akseljoonas HF Staff commited on
Commit
4f87cc3
·
verified ·
1 Parent(s): 87edf04

Model save

Browse files
Files changed (5) hide show
  1. README.md +58 -0
  2. all_results.json +8 -0
  3. generation_config.json +14 -0
  4. train_results.json +8 -0
  5. trainer_state.json +1620 -0
README.md ADDED
@@ -0,0 +1,58 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: Qwen/Qwen2.5-7B-Instruct
3
+ library_name: transformers
4
+ model_name: Agentic-Qwen2.5-7B-e2-lr2-b8
5
+ tags:
6
+ - generated_from_trainer
7
+ - trl
8
+ - sft
9
+ licence: license
10
+ ---
11
+
12
+ # Model Card for Agentic-Qwen2.5-7B-e2-lr2-b8
13
+
14
+ This model is a fine-tuned version of [Qwen/Qwen2.5-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-7B-Instruct).
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="akseljoonas/Agentic-Qwen2.5-7B-e2-lr2-b8", 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/akseljoonas-university-of-groningen/huggingface/runs/wa778klj)
31
+
32
+
33
+ This model was trained with SFT.
34
+
35
+ ### Framework versions
36
+
37
+ - TRL: 0.18.1
38
+ - Transformers: 4.52.4
39
+ - Pytorch: 2.6.0
40
+ - Datasets: 3.6.0
41
+ - Tokenizers: 0.21.1
42
+
43
+ ## Citations
44
+
45
+
46
+
47
+ Cite TRL as:
48
+
49
+ ```bibtex
50
+ @misc{vonwerra2022trl,
51
+ title = {{TRL: Transformer Reinforcement Learning}},
52
+ author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
53
+ year = 2020,
54
+ journal = {GitHub repository},
55
+ publisher = {GitHub},
56
+ howpublished = {\url{https://github.com/huggingface/trl}}
57
+ }
58
+ ```
all_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "total_flos": 114029121830912.0,
3
+ "train_loss": 0.5227763477253587,
4
+ "train_runtime": 1326.8074,
5
+ "train_samples": 9217,
6
+ "train_samples_per_second": 5.273,
7
+ "train_steps_per_second": 0.66
8
+ }
generation_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 151643,
3
+ "do_sample": true,
4
+ "eos_token_id": [
5
+ 151645,
6
+ 151643
7
+ ],
8
+ "pad_token_id": 151643,
9
+ "repetition_penalty": 1.05,
10
+ "temperature": 0.7,
11
+ "top_k": 20,
12
+ "top_p": 0.8,
13
+ "transformers_version": "4.52.4"
14
+ }
train_results.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "total_flos": 114029121830912.0,
3
+ "train_loss": 0.5227763477253587,
4
+ "train_runtime": 1326.8074,
5
+ "train_samples": 9217,
6
+ "train_samples_per_second": 5.273,
7
+ "train_steps_per_second": 0.66
8
+ }
trainer_state.json ADDED
@@ -0,0 +1,1620 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 2.0,
6
+ "eval_steps": 500,
7
+ "global_step": 876,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.01141552511415525,
14
+ "grad_norm": 6.992737402065142,
15
+ "learning_rate": 9.090909090909091e-07,
16
+ "loss": 1.1455,
17
+ "mean_token_accuracy": 0.7416280031204223,
18
+ "num_tokens": 327680.0,
19
+ "step": 5
20
+ },
21
+ {
22
+ "epoch": 0.0228310502283105,
23
+ "grad_norm": 2.415629659704594,
24
+ "learning_rate": 2.0454545454545457e-06,
25
+ "loss": 1.235,
26
+ "mean_token_accuracy": 0.7398699760437012,
27
+ "num_tokens": 655360.0,
28
+ "step": 10
29
+ },
30
+ {
31
+ "epoch": 0.03424657534246575,
32
+ "grad_norm": 1.4642058363745276,
33
+ "learning_rate": 3.181818181818182e-06,
34
+ "loss": 1.0538,
35
+ "mean_token_accuracy": 0.7665150761604309,
36
+ "num_tokens": 983040.0,
37
+ "step": 15
38
+ },
39
+ {
40
+ "epoch": 0.045662100456621,
41
+ "grad_norm": 1.533023139818528,
42
+ "learning_rate": 4.3181818181818185e-06,
43
+ "loss": 0.9744,
44
+ "mean_token_accuracy": 0.7759034276008606,
45
+ "num_tokens": 1310720.0,
46
+ "step": 20
47
+ },
48
+ {
49
+ "epoch": 0.05707762557077625,
50
+ "grad_norm": 1.1766831732403387,
51
+ "learning_rate": 5.4545454545454545e-06,
52
+ "loss": 1.0618,
53
+ "mean_token_accuracy": 0.7671541452407837,
54
+ "num_tokens": 1633950.0,
55
+ "step": 25
56
+ },
57
+ {
58
+ "epoch": 0.0684931506849315,
59
+ "grad_norm": 1.1628361612953064,
60
+ "learning_rate": 6.590909090909091e-06,
61
+ "loss": 0.8032,
62
+ "mean_token_accuracy": 0.8042577266693115,
63
+ "num_tokens": 1961630.0,
64
+ "step": 30
65
+ },
66
+ {
67
+ "epoch": 0.07990867579908675,
68
+ "grad_norm": 0.9724642240958964,
69
+ "learning_rate": 7.727272727272727e-06,
70
+ "loss": 1.1505,
71
+ "mean_token_accuracy": 0.746834933757782,
72
+ "num_tokens": 2289310.0,
73
+ "step": 35
74
+ },
75
+ {
76
+ "epoch": 0.091324200913242,
77
+ "grad_norm": 0.9384838459126222,
78
+ "learning_rate": 8.863636363636365e-06,
79
+ "loss": 0.7937,
80
+ "mean_token_accuracy": 0.805460262298584,
81
+ "num_tokens": 2616990.0,
82
+ "step": 40
83
+ },
84
+ {
85
+ "epoch": 0.10273972602739725,
86
+ "grad_norm": 1.1000363909812299,
87
+ "learning_rate": 1e-05,
88
+ "loss": 0.9913,
89
+ "mean_token_accuracy": 0.776715087890625,
90
+ "num_tokens": 2943724.0,
91
+ "step": 45
92
+ },
93
+ {
94
+ "epoch": 0.1141552511415525,
95
+ "grad_norm": 1.0168831507859384,
96
+ "learning_rate": 1.1136363636363637e-05,
97
+ "loss": 0.9996,
98
+ "mean_token_accuracy": 0.7746000409126281,
99
+ "num_tokens": 3265203.0,
100
+ "step": 50
101
+ },
102
+ {
103
+ "epoch": 0.12557077625570776,
104
+ "grad_norm": 1.0053101286866857,
105
+ "learning_rate": 1.2272727272727274e-05,
106
+ "loss": 1.0377,
107
+ "mean_token_accuracy": 0.7730161190032959,
108
+ "num_tokens": 3592883.0,
109
+ "step": 55
110
+ },
111
+ {
112
+ "epoch": 0.136986301369863,
113
+ "grad_norm": 1.2044011110569786,
114
+ "learning_rate": 1.3409090909090909e-05,
115
+ "loss": 1.046,
116
+ "mean_token_accuracy": 0.7707313656806946,
117
+ "num_tokens": 3917738.0,
118
+ "step": 60
119
+ },
120
+ {
121
+ "epoch": 0.14840182648401826,
122
+ "grad_norm": 1.1613218830446566,
123
+ "learning_rate": 1.4545454545454546e-05,
124
+ "loss": 0.7454,
125
+ "mean_token_accuracy": 0.8163699388504029,
126
+ "num_tokens": 4239105.0,
127
+ "step": 65
128
+ },
129
+ {
130
+ "epoch": 0.1598173515981735,
131
+ "grad_norm": 1.5647320460476475,
132
+ "learning_rate": 1.5681818181818182e-05,
133
+ "loss": 0.8115,
134
+ "mean_token_accuracy": 0.8110944986343384,
135
+ "num_tokens": 4566785.0,
136
+ "step": 70
137
+ },
138
+ {
139
+ "epoch": 0.17123287671232876,
140
+ "grad_norm": 0.9484598379066953,
141
+ "learning_rate": 1.681818181818182e-05,
142
+ "loss": 0.7076,
143
+ "mean_token_accuracy": 0.8221047043800354,
144
+ "num_tokens": 4889208.0,
145
+ "step": 75
146
+ },
147
+ {
148
+ "epoch": 0.182648401826484,
149
+ "grad_norm": 4.152679184463501,
150
+ "learning_rate": 1.7954545454545454e-05,
151
+ "loss": 0.8595,
152
+ "mean_token_accuracy": 0.7962550401687623,
153
+ "num_tokens": 5216888.0,
154
+ "step": 80
155
+ },
156
+ {
157
+ "epoch": 0.19406392694063926,
158
+ "grad_norm": 1.0527043778305278,
159
+ "learning_rate": 1.9090909090909094e-05,
160
+ "loss": 0.793,
161
+ "mean_token_accuracy": 0.8105325341224671,
162
+ "num_tokens": 5542736.0,
163
+ "step": 85
164
+ },
165
+ {
166
+ "epoch": 0.2054794520547945,
167
+ "grad_norm": 3.756868700105614,
168
+ "learning_rate": 1.9974619289340104e-05,
169
+ "loss": 0.6592,
170
+ "mean_token_accuracy": 0.8405689120292663,
171
+ "num_tokens": 5870416.0,
172
+ "step": 90
173
+ },
174
+ {
175
+ "epoch": 0.21689497716894976,
176
+ "grad_norm": 0.9345450078461636,
177
+ "learning_rate": 1.984771573604061e-05,
178
+ "loss": 0.7805,
179
+ "mean_token_accuracy": 0.8161181330680847,
180
+ "num_tokens": 6192634.0,
181
+ "step": 95
182
+ },
183
+ {
184
+ "epoch": 0.228310502283105,
185
+ "grad_norm": 0.9975708882682292,
186
+ "learning_rate": 1.972081218274112e-05,
187
+ "loss": 0.6786,
188
+ "mean_token_accuracy": 0.8344534754753112,
189
+ "num_tokens": 6513849.0,
190
+ "step": 100
191
+ },
192
+ {
193
+ "epoch": 0.23972602739726026,
194
+ "grad_norm": 4.172984059136262,
195
+ "learning_rate": 1.9593908629441626e-05,
196
+ "loss": 0.839,
197
+ "mean_token_accuracy": 0.8037724494934082,
198
+ "num_tokens": 6841529.0,
199
+ "step": 105
200
+ },
201
+ {
202
+ "epoch": 0.2511415525114155,
203
+ "grad_norm": 1.0270411941758375,
204
+ "learning_rate": 1.9467005076142135e-05,
205
+ "loss": 0.5501,
206
+ "mean_token_accuracy": 0.8536208033561706,
207
+ "num_tokens": 7162666.0,
208
+ "step": 110
209
+ },
210
+ {
211
+ "epoch": 0.2625570776255708,
212
+ "grad_norm": 1.029010833438825,
213
+ "learning_rate": 1.934010152284264e-05,
214
+ "loss": 0.8554,
215
+ "mean_token_accuracy": 0.8019899964332581,
216
+ "num_tokens": 7490346.0,
217
+ "step": 115
218
+ },
219
+ {
220
+ "epoch": 0.273972602739726,
221
+ "grad_norm": 0.9572862665009986,
222
+ "learning_rate": 1.921319796954315e-05,
223
+ "loss": 0.6513,
224
+ "mean_token_accuracy": 0.8454767465591431,
225
+ "num_tokens": 7818026.0,
226
+ "step": 120
227
+ },
228
+ {
229
+ "epoch": 0.2853881278538813,
230
+ "grad_norm": 0.7778398380031003,
231
+ "learning_rate": 1.9086294416243656e-05,
232
+ "loss": 0.7328,
233
+ "mean_token_accuracy": 0.8308659195899963,
234
+ "num_tokens": 8138367.0,
235
+ "step": 125
236
+ },
237
+ {
238
+ "epoch": 0.2968036529680365,
239
+ "grad_norm": 4.622897980437575,
240
+ "learning_rate": 1.8959390862944166e-05,
241
+ "loss": 0.7628,
242
+ "mean_token_accuracy": 0.8257263779640198,
243
+ "num_tokens": 8454930.0,
244
+ "step": 130
245
+ },
246
+ {
247
+ "epoch": 0.3082191780821918,
248
+ "grad_norm": 0.8740524230221044,
249
+ "learning_rate": 1.8832487309644672e-05,
250
+ "loss": 0.6121,
251
+ "mean_token_accuracy": 0.8443840742111206,
252
+ "num_tokens": 8782610.0,
253
+ "step": 135
254
+ },
255
+ {
256
+ "epoch": 0.319634703196347,
257
+ "grad_norm": 2.209549198454251,
258
+ "learning_rate": 1.8705583756345178e-05,
259
+ "loss": 0.7859,
260
+ "mean_token_accuracy": 0.8160686254501343,
261
+ "num_tokens": 9107408.0,
262
+ "step": 140
263
+ },
264
+ {
265
+ "epoch": 0.3310502283105023,
266
+ "grad_norm": 0.7958690086371107,
267
+ "learning_rate": 1.8578680203045687e-05,
268
+ "loss": 0.6972,
269
+ "mean_token_accuracy": 0.8361982583999634,
270
+ "num_tokens": 9435088.0,
271
+ "step": 145
272
+ },
273
+ {
274
+ "epoch": 0.3424657534246575,
275
+ "grad_norm": 0.9066549747651292,
276
+ "learning_rate": 1.8451776649746193e-05,
277
+ "loss": 0.7197,
278
+ "mean_token_accuracy": 0.8286533951759338,
279
+ "num_tokens": 9762768.0,
280
+ "step": 150
281
+ },
282
+ {
283
+ "epoch": 0.3538812785388128,
284
+ "grad_norm": 0.8767014153036071,
285
+ "learning_rate": 1.8324873096446703e-05,
286
+ "loss": 0.6008,
287
+ "mean_token_accuracy": 0.8508149266242981,
288
+ "num_tokens": 10090448.0,
289
+ "step": 155
290
+ },
291
+ {
292
+ "epoch": 0.365296803652968,
293
+ "grad_norm": 0.91485360531879,
294
+ "learning_rate": 1.819796954314721e-05,
295
+ "loss": 0.6953,
296
+ "mean_token_accuracy": 0.8278628945350647,
297
+ "num_tokens": 10418128.0,
298
+ "step": 160
299
+ },
300
+ {
301
+ "epoch": 0.3767123287671233,
302
+ "grad_norm": 3.4107484781219566,
303
+ "learning_rate": 1.8071065989847718e-05,
304
+ "loss": 0.6957,
305
+ "mean_token_accuracy": 0.8255402326583863,
306
+ "num_tokens": 10745808.0,
307
+ "step": 165
308
+ },
309
+ {
310
+ "epoch": 0.3881278538812785,
311
+ "grad_norm": 0.7941767715954852,
312
+ "learning_rate": 1.7944162436548224e-05,
313
+ "loss": 0.5556,
314
+ "mean_token_accuracy": 0.8614760160446167,
315
+ "num_tokens": 11073488.0,
316
+ "step": 170
317
+ },
318
+ {
319
+ "epoch": 0.3995433789954338,
320
+ "grad_norm": 0.921383074036204,
321
+ "learning_rate": 1.781725888324873e-05,
322
+ "loss": 0.9175,
323
+ "mean_token_accuracy": 0.7995509386062623,
324
+ "num_tokens": 11397050.0,
325
+ "step": 175
326
+ },
327
+ {
328
+ "epoch": 0.410958904109589,
329
+ "grad_norm": 9.975806692678526,
330
+ "learning_rate": 1.769035532994924e-05,
331
+ "loss": 0.6974,
332
+ "mean_token_accuracy": 0.8399378895759583,
333
+ "num_tokens": 11722000.0,
334
+ "step": 180
335
+ },
336
+ {
337
+ "epoch": 0.4223744292237443,
338
+ "grad_norm": 1.714112308296964,
339
+ "learning_rate": 1.7563451776649745e-05,
340
+ "loss": 0.8606,
341
+ "mean_token_accuracy": 0.807651698589325,
342
+ "num_tokens": 12049680.0,
343
+ "step": 185
344
+ },
345
+ {
346
+ "epoch": 0.4337899543378995,
347
+ "grad_norm": 0.802912357071434,
348
+ "learning_rate": 1.7436548223350255e-05,
349
+ "loss": 0.6242,
350
+ "mean_token_accuracy": 0.8502655267715454,
351
+ "num_tokens": 12377360.0,
352
+ "step": 190
353
+ },
354
+ {
355
+ "epoch": 0.4452054794520548,
356
+ "grad_norm": 0.898460488405579,
357
+ "learning_rate": 1.730964467005076e-05,
358
+ "loss": 0.6842,
359
+ "mean_token_accuracy": 0.8378538608551025,
360
+ "num_tokens": 12699376.0,
361
+ "step": 195
362
+ },
363
+ {
364
+ "epoch": 0.45662100456621,
365
+ "grad_norm": 0.8390361124871176,
366
+ "learning_rate": 1.718274111675127e-05,
367
+ "loss": 0.6723,
368
+ "mean_token_accuracy": 0.8452905654907227,
369
+ "num_tokens": 13027056.0,
370
+ "step": 200
371
+ },
372
+ {
373
+ "epoch": 0.4680365296803653,
374
+ "grad_norm": 0.8712620260174776,
375
+ "learning_rate": 1.705583756345178e-05,
376
+ "loss": 0.5404,
377
+ "mean_token_accuracy": 0.8614973783493042,
378
+ "num_tokens": 13354736.0,
379
+ "step": 205
380
+ },
381
+ {
382
+ "epoch": 0.4794520547945205,
383
+ "grad_norm": 0.8567527998760598,
384
+ "learning_rate": 1.6928934010152286e-05,
385
+ "loss": 0.5684,
386
+ "mean_token_accuracy": 0.8579599499702454,
387
+ "num_tokens": 13682416.0,
388
+ "step": 210
389
+ },
390
+ {
391
+ "epoch": 0.4908675799086758,
392
+ "grad_norm": 0.9888830651281572,
393
+ "learning_rate": 1.6802030456852795e-05,
394
+ "loss": 0.5698,
395
+ "mean_token_accuracy": 0.856434679031372,
396
+ "num_tokens": 14003534.0,
397
+ "step": 215
398
+ },
399
+ {
400
+ "epoch": 0.502283105022831,
401
+ "grad_norm": 0.9572108430973989,
402
+ "learning_rate": 1.66751269035533e-05,
403
+ "loss": 0.5145,
404
+ "mean_token_accuracy": 0.8691582202911377,
405
+ "num_tokens": 14331214.0,
406
+ "step": 220
407
+ },
408
+ {
409
+ "epoch": 0.5136986301369864,
410
+ "grad_norm": 0.9872437195664178,
411
+ "learning_rate": 1.654822335025381e-05,
412
+ "loss": 0.6956,
413
+ "mean_token_accuracy": 0.8335458040237427,
414
+ "num_tokens": 14653317.0,
415
+ "step": 225
416
+ },
417
+ {
418
+ "epoch": 0.5251141552511416,
419
+ "grad_norm": 0.808238593307361,
420
+ "learning_rate": 1.6421319796954316e-05,
421
+ "loss": 0.8908,
422
+ "mean_token_accuracy": 0.8048653244972229,
423
+ "num_tokens": 14976261.0,
424
+ "step": 230
425
+ },
426
+ {
427
+ "epoch": 0.5365296803652968,
428
+ "grad_norm": 1.8900423969196503,
429
+ "learning_rate": 1.6294416243654826e-05,
430
+ "loss": 0.6421,
431
+ "mean_token_accuracy": 0.8384446501731873,
432
+ "num_tokens": 15303941.0,
433
+ "step": 235
434
+ },
435
+ {
436
+ "epoch": 0.547945205479452,
437
+ "grad_norm": 3.008411378596028,
438
+ "learning_rate": 1.6167512690355332e-05,
439
+ "loss": 0.6225,
440
+ "mean_token_accuracy": 0.859709620475769,
441
+ "num_tokens": 15622106.0,
442
+ "step": 240
443
+ },
444
+ {
445
+ "epoch": 0.5593607305936074,
446
+ "grad_norm": 0.7985333046643842,
447
+ "learning_rate": 1.6040609137055838e-05,
448
+ "loss": 0.5424,
449
+ "mean_token_accuracy": 0.8691208839416504,
450
+ "num_tokens": 15939817.0,
451
+ "step": 245
452
+ },
453
+ {
454
+ "epoch": 0.5707762557077626,
455
+ "grad_norm": 0.8112819291065752,
456
+ "learning_rate": 1.5913705583756347e-05,
457
+ "loss": 0.5527,
458
+ "mean_token_accuracy": 0.8647384285926819,
459
+ "num_tokens": 16260487.0,
460
+ "step": 250
461
+ },
462
+ {
463
+ "epoch": 0.5821917808219178,
464
+ "grad_norm": 0.7922206625194624,
465
+ "learning_rate": 1.5786802030456853e-05,
466
+ "loss": 0.4881,
467
+ "mean_token_accuracy": 0.8762391686439515,
468
+ "num_tokens": 16588167.0,
469
+ "step": 255
470
+ },
471
+ {
472
+ "epoch": 0.593607305936073,
473
+ "grad_norm": 0.6279625943118422,
474
+ "learning_rate": 1.5659898477157363e-05,
475
+ "loss": 0.4119,
476
+ "mean_token_accuracy": 0.8932731032371521,
477
+ "num_tokens": 16915847.0,
478
+ "step": 260
479
+ },
480
+ {
481
+ "epoch": 0.6050228310502284,
482
+ "grad_norm": 0.7818186445390249,
483
+ "learning_rate": 1.553299492385787e-05,
484
+ "loss": 0.588,
485
+ "mean_token_accuracy": 0.8551062345504761,
486
+ "num_tokens": 17243527.0,
487
+ "step": 265
488
+ },
489
+ {
490
+ "epoch": 0.6164383561643836,
491
+ "grad_norm": 1.9361225263939912,
492
+ "learning_rate": 1.5406091370558375e-05,
493
+ "loss": 0.637,
494
+ "mean_token_accuracy": 0.8508088111877441,
495
+ "num_tokens": 17571207.0,
496
+ "step": 270
497
+ },
498
+ {
499
+ "epoch": 0.6278538812785388,
500
+ "grad_norm": 1.8209406442195541,
501
+ "learning_rate": 1.5279187817258884e-05,
502
+ "loss": 0.551,
503
+ "mean_token_accuracy": 0.8664357185363769,
504
+ "num_tokens": 17898887.0,
505
+ "step": 275
506
+ },
507
+ {
508
+ "epoch": 0.639269406392694,
509
+ "grad_norm": 1.310483798334614,
510
+ "learning_rate": 1.5152284263959392e-05,
511
+ "loss": 0.5156,
512
+ "mean_token_accuracy": 0.8750967144966125,
513
+ "num_tokens": 18222804.0,
514
+ "step": 280
515
+ },
516
+ {
517
+ "epoch": 0.6506849315068494,
518
+ "grad_norm": 0.7068405371324017,
519
+ "learning_rate": 1.50253807106599e-05,
520
+ "loss": 0.6022,
521
+ "mean_token_accuracy": 0.8513246297836303,
522
+ "num_tokens": 18550484.0,
523
+ "step": 285
524
+ },
525
+ {
526
+ "epoch": 0.6621004566210046,
527
+ "grad_norm": 0.6989325786526743,
528
+ "learning_rate": 1.4898477157360407e-05,
529
+ "loss": 0.533,
530
+ "mean_token_accuracy": 0.8711689591407776,
531
+ "num_tokens": 18871490.0,
532
+ "step": 290
533
+ },
534
+ {
535
+ "epoch": 0.6735159817351598,
536
+ "grad_norm": 0.6619420062510123,
537
+ "learning_rate": 1.4771573604060917e-05,
538
+ "loss": 0.4451,
539
+ "mean_token_accuracy": 0.8861830115318299,
540
+ "num_tokens": 19199170.0,
541
+ "step": 295
542
+ },
543
+ {
544
+ "epoch": 0.684931506849315,
545
+ "grad_norm": 0.7553279039784921,
546
+ "learning_rate": 1.4644670050761423e-05,
547
+ "loss": 0.7542,
548
+ "mean_token_accuracy": 0.8355847954750061,
549
+ "num_tokens": 19526850.0,
550
+ "step": 300
551
+ },
552
+ {
553
+ "epoch": 0.6963470319634704,
554
+ "grad_norm": 1.5890699195218203,
555
+ "learning_rate": 1.4517766497461929e-05,
556
+ "loss": 0.6959,
557
+ "mean_token_accuracy": 0.8418050169944763,
558
+ "num_tokens": 19854530.0,
559
+ "step": 305
560
+ },
561
+ {
562
+ "epoch": 0.7077625570776256,
563
+ "grad_norm": 0.7076959599117634,
564
+ "learning_rate": 1.4390862944162438e-05,
565
+ "loss": 0.424,
566
+ "mean_token_accuracy": 0.8922991752624512,
567
+ "num_tokens": 20180551.0,
568
+ "step": 310
569
+ },
570
+ {
571
+ "epoch": 0.7191780821917808,
572
+ "grad_norm": 0.7685020765926709,
573
+ "learning_rate": 1.4263959390862944e-05,
574
+ "loss": 0.4324,
575
+ "mean_token_accuracy": 0.8872226476669312,
576
+ "num_tokens": 20502709.0,
577
+ "step": 315
578
+ },
579
+ {
580
+ "epoch": 0.730593607305936,
581
+ "grad_norm": 0.8329200051413271,
582
+ "learning_rate": 1.4137055837563453e-05,
583
+ "loss": 0.4938,
584
+ "mean_token_accuracy": 0.8742125391960144,
585
+ "num_tokens": 20830389.0,
586
+ "step": 320
587
+ },
588
+ {
589
+ "epoch": 0.7420091324200914,
590
+ "grad_norm": 0.7213569799331959,
591
+ "learning_rate": 1.4010152284263961e-05,
592
+ "loss": 0.5128,
593
+ "mean_token_accuracy": 0.8703390240669251,
594
+ "num_tokens": 21154626.0,
595
+ "step": 325
596
+ },
597
+ {
598
+ "epoch": 0.7534246575342466,
599
+ "grad_norm": 0.8855772820793749,
600
+ "learning_rate": 1.3883248730964469e-05,
601
+ "loss": 0.402,
602
+ "mean_token_accuracy": 0.8942864060401916,
603
+ "num_tokens": 21482306.0,
604
+ "step": 330
605
+ },
606
+ {
607
+ "epoch": 0.7648401826484018,
608
+ "grad_norm": 2.621149129137716,
609
+ "learning_rate": 1.3756345177664977e-05,
610
+ "loss": 0.4577,
611
+ "mean_token_accuracy": 0.8836394309997558,
612
+ "num_tokens": 21803980.0,
613
+ "step": 335
614
+ },
615
+ {
616
+ "epoch": 0.776255707762557,
617
+ "grad_norm": 0.7585062993749856,
618
+ "learning_rate": 1.3629441624365483e-05,
619
+ "loss": 0.4665,
620
+ "mean_token_accuracy": 0.8822152256965637,
621
+ "num_tokens": 22131660.0,
622
+ "step": 340
623
+ },
624
+ {
625
+ "epoch": 0.7876712328767124,
626
+ "grad_norm": 1.1354904461607713,
627
+ "learning_rate": 1.3502538071065992e-05,
628
+ "loss": 0.654,
629
+ "mean_token_accuracy": 0.8456507325172424,
630
+ "num_tokens": 22459340.0,
631
+ "step": 345
632
+ },
633
+ {
634
+ "epoch": 0.7990867579908676,
635
+ "grad_norm": 0.6478840981965673,
636
+ "learning_rate": 1.3375634517766498e-05,
637
+ "loss": 0.5003,
638
+ "mean_token_accuracy": 0.8774447679519654,
639
+ "num_tokens": 22787020.0,
640
+ "step": 350
641
+ },
642
+ {
643
+ "epoch": 0.8105022831050228,
644
+ "grad_norm": 0.7967917061473204,
645
+ "learning_rate": 1.3248730964467007e-05,
646
+ "loss": 0.8532,
647
+ "mean_token_accuracy": 0.8107648730278015,
648
+ "num_tokens": 23114700.0,
649
+ "step": 355
650
+ },
651
+ {
652
+ "epoch": 0.821917808219178,
653
+ "grad_norm": 0.6340224155697725,
654
+ "learning_rate": 1.3121827411167513e-05,
655
+ "loss": 0.5025,
656
+ "mean_token_accuracy": 0.8769076704978943,
657
+ "num_tokens": 23441899.0,
658
+ "step": 360
659
+ },
660
+ {
661
+ "epoch": 0.8333333333333334,
662
+ "grad_norm": 3.460123070812393,
663
+ "learning_rate": 1.2994923857868021e-05,
664
+ "loss": 0.5892,
665
+ "mean_token_accuracy": 0.854703426361084,
666
+ "num_tokens": 23765036.0,
667
+ "step": 365
668
+ },
669
+ {
670
+ "epoch": 0.8447488584474886,
671
+ "grad_norm": 1.1090601927685084,
672
+ "learning_rate": 1.2868020304568529e-05,
673
+ "loss": 0.6692,
674
+ "mean_token_accuracy": 0.8481107592582703,
675
+ "num_tokens": 24086229.0,
676
+ "step": 370
677
+ },
678
+ {
679
+ "epoch": 0.8561643835616438,
680
+ "grad_norm": 0.8194701007896857,
681
+ "learning_rate": 1.2741116751269036e-05,
682
+ "loss": 0.5795,
683
+ "mean_token_accuracy": 0.8683433055877685,
684
+ "num_tokens": 24413909.0,
685
+ "step": 375
686
+ },
687
+ {
688
+ "epoch": 0.867579908675799,
689
+ "grad_norm": 3.6513876578252713,
690
+ "learning_rate": 1.2614213197969544e-05,
691
+ "loss": 0.4849,
692
+ "mean_token_accuracy": 0.8803595423698425,
693
+ "num_tokens": 24741589.0,
694
+ "step": 380
695
+ },
696
+ {
697
+ "epoch": 0.8789954337899544,
698
+ "grad_norm": 0.7086135942173815,
699
+ "learning_rate": 1.2487309644670052e-05,
700
+ "loss": 0.3962,
701
+ "mean_token_accuracy": 0.8968823671340942,
702
+ "num_tokens": 25067428.0,
703
+ "step": 385
704
+ },
705
+ {
706
+ "epoch": 0.8904109589041096,
707
+ "grad_norm": 0.7079243235112125,
708
+ "learning_rate": 1.236040609137056e-05,
709
+ "loss": 0.6134,
710
+ "mean_token_accuracy": 0.8499247193336487,
711
+ "num_tokens": 25385256.0,
712
+ "step": 390
713
+ },
714
+ {
715
+ "epoch": 0.9018264840182648,
716
+ "grad_norm": 0.7961737645515603,
717
+ "learning_rate": 1.2233502538071067e-05,
718
+ "loss": 0.458,
719
+ "mean_token_accuracy": 0.8771883845329285,
720
+ "num_tokens": 25712936.0,
721
+ "step": 395
722
+ },
723
+ {
724
+ "epoch": 0.91324200913242,
725
+ "grad_norm": 0.8924837764546859,
726
+ "learning_rate": 1.2106598984771573e-05,
727
+ "loss": 0.5281,
728
+ "mean_token_accuracy": 0.8718532443046569,
729
+ "num_tokens": 26040616.0,
730
+ "step": 400
731
+ },
732
+ {
733
+ "epoch": 0.9246575342465754,
734
+ "grad_norm": 0.632245097982913,
735
+ "learning_rate": 1.1979695431472083e-05,
736
+ "loss": 0.5217,
737
+ "mean_token_accuracy": 0.8761781215667724,
738
+ "num_tokens": 26368296.0,
739
+ "step": 405
740
+ },
741
+ {
742
+ "epoch": 0.9360730593607306,
743
+ "grad_norm": 1.0801295341120962,
744
+ "learning_rate": 1.1852791878172589e-05,
745
+ "loss": 0.5067,
746
+ "mean_token_accuracy": 0.8805297255516052,
747
+ "num_tokens": 26683654.0,
748
+ "step": 410
749
+ },
750
+ {
751
+ "epoch": 0.9474885844748858,
752
+ "grad_norm": 0.706906455968574,
753
+ "learning_rate": 1.1725888324873098e-05,
754
+ "loss": 0.4439,
755
+ "mean_token_accuracy": 0.8839793562889099,
756
+ "num_tokens": 27011334.0,
757
+ "step": 415
758
+ },
759
+ {
760
+ "epoch": 0.958904109589041,
761
+ "grad_norm": 0.6821497612218076,
762
+ "learning_rate": 1.1598984771573604e-05,
763
+ "loss": 0.6571,
764
+ "mean_token_accuracy": 0.8574594140052796,
765
+ "num_tokens": 27339014.0,
766
+ "step": 420
767
+ },
768
+ {
769
+ "epoch": 0.9703196347031964,
770
+ "grad_norm": 0.7061552939479407,
771
+ "learning_rate": 1.1472081218274113e-05,
772
+ "loss": 0.4375,
773
+ "mean_token_accuracy": 0.8928519010543823,
774
+ "num_tokens": 27666694.0,
775
+ "step": 425
776
+ },
777
+ {
778
+ "epoch": 0.9817351598173516,
779
+ "grad_norm": 1.1886235188151473,
780
+ "learning_rate": 1.1345177664974621e-05,
781
+ "loss": 0.5535,
782
+ "mean_token_accuracy": 0.8669820427894592,
783
+ "num_tokens": 27994374.0,
784
+ "step": 430
785
+ },
786
+ {
787
+ "epoch": 0.9931506849315068,
788
+ "grad_norm": 1.4431539889842626,
789
+ "learning_rate": 1.1218274111675127e-05,
790
+ "loss": 0.6639,
791
+ "mean_token_accuracy": 0.8481354475021362,
792
+ "num_tokens": 28312497.0,
793
+ "step": 435
794
+ },
795
+ {
796
+ "epoch": 1.004566210045662,
797
+ "grad_norm": 1.5391306247642447,
798
+ "learning_rate": 1.1091370558375637e-05,
799
+ "loss": 0.4529,
800
+ "mean_token_accuracy": 0.8871780037879944,
801
+ "num_tokens": 28591025.0,
802
+ "step": 440
803
+ },
804
+ {
805
+ "epoch": 1.0159817351598173,
806
+ "grad_norm": 0.7468130085588452,
807
+ "learning_rate": 1.0964467005076143e-05,
808
+ "loss": 0.6638,
809
+ "mean_token_accuracy": 0.8555121541023254,
810
+ "num_tokens": 28918705.0,
811
+ "step": 445
812
+ },
813
+ {
814
+ "epoch": 1.0273972602739727,
815
+ "grad_norm": 0.8672593588242117,
816
+ "learning_rate": 1.0837563451776652e-05,
817
+ "loss": 0.3973,
818
+ "mean_token_accuracy": 0.9038914561271667,
819
+ "num_tokens": 29246385.0,
820
+ "step": 450
821
+ },
822
+ {
823
+ "epoch": 1.038812785388128,
824
+ "grad_norm": 0.6850118317612971,
825
+ "learning_rate": 1.0710659898477158e-05,
826
+ "loss": 0.3176,
827
+ "mean_token_accuracy": 0.913110101222992,
828
+ "num_tokens": 29570622.0,
829
+ "step": 455
830
+ },
831
+ {
832
+ "epoch": 1.0502283105022832,
833
+ "grad_norm": 0.5820423957625419,
834
+ "learning_rate": 1.0583756345177664e-05,
835
+ "loss": 0.2427,
836
+ "mean_token_accuracy": 0.9367273092269898,
837
+ "num_tokens": 29893759.0,
838
+ "step": 460
839
+ },
840
+ {
841
+ "epoch": 1.0616438356164384,
842
+ "grad_norm": 0.8715978513851912,
843
+ "learning_rate": 1.0456852791878173e-05,
844
+ "loss": 0.3237,
845
+ "mean_token_accuracy": 0.914007818698883,
846
+ "num_tokens": 30217663.0,
847
+ "step": 465
848
+ },
849
+ {
850
+ "epoch": 1.0730593607305936,
851
+ "grad_norm": 0.6797013211475489,
852
+ "learning_rate": 1.0329949238578681e-05,
853
+ "loss": 0.3675,
854
+ "mean_token_accuracy": 0.9028140544891358,
855
+ "num_tokens": 30545343.0,
856
+ "step": 470
857
+ },
858
+ {
859
+ "epoch": 1.0844748858447488,
860
+ "grad_norm": 0.6754570246613306,
861
+ "learning_rate": 1.0203045685279189e-05,
862
+ "loss": 0.2569,
863
+ "mean_token_accuracy": 0.9257373332977294,
864
+ "num_tokens": 30866619.0,
865
+ "step": 475
866
+ },
867
+ {
868
+ "epoch": 1.095890410958904,
869
+ "grad_norm": 0.6546348381500018,
870
+ "learning_rate": 1.0076142131979697e-05,
871
+ "loss": 0.3465,
872
+ "mean_token_accuracy": 0.9099173307418823,
873
+ "num_tokens": 31188635.0,
874
+ "step": 480
875
+ },
876
+ {
877
+ "epoch": 1.1073059360730593,
878
+ "grad_norm": 2.587460038848857,
879
+ "learning_rate": 9.949238578680204e-06,
880
+ "loss": 0.6244,
881
+ "mean_token_accuracy": 0.8603339195251465,
882
+ "num_tokens": 31509305.0,
883
+ "step": 485
884
+ },
885
+ {
886
+ "epoch": 1.1187214611872145,
887
+ "grad_norm": 1.4161831245099408,
888
+ "learning_rate": 9.822335025380712e-06,
889
+ "loss": 0.3955,
890
+ "mean_token_accuracy": 0.9036645412445068,
891
+ "num_tokens": 31830423.0,
892
+ "step": 490
893
+ },
894
+ {
895
+ "epoch": 1.13013698630137,
896
+ "grad_norm": 0.668264002374529,
897
+ "learning_rate": 9.69543147208122e-06,
898
+ "loss": 0.2658,
899
+ "mean_token_accuracy": 0.9273776173591614,
900
+ "num_tokens": 32158103.0,
901
+ "step": 495
902
+ },
903
+ {
904
+ "epoch": 1.1415525114155252,
905
+ "grad_norm": 0.6135235052955457,
906
+ "learning_rate": 9.568527918781727e-06,
907
+ "loss": 0.5066,
908
+ "mean_token_accuracy": 0.8853986024856567,
909
+ "num_tokens": 32485783.0,
910
+ "step": 500
911
+ },
912
+ {
913
+ "epoch": 1.1529680365296804,
914
+ "grad_norm": 1.2046725731942316,
915
+ "learning_rate": 9.441624365482235e-06,
916
+ "loss": 0.3202,
917
+ "mean_token_accuracy": 0.9164499640464783,
918
+ "num_tokens": 32805758.0,
919
+ "step": 505
920
+ },
921
+ {
922
+ "epoch": 1.1643835616438356,
923
+ "grad_norm": 0.7349167612029248,
924
+ "learning_rate": 9.314720812182741e-06,
925
+ "loss": 0.4923,
926
+ "mean_token_accuracy": 0.8750660181045532,
927
+ "num_tokens": 33130741.0,
928
+ "step": 510
929
+ },
930
+ {
931
+ "epoch": 1.1757990867579908,
932
+ "grad_norm": 1.074164219956015,
933
+ "learning_rate": 9.187817258883249e-06,
934
+ "loss": 0.3519,
935
+ "mean_token_accuracy": 0.9124998807907104,
936
+ "num_tokens": 33451878.0,
937
+ "step": 515
938
+ },
939
+ {
940
+ "epoch": 1.187214611872146,
941
+ "grad_norm": 1.3964241755440436,
942
+ "learning_rate": 9.060913705583756e-06,
943
+ "loss": 0.3946,
944
+ "mean_token_accuracy": 0.9055486679077148,
945
+ "num_tokens": 33775440.0,
946
+ "step": 520
947
+ },
948
+ {
949
+ "epoch": 1.1986301369863013,
950
+ "grad_norm": 0.5942042096078787,
951
+ "learning_rate": 8.934010152284264e-06,
952
+ "loss": 0.7269,
953
+ "mean_token_accuracy": 0.8366968870162964,
954
+ "num_tokens": 34096260.0,
955
+ "step": 525
956
+ },
957
+ {
958
+ "epoch": 1.2100456621004567,
959
+ "grad_norm": 0.6349840994578299,
960
+ "learning_rate": 8.807106598984772e-06,
961
+ "loss": 0.3298,
962
+ "mean_token_accuracy": 0.9168631434440613,
963
+ "num_tokens": 34422099.0,
964
+ "step": 530
965
+ },
966
+ {
967
+ "epoch": 1.221461187214612,
968
+ "grad_norm": 0.617680696452804,
969
+ "learning_rate": 8.68020304568528e-06,
970
+ "loss": 0.5785,
971
+ "mean_token_accuracy": 0.8761628627777099,
972
+ "num_tokens": 34749779.0,
973
+ "step": 535
974
+ },
975
+ {
976
+ "epoch": 1.2328767123287672,
977
+ "grad_norm": 0.5912081216640481,
978
+ "learning_rate": 8.553299492385787e-06,
979
+ "loss": 0.4311,
980
+ "mean_token_accuracy": 0.9003967761993408,
981
+ "num_tokens": 35077459.0,
982
+ "step": 540
983
+ },
984
+ {
985
+ "epoch": 1.2442922374429224,
986
+ "grad_norm": 0.6609405907534032,
987
+ "learning_rate": 8.426395939086295e-06,
988
+ "loss": 0.2324,
989
+ "mean_token_accuracy": 0.9359464049339294,
990
+ "num_tokens": 35402314.0,
991
+ "step": 545
992
+ },
993
+ {
994
+ "epoch": 1.2557077625570776,
995
+ "grad_norm": 0.5283056750675951,
996
+ "learning_rate": 8.299492385786803e-06,
997
+ "loss": 0.2821,
998
+ "mean_token_accuracy": 0.9222683548927307,
999
+ "num_tokens": 35729994.0,
1000
+ "step": 550
1001
+ },
1002
+ {
1003
+ "epoch": 1.2671232876712328,
1004
+ "grad_norm": 0.6678530632897957,
1005
+ "learning_rate": 8.17258883248731e-06,
1006
+ "loss": 0.2224,
1007
+ "mean_token_accuracy": 0.9365523099899292,
1008
+ "num_tokens": 36057674.0,
1009
+ "step": 555
1010
+ },
1011
+ {
1012
+ "epoch": 1.278538812785388,
1013
+ "grad_norm": 0.7784807106476361,
1014
+ "learning_rate": 8.045685279187818e-06,
1015
+ "loss": 0.2799,
1016
+ "mean_token_accuracy": 0.9208948850631714,
1017
+ "num_tokens": 36385354.0,
1018
+ "step": 560
1019
+ },
1020
+ {
1021
+ "epoch": 1.2899543378995433,
1022
+ "grad_norm": 0.7633123705001335,
1023
+ "learning_rate": 7.918781725888326e-06,
1024
+ "loss": 0.2822,
1025
+ "mean_token_accuracy": 0.9235532879829407,
1026
+ "num_tokens": 36713034.0,
1027
+ "step": 565
1028
+ },
1029
+ {
1030
+ "epoch": 1.3013698630136985,
1031
+ "grad_norm": 0.7379625295256531,
1032
+ "learning_rate": 7.791878172588832e-06,
1033
+ "loss": 0.3868,
1034
+ "mean_token_accuracy": 0.9011811733245849,
1035
+ "num_tokens": 37040714.0,
1036
+ "step": 570
1037
+ },
1038
+ {
1039
+ "epoch": 1.312785388127854,
1040
+ "grad_norm": 0.5674949104084854,
1041
+ "learning_rate": 7.664974619289341e-06,
1042
+ "loss": 0.2424,
1043
+ "mean_token_accuracy": 0.9339122295379638,
1044
+ "num_tokens": 37368394.0,
1045
+ "step": 575
1046
+ },
1047
+ {
1048
+ "epoch": 1.3242009132420092,
1049
+ "grad_norm": 1.8341220338334356,
1050
+ "learning_rate": 7.538071065989848e-06,
1051
+ "loss": 0.3459,
1052
+ "mean_token_accuracy": 0.9144212961196899,
1053
+ "num_tokens": 37696074.0,
1054
+ "step": 580
1055
+ },
1056
+ {
1057
+ "epoch": 1.3356164383561644,
1058
+ "grad_norm": 0.6734409121936159,
1059
+ "learning_rate": 7.411167512690356e-06,
1060
+ "loss": 0.3816,
1061
+ "mean_token_accuracy": 0.9056525468826294,
1062
+ "num_tokens": 38023754.0,
1063
+ "step": 585
1064
+ },
1065
+ {
1066
+ "epoch": 1.3470319634703196,
1067
+ "grad_norm": 0.9195786782408636,
1068
+ "learning_rate": 7.2842639593908635e-06,
1069
+ "loss": 0.3466,
1070
+ "mean_token_accuracy": 0.9098285913467408,
1071
+ "num_tokens": 38340587.0,
1072
+ "step": 590
1073
+ },
1074
+ {
1075
+ "epoch": 1.3584474885844748,
1076
+ "grad_norm": 0.49825610018486344,
1077
+ "learning_rate": 7.157360406091371e-06,
1078
+ "loss": 0.61,
1079
+ "mean_token_accuracy": 0.8655078768730163,
1080
+ "num_tokens": 38662567.0,
1081
+ "step": 595
1082
+ },
1083
+ {
1084
+ "epoch": 1.36986301369863,
1085
+ "grad_norm": 0.6926219452649162,
1086
+ "learning_rate": 7.03045685279188e-06,
1087
+ "loss": 0.3086,
1088
+ "mean_token_accuracy": 0.9238279938697815,
1089
+ "num_tokens": 38990247.0,
1090
+ "step": 600
1091
+ },
1092
+ {
1093
+ "epoch": 1.3812785388127855,
1094
+ "grad_norm": 0.7095771390111986,
1095
+ "learning_rate": 6.903553299492386e-06,
1096
+ "loss": 0.2727,
1097
+ "mean_token_accuracy": 0.9202539443969726,
1098
+ "num_tokens": 39317927.0,
1099
+ "step": 605
1100
+ },
1101
+ {
1102
+ "epoch": 1.3926940639269407,
1103
+ "grad_norm": 2.037090157837692,
1104
+ "learning_rate": 6.776649746192893e-06,
1105
+ "loss": 0.5876,
1106
+ "mean_token_accuracy": 0.8768282175064087,
1107
+ "num_tokens": 39645607.0,
1108
+ "step": 610
1109
+ },
1110
+ {
1111
+ "epoch": 1.404109589041096,
1112
+ "grad_norm": 0.6087191447134538,
1113
+ "learning_rate": 6.649746192893401e-06,
1114
+ "loss": 0.3792,
1115
+ "mean_token_accuracy": 0.9073647856712341,
1116
+ "num_tokens": 39973287.0,
1117
+ "step": 615
1118
+ },
1119
+ {
1120
+ "epoch": 1.4155251141552512,
1121
+ "grad_norm": 0.7842912562338646,
1122
+ "learning_rate": 6.52284263959391e-06,
1123
+ "loss": 0.4831,
1124
+ "mean_token_accuracy": 0.8894213199615478,
1125
+ "num_tokens": 40300967.0,
1126
+ "step": 620
1127
+ },
1128
+ {
1129
+ "epoch": 1.4269406392694064,
1130
+ "grad_norm": 0.5672064982607272,
1131
+ "learning_rate": 6.395939086294417e-06,
1132
+ "loss": 0.3456,
1133
+ "mean_token_accuracy": 0.9134271621704102,
1134
+ "num_tokens": 40623392.0,
1135
+ "step": 625
1136
+ },
1137
+ {
1138
+ "epoch": 1.4383561643835616,
1139
+ "grad_norm": 1.3807792287581275,
1140
+ "learning_rate": 6.269035532994925e-06,
1141
+ "loss": 0.3724,
1142
+ "mean_token_accuracy": 0.9124893307685852,
1143
+ "num_tokens": 40951072.0,
1144
+ "step": 630
1145
+ },
1146
+ {
1147
+ "epoch": 1.4497716894977168,
1148
+ "grad_norm": 0.6652210706772995,
1149
+ "learning_rate": 6.142131979695432e-06,
1150
+ "loss": 0.3739,
1151
+ "mean_token_accuracy": 0.9046972155570984,
1152
+ "num_tokens": 41278752.0,
1153
+ "step": 635
1154
+ },
1155
+ {
1156
+ "epoch": 1.461187214611872,
1157
+ "grad_norm": 0.5659692708335803,
1158
+ "learning_rate": 6.01522842639594e-06,
1159
+ "loss": 0.3255,
1160
+ "mean_token_accuracy": 0.9139390826225281,
1161
+ "num_tokens": 41606432.0,
1162
+ "step": 640
1163
+ },
1164
+ {
1165
+ "epoch": 1.4726027397260273,
1166
+ "grad_norm": 4.1799658436454985,
1167
+ "learning_rate": 5.888324873096447e-06,
1168
+ "loss": 0.4124,
1169
+ "mean_token_accuracy": 0.9023898243904114,
1170
+ "num_tokens": 41934112.0,
1171
+ "step": 645
1172
+ },
1173
+ {
1174
+ "epoch": 1.4840182648401825,
1175
+ "grad_norm": 0.5135734072884963,
1176
+ "learning_rate": 5.761421319796955e-06,
1177
+ "loss": 0.2561,
1178
+ "mean_token_accuracy": 0.9317985653877259,
1179
+ "num_tokens": 42257011.0,
1180
+ "step": 650
1181
+ },
1182
+ {
1183
+ "epoch": 1.495433789954338,
1184
+ "grad_norm": 0.5638203849802151,
1185
+ "learning_rate": 5.634517766497463e-06,
1186
+ "loss": 0.2365,
1187
+ "mean_token_accuracy": 0.9371708989143371,
1188
+ "num_tokens": 42577082.0,
1189
+ "step": 655
1190
+ },
1191
+ {
1192
+ "epoch": 1.5068493150684932,
1193
+ "grad_norm": 0.6879142760785716,
1194
+ "learning_rate": 5.5076142131979704e-06,
1195
+ "loss": 0.2935,
1196
+ "mean_token_accuracy": 0.9240141749382019,
1197
+ "num_tokens": 42904762.0,
1198
+ "step": 660
1199
+ },
1200
+ {
1201
+ "epoch": 1.5182648401826484,
1202
+ "grad_norm": 0.7750410637469933,
1203
+ "learning_rate": 5.380710659898477e-06,
1204
+ "loss": 0.3203,
1205
+ "mean_token_accuracy": 0.91428701877594,
1206
+ "num_tokens": 43232442.0,
1207
+ "step": 665
1208
+ },
1209
+ {
1210
+ "epoch": 1.5296803652968036,
1211
+ "grad_norm": 0.6015866918949929,
1212
+ "learning_rate": 5.253807106598985e-06,
1213
+ "loss": 0.4269,
1214
+ "mean_token_accuracy": 0.9015169024467469,
1215
+ "num_tokens": 43560122.0,
1216
+ "step": 670
1217
+ },
1218
+ {
1219
+ "epoch": 1.541095890410959,
1220
+ "grad_norm": 0.7419081802783241,
1221
+ "learning_rate": 5.126903553299493e-06,
1222
+ "loss": 0.3791,
1223
+ "mean_token_accuracy": 0.8882760286331177,
1224
+ "num_tokens": 43883352.0,
1225
+ "step": 675
1226
+ },
1227
+ {
1228
+ "epoch": 1.5525114155251143,
1229
+ "grad_norm": 1.0463314698483188,
1230
+ "learning_rate": 5e-06,
1231
+ "loss": 0.3361,
1232
+ "mean_token_accuracy": 0.9164814949035645,
1233
+ "num_tokens": 44211032.0,
1234
+ "step": 680
1235
+ },
1236
+ {
1237
+ "epoch": 1.5639269406392695,
1238
+ "grad_norm": 1.0052530940065705,
1239
+ "learning_rate": 4.873096446700508e-06,
1240
+ "loss": 0.4119,
1241
+ "mean_token_accuracy": 0.9030063629150391,
1242
+ "num_tokens": 44538712.0,
1243
+ "step": 685
1244
+ },
1245
+ {
1246
+ "epoch": 1.5753424657534247,
1247
+ "grad_norm": 2.634004436653959,
1248
+ "learning_rate": 4.746192893401016e-06,
1249
+ "loss": 0.3595,
1250
+ "mean_token_accuracy": 0.9104088068008422,
1251
+ "num_tokens": 44865459.0,
1252
+ "step": 690
1253
+ },
1254
+ {
1255
+ "epoch": 1.58675799086758,
1256
+ "grad_norm": 0.6108457164515184,
1257
+ "learning_rate": 4.6192893401015235e-06,
1258
+ "loss": 0.2628,
1259
+ "mean_token_accuracy": 0.9270998001098633,
1260
+ "num_tokens": 45192081.0,
1261
+ "step": 695
1262
+ },
1263
+ {
1264
+ "epoch": 1.5981735159817352,
1265
+ "grad_norm": 0.5837697582430958,
1266
+ "learning_rate": 4.492385786802031e-06,
1267
+ "loss": 0.4771,
1268
+ "mean_token_accuracy": 0.8941513538360596,
1269
+ "num_tokens": 45517873.0,
1270
+ "step": 700
1271
+ },
1272
+ {
1273
+ "epoch": 1.6095890410958904,
1274
+ "grad_norm": 1.3327839635586716,
1275
+ "learning_rate": 4.365482233502538e-06,
1276
+ "loss": 0.4418,
1277
+ "mean_token_accuracy": 0.9001519441604614,
1278
+ "num_tokens": 45838879.0,
1279
+ "step": 705
1280
+ },
1281
+ {
1282
+ "epoch": 1.6210045662100456,
1283
+ "grad_norm": 0.8761990524355994,
1284
+ "learning_rate": 4.238578680203046e-06,
1285
+ "loss": 0.2878,
1286
+ "mean_token_accuracy": 0.9253254771232605,
1287
+ "num_tokens": 46154753.0,
1288
+ "step": 710
1289
+ },
1290
+ {
1291
+ "epoch": 1.6324200913242009,
1292
+ "grad_norm": 0.7797165241970883,
1293
+ "learning_rate": 4.1116751269035535e-06,
1294
+ "loss": 0.4321,
1295
+ "mean_token_accuracy": 0.8992583274841308,
1296
+ "num_tokens": 46482433.0,
1297
+ "step": 715
1298
+ },
1299
+ {
1300
+ "epoch": 1.643835616438356,
1301
+ "grad_norm": 0.5034043819066758,
1302
+ "learning_rate": 3.984771573604061e-06,
1303
+ "loss": 0.3578,
1304
+ "mean_token_accuracy": 0.9129929184913635,
1305
+ "num_tokens": 46810113.0,
1306
+ "step": 720
1307
+ },
1308
+ {
1309
+ "epoch": 1.6552511415525113,
1310
+ "grad_norm": 0.7169607234713667,
1311
+ "learning_rate": 3.857868020304569e-06,
1312
+ "loss": 0.4198,
1313
+ "mean_token_accuracy": 0.9019655585289001,
1314
+ "num_tokens": 47137793.0,
1315
+ "step": 725
1316
+ },
1317
+ {
1318
+ "epoch": 1.6666666666666665,
1319
+ "grad_norm": 0.558464393934835,
1320
+ "learning_rate": 3.7309644670050766e-06,
1321
+ "loss": 0.3967,
1322
+ "mean_token_accuracy": 0.9081552982330322,
1323
+ "num_tokens": 47465473.0,
1324
+ "step": 730
1325
+ },
1326
+ {
1327
+ "epoch": 1.678082191780822,
1328
+ "grad_norm": 0.7655920104904193,
1329
+ "learning_rate": 3.6040609137055843e-06,
1330
+ "loss": 0.379,
1331
+ "mean_token_accuracy": 0.9107343554496765,
1332
+ "num_tokens": 47793153.0,
1333
+ "step": 735
1334
+ },
1335
+ {
1336
+ "epoch": 1.6894977168949772,
1337
+ "grad_norm": 0.49184028707678545,
1338
+ "learning_rate": 3.4771573604060916e-06,
1339
+ "loss": 0.343,
1340
+ "mean_token_accuracy": 0.9188774347305297,
1341
+ "num_tokens": 48120833.0,
1342
+ "step": 740
1343
+ },
1344
+ {
1345
+ "epoch": 1.7009132420091324,
1346
+ "grad_norm": 0.5716197168139977,
1347
+ "learning_rate": 3.3502538071065993e-06,
1348
+ "loss": 0.477,
1349
+ "mean_token_accuracy": 0.8837767481803894,
1350
+ "num_tokens": 48437001.0,
1351
+ "step": 745
1352
+ },
1353
+ {
1354
+ "epoch": 1.7123287671232876,
1355
+ "grad_norm": 0.6143187237128267,
1356
+ "learning_rate": 3.223350253807107e-06,
1357
+ "loss": 0.3496,
1358
+ "mean_token_accuracy": 0.9140928864479065,
1359
+ "num_tokens": 48764200.0,
1360
+ "step": 750
1361
+ },
1362
+ {
1363
+ "epoch": 1.723744292237443,
1364
+ "grad_norm": 0.7062097692967121,
1365
+ "learning_rate": 3.0964467005076143e-06,
1366
+ "loss": 0.2871,
1367
+ "mean_token_accuracy": 0.9207997918128967,
1368
+ "num_tokens": 49083727.0,
1369
+ "step": 755
1370
+ },
1371
+ {
1372
+ "epoch": 1.7351598173515983,
1373
+ "grad_norm": 0.5967667747400006,
1374
+ "learning_rate": 2.969543147208122e-06,
1375
+ "loss": 0.3075,
1376
+ "mean_token_accuracy": 0.9261842131614685,
1377
+ "num_tokens": 49411407.0,
1378
+ "step": 760
1379
+ },
1380
+ {
1381
+ "epoch": 1.7465753424657535,
1382
+ "grad_norm": 0.6067471737507767,
1383
+ "learning_rate": 2.8426395939086297e-06,
1384
+ "loss": 0.3323,
1385
+ "mean_token_accuracy": 0.9163744807243347,
1386
+ "num_tokens": 49737669.0,
1387
+ "step": 765
1388
+ },
1389
+ {
1390
+ "epoch": 1.7579908675799087,
1391
+ "grad_norm": 0.5038912700364546,
1392
+ "learning_rate": 2.715736040609137e-06,
1393
+ "loss": 0.3435,
1394
+ "mean_token_accuracy": 0.9163585543632508,
1395
+ "num_tokens": 50058884.0,
1396
+ "step": 770
1397
+ },
1398
+ {
1399
+ "epoch": 1.769406392694064,
1400
+ "grad_norm": 0.5841966451349081,
1401
+ "learning_rate": 2.588832487309645e-06,
1402
+ "loss": 0.4916,
1403
+ "mean_token_accuracy": 0.8925225734710693,
1404
+ "num_tokens": 50381828.0,
1405
+ "step": 775
1406
+ },
1407
+ {
1408
+ "epoch": 1.7808219178082192,
1409
+ "grad_norm": 0.7120652585853746,
1410
+ "learning_rate": 2.4619289340101523e-06,
1411
+ "loss": 0.3648,
1412
+ "mean_token_accuracy": 0.9107160329818725,
1413
+ "num_tokens": 50709508.0,
1414
+ "step": 780
1415
+ },
1416
+ {
1417
+ "epoch": 1.7922374429223744,
1418
+ "grad_norm": 0.8338272887961167,
1419
+ "learning_rate": 2.33502538071066e-06,
1420
+ "loss": 0.5049,
1421
+ "mean_token_accuracy": 0.8752775192260742,
1422
+ "num_tokens": 51034458.0,
1423
+ "step": 785
1424
+ },
1425
+ {
1426
+ "epoch": 1.8036529680365296,
1427
+ "grad_norm": 0.6064117509722058,
1428
+ "learning_rate": 2.2081218274111678e-06,
1429
+ "loss": 0.5298,
1430
+ "mean_token_accuracy": 0.8844402432441711,
1431
+ "num_tokens": 51362138.0,
1432
+ "step": 790
1433
+ },
1434
+ {
1435
+ "epoch": 1.8150684931506849,
1436
+ "grad_norm": 0.5624163026096038,
1437
+ "learning_rate": 2.0812182741116755e-06,
1438
+ "loss": 0.1998,
1439
+ "mean_token_accuracy": 0.9429862022399902,
1440
+ "num_tokens": 51689818.0,
1441
+ "step": 795
1442
+ },
1443
+ {
1444
+ "epoch": 1.82648401826484,
1445
+ "grad_norm": 0.6489646064723047,
1446
+ "learning_rate": 1.9543147208121827e-06,
1447
+ "loss": 0.2074,
1448
+ "mean_token_accuracy": 0.9404376745223999,
1449
+ "num_tokens": 52017498.0,
1450
+ "step": 800
1451
+ },
1452
+ {
1453
+ "epoch": 1.8378995433789953,
1454
+ "grad_norm": 0.7078634387879972,
1455
+ "learning_rate": 1.8274111675126904e-06,
1456
+ "loss": 0.2248,
1457
+ "mean_token_accuracy": 0.9358620643615723,
1458
+ "num_tokens": 52344815.0,
1459
+ "step": 805
1460
+ },
1461
+ {
1462
+ "epoch": 1.8493150684931505,
1463
+ "grad_norm": 0.850606532541916,
1464
+ "learning_rate": 1.7005076142131981e-06,
1465
+ "loss": 0.271,
1466
+ "mean_token_accuracy": 0.9241331934928894,
1467
+ "num_tokens": 52672495.0,
1468
+ "step": 810
1469
+ },
1470
+ {
1471
+ "epoch": 1.860730593607306,
1472
+ "grad_norm": 0.5656500358616859,
1473
+ "learning_rate": 1.5736040609137056e-06,
1474
+ "loss": 0.4001,
1475
+ "mean_token_accuracy": 0.9008942723274231,
1476
+ "num_tokens": 53000175.0,
1477
+ "step": 815
1478
+ },
1479
+ {
1480
+ "epoch": 1.8721461187214612,
1481
+ "grad_norm": 0.5188980614065027,
1482
+ "learning_rate": 1.4467005076142131e-06,
1483
+ "loss": 0.2976,
1484
+ "mean_token_accuracy": 0.9234037399291992,
1485
+ "num_tokens": 53327855.0,
1486
+ "step": 820
1487
+ },
1488
+ {
1489
+ "epoch": 1.8835616438356164,
1490
+ "grad_norm": 0.6522132469808263,
1491
+ "learning_rate": 1.319796954314721e-06,
1492
+ "loss": 0.4339,
1493
+ "mean_token_accuracy": 0.8968196868896484,
1494
+ "num_tokens": 53655535.0,
1495
+ "step": 825
1496
+ },
1497
+ {
1498
+ "epoch": 1.8949771689497716,
1499
+ "grad_norm": 0.6480606818026367,
1500
+ "learning_rate": 1.1928934010152285e-06,
1501
+ "loss": 0.3438,
1502
+ "mean_token_accuracy": 0.9182151198387146,
1503
+ "num_tokens": 53983215.0,
1504
+ "step": 830
1505
+ },
1506
+ {
1507
+ "epoch": 1.906392694063927,
1508
+ "grad_norm": 0.6426743884232268,
1509
+ "learning_rate": 1.0659898477157362e-06,
1510
+ "loss": 0.3568,
1511
+ "mean_token_accuracy": 0.9122115731239319,
1512
+ "num_tokens": 54310895.0,
1513
+ "step": 835
1514
+ },
1515
+ {
1516
+ "epoch": 1.9178082191780823,
1517
+ "grad_norm": 0.6461826817344741,
1518
+ "learning_rate": 9.390862944162437e-07,
1519
+ "loss": 0.3381,
1520
+ "mean_token_accuracy": 0.9099621653556824,
1521
+ "num_tokens": 54638575.0,
1522
+ "step": 840
1523
+ },
1524
+ {
1525
+ "epoch": 1.9292237442922375,
1526
+ "grad_norm": 0.8935473667193835,
1527
+ "learning_rate": 8.121827411167513e-07,
1528
+ "loss": 0.4386,
1529
+ "mean_token_accuracy": 0.9003937244415283,
1530
+ "num_tokens": 54966255.0,
1531
+ "step": 845
1532
+ },
1533
+ {
1534
+ "epoch": 1.9406392694063928,
1535
+ "grad_norm": 0.6349104249304525,
1536
+ "learning_rate": 6.852791878172589e-07,
1537
+ "loss": 0.1985,
1538
+ "mean_token_accuracy": 0.9433280467987061,
1539
+ "num_tokens": 55293935.0,
1540
+ "step": 850
1541
+ },
1542
+ {
1543
+ "epoch": 1.952054794520548,
1544
+ "grad_norm": 0.7223041297457694,
1545
+ "learning_rate": 5.583756345177665e-07,
1546
+ "loss": 0.279,
1547
+ "mean_token_accuracy": 0.9235318779945374,
1548
+ "num_tokens": 55610720.0,
1549
+ "step": 855
1550
+ },
1551
+ {
1552
+ "epoch": 1.9634703196347032,
1553
+ "grad_norm": 0.8028128196207174,
1554
+ "learning_rate": 4.314720812182741e-07,
1555
+ "loss": 0.568,
1556
+ "mean_token_accuracy": 0.8745380878448487,
1557
+ "num_tokens": 55932823.0,
1558
+ "step": 860
1559
+ },
1560
+ {
1561
+ "epoch": 1.9748858447488584,
1562
+ "grad_norm": 0.6149079073694556,
1563
+ "learning_rate": 3.0456852791878176e-07,
1564
+ "loss": 0.3471,
1565
+ "mean_token_accuracy": 0.9153918981552124,
1566
+ "num_tokens": 56260503.0,
1567
+ "step": 865
1568
+ },
1569
+ {
1570
+ "epoch": 1.9863013698630136,
1571
+ "grad_norm": 0.6505083827044454,
1572
+ "learning_rate": 1.7766497461928935e-07,
1573
+ "loss": 0.3967,
1574
+ "mean_token_accuracy": 0.9023218989372254,
1575
+ "num_tokens": 56579764.0,
1576
+ "step": 870
1577
+ },
1578
+ {
1579
+ "epoch": 1.9977168949771689,
1580
+ "grad_norm": 0.8472060619906485,
1581
+ "learning_rate": 5.076142131979696e-08,
1582
+ "loss": 0.3635,
1583
+ "mean_token_accuracy": 0.9119231700897217,
1584
+ "num_tokens": 56903522.0,
1585
+ "step": 875
1586
+ },
1587
+ {
1588
+ "epoch": 2.0,
1589
+ "mean_token_accuracy": 0.8880478739738464,
1590
+ "num_tokens": 56919906.0,
1591
+ "step": 876,
1592
+ "total_flos": 114029121830912.0,
1593
+ "train_loss": 0.5227763477253587,
1594
+ "train_runtime": 1326.8074,
1595
+ "train_samples_per_second": 5.273,
1596
+ "train_steps_per_second": 0.66
1597
+ }
1598
+ ],
1599
+ "logging_steps": 5,
1600
+ "max_steps": 876,
1601
+ "num_input_tokens_seen": 0,
1602
+ "num_train_epochs": 2,
1603
+ "save_steps": 500,
1604
+ "stateful_callbacks": {
1605
+ "TrainerControl": {
1606
+ "args": {
1607
+ "should_epoch_stop": false,
1608
+ "should_evaluate": false,
1609
+ "should_log": false,
1610
+ "should_save": true,
1611
+ "should_training_stop": true
1612
+ },
1613
+ "attributes": {}
1614
+ }
1615
+ },
1616
+ "total_flos": 114029121830912.0,
1617
+ "train_batch_size": 1,
1618
+ "trial_name": null,
1619
+ "trial_params": null
1620
+ }