Hugging Face's logo

GaetanMichelet
/
Gemma-2-2B_task-3_120-samples_config-2

PEFT
TensorBoard
Safetensors
gemma2
alignment-handbook
trl
sft
Generated from Trainer
4-bit precision
bitsandbytes
Model card Files Metrics Community
GaetanMichelet commited on
Commit
dc4f89b
•
1 Parent(s): dd3fa94

Model save

Browse files
Files changed (5) hide show
  1. README.md +79 -0
  2. adapter_model.safetensors +1 -1
  3. all_results.json +9 -0
  4. train_results.json +9 -0
  5. trainer_state.json +465 -0
README.md ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: google/gemma-2-2b-it
3
+ library_name: peft
4
+ license: gemma
5
+ tags:
6
+ - trl
7
+ - sft
8
+ - generated_from_trainer
9
+ model-index:
10
+ - name: Gemma-2-2B_task-3_120-samples_config-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
+ # Gemma-2-2B_task-3_120-samples_config-2
18
+
19
+ This model is a fine-tuned version of [google/gemma-2-2b-it](https://huggingface.co/google/gemma-2-2b-it) on the None dataset.
20
+ It achieves the following results on the evaluation set:
21
+ - Loss: 0.8741
22
+
23
+ ## Model description
24
+
25
+ More information needed
26
+
27
+ ## Intended uses & limitations
28
+
29
+ More information needed
30
+
31
+ ## Training and evaluation data
32
+
33
+ More information needed
34
+
35
+ ## Training procedure
36
+
37
+ ### Training hyperparameters
38
+
39
+ The following hyperparameters were used during training:
40
+ - learning_rate: 0.0001
41
+ - train_batch_size: 1
42
+ - eval_batch_size: 1
43
+ - seed: 42
44
+ - distributed_type: multi-GPU
45
+ - gradient_accumulation_steps: 16
46
+ - total_train_batch_size: 16
47
+ - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
48
+ - lr_scheduler_type: cosine
49
+ - lr_scheduler_warmup_ratio: 0.1
50
+ - num_epochs: 50
51
+
52
+ ### Training results
53
+
54
+ | Training Loss | Epoch | Step | Validation Loss |
55
+ |:-------------:|:-------:|:----:|:---------------:|
56
+ | 3.2814 | 0.9091 | 5 | 3.2207 |
57
+ | 2.3853 | 2.0 | 11 | 1.7735 |
58
+ | 1.0517 | 2.9091 | 16 | 0.9646 |
59
+ | 0.6575 | 4.0 | 22 | 0.6096 |
60
+ | 0.4492 | 4.9091 | 27 | 0.5386 |
61
+ | 0.3535 | 6.0 | 33 | 0.4848 |
62
+ | 0.3521 | 6.9091 | 38 | 0.4613 |
63
+ | 0.2821 | 8.0 | 44 | 0.4495 |
64
+ | 0.1878 | 8.9091 | 49 | 0.4496 |
65
+ | 0.1547 | 10.0 | 55 | 0.4928 |
66
+ | 0.1539 | 10.9091 | 60 | 0.4977 |
67
+ | 0.1194 | 12.0 | 66 | 0.6056 |
68
+ | 0.0872 | 12.9091 | 71 | 0.6938 |
69
+ | 0.0274 | 14.0 | 77 | 0.8379 |
70
+ | 0.0195 | 14.9091 | 82 | 0.8741 |
71
+
72
+
73
+ ### Framework versions
74
+
75
+ - PEFT 0.12.0
76
+ - Transformers 4.44.0
77
+ - Pytorch 2.1.2+cu121
78
+ - Datasets 2.20.0
79
+ - Tokenizers 0.19.1
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b669b798b86a8ea6613c9a6be58c0ba2a6009c76e87069b0eadb608c6d062059
3
  size 83115256
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f0e162e7d37906f1db36d1563d0f5bce0d2b11de267818eb9c8a8f7306a66ce
3
  size 83115256
all_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 14.909090909090908,
3
+ "total_flos": 3.5471186917851136e+16,
4
+ "train_loss": 0.693681104949153,
5
+ "train_runtime": 1505.781,
6
+ "train_samples": 88,
7
+ "train_samples_per_second": 2.922,
8
+ "train_steps_per_second": 0.166
9
+ }
train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 14.909090909090908,
3
+ "total_flos": 3.5471186917851136e+16,
4
+ "train_loss": 0.693681104949153,
5
+ "train_runtime": 1505.781,
6
+ "train_samples": 88,
7
+ "train_samples_per_second": 2.922,
8
+ "train_steps_per_second": 0.166
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,465 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.4495352804660797,
3
+ "best_model_checkpoint": "data/Gemma-2-2B_task-3_120-samples_config-2/checkpoint-44",
4
+ "epoch": 14.909090909090908,
5
+ "eval_steps": 500,
6
+ "global_step": 82,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.18181818181818182,
13
+ "grad_norm": 8.376895904541016,
14
+ "learning_rate": 4.000000000000001e-06,
15
+ "loss": 3.9078,
16
+ "step": 1
17
+ },
18
+ {
19
+ "epoch": 0.36363636363636365,
20
+ "grad_norm": 6.967246055603027,
21
+ "learning_rate": 8.000000000000001e-06,
22
+ "loss": 3.4781,
23
+ "step": 2
24
+ },
25
+ {
26
+ "epoch": 0.7272727272727273,
27
+ "grad_norm": 6.186585426330566,
28
+ "learning_rate": 1.6000000000000003e-05,
29
+ "loss": 3.2814,
30
+ "step": 4
31
+ },
32
+ {
33
+ "epoch": 0.9090909090909091,
34
+ "eval_loss": 3.220660924911499,
35
+ "eval_runtime": 7.6385,
36
+ "eval_samples_per_second": 3.142,
37
+ "eval_steps_per_second": 3.142,
38
+ "step": 5
39
+ },
40
+ {
41
+ "epoch": 1.0909090909090908,
42
+ "grad_norm": 6.225273132324219,
43
+ "learning_rate": 2.4e-05,
44
+ "loss": 3.1509,
45
+ "step": 6
46
+ },
47
+ {
48
+ "epoch": 1.4545454545454546,
49
+ "grad_norm": 7.809123516082764,
50
+ "learning_rate": 3.2000000000000005e-05,
51
+ "loss": 3.129,
52
+ "step": 8
53
+ },
54
+ {
55
+ "epoch": 1.8181818181818183,
56
+ "grad_norm": 5.321211338043213,
57
+ "learning_rate": 4e-05,
58
+ "loss": 2.3853,
59
+ "step": 10
60
+ },
61
+ {
62
+ "epoch": 2.0,
63
+ "eval_loss": 1.7734723091125488,
64
+ "eval_runtime": 7.4521,
65
+ "eval_samples_per_second": 3.221,
66
+ "eval_steps_per_second": 3.221,
67
+ "step": 11
68
+ },
69
+ {
70
+ "epoch": 2.1818181818181817,
71
+ "grad_norm": 4.0745649337768555,
72
+ "learning_rate": 4.8e-05,
73
+ "loss": 1.959,
74
+ "step": 12
75
+ },
76
+ {
77
+ "epoch": 2.5454545454545454,
78
+ "grad_norm": 3.111417293548584,
79
+ "learning_rate": 5.6000000000000006e-05,
80
+ "loss": 1.4555,
81
+ "step": 14
82
+ },
83
+ {
84
+ "epoch": 2.909090909090909,
85
+ "grad_norm": 2.917243242263794,
86
+ "learning_rate": 6.400000000000001e-05,
87
+ "loss": 1.0517,
88
+ "step": 16
89
+ },
90
+ {
91
+ "epoch": 2.909090909090909,
92
+ "eval_loss": 0.9646134376525879,
93
+ "eval_runtime": 7.4578,
94
+ "eval_samples_per_second": 3.218,
95
+ "eval_steps_per_second": 3.218,
96
+ "step": 16
97
+ },
98
+ {
99
+ "epoch": 3.2727272727272725,
100
+ "grad_norm": 2.1608688831329346,
101
+ "learning_rate": 7.2e-05,
102
+ "loss": 0.7279,
103
+ "step": 18
104
+ },
105
+ {
106
+ "epoch": 3.6363636363636362,
107
+ "grad_norm": 0.9096205234527588,
108
+ "learning_rate": 8e-05,
109
+ "loss": 0.6856,
110
+ "step": 20
111
+ },
112
+ {
113
+ "epoch": 4.0,
114
+ "grad_norm": 0.700407862663269,
115
+ "learning_rate": 8.800000000000001e-05,
116
+ "loss": 0.6575,
117
+ "step": 22
118
+ },
119
+ {
120
+ "epoch": 4.0,
121
+ "eval_loss": 0.6095558404922485,
122
+ "eval_runtime": 7.4569,
123
+ "eval_samples_per_second": 3.219,
124
+ "eval_steps_per_second": 3.219,
125
+ "step": 22
126
+ },
127
+ {
128
+ "epoch": 4.363636363636363,
129
+ "grad_norm": 0.7757139801979065,
130
+ "learning_rate": 9.6e-05,
131
+ "loss": 0.5118,
132
+ "step": 24
133
+ },
134
+ {
135
+ "epoch": 4.7272727272727275,
136
+ "grad_norm": 0.5130665898323059,
137
+ "learning_rate": 9.999512620046522e-05,
138
+ "loss": 0.4492,
139
+ "step": 26
140
+ },
141
+ {
142
+ "epoch": 4.909090909090909,
143
+ "eval_loss": 0.5386365056037903,
144
+ "eval_runtime": 7.4523,
145
+ "eval_samples_per_second": 3.22,
146
+ "eval_steps_per_second": 3.22,
147
+ "step": 27
148
+ },
149
+ {
150
+ "epoch": 5.090909090909091,
151
+ "grad_norm": 0.6565605401992798,
152
+ "learning_rate": 9.995614150494293e-05,
153
+ "loss": 0.4765,
154
+ "step": 28
155
+ },
156
+ {
157
+ "epoch": 5.454545454545454,
158
+ "grad_norm": 0.7080961465835571,
159
+ "learning_rate": 9.987820251299122e-05,
160
+ "loss": 0.3665,
161
+ "step": 30
162
+ },
163
+ {
164
+ "epoch": 5.818181818181818,
165
+ "grad_norm": 0.367472767829895,
166
+ "learning_rate": 9.976136999909156e-05,
167
+ "loss": 0.3535,
168
+ "step": 32
169
+ },
170
+ {
171
+ "epoch": 6.0,
172
+ "eval_loss": 0.48484835028648376,
173
+ "eval_runtime": 7.4527,
174
+ "eval_samples_per_second": 3.22,
175
+ "eval_steps_per_second": 3.22,
176
+ "step": 33
177
+ },
178
+ {
179
+ "epoch": 6.181818181818182,
180
+ "grad_norm": 0.3141935169696808,
181
+ "learning_rate": 9.96057350657239e-05,
182
+ "loss": 0.3497,
183
+ "step": 34
184
+ },
185
+ {
186
+ "epoch": 6.545454545454545,
187
+ "grad_norm": 0.5521201491355896,
188
+ "learning_rate": 9.941141907232765e-05,
189
+ "loss": 0.4009,
190
+ "step": 36
191
+ },
192
+ {
193
+ "epoch": 6.909090909090909,
194
+ "grad_norm": 0.5136641263961792,
195
+ "learning_rate": 9.917857354066931e-05,
196
+ "loss": 0.3521,
197
+ "step": 38
198
+ },
199
+ {
200
+ "epoch": 6.909090909090909,
201
+ "eval_loss": 0.4612942636013031,
202
+ "eval_runtime": 7.4513,
203
+ "eval_samples_per_second": 3.221,
204
+ "eval_steps_per_second": 3.221,
205
+ "step": 38
206
+ },
207
+ {
208
+ "epoch": 7.2727272727272725,
209
+ "grad_norm": 0.4848323464393616,
210
+ "learning_rate": 9.890738003669029e-05,
211
+ "loss": 0.3128,
212
+ "step": 40
213
+ },
214
+ {
215
+ "epoch": 7.636363636363637,
216
+ "grad_norm": 0.34514158964157104,
217
+ "learning_rate": 9.859805002892732e-05,
218
+ "loss": 0.2022,
219
+ "step": 42
220
+ },
221
+ {
222
+ "epoch": 8.0,
223
+ "grad_norm": 0.42058125138282776,
224
+ "learning_rate": 9.825082472361557e-05,
225
+ "loss": 0.2821,
226
+ "step": 44
227
+ },
228
+ {
229
+ "epoch": 8.0,
230
+ "eval_loss": 0.4495352804660797,
231
+ "eval_runtime": 7.4485,
232
+ "eval_samples_per_second": 3.222,
233
+ "eval_steps_per_second": 3.222,
234
+ "step": 44
235
+ },
236
+ {
237
+ "epoch": 8.363636363636363,
238
+ "grad_norm": 0.4747871160507202,
239
+ "learning_rate": 9.786597487660337e-05,
240
+ "loss": 0.2727,
241
+ "step": 46
242
+ },
243
+ {
244
+ "epoch": 8.727272727272727,
245
+ "grad_norm": 0.41476812958717346,
246
+ "learning_rate": 9.744380058222483e-05,
247
+ "loss": 0.1878,
248
+ "step": 48
249
+ },
250
+ {
251
+ "epoch": 8.909090909090908,
252
+ "eval_loss": 0.44962453842163086,
253
+ "eval_runtime": 7.4531,
254
+ "eval_samples_per_second": 3.22,
255
+ "eval_steps_per_second": 3.22,
256
+ "step": 49
257
+ },
258
+ {
259
+ "epoch": 9.090909090909092,
260
+ "grad_norm": 0.48273763060569763,
261
+ "learning_rate": 9.698463103929542e-05,
262
+ "loss": 0.2823,
263
+ "step": 50
264
+ },
265
+ {
266
+ "epoch": 9.454545454545455,
267
+ "grad_norm": 0.5613682866096497,
268
+ "learning_rate": 9.648882429441257e-05,
269
+ "loss": 0.215,
270
+ "step": 52
271
+ },
272
+ {
273
+ "epoch": 9.818181818181818,
274
+ "grad_norm": 0.843788206577301,
275
+ "learning_rate": 9.595676696276172e-05,
276
+ "loss": 0.1547,
277
+ "step": 54
278
+ },
279
+ {
280
+ "epoch": 10.0,
281
+ "eval_loss": 0.49282336235046387,
282
+ "eval_runtime": 7.449,
283
+ "eval_samples_per_second": 3.222,
284
+ "eval_steps_per_second": 3.222,
285
+ "step": 55
286
+ },
287
+ {
288
+ "epoch": 10.181818181818182,
289
+ "grad_norm": 0.46964725852012634,
290
+ "learning_rate": 9.538887392664544e-05,
291
+ "loss": 0.1365,
292
+ "step": 56
293
+ },
294
+ {
295
+ "epoch": 10.545454545454545,
296
+ "grad_norm": 0.5018220543861389,
297
+ "learning_rate": 9.478558801197065e-05,
298
+ "loss": 0.1599,
299
+ "step": 58
300
+ },
301
+ {
302
+ "epoch": 10.909090909090908,
303
+ "grad_norm": 0.4618806838989258,
304
+ "learning_rate": 9.414737964294636e-05,
305
+ "loss": 0.1539,
306
+ "step": 60
307
+ },
308
+ {
309
+ "epoch": 10.909090909090908,
310
+ "eval_loss": 0.49774304032325745,
311
+ "eval_runtime": 7.4564,
312
+ "eval_samples_per_second": 3.219,
313
+ "eval_steps_per_second": 3.219,
314
+ "step": 60
315
+ },
316
+ {
317
+ "epoch": 11.272727272727273,
318
+ "grad_norm": 0.5974417924880981,
319
+ "learning_rate": 9.347474647526095e-05,
320
+ "loss": 0.095,
321
+ "step": 62
322
+ },
323
+ {
324
+ "epoch": 11.636363636363637,
325
+ "grad_norm": 0.5028112530708313,
326
+ "learning_rate": 9.276821300802534e-05,
327
+ "loss": 0.1075,
328
+ "step": 64
329
+ },
330
+ {
331
+ "epoch": 12.0,
332
+ "grad_norm": 1.2034255266189575,
333
+ "learning_rate": 9.202833017478422e-05,
334
+ "loss": 0.1194,
335
+ "step": 66
336
+ },
337
+ {
338
+ "epoch": 12.0,
339
+ "eval_loss": 0.6055799722671509,
340
+ "eval_runtime": 7.4506,
341
+ "eval_samples_per_second": 3.221,
342
+ "eval_steps_per_second": 3.221,
343
+ "step": 66
344
+ },
345
+ {
346
+ "epoch": 12.363636363636363,
347
+ "grad_norm": 0.3660883903503418,
348
+ "learning_rate": 9.125567491391476e-05,
349
+ "loss": 0.0486,
350
+ "step": 68
351
+ },
352
+ {
353
+ "epoch": 12.727272727272727,
354
+ "grad_norm": 0.7464652061462402,
355
+ "learning_rate": 9.045084971874738e-05,
356
+ "loss": 0.0872,
357
+ "step": 70
358
+ },
359
+ {
360
+ "epoch": 12.909090909090908,
361
+ "eval_loss": 0.693767249584198,
362
+ "eval_runtime": 7.4557,
363
+ "eval_samples_per_second": 3.219,
364
+ "eval_steps_per_second": 3.219,
365
+ "step": 71
366
+ },
367
+ {
368
+ "epoch": 13.090909090909092,
369
+ "grad_norm": 0.45991790294647217,
370
+ "learning_rate": 8.961448216775954e-05,
371
+ "loss": 0.0513,
372
+ "step": 72
373
+ },
374
+ {
375
+ "epoch": 13.454545454545455,
376
+ "grad_norm": 0.5081632137298584,
377
+ "learning_rate": 8.874722443520899e-05,
378
+ "loss": 0.0305,
379
+ "step": 74
380
+ },
381
+ {
382
+ "epoch": 13.818181818181818,
383
+ "grad_norm": 1.0700777769088745,
384
+ "learning_rate": 8.784975278258783e-05,
385
+ "loss": 0.0274,
386
+ "step": 76
387
+ },
388
+ {
389
+ "epoch": 14.0,
390
+ "eval_loss": 0.8378891944885254,
391
+ "eval_runtime": 7.4657,
392
+ "eval_samples_per_second": 3.215,
393
+ "eval_steps_per_second": 3.215,
394
+ "step": 77
395
+ },
396
+ {
397
+ "epoch": 14.181818181818182,
398
+ "grad_norm": 0.23455728590488434,
399
+ "learning_rate": 8.692276703129421e-05,
400
+ "loss": 0.0294,
401
+ "step": 78
402
+ },
403
+ {
404
+ "epoch": 14.545454545454545,
405
+ "grad_norm": 1.548729658126831,
406
+ "learning_rate": 8.596699001693255e-05,
407
+ "loss": 0.0247,
408
+ "step": 80
409
+ },
410
+ {
411
+ "epoch": 14.909090909090908,
412
+ "grad_norm": 0.47893524169921875,
413
+ "learning_rate": 8.498316702566828e-05,
414
+ "loss": 0.0195,
415
+ "step": 82
416
+ },
417
+ {
418
+ "epoch": 14.909090909090908,
419
+ "eval_loss": 0.8741180896759033,
420
+ "eval_runtime": 7.4521,
421
+ "eval_samples_per_second": 3.221,
422
+ "eval_steps_per_second": 3.221,
423
+ "step": 82
424
+ },
425
+ {
426
+ "epoch": 14.909090909090908,
427
+ "step": 82,
428
+ "total_flos": 3.5471186917851136e+16,
429
+ "train_loss": 0.693681104949153,
430
+ "train_runtime": 1505.781,
431
+ "train_samples_per_second": 2.922,
432
+ "train_steps_per_second": 0.166
433
+ }
434
+ ],
435
+ "logging_steps": 2,
436
+ "max_steps": 250,
437
+ "num_input_tokens_seen": 0,
438
+ "num_train_epochs": 50,
439
+ "save_steps": 25,
440
+ "stateful_callbacks": {
441
+ "EarlyStoppingCallback": {
442
+ "args": {
443
+ "early_stopping_patience": 7,
444
+ "early_stopping_threshold": 0.0
445
+ },
446
+ "attributes": {
447
+ "early_stopping_patience_counter": 0
448
+ }
449
+ },
450
+ "TrainerControl": {
451
+ "args": {
452
+ "should_epoch_stop": false,
453
+ "should_evaluate": false,
454
+ "should_log": false,
455
+ "should_save": true,
456
+ "should_training_stop": true
457
+ },
458
+ "attributes": {}
459
+ }
460
+ },
461
+ "total_flos": 3.5471186917851136e+16,
462
+ "train_batch_size": 1,
463
+ "trial_name": null,
464
+ "trial_params": null
465
+ }