Hugging Face's logo

chansung
/
gemma-7b-sft-qlora-1

PEFT
TensorBoard
Safetensors
gemma
alignment-handbook
trl
sft
Generated from Trainer
4-bit precision
bitsandbytes
Model card Files Metrics Community
chansung commited on
Commit
4b5330a
·
verified ·
1 Parent(s): e2d2df9

Model save

Browse files
Files changed (6) hide show
  1. README.md +91 -0
  2. adapter_model.safetensors +1 -1
  3. all_results.json +8 -0
  4. runs/Apr11_22-29-34_deep-diver-main-blue-cougar-1-0-0/events.out.tfevents.1712889077.deep-diver-main-blue-cougar-1-0-0.520.0 +2 -2
  5. train_results.json +8 -0
  6. trainer_state.json +396 -0
README.md ADDED
@@ -0,0 +1,91 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: gemma
3
+ library_name: peft
4
+ tags:
5
+ - trl
6
+ - sft
7
+ - generated_from_trainer
8
+ base_model: google/gemma-7b
9
+ datasets:
10
+ - generator
11
+ model-index:
12
+ - name: gemma-7b-sft-qlora-1
13
+ results: []
14
+ ---
15
+
16
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
17
+ should probably proofread and complete it, then remove this comment. -->
18
+
19
+ # gemma-7b-sft-qlora-1
20
+
21
+ This model is a fine-tuned version of [google/gemma-7b](https://huggingface.co/google/gemma-7b) on the generator dataset.
22
+ It achieves the following results on the evaluation set:
23
+ - Loss: 2.2095
24
+
25
+ ## Model description
26
+
27
+ More information needed
28
+
29
+ ## Intended uses & limitations
30
+
31
+ More information needed
32
+
33
+ ## Training and evaluation data
34
+
35
+ More information needed
36
+
37
+ ## Training procedure
38
+
39
+ ### Training hyperparameters
40
+
41
+ The following hyperparameters were used during training:
42
+ - learning_rate: 0.0002
43
+ - train_batch_size: 2
44
+ - eval_batch_size: 2
45
+ - seed: 42
46
+ - distributed_type: multi-GPU
47
+ - num_devices: 4
48
+ - gradient_accumulation_steps: 2
49
+ - total_train_batch_size: 16
50
+ - total_eval_batch_size: 8
51
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
52
+ - lr_scheduler_type: cosine
53
+ - lr_scheduler_warmup_ratio: 0.1
54
+ - num_epochs: 25
55
+
56
+ ### Training results
57
+
58
+ | Training Loss | Epoch | Step | Validation Loss |
59
+ |:-------------:|:-----:|:----:|:---------------:|
60
+ | 23.6212 | 0.91 | 5 | 8.0020 |
61
+ | 14.6688 | 2.0 | 11 | 6.8099 |
62
+ | 10.8277 | 2.91 | 16 | 6.4585 |
63
+ | 10.965 | 4.0 | 22 | 5.2759 |
64
+ | 8.3233 | 4.91 | 27 | 1.6939 |
65
+ | 2.2795 | 6.0 | 33 | 1.4540 |
66
+ | 1.5047 | 6.91 | 38 | 1.3612 |
67
+ | 1.3243 | 8.0 | 44 | 1.2886 |
68
+ | 1.1264 | 8.91 | 49 | 1.2783 |
69
+ | 0.9122 | 10.0 | 55 | 1.2740 |
70
+ | 0.8184 | 10.91 | 60 | 1.2854 |
71
+ | 0.6918 | 12.0 | 66 | 1.3135 |
72
+ | 0.6194 | 12.91 | 71 | 1.3431 |
73
+ | 0.5176 | 14.0 | 77 | 1.4737 |
74
+ | 0.4514 | 14.91 | 82 | 1.7112 |
75
+ | 0.3759 | 16.0 | 88 | 1.8429 |
76
+ | 0.3464 | 16.91 | 93 | 1.8994 |
77
+ | 0.2681 | 18.0 | 99 | 1.9583 |
78
+ | 0.2487 | 18.91 | 104 | 2.1623 |
79
+ | 0.2122 | 20.0 | 110 | 2.2136 |
80
+ | 0.2036 | 20.91 | 115 | 2.2150 |
81
+ | 0.2098 | 22.0 | 121 | 2.2189 |
82
+ | 0.1955 | 22.73 | 125 | 2.2095 |
83
+
84
+
85
+ ### Framework versions
86
+
87
+ - PEFT 0.7.1
88
+ - Transformers 4.39.3
89
+ - Pytorch 2.2.2+cu121
90
+ - Datasets 2.18.0
91
+ - Tokenizers 0.15.2
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b2d18eaf6448fefa1db82c12822de8961d66f6d5c272f70d78eb59746888ac07
3
  size 100060536
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5df096199927ddbdde200b1c71b8c291651dc6e208429f57711ce6ceef6530b2
3
  size 100060536
all_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 22.73,
3
+ "train_loss": 3.2900945229530336,
4
+ "train_runtime": 472.0874,
5
+ "train_samples": 926,
6
+ "train_samples_per_second": 4.66,
7
+ "train_steps_per_second": 0.265
8
+ }
runs/Apr11_22-29-34_deep-diver-main-blue-cougar-1-0-0/events.out.tfevents.1712889077.deep-diver-main-blue-cougar-1-0-0.520.0 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:30d93495b768e37538b10fcb4c6dc2770d5b599d274a53238f5cc0176024b5c5
3
- size 14483
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:43992c407a6a78bf211428c86efcd2a5c2896815831aeeb05f20acc0153f6a58
3
+ size 17196
train_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 22.73,
3
+ "train_loss": 3.2900945229530336,
4
+ "train_runtime": 472.0874,
5
+ "train_samples": 926,
6
+ "train_samples_per_second": 4.66,
7
+ "train_steps_per_second": 0.265
8
+ }
trainer_state.json ADDED
@@ -0,0 +1,396 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 22.727272727272727,
5
+ "eval_steps": 500,
6
+ "global_step": 125,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.18,
13
+ "grad_norm": 220.0,
14
+ "learning_rate": 1.5384615384615387e-05,
15
+ "loss": 24.9691,
16
+ "step": 1
17
+ },
18
+ {
19
+ "epoch": 0.91,
20
+ "grad_norm": 90.0,
21
+ "learning_rate": 7.692307692307693e-05,
22
+ "loss": 23.6212,
23
+ "step": 5
24
+ },
25
+ {
26
+ "epoch": 0.91,
27
+ "eval_loss": 8.00199031829834,
28
+ "eval_runtime": 0.5608,
29
+ "eval_samples_per_second": 3.566,
30
+ "eval_steps_per_second": 1.783,
31
+ "step": 5
32
+ },
33
+ {
34
+ "epoch": 1.82,
35
+ "grad_norm": 14.8125,
36
+ "learning_rate": 0.00015384615384615385,
37
+ "loss": 14.6688,
38
+ "step": 10
39
+ },
40
+ {
41
+ "epoch": 2.0,
42
+ "eval_loss": 6.809861183166504,
43
+ "eval_runtime": 0.5624,
44
+ "eval_samples_per_second": 3.556,
45
+ "eval_steps_per_second": 1.778,
46
+ "step": 11
47
+ },
48
+ {
49
+ "epoch": 2.73,
50
+ "grad_norm": 5.625,
51
+ "learning_rate": 0.00019984268150178167,
52
+ "loss": 10.8277,
53
+ "step": 15
54
+ },
55
+ {
56
+ "epoch": 2.91,
57
+ "eval_loss": 6.458514213562012,
58
+ "eval_runtime": 0.5773,
59
+ "eval_samples_per_second": 3.465,
60
+ "eval_steps_per_second": 1.732,
61
+ "step": 16
62
+ },
63
+ {
64
+ "epoch": 3.64,
65
+ "grad_norm": 7.65625,
66
+ "learning_rate": 0.00019807852804032305,
67
+ "loss": 10.965,
68
+ "step": 20
69
+ },
70
+ {
71
+ "epoch": 4.0,
72
+ "eval_loss": 5.275852203369141,
73
+ "eval_runtime": 0.5667,
74
+ "eval_samples_per_second": 3.529,
75
+ "eval_steps_per_second": 1.765,
76
+ "step": 22
77
+ },
78
+ {
79
+ "epoch": 4.55,
80
+ "grad_norm": 32.75,
81
+ "learning_rate": 0.00019438833303083678,
82
+ "loss": 8.3233,
83
+ "step": 25
84
+ },
85
+ {
86
+ "epoch": 4.91,
87
+ "eval_loss": 1.6938855648040771,
88
+ "eval_runtime": 0.5862,
89
+ "eval_samples_per_second": 3.412,
90
+ "eval_steps_per_second": 1.706,
91
+ "step": 27
92
+ },
93
+ {
94
+ "epoch": 5.45,
95
+ "grad_norm": 2.890625,
96
+ "learning_rate": 0.00018884456359788724,
97
+ "loss": 2.2795,
98
+ "step": 30
99
+ },
100
+ {
101
+ "epoch": 6.0,
102
+ "eval_loss": 1.4539892673492432,
103
+ "eval_runtime": 0.5675,
104
+ "eval_samples_per_second": 3.524,
105
+ "eval_steps_per_second": 1.762,
106
+ "step": 33
107
+ },
108
+ {
109
+ "epoch": 6.36,
110
+ "grad_norm": 35.25,
111
+ "learning_rate": 0.00018155608689592604,
112
+ "loss": 1.5047,
113
+ "step": 35
114
+ },
115
+ {
116
+ "epoch": 6.91,
117
+ "eval_loss": 1.3612087965011597,
118
+ "eval_runtime": 0.5882,
119
+ "eval_samples_per_second": 3.4,
120
+ "eval_steps_per_second": 1.7,
121
+ "step": 38
122
+ },
123
+ {
124
+ "epoch": 7.27,
125
+ "grad_norm": 1.65625,
126
+ "learning_rate": 0.0001726660322034027,
127
+ "loss": 1.3243,
128
+ "step": 40
129
+ },
130
+ {
131
+ "epoch": 8.0,
132
+ "eval_loss": 1.288640022277832,
133
+ "eval_runtime": 0.5698,
134
+ "eval_samples_per_second": 3.51,
135
+ "eval_steps_per_second": 1.755,
136
+ "step": 44
137
+ },
138
+ {
139
+ "epoch": 8.18,
140
+ "grad_norm": 1.015625,
141
+ "learning_rate": 0.00016234898018587337,
142
+ "loss": 1.1264,
143
+ "step": 45
144
+ },
145
+ {
146
+ "epoch": 8.91,
147
+ "eval_loss": 1.2782788276672363,
148
+ "eval_runtime": 0.59,
149
+ "eval_samples_per_second": 3.39,
150
+ "eval_steps_per_second": 1.695,
151
+ "step": 49
152
+ },
153
+ {
154
+ "epoch": 9.09,
155
+ "grad_norm": 2.1875,
156
+ "learning_rate": 0.00015080753452465296,
157
+ "loss": 1.0387,
158
+ "step": 50
159
+ },
160
+ {
161
+ "epoch": 10.0,
162
+ "grad_norm": 0.9765625,
163
+ "learning_rate": 0.000138268343236509,
164
+ "loss": 0.9122,
165
+ "step": 55
166
+ },
167
+ {
168
+ "epoch": 10.0,
169
+ "eval_loss": 1.2739648818969727,
170
+ "eval_runtime": 0.5659,
171
+ "eval_samples_per_second": 3.534,
172
+ "eval_steps_per_second": 1.767,
173
+ "step": 55
174
+ },
175
+ {
176
+ "epoch": 10.91,
177
+ "grad_norm": 0.6953125,
178
+ "learning_rate": 0.0001249776478167227,
179
+ "loss": 0.8184,
180
+ "step": 60
181
+ },
182
+ {
183
+ "epoch": 10.91,
184
+ "eval_loss": 1.2853541374206543,
185
+ "eval_runtime": 0.5892,
186
+ "eval_samples_per_second": 3.395,
187
+ "eval_steps_per_second": 1.697,
188
+ "step": 60
189
+ },
190
+ {
191
+ "epoch": 11.82,
192
+ "grad_norm": 0.96875,
193
+ "learning_rate": 0.00011119644761033078,
194
+ "loss": 0.6918,
195
+ "step": 65
196
+ },
197
+ {
198
+ "epoch": 12.0,
199
+ "eval_loss": 1.31352698802948,
200
+ "eval_runtime": 0.5674,
201
+ "eval_samples_per_second": 3.525,
202
+ "eval_steps_per_second": 1.762,
203
+ "step": 66
204
+ },
205
+ {
206
+ "epoch": 12.73,
207
+ "grad_norm": 0.92578125,
208
+ "learning_rate": 9.719537437241312e-05,
209
+ "loss": 0.6194,
210
+ "step": 70
211
+ },
212
+ {
213
+ "epoch": 12.91,
214
+ "eval_loss": 1.343058705329895,
215
+ "eval_runtime": 0.7364,
216
+ "eval_samples_per_second": 2.716,
217
+ "eval_steps_per_second": 1.358,
218
+ "step": 71
219
+ },
220
+ {
221
+ "epoch": 13.64,
222
+ "grad_norm": 1.1875,
223
+ "learning_rate": 8.324937766952638e-05,
224
+ "loss": 0.5176,
225
+ "step": 75
226
+ },
227
+ {
228
+ "epoch": 14.0,
229
+ "eval_loss": 1.4736580848693848,
230
+ "eval_runtime": 0.5693,
231
+ "eval_samples_per_second": 3.513,
232
+ "eval_steps_per_second": 1.757,
233
+ "step": 77
234
+ },
235
+ {
236
+ "epoch": 14.55,
237
+ "grad_norm": 0.9296875,
238
+ "learning_rate": 6.963232548903853e-05,
239
+ "loss": 0.4514,
240
+ "step": 80
241
+ },
242
+ {
243
+ "epoch": 14.91,
244
+ "eval_loss": 1.7112184762954712,
245
+ "eval_runtime": 0.6242,
246
+ "eval_samples_per_second": 3.204,
247
+ "eval_steps_per_second": 1.602,
248
+ "step": 82
249
+ },
250
+ {
251
+ "epoch": 15.45,
252
+ "grad_norm": 1.34375,
253
+ "learning_rate": 5.6611626088244194e-05,
254
+ "loss": 0.3759,
255
+ "step": 85
256
+ },
257
+ {
258
+ "epoch": 16.0,
259
+ "eval_loss": 1.8429330587387085,
260
+ "eval_runtime": 0.5694,
261
+ "eval_samples_per_second": 3.513,
262
+ "eval_steps_per_second": 1.756,
263
+ "step": 88
264
+ },
265
+ {
266
+ "epoch": 16.36,
267
+ "grad_norm": 0.84375,
268
+ "learning_rate": 4.444297669803981e-05,
269
+ "loss": 0.3464,
270
+ "step": 90
271
+ },
272
+ {
273
+ "epoch": 16.91,
274
+ "eval_loss": 1.899384617805481,
275
+ "eval_runtime": 0.5863,
276
+ "eval_samples_per_second": 3.411,
277
+ "eval_steps_per_second": 1.706,
278
+ "step": 93
279
+ },
280
+ {
281
+ "epoch": 17.27,
282
+ "grad_norm": 0.8203125,
283
+ "learning_rate": 3.336534220479961e-05,
284
+ "loss": 0.2681,
285
+ "step": 95
286
+ },
287
+ {
288
+ "epoch": 18.0,
289
+ "eval_loss": 1.9583137035369873,
290
+ "eval_runtime": 0.5683,
291
+ "eval_samples_per_second": 3.519,
292
+ "eval_steps_per_second": 1.76,
293
+ "step": 99
294
+ },
295
+ {
296
+ "epoch": 18.18,
297
+ "grad_norm": 0.73828125,
298
+ "learning_rate": 2.3596262417839255e-05,
299
+ "loss": 0.2487,
300
+ "step": 100
301
+ },
302
+ {
303
+ "epoch": 18.91,
304
+ "eval_loss": 2.162316083908081,
305
+ "eval_runtime": 0.5845,
306
+ "eval_samples_per_second": 3.422,
307
+ "eval_steps_per_second": 1.711,
308
+ "step": 104
309
+ },
310
+ {
311
+ "epoch": 19.09,
312
+ "grad_norm": 0.875,
313
+ "learning_rate": 1.5327580077171587e-05,
314
+ "loss": 0.2322,
315
+ "step": 105
316
+ },
317
+ {
318
+ "epoch": 20.0,
319
+ "grad_norm": 0.6484375,
320
+ "learning_rate": 8.72167349386811e-06,
321
+ "loss": 0.2122,
322
+ "step": 110
323
+ },
324
+ {
325
+ "epoch": 20.0,
326
+ "eval_loss": 2.213620901107788,
327
+ "eval_runtime": 0.5651,
328
+ "eval_samples_per_second": 3.539,
329
+ "eval_steps_per_second": 1.769,
330
+ "step": 110
331
+ },
332
+ {
333
+ "epoch": 20.91,
334
+ "grad_norm": 0.50390625,
335
+ "learning_rate": 3.908267805490051e-06,
336
+ "loss": 0.2036,
337
+ "step": 115
338
+ },
339
+ {
340
+ "epoch": 20.91,
341
+ "eval_loss": 2.2149863243103027,
342
+ "eval_runtime": 0.5812,
343
+ "eval_samples_per_second": 3.441,
344
+ "eval_steps_per_second": 1.721,
345
+ "step": 115
346
+ },
347
+ {
348
+ "epoch": 21.82,
349
+ "grad_norm": 0.458984375,
350
+ "learning_rate": 9.818874663554357e-07,
351
+ "loss": 0.2098,
352
+ "step": 120
353
+ },
354
+ {
355
+ "epoch": 22.0,
356
+ "eval_loss": 2.2188880443573,
357
+ "eval_runtime": 0.5673,
358
+ "eval_samples_per_second": 3.525,
359
+ "eval_steps_per_second": 1.763,
360
+ "step": 121
361
+ },
362
+ {
363
+ "epoch": 22.73,
364
+ "grad_norm": 0.6171875,
365
+ "learning_rate": 0.0,
366
+ "loss": 0.1955,
367
+ "step": 125
368
+ },
369
+ {
370
+ "epoch": 22.73,
371
+ "eval_loss": 2.209489345550537,
372
+ "eval_runtime": 0.5657,
373
+ "eval_samples_per_second": 3.536,
374
+ "eval_steps_per_second": 1.768,
375
+ "step": 125
376
+ },
377
+ {
378
+ "epoch": 22.73,
379
+ "step": 125,
380
+ "total_flos": 1.917235948819579e+17,
381
+ "train_loss": 3.2900945229530336,
382
+ "train_runtime": 472.0874,
383
+ "train_samples_per_second": 4.66,
384
+ "train_steps_per_second": 0.265
385
+ }
386
+ ],
387
+ "logging_steps": 5,
388
+ "max_steps": 125,
389
+ "num_input_tokens_seen": 0,
390
+ "num_train_epochs": 25,
391
+ "save_steps": 100,
392
+ "total_flos": 1.917235948819579e+17,
393
+ "train_batch_size": 2,
394
+ "trial_name": null,
395
+ "trial_params": null
396
+ }