,
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
+ }