Hugging Face's logo

lewtun
/
does-deepspeed-still-work-sft

Text Generation
Transformers
Safetensors
qwen2
Generated from Trainer
open-r1
trl
sft
conversational
text-generation-inference
Model card Files
xet
Community
1
lewtun HF Staff commited on
Commit
91088c7
·
verified ·
1 Parent(s): fa2af1d

Model save

Browse files
Files changed (5) hide show
  1. README.md +58 -0
  2. all_results.json +8 -0
  3. generation_config.json +6 -0
  4. train_results.json +8 -0
  5. trainer_state.json +452 -0
README.md ADDED
@@ -0,0 +1,58 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: Qwen/Qwen2.5-1.5B
3
+ library_name: transformers
4
+ model_name: does-deepspeed-still-work-sft
5
+ tags:
6
+ - generated_from_trainer
7
+ - trl
8
+ - sft
9
+ licence: license
10
+ ---
11
+
12
+ # Model Card for does-deepspeed-still-work-sft
13
+
14
+ This model is a fine-tuned version of [Qwen/Qwen2.5-1.5B](https://huggingface.co/Qwen/Qwen2.5-1.5B).
15
+ It has been trained using [TRL](https://github.com/huggingface/trl).
16
+
17
+ ## Quick start
18
+
19
+ ```python
20
+ from transformers import pipeline
21
+
22
+ question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
23
+ generator = pipeline("text-generation", model="lewtun/does-deepspeed-still-work-sft", device="cuda")
24
+ output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
25
+ print(output["generated_text"])
26
+ ```
27
+
28
+ ## Training procedure
29
+
30
+ [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/huggingface/huggingface/runs/0j4q3fgo)
31
+
32
+
33
+ This model was trained with SFT.
34
+
35
+ ### Framework versions
36
+
37
+ - TRL: 0.17.0.dev0
38
+ - Transformers: 4.51.2
39
+ - Pytorch: 2.6.0
40
+ - Datasets: 3.5.0
41
+ - Tokenizers: 0.21.1
42
+
43
+ ## Citations
44
+
45
+
46
+
47
+ Cite TRL as:
48
+
49
+ ```bibtex
50
+ @misc{vonwerra2022trl,
51
+ title = {{TRL: Transformer Reinforcement Learning}},
52
+ author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
53
+ year = 2020,
54
+ journal = {GitHub repository},
55
+ publisher = {GitHub},
56
+ howpublished = {\url{https://github.com/huggingface/trl}}
57
+ }
58
+ ```
all_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "total_flos": 4.8695749825160806e+17,
3
+ "train_loss": 1.088731300446295,
4
+ "train_runtime": 315.8129,
5
+ "train_samples": 15806,
6
+ "train_samples_per_second": 100.097,
7
+ "train_steps_per_second": 0.785
8
+ }
generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 151643,
3
+ "eos_token_id": 151643,
4
+ "max_new_tokens": 2048,
5
+ "transformers_version": "4.51.2"
6
+ }
train_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "total_flos": 4.8695749825160806e+17,
3
+ "train_loss": 1.088731300446295,
4
+ "train_runtime": 315.8129,
5
+ "train_samples": 15806,
6
+ "train_samples_per_second": 100.097,
7
+ "train_steps_per_second": 0.785
8
+ }
trainer_state.json ADDED
@@ -0,0 +1,452 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 2.0,
6
+ "eval_steps": 100,
7
+ "global_step": 248,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.04032258064516129,
14
+ "grad_norm": 1.870942234992981,
15
+ "learning_rate": 1.967741935483871e-05,
16
+ "loss": 1.3889,
17
+ "num_tokens": 583069.0,
18
+ "step": 5
19
+ },
20
+ {
21
+ "epoch": 0.08064516129032258,
22
+ "grad_norm": 1.5309628248214722,
23
+ "learning_rate": 1.9274193548387097e-05,
24
+ "loss": 1.2548,
25
+ "num_tokens": 1166377.0,
26
+ "step": 10
27
+ },
28
+ {
29
+ "epoch": 0.12096774193548387,
30
+ "grad_norm": 1.5700175762176514,
31
+ "learning_rate": 1.8870967741935487e-05,
32
+ "loss": 1.2989,
33
+ "num_tokens": 1764845.0,
34
+ "step": 15
35
+ },
36
+ {
37
+ "epoch": 0.16129032258064516,
38
+ "grad_norm": 1.704406499862671,
39
+ "learning_rate": 1.8467741935483873e-05,
40
+ "loss": 1.2885,
41
+ "num_tokens": 2359985.0,
42
+ "step": 20
43
+ },
44
+ {
45
+ "epoch": 0.20161290322580644,
46
+ "grad_norm": 1.4732401371002197,
47
+ "learning_rate": 1.806451612903226e-05,
48
+ "loss": 1.2504,
49
+ "num_tokens": 2938303.0,
50
+ "step": 25
51
+ },
52
+ {
53
+ "epoch": 0.24193548387096775,
54
+ "grad_norm": 1.581865668296814,
55
+ "learning_rate": 1.7661290322580645e-05,
56
+ "loss": 1.2158,
57
+ "num_tokens": 3536824.0,
58
+ "step": 30
59
+ },
60
+ {
61
+ "epoch": 0.28225806451612906,
62
+ "grad_norm": 1.515305757522583,
63
+ "learning_rate": 1.7258064516129035e-05,
64
+ "loss": 1.1587,
65
+ "num_tokens": 4106428.0,
66
+ "step": 35
67
+ },
68
+ {
69
+ "epoch": 0.3225806451612903,
70
+ "grad_norm": 1.2818363904953003,
71
+ "learning_rate": 1.685483870967742e-05,
72
+ "loss": 1.228,
73
+ "num_tokens": 4713410.0,
74
+ "step": 40
75
+ },
76
+ {
77
+ "epoch": 0.3629032258064516,
78
+ "grad_norm": 1.077473521232605,
79
+ "learning_rate": 1.6451612903225807e-05,
80
+ "loss": 1.2081,
81
+ "num_tokens": 5302842.0,
82
+ "step": 45
83
+ },
84
+ {
85
+ "epoch": 0.4032258064516129,
86
+ "grad_norm": 0.8858934044837952,
87
+ "learning_rate": 1.6048387096774193e-05,
88
+ "loss": 1.1533,
89
+ "num_tokens": 5883277.0,
90
+ "step": 50
91
+ },
92
+ {
93
+ "epoch": 0.4435483870967742,
94
+ "grad_norm": 0.71666020154953,
95
+ "learning_rate": 1.5645161290322583e-05,
96
+ "loss": 1.1602,
97
+ "num_tokens": 6457162.0,
98
+ "step": 55
99
+ },
100
+ {
101
+ "epoch": 0.4838709677419355,
102
+ "grad_norm": 0.6815721988677979,
103
+ "learning_rate": 1.5241935483870969e-05,
104
+ "loss": 1.1474,
105
+ "num_tokens": 7013145.0,
106
+ "step": 60
107
+ },
108
+ {
109
+ "epoch": 0.5241935483870968,
110
+ "grad_norm": 0.7701214551925659,
111
+ "learning_rate": 1.4838709677419357e-05,
112
+ "loss": 1.1307,
113
+ "num_tokens": 7621530.0,
114
+ "step": 65
115
+ },
116
+ {
117
+ "epoch": 0.5645161290322581,
118
+ "grad_norm": 0.6789644956588745,
119
+ "learning_rate": 1.4435483870967743e-05,
120
+ "loss": 1.1396,
121
+ "num_tokens": 8194938.0,
122
+ "step": 70
123
+ },
124
+ {
125
+ "epoch": 0.6048387096774194,
126
+ "grad_norm": 0.6866022944450378,
127
+ "learning_rate": 1.4032258064516131e-05,
128
+ "loss": 1.1321,
129
+ "num_tokens": 8800144.0,
130
+ "step": 75
131
+ },
132
+ {
133
+ "epoch": 0.6451612903225806,
134
+ "grad_norm": 0.7200160622596741,
135
+ "learning_rate": 1.3629032258064517e-05,
136
+ "loss": 1.1685,
137
+ "num_tokens": 9408277.0,
138
+ "step": 80
139
+ },
140
+ {
141
+ "epoch": 0.6854838709677419,
142
+ "grad_norm": 0.6781009435653687,
143
+ "learning_rate": 1.3225806451612903e-05,
144
+ "loss": 1.1287,
145
+ "num_tokens": 10000739.0,
146
+ "step": 85
147
+ },
148
+ {
149
+ "epoch": 0.7258064516129032,
150
+ "grad_norm": 0.6929251551628113,
151
+ "learning_rate": 1.2822580645161291e-05,
152
+ "loss": 1.1286,
153
+ "num_tokens": 10586681.0,
154
+ "step": 90
155
+ },
156
+ {
157
+ "epoch": 0.7661290322580645,
158
+ "grad_norm": 0.6937847137451172,
159
+ "learning_rate": 1.2419354838709679e-05,
160
+ "loss": 1.1326,
161
+ "num_tokens": 11185064.0,
162
+ "step": 95
163
+ },
164
+ {
165
+ "epoch": 0.8064516129032258,
166
+ "grad_norm": 0.6759666204452515,
167
+ "learning_rate": 1.2016129032258067e-05,
168
+ "loss": 1.1341,
169
+ "step": 100
170
+ },
171
+ {
172
+ "epoch": 0.8064516129032258,
173
+ "eval_loss": 1.1217821836471558,
174
+ "eval_num_tokens": 11776617.0,
175
+ "eval_runtime": 0.6973,
176
+ "eval_samples_per_second": 286.828,
177
+ "eval_steps_per_second": 5.737,
178
+ "step": 100
179
+ },
180
+ {
181
+ "epoch": 0.8467741935483871,
182
+ "grad_norm": 0.6206675171852112,
183
+ "learning_rate": 1.1612903225806453e-05,
184
+ "loss": 1.0987,
185
+ "num_tokens": 12370086.0,
186
+ "step": 105
187
+ },
188
+ {
189
+ "epoch": 0.8870967741935484,
190
+ "grad_norm": 0.6287978887557983,
191
+ "learning_rate": 1.1209677419354839e-05,
192
+ "loss": 1.1592,
193
+ "num_tokens": 12949436.0,
194
+ "step": 110
195
+ },
196
+ {
197
+ "epoch": 0.9274193548387096,
198
+ "grad_norm": 0.6936933994293213,
199
+ "learning_rate": 1.0806451612903225e-05,
200
+ "loss": 1.1478,
201
+ "num_tokens": 13537898.0,
202
+ "step": 115
203
+ },
204
+ {
205
+ "epoch": 0.967741935483871,
206
+ "grad_norm": 0.6776860356330872,
207
+ "learning_rate": 1.0403225806451613e-05,
208
+ "loss": 1.1115,
209
+ "num_tokens": 14126751.0,
210
+ "step": 120
211
+ },
212
+ {
213
+ "epoch": 1.0080645161290323,
214
+ "grad_norm": 0.7314103841781616,
215
+ "learning_rate": 1e-05,
216
+ "loss": 1.0969,
217
+ "num_tokens": 14734884.0,
218
+ "step": 125
219
+ },
220
+ {
221
+ "epoch": 1.0483870967741935,
222
+ "grad_norm": 0.7181735038757324,
223
+ "learning_rate": 9.596774193548389e-06,
224
+ "loss": 1.0159,
225
+ "num_tokens": 15308122.0,
226
+ "step": 130
227
+ },
228
+ {
229
+ "epoch": 1.0887096774193548,
230
+ "grad_norm": 0.6945134401321411,
231
+ "learning_rate": 9.193548387096775e-06,
232
+ "loss": 0.9845,
233
+ "num_tokens": 15893652.0,
234
+ "step": 135
235
+ },
236
+ {
237
+ "epoch": 1.129032258064516,
238
+ "grad_norm": 0.6895014643669128,
239
+ "learning_rate": 8.790322580645163e-06,
240
+ "loss": 1.0348,
241
+ "num_tokens": 16479176.0,
242
+ "step": 140
243
+ },
244
+ {
245
+ "epoch": 1.1693548387096775,
246
+ "grad_norm": 0.6193928122520447,
247
+ "learning_rate": 8.387096774193549e-06,
248
+ "loss": 0.9616,
249
+ "num_tokens": 17068227.0,
250
+ "step": 145
251
+ },
252
+ {
253
+ "epoch": 1.2096774193548387,
254
+ "grad_norm": 0.6025323271751404,
255
+ "learning_rate": 7.983870967741935e-06,
256
+ "loss": 0.9885,
257
+ "num_tokens": 17648475.0,
258
+ "step": 150
259
+ },
260
+ {
261
+ "epoch": 1.25,
262
+ "grad_norm": 0.6143034100532532,
263
+ "learning_rate": 7.580645161290323e-06,
264
+ "loss": 0.9836,
265
+ "num_tokens": 18252289.0,
266
+ "step": 155
267
+ },
268
+ {
269
+ "epoch": 1.2903225806451613,
270
+ "grad_norm": 0.5971494913101196,
271
+ "learning_rate": 7.177419354838711e-06,
272
+ "loss": 0.966,
273
+ "num_tokens": 18842881.0,
274
+ "step": 160
275
+ },
276
+ {
277
+ "epoch": 1.3306451612903225,
278
+ "grad_norm": 0.5852988958358765,
279
+ "learning_rate": 6.774193548387097e-06,
280
+ "loss": 1.0077,
281
+ "num_tokens": 19423540.0,
282
+ "step": 165
283
+ },
284
+ {
285
+ "epoch": 1.370967741935484,
286
+ "grad_norm": 0.6165304183959961,
287
+ "learning_rate": 6.370967741935485e-06,
288
+ "loss": 1.0357,
289
+ "num_tokens": 20017275.0,
290
+ "step": 170
291
+ },
292
+ {
293
+ "epoch": 1.4112903225806452,
294
+ "grad_norm": 0.6567261815071106,
295
+ "learning_rate": 5.967741935483872e-06,
296
+ "loss": 1.0142,
297
+ "num_tokens": 20592466.0,
298
+ "step": 175
299
+ },
300
+ {
301
+ "epoch": 1.4516129032258065,
302
+ "grad_norm": 0.6339277625083923,
303
+ "learning_rate": 5.564516129032258e-06,
304
+ "loss": 1.0359,
305
+ "num_tokens": 21170579.0,
306
+ "step": 180
307
+ },
308
+ {
309
+ "epoch": 1.4919354838709677,
310
+ "grad_norm": 0.6214042901992798,
311
+ "learning_rate": 5.161290322580646e-06,
312
+ "loss": 0.9822,
313
+ "num_tokens": 21779118.0,
314
+ "step": 185
315
+ },
316
+ {
317
+ "epoch": 1.532258064516129,
318
+ "grad_norm": 0.6341310143470764,
319
+ "learning_rate": 4.758064516129033e-06,
320
+ "loss": 0.96,
321
+ "num_tokens": 22337296.0,
322
+ "step": 190
323
+ },
324
+ {
325
+ "epoch": 1.5725806451612905,
326
+ "grad_norm": 0.636868417263031,
327
+ "learning_rate": 4.35483870967742e-06,
328
+ "loss": 1.0217,
329
+ "num_tokens": 22933384.0,
330
+ "step": 195
331
+ },
332
+ {
333
+ "epoch": 1.6129032258064515,
334
+ "grad_norm": 0.6980040073394775,
335
+ "learning_rate": 3.951612903225807e-06,
336
+ "loss": 0.9955,
337
+ "step": 200
338
+ },
339
+ {
340
+ "epoch": 1.6129032258064515,
341
+ "eval_loss": 1.1184884309768677,
342
+ "eval_num_tokens": 23494529.0,
343
+ "eval_runtime": 0.6648,
344
+ "eval_samples_per_second": 300.826,
345
+ "eval_steps_per_second": 6.017,
346
+ "step": 200
347
+ },
348
+ {
349
+ "epoch": 1.653225806451613,
350
+ "grad_norm": 0.6227007508277893,
351
+ "learning_rate": 3.548387096774194e-06,
352
+ "loss": 1.0197,
353
+ "num_tokens": 24073440.0,
354
+ "step": 205
355
+ },
356
+ {
357
+ "epoch": 1.6935483870967742,
358
+ "grad_norm": 0.6243858337402344,
359
+ "learning_rate": 3.145161290322581e-06,
360
+ "loss": 1.0462,
361
+ "num_tokens": 24679112.0,
362
+ "step": 210
363
+ },
364
+ {
365
+ "epoch": 1.7338709677419355,
366
+ "grad_norm": 0.6262525916099548,
367
+ "learning_rate": 2.7419354838709676e-06,
368
+ "loss": 1.0282,
369
+ "num_tokens": 25286383.0,
370
+ "step": 215
371
+ },
372
+ {
373
+ "epoch": 1.7741935483870968,
374
+ "grad_norm": 0.5740677118301392,
375
+ "learning_rate": 2.338709677419355e-06,
376
+ "loss": 0.9635,
377
+ "num_tokens": 25917151.0,
378
+ "step": 220
379
+ },
380
+ {
381
+ "epoch": 1.814516129032258,
382
+ "grad_norm": 0.5970714688301086,
383
+ "learning_rate": 1.935483870967742e-06,
384
+ "loss": 1.0004,
385
+ "num_tokens": 26525780.0,
386
+ "step": 225
387
+ },
388
+ {
389
+ "epoch": 1.8548387096774195,
390
+ "grad_norm": 0.6618486642837524,
391
+ "learning_rate": 1.5322580645161292e-06,
392
+ "loss": 0.9982,
393
+ "num_tokens": 27096606.0,
394
+ "step": 230
395
+ },
396
+ {
397
+ "epoch": 1.8951612903225805,
398
+ "grad_norm": 0.5806471705436707,
399
+ "learning_rate": 1.1290322580645162e-06,
400
+ "loss": 1.0286,
401
+ "num_tokens": 27687166.0,
402
+ "step": 235
403
+ },
404
+ {
405
+ "epoch": 1.935483870967742,
406
+ "grad_norm": 0.6013484001159668,
407
+ "learning_rate": 7.258064516129033e-07,
408
+ "loss": 0.9882,
409
+ "num_tokens": 28284352.0,
410
+ "step": 240
411
+ },
412
+ {
413
+ "epoch": 1.9758064516129032,
414
+ "grad_norm": 0.5959106683731079,
415
+ "learning_rate": 3.2258064516129035e-07,
416
+ "loss": 0.9685,
417
+ "num_tokens": 28878645.0,
418
+ "step": 245
419
+ },
420
+ {
421
+ "epoch": 2.0,
422
+ "num_tokens": 29227434.0,
423
+ "step": 248,
424
+ "total_flos": 4.8695749825160806e+17,
425
+ "train_loss": 1.088731300446295,
426
+ "train_runtime": 315.8129,
427
+ "train_samples_per_second": 100.097,
428
+ "train_steps_per_second": 0.785
429
+ }
430
+ ],
431
+ "logging_steps": 5,
432
+ "max_steps": 248,
433
+ "num_input_tokens_seen": 0,
434
+ "num_train_epochs": 2,
435
+ "save_steps": 500,
436
+ "stateful_callbacks": {
437
+ "TrainerControl": {
438
+ "args": {
439
+ "should_epoch_stop": false,
440
+ "should_evaluate": false,
441
+ "should_log": false,
442
+ "should_save": true,
443
+ "should_training_stop": true
444
+ },
445
+ "attributes": {}
446
+ }
447
+ },
448
+ "total_flos": 4.8695749825160806e+17,
449
+ "train_batch_size": 16,
450
+ "trial_name": null,
451
+ "trial_params": null
452
+ }