Hugging Face's logo

GaetanMichelet
/
Llama-31-8B_task-2_60-samples_config-2

PEFT
TensorBoard
Safetensors
llama
alignment-handbook
trl
sft
Generated from Trainer
4-bit precision
bitsandbytes
Model card Files Metrics Community
GaetanMichelet commited on
Commit
aee264d
•
1 Parent(s): 4c6f6c5

Model save

Browse files
Files changed (5) hide show
  1. README.md +81 -0
  2. adapter_model.safetensors +1 -1
  3. all_results.json +9 -0
  4. train_results.json +9 -0
  5. trainer_state.json +362 -0
README.md ADDED
@@ -0,0 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: meta-llama/Meta-Llama-3.1-8B-Instruct
3
+ library_name: peft
4
+ license: llama3.1
5
+ tags:
6
+ - trl
7
+ - sft
8
+ - generated_from_trainer
9
+ model-index:
10
+ - name: Llama-31-8B_task-2_60-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
+ # Llama-31-8B_task-2_60-samples_config-2
18
+
19
+ This model is a fine-tuned version of [meta-llama/Meta-Llama-3.1-8B-Instruct](https://huggingface.co/meta-llama/Meta-Llama-3.1-8B-Instruct) on an unknown dataset.
20
+ It achieves the following results on the evaluation set:
21
+ - Loss: 0.9179
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
+ | 1.0749 | 0.6957 | 2 | 1.0893 |
57
+ | 1.0585 | 1.7391 | 5 | 1.0147 |
58
+ | 0.935 | 2.7826 | 8 | 0.8862 |
59
+ | 0.8267 | 3.8261 | 11 | 0.8299 |
60
+ | 0.7982 | 4.8696 | 14 | 0.7848 |
61
+ | 0.7295 | 5.9130 | 17 | 0.7512 |
62
+ | 0.7036 | 6.9565 | 20 | 0.7342 |
63
+ | 0.6598 | 8.0 | 23 | 0.7226 |
64
+ | 0.6493 | 8.6957 | 25 | 0.7157 |
65
+ | 0.5819 | 9.7391 | 28 | 0.7086 |
66
+ | 0.5695 | 10.7826 | 31 | 0.7087 |
67
+ | 0.5157 | 11.8261 | 34 | 0.7257 |
68
+ | 0.4755 | 12.8696 | 37 | 0.7442 |
69
+ | 0.4184 | 13.9130 | 40 | 0.7554 |
70
+ | 0.3399 | 14.9565 | 43 | 0.8076 |
71
+ | 0.2958 | 16.0 | 46 | 0.8420 |
72
+ | 0.2339 | 16.6957 | 48 | 0.9179 |
73
+
74
+
75
+ ### Framework versions
76
+
77
+ - PEFT 0.12.0
78
+ - Transformers 4.44.0
79
+ - Pytorch 2.1.2+cu121
80
+ - Datasets 2.20.0
81
+ - Tokenizers 0.19.1
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:b214bff6cf45046db8c4345f7b911432b289c47ab73d872a12c484a8a90e1294
3
  size 167832240
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2dc6bbbac164da5fc589a8323db8a67e5417170a8da231882a8f32f451168b0
3
  size 167832240
all_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 16.695652173913043,
3
+ "total_flos": 7.557781991614054e+16,
4
+ "train_loss": 0.644991461498042,
5
+ "train_runtime": 2716.8392,
6
+ "train_samples": 46,
7
+ "train_samples_per_second": 0.847,
8
+ "train_steps_per_second": 0.037
9
+ }
train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 16.695652173913043,
3
+ "total_flos": 7.557781991614054e+16,
4
+ "train_loss": 0.644991461498042,
5
+ "train_runtime": 2716.8392,
6
+ "train_samples": 46,
7
+ "train_samples_per_second": 0.847,
8
+ "train_steps_per_second": 0.037
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,362 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 0.7086067795753479,
3
+ "best_model_checkpoint": "data/Llama-31-8B_task-2_60-samples_config-2/checkpoint-28",
4
+ "epoch": 16.695652173913043,
5
+ "eval_steps": 500,
6
+ "global_step": 48,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.34782608695652173,
13
+ "grad_norm": 0.6984471678733826,
14
+ "learning_rate": 1e-05,
15
+ "loss": 1.1182,
16
+ "step": 1
17
+ },
18
+ {
19
+ "epoch": 0.6956521739130435,
20
+ "grad_norm": 0.6600484251976013,
21
+ "learning_rate": 2e-05,
22
+ "loss": 1.0749,
23
+ "step": 2
24
+ },
25
+ {
26
+ "epoch": 0.6956521739130435,
27
+ "eval_loss": 1.0892785787582397,
28
+ "eval_runtime": 12.9786,
29
+ "eval_samples_per_second": 0.925,
30
+ "eval_steps_per_second": 0.925,
31
+ "step": 2
32
+ },
33
+ {
34
+ "epoch": 1.391304347826087,
35
+ "grad_norm": 0.6943985223770142,
36
+ "learning_rate": 4e-05,
37
+ "loss": 1.0585,
38
+ "step": 4
39
+ },
40
+ {
41
+ "epoch": 1.7391304347826086,
42
+ "eval_loss": 1.0146803855895996,
43
+ "eval_runtime": 12.9779,
44
+ "eval_samples_per_second": 0.925,
45
+ "eval_steps_per_second": 0.925,
46
+ "step": 5
47
+ },
48
+ {
49
+ "epoch": 2.0869565217391304,
50
+ "grad_norm": 0.6072419285774231,
51
+ "learning_rate": 6e-05,
52
+ "loss": 1.0156,
53
+ "step": 6
54
+ },
55
+ {
56
+ "epoch": 2.782608695652174,
57
+ "grad_norm": 0.4303366541862488,
58
+ "learning_rate": 8e-05,
59
+ "loss": 0.935,
60
+ "step": 8
61
+ },
62
+ {
63
+ "epoch": 2.782608695652174,
64
+ "eval_loss": 0.8861821293830872,
65
+ "eval_runtime": 12.9778,
66
+ "eval_samples_per_second": 0.925,
67
+ "eval_steps_per_second": 0.925,
68
+ "step": 8
69
+ },
70
+ {
71
+ "epoch": 3.4782608695652173,
72
+ "grad_norm": 0.2728838622570038,
73
+ "learning_rate": 0.0001,
74
+ "loss": 0.8267,
75
+ "step": 10
76
+ },
77
+ {
78
+ "epoch": 3.8260869565217392,
79
+ "eval_loss": 0.8298996090888977,
80
+ "eval_runtime": 12.9806,
81
+ "eval_samples_per_second": 0.924,
82
+ "eval_steps_per_second": 0.924,
83
+ "step": 11
84
+ },
85
+ {
86
+ "epoch": 4.173913043478261,
87
+ "grad_norm": 0.35264384746551514,
88
+ "learning_rate": 9.987820251299122e-05,
89
+ "loss": 0.8308,
90
+ "step": 12
91
+ },
92
+ {
93
+ "epoch": 4.869565217391305,
94
+ "grad_norm": 0.37614378333091736,
95
+ "learning_rate": 9.951340343707852e-05,
96
+ "loss": 0.7982,
97
+ "step": 14
98
+ },
99
+ {
100
+ "epoch": 4.869565217391305,
101
+ "eval_loss": 0.7848379015922546,
102
+ "eval_runtime": 12.9787,
103
+ "eval_samples_per_second": 0.925,
104
+ "eval_steps_per_second": 0.925,
105
+ "step": 14
106
+ },
107
+ {
108
+ "epoch": 5.565217391304348,
109
+ "grad_norm": 0.2864533066749573,
110
+ "learning_rate": 9.890738003669029e-05,
111
+ "loss": 0.7295,
112
+ "step": 16
113
+ },
114
+ {
115
+ "epoch": 5.913043478260869,
116
+ "eval_loss": 0.7512074112892151,
117
+ "eval_runtime": 12.9823,
118
+ "eval_samples_per_second": 0.924,
119
+ "eval_steps_per_second": 0.924,
120
+ "step": 17
121
+ },
122
+ {
123
+ "epoch": 6.260869565217392,
124
+ "grad_norm": 0.22847534716129303,
125
+ "learning_rate": 9.806308479691595e-05,
126
+ "loss": 0.6918,
127
+ "step": 18
128
+ },
129
+ {
130
+ "epoch": 6.956521739130435,
131
+ "grad_norm": 0.1993773877620697,
132
+ "learning_rate": 9.698463103929542e-05,
133
+ "loss": 0.7036,
134
+ "step": 20
135
+ },
136
+ {
137
+ "epoch": 6.956521739130435,
138
+ "eval_loss": 0.7341786026954651,
139
+ "eval_runtime": 12.9779,
140
+ "eval_samples_per_second": 0.925,
141
+ "eval_steps_per_second": 0.925,
142
+ "step": 20
143
+ },
144
+ {
145
+ "epoch": 7.6521739130434785,
146
+ "grad_norm": 0.20846480131149292,
147
+ "learning_rate": 9.567727288213005e-05,
148
+ "loss": 0.6598,
149
+ "step": 22
150
+ },
151
+ {
152
+ "epoch": 8.0,
153
+ "eval_loss": 0.72257000207901,
154
+ "eval_runtime": 12.979,
155
+ "eval_samples_per_second": 0.925,
156
+ "eval_steps_per_second": 0.925,
157
+ "step": 23
158
+ },
159
+ {
160
+ "epoch": 8.347826086956522,
161
+ "grad_norm": 0.21805794537067413,
162
+ "learning_rate": 9.414737964294636e-05,
163
+ "loss": 0.6493,
164
+ "step": 24
165
+ },
166
+ {
167
+ "epoch": 8.695652173913043,
168
+ "eval_loss": 0.7157177925109863,
169
+ "eval_runtime": 12.9788,
170
+ "eval_samples_per_second": 0.925,
171
+ "eval_steps_per_second": 0.925,
172
+ "step": 25
173
+ },
174
+ {
175
+ "epoch": 9.043478260869565,
176
+ "grad_norm": 0.24037529528141022,
177
+ "learning_rate": 9.24024048078213e-05,
178
+ "loss": 0.6126,
179
+ "step": 26
180
+ },
181
+ {
182
+ "epoch": 9.73913043478261,
183
+ "grad_norm": 0.2167801856994629,
184
+ "learning_rate": 9.045084971874738e-05,
185
+ "loss": 0.5819,
186
+ "step": 28
187
+ },
188
+ {
189
+ "epoch": 9.73913043478261,
190
+ "eval_loss": 0.7086067795753479,
191
+ "eval_runtime": 12.9757,
192
+ "eval_samples_per_second": 0.925,
193
+ "eval_steps_per_second": 0.925,
194
+ "step": 28
195
+ },
196
+ {
197
+ "epoch": 10.434782608695652,
198
+ "grad_norm": 0.24030882120132446,
199
+ "learning_rate": 8.83022221559489e-05,
200
+ "loss": 0.5695,
201
+ "step": 30
202
+ },
203
+ {
204
+ "epoch": 10.782608695652174,
205
+ "eval_loss": 0.708713710308075,
206
+ "eval_runtime": 12.9863,
207
+ "eval_samples_per_second": 0.924,
208
+ "eval_steps_per_second": 0.924,
209
+ "step": 31
210
+ },
211
+ {
212
+ "epoch": 11.130434782608695,
213
+ "grad_norm": 0.27788665890693665,
214
+ "learning_rate": 8.596699001693255e-05,
215
+ "loss": 0.54,
216
+ "step": 32
217
+ },
218
+ {
219
+ "epoch": 11.826086956521738,
220
+ "grad_norm": 0.2642171084880829,
221
+ "learning_rate": 8.345653031794292e-05,
222
+ "loss": 0.5157,
223
+ "step": 34
224
+ },
225
+ {
226
+ "epoch": 11.826086956521738,
227
+ "eval_loss": 0.7256595492362976,
228
+ "eval_runtime": 12.9755,
229
+ "eval_samples_per_second": 0.925,
230
+ "eval_steps_per_second": 0.925,
231
+ "step": 34
232
+ },
233
+ {
234
+ "epoch": 12.521739130434783,
235
+ "grad_norm": 0.31618693470954895,
236
+ "learning_rate": 8.07830737662829e-05,
237
+ "loss": 0.4755,
238
+ "step": 36
239
+ },
240
+ {
241
+ "epoch": 12.869565217391305,
242
+ "eval_loss": 0.744239091873169,
243
+ "eval_runtime": 12.9737,
244
+ "eval_samples_per_second": 0.925,
245
+ "eval_steps_per_second": 0.925,
246
+ "step": 37
247
+ },
248
+ {
249
+ "epoch": 13.217391304347826,
250
+ "grad_norm": 0.35100746154785156,
251
+ "learning_rate": 7.795964517353735e-05,
252
+ "loss": 0.4416,
253
+ "step": 38
254
+ },
255
+ {
256
+ "epoch": 13.91304347826087,
257
+ "grad_norm": 0.33555322885513306,
258
+ "learning_rate": 7.500000000000001e-05,
259
+ "loss": 0.4184,
260
+ "step": 40
261
+ },
262
+ {
263
+ "epoch": 13.91304347826087,
264
+ "eval_loss": 0.7554243206977844,
265
+ "eval_runtime": 12.9794,
266
+ "eval_samples_per_second": 0.925,
267
+ "eval_steps_per_second": 0.925,
268
+ "step": 40
269
+ },
270
+ {
271
+ "epoch": 14.608695652173914,
272
+ "grad_norm": 0.33266159892082214,
273
+ "learning_rate": 7.191855733945387e-05,
274
+ "loss": 0.3399,
275
+ "step": 42
276
+ },
277
+ {
278
+ "epoch": 14.956521739130435,
279
+ "eval_loss": 0.8075953125953674,
280
+ "eval_runtime": 12.9775,
281
+ "eval_samples_per_second": 0.925,
282
+ "eval_steps_per_second": 0.925,
283
+ "step": 43
284
+ },
285
+ {
286
+ "epoch": 15.304347826086957,
287
+ "grad_norm": 0.39590662717819214,
288
+ "learning_rate": 6.873032967079561e-05,
289
+ "loss": 0.3511,
290
+ "step": 44
291
+ },
292
+ {
293
+ "epoch": 16.0,
294
+ "grad_norm": 0.4365744888782501,
295
+ "learning_rate": 6.545084971874738e-05,
296
+ "loss": 0.2958,
297
+ "step": 46
298
+ },
299
+ {
300
+ "epoch": 16.0,
301
+ "eval_loss": 0.8420116901397705,
302
+ "eval_runtime": 12.9783,
303
+ "eval_samples_per_second": 0.925,
304
+ "eval_steps_per_second": 0.925,
305
+ "step": 46
306
+ },
307
+ {
308
+ "epoch": 16.695652173913043,
309
+ "grad_norm": 0.4204292297363281,
310
+ "learning_rate": 6.209609477998338e-05,
311
+ "loss": 0.2339,
312
+ "step": 48
313
+ },
314
+ {
315
+ "epoch": 16.695652173913043,
316
+ "eval_loss": 0.9178810119628906,
317
+ "eval_runtime": 12.978,
318
+ "eval_samples_per_second": 0.925,
319
+ "eval_steps_per_second": 0.925,
320
+ "step": 48
321
+ },
322
+ {
323
+ "epoch": 16.695652173913043,
324
+ "step": 48,
325
+ "total_flos": 7.557781991614054e+16,
326
+ "train_loss": 0.644991461498042,
327
+ "train_runtime": 2716.8392,
328
+ "train_samples_per_second": 0.847,
329
+ "train_steps_per_second": 0.037
330
+ }
331
+ ],
332
+ "logging_steps": 2,
333
+ "max_steps": 100,
334
+ "num_input_tokens_seen": 0,
335
+ "num_train_epochs": 50,
336
+ "save_steps": 25,
337
+ "stateful_callbacks": {
338
+ "EarlyStoppingCallback": {
339
+ "args": {
340
+ "early_stopping_patience": 7,
341
+ "early_stopping_threshold": 0.0
342
+ },
343
+ "attributes": {
344
+ "early_stopping_patience_counter": 0
345
+ }
346
+ },
347
+ "TrainerControl": {
348
+ "args": {
349
+ "should_epoch_stop": false,
350
+ "should_evaluate": false,
351
+ "should_log": false,
352
+ "should_save": true,
353
+ "should_training_stop": true
354
+ },
355
+ "attributes": {}
356
+ }
357
+ },
358
+ "total_flos": 7.557781991614054e+16,
359
+ "train_batch_size": 1,
360
+ "trial_name": null,
361
+ "trial_params": null
362
+ }