antonpolishko commited on
Commit
83b9941
·
verified ·
1 Parent(s): c3951ed

Model save

Browse files
README.md ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: meta-llama/Llama-3.1-8B
3
+ library_name: transformers
4
+ model_name: llama-3_1-8b-tok10k-overfit-ua
5
+ tags:
6
+ - generated_from_trainer
7
+ - trl
8
+ - sft
9
+ licence: license
10
+ ---
11
+
12
+ # Model Card for llama-3_1-8b-tok10k-overfit-ua
13
+
14
+ This model is a fine-tuned version of [meta-llama/Llama-3.1-8B](https://huggingface.co/meta-llama/Llama-3.1-8B).
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="antonpolishko/llama-3_1-8b-tok10k-overfit-ua", 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/polyagent/huggingface/runs/su6e9ufu)
31
+
32
+ This model was trained with SFT.
33
+
34
+ ### Framework versions
35
+
36
+ - TRL: 0.12.1
37
+ - Transformers: 4.46.3
38
+ - Pytorch: 2.6.0.dev20241113+rocm6.2
39
+ - Datasets: 3.1.0
40
+ - Tokenizers: 0.20.3
41
+
42
+ ## Citations
43
+
44
+
45
+
46
+ Cite TRL as:
47
+
48
+ ```bibtex
49
+ @misc{vonwerra2022trl,
50
+ title = {{TRL: Transformer Reinforcement Learning}},
51
+ 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 Gallouédec},
52
+ year = 2020,
53
+ journal = {GitHub repository},
54
+ publisher = {GitHub},
55
+ howpublished = {\url{https://github.com/huggingface/trl}}
56
+ }
57
+ ```
all_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "total_flos": 5.294195477904884e+18,
4
+ "train_loss": 4.635029224514828,
5
+ "train_runtime": 4299.9756,
6
+ "train_samples": 95663,
7
+ "train_samples_per_second": 13.31,
8
+ "train_steps_per_second": 0.209
9
+ }
generation_config.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 128000,
4
+ "do_sample": true,
5
+ "eos_token_id": 128001,
6
+ "temperature": 0.6,
7
+ "top_p": 0.9,
8
+ "transformers_version": "4.46.3"
9
+ }
runs/Dec26_10-52-59_mia1-gpu-110/events.out.tfevents.1735210423.mia1-gpu-110.3598014.0 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1f656e244b5ba9ab58ed92efeb888da9f4ec1781979b60c6b99dd4a5623ab11e
3
- size 44391
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:26050538f112c0ea43d2da119cd40addf2c6b07dfa3f178f877fe28ef9a458dc
3
+ size 44745
train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "total_flos": 5.294195477904884e+18,
4
+ "train_loss": 4.635029224514828,
5
+ "train_runtime": 4299.9756,
6
+ "train_samples": 95663,
7
+ "train_samples_per_second": 13.31,
8
+ "train_steps_per_second": 0.209
9
+ }
trainer_state.json ADDED
@@ -0,0 +1,1318 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 3.0,
5
+ "eval_steps": 300,
6
+ "global_step": 897,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.0033444816053511705,
13
+ "grad_norm": 28.25,
14
+ "learning_rate": 2e-06,
15
+ "loss": 5.9886,
16
+ "step": 1
17
+ },
18
+ {
19
+ "epoch": 0.016722408026755852,
20
+ "grad_norm": 16.375,
21
+ "learning_rate": 1.999901651759575e-06,
22
+ "loss": 5.9178,
23
+ "step": 5
24
+ },
25
+ {
26
+ "epoch": 0.033444816053511704,
27
+ "grad_norm": 11.0625,
28
+ "learning_rate": 1.9995021451869543e-06,
29
+ "loss": 5.7759,
30
+ "step": 10
31
+ },
32
+ {
33
+ "epoch": 0.05016722408026756,
34
+ "grad_norm": 9.0625,
35
+ "learning_rate": 1.9987954562051724e-06,
36
+ "loss": 5.6469,
37
+ "step": 15
38
+ },
39
+ {
40
+ "epoch": 0.06688963210702341,
41
+ "grad_norm": 7.5,
42
+ "learning_rate": 1.9977818020047815e-06,
43
+ "loss": 5.5274,
44
+ "step": 20
45
+ },
46
+ {
47
+ "epoch": 0.08361204013377926,
48
+ "grad_norm": 6.9375,
49
+ "learning_rate": 1.996461494117619e-06,
50
+ "loss": 5.4792,
51
+ "step": 25
52
+ },
53
+ {
54
+ "epoch": 0.10033444816053512,
55
+ "grad_norm": 6.65625,
56
+ "learning_rate": 1.994834938321061e-06,
57
+ "loss": 5.4114,
58
+ "step": 30
59
+ },
60
+ {
61
+ "epoch": 0.11705685618729098,
62
+ "grad_norm": 6.0625,
63
+ "learning_rate": 1.992902634513312e-06,
64
+ "loss": 5.3313,
65
+ "step": 35
66
+ },
67
+ {
68
+ "epoch": 0.13377926421404682,
69
+ "grad_norm": 5.8125,
70
+ "learning_rate": 1.99066517655977e-06,
71
+ "loss": 5.2905,
72
+ "step": 40
73
+ },
74
+ {
75
+ "epoch": 0.1505016722408027,
76
+ "grad_norm": 5.21875,
77
+ "learning_rate": 1.9881232521105087e-06,
78
+ "loss": 5.2384,
79
+ "step": 45
80
+ },
81
+ {
82
+ "epoch": 0.16722408026755853,
83
+ "grad_norm": 4.375,
84
+ "learning_rate": 1.985277642388941e-06,
85
+ "loss": 5.1707,
86
+ "step": 50
87
+ },
88
+ {
89
+ "epoch": 0.18394648829431437,
90
+ "grad_norm": 4.25,
91
+ "learning_rate": 1.982129221951719e-06,
92
+ "loss": 5.1543,
93
+ "step": 55
94
+ },
95
+ {
96
+ "epoch": 0.20066889632107024,
97
+ "grad_norm": 3.953125,
98
+ "learning_rate": 1.978678958419952e-06,
99
+ "loss": 5.1196,
100
+ "step": 60
101
+ },
102
+ {
103
+ "epoch": 0.21739130434782608,
104
+ "grad_norm": 4.25,
105
+ "learning_rate": 1.9749279121818236e-06,
106
+ "loss": 5.0982,
107
+ "step": 65
108
+ },
109
+ {
110
+ "epoch": 0.23411371237458195,
111
+ "grad_norm": 4.6875,
112
+ "learning_rate": 1.9708772360666954e-06,
113
+ "loss": 5.031,
114
+ "step": 70
115
+ },
116
+ {
117
+ "epoch": 0.2508361204013378,
118
+ "grad_norm": 4.28125,
119
+ "learning_rate": 1.9665281749908033e-06,
120
+ "loss": 5.0424,
121
+ "step": 75
122
+ },
123
+ {
124
+ "epoch": 0.26755852842809363,
125
+ "grad_norm": 3.265625,
126
+ "learning_rate": 1.9618820655746486e-06,
127
+ "loss": 4.981,
128
+ "step": 80
129
+ },
130
+ {
131
+ "epoch": 0.2842809364548495,
132
+ "grad_norm": 3.28125,
133
+ "learning_rate": 1.956940335732209e-06,
134
+ "loss": 4.9588,
135
+ "step": 85
136
+ },
137
+ {
138
+ "epoch": 0.3010033444816054,
139
+ "grad_norm": 3.359375,
140
+ "learning_rate": 1.951704504232089e-06,
141
+ "loss": 4.9495,
142
+ "step": 90
143
+ },
144
+ {
145
+ "epoch": 0.3177257525083612,
146
+ "grad_norm": 3.03125,
147
+ "learning_rate": 1.9461761802307492e-06,
148
+ "loss": 4.9255,
149
+ "step": 95
150
+ },
151
+ {
152
+ "epoch": 0.33444816053511706,
153
+ "grad_norm": 3.984375,
154
+ "learning_rate": 1.9403570627779557e-06,
155
+ "loss": 4.908,
156
+ "step": 100
157
+ },
158
+ {
159
+ "epoch": 0.3511705685618729,
160
+ "grad_norm": 3.328125,
161
+ "learning_rate": 1.9342489402945995e-06,
162
+ "loss": 4.8782,
163
+ "step": 105
164
+ },
165
+ {
166
+ "epoch": 0.36789297658862874,
167
+ "grad_norm": 2.890625,
168
+ "learning_rate": 1.9278536900230564e-06,
169
+ "loss": 4.8632,
170
+ "step": 110
171
+ },
172
+ {
173
+ "epoch": 0.38461538461538464,
174
+ "grad_norm": 2.875,
175
+ "learning_rate": 1.921173277450237e-06,
176
+ "loss": 4.8481,
177
+ "step": 115
178
+ },
179
+ {
180
+ "epoch": 0.4013377926421405,
181
+ "grad_norm": 3.59375,
182
+ "learning_rate": 1.9142097557035305e-06,
183
+ "loss": 4.8467,
184
+ "step": 120
185
+ },
186
+ {
187
+ "epoch": 0.4180602006688963,
188
+ "grad_norm": 2.9375,
189
+ "learning_rate": 1.9069652649198002e-06,
190
+ "loss": 4.8267,
191
+ "step": 125
192
+ },
193
+ {
194
+ "epoch": 0.43478260869565216,
195
+ "grad_norm": 3.421875,
196
+ "learning_rate": 1.8994420315876468e-06,
197
+ "loss": 4.8225,
198
+ "step": 130
199
+ },
200
+ {
201
+ "epoch": 0.451505016722408,
202
+ "grad_norm": 3.609375,
203
+ "learning_rate": 1.891642367863127e-06,
204
+ "loss": 4.7949,
205
+ "step": 135
206
+ },
207
+ {
208
+ "epoch": 0.4682274247491639,
209
+ "grad_norm": 4.40625,
210
+ "learning_rate": 1.8835686708591495e-06,
211
+ "loss": 4.817,
212
+ "step": 140
213
+ },
214
+ {
215
+ "epoch": 0.48494983277591974,
216
+ "grad_norm": 4.625,
217
+ "learning_rate": 1.8752234219087537e-06,
218
+ "loss": 4.8144,
219
+ "step": 145
220
+ },
221
+ {
222
+ "epoch": 0.5016722408026756,
223
+ "grad_norm": 2.703125,
224
+ "learning_rate": 1.866609185802511e-06,
225
+ "loss": 4.7662,
226
+ "step": 150
227
+ },
228
+ {
229
+ "epoch": 0.5183946488294314,
230
+ "grad_norm": 3.796875,
231
+ "learning_rate": 1.857728610000272e-06,
232
+ "loss": 4.7743,
233
+ "step": 155
234
+ },
235
+ {
236
+ "epoch": 0.5351170568561873,
237
+ "grad_norm": 2.515625,
238
+ "learning_rate": 1.8485844238175095e-06,
239
+ "loss": 4.7647,
240
+ "step": 160
241
+ },
242
+ {
243
+ "epoch": 0.5518394648829431,
244
+ "grad_norm": 2.75,
245
+ "learning_rate": 1.839179437586502e-06,
246
+ "loss": 4.7506,
247
+ "step": 165
248
+ },
249
+ {
250
+ "epoch": 0.568561872909699,
251
+ "grad_norm": 3.0625,
252
+ "learning_rate": 1.8295165417926206e-06,
253
+ "loss": 4.7598,
254
+ "step": 170
255
+ },
256
+ {
257
+ "epoch": 0.5852842809364549,
258
+ "grad_norm": 2.6875,
259
+ "learning_rate": 1.8195987061859789e-06,
260
+ "loss": 4.7302,
261
+ "step": 175
262
+ },
263
+ {
264
+ "epoch": 0.6020066889632107,
265
+ "grad_norm": 3.71875,
266
+ "learning_rate": 1.8094289788687243e-06,
267
+ "loss": 4.7313,
268
+ "step": 180
269
+ },
270
+ {
271
+ "epoch": 0.6187290969899666,
272
+ "grad_norm": 2.515625,
273
+ "learning_rate": 1.799010485358249e-06,
274
+ "loss": 4.7353,
275
+ "step": 185
276
+ },
277
+ {
278
+ "epoch": 0.6354515050167224,
279
+ "grad_norm": 2.5,
280
+ "learning_rate": 1.7883464276266062e-06,
281
+ "loss": 4.7363,
282
+ "step": 190
283
+ },
284
+ {
285
+ "epoch": 0.6521739130434783,
286
+ "grad_norm": 2.90625,
287
+ "learning_rate": 1.777440083116432e-06,
288
+ "loss": 4.7022,
289
+ "step": 195
290
+ },
291
+ {
292
+ "epoch": 0.6688963210702341,
293
+ "grad_norm": 2.609375,
294
+ "learning_rate": 1.766294803733671e-06,
295
+ "loss": 4.6943,
296
+ "step": 200
297
+ },
298
+ {
299
+ "epoch": 0.68561872909699,
300
+ "grad_norm": 3.03125,
301
+ "learning_rate": 1.754914014817416e-06,
302
+ "loss": 4.6801,
303
+ "step": 205
304
+ },
305
+ {
306
+ "epoch": 0.7023411371237458,
307
+ "grad_norm": 2.546875,
308
+ "learning_rate": 1.743301214087181e-06,
309
+ "loss": 4.6977,
310
+ "step": 210
311
+ },
312
+ {
313
+ "epoch": 0.7190635451505016,
314
+ "grad_norm": 3.359375,
315
+ "learning_rate": 1.7314599705679277e-06,
316
+ "loss": 4.6801,
317
+ "step": 215
318
+ },
319
+ {
320
+ "epoch": 0.7357859531772575,
321
+ "grad_norm": 2.59375,
322
+ "learning_rate": 1.7193939234931776e-06,
323
+ "loss": 4.6643,
324
+ "step": 220
325
+ },
326
+ {
327
+ "epoch": 0.7525083612040134,
328
+ "grad_norm": 2.625,
329
+ "learning_rate": 1.7071067811865474e-06,
330
+ "loss": 4.6589,
331
+ "step": 225
332
+ },
333
+ {
334
+ "epoch": 0.7692307692307693,
335
+ "grad_norm": 2.484375,
336
+ "learning_rate": 1.6946023199220484e-06,
337
+ "loss": 4.6329,
338
+ "step": 230
339
+ },
340
+ {
341
+ "epoch": 0.7859531772575251,
342
+ "grad_norm": 2.515625,
343
+ "learning_rate": 1.681884382763505e-06,
344
+ "loss": 4.666,
345
+ "step": 235
346
+ },
347
+ {
348
+ "epoch": 0.802675585284281,
349
+ "grad_norm": 3.171875,
350
+ "learning_rate": 1.6689568783834447e-06,
351
+ "loss": 4.6603,
352
+ "step": 240
353
+ },
354
+ {
355
+ "epoch": 0.8193979933110368,
356
+ "grad_norm": 3.484375,
357
+ "learning_rate": 1.6558237798618241e-06,
358
+ "loss": 4.6389,
359
+ "step": 245
360
+ },
361
+ {
362
+ "epoch": 0.8361204013377926,
363
+ "grad_norm": 2.546875,
364
+ "learning_rate": 1.6424891234649616e-06,
365
+ "loss": 4.6501,
366
+ "step": 250
367
+ },
368
+ {
369
+ "epoch": 0.8528428093645485,
370
+ "grad_norm": 2.9375,
371
+ "learning_rate": 1.6289570074050492e-06,
372
+ "loss": 4.6093,
373
+ "step": 255
374
+ },
375
+ {
376
+ "epoch": 0.8695652173913043,
377
+ "grad_norm": 3.109375,
378
+ "learning_rate": 1.6152315905806267e-06,
379
+ "loss": 4.6375,
380
+ "step": 260
381
+ },
382
+ {
383
+ "epoch": 0.8862876254180602,
384
+ "grad_norm": 3.46875,
385
+ "learning_rate": 1.6013170912984058e-06,
386
+ "loss": 4.6377,
387
+ "step": 265
388
+ },
389
+ {
390
+ "epoch": 0.903010033444816,
391
+ "grad_norm": 2.46875,
392
+ "learning_rate": 1.5872177859768332e-06,
393
+ "loss": 4.6326,
394
+ "step": 270
395
+ },
396
+ {
397
+ "epoch": 0.919732441471572,
398
+ "grad_norm": 2.71875,
399
+ "learning_rate": 1.572938007831798e-06,
400
+ "loss": 4.6,
401
+ "step": 275
402
+ },
403
+ {
404
+ "epoch": 0.9364548494983278,
405
+ "grad_norm": 2.421875,
406
+ "learning_rate": 1.5584821455448788e-06,
407
+ "loss": 4.6083,
408
+ "step": 280
409
+ },
410
+ {
411
+ "epoch": 0.9531772575250836,
412
+ "grad_norm": 3.65625,
413
+ "learning_rate": 1.5438546419145485e-06,
414
+ "loss": 4.6106,
415
+ "step": 285
416
+ },
417
+ {
418
+ "epoch": 0.9698996655518395,
419
+ "grad_norm": 2.421875,
420
+ "learning_rate": 1.5290599924907432e-06,
421
+ "loss": 4.6087,
422
+ "step": 290
423
+ },
424
+ {
425
+ "epoch": 0.9866220735785953,
426
+ "grad_norm": 2.71875,
427
+ "learning_rate": 1.5141027441932214e-06,
428
+ "loss": 4.597,
429
+ "step": 295
430
+ },
431
+ {
432
+ "epoch": 1.0033444816053512,
433
+ "grad_norm": 2.75,
434
+ "learning_rate": 1.498987493914135e-06,
435
+ "loss": 4.5658,
436
+ "step": 300
437
+ },
438
+ {
439
+ "epoch": 1.0033444816053512,
440
+ "eval_loss": 4.585446834564209,
441
+ "eval_runtime": 7.2206,
442
+ "eval_samples_per_second": 80.188,
443
+ "eval_steps_per_second": 2.631,
444
+ "step": 300
445
+ },
446
+ {
447
+ "epoch": 1.020066889632107,
448
+ "grad_norm": 2.90625,
449
+ "learning_rate": 1.4837188871052397e-06,
450
+ "loss": 4.5897,
451
+ "step": 305
452
+ },
453
+ {
454
+ "epoch": 1.0367892976588629,
455
+ "grad_norm": 2.734375,
456
+ "learning_rate": 1.4683016163501854e-06,
457
+ "loss": 4.5885,
458
+ "step": 310
459
+ },
460
+ {
461
+ "epoch": 1.0535117056856187,
462
+ "grad_norm": 2.75,
463
+ "learning_rate": 1.452740419922317e-06,
464
+ "loss": 4.6009,
465
+ "step": 315
466
+ },
467
+ {
468
+ "epoch": 1.0702341137123745,
469
+ "grad_norm": 2.96875,
470
+ "learning_rate": 1.4370400803284373e-06,
471
+ "loss": 4.5789,
472
+ "step": 320
473
+ },
474
+ {
475
+ "epoch": 1.0869565217391304,
476
+ "grad_norm": 3.4375,
477
+ "learning_rate": 1.421205422838971e-06,
478
+ "loss": 4.5435,
479
+ "step": 325
480
+ },
481
+ {
482
+ "epoch": 1.1036789297658862,
483
+ "grad_norm": 2.734375,
484
+ "learning_rate": 1.4052413140049897e-06,
485
+ "loss": 4.566,
486
+ "step": 330
487
+ },
488
+ {
489
+ "epoch": 1.120401337792642,
490
+ "grad_norm": 2.59375,
491
+ "learning_rate": 1.389152660162549e-06,
492
+ "loss": 4.5621,
493
+ "step": 335
494
+ },
495
+ {
496
+ "epoch": 1.137123745819398,
497
+ "grad_norm": 2.65625,
498
+ "learning_rate": 1.3729444059247953e-06,
499
+ "loss": 4.5505,
500
+ "step": 340
501
+ },
502
+ {
503
+ "epoch": 1.1538461538461537,
504
+ "grad_norm": 2.40625,
505
+ "learning_rate": 1.3566215326623129e-06,
506
+ "loss": 4.5709,
507
+ "step": 345
508
+ },
509
+ {
510
+ "epoch": 1.1705685618729098,
511
+ "grad_norm": 2.390625,
512
+ "learning_rate": 1.3401890569721723e-06,
513
+ "loss": 4.5569,
514
+ "step": 350
515
+ },
516
+ {
517
+ "epoch": 1.1872909698996654,
518
+ "grad_norm": 2.46875,
519
+ "learning_rate": 1.3236520291361515e-06,
520
+ "loss": 4.5456,
521
+ "step": 355
522
+ },
523
+ {
524
+ "epoch": 1.2040133779264215,
525
+ "grad_norm": 2.5625,
526
+ "learning_rate": 1.3070155315686059e-06,
527
+ "loss": 4.5543,
528
+ "step": 360
529
+ },
530
+ {
531
+ "epoch": 1.2207357859531773,
532
+ "grad_norm": 2.65625,
533
+ "learning_rate": 1.2902846772544622e-06,
534
+ "loss": 4.5408,
535
+ "step": 365
536
+ },
537
+ {
538
+ "epoch": 1.2374581939799332,
539
+ "grad_norm": 2.90625,
540
+ "learning_rate": 1.273464608177818e-06,
541
+ "loss": 4.5435,
542
+ "step": 370
543
+ },
544
+ {
545
+ "epoch": 1.254180602006689,
546
+ "grad_norm": 2.578125,
547
+ "learning_rate": 1.2565604937416266e-06,
548
+ "loss": 4.5436,
549
+ "step": 375
550
+ },
551
+ {
552
+ "epoch": 1.2709030100334449,
553
+ "grad_norm": 2.484375,
554
+ "learning_rate": 1.2395775291789567e-06,
555
+ "loss": 4.5448,
556
+ "step": 380
557
+ },
558
+ {
559
+ "epoch": 1.2876254180602007,
560
+ "grad_norm": 2.375,
561
+ "learning_rate": 1.2225209339563143e-06,
562
+ "loss": 4.5398,
563
+ "step": 385
564
+ },
565
+ {
566
+ "epoch": 1.3043478260869565,
567
+ "grad_norm": 2.484375,
568
+ "learning_rate": 1.2053959501695144e-06,
569
+ "loss": 4.5383,
570
+ "step": 390
571
+ },
572
+ {
573
+ "epoch": 1.3210702341137124,
574
+ "grad_norm": 2.421875,
575
+ "learning_rate": 1.1882078409326002e-06,
576
+ "loss": 4.5556,
577
+ "step": 395
578
+ },
579
+ {
580
+ "epoch": 1.3377926421404682,
581
+ "grad_norm": 2.703125,
582
+ "learning_rate": 1.1709618887603011e-06,
583
+ "loss": 4.5273,
584
+ "step": 400
585
+ },
586
+ {
587
+ "epoch": 1.354515050167224,
588
+ "grad_norm": 2.5,
589
+ "learning_rate": 1.15366339394453e-06,
590
+ "loss": 4.5262,
591
+ "step": 405
592
+ },
593
+ {
594
+ "epoch": 1.37123745819398,
595
+ "grad_norm": 2.46875,
596
+ "learning_rate": 1.1363176729254144e-06,
597
+ "loss": 4.5391,
598
+ "step": 410
599
+ },
600
+ {
601
+ "epoch": 1.3879598662207357,
602
+ "grad_norm": 2.953125,
603
+ "learning_rate": 1.118930056657367e-06,
604
+ "loss": 4.5218,
605
+ "step": 415
606
+ },
607
+ {
608
+ "epoch": 1.4046822742474916,
609
+ "grad_norm": 2.421875,
610
+ "learning_rate": 1.1015058889706942e-06,
611
+ "loss": 4.5255,
612
+ "step": 420
613
+ },
614
+ {
615
+ "epoch": 1.4214046822742474,
616
+ "grad_norm": 2.53125,
617
+ "learning_rate": 1.0840505249292475e-06,
618
+ "loss": 4.5304,
619
+ "step": 425
620
+ },
621
+ {
622
+ "epoch": 1.4381270903010033,
623
+ "grad_norm": 2.5,
624
+ "learning_rate": 1.0665693291846243e-06,
625
+ "loss": 4.5418,
626
+ "step": 430
627
+ },
628
+ {
629
+ "epoch": 1.4548494983277591,
630
+ "grad_norm": 2.40625,
631
+ "learning_rate": 1.0490676743274181e-06,
632
+ "loss": 4.522,
633
+ "step": 435
634
+ },
635
+ {
636
+ "epoch": 1.471571906354515,
637
+ "grad_norm": 2.5,
638
+ "learning_rate": 1.031550939236033e-06,
639
+ "loss": 4.5346,
640
+ "step": 440
641
+ },
642
+ {
643
+ "epoch": 1.488294314381271,
644
+ "grad_norm": 2.296875,
645
+ "learning_rate": 1.0140245074235622e-06,
646
+ "loss": 4.496,
647
+ "step": 445
648
+ },
649
+ {
650
+ "epoch": 1.5050167224080266,
651
+ "grad_norm": 2.390625,
652
+ "learning_rate": 9.964937653832469e-07,
653
+ "loss": 4.5212,
654
+ "step": 450
655
+ },
656
+ {
657
+ "epoch": 1.5217391304347827,
658
+ "grad_norm": 2.3125,
659
+ "learning_rate": 9.78964100933011e-07,
660
+ "loss": 4.5069,
661
+ "step": 455
662
+ },
663
+ {
664
+ "epoch": 1.5384615384615383,
665
+ "grad_norm": 2.546875,
666
+ "learning_rate": 9.614409015595994e-07,
667
+ "loss": 4.5124,
668
+ "step": 460
669
+ },
670
+ {
671
+ "epoch": 1.5551839464882944,
672
+ "grad_norm": 2.515625,
673
+ "learning_rate": 9.43929552762808e-07,
674
+ "loss": 4.5155,
675
+ "step": 465
676
+ },
677
+ {
678
+ "epoch": 1.57190635451505,
679
+ "grad_norm": 2.46875,
680
+ "learning_rate": 9.264354364003326e-07,
681
+ "loss": 4.5143,
682
+ "step": 470
683
+ },
684
+ {
685
+ "epoch": 1.588628762541806,
686
+ "grad_norm": 2.25,
687
+ "learning_rate": 9.089639290337298e-07,
688
+ "loss": 4.4947,
689
+ "step": 475
690
+ },
691
+ {
692
+ "epoch": 1.605351170568562,
693
+ "grad_norm": 2.53125,
694
+ "learning_rate": 8.915204002760122e-07,
695
+ "loss": 4.5113,
696
+ "step": 480
697
+ },
698
+ {
699
+ "epoch": 1.6220735785953178,
700
+ "grad_norm": 2.359375,
701
+ "learning_rate": 8.741102111413748e-07,
702
+ "loss": 4.5215,
703
+ "step": 485
704
+ },
705
+ {
706
+ "epoch": 1.6387959866220736,
707
+ "grad_norm": 2.390625,
708
+ "learning_rate": 8.567387123975647e-07,
709
+ "loss": 4.4991,
710
+ "step": 490
711
+ },
712
+ {
713
+ "epoch": 1.6555183946488294,
714
+ "grad_norm": 3.296875,
715
+ "learning_rate": 8.394112429214029e-07,
716
+ "loss": 4.5263,
717
+ "step": 495
718
+ },
719
+ {
720
+ "epoch": 1.6722408026755853,
721
+ "grad_norm": 2.28125,
722
+ "learning_rate": 8.221331280579564e-07,
723
+ "loss": 4.5039,
724
+ "step": 500
725
+ },
726
+ {
727
+ "epoch": 1.6889632107023411,
728
+ "grad_norm": 2.46875,
729
+ "learning_rate": 8.049096779838717e-07,
730
+ "loss": 4.5294,
731
+ "step": 505
732
+ },
733
+ {
734
+ "epoch": 1.705685618729097,
735
+ "grad_norm": 2.734375,
736
+ "learning_rate": 7.877461860753696e-07,
737
+ "loss": 4.4868,
738
+ "step": 510
739
+ },
740
+ {
741
+ "epoch": 1.7224080267558528,
742
+ "grad_norm": 2.484375,
743
+ "learning_rate": 7.706479272814022e-07,
744
+ "loss": 4.4988,
745
+ "step": 515
746
+ },
747
+ {
748
+ "epoch": 1.7391304347826086,
749
+ "grad_norm": 2.3125,
750
+ "learning_rate": 7.536201565024767e-07,
751
+ "loss": 4.4843,
752
+ "step": 520
753
+ },
754
+ {
755
+ "epoch": 1.7558528428093645,
756
+ "grad_norm": 2.359375,
757
+ "learning_rate": 7.366681069756351e-07,
758
+ "loss": 4.4878,
759
+ "step": 525
760
+ },
761
+ {
762
+ "epoch": 1.7725752508361206,
763
+ "grad_norm": 2.578125,
764
+ "learning_rate": 7.197969886660984e-07,
765
+ "loss": 4.4925,
766
+ "step": 530
767
+ },
768
+ {
769
+ "epoch": 1.7892976588628762,
770
+ "grad_norm": 2.65625,
771
+ "learning_rate": 7.030119866660565e-07,
772
+ "loss": 4.5185,
773
+ "step": 535
774
+ },
775
+ {
776
+ "epoch": 1.8060200668896322,
777
+ "grad_norm": 2.375,
778
+ "learning_rate": 6.863182596011085e-07,
779
+ "loss": 4.4988,
780
+ "step": 540
781
+ },
782
+ {
783
+ "epoch": 1.8227424749163879,
784
+ "grad_norm": 2.640625,
785
+ "learning_rate": 6.697209380448332e-07,
786
+ "loss": 4.4764,
787
+ "step": 545
788
+ },
789
+ {
790
+ "epoch": 1.839464882943144,
791
+ "grad_norm": 2.296875,
792
+ "learning_rate": 6.532251229419809e-07,
793
+ "loss": 4.4987,
794
+ "step": 550
795
+ },
796
+ {
797
+ "epoch": 1.8561872909698995,
798
+ "grad_norm": 2.46875,
799
+ "learning_rate": 6.368358840407752e-07,
800
+ "loss": 4.5123,
801
+ "step": 555
802
+ },
803
+ {
804
+ "epoch": 1.8729096989966556,
805
+ "grad_norm": 2.375,
806
+ "learning_rate": 6.205582583347973e-07,
807
+ "loss": 4.5019,
808
+ "step": 560
809
+ },
810
+ {
811
+ "epoch": 1.8896321070234112,
812
+ "grad_norm": 2.28125,
813
+ "learning_rate": 6.043972485149414e-07,
814
+ "loss": 4.5041,
815
+ "step": 565
816
+ },
817
+ {
818
+ "epoch": 1.9063545150501673,
819
+ "grad_norm": 2.453125,
820
+ "learning_rate": 5.88357821431908e-07,
821
+ "loss": 4.485,
822
+ "step": 570
823
+ },
824
+ {
825
+ "epoch": 1.9230769230769231,
826
+ "grad_norm": 2.4375,
827
+ "learning_rate": 5.724449065697181e-07,
828
+ "loss": 4.4854,
829
+ "step": 575
830
+ },
831
+ {
832
+ "epoch": 1.939799331103679,
833
+ "grad_norm": 2.5625,
834
+ "learning_rate": 5.566633945307052e-07,
835
+ "loss": 4.5039,
836
+ "step": 580
837
+ },
838
+ {
839
+ "epoch": 1.9565217391304348,
840
+ "grad_norm": 2.359375,
841
+ "learning_rate": 5.410181355324621e-07,
842
+ "loss": 4.507,
843
+ "step": 585
844
+ },
845
+ {
846
+ "epoch": 1.9732441471571907,
847
+ "grad_norm": 2.3125,
848
+ "learning_rate": 5.255139379171966e-07,
849
+ "loss": 4.5087,
850
+ "step": 590
851
+ },
852
+ {
853
+ "epoch": 1.9899665551839465,
854
+ "grad_norm": 2.359375,
855
+ "learning_rate": 5.101555666739563e-07,
856
+ "loss": 4.5007,
857
+ "step": 595
858
+ },
859
+ {
860
+ "epoch": 2.0066889632107023,
861
+ "grad_norm": 2.25,
862
+ "learning_rate": 4.949477419741814e-07,
863
+ "loss": 4.4832,
864
+ "step": 600
865
+ },
866
+ {
867
+ "epoch": 2.0066889632107023,
868
+ "eval_loss": 4.490821361541748,
869
+ "eval_runtime": 13.8089,
870
+ "eval_samples_per_second": 41.929,
871
+ "eval_steps_per_second": 1.376,
872
+ "step": 600
873
+ },
874
+ {
875
+ "epoch": 2.0234113712374584,
876
+ "grad_norm": 2.25,
877
+ "learning_rate": 4.798951377210254e-07,
878
+ "loss": 4.481,
879
+ "step": 605
880
+ },
881
+ {
882
+ "epoch": 2.040133779264214,
883
+ "grad_norm": 2.4375,
884
+ "learning_rate": 4.6500238011290293e-07,
885
+ "loss": 4.5015,
886
+ "step": 610
887
+ },
888
+ {
889
+ "epoch": 2.05685618729097,
890
+ "grad_norm": 2.265625,
891
+ "learning_rate": 4.5027404622169185e-07,
892
+ "loss": 4.4934,
893
+ "step": 615
894
+ },
895
+ {
896
+ "epoch": 2.0735785953177257,
897
+ "grad_norm": 2.5,
898
+ "learning_rate": 4.3571466258603907e-07,
899
+ "loss": 4.5015,
900
+ "step": 620
901
+ },
902
+ {
903
+ "epoch": 2.0903010033444818,
904
+ "grad_norm": 2.515625,
905
+ "learning_rate": 4.2132870382019427e-07,
906
+ "loss": 4.5043,
907
+ "step": 625
908
+ },
909
+ {
910
+ "epoch": 2.1070234113712374,
911
+ "grad_norm": 2.390625,
912
+ "learning_rate": 4.071205912388015e-07,
913
+ "loss": 4.5132,
914
+ "step": 630
915
+ },
916
+ {
917
+ "epoch": 2.1237458193979935,
918
+ "grad_norm": 2.484375,
919
+ "learning_rate": 3.930946914980744e-07,
920
+ "loss": 4.5032,
921
+ "step": 635
922
+ },
923
+ {
924
+ "epoch": 2.140468227424749,
925
+ "grad_norm": 2.5625,
926
+ "learning_rate": 3.7925531525376617e-07,
927
+ "loss": 4.487,
928
+ "step": 640
929
+ },
930
+ {
931
+ "epoch": 2.157190635451505,
932
+ "grad_norm": 2.453125,
933
+ "learning_rate": 3.656067158363546e-07,
934
+ "loss": 4.4899,
935
+ "step": 645
936
+ },
937
+ {
938
+ "epoch": 2.1739130434782608,
939
+ "grad_norm": 2.296875,
940
+ "learning_rate": 3.521530879438407e-07,
941
+ "loss": 4.4797,
942
+ "step": 650
943
+ },
944
+ {
945
+ "epoch": 2.190635451505017,
946
+ "grad_norm": 2.34375,
947
+ "learning_rate": 3.388985663525702e-07,
948
+ "loss": 4.4826,
949
+ "step": 655
950
+ },
951
+ {
952
+ "epoch": 2.2073578595317724,
953
+ "grad_norm": 2.375,
954
+ "learning_rate": 3.2584722464647165e-07,
955
+ "loss": 4.4947,
956
+ "step": 660
957
+ },
958
+ {
959
+ "epoch": 2.2240802675585285,
960
+ "grad_norm": 2.328125,
961
+ "learning_rate": 3.1300307396509826e-07,
962
+ "loss": 4.4985,
963
+ "step": 665
964
+ },
965
+ {
966
+ "epoch": 2.240802675585284,
967
+ "grad_norm": 2.203125,
968
+ "learning_rate": 3.0037006177086345e-07,
969
+ "loss": 4.4972,
970
+ "step": 670
971
+ },
972
+ {
973
+ "epoch": 2.25752508361204,
974
+ "grad_norm": 2.75,
975
+ "learning_rate": 2.879520706358446e-07,
976
+ "loss": 4.4785,
977
+ "step": 675
978
+ },
979
+ {
980
+ "epoch": 2.274247491638796,
981
+ "grad_norm": 2.28125,
982
+ "learning_rate": 2.757529170485332e-07,
983
+ "loss": 4.4973,
984
+ "step": 680
985
+ },
986
+ {
987
+ "epoch": 2.290969899665552,
988
+ "grad_norm": 2.359375,
989
+ "learning_rate": 2.6377635024089087e-07,
990
+ "loss": 4.4967,
991
+ "step": 685
992
+ },
993
+ {
994
+ "epoch": 2.3076923076923075,
995
+ "grad_norm": 2.1875,
996
+ "learning_rate": 2.5202605103607835e-07,
997
+ "loss": 4.4944,
998
+ "step": 690
999
+ },
1000
+ {
1001
+ "epoch": 2.3244147157190636,
1002
+ "grad_norm": 2.265625,
1003
+ "learning_rate": 2.4050563071720864e-07,
1004
+ "loss": 4.5062,
1005
+ "step": 695
1006
+ },
1007
+ {
1008
+ "epoch": 2.3411371237458196,
1009
+ "grad_norm": 2.421875,
1010
+ "learning_rate": 2.2921862991747121e-07,
1011
+ "loss": 4.4884,
1012
+ "step": 700
1013
+ },
1014
+ {
1015
+ "epoch": 2.3578595317725752,
1016
+ "grad_norm": 2.296875,
1017
+ "learning_rate": 2.181685175319702e-07,
1018
+ "loss": 4.4793,
1019
+ "step": 705
1020
+ },
1021
+ {
1022
+ "epoch": 2.374581939799331,
1023
+ "grad_norm": 2.46875,
1024
+ "learning_rate": 2.073586896516095e-07,
1025
+ "loss": 4.4759,
1026
+ "step": 710
1027
+ },
1028
+ {
1029
+ "epoch": 2.391304347826087,
1030
+ "grad_norm": 2.46875,
1031
+ "learning_rate": 1.9679246851935516e-07,
1032
+ "loss": 4.4862,
1033
+ "step": 715
1034
+ },
1035
+ {
1036
+ "epoch": 2.408026755852843,
1037
+ "grad_norm": 2.3125,
1038
+ "learning_rate": 1.864731015091908e-07,
1039
+ "loss": 4.4886,
1040
+ "step": 720
1041
+ },
1042
+ {
1043
+ "epoch": 2.4247491638795986,
1044
+ "grad_norm": 2.34375,
1045
+ "learning_rate": 1.7640376012808533e-07,
1046
+ "loss": 4.5015,
1047
+ "step": 725
1048
+ },
1049
+ {
1050
+ "epoch": 2.4414715719063547,
1051
+ "grad_norm": 2.234375,
1052
+ "learning_rate": 1.6658753904127732e-07,
1053
+ "loss": 4.4842,
1054
+ "step": 730
1055
+ },
1056
+ {
1057
+ "epoch": 2.4581939799331103,
1058
+ "grad_norm": 2.34375,
1059
+ "learning_rate": 1.570274551211732e-07,
1060
+ "loss": 4.5025,
1061
+ "step": 735
1062
+ },
1063
+ {
1064
+ "epoch": 2.4749163879598663,
1065
+ "grad_norm": 2.359375,
1066
+ "learning_rate": 1.477264465201572e-07,
1067
+ "loss": 4.492,
1068
+ "step": 740
1069
+ },
1070
+ {
1071
+ "epoch": 2.491638795986622,
1072
+ "grad_norm": 2.4375,
1073
+ "learning_rate": 1.3868737176759105e-07,
1074
+ "loss": 4.5102,
1075
+ "step": 745
1076
+ },
1077
+ {
1078
+ "epoch": 2.508361204013378,
1079
+ "grad_norm": 2.34375,
1080
+ "learning_rate": 1.2991300889128865e-07,
1081
+ "loss": 4.4899,
1082
+ "step": 750
1083
+ },
1084
+ {
1085
+ "epoch": 2.5250836120401337,
1086
+ "grad_norm": 2.296875,
1087
+ "learning_rate": 1.2140605456372854e-07,
1088
+ "loss": 4.5076,
1089
+ "step": 755
1090
+ },
1091
+ {
1092
+ "epoch": 2.5418060200668897,
1093
+ "grad_norm": 2.21875,
1094
+ "learning_rate": 1.13169123273271e-07,
1095
+ "loss": 4.4919,
1096
+ "step": 760
1097
+ },
1098
+ {
1099
+ "epoch": 2.5585284280936453,
1100
+ "grad_norm": 2.40625,
1101
+ "learning_rate": 1.0520474652063394e-07,
1102
+ "loss": 4.4859,
1103
+ "step": 765
1104
+ },
1105
+ {
1106
+ "epoch": 2.5752508361204014,
1107
+ "grad_norm": 2.453125,
1108
+ "learning_rate": 9.751537204087257e-08,
1109
+ "loss": 4.4933,
1110
+ "step": 770
1111
+ },
1112
+ {
1113
+ "epoch": 2.591973244147157,
1114
+ "grad_norm": 2.359375,
1115
+ "learning_rate": 9.010336305110345e-08,
1116
+ "loss": 4.493,
1117
+ "step": 775
1118
+ },
1119
+ {
1120
+ "epoch": 2.608695652173913,
1121
+ "grad_norm": 2.328125,
1122
+ "learning_rate": 8.297099752420444e-08,
1123
+ "loss": 4.4685,
1124
+ "step": 780
1125
+ },
1126
+ {
1127
+ "epoch": 2.625418060200669,
1128
+ "grad_norm": 2.328125,
1129
+ "learning_rate": 7.612046748871326e-08,
1130
+ "loss": 4.4856,
1131
+ "step": 785
1132
+ },
1133
+ {
1134
+ "epoch": 2.6421404682274248,
1135
+ "grad_norm": 2.375,
1136
+ "learning_rate": 6.955387835513893e-08,
1137
+ "loss": 4.4676,
1138
+ "step": 790
1139
+ },
1140
+ {
1141
+ "epoch": 2.6588628762541804,
1142
+ "grad_norm": 2.25,
1143
+ "learning_rate": 6.327324826889469e-08,
1144
+ "loss": 4.4745,
1145
+ "step": 795
1146
+ },
1147
+ {
1148
+ "epoch": 2.6755852842809364,
1149
+ "grad_norm": 2.40625,
1150
+ "learning_rate": 5.728050749005098e-08,
1151
+ "loss": 4.4856,
1152
+ "step": 800
1153
+ },
1154
+ {
1155
+ "epoch": 2.6923076923076925,
1156
+ "grad_norm": 2.234375,
1157
+ "learning_rate": 5.1577497800097345e-08,
1158
+ "loss": 4.5096,
1159
+ "step": 805
1160
+ },
1161
+ {
1162
+ "epoch": 2.709030100334448,
1163
+ "grad_norm": 2.484375,
1164
+ "learning_rate": 4.616597193589833e-08,
1165
+ "loss": 4.4839,
1166
+ "step": 810
1167
+ },
1168
+ {
1169
+ "epoch": 2.7257525083612038,
1170
+ "grad_norm": 2.328125,
1171
+ "learning_rate": 4.1047593051015237e-08,
1172
+ "loss": 4.5145,
1173
+ "step": 815
1174
+ },
1175
+ {
1176
+ "epoch": 2.74247491638796,
1177
+ "grad_norm": 2.4375,
1178
+ "learning_rate": 3.622393420456016e-08,
1179
+ "loss": 4.4835,
1180
+ "step": 820
1181
+ },
1182
+ {
1183
+ "epoch": 2.759197324414716,
1184
+ "grad_norm": 2.234375,
1185
+ "learning_rate": 3.169647787773866e-08,
1186
+ "loss": 4.4983,
1187
+ "step": 825
1188
+ },
1189
+ {
1190
+ "epoch": 2.7759197324414715,
1191
+ "grad_norm": 2.21875,
1192
+ "learning_rate": 2.7466615518231483e-08,
1193
+ "loss": 4.4734,
1194
+ "step": 830
1195
+ },
1196
+ {
1197
+ "epoch": 2.7926421404682276,
1198
+ "grad_norm": 2.390625,
1199
+ "learning_rate": 2.353564711255329e-08,
1200
+ "loss": 4.4825,
1201
+ "step": 835
1202
+ },
1203
+ {
1204
+ "epoch": 2.809364548494983,
1205
+ "grad_norm": 2.328125,
1206
+ "learning_rate": 1.990478078652047e-08,
1207
+ "loss": 4.4676,
1208
+ "step": 840
1209
+ },
1210
+ {
1211
+ "epoch": 2.8260869565217392,
1212
+ "grad_norm": 2.421875,
1213
+ "learning_rate": 1.657513243395159e-08,
1214
+ "loss": 4.4944,
1215
+ "step": 845
1216
+ },
1217
+ {
1218
+ "epoch": 2.842809364548495,
1219
+ "grad_norm": 2.203125,
1220
+ "learning_rate": 1.3547725373713403e-08,
1221
+ "loss": 4.4819,
1222
+ "step": 850
1223
+ },
1224
+ {
1225
+ "epoch": 2.859531772575251,
1226
+ "grad_norm": 2.34375,
1227
+ "learning_rate": 1.0823490035218986e-08,
1228
+ "loss": 4.4775,
1229
+ "step": 855
1230
+ },
1231
+ {
1232
+ "epoch": 2.8762541806020065,
1233
+ "grad_norm": 2.3125,
1234
+ "learning_rate": 8.403263672473792e-09,
1235
+ "loss": 4.4999,
1236
+ "step": 860
1237
+ },
1238
+ {
1239
+ "epoch": 2.8929765886287626,
1240
+ "grad_norm": 2.390625,
1241
+ "learning_rate": 6.2877901067573955e-09,
1242
+ "loss": 4.4673,
1243
+ "step": 865
1244
+ },
1245
+ {
1246
+ "epoch": 2.9096989966555182,
1247
+ "grad_norm": 2.53125,
1248
+ "learning_rate": 4.477719498021781e-09,
1249
+ "loss": 4.4816,
1250
+ "step": 870
1251
+ },
1252
+ {
1253
+ "epoch": 2.9264214046822743,
1254
+ "grad_norm": 2.265625,
1255
+ "learning_rate": 2.973608145073081e-09,
1256
+ "loss": 4.4942,
1257
+ "step": 875
1258
+ },
1259
+ {
1260
+ "epoch": 2.94314381270903,
1261
+ "grad_norm": 2.390625,
1262
+ "learning_rate": 1.7759183146021094e-09,
1263
+ "loss": 4.5004,
1264
+ "step": 880
1265
+ },
1266
+ {
1267
+ "epoch": 2.959866220735786,
1268
+ "grad_norm": 2.28125,
1269
+ "learning_rate": 8.850180991131218e-10,
1270
+ "loss": 4.4763,
1271
+ "step": 885
1272
+ },
1273
+ {
1274
+ "epoch": 2.976588628762542,
1275
+ "grad_norm": 2.515625,
1276
+ "learning_rate": 3.0118130379575e-10,
1277
+ "loss": 4.4912,
1278
+ "step": 890
1279
+ },
1280
+ {
1281
+ "epoch": 2.9933110367892977,
1282
+ "grad_norm": 2.4375,
1283
+ "learning_rate": 2.45873623754278e-11,
1284
+ "loss": 4.473,
1285
+ "step": 895
1286
+ },
1287
+ {
1288
+ "epoch": 3.0,
1289
+ "step": 897,
1290
+ "total_flos": 5.294195477904884e+18,
1291
+ "train_loss": 4.635029224514828,
1292
+ "train_runtime": 4299.9756,
1293
+ "train_samples_per_second": 13.31,
1294
+ "train_steps_per_second": 0.209
1295
+ }
1296
+ ],
1297
+ "logging_steps": 5,
1298
+ "max_steps": 897,
1299
+ "num_input_tokens_seen": 0,
1300
+ "num_train_epochs": 3,
1301
+ "save_steps": 300,
1302
+ "stateful_callbacks": {
1303
+ "TrainerControl": {
1304
+ "args": {
1305
+ "should_epoch_stop": false,
1306
+ "should_evaluate": false,
1307
+ "should_log": false,
1308
+ "should_save": true,
1309
+ "should_training_stop": true
1310
+ },
1311
+ "attributes": {}
1312
+ }
1313
+ },
1314
+ "total_flos": 5.294195477904884e+18,
1315
+ "train_batch_size": 8,
1316
+ "trial_name": null,
1317
+ "trial_params": null
1318
+ }