Hugging Face's logo

ilharco
/
altruistic-zephyr-7b-sft-qlora

PEFT
TensorBoard
Safetensors
mistral
alignment-handbook
Generated from Trainer
trl
dpo
4-bit precision
bitsandbytes
Model card Files
xet
Metrics Community
ilharco commited on
Commit
df97976
·
verified ·
1 Parent(s): 0efdd62

Model save

Browse files
Files changed (8) hide show
  1. README.md +73 -0
  2. adapter_model.safetensors +1 -1
  3. all_results.json +21 -0
  4. eval_results.json +16 -0
  5. runs/Feb12_23-07-01_debian/events.out.tfevents.1707775659.debian.74814.0 +2 -2
  6. runs/Feb12_23-07-01_debian/events.out.tfevents.1707777120.debian.74814.1 +3 -0
  7. train_results.json +8 -0
  8. trainer_state.json +326 -0
README.md ADDED
@@ -0,0 +1,73 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ library_name: peft
4
+ tags:
5
+ - trl
6
+ - dpo
7
+ - generated_from_trainer
8
+ base_model: mistralai/Mistral-7B-v0.1
9
+ model-index:
10
+ - name: zephyr-7b-dpo-qlora
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
+ # zephyr-7b-dpo-qlora
18
+
19
+ This model is a fine-tuned version of [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) on the None dataset.
20
+ It achieves the following results on the evaluation set:
21
+ - Loss: 0.0420
22
+ - Rewards/chosen: -5.5143
23
+ - Rewards/rejected: -13.0608
24
+ - Rewards/accuracies: 0.9949
25
+ - Rewards/margins: 7.5465
26
+ - Logps/rejected: -1865.7168
27
+ - Logps/chosen: -1348.6318
28
+ - Logits/rejected: -2.6274
29
+ - Logits/chosen: -2.6531
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: 1
50
+ - eval_batch_size: 1
51
+ - seed: 42
52
+ - distributed_type: multi-GPU
53
+ - gradient_accumulation_steps: 4
54
+ - total_train_batch_size: 4
55
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
56
+ - lr_scheduler_type: cosine
57
+ - lr_scheduler_warmup_ratio: 0.1
58
+ - num_epochs: 1
59
+
60
+ ### Training results
61
+
62
+ | Training Loss | Epoch | Step | Validation Loss | Rewards/chosen | Rewards/rejected | Rewards/accuracies | Rewards/margins | Logps/rejected | Logps/chosen | Logits/rejected | Logits/chosen |
63
+ |:-------------:|:-----:|:----:|:---------------:|:--------------:|:----------------:|:------------------:|:---------------:|:--------------:|:------------:|:---------------:|:-------------:|
64
+ | 0.1963 | 0.51 | 100 | 0.1328 | -1.0972 | -3.5537 | 1.0 | 2.4565 | -915.0070 | -906.9143 | -2.8398 | -2.8256 |
65
+
66
+
67
+ ### Framework versions
68
+
69
+ - PEFT 0.7.1
70
+ - Transformers 4.36.2
71
+ - Pytorch 2.1.2
72
+ - Datasets 2.14.6
73
+ - Tokenizers 0.15.2
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:4856bbac36de6133d8455822de6f8c48cd1529126083f4d8460d31ba45e6e5aa
3
  size 671150064
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d338e451507babbf5350fe3946b82561509286790fbbfe748acd297ddde137cc
3
  size 671150064
all_results.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "eval_logits/chosen": -2.6530776023864746,
4
+ "eval_logits/rejected": -2.627399444580078,
5
+ "eval_logps/chosen": -1348.6318359375,
6
+ "eval_logps/rejected": -1865.716796875,
7
+ "eval_loss": 0.041974812746047974,
8
+ "eval_rewards/accuracies": 0.9949495196342468,
9
+ "eval_rewards/chosen": -5.514344692230225,
10
+ "eval_rewards/margins": 7.546464920043945,
11
+ "eval_rewards/rejected": -13.060811042785645,
12
+ "eval_runtime": 138.5776,
13
+ "eval_samples": 198,
14
+ "eval_samples_per_second": 1.429,
15
+ "eval_steps_per_second": 1.429,
16
+ "train_loss": 0.24469107510474733,
17
+ "train_runtime": 1322.9236,
18
+ "train_samples": 791,
19
+ "train_samples_per_second": 0.598,
20
+ "train_steps_per_second": 0.149
21
+ }
eval_results.json ADDED
@@ -0,0 +1,16 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "eval_logits/chosen": -2.6530776023864746,
4
+ "eval_logits/rejected": -2.627399444580078,
5
+ "eval_logps/chosen": -1348.6318359375,
6
+ "eval_logps/rejected": -1865.716796875,
7
+ "eval_loss": 0.041974812746047974,
8
+ "eval_rewards/accuracies": 0.9949495196342468,
9
+ "eval_rewards/chosen": -5.514344692230225,
10
+ "eval_rewards/margins": 7.546464920043945,
11
+ "eval_rewards/rejected": -13.060811042785645,
12
+ "eval_runtime": 138.5776,
13
+ "eval_samples": 198,
14
+ "eval_samples_per_second": 1.429,
15
+ "eval_steps_per_second": 1.429
16
+ }
runs/Feb12_23-07-01_debian/events.out.tfevents.1707775659.debian.74814.0 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6bfbbfbdc523014b528d282c159b95eca81a763a3814912cb2090ca0625d0027
3
- size 12532
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34c3e0e8a7f098202f7e3ca19c50c400df8c41f5ca6925d3af66c28fef761234
3
+ size 18570
runs/Feb12_23-07-01_debian/events.out.tfevents.1707777120.debian.74814.1 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ee613b8242112671d4a16bad499ca2324e8370908b84a82f4f0b399da34d1c5e
3
+ size 828
train_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 1.0,
3
+ "train_loss": 0.24469107510474733,
4
+ "train_runtime": 1322.9236,
5
+ "train_samples": 791,
6
+ "train_samples_per_second": 0.598,
7
+ "train_steps_per_second": 0.149
8
+ }
trainer_state.json ADDED
@@ -0,0 +1,326 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.9962073324905183,
5
+ "eval_steps": 100,
6
+ "global_step": 197,
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": 2.5000000000000004e-07,
14
+ "logits/chosen": -2.867753744125366,
15
+ "logits/rejected": -3.003779411315918,
16
+ "logps/chosen": -502.99774169921875,
17
+ "logps/rejected": -334.49993896484375,
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": 0.05,
27
+ "learning_rate": 2.5e-06,
28
+ "logits/chosen": -2.9846487045288086,
29
+ "logits/rejected": -3.0769829750061035,
30
+ "logps/chosen": -746.551025390625,
31
+ "logps/rejected": -522.1072998046875,
32
+ "loss": 0.6885,
33
+ "rewards/accuracies": 0.7222222089767456,
34
+ "rewards/chosen": 0.011263555847108364,
35
+ "rewards/margins": 0.009397478774189949,
36
+ "rewards/rejected": 0.0018660753266885877,
37
+ "step": 10
38
+ },
39
+ {
40
+ "epoch": 0.1,
41
+ "learning_rate": 5e-06,
42
+ "logits/chosen": -2.9451098442077637,
43
+ "logits/rejected": -3.023451805114746,
44
+ "logps/chosen": -828.2189331054688,
45
+ "logps/rejected": -592.4244384765625,
46
+ "loss": 0.6355,
47
+ "rewards/accuracies": 0.9750000238418579,
48
+ "rewards/chosen": 0.11439962685108185,
49
+ "rewards/margins": 0.1205785870552063,
50
+ "rewards/rejected": -0.006178960669785738,
51
+ "step": 20
52
+ },
53
+ {
54
+ "epoch": 0.15,
55
+ "learning_rate": 4.960724403357315e-06,
56
+ "logits/chosen": -2.925408124923706,
57
+ "logits/rejected": -2.9945900440216064,
58
+ "logps/chosen": -789.0283203125,
59
+ "logps/rejected": -608.7799072265625,
60
+ "loss": 0.5514,
61
+ "rewards/accuracies": 1.0,
62
+ "rewards/chosen": 0.2579498291015625,
63
+ "rewards/margins": 0.31374359130859375,
64
+ "rewards/rejected": -0.055793724954128265,
65
+ "step": 30
66
+ },
67
+ {
68
+ "epoch": 0.2,
69
+ "learning_rate": 4.84413167142257e-06,
70
+ "logits/chosen": -2.9325766563415527,
71
+ "logits/rejected": -2.995227813720703,
72
+ "logps/chosen": -765.6848754882812,
73
+ "logps/rejected": -547.8126220703125,
74
+ "loss": 0.4868,
75
+ "rewards/accuracies": 0.9750000238418579,
76
+ "rewards/chosen": 0.3068661689758301,
77
+ "rewards/margins": 0.48913106322288513,
78
+ "rewards/rejected": -0.18226495385169983,
79
+ "step": 40
80
+ },
81
+ {
82
+ "epoch": 0.25,
83
+ "learning_rate": 4.653885203484516e-06,
84
+ "logits/chosen": -2.9359793663024902,
85
+ "logits/rejected": -3.0224156379699707,
86
+ "logps/chosen": -789.9358520507812,
87
+ "logps/rejected": -607.2333984375,
88
+ "loss": 0.4211,
89
+ "rewards/accuracies": 1.0,
90
+ "rewards/chosen": 0.32125765085220337,
91
+ "rewards/margins": 0.6743311285972595,
92
+ "rewards/rejected": -0.35307347774505615,
93
+ "step": 50
94
+ },
95
+ {
96
+ "epoch": 0.3,
97
+ "learning_rate": 4.3959626343730965e-06,
98
+ "logits/chosen": -2.9203407764434814,
99
+ "logits/rejected": -2.9182143211364746,
100
+ "logps/chosen": -825.3511962890625,
101
+ "logps/rejected": -596.4517211914062,
102
+ "loss": 0.3456,
103
+ "rewards/accuracies": 1.0,
104
+ "rewards/chosen": 0.39548319578170776,
105
+ "rewards/margins": 0.9407623410224915,
106
+ "rewards/rejected": -0.5452791452407837,
107
+ "step": 60
108
+ },
109
+ {
110
+ "epoch": 0.35,
111
+ "learning_rate": 4.078468014319884e-06,
112
+ "logits/chosen": -2.883965015411377,
113
+ "logits/rejected": -2.8747031688690186,
114
+ "logps/chosen": -732.6103515625,
115
+ "logps/rejected": -626.3192138671875,
116
+ "loss": 0.3458,
117
+ "rewards/accuracies": 1.0,
118
+ "rewards/chosen": 0.10843221098184586,
119
+ "rewards/margins": 0.9775202870368958,
120
+ "rewards/rejected": -0.8690881729125977,
121
+ "step": 70
122
+ },
123
+ {
124
+ "epoch": 0.4,
125
+ "learning_rate": 3.711377175831626e-06,
126
+ "logits/chosen": -2.9415102005004883,
127
+ "logits/rejected": -2.9699349403381348,
128
+ "logps/chosen": -809.07568359375,
129
+ "logps/rejected": -704.0150756835938,
130
+ "loss": 0.2466,
131
+ "rewards/accuracies": 1.0,
132
+ "rewards/chosen": 0.03153586387634277,
133
+ "rewards/margins": 1.388339877128601,
134
+ "rewards/rejected": -1.3568041324615479,
135
+ "step": 80
136
+ },
137
+ {
138
+ "epoch": 0.46,
139
+ "learning_rate": 3.3062242882712727e-06,
140
+ "logits/chosen": -2.8811752796173096,
141
+ "logits/rejected": -2.9307639598846436,
142
+ "logps/chosen": -868.4812622070312,
143
+ "logps/rejected": -745.4024658203125,
144
+ "loss": 0.2185,
145
+ "rewards/accuracies": 1.0,
146
+ "rewards/chosen": -0.1719144582748413,
147
+ "rewards/margins": 1.6975847482681274,
148
+ "rewards/rejected": -1.8694992065429688,
149
+ "step": 90
150
+ },
151
+ {
152
+ "epoch": 0.51,
153
+ "learning_rate": 2.8757394487511766e-06,
154
+ "logits/chosen": -2.8607354164123535,
155
+ "logits/rejected": -2.861823558807373,
156
+ "logps/chosen": -892.8031005859375,
157
+ "logps/rejected": -857.4615478515625,
158
+ "loss": 0.1963,
159
+ "rewards/accuracies": 0.9750000238418579,
160
+ "rewards/chosen": -0.9989568591117859,
161
+ "rewards/margins": 1.9792629480361938,
162
+ "rewards/rejected": -2.978219509124756,
163
+ "step": 100
164
+ },
165
+ {
166
+ "epoch": 0.51,
167
+ "eval_logits/chosen": -2.8255772590637207,
168
+ "eval_logits/rejected": -2.8398427963256836,
169
+ "eval_logps/chosen": -906.914306640625,
170
+ "eval_logps/rejected": -915.0069580078125,
171
+ "eval_loss": 0.1327642947435379,
172
+ "eval_rewards/accuracies": 1.0,
173
+ "eval_rewards/chosen": -1.0971715450286865,
174
+ "eval_rewards/margins": 2.45654034614563,
175
+ "eval_rewards/rejected": -3.5537116527557373,
176
+ "eval_runtime": 138.8088,
177
+ "eval_samples_per_second": 1.426,
178
+ "eval_steps_per_second": 1.426,
179
+ "step": 100
180
+ },
181
+ {
182
+ "epoch": 0.56,
183
+ "learning_rate": 2.4334486964055635e-06,
184
+ "logits/chosen": -2.7996840476989746,
185
+ "logits/rejected": -2.765969753265381,
186
+ "logps/chosen": -926.7333984375,
187
+ "logps/rejected": -968.1886596679688,
188
+ "loss": 0.1226,
189
+ "rewards/accuracies": 0.9750000238418579,
190
+ "rewards/chosen": -1.7008222341537476,
191
+ "rewards/margins": 2.827996015548706,
192
+ "rewards/rejected": -4.528818130493164,
193
+ "step": 110
194
+ },
195
+ {
196
+ "epoch": 0.61,
197
+ "learning_rate": 1.9932490177847664e-06,
198
+ "logits/chosen": -2.7951323986053467,
199
+ "logits/rejected": -2.793102741241455,
200
+ "logps/chosen": -1156.3507080078125,
201
+ "logps/rejected": -1393.317626953125,
202
+ "loss": 0.0599,
203
+ "rewards/accuracies": 1.0,
204
+ "rewards/chosen": -3.1200733184814453,
205
+ "rewards/margins": 4.863899230957031,
206
+ "rewards/rejected": -7.983973026275635,
207
+ "step": 120
208
+ },
209
+ {
210
+ "epoch": 0.66,
211
+ "learning_rate": 1.5689716969045848e-06,
212
+ "logits/chosen": -2.722018003463745,
213
+ "logits/rejected": -2.7178285121917725,
214
+ "logps/chosen": -1150.283203125,
215
+ "logps/rejected": -1416.4925537109375,
216
+ "loss": 0.1767,
217
+ "rewards/accuracies": 0.949999988079071,
218
+ "rewards/chosen": -4.09153413772583,
219
+ "rewards/margins": 4.929880142211914,
220
+ "rewards/rejected": -9.021414756774902,
221
+ "step": 130
222
+ },
223
+ {
224
+ "epoch": 0.71,
225
+ "learning_rate": 1.1739477297006442e-06,
226
+ "logits/chosen": -2.7221176624298096,
227
+ "logits/rejected": -2.6924355030059814,
228
+ "logps/chosen": -1250.833984375,
229
+ "logps/rejected": -1684.3460693359375,
230
+ "loss": 0.0761,
231
+ "rewards/accuracies": 0.9750000238418579,
232
+ "rewards/chosen": -4.285839080810547,
233
+ "rewards/margins": 6.7211456298828125,
234
+ "rewards/rejected": -11.00698471069336,
235
+ "step": 140
236
+ },
237
+ {
238
+ "epoch": 0.76,
239
+ "learning_rate": 8.20588957773018e-07,
240
+ "logits/chosen": -2.6774404048919678,
241
+ "logits/rejected": -2.671726942062378,
242
+ "logps/chosen": -1310.638427734375,
243
+ "logps/rejected": -1781.271240234375,
244
+ "loss": 0.0435,
245
+ "rewards/accuracies": 1.0,
246
+ "rewards/chosen": -5.060001850128174,
247
+ "rewards/margins": 7.21289587020874,
248
+ "rewards/rejected": -12.272897720336914,
249
+ "step": 150
250
+ },
251
+ {
252
+ "epoch": 0.81,
253
+ "learning_rate": 5.199980823988158e-07,
254
+ "logits/chosen": -2.670541286468506,
255
+ "logits/rejected": -2.615200996398926,
256
+ "logps/chosen": -1342.9185791015625,
257
+ "logps/rejected": -1865.0504150390625,
258
+ "loss": 0.1071,
259
+ "rewards/accuracies": 0.949999988079071,
260
+ "rewards/chosen": -5.760629653930664,
261
+ "rewards/margins": 7.400970458984375,
262
+ "rewards/rejected": -13.161600112915039,
263
+ "step": 160
264
+ },
265
+ {
266
+ "epoch": 0.86,
267
+ "learning_rate": 2.8161981235857145e-07,
268
+ "logits/chosen": -2.6611950397491455,
269
+ "logits/rejected": -2.6389710903167725,
270
+ "logps/chosen": -1471.225341796875,
271
+ "logps/rejected": -1975.236083984375,
272
+ "loss": 0.0247,
273
+ "rewards/accuracies": 1.0,
274
+ "rewards/chosen": -6.465508460998535,
275
+ "rewards/margins": 7.263296604156494,
276
+ "rewards/rejected": -13.728803634643555,
277
+ "step": 170
278
+ },
279
+ {
280
+ "epoch": 0.91,
281
+ "learning_rate": 1.1294410667827022e-07,
282
+ "logits/chosen": -2.6490066051483154,
283
+ "logits/rejected": -2.6506030559539795,
284
+ "logps/chosen": -1516.554443359375,
285
+ "logps/rejected": -2156.269287109375,
286
+ "loss": 0.024,
287
+ "rewards/accuracies": 1.0,
288
+ "rewards/chosen": -7.031141757965088,
289
+ "rewards/margins": 8.472519874572754,
290
+ "rewards/rejected": -15.503662109375,
291
+ "step": 180
292
+ },
293
+ {
294
+ "epoch": 0.96,
295
+ "learning_rate": 1.9270836541688543e-08,
296
+ "logits/chosen": -2.681256055831909,
297
+ "logits/rejected": -2.657177209854126,
298
+ "logps/chosen": -1404.041015625,
299
+ "logps/rejected": -1850.4908447265625,
300
+ "loss": 0.0397,
301
+ "rewards/accuracies": 1.0,
302
+ "rewards/chosen": -6.127170562744141,
303
+ "rewards/margins": 6.76980447769165,
304
+ "rewards/rejected": -12.896974563598633,
305
+ "step": 190
306
+ },
307
+ {
308
+ "epoch": 1.0,
309
+ "step": 197,
310
+ "total_flos": 0.0,
311
+ "train_loss": 0.24469107510474733,
312
+ "train_runtime": 1322.9236,
313
+ "train_samples_per_second": 0.598,
314
+ "train_steps_per_second": 0.149
315
+ }
316
+ ],
317
+ "logging_steps": 10,
318
+ "max_steps": 197,
319
+ "num_input_tokens_seen": 0,
320
+ "num_train_epochs": 1,
321
+ "save_steps": 100,
322
+ "total_flos": 0.0,
323
+ "train_batch_size": 1,
324
+ "trial_name": null,
325
+ "trial_params": null
326
+ }