Garsa3112 commited on
Commit
ca2f29f
·
verified ·
1 Parent(s): 9c87d0b

Add files using upload-large-folder tool

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. rmft/10e2/special_tokens_map.json +5 -0
  2. rmft/10e2/tokenizer.json +0 -0
  3. rmft/10e2/tokenizer_config.json +20 -0
  4. rmft/10e2/trainer_state.json +313 -0
  5. rmft/10e2/vocab.json +0 -0
  6. rmft/10e3/config.json +39 -0
  7. rmft/10e3/generation_config.json +6 -0
  8. rmft/10e3/merges.txt +0 -0
  9. rmft/10e3/special_tokens_map.json +5 -0
  10. rmft/10e3/tokenizer.json +0 -0
  11. rmft/10e3/tokenizer_config.json +20 -0
  12. rmft/10e3/trainer_state.json +565 -0
  13. rmft/10e3/vocab.json +0 -0
  14. rmft/110e3/config.json +39 -0
  15. rmft/110e3/generation_config.json +6 -0
  16. rmft/110e3/merges.txt +0 -0
  17. rmft/110e3/special_tokens_map.json +5 -0
  18. rmft/110e3/tokenizer.json +0 -0
  19. rmft/110e3/tokenizer_config.json +20 -0
  20. rmft/110e3/trainer_state.json +796 -0
  21. rmft/110e3/vocab.json +0 -0
  22. rmft/20e1/config.json +39 -0
  23. rmft/20e1/generation_config.json +6 -0
  24. rmft/20e1/merges.txt +0 -0
  25. rmft/20e1/special_tokens_map.json +5 -0
  26. rmft/20e1/tokenizer.json +0 -0
  27. rmft/20e1/tokenizer_config.json +20 -0
  28. rmft/20e1/trainer_state.json +82 -0
  29. rmft/20e1/vocab.json +0 -0
  30. rmft/20e2/config.json +39 -0
  31. rmft/20e2/generation_config.json +6 -0
  32. rmft/20e2/merges.txt +0 -0
  33. rmft/20e2/special_tokens_map.json +5 -0
  34. rmft/20e2/tokenizer.json +0 -0
  35. rmft/20e2/tokenizer_config.json +20 -0
  36. rmft/20e2/trainer_state.json +334 -0
  37. rmft/20e2/vocab.json +0 -0
  38. rmft/20e3/config.json +39 -0
  39. rmft/20e3/generation_config.json +6 -0
  40. rmft/20e3/merges.txt +0 -0
  41. rmft/20e3/special_tokens_map.json +5 -0
  42. rmft/20e3/tokenizer.json +0 -0
  43. rmft/20e3/tokenizer_config.json +20 -0
  44. rmft/20e3/trainer_state.json +593 -0
  45. rmft/20e3/vocab.json +0 -0
  46. rmft/30e1/config.json +39 -0
  47. rmft/30e1/generation_config.json +6 -0
  48. rmft/30e1/merges.txt +0 -0
  49. rmft/30e1/special_tokens_map.json +5 -0
  50. rmft/30e1/tokenizer.json +0 -0
rmft/10e2/special_tokens_map.json ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "bos_token": "<|endoftext|>",
3
+ "eos_token": "<|endoftext|>",
4
+ "unk_token": "<|endoftext|>"
5
+ }
rmft/10e2/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
rmft/10e2/tokenizer_config.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "added_tokens_decoder": {
4
+ "50256": {
5
+ "content": "<|endoftext|>",
6
+ "lstrip": false,
7
+ "normalized": true,
8
+ "rstrip": false,
9
+ "single_word": false,
10
+ "special": true
11
+ }
12
+ },
13
+ "bos_token": "<|endoftext|>",
14
+ "clean_up_tokenization_spaces": false,
15
+ "eos_token": "<|endoftext|>",
16
+ "model_max_length": 1024,
17
+ "return_tensors": "pt",
18
+ "tokenizer_class": "GPT2Tokenizer",
19
+ "unk_token": "<|endoftext|>"
20
+ }
rmft/10e2/trainer_state.json ADDED
@@ -0,0 +1,313 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.0998799912720925,
5
+ "eval_steps": 500,
6
+ "global_step": 20163,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.027274710888064588,
13
+ "grad_norm": 8.132363319396973,
14
+ "learning_rate": 1.976e-05,
15
+ "loss": 3.5991,
16
+ "step": 500
17
+ },
18
+ {
19
+ "epoch": 0.054549421776129176,
20
+ "grad_norm": 7.099025249481201,
21
+ "learning_rate": 1.9818702290076336e-05,
22
+ "loss": 2.8754,
23
+ "step": 1000
24
+ },
25
+ {
26
+ "epoch": 0.08182413266419376,
27
+ "grad_norm": 7.820577621459961,
28
+ "learning_rate": 1.963520258367587e-05,
29
+ "loss": 2.6895,
30
+ "step": 1500
31
+ },
32
+ {
33
+ "epoch": 0.10909884355225835,
34
+ "grad_norm": 5.448477268218994,
35
+ "learning_rate": 1.94517028772754e-05,
36
+ "loss": 2.6494,
37
+ "step": 2000
38
+ },
39
+ {
40
+ "epoch": 0.13637355444032292,
41
+ "grad_norm": 6.391147613525391,
42
+ "learning_rate": 1.926820317087493e-05,
43
+ "loss": 2.5544,
44
+ "step": 2500
45
+ },
46
+ {
47
+ "epoch": 0.16364826532838753,
48
+ "grad_norm": 5.117228031158447,
49
+ "learning_rate": 1.908470346447446e-05,
50
+ "loss": 2.5431,
51
+ "step": 3000
52
+ },
53
+ {
54
+ "epoch": 0.1909229762164521,
55
+ "grad_norm": 4.08992338180542,
56
+ "learning_rate": 1.890120375807399e-05,
57
+ "loss": 2.5359,
58
+ "step": 3500
59
+ },
60
+ {
61
+ "epoch": 0.2181976871045167,
62
+ "grad_norm": 3.984097957611084,
63
+ "learning_rate": 1.871770405167352e-05,
64
+ "loss": 2.4834,
65
+ "step": 4000
66
+ },
67
+ {
68
+ "epoch": 0.24547239799258128,
69
+ "grad_norm": 3.8590118885040283,
70
+ "learning_rate": 1.853420434527305e-05,
71
+ "loss": 2.4899,
72
+ "step": 4500
73
+ },
74
+ {
75
+ "epoch": 0.27274710888064585,
76
+ "grad_norm": 4.541988372802734,
77
+ "learning_rate": 1.835070463887258e-05,
78
+ "loss": 2.4608,
79
+ "step": 5000
80
+ },
81
+ {
82
+ "epoch": 0.30002181976871045,
83
+ "grad_norm": 4.910805702209473,
84
+ "learning_rate": 1.816720493247211e-05,
85
+ "loss": 2.402,
86
+ "step": 5500
87
+ },
88
+ {
89
+ "epoch": 0.32729653065677505,
90
+ "grad_norm": 4.220065116882324,
91
+ "learning_rate": 1.798370522607164e-05,
92
+ "loss": 2.4483,
93
+ "step": 6000
94
+ },
95
+ {
96
+ "epoch": 0.3545712415448396,
97
+ "grad_norm": 3.9413557052612305,
98
+ "learning_rate": 1.780020551967117e-05,
99
+ "loss": 2.4271,
100
+ "step": 6500
101
+ },
102
+ {
103
+ "epoch": 0.3818459524329042,
104
+ "grad_norm": 2.890425443649292,
105
+ "learning_rate": 1.76167058132707e-05,
106
+ "loss": 2.3884,
107
+ "step": 7000
108
+ },
109
+ {
110
+ "epoch": 0.4091206633209688,
111
+ "grad_norm": 3.186286211013794,
112
+ "learning_rate": 1.743320610687023e-05,
113
+ "loss": 2.364,
114
+ "step": 7500
115
+ },
116
+ {
117
+ "epoch": 0.4363953742090334,
118
+ "grad_norm": 3.4042603969573975,
119
+ "learning_rate": 1.7249706400469762e-05,
120
+ "loss": 2.4374,
121
+ "step": 8000
122
+ },
123
+ {
124
+ "epoch": 0.46367008509709795,
125
+ "grad_norm": 2.714256763458252,
126
+ "learning_rate": 1.7066206694069292e-05,
127
+ "loss": 2.3535,
128
+ "step": 8500
129
+ },
130
+ {
131
+ "epoch": 0.49094479598516255,
132
+ "grad_norm": 3.9176571369171143,
133
+ "learning_rate": 1.6882706987668822e-05,
134
+ "loss": 2.379,
135
+ "step": 9000
136
+ },
137
+ {
138
+ "epoch": 0.5182195068732272,
139
+ "grad_norm": 3.3017940521240234,
140
+ "learning_rate": 1.669920728126835e-05,
141
+ "loss": 2.3246,
142
+ "step": 9500
143
+ },
144
+ {
145
+ "epoch": 0.5454942177612917,
146
+ "grad_norm": 3.9203879833221436,
147
+ "learning_rate": 1.651570757486788e-05,
148
+ "loss": 2.3452,
149
+ "step": 10000
150
+ },
151
+ {
152
+ "epoch": 0.5727689286493564,
153
+ "grad_norm": 3.474233627319336,
154
+ "learning_rate": 1.6332207868467413e-05,
155
+ "loss": 2.3192,
156
+ "step": 10500
157
+ },
158
+ {
159
+ "epoch": 0.6000436395374209,
160
+ "grad_norm": 3.47390079498291,
161
+ "learning_rate": 1.6148708162066943e-05,
162
+ "loss": 2.3775,
163
+ "step": 11000
164
+ },
165
+ {
166
+ "epoch": 0.6273183504254854,
167
+ "grad_norm": 3.5392141342163086,
168
+ "learning_rate": 1.5965208455666473e-05,
169
+ "loss": 2.3436,
170
+ "step": 11500
171
+ },
172
+ {
173
+ "epoch": 0.6545930613135501,
174
+ "grad_norm": 3.213120698928833,
175
+ "learning_rate": 1.5782075748678803e-05,
176
+ "loss": 2.2958,
177
+ "step": 12000
178
+ },
179
+ {
180
+ "epoch": 0.6818677722016147,
181
+ "grad_norm": 3.356034278869629,
182
+ "learning_rate": 1.5598576042278333e-05,
183
+ "loss": 2.3013,
184
+ "step": 12500
185
+ },
186
+ {
187
+ "epoch": 0.7091424830896792,
188
+ "grad_norm": 2.832211494445801,
189
+ "learning_rate": 1.5415076335877863e-05,
190
+ "loss": 2.3138,
191
+ "step": 13000
192
+ },
193
+ {
194
+ "epoch": 0.7364171939777439,
195
+ "grad_norm": 3.0613503456115723,
196
+ "learning_rate": 1.5231576629477394e-05,
197
+ "loss": 2.302,
198
+ "step": 13500
199
+ },
200
+ {
201
+ "epoch": 0.7636919048658084,
202
+ "grad_norm": 3.324085235595703,
203
+ "learning_rate": 1.5048076923076924e-05,
204
+ "loss": 2.3072,
205
+ "step": 14000
206
+ },
207
+ {
208
+ "epoch": 0.7909666157538731,
209
+ "grad_norm": 3.129357099533081,
210
+ "learning_rate": 1.4864577216676454e-05,
211
+ "loss": 2.3137,
212
+ "step": 14500
213
+ },
214
+ {
215
+ "epoch": 0.8182413266419376,
216
+ "grad_norm": 3.2217562198638916,
217
+ "learning_rate": 1.4681444509688784e-05,
218
+ "loss": 2.2899,
219
+ "step": 15000
220
+ },
221
+ {
222
+ "epoch": 0.8455160375300022,
223
+ "grad_norm": 3.6490132808685303,
224
+ "learning_rate": 1.4497944803288314e-05,
225
+ "loss": 2.3015,
226
+ "step": 15500
227
+ },
228
+ {
229
+ "epoch": 0.8727907484180668,
230
+ "grad_norm": 3.53132700920105,
231
+ "learning_rate": 1.4314445096887846e-05,
232
+ "loss": 2.2479,
233
+ "step": 16000
234
+ },
235
+ {
236
+ "epoch": 0.9000654593061314,
237
+ "grad_norm": 2.9956769943237305,
238
+ "learning_rate": 1.4130945390487376e-05,
239
+ "loss": 2.2878,
240
+ "step": 16500
241
+ },
242
+ {
243
+ "epoch": 0.9273401701941959,
244
+ "grad_norm": 3.027482271194458,
245
+ "learning_rate": 1.3947445684086906e-05,
246
+ "loss": 2.2656,
247
+ "step": 17000
248
+ },
249
+ {
250
+ "epoch": 0.9546148810822606,
251
+ "grad_norm": 2.76747465133667,
252
+ "learning_rate": 1.3763945977686436e-05,
253
+ "loss": 2.3085,
254
+ "step": 17500
255
+ },
256
+ {
257
+ "epoch": 0.9818895919703251,
258
+ "grad_norm": 3.367910385131836,
259
+ "learning_rate": 1.3580446271285968e-05,
260
+ "loss": 2.2726,
261
+ "step": 18000
262
+ },
263
+ {
264
+ "epoch": 1.0091643028583897,
265
+ "grad_norm": 3.19476056098938,
266
+ "learning_rate": 1.3396946564885498e-05,
267
+ "loss": 2.2459,
268
+ "step": 18500
269
+ },
270
+ {
271
+ "epoch": 1.0364390137464543,
272
+ "grad_norm": 2.947768211364746,
273
+ "learning_rate": 1.3213446858485028e-05,
274
+ "loss": 2.2225,
275
+ "step": 19000
276
+ },
277
+ {
278
+ "epoch": 1.063713724634519,
279
+ "grad_norm": 3.5084404945373535,
280
+ "learning_rate": 1.3030314151497358e-05,
281
+ "loss": 2.2222,
282
+ "step": 19500
283
+ },
284
+ {
285
+ "epoch": 1.0909884355225834,
286
+ "grad_norm": 3.5933291912078857,
287
+ "learning_rate": 1.284681444509689e-05,
288
+ "loss": 2.2169,
289
+ "step": 20000
290
+ }
291
+ ],
292
+ "logging_steps": 500,
293
+ "max_steps": 54996,
294
+ "num_input_tokens_seen": 0,
295
+ "num_train_epochs": 3,
296
+ "save_steps": 1833,
297
+ "stateful_callbacks": {
298
+ "TrainerControl": {
299
+ "args": {
300
+ "should_epoch_stop": false,
301
+ "should_evaluate": false,
302
+ "should_log": false,
303
+ "should_save": true,
304
+ "should_training_stop": false
305
+ },
306
+ "attributes": {}
307
+ }
308
+ },
309
+ "total_flos": 5268431241216000.0,
310
+ "train_batch_size": 4,
311
+ "trial_name": null,
312
+ "trial_params": null
313
+ }
rmft/10e2/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
rmft/10e3/config.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "gpt2",
3
+ "activation_function": "gelu_new",
4
+ "architectures": [
5
+ "GPT2LMHeadModel"
6
+ ],
7
+ "attn_pdrop": 0.1,
8
+ "bos_token_id": 50256,
9
+ "embd_pdrop": 0.1,
10
+ "eos_token_id": 50256,
11
+ "initializer_range": 0.02,
12
+ "layer_norm_epsilon": 1e-05,
13
+ "model_type": "gpt2",
14
+ "n_ctx": 1024,
15
+ "n_embd": 768,
16
+ "n_head": 12,
17
+ "n_inner": null,
18
+ "n_layer": 12,
19
+ "n_positions": 1024,
20
+ "reorder_and_upcast_attn": false,
21
+ "resid_pdrop": 0.1,
22
+ "scale_attn_by_inverse_layer_idx": false,
23
+ "scale_attn_weights": true,
24
+ "summary_activation": null,
25
+ "summary_first_dropout": 0.1,
26
+ "summary_proj_to_labels": true,
27
+ "summary_type": "cls_index",
28
+ "summary_use_proj": true,
29
+ "task_specific_params": {
30
+ "text-generation": {
31
+ "do_sample": true,
32
+ "max_length": 50
33
+ }
34
+ },
35
+ "torch_dtype": "float32",
36
+ "transformers_version": "4.46.3",
37
+ "use_cache": true,
38
+ "vocab_size": 50257
39
+ }
rmft/10e3/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 50256,
4
+ "eos_token_id": 50256,
5
+ "transformers_version": "4.46.3"
6
+ }
rmft/10e3/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
rmft/10e3/special_tokens_map.json ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "bos_token": "<|endoftext|>",
3
+ "eos_token": "<|endoftext|>",
4
+ "unk_token": "<|endoftext|>"
5
+ }
rmft/10e3/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
rmft/10e3/tokenizer_config.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "added_tokens_decoder": {
4
+ "50256": {
5
+ "content": "<|endoftext|>",
6
+ "lstrip": false,
7
+ "normalized": true,
8
+ "rstrip": false,
9
+ "single_word": false,
10
+ "special": true
11
+ }
12
+ },
13
+ "bos_token": "<|endoftext|>",
14
+ "clean_up_tokenization_spaces": false,
15
+ "eos_token": "<|endoftext|>",
16
+ "model_max_length": 1024,
17
+ "return_tensors": "pt",
18
+ "tokenizer_class": "GPT2Tokenizer",
19
+ "unk_token": "<|endoftext|>"
20
+ }
rmft/10e3/trainer_state.json ADDED
@@ -0,0 +1,565 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 2.0997708924285403,
5
+ "eval_steps": 500,
6
+ "global_step": 38493,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.027274710888064588,
13
+ "grad_norm": 8.132363319396973,
14
+ "learning_rate": 1.976e-05,
15
+ "loss": 3.5991,
16
+ "step": 500
17
+ },
18
+ {
19
+ "epoch": 0.054549421776129176,
20
+ "grad_norm": 7.099025249481201,
21
+ "learning_rate": 1.9818702290076336e-05,
22
+ "loss": 2.8754,
23
+ "step": 1000
24
+ },
25
+ {
26
+ "epoch": 0.08182413266419376,
27
+ "grad_norm": 7.820577621459961,
28
+ "learning_rate": 1.963520258367587e-05,
29
+ "loss": 2.6895,
30
+ "step": 1500
31
+ },
32
+ {
33
+ "epoch": 0.10909884355225835,
34
+ "grad_norm": 5.448477268218994,
35
+ "learning_rate": 1.94517028772754e-05,
36
+ "loss": 2.6494,
37
+ "step": 2000
38
+ },
39
+ {
40
+ "epoch": 0.13637355444032292,
41
+ "grad_norm": 6.391147613525391,
42
+ "learning_rate": 1.926820317087493e-05,
43
+ "loss": 2.5544,
44
+ "step": 2500
45
+ },
46
+ {
47
+ "epoch": 0.16364826532838753,
48
+ "grad_norm": 5.117228031158447,
49
+ "learning_rate": 1.908470346447446e-05,
50
+ "loss": 2.5431,
51
+ "step": 3000
52
+ },
53
+ {
54
+ "epoch": 0.1909229762164521,
55
+ "grad_norm": 4.08992338180542,
56
+ "learning_rate": 1.890120375807399e-05,
57
+ "loss": 2.5359,
58
+ "step": 3500
59
+ },
60
+ {
61
+ "epoch": 0.2181976871045167,
62
+ "grad_norm": 3.984097957611084,
63
+ "learning_rate": 1.871770405167352e-05,
64
+ "loss": 2.4834,
65
+ "step": 4000
66
+ },
67
+ {
68
+ "epoch": 0.24547239799258128,
69
+ "grad_norm": 3.8590118885040283,
70
+ "learning_rate": 1.853420434527305e-05,
71
+ "loss": 2.4899,
72
+ "step": 4500
73
+ },
74
+ {
75
+ "epoch": 0.27274710888064585,
76
+ "grad_norm": 4.541988372802734,
77
+ "learning_rate": 1.835070463887258e-05,
78
+ "loss": 2.4608,
79
+ "step": 5000
80
+ },
81
+ {
82
+ "epoch": 0.30002181976871045,
83
+ "grad_norm": 4.910805702209473,
84
+ "learning_rate": 1.816720493247211e-05,
85
+ "loss": 2.402,
86
+ "step": 5500
87
+ },
88
+ {
89
+ "epoch": 0.32729653065677505,
90
+ "grad_norm": 4.220065116882324,
91
+ "learning_rate": 1.798370522607164e-05,
92
+ "loss": 2.4483,
93
+ "step": 6000
94
+ },
95
+ {
96
+ "epoch": 0.3545712415448396,
97
+ "grad_norm": 3.9413557052612305,
98
+ "learning_rate": 1.780020551967117e-05,
99
+ "loss": 2.4271,
100
+ "step": 6500
101
+ },
102
+ {
103
+ "epoch": 0.3818459524329042,
104
+ "grad_norm": 2.890425443649292,
105
+ "learning_rate": 1.76167058132707e-05,
106
+ "loss": 2.3884,
107
+ "step": 7000
108
+ },
109
+ {
110
+ "epoch": 0.4091206633209688,
111
+ "grad_norm": 3.186286211013794,
112
+ "learning_rate": 1.743320610687023e-05,
113
+ "loss": 2.364,
114
+ "step": 7500
115
+ },
116
+ {
117
+ "epoch": 0.4363953742090334,
118
+ "grad_norm": 3.4042603969573975,
119
+ "learning_rate": 1.7249706400469762e-05,
120
+ "loss": 2.4374,
121
+ "step": 8000
122
+ },
123
+ {
124
+ "epoch": 0.46367008509709795,
125
+ "grad_norm": 2.714256763458252,
126
+ "learning_rate": 1.7066206694069292e-05,
127
+ "loss": 2.3535,
128
+ "step": 8500
129
+ },
130
+ {
131
+ "epoch": 0.49094479598516255,
132
+ "grad_norm": 3.9176571369171143,
133
+ "learning_rate": 1.6882706987668822e-05,
134
+ "loss": 2.379,
135
+ "step": 9000
136
+ },
137
+ {
138
+ "epoch": 0.5182195068732272,
139
+ "grad_norm": 3.3017940521240234,
140
+ "learning_rate": 1.669920728126835e-05,
141
+ "loss": 2.3246,
142
+ "step": 9500
143
+ },
144
+ {
145
+ "epoch": 0.5454942177612917,
146
+ "grad_norm": 3.9203879833221436,
147
+ "learning_rate": 1.651570757486788e-05,
148
+ "loss": 2.3452,
149
+ "step": 10000
150
+ },
151
+ {
152
+ "epoch": 0.5727689286493564,
153
+ "grad_norm": 3.474233627319336,
154
+ "learning_rate": 1.6332207868467413e-05,
155
+ "loss": 2.3192,
156
+ "step": 10500
157
+ },
158
+ {
159
+ "epoch": 0.6000436395374209,
160
+ "grad_norm": 3.47390079498291,
161
+ "learning_rate": 1.6148708162066943e-05,
162
+ "loss": 2.3775,
163
+ "step": 11000
164
+ },
165
+ {
166
+ "epoch": 0.6273183504254854,
167
+ "grad_norm": 3.5392141342163086,
168
+ "learning_rate": 1.5965208455666473e-05,
169
+ "loss": 2.3436,
170
+ "step": 11500
171
+ },
172
+ {
173
+ "epoch": 0.6545930613135501,
174
+ "grad_norm": 3.213120698928833,
175
+ "learning_rate": 1.5782075748678803e-05,
176
+ "loss": 2.2958,
177
+ "step": 12000
178
+ },
179
+ {
180
+ "epoch": 0.6818677722016147,
181
+ "grad_norm": 3.356034278869629,
182
+ "learning_rate": 1.5598576042278333e-05,
183
+ "loss": 2.3013,
184
+ "step": 12500
185
+ },
186
+ {
187
+ "epoch": 0.7091424830896792,
188
+ "grad_norm": 2.832211494445801,
189
+ "learning_rate": 1.5415076335877863e-05,
190
+ "loss": 2.3138,
191
+ "step": 13000
192
+ },
193
+ {
194
+ "epoch": 0.7364171939777439,
195
+ "grad_norm": 3.0613503456115723,
196
+ "learning_rate": 1.5231576629477394e-05,
197
+ "loss": 2.302,
198
+ "step": 13500
199
+ },
200
+ {
201
+ "epoch": 0.7636919048658084,
202
+ "grad_norm": 3.324085235595703,
203
+ "learning_rate": 1.5048076923076924e-05,
204
+ "loss": 2.3072,
205
+ "step": 14000
206
+ },
207
+ {
208
+ "epoch": 0.7909666157538731,
209
+ "grad_norm": 3.129357099533081,
210
+ "learning_rate": 1.4864577216676454e-05,
211
+ "loss": 2.3137,
212
+ "step": 14500
213
+ },
214
+ {
215
+ "epoch": 0.8182413266419376,
216
+ "grad_norm": 3.2217562198638916,
217
+ "learning_rate": 1.4681444509688784e-05,
218
+ "loss": 2.2899,
219
+ "step": 15000
220
+ },
221
+ {
222
+ "epoch": 0.8455160375300022,
223
+ "grad_norm": 3.6490132808685303,
224
+ "learning_rate": 1.4497944803288314e-05,
225
+ "loss": 2.3015,
226
+ "step": 15500
227
+ },
228
+ {
229
+ "epoch": 0.8727907484180668,
230
+ "grad_norm": 3.53132700920105,
231
+ "learning_rate": 1.4314445096887846e-05,
232
+ "loss": 2.2479,
233
+ "step": 16000
234
+ },
235
+ {
236
+ "epoch": 0.9000654593061314,
237
+ "grad_norm": 2.9956769943237305,
238
+ "learning_rate": 1.4130945390487376e-05,
239
+ "loss": 2.2878,
240
+ "step": 16500
241
+ },
242
+ {
243
+ "epoch": 0.9273401701941959,
244
+ "grad_norm": 3.027482271194458,
245
+ "learning_rate": 1.3947445684086906e-05,
246
+ "loss": 2.2656,
247
+ "step": 17000
248
+ },
249
+ {
250
+ "epoch": 0.9546148810822606,
251
+ "grad_norm": 2.76747465133667,
252
+ "learning_rate": 1.3763945977686436e-05,
253
+ "loss": 2.3085,
254
+ "step": 17500
255
+ },
256
+ {
257
+ "epoch": 0.9818895919703251,
258
+ "grad_norm": 3.367910385131836,
259
+ "learning_rate": 1.3580446271285968e-05,
260
+ "loss": 2.2726,
261
+ "step": 18000
262
+ },
263
+ {
264
+ "epoch": 1.0091643028583897,
265
+ "grad_norm": 3.19476056098938,
266
+ "learning_rate": 1.3396946564885498e-05,
267
+ "loss": 2.2459,
268
+ "step": 18500
269
+ },
270
+ {
271
+ "epoch": 1.0364390137464543,
272
+ "grad_norm": 2.947768211364746,
273
+ "learning_rate": 1.3213446858485028e-05,
274
+ "loss": 2.2225,
275
+ "step": 19000
276
+ },
277
+ {
278
+ "epoch": 1.063713724634519,
279
+ "grad_norm": 3.5084404945373535,
280
+ "learning_rate": 1.3030314151497358e-05,
281
+ "loss": 2.2222,
282
+ "step": 19500
283
+ },
284
+ {
285
+ "epoch": 1.0909884355225834,
286
+ "grad_norm": 3.5933291912078857,
287
+ "learning_rate": 1.284681444509689e-05,
288
+ "loss": 2.2169,
289
+ "step": 20000
290
+ },
291
+ {
292
+ "epoch": 1.118263146410648,
293
+ "grad_norm": 3.1414079666137695,
294
+ "learning_rate": 1.266331473869642e-05,
295
+ "loss": 2.2104,
296
+ "step": 20500
297
+ },
298
+ {
299
+ "epoch": 1.1455378572987127,
300
+ "grad_norm": 3.1075944900512695,
301
+ "learning_rate": 1.247981503229595e-05,
302
+ "loss": 2.1874,
303
+ "step": 21000
304
+ },
305
+ {
306
+ "epoch": 1.1728125681867771,
307
+ "grad_norm": 3.304654598236084,
308
+ "learning_rate": 1.229668232530828e-05,
309
+ "loss": 2.2282,
310
+ "step": 21500
311
+ },
312
+ {
313
+ "epoch": 1.2000872790748418,
314
+ "grad_norm": 2.92022705078125,
315
+ "learning_rate": 1.211318261890781e-05,
316
+ "loss": 2.2337,
317
+ "step": 22000
318
+ },
319
+ {
320
+ "epoch": 1.2273619899629065,
321
+ "grad_norm": 3.450542449951172,
322
+ "learning_rate": 1.1929682912507342e-05,
323
+ "loss": 2.1784,
324
+ "step": 22500
325
+ },
326
+ {
327
+ "epoch": 1.254636700850971,
328
+ "grad_norm": 2.8238916397094727,
329
+ "learning_rate": 1.1746183206106872e-05,
330
+ "loss": 2.2287,
331
+ "step": 23000
332
+ },
333
+ {
334
+ "epoch": 1.2819114117390356,
335
+ "grad_norm": 2.935835123062134,
336
+ "learning_rate": 1.1563050499119202e-05,
337
+ "loss": 2.2062,
338
+ "step": 23500
339
+ },
340
+ {
341
+ "epoch": 1.3091861226271002,
342
+ "grad_norm": 2.8934576511383057,
343
+ "learning_rate": 1.1379550792718732e-05,
344
+ "loss": 2.1836,
345
+ "step": 24000
346
+ },
347
+ {
348
+ "epoch": 1.3364608335151646,
349
+ "grad_norm": 2.884962797164917,
350
+ "learning_rate": 1.1196051086318263e-05,
351
+ "loss": 2.1632,
352
+ "step": 24500
353
+ },
354
+ {
355
+ "epoch": 1.3637355444032293,
356
+ "grad_norm": 3.163294792175293,
357
+ "learning_rate": 1.1012551379917793e-05,
358
+ "loss": 2.1715,
359
+ "step": 25000
360
+ },
361
+ {
362
+ "epoch": 1.391010255291294,
363
+ "grad_norm": 3.31972599029541,
364
+ "learning_rate": 1.0829051673517323e-05,
365
+ "loss": 2.1867,
366
+ "step": 25500
367
+ },
368
+ {
369
+ "epoch": 1.4182849661793586,
370
+ "grad_norm": 2.9578099250793457,
371
+ "learning_rate": 1.0645918966529653e-05,
372
+ "loss": 2.1851,
373
+ "step": 26000
374
+ },
375
+ {
376
+ "epoch": 1.445559677067423,
377
+ "grad_norm": 3.1862030029296875,
378
+ "learning_rate": 1.0462419260129184e-05,
379
+ "loss": 2.1805,
380
+ "step": 26500
381
+ },
382
+ {
383
+ "epoch": 1.4728343879554877,
384
+ "grad_norm": 3.1671643257141113,
385
+ "learning_rate": 1.0278919553728715e-05,
386
+ "loss": 2.1129,
387
+ "step": 27000
388
+ },
389
+ {
390
+ "epoch": 1.5001090988435521,
391
+ "grad_norm": 2.691154718399048,
392
+ "learning_rate": 1.0095786846741047e-05,
393
+ "loss": 2.1386,
394
+ "step": 27500
395
+ },
396
+ {
397
+ "epoch": 1.5273838097316168,
398
+ "grad_norm": 3.2530946731567383,
399
+ "learning_rate": 9.912287140340577e-06,
400
+ "loss": 2.2013,
401
+ "step": 28000
402
+ },
403
+ {
404
+ "epoch": 1.5546585206196815,
405
+ "grad_norm": 2.7970876693725586,
406
+ "learning_rate": 9.728787433940107e-06,
407
+ "loss": 2.1913,
408
+ "step": 28500
409
+ },
410
+ {
411
+ "epoch": 1.5819332315077461,
412
+ "grad_norm": 3.6721036434173584,
413
+ "learning_rate": 9.545287727539637e-06,
414
+ "loss": 2.1596,
415
+ "step": 29000
416
+ },
417
+ {
418
+ "epoch": 1.6092079423958106,
419
+ "grad_norm": 3.152144432067871,
420
+ "learning_rate": 9.361788021139167e-06,
421
+ "loss": 2.1486,
422
+ "step": 29500
423
+ },
424
+ {
425
+ "epoch": 1.6364826532838752,
426
+ "grad_norm": 3.174539089202881,
427
+ "learning_rate": 9.178288314738697e-06,
428
+ "loss": 2.169,
429
+ "step": 30000
430
+ },
431
+ {
432
+ "epoch": 1.6637573641719396,
433
+ "grad_norm": 2.7281394004821777,
434
+ "learning_rate": 8.994788608338227e-06,
435
+ "loss": 2.1478,
436
+ "step": 30500
437
+ },
438
+ {
439
+ "epoch": 1.6910320750600043,
440
+ "grad_norm": 3.2886815071105957,
441
+ "learning_rate": 8.811288901937757e-06,
442
+ "loss": 2.1372,
443
+ "step": 31000
444
+ },
445
+ {
446
+ "epoch": 1.718306785948069,
447
+ "grad_norm": 2.9683258533477783,
448
+ "learning_rate": 8.628156194950089e-06,
449
+ "loss": 2.1335,
450
+ "step": 31500
451
+ },
452
+ {
453
+ "epoch": 1.7455814968361336,
454
+ "grad_norm": 2.619202136993408,
455
+ "learning_rate": 8.444656488549619e-06,
456
+ "loss": 2.162,
457
+ "step": 32000
458
+ },
459
+ {
460
+ "epoch": 1.7728562077241983,
461
+ "grad_norm": 2.4466633796691895,
462
+ "learning_rate": 8.261156782149149e-06,
463
+ "loss": 2.1595,
464
+ "step": 32500
465
+ },
466
+ {
467
+ "epoch": 1.8001309186122627,
468
+ "grad_norm": 3.0279412269592285,
469
+ "learning_rate": 8.077657075748679e-06,
470
+ "loss": 2.1199,
471
+ "step": 33000
472
+ },
473
+ {
474
+ "epoch": 1.8274056295003271,
475
+ "grad_norm": 2.4983465671539307,
476
+ "learning_rate": 7.89415736934821e-06,
477
+ "loss": 2.1368,
478
+ "step": 33500
479
+ },
480
+ {
481
+ "epoch": 1.8546803403883918,
482
+ "grad_norm": 3.1389667987823486,
483
+ "learning_rate": 7.71102466236054e-06,
484
+ "loss": 2.1801,
485
+ "step": 34000
486
+ },
487
+ {
488
+ "epoch": 1.8819550512764565,
489
+ "grad_norm": 3.049088716506958,
490
+ "learning_rate": 7.5275249559600715e-06,
491
+ "loss": 2.151,
492
+ "step": 34500
493
+ },
494
+ {
495
+ "epoch": 1.9092297621645211,
496
+ "grad_norm": 2.899380922317505,
497
+ "learning_rate": 7.344025249559601e-06,
498
+ "loss": 2.1363,
499
+ "step": 35000
500
+ },
501
+ {
502
+ "epoch": 1.9365044730525858,
503
+ "grad_norm": 3.1051132678985596,
504
+ "learning_rate": 7.1605255431591316e-06,
505
+ "loss": 2.1412,
506
+ "step": 35500
507
+ },
508
+ {
509
+ "epoch": 1.9637791839406502,
510
+ "grad_norm": 2.9140512943267822,
511
+ "learning_rate": 6.9773928361714625e-06,
512
+ "loss": 2.1659,
513
+ "step": 36000
514
+ },
515
+ {
516
+ "epoch": 1.9910538948287149,
517
+ "grad_norm": 2.7826218605041504,
518
+ "learning_rate": 6.793893129770993e-06,
519
+ "loss": 2.1352,
520
+ "step": 36500
521
+ },
522
+ {
523
+ "epoch": 2.0183286057167793,
524
+ "grad_norm": 2.935887575149536,
525
+ "learning_rate": 6.610393423370523e-06,
526
+ "loss": 2.1212,
527
+ "step": 37000
528
+ },
529
+ {
530
+ "epoch": 2.045603316604844,
531
+ "grad_norm": 2.8185176849365234,
532
+ "learning_rate": 6.4268937169700525e-06,
533
+ "loss": 2.1403,
534
+ "step": 37500
535
+ },
536
+ {
537
+ "epoch": 2.0728780274929086,
538
+ "grad_norm": 3.057570219039917,
539
+ "learning_rate": 6.243761009982384e-06,
540
+ "loss": 2.1073,
541
+ "step": 38000
542
+ }
543
+ ],
544
+ "logging_steps": 500,
545
+ "max_steps": 54996,
546
+ "num_input_tokens_seen": 0,
547
+ "num_train_epochs": 3,
548
+ "save_steps": 1833,
549
+ "stateful_callbacks": {
550
+ "TrainerControl": {
551
+ "args": {
552
+ "should_epoch_stop": false,
553
+ "should_evaluate": false,
554
+ "should_log": false,
555
+ "should_save": true,
556
+ "should_training_stop": false
557
+ },
558
+ "attributes": {}
559
+ }
560
+ },
561
+ "total_flos": 1.0057914187776e+16,
562
+ "train_batch_size": 4,
563
+ "trial_name": null,
564
+ "trial_params": null
565
+ }
rmft/10e3/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
rmft/110e3/config.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "gpt2",
3
+ "activation_function": "gelu_new",
4
+ "architectures": [
5
+ "GPT2LMHeadModel"
6
+ ],
7
+ "attn_pdrop": 0.1,
8
+ "bos_token_id": 50256,
9
+ "embd_pdrop": 0.1,
10
+ "eos_token_id": 50256,
11
+ "initializer_range": 0.02,
12
+ "layer_norm_epsilon": 1e-05,
13
+ "model_type": "gpt2",
14
+ "n_ctx": 1024,
15
+ "n_embd": 768,
16
+ "n_head": 12,
17
+ "n_inner": null,
18
+ "n_layer": 12,
19
+ "n_positions": 1024,
20
+ "reorder_and_upcast_attn": false,
21
+ "resid_pdrop": 0.1,
22
+ "scale_attn_by_inverse_layer_idx": false,
23
+ "scale_attn_weights": true,
24
+ "summary_activation": null,
25
+ "summary_first_dropout": 0.1,
26
+ "summary_proj_to_labels": true,
27
+ "summary_type": "cls_index",
28
+ "summary_use_proj": true,
29
+ "task_specific_params": {
30
+ "text-generation": {
31
+ "do_sample": true,
32
+ "max_length": 50
33
+ }
34
+ },
35
+ "torch_dtype": "float32",
36
+ "transformers_version": "4.46.3",
37
+ "use_cache": true,
38
+ "vocab_size": 50257
39
+ }
rmft/110e3/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 50256,
4
+ "eos_token_id": 50256,
5
+ "transformers_version": "4.46.3"
6
+ }
rmft/110e3/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
rmft/110e3/special_tokens_map.json ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "bos_token": "<|endoftext|>",
3
+ "eos_token": "<|endoftext|>",
4
+ "unk_token": "<|endoftext|>"
5
+ }
rmft/110e3/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
rmft/110e3/tokenizer_config.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "added_tokens_decoder": {
4
+ "50256": {
5
+ "content": "<|endoftext|>",
6
+ "lstrip": false,
7
+ "normalized": true,
8
+ "rstrip": false,
9
+ "single_word": false,
10
+ "special": true
11
+ }
12
+ },
13
+ "bos_token": "<|endoftext|>",
14
+ "clean_up_tokenization_spaces": false,
15
+ "eos_token": "<|endoftext|>",
16
+ "model_max_length": 1024,
17
+ "return_tensors": "pt",
18
+ "tokenizer_class": "GPT2Tokenizer",
19
+ "unk_token": "<|endoftext|>"
20
+ }
rmft/110e3/trainer_state.json ADDED
@@ -0,0 +1,796 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 3.0,
5
+ "eval_steps": 500,
6
+ "global_step": 54996,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.027274710888064588,
13
+ "grad_norm": 8.132363319396973,
14
+ "learning_rate": 1.976e-05,
15
+ "loss": 3.5991,
16
+ "step": 500
17
+ },
18
+ {
19
+ "epoch": 0.054549421776129176,
20
+ "grad_norm": 7.099025249481201,
21
+ "learning_rate": 1.9818702290076336e-05,
22
+ "loss": 2.8754,
23
+ "step": 1000
24
+ },
25
+ {
26
+ "epoch": 0.08182413266419376,
27
+ "grad_norm": 7.820577621459961,
28
+ "learning_rate": 1.963520258367587e-05,
29
+ "loss": 2.6895,
30
+ "step": 1500
31
+ },
32
+ {
33
+ "epoch": 0.10909884355225835,
34
+ "grad_norm": 5.448477268218994,
35
+ "learning_rate": 1.94517028772754e-05,
36
+ "loss": 2.6494,
37
+ "step": 2000
38
+ },
39
+ {
40
+ "epoch": 0.13637355444032292,
41
+ "grad_norm": 6.391147613525391,
42
+ "learning_rate": 1.926820317087493e-05,
43
+ "loss": 2.5544,
44
+ "step": 2500
45
+ },
46
+ {
47
+ "epoch": 0.16364826532838753,
48
+ "grad_norm": 5.117228031158447,
49
+ "learning_rate": 1.908470346447446e-05,
50
+ "loss": 2.5431,
51
+ "step": 3000
52
+ },
53
+ {
54
+ "epoch": 0.1909229762164521,
55
+ "grad_norm": 4.08992338180542,
56
+ "learning_rate": 1.890120375807399e-05,
57
+ "loss": 2.5359,
58
+ "step": 3500
59
+ },
60
+ {
61
+ "epoch": 0.2181976871045167,
62
+ "grad_norm": 3.984097957611084,
63
+ "learning_rate": 1.871770405167352e-05,
64
+ "loss": 2.4834,
65
+ "step": 4000
66
+ },
67
+ {
68
+ "epoch": 0.24547239799258128,
69
+ "grad_norm": 3.8590118885040283,
70
+ "learning_rate": 1.853420434527305e-05,
71
+ "loss": 2.4899,
72
+ "step": 4500
73
+ },
74
+ {
75
+ "epoch": 0.27274710888064585,
76
+ "grad_norm": 4.541988372802734,
77
+ "learning_rate": 1.835070463887258e-05,
78
+ "loss": 2.4608,
79
+ "step": 5000
80
+ },
81
+ {
82
+ "epoch": 0.30002181976871045,
83
+ "grad_norm": 4.910805702209473,
84
+ "learning_rate": 1.816720493247211e-05,
85
+ "loss": 2.402,
86
+ "step": 5500
87
+ },
88
+ {
89
+ "epoch": 0.32729653065677505,
90
+ "grad_norm": 4.220065116882324,
91
+ "learning_rate": 1.798370522607164e-05,
92
+ "loss": 2.4483,
93
+ "step": 6000
94
+ },
95
+ {
96
+ "epoch": 0.3545712415448396,
97
+ "grad_norm": 3.9413557052612305,
98
+ "learning_rate": 1.780020551967117e-05,
99
+ "loss": 2.4271,
100
+ "step": 6500
101
+ },
102
+ {
103
+ "epoch": 0.3818459524329042,
104
+ "grad_norm": 2.890425443649292,
105
+ "learning_rate": 1.76167058132707e-05,
106
+ "loss": 2.3884,
107
+ "step": 7000
108
+ },
109
+ {
110
+ "epoch": 0.4091206633209688,
111
+ "grad_norm": 3.186286211013794,
112
+ "learning_rate": 1.743320610687023e-05,
113
+ "loss": 2.364,
114
+ "step": 7500
115
+ },
116
+ {
117
+ "epoch": 0.4363953742090334,
118
+ "grad_norm": 3.4042603969573975,
119
+ "learning_rate": 1.7249706400469762e-05,
120
+ "loss": 2.4374,
121
+ "step": 8000
122
+ },
123
+ {
124
+ "epoch": 0.46367008509709795,
125
+ "grad_norm": 2.714256763458252,
126
+ "learning_rate": 1.7066206694069292e-05,
127
+ "loss": 2.3535,
128
+ "step": 8500
129
+ },
130
+ {
131
+ "epoch": 0.49094479598516255,
132
+ "grad_norm": 3.9176571369171143,
133
+ "learning_rate": 1.6882706987668822e-05,
134
+ "loss": 2.379,
135
+ "step": 9000
136
+ },
137
+ {
138
+ "epoch": 0.5182195068732272,
139
+ "grad_norm": 3.3017940521240234,
140
+ "learning_rate": 1.669920728126835e-05,
141
+ "loss": 2.3246,
142
+ "step": 9500
143
+ },
144
+ {
145
+ "epoch": 0.5454942177612917,
146
+ "grad_norm": 3.9203879833221436,
147
+ "learning_rate": 1.651570757486788e-05,
148
+ "loss": 2.3452,
149
+ "step": 10000
150
+ },
151
+ {
152
+ "epoch": 0.5727689286493564,
153
+ "grad_norm": 3.474233627319336,
154
+ "learning_rate": 1.6332207868467413e-05,
155
+ "loss": 2.3192,
156
+ "step": 10500
157
+ },
158
+ {
159
+ "epoch": 0.6000436395374209,
160
+ "grad_norm": 3.47390079498291,
161
+ "learning_rate": 1.6148708162066943e-05,
162
+ "loss": 2.3775,
163
+ "step": 11000
164
+ },
165
+ {
166
+ "epoch": 0.6273183504254854,
167
+ "grad_norm": 3.5392141342163086,
168
+ "learning_rate": 1.5965208455666473e-05,
169
+ "loss": 2.3436,
170
+ "step": 11500
171
+ },
172
+ {
173
+ "epoch": 0.6545930613135501,
174
+ "grad_norm": 3.213120698928833,
175
+ "learning_rate": 1.5782075748678803e-05,
176
+ "loss": 2.2958,
177
+ "step": 12000
178
+ },
179
+ {
180
+ "epoch": 0.6818677722016147,
181
+ "grad_norm": 3.356034278869629,
182
+ "learning_rate": 1.5598576042278333e-05,
183
+ "loss": 2.3013,
184
+ "step": 12500
185
+ },
186
+ {
187
+ "epoch": 0.7091424830896792,
188
+ "grad_norm": 2.832211494445801,
189
+ "learning_rate": 1.5415076335877863e-05,
190
+ "loss": 2.3138,
191
+ "step": 13000
192
+ },
193
+ {
194
+ "epoch": 0.7364171939777439,
195
+ "grad_norm": 3.0613503456115723,
196
+ "learning_rate": 1.5231576629477394e-05,
197
+ "loss": 2.302,
198
+ "step": 13500
199
+ },
200
+ {
201
+ "epoch": 0.7636919048658084,
202
+ "grad_norm": 3.324085235595703,
203
+ "learning_rate": 1.5048076923076924e-05,
204
+ "loss": 2.3072,
205
+ "step": 14000
206
+ },
207
+ {
208
+ "epoch": 0.7909666157538731,
209
+ "grad_norm": 3.129357099533081,
210
+ "learning_rate": 1.4864577216676454e-05,
211
+ "loss": 2.3137,
212
+ "step": 14500
213
+ },
214
+ {
215
+ "epoch": 0.8182413266419376,
216
+ "grad_norm": 3.2217562198638916,
217
+ "learning_rate": 1.4681444509688784e-05,
218
+ "loss": 2.2899,
219
+ "step": 15000
220
+ },
221
+ {
222
+ "epoch": 0.8455160375300022,
223
+ "grad_norm": 3.6490132808685303,
224
+ "learning_rate": 1.4497944803288314e-05,
225
+ "loss": 2.3015,
226
+ "step": 15500
227
+ },
228
+ {
229
+ "epoch": 0.8727907484180668,
230
+ "grad_norm": 3.53132700920105,
231
+ "learning_rate": 1.4314445096887846e-05,
232
+ "loss": 2.2479,
233
+ "step": 16000
234
+ },
235
+ {
236
+ "epoch": 0.9000654593061314,
237
+ "grad_norm": 2.9956769943237305,
238
+ "learning_rate": 1.4130945390487376e-05,
239
+ "loss": 2.2878,
240
+ "step": 16500
241
+ },
242
+ {
243
+ "epoch": 0.9273401701941959,
244
+ "grad_norm": 3.027482271194458,
245
+ "learning_rate": 1.3947445684086906e-05,
246
+ "loss": 2.2656,
247
+ "step": 17000
248
+ },
249
+ {
250
+ "epoch": 0.9546148810822606,
251
+ "grad_norm": 2.76747465133667,
252
+ "learning_rate": 1.3763945977686436e-05,
253
+ "loss": 2.3085,
254
+ "step": 17500
255
+ },
256
+ {
257
+ "epoch": 0.9818895919703251,
258
+ "grad_norm": 3.367910385131836,
259
+ "learning_rate": 1.3580446271285968e-05,
260
+ "loss": 2.2726,
261
+ "step": 18000
262
+ },
263
+ {
264
+ "epoch": 1.0091643028583897,
265
+ "grad_norm": 3.19476056098938,
266
+ "learning_rate": 1.3396946564885498e-05,
267
+ "loss": 2.2459,
268
+ "step": 18500
269
+ },
270
+ {
271
+ "epoch": 1.0364390137464543,
272
+ "grad_norm": 2.947768211364746,
273
+ "learning_rate": 1.3213446858485028e-05,
274
+ "loss": 2.2225,
275
+ "step": 19000
276
+ },
277
+ {
278
+ "epoch": 1.063713724634519,
279
+ "grad_norm": 3.5084404945373535,
280
+ "learning_rate": 1.3030314151497358e-05,
281
+ "loss": 2.2222,
282
+ "step": 19500
283
+ },
284
+ {
285
+ "epoch": 1.0909884355225834,
286
+ "grad_norm": 3.5933291912078857,
287
+ "learning_rate": 1.284681444509689e-05,
288
+ "loss": 2.2169,
289
+ "step": 20000
290
+ },
291
+ {
292
+ "epoch": 1.118263146410648,
293
+ "grad_norm": 3.1414079666137695,
294
+ "learning_rate": 1.266331473869642e-05,
295
+ "loss": 2.2104,
296
+ "step": 20500
297
+ },
298
+ {
299
+ "epoch": 1.1455378572987127,
300
+ "grad_norm": 3.1075944900512695,
301
+ "learning_rate": 1.247981503229595e-05,
302
+ "loss": 2.1874,
303
+ "step": 21000
304
+ },
305
+ {
306
+ "epoch": 1.1728125681867771,
307
+ "grad_norm": 3.304654598236084,
308
+ "learning_rate": 1.229668232530828e-05,
309
+ "loss": 2.2282,
310
+ "step": 21500
311
+ },
312
+ {
313
+ "epoch": 1.2000872790748418,
314
+ "grad_norm": 2.92022705078125,
315
+ "learning_rate": 1.211318261890781e-05,
316
+ "loss": 2.2337,
317
+ "step": 22000
318
+ },
319
+ {
320
+ "epoch": 1.2273619899629065,
321
+ "grad_norm": 3.450542449951172,
322
+ "learning_rate": 1.1929682912507342e-05,
323
+ "loss": 2.1784,
324
+ "step": 22500
325
+ },
326
+ {
327
+ "epoch": 1.254636700850971,
328
+ "grad_norm": 2.8238916397094727,
329
+ "learning_rate": 1.1746183206106872e-05,
330
+ "loss": 2.2287,
331
+ "step": 23000
332
+ },
333
+ {
334
+ "epoch": 1.2819114117390356,
335
+ "grad_norm": 2.935835123062134,
336
+ "learning_rate": 1.1563050499119202e-05,
337
+ "loss": 2.2062,
338
+ "step": 23500
339
+ },
340
+ {
341
+ "epoch": 1.3091861226271002,
342
+ "grad_norm": 2.8934576511383057,
343
+ "learning_rate": 1.1379550792718732e-05,
344
+ "loss": 2.1836,
345
+ "step": 24000
346
+ },
347
+ {
348
+ "epoch": 1.3364608335151646,
349
+ "grad_norm": 2.884962797164917,
350
+ "learning_rate": 1.1196051086318263e-05,
351
+ "loss": 2.1632,
352
+ "step": 24500
353
+ },
354
+ {
355
+ "epoch": 1.3637355444032293,
356
+ "grad_norm": 3.163294792175293,
357
+ "learning_rate": 1.1012551379917793e-05,
358
+ "loss": 2.1715,
359
+ "step": 25000
360
+ },
361
+ {
362
+ "epoch": 1.391010255291294,
363
+ "grad_norm": 3.31972599029541,
364
+ "learning_rate": 1.0829051673517323e-05,
365
+ "loss": 2.1867,
366
+ "step": 25500
367
+ },
368
+ {
369
+ "epoch": 1.4182849661793586,
370
+ "grad_norm": 2.9578099250793457,
371
+ "learning_rate": 1.0645918966529653e-05,
372
+ "loss": 2.1851,
373
+ "step": 26000
374
+ },
375
+ {
376
+ "epoch": 1.445559677067423,
377
+ "grad_norm": 3.1862030029296875,
378
+ "learning_rate": 1.0462419260129184e-05,
379
+ "loss": 2.1805,
380
+ "step": 26500
381
+ },
382
+ {
383
+ "epoch": 1.4728343879554877,
384
+ "grad_norm": 3.1671643257141113,
385
+ "learning_rate": 1.0278919553728715e-05,
386
+ "loss": 2.1129,
387
+ "step": 27000
388
+ },
389
+ {
390
+ "epoch": 1.5001090988435521,
391
+ "grad_norm": 2.691154718399048,
392
+ "learning_rate": 1.0095786846741047e-05,
393
+ "loss": 2.1386,
394
+ "step": 27500
395
+ },
396
+ {
397
+ "epoch": 1.5273838097316168,
398
+ "grad_norm": 3.2530946731567383,
399
+ "learning_rate": 9.912287140340577e-06,
400
+ "loss": 2.2013,
401
+ "step": 28000
402
+ },
403
+ {
404
+ "epoch": 1.5546585206196815,
405
+ "grad_norm": 2.7970876693725586,
406
+ "learning_rate": 9.728787433940107e-06,
407
+ "loss": 2.1913,
408
+ "step": 28500
409
+ },
410
+ {
411
+ "epoch": 1.5819332315077461,
412
+ "grad_norm": 3.6721036434173584,
413
+ "learning_rate": 9.545287727539637e-06,
414
+ "loss": 2.1596,
415
+ "step": 29000
416
+ },
417
+ {
418
+ "epoch": 1.6092079423958106,
419
+ "grad_norm": 3.152144432067871,
420
+ "learning_rate": 9.361788021139167e-06,
421
+ "loss": 2.1486,
422
+ "step": 29500
423
+ },
424
+ {
425
+ "epoch": 1.6364826532838752,
426
+ "grad_norm": 3.174539089202881,
427
+ "learning_rate": 9.178288314738697e-06,
428
+ "loss": 2.169,
429
+ "step": 30000
430
+ },
431
+ {
432
+ "epoch": 1.6637573641719396,
433
+ "grad_norm": 2.7281394004821777,
434
+ "learning_rate": 8.994788608338227e-06,
435
+ "loss": 2.1478,
436
+ "step": 30500
437
+ },
438
+ {
439
+ "epoch": 1.6910320750600043,
440
+ "grad_norm": 3.2886815071105957,
441
+ "learning_rate": 8.811288901937757e-06,
442
+ "loss": 2.1372,
443
+ "step": 31000
444
+ },
445
+ {
446
+ "epoch": 1.718306785948069,
447
+ "grad_norm": 2.9683258533477783,
448
+ "learning_rate": 8.628156194950089e-06,
449
+ "loss": 2.1335,
450
+ "step": 31500
451
+ },
452
+ {
453
+ "epoch": 1.7455814968361336,
454
+ "grad_norm": 2.619202136993408,
455
+ "learning_rate": 8.444656488549619e-06,
456
+ "loss": 2.162,
457
+ "step": 32000
458
+ },
459
+ {
460
+ "epoch": 1.7728562077241983,
461
+ "grad_norm": 2.4466633796691895,
462
+ "learning_rate": 8.261156782149149e-06,
463
+ "loss": 2.1595,
464
+ "step": 32500
465
+ },
466
+ {
467
+ "epoch": 1.8001309186122627,
468
+ "grad_norm": 3.0279412269592285,
469
+ "learning_rate": 8.077657075748679e-06,
470
+ "loss": 2.1199,
471
+ "step": 33000
472
+ },
473
+ {
474
+ "epoch": 1.8274056295003271,
475
+ "grad_norm": 2.4983465671539307,
476
+ "learning_rate": 7.89415736934821e-06,
477
+ "loss": 2.1368,
478
+ "step": 33500
479
+ },
480
+ {
481
+ "epoch": 1.8546803403883918,
482
+ "grad_norm": 3.1389667987823486,
483
+ "learning_rate": 7.71102466236054e-06,
484
+ "loss": 2.1801,
485
+ "step": 34000
486
+ },
487
+ {
488
+ "epoch": 1.8819550512764565,
489
+ "grad_norm": 3.049088716506958,
490
+ "learning_rate": 7.5275249559600715e-06,
491
+ "loss": 2.151,
492
+ "step": 34500
493
+ },
494
+ {
495
+ "epoch": 1.9092297621645211,
496
+ "grad_norm": 2.899380922317505,
497
+ "learning_rate": 7.344025249559601e-06,
498
+ "loss": 2.1363,
499
+ "step": 35000
500
+ },
501
+ {
502
+ "epoch": 1.9365044730525858,
503
+ "grad_norm": 3.1051132678985596,
504
+ "learning_rate": 7.1605255431591316e-06,
505
+ "loss": 2.1412,
506
+ "step": 35500
507
+ },
508
+ {
509
+ "epoch": 1.9637791839406502,
510
+ "grad_norm": 2.9140512943267822,
511
+ "learning_rate": 6.9773928361714625e-06,
512
+ "loss": 2.1659,
513
+ "step": 36000
514
+ },
515
+ {
516
+ "epoch": 1.9910538948287149,
517
+ "grad_norm": 2.7826218605041504,
518
+ "learning_rate": 6.793893129770993e-06,
519
+ "loss": 2.1352,
520
+ "step": 36500
521
+ },
522
+ {
523
+ "epoch": 2.0183286057167793,
524
+ "grad_norm": 2.935887575149536,
525
+ "learning_rate": 6.610393423370523e-06,
526
+ "loss": 2.1212,
527
+ "step": 37000
528
+ },
529
+ {
530
+ "epoch": 2.045603316604844,
531
+ "grad_norm": 2.8185176849365234,
532
+ "learning_rate": 6.4268937169700525e-06,
533
+ "loss": 2.1403,
534
+ "step": 37500
535
+ },
536
+ {
537
+ "epoch": 2.0728780274929086,
538
+ "grad_norm": 3.057570219039917,
539
+ "learning_rate": 6.243761009982384e-06,
540
+ "loss": 2.1073,
541
+ "step": 38000
542
+ },
543
+ {
544
+ "epoch": 2.1001527383809733,
545
+ "grad_norm": 2.8996992111206055,
546
+ "learning_rate": 6.060261303581914e-06,
547
+ "loss": 2.115,
548
+ "step": 38500
549
+ },
550
+ {
551
+ "epoch": 2.127427449269038,
552
+ "grad_norm": 2.9682915210723877,
553
+ "learning_rate": 5.876761597181445e-06,
554
+ "loss": 2.1284,
555
+ "step": 39000
556
+ },
557
+ {
558
+ "epoch": 2.154702160157102,
559
+ "grad_norm": 3.1985812187194824,
560
+ "learning_rate": 5.693261890780975e-06,
561
+ "loss": 2.0682,
562
+ "step": 39500
563
+ },
564
+ {
565
+ "epoch": 2.181976871045167,
566
+ "grad_norm": 2.996978282928467,
567
+ "learning_rate": 5.510129183793307e-06,
568
+ "loss": 2.1098,
569
+ "step": 40000
570
+ },
571
+ {
572
+ "epoch": 2.2092515819332315,
573
+ "grad_norm": 3.008801221847534,
574
+ "learning_rate": 5.326629477392836e-06,
575
+ "loss": 2.0958,
576
+ "step": 40500
577
+ },
578
+ {
579
+ "epoch": 2.236526292821296,
580
+ "grad_norm": 3.4560887813568115,
581
+ "learning_rate": 5.143129770992367e-06,
582
+ "loss": 2.1124,
583
+ "step": 41000
584
+ },
585
+ {
586
+ "epoch": 2.2638010037093608,
587
+ "grad_norm": 2.72576642036438,
588
+ "learning_rate": 4.959630064591897e-06,
589
+ "loss": 2.0714,
590
+ "step": 41500
591
+ },
592
+ {
593
+ "epoch": 2.2910757145974254,
594
+ "grad_norm": 3.265460968017578,
595
+ "learning_rate": 4.776497357604229e-06,
596
+ "loss": 2.1251,
597
+ "step": 42000
598
+ },
599
+ {
600
+ "epoch": 2.31835042548549,
601
+ "grad_norm": 3.2544667720794678,
602
+ "learning_rate": 4.592997651203758e-06,
603
+ "loss": 2.0752,
604
+ "step": 42500
605
+ },
606
+ {
607
+ "epoch": 2.3456251363735543,
608
+ "grad_norm": 2.7102246284484863,
609
+ "learning_rate": 4.409497944803289e-06,
610
+ "loss": 2.1087,
611
+ "step": 43000
612
+ },
613
+ {
614
+ "epoch": 2.372899847261619,
615
+ "grad_norm": 2.5479111671447754,
616
+ "learning_rate": 4.225998238402819e-06,
617
+ "loss": 2.0828,
618
+ "step": 43500
619
+ },
620
+ {
621
+ "epoch": 2.4001745581496836,
622
+ "grad_norm": 3.278614044189453,
623
+ "learning_rate": 4.042498532002349e-06,
624
+ "loss": 2.1031,
625
+ "step": 44000
626
+ },
627
+ {
628
+ "epoch": 2.4274492690377483,
629
+ "grad_norm": 2.89288330078125,
630
+ "learning_rate": 3.859365825014681e-06,
631
+ "loss": 2.0895,
632
+ "step": 44500
633
+ },
634
+ {
635
+ "epoch": 2.454723979925813,
636
+ "grad_norm": 4.075788497924805,
637
+ "learning_rate": 3.6758661186142102e-06,
638
+ "loss": 2.1068,
639
+ "step": 45000
640
+ },
641
+ {
642
+ "epoch": 2.481998690813877,
643
+ "grad_norm": 3.624413251876831,
644
+ "learning_rate": 3.4923664122137406e-06,
645
+ "loss": 2.0892,
646
+ "step": 45500
647
+ },
648
+ {
649
+ "epoch": 2.509273401701942,
650
+ "grad_norm": 3.3168036937713623,
651
+ "learning_rate": 3.308866705813271e-06,
652
+ "loss": 2.0817,
653
+ "step": 46000
654
+ },
655
+ {
656
+ "epoch": 2.5365481125900065,
657
+ "grad_norm": 3.027834892272949,
658
+ "learning_rate": 3.125733998825602e-06,
659
+ "loss": 2.0769,
660
+ "step": 46500
661
+ },
662
+ {
663
+ "epoch": 2.563822823478071,
664
+ "grad_norm": 3.3732242584228516,
665
+ "learning_rate": 2.9422342924251325e-06,
666
+ "loss": 2.09,
667
+ "step": 47000
668
+ },
669
+ {
670
+ "epoch": 2.5910975343661358,
671
+ "grad_norm": 3.3527138233184814,
672
+ "learning_rate": 2.758734586024663e-06,
673
+ "loss": 2.0494,
674
+ "step": 47500
675
+ },
676
+ {
677
+ "epoch": 2.6183722452542004,
678
+ "grad_norm": 3.75742769241333,
679
+ "learning_rate": 2.575234879624193e-06,
680
+ "loss": 2.088,
681
+ "step": 48000
682
+ },
683
+ {
684
+ "epoch": 2.645646956142265,
685
+ "grad_norm": 3.155179977416992,
686
+ "learning_rate": 2.392102172636524e-06,
687
+ "loss": 2.0849,
688
+ "step": 48500
689
+ },
690
+ {
691
+ "epoch": 2.6729216670303293,
692
+ "grad_norm": 3.3274827003479004,
693
+ "learning_rate": 2.2086024662360543e-06,
694
+ "loss": 2.1262,
695
+ "step": 49000
696
+ },
697
+ {
698
+ "epoch": 2.700196377918394,
699
+ "grad_norm": 3.319843292236328,
700
+ "learning_rate": 2.0251027598355843e-06,
701
+ "loss": 2.0734,
702
+ "step": 49500
703
+ },
704
+ {
705
+ "epoch": 2.7274710888064586,
706
+ "grad_norm": 2.9508001804351807,
707
+ "learning_rate": 1.8416030534351145e-06,
708
+ "loss": 2.0847,
709
+ "step": 50000
710
+ },
711
+ {
712
+ "epoch": 2.7547457996945233,
713
+ "grad_norm": 2.934704303741455,
714
+ "learning_rate": 1.6584703464474459e-06,
715
+ "loss": 2.1143,
716
+ "step": 50500
717
+ },
718
+ {
719
+ "epoch": 2.782020510582588,
720
+ "grad_norm": 2.870739221572876,
721
+ "learning_rate": 1.4749706400469759e-06,
722
+ "loss": 2.0671,
723
+ "step": 51000
724
+ },
725
+ {
726
+ "epoch": 2.809295221470652,
727
+ "grad_norm": 3.7797598838806152,
728
+ "learning_rate": 1.2914709336465061e-06,
729
+ "loss": 2.1031,
730
+ "step": 51500
731
+ },
732
+ {
733
+ "epoch": 2.8365699323587172,
734
+ "grad_norm": 2.9997830390930176,
735
+ "learning_rate": 1.1079712272460365e-06,
736
+ "loss": 2.0789,
737
+ "step": 52000
738
+ },
739
+ {
740
+ "epoch": 2.8638446432467815,
741
+ "grad_norm": 2.9617133140563965,
742
+ "learning_rate": 9.248385202583676e-07,
743
+ "loss": 2.122,
744
+ "step": 52500
745
+ },
746
+ {
747
+ "epoch": 2.891119354134846,
748
+ "grad_norm": 2.7891688346862793,
749
+ "learning_rate": 7.413388138578979e-07,
750
+ "loss": 2.0962,
751
+ "step": 53000
752
+ },
753
+ {
754
+ "epoch": 2.9183940650229108,
755
+ "grad_norm": 3.2359707355499268,
756
+ "learning_rate": 5.578391074574281e-07,
757
+ "loss": 2.1573,
758
+ "step": 53500
759
+ },
760
+ {
761
+ "epoch": 2.9456687759109754,
762
+ "grad_norm": 3.748823642730713,
763
+ "learning_rate": 3.7433940105695836e-07,
764
+ "loss": 2.0932,
765
+ "step": 54000
766
+ },
767
+ {
768
+ "epoch": 2.97294348679904,
769
+ "grad_norm": 3.199406147003174,
770
+ "learning_rate": 1.9083969465648858e-07,
771
+ "loss": 2.1349,
772
+ "step": 54500
773
+ }
774
+ ],
775
+ "logging_steps": 500,
776
+ "max_steps": 54996,
777
+ "num_input_tokens_seen": 0,
778
+ "num_train_epochs": 3,
779
+ "save_steps": 1833,
780
+ "stateful_callbacks": {
781
+ "TrainerControl": {
782
+ "args": {
783
+ "should_epoch_stop": false,
784
+ "should_evaluate": false,
785
+ "should_log": false,
786
+ "should_save": true,
787
+ "should_training_stop": true
788
+ },
789
+ "attributes": {}
790
+ }
791
+ },
792
+ "total_flos": 1.4370016591872e+16,
793
+ "train_batch_size": 4,
794
+ "trial_name": null,
795
+ "trial_params": null
796
+ }
rmft/110e3/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
rmft/20e1/config.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "gpt2",
3
+ "activation_function": "gelu_new",
4
+ "architectures": [
5
+ "GPT2LMHeadModel"
6
+ ],
7
+ "attn_pdrop": 0.1,
8
+ "bos_token_id": 50256,
9
+ "embd_pdrop": 0.1,
10
+ "eos_token_id": 50256,
11
+ "initializer_range": 0.02,
12
+ "layer_norm_epsilon": 1e-05,
13
+ "model_type": "gpt2",
14
+ "n_ctx": 1024,
15
+ "n_embd": 768,
16
+ "n_head": 12,
17
+ "n_inner": null,
18
+ "n_layer": 12,
19
+ "n_positions": 1024,
20
+ "reorder_and_upcast_attn": false,
21
+ "resid_pdrop": 0.1,
22
+ "scale_attn_by_inverse_layer_idx": false,
23
+ "scale_attn_weights": true,
24
+ "summary_activation": null,
25
+ "summary_first_dropout": 0.1,
26
+ "summary_proj_to_labels": true,
27
+ "summary_type": "cls_index",
28
+ "summary_use_proj": true,
29
+ "task_specific_params": {
30
+ "text-generation": {
31
+ "do_sample": true,
32
+ "max_length": 50
33
+ }
34
+ },
35
+ "torch_dtype": "float32",
36
+ "transformers_version": "4.46.3",
37
+ "use_cache": true,
38
+ "vocab_size": 50257
39
+ }
rmft/20e1/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 50256,
4
+ "eos_token_id": 50256,
5
+ "transformers_version": "4.46.3"
6
+ }
rmft/20e1/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
rmft/20e1/special_tokens_map.json ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "bos_token": "<|endoftext|>",
3
+ "eos_token": "<|endoftext|>",
4
+ "unk_token": "<|endoftext|>"
5
+ }
rmft/20e1/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
rmft/20e1/tokenizer_config.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "added_tokens_decoder": {
4
+ "50256": {
5
+ "content": "<|endoftext|>",
6
+ "lstrip": false,
7
+ "normalized": true,
8
+ "rstrip": false,
9
+ "single_word": false,
10
+ "special": true
11
+ }
12
+ },
13
+ "bos_token": "<|endoftext|>",
14
+ "clean_up_tokenization_spaces": false,
15
+ "eos_token": "<|endoftext|>",
16
+ "model_max_length": 1024,
17
+ "return_tensors": "pt",
18
+ "tokenizer_class": "GPT2Tokenizer",
19
+ "unk_token": "<|endoftext|>"
20
+ }
rmft/20e1/trainer_state.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 0.19997818023128955,
5
+ "eval_steps": 500,
6
+ "global_step": 3666,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.027274710888064588,
13
+ "grad_norm": 8.132363319396973,
14
+ "learning_rate": 1.976e-05,
15
+ "loss": 3.5991,
16
+ "step": 500
17
+ },
18
+ {
19
+ "epoch": 0.054549421776129176,
20
+ "grad_norm": 7.099025249481201,
21
+ "learning_rate": 1.9818702290076336e-05,
22
+ "loss": 2.8754,
23
+ "step": 1000
24
+ },
25
+ {
26
+ "epoch": 0.08182413266419376,
27
+ "grad_norm": 7.820577621459961,
28
+ "learning_rate": 1.963520258367587e-05,
29
+ "loss": 2.6895,
30
+ "step": 1500
31
+ },
32
+ {
33
+ "epoch": 0.10909884355225835,
34
+ "grad_norm": 5.448477268218994,
35
+ "learning_rate": 1.94517028772754e-05,
36
+ "loss": 2.6494,
37
+ "step": 2000
38
+ },
39
+ {
40
+ "epoch": 0.13637355444032292,
41
+ "grad_norm": 6.391147613525391,
42
+ "learning_rate": 1.926820317087493e-05,
43
+ "loss": 2.5544,
44
+ "step": 2500
45
+ },
46
+ {
47
+ "epoch": 0.16364826532838753,
48
+ "grad_norm": 5.117228031158447,
49
+ "learning_rate": 1.908470346447446e-05,
50
+ "loss": 2.5431,
51
+ "step": 3000
52
+ },
53
+ {
54
+ "epoch": 0.1909229762164521,
55
+ "grad_norm": 4.08992338180542,
56
+ "learning_rate": 1.890120375807399e-05,
57
+ "loss": 2.5359,
58
+ "step": 3500
59
+ }
60
+ ],
61
+ "logging_steps": 500,
62
+ "max_steps": 54996,
63
+ "num_input_tokens_seen": 0,
64
+ "num_train_epochs": 3,
65
+ "save_steps": 1833,
66
+ "stateful_callbacks": {
67
+ "TrainerControl": {
68
+ "args": {
69
+ "should_epoch_stop": false,
70
+ "should_evaluate": false,
71
+ "should_log": false,
72
+ "should_save": true,
73
+ "should_training_stop": false
74
+ },
75
+ "attributes": {}
76
+ }
77
+ },
78
+ "total_flos": 957896589312000.0,
79
+ "train_batch_size": 4,
80
+ "trial_name": null,
81
+ "trial_params": null
82
+ }
rmft/20e1/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
rmft/20e2/config.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "gpt2",
3
+ "activation_function": "gelu_new",
4
+ "architectures": [
5
+ "GPT2LMHeadModel"
6
+ ],
7
+ "attn_pdrop": 0.1,
8
+ "bos_token_id": 50256,
9
+ "embd_pdrop": 0.1,
10
+ "eos_token_id": 50256,
11
+ "initializer_range": 0.02,
12
+ "layer_norm_epsilon": 1e-05,
13
+ "model_type": "gpt2",
14
+ "n_ctx": 1024,
15
+ "n_embd": 768,
16
+ "n_head": 12,
17
+ "n_inner": null,
18
+ "n_layer": 12,
19
+ "n_positions": 1024,
20
+ "reorder_and_upcast_attn": false,
21
+ "resid_pdrop": 0.1,
22
+ "scale_attn_by_inverse_layer_idx": false,
23
+ "scale_attn_weights": true,
24
+ "summary_activation": null,
25
+ "summary_first_dropout": 0.1,
26
+ "summary_proj_to_labels": true,
27
+ "summary_type": "cls_index",
28
+ "summary_use_proj": true,
29
+ "task_specific_params": {
30
+ "text-generation": {
31
+ "do_sample": true,
32
+ "max_length": 50
33
+ }
34
+ },
35
+ "torch_dtype": "float32",
36
+ "transformers_version": "4.46.3",
37
+ "use_cache": true,
38
+ "vocab_size": 50257
39
+ }
rmft/20e2/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 50256,
4
+ "eos_token_id": 50256,
5
+ "transformers_version": "4.46.3"
6
+ }
rmft/20e2/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
rmft/20e2/special_tokens_map.json ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "bos_token": "<|endoftext|>",
3
+ "eos_token": "<|endoftext|>",
4
+ "unk_token": "<|endoftext|>"
5
+ }
rmft/20e2/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
rmft/20e2/tokenizer_config.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "added_tokens_decoder": {
4
+ "50256": {
5
+ "content": "<|endoftext|>",
6
+ "lstrip": false,
7
+ "normalized": true,
8
+ "rstrip": false,
9
+ "single_word": false,
10
+ "special": true
11
+ }
12
+ },
13
+ "bos_token": "<|endoftext|>",
14
+ "clean_up_tokenization_spaces": false,
15
+ "eos_token": "<|endoftext|>",
16
+ "model_max_length": 1024,
17
+ "return_tensors": "pt",
18
+ "tokenizer_class": "GPT2Tokenizer",
19
+ "unk_token": "<|endoftext|>"
20
+ }
rmft/20e2/trainer_state.json ADDED
@@ -0,0 +1,334 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 1.1998690813877373,
5
+ "eval_steps": 500,
6
+ "global_step": 21996,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.027274710888064588,
13
+ "grad_norm": 8.132363319396973,
14
+ "learning_rate": 1.976e-05,
15
+ "loss": 3.5991,
16
+ "step": 500
17
+ },
18
+ {
19
+ "epoch": 0.054549421776129176,
20
+ "grad_norm": 7.099025249481201,
21
+ "learning_rate": 1.9818702290076336e-05,
22
+ "loss": 2.8754,
23
+ "step": 1000
24
+ },
25
+ {
26
+ "epoch": 0.08182413266419376,
27
+ "grad_norm": 7.820577621459961,
28
+ "learning_rate": 1.963520258367587e-05,
29
+ "loss": 2.6895,
30
+ "step": 1500
31
+ },
32
+ {
33
+ "epoch": 0.10909884355225835,
34
+ "grad_norm": 5.448477268218994,
35
+ "learning_rate": 1.94517028772754e-05,
36
+ "loss": 2.6494,
37
+ "step": 2000
38
+ },
39
+ {
40
+ "epoch": 0.13637355444032292,
41
+ "grad_norm": 6.391147613525391,
42
+ "learning_rate": 1.926820317087493e-05,
43
+ "loss": 2.5544,
44
+ "step": 2500
45
+ },
46
+ {
47
+ "epoch": 0.16364826532838753,
48
+ "grad_norm": 5.117228031158447,
49
+ "learning_rate": 1.908470346447446e-05,
50
+ "loss": 2.5431,
51
+ "step": 3000
52
+ },
53
+ {
54
+ "epoch": 0.1909229762164521,
55
+ "grad_norm": 4.08992338180542,
56
+ "learning_rate": 1.890120375807399e-05,
57
+ "loss": 2.5359,
58
+ "step": 3500
59
+ },
60
+ {
61
+ "epoch": 0.2181976871045167,
62
+ "grad_norm": 3.984097957611084,
63
+ "learning_rate": 1.871770405167352e-05,
64
+ "loss": 2.4834,
65
+ "step": 4000
66
+ },
67
+ {
68
+ "epoch": 0.24547239799258128,
69
+ "grad_norm": 3.8590118885040283,
70
+ "learning_rate": 1.853420434527305e-05,
71
+ "loss": 2.4899,
72
+ "step": 4500
73
+ },
74
+ {
75
+ "epoch": 0.27274710888064585,
76
+ "grad_norm": 4.541988372802734,
77
+ "learning_rate": 1.835070463887258e-05,
78
+ "loss": 2.4608,
79
+ "step": 5000
80
+ },
81
+ {
82
+ "epoch": 0.30002181976871045,
83
+ "grad_norm": 4.910805702209473,
84
+ "learning_rate": 1.816720493247211e-05,
85
+ "loss": 2.402,
86
+ "step": 5500
87
+ },
88
+ {
89
+ "epoch": 0.32729653065677505,
90
+ "grad_norm": 4.220065116882324,
91
+ "learning_rate": 1.798370522607164e-05,
92
+ "loss": 2.4483,
93
+ "step": 6000
94
+ },
95
+ {
96
+ "epoch": 0.3545712415448396,
97
+ "grad_norm": 3.9413557052612305,
98
+ "learning_rate": 1.780020551967117e-05,
99
+ "loss": 2.4271,
100
+ "step": 6500
101
+ },
102
+ {
103
+ "epoch": 0.3818459524329042,
104
+ "grad_norm": 2.890425443649292,
105
+ "learning_rate": 1.76167058132707e-05,
106
+ "loss": 2.3884,
107
+ "step": 7000
108
+ },
109
+ {
110
+ "epoch": 0.4091206633209688,
111
+ "grad_norm": 3.186286211013794,
112
+ "learning_rate": 1.743320610687023e-05,
113
+ "loss": 2.364,
114
+ "step": 7500
115
+ },
116
+ {
117
+ "epoch": 0.4363953742090334,
118
+ "grad_norm": 3.4042603969573975,
119
+ "learning_rate": 1.7249706400469762e-05,
120
+ "loss": 2.4374,
121
+ "step": 8000
122
+ },
123
+ {
124
+ "epoch": 0.46367008509709795,
125
+ "grad_norm": 2.714256763458252,
126
+ "learning_rate": 1.7066206694069292e-05,
127
+ "loss": 2.3535,
128
+ "step": 8500
129
+ },
130
+ {
131
+ "epoch": 0.49094479598516255,
132
+ "grad_norm": 3.9176571369171143,
133
+ "learning_rate": 1.6882706987668822e-05,
134
+ "loss": 2.379,
135
+ "step": 9000
136
+ },
137
+ {
138
+ "epoch": 0.5182195068732272,
139
+ "grad_norm": 3.3017940521240234,
140
+ "learning_rate": 1.669920728126835e-05,
141
+ "loss": 2.3246,
142
+ "step": 9500
143
+ },
144
+ {
145
+ "epoch": 0.5454942177612917,
146
+ "grad_norm": 3.9203879833221436,
147
+ "learning_rate": 1.651570757486788e-05,
148
+ "loss": 2.3452,
149
+ "step": 10000
150
+ },
151
+ {
152
+ "epoch": 0.5727689286493564,
153
+ "grad_norm": 3.474233627319336,
154
+ "learning_rate": 1.6332207868467413e-05,
155
+ "loss": 2.3192,
156
+ "step": 10500
157
+ },
158
+ {
159
+ "epoch": 0.6000436395374209,
160
+ "grad_norm": 3.47390079498291,
161
+ "learning_rate": 1.6148708162066943e-05,
162
+ "loss": 2.3775,
163
+ "step": 11000
164
+ },
165
+ {
166
+ "epoch": 0.6273183504254854,
167
+ "grad_norm": 3.5392141342163086,
168
+ "learning_rate": 1.5965208455666473e-05,
169
+ "loss": 2.3436,
170
+ "step": 11500
171
+ },
172
+ {
173
+ "epoch": 0.6545930613135501,
174
+ "grad_norm": 3.213120698928833,
175
+ "learning_rate": 1.5782075748678803e-05,
176
+ "loss": 2.2958,
177
+ "step": 12000
178
+ },
179
+ {
180
+ "epoch": 0.6818677722016147,
181
+ "grad_norm": 3.356034278869629,
182
+ "learning_rate": 1.5598576042278333e-05,
183
+ "loss": 2.3013,
184
+ "step": 12500
185
+ },
186
+ {
187
+ "epoch": 0.7091424830896792,
188
+ "grad_norm": 2.832211494445801,
189
+ "learning_rate": 1.5415076335877863e-05,
190
+ "loss": 2.3138,
191
+ "step": 13000
192
+ },
193
+ {
194
+ "epoch": 0.7364171939777439,
195
+ "grad_norm": 3.0613503456115723,
196
+ "learning_rate": 1.5231576629477394e-05,
197
+ "loss": 2.302,
198
+ "step": 13500
199
+ },
200
+ {
201
+ "epoch": 0.7636919048658084,
202
+ "grad_norm": 3.324085235595703,
203
+ "learning_rate": 1.5048076923076924e-05,
204
+ "loss": 2.3072,
205
+ "step": 14000
206
+ },
207
+ {
208
+ "epoch": 0.7909666157538731,
209
+ "grad_norm": 3.129357099533081,
210
+ "learning_rate": 1.4864577216676454e-05,
211
+ "loss": 2.3137,
212
+ "step": 14500
213
+ },
214
+ {
215
+ "epoch": 0.8182413266419376,
216
+ "grad_norm": 3.2217562198638916,
217
+ "learning_rate": 1.4681444509688784e-05,
218
+ "loss": 2.2899,
219
+ "step": 15000
220
+ },
221
+ {
222
+ "epoch": 0.8455160375300022,
223
+ "grad_norm": 3.6490132808685303,
224
+ "learning_rate": 1.4497944803288314e-05,
225
+ "loss": 2.3015,
226
+ "step": 15500
227
+ },
228
+ {
229
+ "epoch": 0.8727907484180668,
230
+ "grad_norm": 3.53132700920105,
231
+ "learning_rate": 1.4314445096887846e-05,
232
+ "loss": 2.2479,
233
+ "step": 16000
234
+ },
235
+ {
236
+ "epoch": 0.9000654593061314,
237
+ "grad_norm": 2.9956769943237305,
238
+ "learning_rate": 1.4130945390487376e-05,
239
+ "loss": 2.2878,
240
+ "step": 16500
241
+ },
242
+ {
243
+ "epoch": 0.9273401701941959,
244
+ "grad_norm": 3.027482271194458,
245
+ "learning_rate": 1.3947445684086906e-05,
246
+ "loss": 2.2656,
247
+ "step": 17000
248
+ },
249
+ {
250
+ "epoch": 0.9546148810822606,
251
+ "grad_norm": 2.76747465133667,
252
+ "learning_rate": 1.3763945977686436e-05,
253
+ "loss": 2.3085,
254
+ "step": 17500
255
+ },
256
+ {
257
+ "epoch": 0.9818895919703251,
258
+ "grad_norm": 3.367910385131836,
259
+ "learning_rate": 1.3580446271285968e-05,
260
+ "loss": 2.2726,
261
+ "step": 18000
262
+ },
263
+ {
264
+ "epoch": 1.0091643028583897,
265
+ "grad_norm": 3.19476056098938,
266
+ "learning_rate": 1.3396946564885498e-05,
267
+ "loss": 2.2459,
268
+ "step": 18500
269
+ },
270
+ {
271
+ "epoch": 1.0364390137464543,
272
+ "grad_norm": 2.947768211364746,
273
+ "learning_rate": 1.3213446858485028e-05,
274
+ "loss": 2.2225,
275
+ "step": 19000
276
+ },
277
+ {
278
+ "epoch": 1.063713724634519,
279
+ "grad_norm": 3.5084404945373535,
280
+ "learning_rate": 1.3030314151497358e-05,
281
+ "loss": 2.2222,
282
+ "step": 19500
283
+ },
284
+ {
285
+ "epoch": 1.0909884355225834,
286
+ "grad_norm": 3.5933291912078857,
287
+ "learning_rate": 1.284681444509689e-05,
288
+ "loss": 2.2169,
289
+ "step": 20000
290
+ },
291
+ {
292
+ "epoch": 1.118263146410648,
293
+ "grad_norm": 3.1414079666137695,
294
+ "learning_rate": 1.266331473869642e-05,
295
+ "loss": 2.2104,
296
+ "step": 20500
297
+ },
298
+ {
299
+ "epoch": 1.1455378572987127,
300
+ "grad_norm": 3.1075944900512695,
301
+ "learning_rate": 1.247981503229595e-05,
302
+ "loss": 2.1874,
303
+ "step": 21000
304
+ },
305
+ {
306
+ "epoch": 1.1728125681867771,
307
+ "grad_norm": 3.304654598236084,
308
+ "learning_rate": 1.229668232530828e-05,
309
+ "loss": 2.2282,
310
+ "step": 21500
311
+ }
312
+ ],
313
+ "logging_steps": 500,
314
+ "max_steps": 54996,
315
+ "num_input_tokens_seen": 0,
316
+ "num_train_epochs": 3,
317
+ "save_steps": 1833,
318
+ "stateful_callbacks": {
319
+ "TrainerControl": {
320
+ "args": {
321
+ "should_epoch_stop": false,
322
+ "should_evaluate": false,
323
+ "should_log": false,
324
+ "should_save": true,
325
+ "should_training_stop": false
326
+ },
327
+ "attributes": {}
328
+ }
329
+ },
330
+ "total_flos": 5747379535872000.0,
331
+ "train_batch_size": 4,
332
+ "trial_name": null,
333
+ "trial_params": null
334
+ }
rmft/20e2/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
rmft/20e3/config.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "gpt2",
3
+ "activation_function": "gelu_new",
4
+ "architectures": [
5
+ "GPT2LMHeadModel"
6
+ ],
7
+ "attn_pdrop": 0.1,
8
+ "bos_token_id": 50256,
9
+ "embd_pdrop": 0.1,
10
+ "eos_token_id": 50256,
11
+ "initializer_range": 0.02,
12
+ "layer_norm_epsilon": 1e-05,
13
+ "model_type": "gpt2",
14
+ "n_ctx": 1024,
15
+ "n_embd": 768,
16
+ "n_head": 12,
17
+ "n_inner": null,
18
+ "n_layer": 12,
19
+ "n_positions": 1024,
20
+ "reorder_and_upcast_attn": false,
21
+ "resid_pdrop": 0.1,
22
+ "scale_attn_by_inverse_layer_idx": false,
23
+ "scale_attn_weights": true,
24
+ "summary_activation": null,
25
+ "summary_first_dropout": 0.1,
26
+ "summary_proj_to_labels": true,
27
+ "summary_type": "cls_index",
28
+ "summary_use_proj": true,
29
+ "task_specific_params": {
30
+ "text-generation": {
31
+ "do_sample": true,
32
+ "max_length": 50
33
+ }
34
+ },
35
+ "torch_dtype": "float32",
36
+ "transformers_version": "4.46.3",
37
+ "use_cache": true,
38
+ "vocab_size": 50257
39
+ }
rmft/20e3/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 50256,
4
+ "eos_token_id": 50256,
5
+ "transformers_version": "4.46.3"
6
+ }
rmft/20e3/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
rmft/20e3/special_tokens_map.json ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "bos_token": "<|endoftext|>",
3
+ "eos_token": "<|endoftext|>",
4
+ "unk_token": "<|endoftext|>"
5
+ }
rmft/20e3/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
rmft/20e3/tokenizer_config.json ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "added_tokens_decoder": {
4
+ "50256": {
5
+ "content": "<|endoftext|>",
6
+ "lstrip": false,
7
+ "normalized": true,
8
+ "rstrip": false,
9
+ "single_word": false,
10
+ "special": true
11
+ }
12
+ },
13
+ "bos_token": "<|endoftext|>",
14
+ "clean_up_tokenization_spaces": false,
15
+ "eos_token": "<|endoftext|>",
16
+ "model_max_length": 1024,
17
+ "return_tensors": "pt",
18
+ "tokenizer_class": "GPT2Tokenizer",
19
+ "unk_token": "<|endoftext|>"
20
+ }
rmft/20e3/trainer_state.json ADDED
@@ -0,0 +1,593 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 2.199759982544185,
5
+ "eval_steps": 500,
6
+ "global_step": 40326,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.027274710888064588,
13
+ "grad_norm": 8.132363319396973,
14
+ "learning_rate": 1.976e-05,
15
+ "loss": 3.5991,
16
+ "step": 500
17
+ },
18
+ {
19
+ "epoch": 0.054549421776129176,
20
+ "grad_norm": 7.099025249481201,
21
+ "learning_rate": 1.9818702290076336e-05,
22
+ "loss": 2.8754,
23
+ "step": 1000
24
+ },
25
+ {
26
+ "epoch": 0.08182413266419376,
27
+ "grad_norm": 7.820577621459961,
28
+ "learning_rate": 1.963520258367587e-05,
29
+ "loss": 2.6895,
30
+ "step": 1500
31
+ },
32
+ {
33
+ "epoch": 0.10909884355225835,
34
+ "grad_norm": 5.448477268218994,
35
+ "learning_rate": 1.94517028772754e-05,
36
+ "loss": 2.6494,
37
+ "step": 2000
38
+ },
39
+ {
40
+ "epoch": 0.13637355444032292,
41
+ "grad_norm": 6.391147613525391,
42
+ "learning_rate": 1.926820317087493e-05,
43
+ "loss": 2.5544,
44
+ "step": 2500
45
+ },
46
+ {
47
+ "epoch": 0.16364826532838753,
48
+ "grad_norm": 5.117228031158447,
49
+ "learning_rate": 1.908470346447446e-05,
50
+ "loss": 2.5431,
51
+ "step": 3000
52
+ },
53
+ {
54
+ "epoch": 0.1909229762164521,
55
+ "grad_norm": 4.08992338180542,
56
+ "learning_rate": 1.890120375807399e-05,
57
+ "loss": 2.5359,
58
+ "step": 3500
59
+ },
60
+ {
61
+ "epoch": 0.2181976871045167,
62
+ "grad_norm": 3.984097957611084,
63
+ "learning_rate": 1.871770405167352e-05,
64
+ "loss": 2.4834,
65
+ "step": 4000
66
+ },
67
+ {
68
+ "epoch": 0.24547239799258128,
69
+ "grad_norm": 3.8590118885040283,
70
+ "learning_rate": 1.853420434527305e-05,
71
+ "loss": 2.4899,
72
+ "step": 4500
73
+ },
74
+ {
75
+ "epoch": 0.27274710888064585,
76
+ "grad_norm": 4.541988372802734,
77
+ "learning_rate": 1.835070463887258e-05,
78
+ "loss": 2.4608,
79
+ "step": 5000
80
+ },
81
+ {
82
+ "epoch": 0.30002181976871045,
83
+ "grad_norm": 4.910805702209473,
84
+ "learning_rate": 1.816720493247211e-05,
85
+ "loss": 2.402,
86
+ "step": 5500
87
+ },
88
+ {
89
+ "epoch": 0.32729653065677505,
90
+ "grad_norm": 4.220065116882324,
91
+ "learning_rate": 1.798370522607164e-05,
92
+ "loss": 2.4483,
93
+ "step": 6000
94
+ },
95
+ {
96
+ "epoch": 0.3545712415448396,
97
+ "grad_norm": 3.9413557052612305,
98
+ "learning_rate": 1.780020551967117e-05,
99
+ "loss": 2.4271,
100
+ "step": 6500
101
+ },
102
+ {
103
+ "epoch": 0.3818459524329042,
104
+ "grad_norm": 2.890425443649292,
105
+ "learning_rate": 1.76167058132707e-05,
106
+ "loss": 2.3884,
107
+ "step": 7000
108
+ },
109
+ {
110
+ "epoch": 0.4091206633209688,
111
+ "grad_norm": 3.186286211013794,
112
+ "learning_rate": 1.743320610687023e-05,
113
+ "loss": 2.364,
114
+ "step": 7500
115
+ },
116
+ {
117
+ "epoch": 0.4363953742090334,
118
+ "grad_norm": 3.4042603969573975,
119
+ "learning_rate": 1.7249706400469762e-05,
120
+ "loss": 2.4374,
121
+ "step": 8000
122
+ },
123
+ {
124
+ "epoch": 0.46367008509709795,
125
+ "grad_norm": 2.714256763458252,
126
+ "learning_rate": 1.7066206694069292e-05,
127
+ "loss": 2.3535,
128
+ "step": 8500
129
+ },
130
+ {
131
+ "epoch": 0.49094479598516255,
132
+ "grad_norm": 3.9176571369171143,
133
+ "learning_rate": 1.6882706987668822e-05,
134
+ "loss": 2.379,
135
+ "step": 9000
136
+ },
137
+ {
138
+ "epoch": 0.5182195068732272,
139
+ "grad_norm": 3.3017940521240234,
140
+ "learning_rate": 1.669920728126835e-05,
141
+ "loss": 2.3246,
142
+ "step": 9500
143
+ },
144
+ {
145
+ "epoch": 0.5454942177612917,
146
+ "grad_norm": 3.9203879833221436,
147
+ "learning_rate": 1.651570757486788e-05,
148
+ "loss": 2.3452,
149
+ "step": 10000
150
+ },
151
+ {
152
+ "epoch": 0.5727689286493564,
153
+ "grad_norm": 3.474233627319336,
154
+ "learning_rate": 1.6332207868467413e-05,
155
+ "loss": 2.3192,
156
+ "step": 10500
157
+ },
158
+ {
159
+ "epoch": 0.6000436395374209,
160
+ "grad_norm": 3.47390079498291,
161
+ "learning_rate": 1.6148708162066943e-05,
162
+ "loss": 2.3775,
163
+ "step": 11000
164
+ },
165
+ {
166
+ "epoch": 0.6273183504254854,
167
+ "grad_norm": 3.5392141342163086,
168
+ "learning_rate": 1.5965208455666473e-05,
169
+ "loss": 2.3436,
170
+ "step": 11500
171
+ },
172
+ {
173
+ "epoch": 0.6545930613135501,
174
+ "grad_norm": 3.213120698928833,
175
+ "learning_rate": 1.5782075748678803e-05,
176
+ "loss": 2.2958,
177
+ "step": 12000
178
+ },
179
+ {
180
+ "epoch": 0.6818677722016147,
181
+ "grad_norm": 3.356034278869629,
182
+ "learning_rate": 1.5598576042278333e-05,
183
+ "loss": 2.3013,
184
+ "step": 12500
185
+ },
186
+ {
187
+ "epoch": 0.7091424830896792,
188
+ "grad_norm": 2.832211494445801,
189
+ "learning_rate": 1.5415076335877863e-05,
190
+ "loss": 2.3138,
191
+ "step": 13000
192
+ },
193
+ {
194
+ "epoch": 0.7364171939777439,
195
+ "grad_norm": 3.0613503456115723,
196
+ "learning_rate": 1.5231576629477394e-05,
197
+ "loss": 2.302,
198
+ "step": 13500
199
+ },
200
+ {
201
+ "epoch": 0.7636919048658084,
202
+ "grad_norm": 3.324085235595703,
203
+ "learning_rate": 1.5048076923076924e-05,
204
+ "loss": 2.3072,
205
+ "step": 14000
206
+ },
207
+ {
208
+ "epoch": 0.7909666157538731,
209
+ "grad_norm": 3.129357099533081,
210
+ "learning_rate": 1.4864577216676454e-05,
211
+ "loss": 2.3137,
212
+ "step": 14500
213
+ },
214
+ {
215
+ "epoch": 0.8182413266419376,
216
+ "grad_norm": 3.2217562198638916,
217
+ "learning_rate": 1.4681444509688784e-05,
218
+ "loss": 2.2899,
219
+ "step": 15000
220
+ },
221
+ {
222
+ "epoch": 0.8455160375300022,
223
+ "grad_norm": 3.6490132808685303,
224
+ "learning_rate": 1.4497944803288314e-05,
225
+ "loss": 2.3015,
226
+ "step": 15500
227
+ },
228
+ {
229
+ "epoch": 0.8727907484180668,
230
+ "grad_norm": 3.53132700920105,
231
+ "learning_rate": 1.4314445096887846e-05,
232
+ "loss": 2.2479,
233
+ "step": 16000
234
+ },
235
+ {
236
+ "epoch": 0.9000654593061314,
237
+ "grad_norm": 2.9956769943237305,
238
+ "learning_rate": 1.4130945390487376e-05,
239
+ "loss": 2.2878,
240
+ "step": 16500
241
+ },
242
+ {
243
+ "epoch": 0.9273401701941959,
244
+ "grad_norm": 3.027482271194458,
245
+ "learning_rate": 1.3947445684086906e-05,
246
+ "loss": 2.2656,
247
+ "step": 17000
248
+ },
249
+ {
250
+ "epoch": 0.9546148810822606,
251
+ "grad_norm": 2.76747465133667,
252
+ "learning_rate": 1.3763945977686436e-05,
253
+ "loss": 2.3085,
254
+ "step": 17500
255
+ },
256
+ {
257
+ "epoch": 0.9818895919703251,
258
+ "grad_norm": 3.367910385131836,
259
+ "learning_rate": 1.3580446271285968e-05,
260
+ "loss": 2.2726,
261
+ "step": 18000
262
+ },
263
+ {
264
+ "epoch": 1.0091643028583897,
265
+ "grad_norm": 3.19476056098938,
266
+ "learning_rate": 1.3396946564885498e-05,
267
+ "loss": 2.2459,
268
+ "step": 18500
269
+ },
270
+ {
271
+ "epoch": 1.0364390137464543,
272
+ "grad_norm": 2.947768211364746,
273
+ "learning_rate": 1.3213446858485028e-05,
274
+ "loss": 2.2225,
275
+ "step": 19000
276
+ },
277
+ {
278
+ "epoch": 1.063713724634519,
279
+ "grad_norm": 3.5084404945373535,
280
+ "learning_rate": 1.3030314151497358e-05,
281
+ "loss": 2.2222,
282
+ "step": 19500
283
+ },
284
+ {
285
+ "epoch": 1.0909884355225834,
286
+ "grad_norm": 3.5933291912078857,
287
+ "learning_rate": 1.284681444509689e-05,
288
+ "loss": 2.2169,
289
+ "step": 20000
290
+ },
291
+ {
292
+ "epoch": 1.118263146410648,
293
+ "grad_norm": 3.1414079666137695,
294
+ "learning_rate": 1.266331473869642e-05,
295
+ "loss": 2.2104,
296
+ "step": 20500
297
+ },
298
+ {
299
+ "epoch": 1.1455378572987127,
300
+ "grad_norm": 3.1075944900512695,
301
+ "learning_rate": 1.247981503229595e-05,
302
+ "loss": 2.1874,
303
+ "step": 21000
304
+ },
305
+ {
306
+ "epoch": 1.1728125681867771,
307
+ "grad_norm": 3.304654598236084,
308
+ "learning_rate": 1.229668232530828e-05,
309
+ "loss": 2.2282,
310
+ "step": 21500
311
+ },
312
+ {
313
+ "epoch": 1.2000872790748418,
314
+ "grad_norm": 2.92022705078125,
315
+ "learning_rate": 1.211318261890781e-05,
316
+ "loss": 2.2337,
317
+ "step": 22000
318
+ },
319
+ {
320
+ "epoch": 1.2273619899629065,
321
+ "grad_norm": 3.450542449951172,
322
+ "learning_rate": 1.1929682912507342e-05,
323
+ "loss": 2.1784,
324
+ "step": 22500
325
+ },
326
+ {
327
+ "epoch": 1.254636700850971,
328
+ "grad_norm": 2.8238916397094727,
329
+ "learning_rate": 1.1746183206106872e-05,
330
+ "loss": 2.2287,
331
+ "step": 23000
332
+ },
333
+ {
334
+ "epoch": 1.2819114117390356,
335
+ "grad_norm": 2.935835123062134,
336
+ "learning_rate": 1.1563050499119202e-05,
337
+ "loss": 2.2062,
338
+ "step": 23500
339
+ },
340
+ {
341
+ "epoch": 1.3091861226271002,
342
+ "grad_norm": 2.8934576511383057,
343
+ "learning_rate": 1.1379550792718732e-05,
344
+ "loss": 2.1836,
345
+ "step": 24000
346
+ },
347
+ {
348
+ "epoch": 1.3364608335151646,
349
+ "grad_norm": 2.884962797164917,
350
+ "learning_rate": 1.1196051086318263e-05,
351
+ "loss": 2.1632,
352
+ "step": 24500
353
+ },
354
+ {
355
+ "epoch": 1.3637355444032293,
356
+ "grad_norm": 3.163294792175293,
357
+ "learning_rate": 1.1012551379917793e-05,
358
+ "loss": 2.1715,
359
+ "step": 25000
360
+ },
361
+ {
362
+ "epoch": 1.391010255291294,
363
+ "grad_norm": 3.31972599029541,
364
+ "learning_rate": 1.0829051673517323e-05,
365
+ "loss": 2.1867,
366
+ "step": 25500
367
+ },
368
+ {
369
+ "epoch": 1.4182849661793586,
370
+ "grad_norm": 2.9578099250793457,
371
+ "learning_rate": 1.0645918966529653e-05,
372
+ "loss": 2.1851,
373
+ "step": 26000
374
+ },
375
+ {
376
+ "epoch": 1.445559677067423,
377
+ "grad_norm": 3.1862030029296875,
378
+ "learning_rate": 1.0462419260129184e-05,
379
+ "loss": 2.1805,
380
+ "step": 26500
381
+ },
382
+ {
383
+ "epoch": 1.4728343879554877,
384
+ "grad_norm": 3.1671643257141113,
385
+ "learning_rate": 1.0278919553728715e-05,
386
+ "loss": 2.1129,
387
+ "step": 27000
388
+ },
389
+ {
390
+ "epoch": 1.5001090988435521,
391
+ "grad_norm": 2.691154718399048,
392
+ "learning_rate": 1.0095786846741047e-05,
393
+ "loss": 2.1386,
394
+ "step": 27500
395
+ },
396
+ {
397
+ "epoch": 1.5273838097316168,
398
+ "grad_norm": 3.2530946731567383,
399
+ "learning_rate": 9.912287140340577e-06,
400
+ "loss": 2.2013,
401
+ "step": 28000
402
+ },
403
+ {
404
+ "epoch": 1.5546585206196815,
405
+ "grad_norm": 2.7970876693725586,
406
+ "learning_rate": 9.728787433940107e-06,
407
+ "loss": 2.1913,
408
+ "step": 28500
409
+ },
410
+ {
411
+ "epoch": 1.5819332315077461,
412
+ "grad_norm": 3.6721036434173584,
413
+ "learning_rate": 9.545287727539637e-06,
414
+ "loss": 2.1596,
415
+ "step": 29000
416
+ },
417
+ {
418
+ "epoch": 1.6092079423958106,
419
+ "grad_norm": 3.152144432067871,
420
+ "learning_rate": 9.361788021139167e-06,
421
+ "loss": 2.1486,
422
+ "step": 29500
423
+ },
424
+ {
425
+ "epoch": 1.6364826532838752,
426
+ "grad_norm": 3.174539089202881,
427
+ "learning_rate": 9.178288314738697e-06,
428
+ "loss": 2.169,
429
+ "step": 30000
430
+ },
431
+ {
432
+ "epoch": 1.6637573641719396,
433
+ "grad_norm": 2.7281394004821777,
434
+ "learning_rate": 8.994788608338227e-06,
435
+ "loss": 2.1478,
436
+ "step": 30500
437
+ },
438
+ {
439
+ "epoch": 1.6910320750600043,
440
+ "grad_norm": 3.2886815071105957,
441
+ "learning_rate": 8.811288901937757e-06,
442
+ "loss": 2.1372,
443
+ "step": 31000
444
+ },
445
+ {
446
+ "epoch": 1.718306785948069,
447
+ "grad_norm": 2.9683258533477783,
448
+ "learning_rate": 8.628156194950089e-06,
449
+ "loss": 2.1335,
450
+ "step": 31500
451
+ },
452
+ {
453
+ "epoch": 1.7455814968361336,
454
+ "grad_norm": 2.619202136993408,
455
+ "learning_rate": 8.444656488549619e-06,
456
+ "loss": 2.162,
457
+ "step": 32000
458
+ },
459
+ {
460
+ "epoch": 1.7728562077241983,
461
+ "grad_norm": 2.4466633796691895,
462
+ "learning_rate": 8.261156782149149e-06,
463
+ "loss": 2.1595,
464
+ "step": 32500
465
+ },
466
+ {
467
+ "epoch": 1.8001309186122627,
468
+ "grad_norm": 3.0279412269592285,
469
+ "learning_rate": 8.077657075748679e-06,
470
+ "loss": 2.1199,
471
+ "step": 33000
472
+ },
473
+ {
474
+ "epoch": 1.8274056295003271,
475
+ "grad_norm": 2.4983465671539307,
476
+ "learning_rate": 7.89415736934821e-06,
477
+ "loss": 2.1368,
478
+ "step": 33500
479
+ },
480
+ {
481
+ "epoch": 1.8546803403883918,
482
+ "grad_norm": 3.1389667987823486,
483
+ "learning_rate": 7.71102466236054e-06,
484
+ "loss": 2.1801,
485
+ "step": 34000
486
+ },
487
+ {
488
+ "epoch": 1.8819550512764565,
489
+ "grad_norm": 3.049088716506958,
490
+ "learning_rate": 7.5275249559600715e-06,
491
+ "loss": 2.151,
492
+ "step": 34500
493
+ },
494
+ {
495
+ "epoch": 1.9092297621645211,
496
+ "grad_norm": 2.899380922317505,
497
+ "learning_rate": 7.344025249559601e-06,
498
+ "loss": 2.1363,
499
+ "step": 35000
500
+ },
501
+ {
502
+ "epoch": 1.9365044730525858,
503
+ "grad_norm": 3.1051132678985596,
504
+ "learning_rate": 7.1605255431591316e-06,
505
+ "loss": 2.1412,
506
+ "step": 35500
507
+ },
508
+ {
509
+ "epoch": 1.9637791839406502,
510
+ "grad_norm": 2.9140512943267822,
511
+ "learning_rate": 6.9773928361714625e-06,
512
+ "loss": 2.1659,
513
+ "step": 36000
514
+ },
515
+ {
516
+ "epoch": 1.9910538948287149,
517
+ "grad_norm": 2.7826218605041504,
518
+ "learning_rate": 6.793893129770993e-06,
519
+ "loss": 2.1352,
520
+ "step": 36500
521
+ },
522
+ {
523
+ "epoch": 2.0183286057167793,
524
+ "grad_norm": 2.935887575149536,
525
+ "learning_rate": 6.610393423370523e-06,
526
+ "loss": 2.1212,
527
+ "step": 37000
528
+ },
529
+ {
530
+ "epoch": 2.045603316604844,
531
+ "grad_norm": 2.8185176849365234,
532
+ "learning_rate": 6.4268937169700525e-06,
533
+ "loss": 2.1403,
534
+ "step": 37500
535
+ },
536
+ {
537
+ "epoch": 2.0728780274929086,
538
+ "grad_norm": 3.057570219039917,
539
+ "learning_rate": 6.243761009982384e-06,
540
+ "loss": 2.1073,
541
+ "step": 38000
542
+ },
543
+ {
544
+ "epoch": 2.1001527383809733,
545
+ "grad_norm": 2.8996992111206055,
546
+ "learning_rate": 6.060261303581914e-06,
547
+ "loss": 2.115,
548
+ "step": 38500
549
+ },
550
+ {
551
+ "epoch": 2.127427449269038,
552
+ "grad_norm": 2.9682915210723877,
553
+ "learning_rate": 5.876761597181445e-06,
554
+ "loss": 2.1284,
555
+ "step": 39000
556
+ },
557
+ {
558
+ "epoch": 2.154702160157102,
559
+ "grad_norm": 3.1985812187194824,
560
+ "learning_rate": 5.693261890780975e-06,
561
+ "loss": 2.0682,
562
+ "step": 39500
563
+ },
564
+ {
565
+ "epoch": 2.181976871045167,
566
+ "grad_norm": 2.996978282928467,
567
+ "learning_rate": 5.510129183793307e-06,
568
+ "loss": 2.1098,
569
+ "step": 40000
570
+ }
571
+ ],
572
+ "logging_steps": 500,
573
+ "max_steps": 54996,
574
+ "num_input_tokens_seen": 0,
575
+ "num_train_epochs": 3,
576
+ "save_steps": 1833,
577
+ "stateful_callbacks": {
578
+ "TrainerControl": {
579
+ "args": {
580
+ "should_epoch_stop": false,
581
+ "should_evaluate": false,
582
+ "should_log": false,
583
+ "should_save": true,
584
+ "should_training_stop": false
585
+ },
586
+ "attributes": {}
587
+ }
588
+ },
589
+ "total_flos": 1.0536862482432e+16,
590
+ "train_batch_size": 4,
591
+ "trial_name": null,
592
+ "trial_params": null
593
+ }
rmft/20e3/vocab.json ADDED
The diff for this file is too large to render. See raw diff
 
rmft/30e1/config.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "gpt2",
3
+ "activation_function": "gelu_new",
4
+ "architectures": [
5
+ "GPT2LMHeadModel"
6
+ ],
7
+ "attn_pdrop": 0.1,
8
+ "bos_token_id": 50256,
9
+ "embd_pdrop": 0.1,
10
+ "eos_token_id": 50256,
11
+ "initializer_range": 0.02,
12
+ "layer_norm_epsilon": 1e-05,
13
+ "model_type": "gpt2",
14
+ "n_ctx": 1024,
15
+ "n_embd": 768,
16
+ "n_head": 12,
17
+ "n_inner": null,
18
+ "n_layer": 12,
19
+ "n_positions": 1024,
20
+ "reorder_and_upcast_attn": false,
21
+ "resid_pdrop": 0.1,
22
+ "scale_attn_by_inverse_layer_idx": false,
23
+ "scale_attn_weights": true,
24
+ "summary_activation": null,
25
+ "summary_first_dropout": 0.1,
26
+ "summary_proj_to_labels": true,
27
+ "summary_type": "cls_index",
28
+ "summary_use_proj": true,
29
+ "task_specific_params": {
30
+ "text-generation": {
31
+ "do_sample": true,
32
+ "max_length": 50
33
+ }
34
+ },
35
+ "torch_dtype": "float32",
36
+ "transformers_version": "4.46.3",
37
+ "use_cache": true,
38
+ "vocab_size": 50257
39
+ }
rmft/30e1/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 50256,
4
+ "eos_token_id": 50256,
5
+ "transformers_version": "4.46.3"
6
+ }
rmft/30e1/merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
rmft/30e1/special_tokens_map.json ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ {
2
+ "bos_token": "<|endoftext|>",
3
+ "eos_token": "<|endoftext|>",
4
+ "unk_token": "<|endoftext|>"
5
+ }
rmft/30e1/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff