w-ahmad commited on
Commit
a4e3fcc
·
verified ·
1 Parent(s): d782871

Auto upload 2026-08-06T20:50:17.909881

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. out/glu-linear-9L_run/README.md +65 -0
  2. out/glu-linear-9L_run/checkpoint-1500/config.json +35 -0
  3. out/glu-linear-9L_run/checkpoint-1500/model.safetensors +3 -0
  4. out/glu-linear-9L_run/checkpoint-1500/optimizer.pt +3 -0
  5. out/glu-linear-9L_run/checkpoint-1500/rng_state.pth +3 -0
  6. out/glu-linear-9L_run/checkpoint-1500/scheduler.pt +3 -0
  7. out/glu-linear-9L_run/checkpoint-1500/tokenizer.json +0 -0
  8. out/glu-linear-9L_run/checkpoint-1500/tokenizer_config.json +13 -0
  9. out/glu-linear-9L_run/checkpoint-1500/trainer_state.json +615 -0
  10. out/glu-linear-9L_run/checkpoint-1500/training_args.bin +3 -0
  11. out/glu-linear-9L_run/checkpoint-2000/config.json +35 -0
  12. out/glu-linear-9L_run/checkpoint-2000/model.safetensors +3 -0
  13. out/glu-linear-9L_run/checkpoint-2000/optimizer.pt +3 -0
  14. out/glu-linear-9L_run/checkpoint-2000/rng_state.pth +3 -0
  15. out/glu-linear-9L_run/checkpoint-2000/scheduler.pt +3 -0
  16. out/glu-linear-9L_run/checkpoint-2000/tokenizer.json +0 -0
  17. out/glu-linear-9L_run/checkpoint-2000/tokenizer_config.json +13 -0
  18. out/glu-linear-9L_run/checkpoint-2000/trainer_state.json +814 -0
  19. out/glu-linear-9L_run/checkpoint-2000/training_args.bin +3 -0
  20. out/glu-linear-9L_run/model.safetensors +1 -1
  21. out/glu-linear-9L_run/tokenizer.json +0 -0
  22. out/glu-linear-9L_run/tokenizer_config.json +13 -0
  23. out/glu-linear-9L_run/training_log.jsonl +0 -0
  24. out/glu-situglu-9L_run/checkpoint-1000/config.json +35 -0
  25. out/glu-situglu-9L_run/checkpoint-1000/model.safetensors +3 -0
  26. out/glu-situglu-9L_run/checkpoint-1000/optimizer.pt +3 -0
  27. out/glu-situglu-9L_run/checkpoint-1000/rng_state.pth +3 -0
  28. out/glu-situglu-9L_run/checkpoint-1000/scheduler.pt +3 -0
  29. out/glu-situglu-9L_run/checkpoint-1000/tokenizer.json +0 -0
  30. out/glu-situglu-9L_run/checkpoint-1000/tokenizer_config.json +13 -0
  31. out/glu-situglu-9L_run/checkpoint-1000/trainer_state.json +424 -0
  32. out/glu-situglu-9L_run/checkpoint-1000/training_args.bin +3 -0
  33. out/glu-situglu-9L_run/checkpoint-1500/config.json +35 -0
  34. out/glu-situglu-9L_run/checkpoint-1500/model.safetensors +3 -0
  35. out/glu-situglu-9L_run/checkpoint-1500/optimizer.pt +3 -0
  36. out/glu-situglu-9L_run/checkpoint-1500/rng_state.pth +3 -0
  37. out/glu-situglu-9L_run/checkpoint-1500/scheduler.pt +3 -0
  38. out/glu-situglu-9L_run/checkpoint-1500/tokenizer.json +0 -0
  39. out/glu-situglu-9L_run/checkpoint-1500/tokenizer_config.json +13 -0
  40. out/glu-situglu-9L_run/checkpoint-1500/trainer_state.json +615 -0
  41. out/glu-situglu-9L_run/checkpoint-1500/training_args.bin +3 -0
  42. out/glu-situglu-9L_run/checkpoint-500/config.json +35 -0
  43. out/glu-situglu-9L_run/checkpoint-500/model.safetensors +3 -0
  44. out/glu-situglu-9L_run/checkpoint-500/optimizer.pt +3 -0
  45. out/glu-situglu-9L_run/checkpoint-500/rng_state.pth +3 -0
  46. out/glu-situglu-9L_run/checkpoint-500/scheduler.pt +3 -0
  47. out/glu-situglu-9L_run/checkpoint-500/tokenizer.json +0 -0
  48. out/glu-situglu-9L_run/checkpoint-500/tokenizer_config.json +13 -0
  49. out/glu-situglu-9L_run/checkpoint-500/trainer_state.json +225 -0
  50. out/glu-situglu-9L_run/checkpoint-500/training_args.bin +3 -0
out/glu-linear-9L_run/README.md ADDED
@@ -0,0 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers
3
+ tags:
4
+ - generated_from_trainer
5
+ model-index:
6
+ - name: ACTio-glu-linear-9L
7
+ results: []
8
+ ---
9
+
10
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
11
+ should probably proofread and complete it, then remove this comment. -->
12
+
13
+ # ACTio-glu-linear-9L
14
+
15
+ This model is a fine-tuned version of [](https://huggingface.co/) on an unknown dataset.
16
+ It achieves the following results on the evaluation set:
17
+ - Loss: 0.2028
18
+
19
+ ## Model description
20
+
21
+ More information needed
22
+
23
+ ## Intended uses & limitations
24
+
25
+ More information needed
26
+
27
+ ## Training and evaluation data
28
+
29
+ More information needed
30
+
31
+ ## Training procedure
32
+
33
+ ### Training hyperparameters
34
+
35
+ The following hyperparameters were used during training:
36
+ - learning_rate: 0.0003
37
+ - train_batch_size: 64
38
+ - eval_batch_size: 64
39
+ - seed: 42
40
+ - optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
41
+ - lr_scheduler_type: constant
42
+ - training_steps: 2000
43
+
44
+ ### Training results
45
+
46
+ | Training Loss | Epoch | Step | Validation Loss |
47
+ |:-------------:|:------:|:----:|:---------------:|
48
+ | 3.2290 | 0.0135 | 200 | 3.0630 |
49
+ | 0.9419 | 0.0270 | 400 | 0.9078 |
50
+ | 0.4119 | 0.0404 | 600 | 0.4039 |
51
+ | 0.3213 | 0.0539 | 800 | 0.3212 |
52
+ | 0.2753 | 0.0674 | 1000 | 0.2748 |
53
+ | 0.2577 | 0.0809 | 1200 | 0.2531 |
54
+ | 0.2275 | 0.0944 | 1400 | 0.2294 |
55
+ | 0.2254 | 0.1079 | 1600 | 0.2337 |
56
+ | 0.2096 | 0.1213 | 1800 | 0.2087 |
57
+ | 0.2011 | 0.1348 | 2000 | 0.2028 |
58
+
59
+
60
+ ### Framework versions
61
+
62
+ - Transformers 5.15.0.dev0
63
+ - Pytorch 2.6.0+cu124
64
+ - Datasets 5.0.1
65
+ - Tokenizers 0.22.2
out/glu-linear-9L_run/checkpoint-1500/config.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation": "linear",
3
+ "architectures": [
4
+ "TinyLlamaForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "bos_token_id": 1,
9
+ "dtype": "bfloat16",
10
+ "eos_token_id": 2,
11
+ "head_dim": 32,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 128,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 256,
16
+ "max_position_embeddings": 512,
17
+ "mlp_bias": false,
18
+ "mlp_type": "glu",
19
+ "model_type": "tiny_llama",
20
+ "num_attention_heads": 4,
21
+ "num_hidden_layers": 9,
22
+ "num_key_value_heads": 4,
23
+ "pad_token_id": 0,
24
+ "pretraining_tp": 1,
25
+ "rms_norm_eps": 1e-06,
26
+ "rope_parameters": {
27
+ "rope_theta": 10000.0,
28
+ "rope_type": "default"
29
+ },
30
+ "tie_word_embeddings": true,
31
+ "tokenizer_name": "w-ahmad/tiny-stories-tokenizer",
32
+ "transformers_version": "5.15.0.dev0",
33
+ "use_cache": false,
34
+ "vocab_size": 4096
35
+ }
out/glu-linear-9L_run/checkpoint-1500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92d9ce7258239bf24eb6172496b5928c3ab96490ba932977e4a0cac5a06a0cf6
3
+ size 4011496
out/glu-linear-9L_run/checkpoint-1500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3caaeb2adc29f6465e3fc74df8a80628aa0cb70252170899cc321a561f3d5b06
3
+ size 8074746
out/glu-linear-9L_run/checkpoint-1500/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2768285b45b2a0c05f6f50bbb8c0287fca6f62a8cde6d1b1f02151ac72ee8dc
3
+ size 14244
out/glu-linear-9L_run/checkpoint-1500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b26a759cc207e28dea06eae7ad531c2dd78a09978f6972edbfc1566adccd7142
3
+ size 1064
out/glu-linear-9L_run/checkpoint-1500/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
out/glu-linear-9L_run/checkpoint-1500/tokenizer_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "backend": "tokenizers",
4
+ "bos_token": "<|endoftext|>",
5
+ "eos_token": "<|endoftext|>",
6
+ "errors": "replace",
7
+ "is_local": false,
8
+ "local_files_only": false,
9
+ "model_max_length": 1024,
10
+ "pad_token": "<|endoftext|>",
11
+ "tokenizer_class": "GPT2Tokenizer",
12
+ "unk_token": "<|endoftext|>"
13
+ }
out/glu-linear-9L_run/checkpoint-1500/trainer_state.json ADDED
@@ -0,0 +1,615 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.10111223458038422,
6
+ "eval_steps": 200,
7
+ "global_step": 1500,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.0013481631277384564,
14
+ "grad_norm": 1.25,
15
+ "learning_rate": 0.0003,
16
+ "loss": 7.778378295898437,
17
+ "step": 20
18
+ },
19
+ {
20
+ "epoch": 0.002696326255476913,
21
+ "grad_norm": 1.1875,
22
+ "learning_rate": 0.0003,
23
+ "loss": 6.979072570800781,
24
+ "step": 40
25
+ },
26
+ {
27
+ "epoch": 0.004044489383215369,
28
+ "grad_norm": 1.09375,
29
+ "learning_rate": 0.0003,
30
+ "loss": 6.366487884521485,
31
+ "step": 60
32
+ },
33
+ {
34
+ "epoch": 0.005392652510953826,
35
+ "grad_norm": 2.328125,
36
+ "learning_rate": 0.0003,
37
+ "loss": 5.798646926879883,
38
+ "step": 80
39
+ },
40
+ {
41
+ "epoch": 0.006740815638692282,
42
+ "grad_norm": 4.78125,
43
+ "learning_rate": 0.0003,
44
+ "loss": 5.200259017944336,
45
+ "step": 100
46
+ },
47
+ {
48
+ "epoch": 0.008088978766430738,
49
+ "grad_norm": 3.421875,
50
+ "learning_rate": 0.0003,
51
+ "loss": 4.721379089355469,
52
+ "step": 120
53
+ },
54
+ {
55
+ "epoch": 0.009437141894169195,
56
+ "grad_norm": 3.171875,
57
+ "learning_rate": 0.0003,
58
+ "loss": 4.299360275268555,
59
+ "step": 140
60
+ },
61
+ {
62
+ "epoch": 0.010785305021907651,
63
+ "grad_norm": 2.328125,
64
+ "learning_rate": 0.0003,
65
+ "loss": 3.917990493774414,
66
+ "step": 160
67
+ },
68
+ {
69
+ "epoch": 0.012133468149646108,
70
+ "grad_norm": 2.359375,
71
+ "learning_rate": 0.0003,
72
+ "loss": 3.5716182708740236,
73
+ "step": 180
74
+ },
75
+ {
76
+ "epoch": 0.013481631277384564,
77
+ "grad_norm": 2.890625,
78
+ "learning_rate": 0.0003,
79
+ "loss": 3.2290443420410155,
80
+ "step": 200
81
+ },
82
+ {
83
+ "epoch": 0.013481631277384564,
84
+ "eval_loss": 3.0630435943603516,
85
+ "eval_runtime": 8.4031,
86
+ "eval_samples_per_second": 1133.751,
87
+ "eval_steps_per_second": 17.732,
88
+ "step": 200
89
+ },
90
+ {
91
+ "epoch": 0.01482979440512302,
92
+ "grad_norm": 2.84375,
93
+ "learning_rate": 0.0003,
94
+ "loss": 2.9105119705200195,
95
+ "step": 220
96
+ },
97
+ {
98
+ "epoch": 0.016177957532861477,
99
+ "grad_norm": 2.21875,
100
+ "learning_rate": 0.0003,
101
+ "loss": 2.5742773056030273,
102
+ "step": 240
103
+ },
104
+ {
105
+ "epoch": 0.01752612066059993,
106
+ "grad_norm": 2.875,
107
+ "learning_rate": 0.0003,
108
+ "loss": 2.2510210037231446,
109
+ "step": 260
110
+ },
111
+ {
112
+ "epoch": 0.01887428378833839,
113
+ "grad_norm": 1.8359375,
114
+ "learning_rate": 0.0003,
115
+ "loss": 1.948414421081543,
116
+ "step": 280
117
+ },
118
+ {
119
+ "epoch": 0.020222446916076844,
120
+ "grad_norm": 2.828125,
121
+ "learning_rate": 0.0003,
122
+ "loss": 1.6830488204956056,
123
+ "step": 300
124
+ },
125
+ {
126
+ "epoch": 0.021570610043815303,
127
+ "grad_norm": 2.03125,
128
+ "learning_rate": 0.0003,
129
+ "loss": 1.488680934906006,
130
+ "step": 320
131
+ },
132
+ {
133
+ "epoch": 0.022918773171553757,
134
+ "grad_norm": 1.796875,
135
+ "learning_rate": 0.0003,
136
+ "loss": 1.3138065338134766,
137
+ "step": 340
138
+ },
139
+ {
140
+ "epoch": 0.024266936299292215,
141
+ "grad_norm": 2.515625,
142
+ "learning_rate": 0.0003,
143
+ "loss": 1.1533904075622559,
144
+ "step": 360
145
+ },
146
+ {
147
+ "epoch": 0.02561509942703067,
148
+ "grad_norm": 1.4921875,
149
+ "learning_rate": 0.0003,
150
+ "loss": 1.040479850769043,
151
+ "step": 380
152
+ },
153
+ {
154
+ "epoch": 0.026963262554769128,
155
+ "grad_norm": 1.953125,
156
+ "learning_rate": 0.0003,
157
+ "loss": 0.9418865203857422,
158
+ "step": 400
159
+ },
160
+ {
161
+ "epoch": 0.026963262554769128,
162
+ "eval_loss": 0.9077715873718262,
163
+ "eval_runtime": 8.327,
164
+ "eval_samples_per_second": 1144.104,
165
+ "eval_steps_per_second": 17.894,
166
+ "step": 400
167
+ },
168
+ {
169
+ "epoch": 0.028311425682507583,
170
+ "grad_norm": 1.0625,
171
+ "learning_rate": 0.0003,
172
+ "loss": 0.866397476196289,
173
+ "step": 420
174
+ },
175
+ {
176
+ "epoch": 0.02965958881024604,
177
+ "grad_norm": 1.6015625,
178
+ "learning_rate": 0.0003,
179
+ "loss": 0.8063242912292481,
180
+ "step": 440
181
+ },
182
+ {
183
+ "epoch": 0.031007751937984496,
184
+ "grad_norm": 1.171875,
185
+ "learning_rate": 0.0003,
186
+ "loss": 0.7396225929260254,
187
+ "step": 460
188
+ },
189
+ {
190
+ "epoch": 0.032355915065722954,
191
+ "grad_norm": 0.8125,
192
+ "learning_rate": 0.0003,
193
+ "loss": 0.679314136505127,
194
+ "step": 480
195
+ },
196
+ {
197
+ "epoch": 0.03370407819346141,
198
+ "grad_norm": 0.453125,
199
+ "learning_rate": 0.0003,
200
+ "loss": 0.6011962890625,
201
+ "step": 500
202
+ },
203
+ {
204
+ "epoch": 0.03505224132119986,
205
+ "grad_norm": 0.423828125,
206
+ "learning_rate": 0.0003,
207
+ "loss": 0.5412269592285156,
208
+ "step": 520
209
+ },
210
+ {
211
+ "epoch": 0.03640040444893832,
212
+ "grad_norm": 0.447265625,
213
+ "learning_rate": 0.0003,
214
+ "loss": 0.49077553749084474,
215
+ "step": 540
216
+ },
217
+ {
218
+ "epoch": 0.03774856757667678,
219
+ "grad_norm": 0.396484375,
220
+ "learning_rate": 0.0003,
221
+ "loss": 0.45958123207092283,
222
+ "step": 560
223
+ },
224
+ {
225
+ "epoch": 0.03909673070441524,
226
+ "grad_norm": 0.390625,
227
+ "learning_rate": 0.0003,
228
+ "loss": 0.43019986152648926,
229
+ "step": 580
230
+ },
231
+ {
232
+ "epoch": 0.04044489383215369,
233
+ "grad_norm": 0.373046875,
234
+ "learning_rate": 0.0003,
235
+ "loss": 0.4119072914123535,
236
+ "step": 600
237
+ },
238
+ {
239
+ "epoch": 0.04044489383215369,
240
+ "eval_loss": 0.4039357006549835,
241
+ "eval_runtime": 8.4391,
242
+ "eval_samples_per_second": 1128.917,
243
+ "eval_steps_per_second": 17.656,
244
+ "step": 600
245
+ },
246
+ {
247
+ "epoch": 0.04179305695989215,
248
+ "grad_norm": 0.34375,
249
+ "learning_rate": 0.0003,
250
+ "loss": 0.39308140277862547,
251
+ "step": 620
252
+ },
253
+ {
254
+ "epoch": 0.043141220087630605,
255
+ "grad_norm": 0.365234375,
256
+ "learning_rate": 0.0003,
257
+ "loss": 0.3809913158416748,
258
+ "step": 640
259
+ },
260
+ {
261
+ "epoch": 0.044489383215369056,
262
+ "grad_norm": 0.365234375,
263
+ "learning_rate": 0.0003,
264
+ "loss": 0.3649900913238525,
265
+ "step": 660
266
+ },
267
+ {
268
+ "epoch": 0.045837546343107514,
269
+ "grad_norm": 0.44140625,
270
+ "learning_rate": 0.0003,
271
+ "loss": 0.3559495210647583,
272
+ "step": 680
273
+ },
274
+ {
275
+ "epoch": 0.04718570947084597,
276
+ "grad_norm": 0.341796875,
277
+ "learning_rate": 0.0003,
278
+ "loss": 0.3457839012145996,
279
+ "step": 700
280
+ },
281
+ {
282
+ "epoch": 0.04853387259858443,
283
+ "grad_norm": 1.5546875,
284
+ "learning_rate": 0.0003,
285
+ "loss": 0.3411240577697754,
286
+ "step": 720
287
+ },
288
+ {
289
+ "epoch": 0.04988203572632288,
290
+ "grad_norm": 1.0,
291
+ "learning_rate": 0.0003,
292
+ "loss": 0.3416250228881836,
293
+ "step": 740
294
+ },
295
+ {
296
+ "epoch": 0.05123019885406134,
297
+ "grad_norm": 1.2890625,
298
+ "learning_rate": 0.0003,
299
+ "loss": 0.3455547094345093,
300
+ "step": 760
301
+ },
302
+ {
303
+ "epoch": 0.0525783619817998,
304
+ "grad_norm": 0.447265625,
305
+ "learning_rate": 0.0003,
306
+ "loss": 0.3340980291366577,
307
+ "step": 780
308
+ },
309
+ {
310
+ "epoch": 0.053926525109538256,
311
+ "grad_norm": 0.3125,
312
+ "learning_rate": 0.0003,
313
+ "loss": 0.3213367223739624,
314
+ "step": 800
315
+ },
316
+ {
317
+ "epoch": 0.053926525109538256,
318
+ "eval_loss": 0.32119977474212646,
319
+ "eval_runtime": 8.4,
320
+ "eval_samples_per_second": 1134.168,
321
+ "eval_steps_per_second": 17.738,
322
+ "step": 800
323
+ },
324
+ {
325
+ "epoch": 0.05527468823727671,
326
+ "grad_norm": 0.271484375,
327
+ "learning_rate": 0.0003,
328
+ "loss": 0.31615324020385743,
329
+ "step": 820
330
+ },
331
+ {
332
+ "epoch": 0.056622851365015166,
333
+ "grad_norm": 0.267578125,
334
+ "learning_rate": 0.0003,
335
+ "loss": 0.3102609634399414,
336
+ "step": 840
337
+ },
338
+ {
339
+ "epoch": 0.057971014492753624,
340
+ "grad_norm": 0.26171875,
341
+ "learning_rate": 0.0003,
342
+ "loss": 0.30272841453552246,
343
+ "step": 860
344
+ },
345
+ {
346
+ "epoch": 0.05931917762049208,
347
+ "grad_norm": 0.265625,
348
+ "learning_rate": 0.0003,
349
+ "loss": 0.2957427501678467,
350
+ "step": 880
351
+ },
352
+ {
353
+ "epoch": 0.06066734074823053,
354
+ "grad_norm": 0.265625,
355
+ "learning_rate": 0.0003,
356
+ "loss": 0.29436583518981935,
357
+ "step": 900
358
+ },
359
+ {
360
+ "epoch": 0.06201550387596899,
361
+ "grad_norm": 0.267578125,
362
+ "learning_rate": 0.0003,
363
+ "loss": 0.2905875205993652,
364
+ "step": 920
365
+ },
366
+ {
367
+ "epoch": 0.06336366700370745,
368
+ "grad_norm": 0.275390625,
369
+ "learning_rate": 0.0003,
370
+ "loss": 0.28632206916809083,
371
+ "step": 940
372
+ },
373
+ {
374
+ "epoch": 0.06471183013144591,
375
+ "grad_norm": 0.255859375,
376
+ "learning_rate": 0.0003,
377
+ "loss": 0.28044116497039795,
378
+ "step": 960
379
+ },
380
+ {
381
+ "epoch": 0.06605999325918437,
382
+ "grad_norm": 0.2890625,
383
+ "learning_rate": 0.0003,
384
+ "loss": 0.278120231628418,
385
+ "step": 980
386
+ },
387
+ {
388
+ "epoch": 0.06740815638692282,
389
+ "grad_norm": 0.28125,
390
+ "learning_rate": 0.0003,
391
+ "loss": 0.2752914667129517,
392
+ "step": 1000
393
+ },
394
+ {
395
+ "epoch": 0.06740815638692282,
396
+ "eval_loss": 0.2748347520828247,
397
+ "eval_runtime": 8.4469,
398
+ "eval_samples_per_second": 1127.864,
399
+ "eval_steps_per_second": 17.64,
400
+ "step": 1000
401
+ },
402
+ {
403
+ "epoch": 0.06875631951466127,
404
+ "grad_norm": 0.267578125,
405
+ "learning_rate": 0.0003,
406
+ "loss": 0.2703357934951782,
407
+ "step": 1020
408
+ },
409
+ {
410
+ "epoch": 0.07010448264239973,
411
+ "grad_norm": 0.2578125,
412
+ "learning_rate": 0.0003,
413
+ "loss": 0.2655144691467285,
414
+ "step": 1040
415
+ },
416
+ {
417
+ "epoch": 0.07145264577013818,
418
+ "grad_norm": 0.326171875,
419
+ "learning_rate": 0.0003,
420
+ "loss": 0.26470656394958497,
421
+ "step": 1060
422
+ },
423
+ {
424
+ "epoch": 0.07280080889787664,
425
+ "grad_norm": 0.337890625,
426
+ "learning_rate": 0.0003,
427
+ "loss": 0.26000075340270995,
428
+ "step": 1080
429
+ },
430
+ {
431
+ "epoch": 0.0741489720256151,
432
+ "grad_norm": 0.294921875,
433
+ "learning_rate": 0.0003,
434
+ "loss": 0.2588791847229004,
435
+ "step": 1100
436
+ },
437
+ {
438
+ "epoch": 0.07549713515335356,
439
+ "grad_norm": 0.3125,
440
+ "learning_rate": 0.0003,
441
+ "loss": 0.2555199146270752,
442
+ "step": 1120
443
+ },
444
+ {
445
+ "epoch": 0.07684529828109202,
446
+ "grad_norm": 0.326171875,
447
+ "learning_rate": 0.0003,
448
+ "loss": 0.25160832405090333,
449
+ "step": 1140
450
+ },
451
+ {
452
+ "epoch": 0.07819346140883048,
453
+ "grad_norm": 0.287109375,
454
+ "learning_rate": 0.0003,
455
+ "loss": 0.2508568048477173,
456
+ "step": 1160
457
+ },
458
+ {
459
+ "epoch": 0.07954162453656892,
460
+ "grad_norm": 0.8359375,
461
+ "learning_rate": 0.0003,
462
+ "loss": 0.26505236625671386,
463
+ "step": 1180
464
+ },
465
+ {
466
+ "epoch": 0.08088978766430738,
467
+ "grad_norm": 0.63671875,
468
+ "learning_rate": 0.0003,
469
+ "loss": 0.2577481269836426,
470
+ "step": 1200
471
+ },
472
+ {
473
+ "epoch": 0.08088978766430738,
474
+ "eval_loss": 0.25307410955429077,
475
+ "eval_runtime": 8.4059,
476
+ "eval_samples_per_second": 1133.37,
477
+ "eval_steps_per_second": 17.726,
478
+ "step": 1200
479
+ },
480
+ {
481
+ "epoch": 0.08223795079204584,
482
+ "grad_norm": 0.2890625,
483
+ "learning_rate": 0.0003,
484
+ "loss": 0.248107647895813,
485
+ "step": 1220
486
+ },
487
+ {
488
+ "epoch": 0.0835861139197843,
489
+ "grad_norm": 0.244140625,
490
+ "learning_rate": 0.0003,
491
+ "loss": 0.24457168579101562,
492
+ "step": 1240
493
+ },
494
+ {
495
+ "epoch": 0.08493427704752275,
496
+ "grad_norm": 0.23828125,
497
+ "learning_rate": 0.0003,
498
+ "loss": 0.24279680252075195,
499
+ "step": 1260
500
+ },
501
+ {
502
+ "epoch": 0.08628244017526121,
503
+ "grad_norm": 0.2314453125,
504
+ "learning_rate": 0.0003,
505
+ "loss": 0.23819704055786134,
506
+ "step": 1280
507
+ },
508
+ {
509
+ "epoch": 0.08763060330299967,
510
+ "grad_norm": 0.2353515625,
511
+ "learning_rate": 0.0003,
512
+ "loss": 0.23753654956817627,
513
+ "step": 1300
514
+ },
515
+ {
516
+ "epoch": 0.08897876643073811,
517
+ "grad_norm": 0.2275390625,
518
+ "learning_rate": 0.0003,
519
+ "loss": 0.23422625064849853,
520
+ "step": 1320
521
+ },
522
+ {
523
+ "epoch": 0.09032692955847657,
524
+ "grad_norm": 0.2255859375,
525
+ "learning_rate": 0.0003,
526
+ "loss": 0.23316617012023927,
527
+ "step": 1340
528
+ },
529
+ {
530
+ "epoch": 0.09167509268621503,
531
+ "grad_norm": 0.22265625,
532
+ "learning_rate": 0.0003,
533
+ "loss": 0.23011410236358643,
534
+ "step": 1360
535
+ },
536
+ {
537
+ "epoch": 0.09302325581395349,
538
+ "grad_norm": 0.2177734375,
539
+ "learning_rate": 0.0003,
540
+ "loss": 0.22971732616424562,
541
+ "step": 1380
542
+ },
543
+ {
544
+ "epoch": 0.09437141894169195,
545
+ "grad_norm": 0.2236328125,
546
+ "learning_rate": 0.0003,
547
+ "loss": 0.22751867771148682,
548
+ "step": 1400
549
+ },
550
+ {
551
+ "epoch": 0.09437141894169195,
552
+ "eval_loss": 0.22943033277988434,
553
+ "eval_runtime": 8.4608,
554
+ "eval_samples_per_second": 1126.015,
555
+ "eval_steps_per_second": 17.611,
556
+ "step": 1400
557
+ },
558
+ {
559
+ "epoch": 0.0957195820694304,
560
+ "grad_norm": 0.2294921875,
561
+ "learning_rate": 0.0003,
562
+ "loss": 0.22718305587768556,
563
+ "step": 1420
564
+ },
565
+ {
566
+ "epoch": 0.09706774519716886,
567
+ "grad_norm": 0.2138671875,
568
+ "learning_rate": 0.0003,
569
+ "loss": 0.22553596496582032,
570
+ "step": 1440
571
+ },
572
+ {
573
+ "epoch": 0.09841590832490732,
574
+ "grad_norm": 0.2373046875,
575
+ "learning_rate": 0.0003,
576
+ "loss": 0.22326209545135497,
577
+ "step": 1460
578
+ },
579
+ {
580
+ "epoch": 0.09976407145264576,
581
+ "grad_norm": 0.212890625,
582
+ "learning_rate": 0.0003,
583
+ "loss": 0.2213585376739502,
584
+ "step": 1480
585
+ },
586
+ {
587
+ "epoch": 0.10111223458038422,
588
+ "grad_norm": 0.21875,
589
+ "learning_rate": 0.0003,
590
+ "loss": 0.22035691738128663,
591
+ "step": 1500
592
+ }
593
+ ],
594
+ "logging_steps": 20,
595
+ "max_steps": 2000,
596
+ "num_input_tokens_seen": 0,
597
+ "num_train_epochs": 1,
598
+ "save_steps": 500,
599
+ "stateful_callbacks": {
600
+ "TrainerControl": {
601
+ "args": {
602
+ "should_epoch_stop": false,
603
+ "should_evaluate": false,
604
+ "should_log": false,
605
+ "should_save": true,
606
+ "should_training_stop": false
607
+ },
608
+ "attributes": {}
609
+ }
610
+ },
611
+ "total_flos": 435582664704000.0,
612
+ "train_batch_size": 64,
613
+ "trial_name": null,
614
+ "trial_params": null
615
+ }
out/glu-linear-9L_run/checkpoint-1500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf92dabf7b603576d3b82f551073a92c3984245e43d5617ff63f7d3239a77c48
3
+ size 4920
out/glu-linear-9L_run/checkpoint-2000/config.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation": "linear",
3
+ "architectures": [
4
+ "TinyLlamaForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "bos_token_id": 1,
9
+ "dtype": "bfloat16",
10
+ "eos_token_id": 2,
11
+ "head_dim": 32,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 128,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 256,
16
+ "max_position_embeddings": 512,
17
+ "mlp_bias": false,
18
+ "mlp_type": "glu",
19
+ "model_type": "tiny_llama",
20
+ "num_attention_heads": 4,
21
+ "num_hidden_layers": 9,
22
+ "num_key_value_heads": 4,
23
+ "pad_token_id": 0,
24
+ "pretraining_tp": 1,
25
+ "rms_norm_eps": 1e-06,
26
+ "rope_parameters": {
27
+ "rope_theta": 10000.0,
28
+ "rope_type": "default"
29
+ },
30
+ "tie_word_embeddings": true,
31
+ "tokenizer_name": "w-ahmad/tiny-stories-tokenizer",
32
+ "transformers_version": "5.15.0.dev0",
33
+ "use_cache": false,
34
+ "vocab_size": 4096
35
+ }
out/glu-linear-9L_run/checkpoint-2000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c5bce0c90d5c105f9f425e0f61e8e55cbb5a8123aab308fa9ba9f925ad1e1ee1
3
+ size 4011496
out/glu-linear-9L_run/checkpoint-2000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:56ffc0ee4f7d59dd478948f78ef6cd956f9aef5ab08092267009b2b82a07fc60
3
+ size 8074746
out/glu-linear-9L_run/checkpoint-2000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b1a97db8e41139aa1239ba7fb79ddeb0af5998c6305a440c1fe182e6ad02f2f5
3
+ size 14244
out/glu-linear-9L_run/checkpoint-2000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e91519ad666af6acc4beafd44aa617aefb90f632388a0f75cbf88f957400c190
3
+ size 1064
out/glu-linear-9L_run/checkpoint-2000/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
out/glu-linear-9L_run/checkpoint-2000/tokenizer_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "backend": "tokenizers",
4
+ "bos_token": "<|endoftext|>",
5
+ "eos_token": "<|endoftext|>",
6
+ "errors": "replace",
7
+ "is_local": false,
8
+ "local_files_only": false,
9
+ "model_max_length": 1024,
10
+ "pad_token": "<|endoftext|>",
11
+ "tokenizer_class": "GPT2Tokenizer",
12
+ "unk_token": "<|endoftext|>"
13
+ }
out/glu-linear-9L_run/checkpoint-2000/trainer_state.json ADDED
@@ -0,0 +1,814 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.13481631277384565,
6
+ "eval_steps": 200,
7
+ "global_step": 2000,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.0013481631277384564,
14
+ "grad_norm": 1.25,
15
+ "learning_rate": 0.0003,
16
+ "loss": 7.778378295898437,
17
+ "step": 20
18
+ },
19
+ {
20
+ "epoch": 0.002696326255476913,
21
+ "grad_norm": 1.1875,
22
+ "learning_rate": 0.0003,
23
+ "loss": 6.979072570800781,
24
+ "step": 40
25
+ },
26
+ {
27
+ "epoch": 0.004044489383215369,
28
+ "grad_norm": 1.09375,
29
+ "learning_rate": 0.0003,
30
+ "loss": 6.366487884521485,
31
+ "step": 60
32
+ },
33
+ {
34
+ "epoch": 0.005392652510953826,
35
+ "grad_norm": 2.328125,
36
+ "learning_rate": 0.0003,
37
+ "loss": 5.798646926879883,
38
+ "step": 80
39
+ },
40
+ {
41
+ "epoch": 0.006740815638692282,
42
+ "grad_norm": 4.78125,
43
+ "learning_rate": 0.0003,
44
+ "loss": 5.200259017944336,
45
+ "step": 100
46
+ },
47
+ {
48
+ "epoch": 0.008088978766430738,
49
+ "grad_norm": 3.421875,
50
+ "learning_rate": 0.0003,
51
+ "loss": 4.721379089355469,
52
+ "step": 120
53
+ },
54
+ {
55
+ "epoch": 0.009437141894169195,
56
+ "grad_norm": 3.171875,
57
+ "learning_rate": 0.0003,
58
+ "loss": 4.299360275268555,
59
+ "step": 140
60
+ },
61
+ {
62
+ "epoch": 0.010785305021907651,
63
+ "grad_norm": 2.328125,
64
+ "learning_rate": 0.0003,
65
+ "loss": 3.917990493774414,
66
+ "step": 160
67
+ },
68
+ {
69
+ "epoch": 0.012133468149646108,
70
+ "grad_norm": 2.359375,
71
+ "learning_rate": 0.0003,
72
+ "loss": 3.5716182708740236,
73
+ "step": 180
74
+ },
75
+ {
76
+ "epoch": 0.013481631277384564,
77
+ "grad_norm": 2.890625,
78
+ "learning_rate": 0.0003,
79
+ "loss": 3.2290443420410155,
80
+ "step": 200
81
+ },
82
+ {
83
+ "epoch": 0.013481631277384564,
84
+ "eval_loss": 3.0630435943603516,
85
+ "eval_runtime": 8.4031,
86
+ "eval_samples_per_second": 1133.751,
87
+ "eval_steps_per_second": 17.732,
88
+ "step": 200
89
+ },
90
+ {
91
+ "epoch": 0.01482979440512302,
92
+ "grad_norm": 2.84375,
93
+ "learning_rate": 0.0003,
94
+ "loss": 2.9105119705200195,
95
+ "step": 220
96
+ },
97
+ {
98
+ "epoch": 0.016177957532861477,
99
+ "grad_norm": 2.21875,
100
+ "learning_rate": 0.0003,
101
+ "loss": 2.5742773056030273,
102
+ "step": 240
103
+ },
104
+ {
105
+ "epoch": 0.01752612066059993,
106
+ "grad_norm": 2.875,
107
+ "learning_rate": 0.0003,
108
+ "loss": 2.2510210037231446,
109
+ "step": 260
110
+ },
111
+ {
112
+ "epoch": 0.01887428378833839,
113
+ "grad_norm": 1.8359375,
114
+ "learning_rate": 0.0003,
115
+ "loss": 1.948414421081543,
116
+ "step": 280
117
+ },
118
+ {
119
+ "epoch": 0.020222446916076844,
120
+ "grad_norm": 2.828125,
121
+ "learning_rate": 0.0003,
122
+ "loss": 1.6830488204956056,
123
+ "step": 300
124
+ },
125
+ {
126
+ "epoch": 0.021570610043815303,
127
+ "grad_norm": 2.03125,
128
+ "learning_rate": 0.0003,
129
+ "loss": 1.488680934906006,
130
+ "step": 320
131
+ },
132
+ {
133
+ "epoch": 0.022918773171553757,
134
+ "grad_norm": 1.796875,
135
+ "learning_rate": 0.0003,
136
+ "loss": 1.3138065338134766,
137
+ "step": 340
138
+ },
139
+ {
140
+ "epoch": 0.024266936299292215,
141
+ "grad_norm": 2.515625,
142
+ "learning_rate": 0.0003,
143
+ "loss": 1.1533904075622559,
144
+ "step": 360
145
+ },
146
+ {
147
+ "epoch": 0.02561509942703067,
148
+ "grad_norm": 1.4921875,
149
+ "learning_rate": 0.0003,
150
+ "loss": 1.040479850769043,
151
+ "step": 380
152
+ },
153
+ {
154
+ "epoch": 0.026963262554769128,
155
+ "grad_norm": 1.953125,
156
+ "learning_rate": 0.0003,
157
+ "loss": 0.9418865203857422,
158
+ "step": 400
159
+ },
160
+ {
161
+ "epoch": 0.026963262554769128,
162
+ "eval_loss": 0.9077715873718262,
163
+ "eval_runtime": 8.327,
164
+ "eval_samples_per_second": 1144.104,
165
+ "eval_steps_per_second": 17.894,
166
+ "step": 400
167
+ },
168
+ {
169
+ "epoch": 0.028311425682507583,
170
+ "grad_norm": 1.0625,
171
+ "learning_rate": 0.0003,
172
+ "loss": 0.866397476196289,
173
+ "step": 420
174
+ },
175
+ {
176
+ "epoch": 0.02965958881024604,
177
+ "grad_norm": 1.6015625,
178
+ "learning_rate": 0.0003,
179
+ "loss": 0.8063242912292481,
180
+ "step": 440
181
+ },
182
+ {
183
+ "epoch": 0.031007751937984496,
184
+ "grad_norm": 1.171875,
185
+ "learning_rate": 0.0003,
186
+ "loss": 0.7396225929260254,
187
+ "step": 460
188
+ },
189
+ {
190
+ "epoch": 0.032355915065722954,
191
+ "grad_norm": 0.8125,
192
+ "learning_rate": 0.0003,
193
+ "loss": 0.679314136505127,
194
+ "step": 480
195
+ },
196
+ {
197
+ "epoch": 0.03370407819346141,
198
+ "grad_norm": 0.453125,
199
+ "learning_rate": 0.0003,
200
+ "loss": 0.6011962890625,
201
+ "step": 500
202
+ },
203
+ {
204
+ "epoch": 0.03505224132119986,
205
+ "grad_norm": 0.423828125,
206
+ "learning_rate": 0.0003,
207
+ "loss": 0.5412269592285156,
208
+ "step": 520
209
+ },
210
+ {
211
+ "epoch": 0.03640040444893832,
212
+ "grad_norm": 0.447265625,
213
+ "learning_rate": 0.0003,
214
+ "loss": 0.49077553749084474,
215
+ "step": 540
216
+ },
217
+ {
218
+ "epoch": 0.03774856757667678,
219
+ "grad_norm": 0.396484375,
220
+ "learning_rate": 0.0003,
221
+ "loss": 0.45958123207092283,
222
+ "step": 560
223
+ },
224
+ {
225
+ "epoch": 0.03909673070441524,
226
+ "grad_norm": 0.390625,
227
+ "learning_rate": 0.0003,
228
+ "loss": 0.43019986152648926,
229
+ "step": 580
230
+ },
231
+ {
232
+ "epoch": 0.04044489383215369,
233
+ "grad_norm": 0.373046875,
234
+ "learning_rate": 0.0003,
235
+ "loss": 0.4119072914123535,
236
+ "step": 600
237
+ },
238
+ {
239
+ "epoch": 0.04044489383215369,
240
+ "eval_loss": 0.4039357006549835,
241
+ "eval_runtime": 8.4391,
242
+ "eval_samples_per_second": 1128.917,
243
+ "eval_steps_per_second": 17.656,
244
+ "step": 600
245
+ },
246
+ {
247
+ "epoch": 0.04179305695989215,
248
+ "grad_norm": 0.34375,
249
+ "learning_rate": 0.0003,
250
+ "loss": 0.39308140277862547,
251
+ "step": 620
252
+ },
253
+ {
254
+ "epoch": 0.043141220087630605,
255
+ "grad_norm": 0.365234375,
256
+ "learning_rate": 0.0003,
257
+ "loss": 0.3809913158416748,
258
+ "step": 640
259
+ },
260
+ {
261
+ "epoch": 0.044489383215369056,
262
+ "grad_norm": 0.365234375,
263
+ "learning_rate": 0.0003,
264
+ "loss": 0.3649900913238525,
265
+ "step": 660
266
+ },
267
+ {
268
+ "epoch": 0.045837546343107514,
269
+ "grad_norm": 0.44140625,
270
+ "learning_rate": 0.0003,
271
+ "loss": 0.3559495210647583,
272
+ "step": 680
273
+ },
274
+ {
275
+ "epoch": 0.04718570947084597,
276
+ "grad_norm": 0.341796875,
277
+ "learning_rate": 0.0003,
278
+ "loss": 0.3457839012145996,
279
+ "step": 700
280
+ },
281
+ {
282
+ "epoch": 0.04853387259858443,
283
+ "grad_norm": 1.5546875,
284
+ "learning_rate": 0.0003,
285
+ "loss": 0.3411240577697754,
286
+ "step": 720
287
+ },
288
+ {
289
+ "epoch": 0.04988203572632288,
290
+ "grad_norm": 1.0,
291
+ "learning_rate": 0.0003,
292
+ "loss": 0.3416250228881836,
293
+ "step": 740
294
+ },
295
+ {
296
+ "epoch": 0.05123019885406134,
297
+ "grad_norm": 1.2890625,
298
+ "learning_rate": 0.0003,
299
+ "loss": 0.3455547094345093,
300
+ "step": 760
301
+ },
302
+ {
303
+ "epoch": 0.0525783619817998,
304
+ "grad_norm": 0.447265625,
305
+ "learning_rate": 0.0003,
306
+ "loss": 0.3340980291366577,
307
+ "step": 780
308
+ },
309
+ {
310
+ "epoch": 0.053926525109538256,
311
+ "grad_norm": 0.3125,
312
+ "learning_rate": 0.0003,
313
+ "loss": 0.3213367223739624,
314
+ "step": 800
315
+ },
316
+ {
317
+ "epoch": 0.053926525109538256,
318
+ "eval_loss": 0.32119977474212646,
319
+ "eval_runtime": 8.4,
320
+ "eval_samples_per_second": 1134.168,
321
+ "eval_steps_per_second": 17.738,
322
+ "step": 800
323
+ },
324
+ {
325
+ "epoch": 0.05527468823727671,
326
+ "grad_norm": 0.271484375,
327
+ "learning_rate": 0.0003,
328
+ "loss": 0.31615324020385743,
329
+ "step": 820
330
+ },
331
+ {
332
+ "epoch": 0.056622851365015166,
333
+ "grad_norm": 0.267578125,
334
+ "learning_rate": 0.0003,
335
+ "loss": 0.3102609634399414,
336
+ "step": 840
337
+ },
338
+ {
339
+ "epoch": 0.057971014492753624,
340
+ "grad_norm": 0.26171875,
341
+ "learning_rate": 0.0003,
342
+ "loss": 0.30272841453552246,
343
+ "step": 860
344
+ },
345
+ {
346
+ "epoch": 0.05931917762049208,
347
+ "grad_norm": 0.265625,
348
+ "learning_rate": 0.0003,
349
+ "loss": 0.2957427501678467,
350
+ "step": 880
351
+ },
352
+ {
353
+ "epoch": 0.06066734074823053,
354
+ "grad_norm": 0.265625,
355
+ "learning_rate": 0.0003,
356
+ "loss": 0.29436583518981935,
357
+ "step": 900
358
+ },
359
+ {
360
+ "epoch": 0.06201550387596899,
361
+ "grad_norm": 0.267578125,
362
+ "learning_rate": 0.0003,
363
+ "loss": 0.2905875205993652,
364
+ "step": 920
365
+ },
366
+ {
367
+ "epoch": 0.06336366700370745,
368
+ "grad_norm": 0.275390625,
369
+ "learning_rate": 0.0003,
370
+ "loss": 0.28632206916809083,
371
+ "step": 940
372
+ },
373
+ {
374
+ "epoch": 0.06471183013144591,
375
+ "grad_norm": 0.255859375,
376
+ "learning_rate": 0.0003,
377
+ "loss": 0.28044116497039795,
378
+ "step": 960
379
+ },
380
+ {
381
+ "epoch": 0.06605999325918437,
382
+ "grad_norm": 0.2890625,
383
+ "learning_rate": 0.0003,
384
+ "loss": 0.278120231628418,
385
+ "step": 980
386
+ },
387
+ {
388
+ "epoch": 0.06740815638692282,
389
+ "grad_norm": 0.28125,
390
+ "learning_rate": 0.0003,
391
+ "loss": 0.2752914667129517,
392
+ "step": 1000
393
+ },
394
+ {
395
+ "epoch": 0.06740815638692282,
396
+ "eval_loss": 0.2748347520828247,
397
+ "eval_runtime": 8.4469,
398
+ "eval_samples_per_second": 1127.864,
399
+ "eval_steps_per_second": 17.64,
400
+ "step": 1000
401
+ },
402
+ {
403
+ "epoch": 0.06875631951466127,
404
+ "grad_norm": 0.267578125,
405
+ "learning_rate": 0.0003,
406
+ "loss": 0.2703357934951782,
407
+ "step": 1020
408
+ },
409
+ {
410
+ "epoch": 0.07010448264239973,
411
+ "grad_norm": 0.2578125,
412
+ "learning_rate": 0.0003,
413
+ "loss": 0.2655144691467285,
414
+ "step": 1040
415
+ },
416
+ {
417
+ "epoch": 0.07145264577013818,
418
+ "grad_norm": 0.326171875,
419
+ "learning_rate": 0.0003,
420
+ "loss": 0.26470656394958497,
421
+ "step": 1060
422
+ },
423
+ {
424
+ "epoch": 0.07280080889787664,
425
+ "grad_norm": 0.337890625,
426
+ "learning_rate": 0.0003,
427
+ "loss": 0.26000075340270995,
428
+ "step": 1080
429
+ },
430
+ {
431
+ "epoch": 0.0741489720256151,
432
+ "grad_norm": 0.294921875,
433
+ "learning_rate": 0.0003,
434
+ "loss": 0.2588791847229004,
435
+ "step": 1100
436
+ },
437
+ {
438
+ "epoch": 0.07549713515335356,
439
+ "grad_norm": 0.3125,
440
+ "learning_rate": 0.0003,
441
+ "loss": 0.2555199146270752,
442
+ "step": 1120
443
+ },
444
+ {
445
+ "epoch": 0.07684529828109202,
446
+ "grad_norm": 0.326171875,
447
+ "learning_rate": 0.0003,
448
+ "loss": 0.25160832405090333,
449
+ "step": 1140
450
+ },
451
+ {
452
+ "epoch": 0.07819346140883048,
453
+ "grad_norm": 0.287109375,
454
+ "learning_rate": 0.0003,
455
+ "loss": 0.2508568048477173,
456
+ "step": 1160
457
+ },
458
+ {
459
+ "epoch": 0.07954162453656892,
460
+ "grad_norm": 0.8359375,
461
+ "learning_rate": 0.0003,
462
+ "loss": 0.26505236625671386,
463
+ "step": 1180
464
+ },
465
+ {
466
+ "epoch": 0.08088978766430738,
467
+ "grad_norm": 0.63671875,
468
+ "learning_rate": 0.0003,
469
+ "loss": 0.2577481269836426,
470
+ "step": 1200
471
+ },
472
+ {
473
+ "epoch": 0.08088978766430738,
474
+ "eval_loss": 0.25307410955429077,
475
+ "eval_runtime": 8.4059,
476
+ "eval_samples_per_second": 1133.37,
477
+ "eval_steps_per_second": 17.726,
478
+ "step": 1200
479
+ },
480
+ {
481
+ "epoch": 0.08223795079204584,
482
+ "grad_norm": 0.2890625,
483
+ "learning_rate": 0.0003,
484
+ "loss": 0.248107647895813,
485
+ "step": 1220
486
+ },
487
+ {
488
+ "epoch": 0.0835861139197843,
489
+ "grad_norm": 0.244140625,
490
+ "learning_rate": 0.0003,
491
+ "loss": 0.24457168579101562,
492
+ "step": 1240
493
+ },
494
+ {
495
+ "epoch": 0.08493427704752275,
496
+ "grad_norm": 0.23828125,
497
+ "learning_rate": 0.0003,
498
+ "loss": 0.24279680252075195,
499
+ "step": 1260
500
+ },
501
+ {
502
+ "epoch": 0.08628244017526121,
503
+ "grad_norm": 0.2314453125,
504
+ "learning_rate": 0.0003,
505
+ "loss": 0.23819704055786134,
506
+ "step": 1280
507
+ },
508
+ {
509
+ "epoch": 0.08763060330299967,
510
+ "grad_norm": 0.2353515625,
511
+ "learning_rate": 0.0003,
512
+ "loss": 0.23753654956817627,
513
+ "step": 1300
514
+ },
515
+ {
516
+ "epoch": 0.08897876643073811,
517
+ "grad_norm": 0.2275390625,
518
+ "learning_rate": 0.0003,
519
+ "loss": 0.23422625064849853,
520
+ "step": 1320
521
+ },
522
+ {
523
+ "epoch": 0.09032692955847657,
524
+ "grad_norm": 0.2255859375,
525
+ "learning_rate": 0.0003,
526
+ "loss": 0.23316617012023927,
527
+ "step": 1340
528
+ },
529
+ {
530
+ "epoch": 0.09167509268621503,
531
+ "grad_norm": 0.22265625,
532
+ "learning_rate": 0.0003,
533
+ "loss": 0.23011410236358643,
534
+ "step": 1360
535
+ },
536
+ {
537
+ "epoch": 0.09302325581395349,
538
+ "grad_norm": 0.2177734375,
539
+ "learning_rate": 0.0003,
540
+ "loss": 0.22971732616424562,
541
+ "step": 1380
542
+ },
543
+ {
544
+ "epoch": 0.09437141894169195,
545
+ "grad_norm": 0.2236328125,
546
+ "learning_rate": 0.0003,
547
+ "loss": 0.22751867771148682,
548
+ "step": 1400
549
+ },
550
+ {
551
+ "epoch": 0.09437141894169195,
552
+ "eval_loss": 0.22943033277988434,
553
+ "eval_runtime": 8.4608,
554
+ "eval_samples_per_second": 1126.015,
555
+ "eval_steps_per_second": 17.611,
556
+ "step": 1400
557
+ },
558
+ {
559
+ "epoch": 0.0957195820694304,
560
+ "grad_norm": 0.2294921875,
561
+ "learning_rate": 0.0003,
562
+ "loss": 0.22718305587768556,
563
+ "step": 1420
564
+ },
565
+ {
566
+ "epoch": 0.09706774519716886,
567
+ "grad_norm": 0.2138671875,
568
+ "learning_rate": 0.0003,
569
+ "loss": 0.22553596496582032,
570
+ "step": 1440
571
+ },
572
+ {
573
+ "epoch": 0.09841590832490732,
574
+ "grad_norm": 0.2373046875,
575
+ "learning_rate": 0.0003,
576
+ "loss": 0.22326209545135497,
577
+ "step": 1460
578
+ },
579
+ {
580
+ "epoch": 0.09976407145264576,
581
+ "grad_norm": 0.212890625,
582
+ "learning_rate": 0.0003,
583
+ "loss": 0.2213585376739502,
584
+ "step": 1480
585
+ },
586
+ {
587
+ "epoch": 0.10111223458038422,
588
+ "grad_norm": 0.21875,
589
+ "learning_rate": 0.0003,
590
+ "loss": 0.22035691738128663,
591
+ "step": 1500
592
+ },
593
+ {
594
+ "epoch": 0.10246039770812268,
595
+ "grad_norm": 0.2275390625,
596
+ "learning_rate": 0.0003,
597
+ "loss": 0.21844491958618165,
598
+ "step": 1520
599
+ },
600
+ {
601
+ "epoch": 0.10380856083586114,
602
+ "grad_norm": 0.2216796875,
603
+ "learning_rate": 0.0003,
604
+ "loss": 0.2179863691329956,
605
+ "step": 1540
606
+ },
607
+ {
608
+ "epoch": 0.1051567239635996,
609
+ "grad_norm": 0.2294921875,
610
+ "learning_rate": 0.0003,
611
+ "loss": 0.217010498046875,
612
+ "step": 1560
613
+ },
614
+ {
615
+ "epoch": 0.10650488709133805,
616
+ "grad_norm": 0.22265625,
617
+ "learning_rate": 0.0003,
618
+ "loss": 0.2174626350402832,
619
+ "step": 1580
620
+ },
621
+ {
622
+ "epoch": 0.10785305021907651,
623
+ "grad_norm": 1.109375,
624
+ "learning_rate": 0.0003,
625
+ "loss": 0.2253711462020874,
626
+ "step": 1600
627
+ },
628
+ {
629
+ "epoch": 0.10785305021907651,
630
+ "eval_loss": 0.23372706770896912,
631
+ "eval_runtime": 8.4084,
632
+ "eval_samples_per_second": 1133.035,
633
+ "eval_steps_per_second": 17.72,
634
+ "step": 1600
635
+ },
636
+ {
637
+ "epoch": 0.10920121334681497,
638
+ "grad_norm": 0.494140625,
639
+ "learning_rate": 0.0003,
640
+ "loss": 0.22224326133728028,
641
+ "step": 1620
642
+ },
643
+ {
644
+ "epoch": 0.11054937647455342,
645
+ "grad_norm": 0.2451171875,
646
+ "learning_rate": 0.0003,
647
+ "loss": 0.21563270092010497,
648
+ "step": 1640
649
+ },
650
+ {
651
+ "epoch": 0.11189753960229187,
652
+ "grad_norm": 0.21484375,
653
+ "learning_rate": 0.0003,
654
+ "loss": 0.21244332790374756,
655
+ "step": 1660
656
+ },
657
+ {
658
+ "epoch": 0.11324570273003033,
659
+ "grad_norm": 0.2080078125,
660
+ "learning_rate": 0.0003,
661
+ "loss": 0.21173064708709716,
662
+ "step": 1680
663
+ },
664
+ {
665
+ "epoch": 0.11459386585776879,
666
+ "grad_norm": 0.208984375,
667
+ "learning_rate": 0.0003,
668
+ "loss": 0.2120938777923584,
669
+ "step": 1700
670
+ },
671
+ {
672
+ "epoch": 0.11594202898550725,
673
+ "grad_norm": 0.2060546875,
674
+ "learning_rate": 0.0003,
675
+ "loss": 0.20998167991638184,
676
+ "step": 1720
677
+ },
678
+ {
679
+ "epoch": 0.1172901921132457,
680
+ "grad_norm": 0.205078125,
681
+ "learning_rate": 0.0003,
682
+ "loss": 0.2100886106491089,
683
+ "step": 1740
684
+ },
685
+ {
686
+ "epoch": 0.11863835524098416,
687
+ "grad_norm": 0.2119140625,
688
+ "learning_rate": 0.0003,
689
+ "loss": 0.2085808038711548,
690
+ "step": 1760
691
+ },
692
+ {
693
+ "epoch": 0.11998651836872262,
694
+ "grad_norm": 0.201171875,
695
+ "learning_rate": 0.0003,
696
+ "loss": 0.20966055393218994,
697
+ "step": 1780
698
+ },
699
+ {
700
+ "epoch": 0.12133468149646107,
701
+ "grad_norm": 0.205078125,
702
+ "learning_rate": 0.0003,
703
+ "loss": 0.20960137844085694,
704
+ "step": 1800
705
+ },
706
+ {
707
+ "epoch": 0.12133468149646107,
708
+ "eval_loss": 0.20872168242931366,
709
+ "eval_runtime": 8.4554,
710
+ "eval_samples_per_second": 1126.736,
711
+ "eval_steps_per_second": 17.622,
712
+ "step": 1800
713
+ },
714
+ {
715
+ "epoch": 0.12268284462419952,
716
+ "grad_norm": 0.2158203125,
717
+ "learning_rate": 0.0003,
718
+ "loss": 0.20879044532775878,
719
+ "step": 1820
720
+ },
721
+ {
722
+ "epoch": 0.12403100775193798,
723
+ "grad_norm": 0.2060546875,
724
+ "learning_rate": 0.0003,
725
+ "loss": 0.20482723712921141,
726
+ "step": 1840
727
+ },
728
+ {
729
+ "epoch": 0.12537917087967643,
730
+ "grad_norm": 0.220703125,
731
+ "learning_rate": 0.0003,
732
+ "loss": 0.20474963188171386,
733
+ "step": 1860
734
+ },
735
+ {
736
+ "epoch": 0.1267273340074149,
737
+ "grad_norm": 0.2109375,
738
+ "learning_rate": 0.0003,
739
+ "loss": 0.20423946380615235,
740
+ "step": 1880
741
+ },
742
+ {
743
+ "epoch": 0.12807549713515334,
744
+ "grad_norm": 0.216796875,
745
+ "learning_rate": 0.0003,
746
+ "loss": 0.20402367115020753,
747
+ "step": 1900
748
+ },
749
+ {
750
+ "epoch": 0.12942366026289182,
751
+ "grad_norm": 0.2119140625,
752
+ "learning_rate": 0.0003,
753
+ "loss": 0.20232999324798584,
754
+ "step": 1920
755
+ },
756
+ {
757
+ "epoch": 0.13077182339063026,
758
+ "grad_norm": 0.203125,
759
+ "learning_rate": 0.0003,
760
+ "loss": 0.20286252498626708,
761
+ "step": 1940
762
+ },
763
+ {
764
+ "epoch": 0.13211998651836873,
765
+ "grad_norm": 0.20703125,
766
+ "learning_rate": 0.0003,
767
+ "loss": 0.20078661441802978,
768
+ "step": 1960
769
+ },
770
+ {
771
+ "epoch": 0.13346814964610718,
772
+ "grad_norm": 0.2080078125,
773
+ "learning_rate": 0.0003,
774
+ "loss": 0.20174212455749513,
775
+ "step": 1980
776
+ },
777
+ {
778
+ "epoch": 0.13481631277384565,
779
+ "grad_norm": 0.208984375,
780
+ "learning_rate": 0.0003,
781
+ "loss": 0.20114669799804688,
782
+ "step": 2000
783
+ },
784
+ {
785
+ "epoch": 0.13481631277384565,
786
+ "eval_loss": 0.20278693735599518,
787
+ "eval_runtime": 8.4147,
788
+ "eval_samples_per_second": 1132.186,
789
+ "eval_steps_per_second": 17.707,
790
+ "step": 2000
791
+ }
792
+ ],
793
+ "logging_steps": 20,
794
+ "max_steps": 2000,
795
+ "num_input_tokens_seen": 0,
796
+ "num_train_epochs": 1,
797
+ "save_steps": 500,
798
+ "stateful_callbacks": {
799
+ "TrainerControl": {
800
+ "args": {
801
+ "should_epoch_stop": false,
802
+ "should_evaluate": false,
803
+ "should_log": false,
804
+ "should_save": true,
805
+ "should_training_stop": true
806
+ },
807
+ "attributes": {}
808
+ }
809
+ },
810
+ "total_flos": 580776886272000.0,
811
+ "train_batch_size": 64,
812
+ "trial_name": null,
813
+ "trial_params": null
814
+ }
out/glu-linear-9L_run/checkpoint-2000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf92dabf7b603576d3b82f551073a92c3984245e43d5617ff63f7d3239a77c48
3
+ size 4920
out/glu-linear-9L_run/model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:920847be41cb9f066360b9a2fdbbb06c94b8396b086f977d97d4ab5c5e8c5873
3
  size 4011496
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c5bce0c90d5c105f9f425e0f61e8e55cbb5a8123aab308fa9ba9f925ad1e1ee1
3
  size 4011496
out/glu-linear-9L_run/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
out/glu-linear-9L_run/tokenizer_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "backend": "tokenizers",
4
+ "bos_token": "<|endoftext|>",
5
+ "eos_token": "<|endoftext|>",
6
+ "errors": "replace",
7
+ "is_local": false,
8
+ "local_files_only": false,
9
+ "model_max_length": 1024,
10
+ "pad_token": "<|endoftext|>",
11
+ "tokenizer_class": "GPT2Tokenizer",
12
+ "unk_token": "<|endoftext|>"
13
+ }
out/glu-linear-9L_run/training_log.jsonl CHANGED
The diff for this file is too large to render. See raw diff
 
out/glu-situglu-9L_run/checkpoint-1000/config.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation": "situglu",
3
+ "architectures": [
4
+ "TinyLlamaForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "bos_token_id": 1,
9
+ "dtype": "bfloat16",
10
+ "eos_token_id": 2,
11
+ "head_dim": 32,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 128,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 256,
16
+ "max_position_embeddings": 512,
17
+ "mlp_bias": false,
18
+ "mlp_type": "glu",
19
+ "model_type": "tiny_llama",
20
+ "num_attention_heads": 4,
21
+ "num_hidden_layers": 9,
22
+ "num_key_value_heads": 4,
23
+ "pad_token_id": 0,
24
+ "pretraining_tp": 1,
25
+ "rms_norm_eps": 1e-06,
26
+ "rope_parameters": {
27
+ "rope_theta": 10000.0,
28
+ "rope_type": "default"
29
+ },
30
+ "tie_word_embeddings": true,
31
+ "tokenizer_name": "w-ahmad/tiny-stories-tokenizer",
32
+ "transformers_version": "5.15.0.dev0",
33
+ "use_cache": false,
34
+ "vocab_size": 4096
35
+ }
out/glu-situglu-9L_run/checkpoint-1000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d2cd5c8ad2c757d75a6418d9320642b878894b56597a6142c7aad179394f4730
3
+ size 4011552
out/glu-situglu-9L_run/checkpoint-1000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d0a3ef2ff52b5b829bbd3c1b9b2aefe33afee3bd104dc4d7a66b1a42a299506f
3
+ size 8074746
out/glu-situglu-9L_run/checkpoint-1000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:95b6047bd8cc6f4cdf7c46dea47edb8e542435510070c6cd1e0a7d9ccf5fd7da
3
+ size 14244
out/glu-situglu-9L_run/checkpoint-1000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d5f71c5fa69d56807b14256b78700be6088fcb1bda0ff8e0ea6f8709a7df2fde
3
+ size 1064
out/glu-situglu-9L_run/checkpoint-1000/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
out/glu-situglu-9L_run/checkpoint-1000/tokenizer_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "backend": "tokenizers",
4
+ "bos_token": "<|endoftext|>",
5
+ "eos_token": "<|endoftext|>",
6
+ "errors": "replace",
7
+ "is_local": false,
8
+ "local_files_only": false,
9
+ "model_max_length": 1024,
10
+ "pad_token": "<|endoftext|>",
11
+ "tokenizer_class": "GPT2Tokenizer",
12
+ "unk_token": "<|endoftext|>"
13
+ }
out/glu-situglu-9L_run/checkpoint-1000/trainer_state.json ADDED
@@ -0,0 +1,424 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.06740815638692282,
6
+ "eval_steps": 200,
7
+ "global_step": 1000,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.0013481631277384564,
14
+ "grad_norm": 1.2578125,
15
+ "learning_rate": 0.0003,
16
+ "loss": 7.793804931640625,
17
+ "step": 20
18
+ },
19
+ {
20
+ "epoch": 0.002696326255476913,
21
+ "grad_norm": 1.171875,
22
+ "learning_rate": 0.0003,
23
+ "loss": 7.022835540771484,
24
+ "step": 40
25
+ },
26
+ {
27
+ "epoch": 0.004044489383215369,
28
+ "grad_norm": 1.0,
29
+ "learning_rate": 0.0003,
30
+ "loss": 6.450706481933594,
31
+ "step": 60
32
+ },
33
+ {
34
+ "epoch": 0.005392652510953826,
35
+ "grad_norm": 1.9296875,
36
+ "learning_rate": 0.0003,
37
+ "loss": 6.003890609741211,
38
+ "step": 80
39
+ },
40
+ {
41
+ "epoch": 0.006740815638692282,
42
+ "grad_norm": 3.53125,
43
+ "learning_rate": 0.0003,
44
+ "loss": 5.4485828399658205,
45
+ "step": 100
46
+ },
47
+ {
48
+ "epoch": 0.008088978766430738,
49
+ "grad_norm": 2.9375,
50
+ "learning_rate": 0.0003,
51
+ "loss": 4.944174575805664,
52
+ "step": 120
53
+ },
54
+ {
55
+ "epoch": 0.009437141894169195,
56
+ "grad_norm": 3.3125,
57
+ "learning_rate": 0.0003,
58
+ "loss": 4.507948684692383,
59
+ "step": 140
60
+ },
61
+ {
62
+ "epoch": 0.010785305021907651,
63
+ "grad_norm": 2.859375,
64
+ "learning_rate": 0.0003,
65
+ "loss": 4.094369125366211,
66
+ "step": 160
67
+ },
68
+ {
69
+ "epoch": 0.012133468149646108,
70
+ "grad_norm": 2.703125,
71
+ "learning_rate": 0.0003,
72
+ "loss": 3.715929412841797,
73
+ "step": 180
74
+ },
75
+ {
76
+ "epoch": 0.013481631277384564,
77
+ "grad_norm": 2.203125,
78
+ "learning_rate": 0.0003,
79
+ "loss": 3.352855682373047,
80
+ "step": 200
81
+ },
82
+ {
83
+ "epoch": 0.013481631277384564,
84
+ "eval_loss": 3.1672160625457764,
85
+ "eval_runtime": 8.5158,
86
+ "eval_samples_per_second": 1118.739,
87
+ "eval_steps_per_second": 17.497,
88
+ "step": 200
89
+ },
90
+ {
91
+ "epoch": 0.01482979440512302,
92
+ "grad_norm": 2.203125,
93
+ "learning_rate": 0.0003,
94
+ "loss": 2.9979101181030274,
95
+ "step": 220
96
+ },
97
+ {
98
+ "epoch": 0.016177957532861477,
99
+ "grad_norm": 1.9140625,
100
+ "learning_rate": 0.0003,
101
+ "loss": 2.638140106201172,
102
+ "step": 240
103
+ },
104
+ {
105
+ "epoch": 0.01752612066059993,
106
+ "grad_norm": 2.140625,
107
+ "learning_rate": 0.0003,
108
+ "loss": 2.2962114334106447,
109
+ "step": 260
110
+ },
111
+ {
112
+ "epoch": 0.01887428378833839,
113
+ "grad_norm": 1.84375,
114
+ "learning_rate": 0.0003,
115
+ "loss": 1.9714105606079102,
116
+ "step": 280
117
+ },
118
+ {
119
+ "epoch": 0.020222446916076844,
120
+ "grad_norm": 2.0625,
121
+ "learning_rate": 0.0003,
122
+ "loss": 1.6777704238891602,
123
+ "step": 300
124
+ },
125
+ {
126
+ "epoch": 0.021570610043815303,
127
+ "grad_norm": 1.4765625,
128
+ "learning_rate": 0.0003,
129
+ "loss": 1.4184170722961427,
130
+ "step": 320
131
+ },
132
+ {
133
+ "epoch": 0.022918773171553757,
134
+ "grad_norm": 1.4375,
135
+ "learning_rate": 0.0003,
136
+ "loss": 1.2252904891967773,
137
+ "step": 340
138
+ },
139
+ {
140
+ "epoch": 0.024266936299292215,
141
+ "grad_norm": 1.0859375,
142
+ "learning_rate": 0.0003,
143
+ "loss": 1.0854849815368652,
144
+ "step": 360
145
+ },
146
+ {
147
+ "epoch": 0.02561509942703067,
148
+ "grad_norm": 1.109375,
149
+ "learning_rate": 0.0003,
150
+ "loss": 0.9601390838623047,
151
+ "step": 380
152
+ },
153
+ {
154
+ "epoch": 0.026963262554769128,
155
+ "grad_norm": 0.60546875,
156
+ "learning_rate": 0.0003,
157
+ "loss": 0.8285150527954102,
158
+ "step": 400
159
+ },
160
+ {
161
+ "epoch": 0.026963262554769128,
162
+ "eval_loss": 0.77406907081604,
163
+ "eval_runtime": 8.4577,
164
+ "eval_samples_per_second": 1126.423,
165
+ "eval_steps_per_second": 17.617,
166
+ "step": 400
167
+ },
168
+ {
169
+ "epoch": 0.028311425682507583,
170
+ "grad_norm": 0.5234375,
171
+ "learning_rate": 0.0003,
172
+ "loss": 0.7167386531829834,
173
+ "step": 420
174
+ },
175
+ {
176
+ "epoch": 0.02965958881024604,
177
+ "grad_norm": 0.498046875,
178
+ "learning_rate": 0.0003,
179
+ "loss": 0.6465096950531006,
180
+ "step": 440
181
+ },
182
+ {
183
+ "epoch": 0.031007751937984496,
184
+ "grad_norm": 0.51953125,
185
+ "learning_rate": 0.0003,
186
+ "loss": 0.5791397094726562,
187
+ "step": 460
188
+ },
189
+ {
190
+ "epoch": 0.032355915065722954,
191
+ "grad_norm": 0.4375,
192
+ "learning_rate": 0.0003,
193
+ "loss": 0.5380198001861572,
194
+ "step": 480
195
+ },
196
+ {
197
+ "epoch": 0.03370407819346141,
198
+ "grad_norm": 0.447265625,
199
+ "learning_rate": 0.0003,
200
+ "loss": 0.49995784759521483,
201
+ "step": 500
202
+ },
203
+ {
204
+ "epoch": 0.03505224132119986,
205
+ "grad_norm": 0.35546875,
206
+ "learning_rate": 0.0003,
207
+ "loss": 0.47292389869689944,
208
+ "step": 520
209
+ },
210
+ {
211
+ "epoch": 0.03640040444893832,
212
+ "grad_norm": 0.361328125,
213
+ "learning_rate": 0.0003,
214
+ "loss": 0.44700446128845217,
215
+ "step": 540
216
+ },
217
+ {
218
+ "epoch": 0.03774856757667678,
219
+ "grad_norm": 0.408203125,
220
+ "learning_rate": 0.0003,
221
+ "loss": 0.4330583572387695,
222
+ "step": 560
223
+ },
224
+ {
225
+ "epoch": 0.03909673070441524,
226
+ "grad_norm": 0.365234375,
227
+ "learning_rate": 0.0003,
228
+ "loss": 0.4167480945587158,
229
+ "step": 580
230
+ },
231
+ {
232
+ "epoch": 0.04044489383215369,
233
+ "grad_norm": 0.365234375,
234
+ "learning_rate": 0.0003,
235
+ "loss": 0.40725293159484866,
236
+ "step": 600
237
+ },
238
+ {
239
+ "epoch": 0.04044489383215369,
240
+ "eval_loss": 0.40257543325424194,
241
+ "eval_runtime": 8.4955,
242
+ "eval_samples_per_second": 1121.412,
243
+ "eval_steps_per_second": 17.539,
244
+ "step": 600
245
+ },
246
+ {
247
+ "epoch": 0.04179305695989215,
248
+ "grad_norm": 0.380859375,
249
+ "learning_rate": 0.0003,
250
+ "loss": 0.39375302791595457,
251
+ "step": 620
252
+ },
253
+ {
254
+ "epoch": 0.043141220087630605,
255
+ "grad_norm": 2.453125,
256
+ "learning_rate": 0.0003,
257
+ "loss": 0.3863722324371338,
258
+ "step": 640
259
+ },
260
+ {
261
+ "epoch": 0.044489383215369056,
262
+ "grad_norm": 0.9453125,
263
+ "learning_rate": 0.0003,
264
+ "loss": 0.39115495681762696,
265
+ "step": 660
266
+ },
267
+ {
268
+ "epoch": 0.045837546343107514,
269
+ "grad_norm": 0.408203125,
270
+ "learning_rate": 0.0003,
271
+ "loss": 0.3735619783401489,
272
+ "step": 680
273
+ },
274
+ {
275
+ "epoch": 0.04718570947084597,
276
+ "grad_norm": 0.306640625,
277
+ "learning_rate": 0.0003,
278
+ "loss": 0.360651421546936,
279
+ "step": 700
280
+ },
281
+ {
282
+ "epoch": 0.04853387259858443,
283
+ "grad_norm": 0.279296875,
284
+ "learning_rate": 0.0003,
285
+ "loss": 0.3540374279022217,
286
+ "step": 720
287
+ },
288
+ {
289
+ "epoch": 0.04988203572632288,
290
+ "grad_norm": 0.27734375,
291
+ "learning_rate": 0.0003,
292
+ "loss": 0.3440989971160889,
293
+ "step": 740
294
+ },
295
+ {
296
+ "epoch": 0.05123019885406134,
297
+ "grad_norm": 0.2734375,
298
+ "learning_rate": 0.0003,
299
+ "loss": 0.34135589599609373,
300
+ "step": 760
301
+ },
302
+ {
303
+ "epoch": 0.0525783619817998,
304
+ "grad_norm": 0.29296875,
305
+ "learning_rate": 0.0003,
306
+ "loss": 0.3341085910797119,
307
+ "step": 780
308
+ },
309
+ {
310
+ "epoch": 0.053926525109538256,
311
+ "grad_norm": 0.294921875,
312
+ "learning_rate": 0.0003,
313
+ "loss": 0.3262310266494751,
314
+ "step": 800
315
+ },
316
+ {
317
+ "epoch": 0.053926525109538256,
318
+ "eval_loss": 0.3270248770713806,
319
+ "eval_runtime": 8.5141,
320
+ "eval_samples_per_second": 1118.961,
321
+ "eval_steps_per_second": 17.5,
322
+ "step": 800
323
+ },
324
+ {
325
+ "epoch": 0.05527468823727671,
326
+ "grad_norm": 0.32421875,
327
+ "learning_rate": 0.0003,
328
+ "loss": 0.3224670648574829,
329
+ "step": 820
330
+ },
331
+ {
332
+ "epoch": 0.056622851365015166,
333
+ "grad_norm": 0.291015625,
334
+ "learning_rate": 0.0003,
335
+ "loss": 0.31727190017700196,
336
+ "step": 840
337
+ },
338
+ {
339
+ "epoch": 0.057971014492753624,
340
+ "grad_norm": 0.298828125,
341
+ "learning_rate": 0.0003,
342
+ "loss": 0.31037948131561277,
343
+ "step": 860
344
+ },
345
+ {
346
+ "epoch": 0.05931917762049208,
347
+ "grad_norm": 0.283203125,
348
+ "learning_rate": 0.0003,
349
+ "loss": 0.30405306816101074,
350
+ "step": 880
351
+ },
352
+ {
353
+ "epoch": 0.06066734074823053,
354
+ "grad_norm": 0.306640625,
355
+ "learning_rate": 0.0003,
356
+ "loss": 0.30279972553253176,
357
+ "step": 900
358
+ },
359
+ {
360
+ "epoch": 0.06201550387596899,
361
+ "grad_norm": 0.328125,
362
+ "learning_rate": 0.0003,
363
+ "loss": 0.2994309663772583,
364
+ "step": 920
365
+ },
366
+ {
367
+ "epoch": 0.06336366700370745,
368
+ "grad_norm": 0.318359375,
369
+ "learning_rate": 0.0003,
370
+ "loss": 0.2952401638031006,
371
+ "step": 940
372
+ },
373
+ {
374
+ "epoch": 0.06471183013144591,
375
+ "grad_norm": 0.330078125,
376
+ "learning_rate": 0.0003,
377
+ "loss": 0.2905283451080322,
378
+ "step": 960
379
+ },
380
+ {
381
+ "epoch": 0.06605999325918437,
382
+ "grad_norm": 0.365234375,
383
+ "learning_rate": 0.0003,
384
+ "loss": 0.28865833282470704,
385
+ "step": 980
386
+ },
387
+ {
388
+ "epoch": 0.06740815638692282,
389
+ "grad_norm": 0.3046875,
390
+ "learning_rate": 0.0003,
391
+ "loss": 0.2860438585281372,
392
+ "step": 1000
393
+ },
394
+ {
395
+ "epoch": 0.06740815638692282,
396
+ "eval_loss": 0.28539252281188965,
397
+ "eval_runtime": 8.5689,
398
+ "eval_samples_per_second": 1111.809,
399
+ "eval_steps_per_second": 17.388,
400
+ "step": 1000
401
+ }
402
+ ],
403
+ "logging_steps": 20,
404
+ "max_steps": 2000,
405
+ "num_input_tokens_seen": 0,
406
+ "num_train_epochs": 1,
407
+ "save_steps": 500,
408
+ "stateful_callbacks": {
409
+ "TrainerControl": {
410
+ "args": {
411
+ "should_epoch_stop": false,
412
+ "should_evaluate": false,
413
+ "should_log": false,
414
+ "should_save": true,
415
+ "should_training_stop": false
416
+ },
417
+ "attributes": {}
418
+ }
419
+ },
420
+ "total_flos": 290388443136000.0,
421
+ "train_batch_size": 64,
422
+ "trial_name": null,
423
+ "trial_params": null
424
+ }
out/glu-situglu-9L_run/checkpoint-1000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f2d9b411968661617d32794fc4981dd80c49ba782deaa6cc247aefafa8768d2e
3
+ size 4920
out/glu-situglu-9L_run/checkpoint-1500/config.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation": "situglu",
3
+ "architectures": [
4
+ "TinyLlamaForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "bos_token_id": 1,
9
+ "dtype": "bfloat16",
10
+ "eos_token_id": 2,
11
+ "head_dim": 32,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 128,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 256,
16
+ "max_position_embeddings": 512,
17
+ "mlp_bias": false,
18
+ "mlp_type": "glu",
19
+ "model_type": "tiny_llama",
20
+ "num_attention_heads": 4,
21
+ "num_hidden_layers": 9,
22
+ "num_key_value_heads": 4,
23
+ "pad_token_id": 0,
24
+ "pretraining_tp": 1,
25
+ "rms_norm_eps": 1e-06,
26
+ "rope_parameters": {
27
+ "rope_theta": 10000.0,
28
+ "rope_type": "default"
29
+ },
30
+ "tie_word_embeddings": true,
31
+ "tokenizer_name": "w-ahmad/tiny-stories-tokenizer",
32
+ "transformers_version": "5.15.0.dev0",
33
+ "use_cache": false,
34
+ "vocab_size": 4096
35
+ }
out/glu-situglu-9L_run/checkpoint-1500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cb8dcf9f3e4206d2f3e5384981dd51034aa0869564b18933bb1057f78822efd6
3
+ size 4011552
out/glu-situglu-9L_run/checkpoint-1500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:64b910f3c45cceca5c8cda6fc10f98650287bd06597ec92b3d3cd75e68d8a960
3
+ size 8074746
out/glu-situglu-9L_run/checkpoint-1500/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2768285b45b2a0c05f6f50bbb8c0287fca6f62a8cde6d1b1f02151ac72ee8dc
3
+ size 14244
out/glu-situglu-9L_run/checkpoint-1500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b26a759cc207e28dea06eae7ad531c2dd78a09978f6972edbfc1566adccd7142
3
+ size 1064
out/glu-situglu-9L_run/checkpoint-1500/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
out/glu-situglu-9L_run/checkpoint-1500/tokenizer_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "backend": "tokenizers",
4
+ "bos_token": "<|endoftext|>",
5
+ "eos_token": "<|endoftext|>",
6
+ "errors": "replace",
7
+ "is_local": false,
8
+ "local_files_only": false,
9
+ "model_max_length": 1024,
10
+ "pad_token": "<|endoftext|>",
11
+ "tokenizer_class": "GPT2Tokenizer",
12
+ "unk_token": "<|endoftext|>"
13
+ }
out/glu-situglu-9L_run/checkpoint-1500/trainer_state.json ADDED
@@ -0,0 +1,615 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.10111223458038422,
6
+ "eval_steps": 200,
7
+ "global_step": 1500,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.0013481631277384564,
14
+ "grad_norm": 1.2578125,
15
+ "learning_rate": 0.0003,
16
+ "loss": 7.793804931640625,
17
+ "step": 20
18
+ },
19
+ {
20
+ "epoch": 0.002696326255476913,
21
+ "grad_norm": 1.171875,
22
+ "learning_rate": 0.0003,
23
+ "loss": 7.022835540771484,
24
+ "step": 40
25
+ },
26
+ {
27
+ "epoch": 0.004044489383215369,
28
+ "grad_norm": 1.0,
29
+ "learning_rate": 0.0003,
30
+ "loss": 6.450706481933594,
31
+ "step": 60
32
+ },
33
+ {
34
+ "epoch": 0.005392652510953826,
35
+ "grad_norm": 1.9296875,
36
+ "learning_rate": 0.0003,
37
+ "loss": 6.003890609741211,
38
+ "step": 80
39
+ },
40
+ {
41
+ "epoch": 0.006740815638692282,
42
+ "grad_norm": 3.53125,
43
+ "learning_rate": 0.0003,
44
+ "loss": 5.4485828399658205,
45
+ "step": 100
46
+ },
47
+ {
48
+ "epoch": 0.008088978766430738,
49
+ "grad_norm": 2.9375,
50
+ "learning_rate": 0.0003,
51
+ "loss": 4.944174575805664,
52
+ "step": 120
53
+ },
54
+ {
55
+ "epoch": 0.009437141894169195,
56
+ "grad_norm": 3.3125,
57
+ "learning_rate": 0.0003,
58
+ "loss": 4.507948684692383,
59
+ "step": 140
60
+ },
61
+ {
62
+ "epoch": 0.010785305021907651,
63
+ "grad_norm": 2.859375,
64
+ "learning_rate": 0.0003,
65
+ "loss": 4.094369125366211,
66
+ "step": 160
67
+ },
68
+ {
69
+ "epoch": 0.012133468149646108,
70
+ "grad_norm": 2.703125,
71
+ "learning_rate": 0.0003,
72
+ "loss": 3.715929412841797,
73
+ "step": 180
74
+ },
75
+ {
76
+ "epoch": 0.013481631277384564,
77
+ "grad_norm": 2.203125,
78
+ "learning_rate": 0.0003,
79
+ "loss": 3.352855682373047,
80
+ "step": 200
81
+ },
82
+ {
83
+ "epoch": 0.013481631277384564,
84
+ "eval_loss": 3.1672160625457764,
85
+ "eval_runtime": 8.5158,
86
+ "eval_samples_per_second": 1118.739,
87
+ "eval_steps_per_second": 17.497,
88
+ "step": 200
89
+ },
90
+ {
91
+ "epoch": 0.01482979440512302,
92
+ "grad_norm": 2.203125,
93
+ "learning_rate": 0.0003,
94
+ "loss": 2.9979101181030274,
95
+ "step": 220
96
+ },
97
+ {
98
+ "epoch": 0.016177957532861477,
99
+ "grad_norm": 1.9140625,
100
+ "learning_rate": 0.0003,
101
+ "loss": 2.638140106201172,
102
+ "step": 240
103
+ },
104
+ {
105
+ "epoch": 0.01752612066059993,
106
+ "grad_norm": 2.140625,
107
+ "learning_rate": 0.0003,
108
+ "loss": 2.2962114334106447,
109
+ "step": 260
110
+ },
111
+ {
112
+ "epoch": 0.01887428378833839,
113
+ "grad_norm": 1.84375,
114
+ "learning_rate": 0.0003,
115
+ "loss": 1.9714105606079102,
116
+ "step": 280
117
+ },
118
+ {
119
+ "epoch": 0.020222446916076844,
120
+ "grad_norm": 2.0625,
121
+ "learning_rate": 0.0003,
122
+ "loss": 1.6777704238891602,
123
+ "step": 300
124
+ },
125
+ {
126
+ "epoch": 0.021570610043815303,
127
+ "grad_norm": 1.4765625,
128
+ "learning_rate": 0.0003,
129
+ "loss": 1.4184170722961427,
130
+ "step": 320
131
+ },
132
+ {
133
+ "epoch": 0.022918773171553757,
134
+ "grad_norm": 1.4375,
135
+ "learning_rate": 0.0003,
136
+ "loss": 1.2252904891967773,
137
+ "step": 340
138
+ },
139
+ {
140
+ "epoch": 0.024266936299292215,
141
+ "grad_norm": 1.0859375,
142
+ "learning_rate": 0.0003,
143
+ "loss": 1.0854849815368652,
144
+ "step": 360
145
+ },
146
+ {
147
+ "epoch": 0.02561509942703067,
148
+ "grad_norm": 1.109375,
149
+ "learning_rate": 0.0003,
150
+ "loss": 0.9601390838623047,
151
+ "step": 380
152
+ },
153
+ {
154
+ "epoch": 0.026963262554769128,
155
+ "grad_norm": 0.60546875,
156
+ "learning_rate": 0.0003,
157
+ "loss": 0.8285150527954102,
158
+ "step": 400
159
+ },
160
+ {
161
+ "epoch": 0.026963262554769128,
162
+ "eval_loss": 0.77406907081604,
163
+ "eval_runtime": 8.4577,
164
+ "eval_samples_per_second": 1126.423,
165
+ "eval_steps_per_second": 17.617,
166
+ "step": 400
167
+ },
168
+ {
169
+ "epoch": 0.028311425682507583,
170
+ "grad_norm": 0.5234375,
171
+ "learning_rate": 0.0003,
172
+ "loss": 0.7167386531829834,
173
+ "step": 420
174
+ },
175
+ {
176
+ "epoch": 0.02965958881024604,
177
+ "grad_norm": 0.498046875,
178
+ "learning_rate": 0.0003,
179
+ "loss": 0.6465096950531006,
180
+ "step": 440
181
+ },
182
+ {
183
+ "epoch": 0.031007751937984496,
184
+ "grad_norm": 0.51953125,
185
+ "learning_rate": 0.0003,
186
+ "loss": 0.5791397094726562,
187
+ "step": 460
188
+ },
189
+ {
190
+ "epoch": 0.032355915065722954,
191
+ "grad_norm": 0.4375,
192
+ "learning_rate": 0.0003,
193
+ "loss": 0.5380198001861572,
194
+ "step": 480
195
+ },
196
+ {
197
+ "epoch": 0.03370407819346141,
198
+ "grad_norm": 0.447265625,
199
+ "learning_rate": 0.0003,
200
+ "loss": 0.49995784759521483,
201
+ "step": 500
202
+ },
203
+ {
204
+ "epoch": 0.03505224132119986,
205
+ "grad_norm": 0.35546875,
206
+ "learning_rate": 0.0003,
207
+ "loss": 0.47292389869689944,
208
+ "step": 520
209
+ },
210
+ {
211
+ "epoch": 0.03640040444893832,
212
+ "grad_norm": 0.361328125,
213
+ "learning_rate": 0.0003,
214
+ "loss": 0.44700446128845217,
215
+ "step": 540
216
+ },
217
+ {
218
+ "epoch": 0.03774856757667678,
219
+ "grad_norm": 0.408203125,
220
+ "learning_rate": 0.0003,
221
+ "loss": 0.4330583572387695,
222
+ "step": 560
223
+ },
224
+ {
225
+ "epoch": 0.03909673070441524,
226
+ "grad_norm": 0.365234375,
227
+ "learning_rate": 0.0003,
228
+ "loss": 0.4167480945587158,
229
+ "step": 580
230
+ },
231
+ {
232
+ "epoch": 0.04044489383215369,
233
+ "grad_norm": 0.365234375,
234
+ "learning_rate": 0.0003,
235
+ "loss": 0.40725293159484866,
236
+ "step": 600
237
+ },
238
+ {
239
+ "epoch": 0.04044489383215369,
240
+ "eval_loss": 0.40257543325424194,
241
+ "eval_runtime": 8.4955,
242
+ "eval_samples_per_second": 1121.412,
243
+ "eval_steps_per_second": 17.539,
244
+ "step": 600
245
+ },
246
+ {
247
+ "epoch": 0.04179305695989215,
248
+ "grad_norm": 0.380859375,
249
+ "learning_rate": 0.0003,
250
+ "loss": 0.39375302791595457,
251
+ "step": 620
252
+ },
253
+ {
254
+ "epoch": 0.043141220087630605,
255
+ "grad_norm": 2.453125,
256
+ "learning_rate": 0.0003,
257
+ "loss": 0.3863722324371338,
258
+ "step": 640
259
+ },
260
+ {
261
+ "epoch": 0.044489383215369056,
262
+ "grad_norm": 0.9453125,
263
+ "learning_rate": 0.0003,
264
+ "loss": 0.39115495681762696,
265
+ "step": 660
266
+ },
267
+ {
268
+ "epoch": 0.045837546343107514,
269
+ "grad_norm": 0.408203125,
270
+ "learning_rate": 0.0003,
271
+ "loss": 0.3735619783401489,
272
+ "step": 680
273
+ },
274
+ {
275
+ "epoch": 0.04718570947084597,
276
+ "grad_norm": 0.306640625,
277
+ "learning_rate": 0.0003,
278
+ "loss": 0.360651421546936,
279
+ "step": 700
280
+ },
281
+ {
282
+ "epoch": 0.04853387259858443,
283
+ "grad_norm": 0.279296875,
284
+ "learning_rate": 0.0003,
285
+ "loss": 0.3540374279022217,
286
+ "step": 720
287
+ },
288
+ {
289
+ "epoch": 0.04988203572632288,
290
+ "grad_norm": 0.27734375,
291
+ "learning_rate": 0.0003,
292
+ "loss": 0.3440989971160889,
293
+ "step": 740
294
+ },
295
+ {
296
+ "epoch": 0.05123019885406134,
297
+ "grad_norm": 0.2734375,
298
+ "learning_rate": 0.0003,
299
+ "loss": 0.34135589599609373,
300
+ "step": 760
301
+ },
302
+ {
303
+ "epoch": 0.0525783619817998,
304
+ "grad_norm": 0.29296875,
305
+ "learning_rate": 0.0003,
306
+ "loss": 0.3341085910797119,
307
+ "step": 780
308
+ },
309
+ {
310
+ "epoch": 0.053926525109538256,
311
+ "grad_norm": 0.294921875,
312
+ "learning_rate": 0.0003,
313
+ "loss": 0.3262310266494751,
314
+ "step": 800
315
+ },
316
+ {
317
+ "epoch": 0.053926525109538256,
318
+ "eval_loss": 0.3270248770713806,
319
+ "eval_runtime": 8.5141,
320
+ "eval_samples_per_second": 1118.961,
321
+ "eval_steps_per_second": 17.5,
322
+ "step": 800
323
+ },
324
+ {
325
+ "epoch": 0.05527468823727671,
326
+ "grad_norm": 0.32421875,
327
+ "learning_rate": 0.0003,
328
+ "loss": 0.3224670648574829,
329
+ "step": 820
330
+ },
331
+ {
332
+ "epoch": 0.056622851365015166,
333
+ "grad_norm": 0.291015625,
334
+ "learning_rate": 0.0003,
335
+ "loss": 0.31727190017700196,
336
+ "step": 840
337
+ },
338
+ {
339
+ "epoch": 0.057971014492753624,
340
+ "grad_norm": 0.298828125,
341
+ "learning_rate": 0.0003,
342
+ "loss": 0.31037948131561277,
343
+ "step": 860
344
+ },
345
+ {
346
+ "epoch": 0.05931917762049208,
347
+ "grad_norm": 0.283203125,
348
+ "learning_rate": 0.0003,
349
+ "loss": 0.30405306816101074,
350
+ "step": 880
351
+ },
352
+ {
353
+ "epoch": 0.06066734074823053,
354
+ "grad_norm": 0.306640625,
355
+ "learning_rate": 0.0003,
356
+ "loss": 0.30279972553253176,
357
+ "step": 900
358
+ },
359
+ {
360
+ "epoch": 0.06201550387596899,
361
+ "grad_norm": 0.328125,
362
+ "learning_rate": 0.0003,
363
+ "loss": 0.2994309663772583,
364
+ "step": 920
365
+ },
366
+ {
367
+ "epoch": 0.06336366700370745,
368
+ "grad_norm": 0.318359375,
369
+ "learning_rate": 0.0003,
370
+ "loss": 0.2952401638031006,
371
+ "step": 940
372
+ },
373
+ {
374
+ "epoch": 0.06471183013144591,
375
+ "grad_norm": 0.330078125,
376
+ "learning_rate": 0.0003,
377
+ "loss": 0.2905283451080322,
378
+ "step": 960
379
+ },
380
+ {
381
+ "epoch": 0.06605999325918437,
382
+ "grad_norm": 0.365234375,
383
+ "learning_rate": 0.0003,
384
+ "loss": 0.28865833282470704,
385
+ "step": 980
386
+ },
387
+ {
388
+ "epoch": 0.06740815638692282,
389
+ "grad_norm": 0.3046875,
390
+ "learning_rate": 0.0003,
391
+ "loss": 0.2860438585281372,
392
+ "step": 1000
393
+ },
394
+ {
395
+ "epoch": 0.06740815638692282,
396
+ "eval_loss": 0.28539252281188965,
397
+ "eval_runtime": 8.5689,
398
+ "eval_samples_per_second": 1111.809,
399
+ "eval_steps_per_second": 17.388,
400
+ "step": 1000
401
+ },
402
+ {
403
+ "epoch": 0.06875631951466127,
404
+ "grad_norm": 0.333984375,
405
+ "learning_rate": 0.0003,
406
+ "loss": 0.2810789108276367,
407
+ "step": 1020
408
+ },
409
+ {
410
+ "epoch": 0.07010448264239973,
411
+ "grad_norm": 0.28125,
412
+ "learning_rate": 0.0003,
413
+ "loss": 0.2767570734024048,
414
+ "step": 1040
415
+ },
416
+ {
417
+ "epoch": 0.07145264577013818,
418
+ "grad_norm": 0.333984375,
419
+ "learning_rate": 0.0003,
420
+ "loss": 0.2760180950164795,
421
+ "step": 1060
422
+ },
423
+ {
424
+ "epoch": 0.07280080889787664,
425
+ "grad_norm": 0.306640625,
426
+ "learning_rate": 0.0003,
427
+ "loss": 0.2714431047439575,
428
+ "step": 1080
429
+ },
430
+ {
431
+ "epoch": 0.0741489720256151,
432
+ "grad_norm": 0.32421875,
433
+ "learning_rate": 0.0003,
434
+ "loss": 0.2704152584075928,
435
+ "step": 1100
436
+ },
437
+ {
438
+ "epoch": 0.07549713515335356,
439
+ "grad_norm": 0.37890625,
440
+ "learning_rate": 0.0003,
441
+ "loss": 0.26700942516326903,
442
+ "step": 1120
443
+ },
444
+ {
445
+ "epoch": 0.07684529828109202,
446
+ "grad_norm": 1.140625,
447
+ "learning_rate": 0.0003,
448
+ "loss": 0.29485843181610105,
449
+ "step": 1140
450
+ },
451
+ {
452
+ "epoch": 0.07819346140883048,
453
+ "grad_norm": 0.400390625,
454
+ "learning_rate": 0.0003,
455
+ "loss": 0.27065248489379884,
456
+ "step": 1160
457
+ },
458
+ {
459
+ "epoch": 0.07954162453656892,
460
+ "grad_norm": 0.333984375,
461
+ "learning_rate": 0.0003,
462
+ "loss": 0.26451032161712645,
463
+ "step": 1180
464
+ },
465
+ {
466
+ "epoch": 0.08088978766430738,
467
+ "grad_norm": 0.267578125,
468
+ "learning_rate": 0.0003,
469
+ "loss": 0.26219322681427004,
470
+ "step": 1200
471
+ },
472
+ {
473
+ "epoch": 0.08088978766430738,
474
+ "eval_loss": 0.2612524628639221,
475
+ "eval_runtime": 8.6079,
476
+ "eval_samples_per_second": 1106.779,
477
+ "eval_steps_per_second": 17.31,
478
+ "step": 1200
479
+ },
480
+ {
481
+ "epoch": 0.08223795079204584,
482
+ "grad_norm": 0.25,
483
+ "learning_rate": 0.0003,
484
+ "loss": 0.2575028896331787,
485
+ "step": 1220
486
+ },
487
+ {
488
+ "epoch": 0.0835861139197843,
489
+ "grad_norm": 0.248046875,
490
+ "learning_rate": 0.0003,
491
+ "loss": 0.25518064498901366,
492
+ "step": 1240
493
+ },
494
+ {
495
+ "epoch": 0.08493427704752275,
496
+ "grad_norm": 0.2490234375,
497
+ "learning_rate": 0.0003,
498
+ "loss": 0.2539027452468872,
499
+ "step": 1260
500
+ },
501
+ {
502
+ "epoch": 0.08628244017526121,
503
+ "grad_norm": 0.25,
504
+ "learning_rate": 0.0003,
505
+ "loss": 0.24920670986175536,
506
+ "step": 1280
507
+ },
508
+ {
509
+ "epoch": 0.08763060330299967,
510
+ "grad_norm": 0.2470703125,
511
+ "learning_rate": 0.0003,
512
+ "loss": 0.24917314052581788,
513
+ "step": 1300
514
+ },
515
+ {
516
+ "epoch": 0.08897876643073811,
517
+ "grad_norm": 0.240234375,
518
+ "learning_rate": 0.0003,
519
+ "loss": 0.24632787704467773,
520
+ "step": 1320
521
+ },
522
+ {
523
+ "epoch": 0.09032692955847657,
524
+ "grad_norm": 0.25,
525
+ "learning_rate": 0.0003,
526
+ "loss": 0.24541354179382324,
527
+ "step": 1340
528
+ },
529
+ {
530
+ "epoch": 0.09167509268621503,
531
+ "grad_norm": 0.2490234375,
532
+ "learning_rate": 0.0003,
533
+ "loss": 0.2430295467376709,
534
+ "step": 1360
535
+ },
536
+ {
537
+ "epoch": 0.09302325581395349,
538
+ "grad_norm": 0.26171875,
539
+ "learning_rate": 0.0003,
540
+ "loss": 0.24261221885681153,
541
+ "step": 1380
542
+ },
543
+ {
544
+ "epoch": 0.09437141894169195,
545
+ "grad_norm": 0.28515625,
546
+ "learning_rate": 0.0003,
547
+ "loss": 0.24039824008941652,
548
+ "step": 1400
549
+ },
550
+ {
551
+ "epoch": 0.09437141894169195,
552
+ "eval_loss": 0.24237078428268433,
553
+ "eval_runtime": 8.6343,
554
+ "eval_samples_per_second": 1103.385,
555
+ "eval_steps_per_second": 17.257,
556
+ "step": 1400
557
+ },
558
+ {
559
+ "epoch": 0.0957195820694304,
560
+ "grad_norm": 0.259765625,
561
+ "learning_rate": 0.0003,
562
+ "loss": 0.24012534618377684,
563
+ "step": 1420
564
+ },
565
+ {
566
+ "epoch": 0.09706774519716886,
567
+ "grad_norm": 0.248046875,
568
+ "learning_rate": 0.0003,
569
+ "loss": 0.2381824731826782,
570
+ "step": 1440
571
+ },
572
+ {
573
+ "epoch": 0.09841590832490732,
574
+ "grad_norm": 0.251953125,
575
+ "learning_rate": 0.0003,
576
+ "loss": 0.23610148429870606,
577
+ "step": 1460
578
+ },
579
+ {
580
+ "epoch": 0.09976407145264576,
581
+ "grad_norm": 0.314453125,
582
+ "learning_rate": 0.0003,
583
+ "loss": 0.2342632532119751,
584
+ "step": 1480
585
+ },
586
+ {
587
+ "epoch": 0.10111223458038422,
588
+ "grad_norm": 0.2451171875,
589
+ "learning_rate": 0.0003,
590
+ "loss": 0.23345742225646973,
591
+ "step": 1500
592
+ }
593
+ ],
594
+ "logging_steps": 20,
595
+ "max_steps": 2000,
596
+ "num_input_tokens_seen": 0,
597
+ "num_train_epochs": 1,
598
+ "save_steps": 500,
599
+ "stateful_callbacks": {
600
+ "TrainerControl": {
601
+ "args": {
602
+ "should_epoch_stop": false,
603
+ "should_evaluate": false,
604
+ "should_log": false,
605
+ "should_save": true,
606
+ "should_training_stop": false
607
+ },
608
+ "attributes": {}
609
+ }
610
+ },
611
+ "total_flos": 435582664704000.0,
612
+ "train_batch_size": 64,
613
+ "trial_name": null,
614
+ "trial_params": null
615
+ }
out/glu-situglu-9L_run/checkpoint-1500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f2d9b411968661617d32794fc4981dd80c49ba782deaa6cc247aefafa8768d2e
3
+ size 4920
out/glu-situglu-9L_run/checkpoint-500/config.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation": "situglu",
3
+ "architectures": [
4
+ "TinyLlamaForCausalLM"
5
+ ],
6
+ "attention_bias": false,
7
+ "attention_dropout": 0.0,
8
+ "bos_token_id": 1,
9
+ "dtype": "bfloat16",
10
+ "eos_token_id": 2,
11
+ "head_dim": 32,
12
+ "hidden_act": "silu",
13
+ "hidden_size": 128,
14
+ "initializer_range": 0.02,
15
+ "intermediate_size": 256,
16
+ "max_position_embeddings": 512,
17
+ "mlp_bias": false,
18
+ "mlp_type": "glu",
19
+ "model_type": "tiny_llama",
20
+ "num_attention_heads": 4,
21
+ "num_hidden_layers": 9,
22
+ "num_key_value_heads": 4,
23
+ "pad_token_id": 0,
24
+ "pretraining_tp": 1,
25
+ "rms_norm_eps": 1e-06,
26
+ "rope_parameters": {
27
+ "rope_theta": 10000.0,
28
+ "rope_type": "default"
29
+ },
30
+ "tie_word_embeddings": true,
31
+ "tokenizer_name": "w-ahmad/tiny-stories-tokenizer",
32
+ "transformers_version": "5.15.0.dev0",
33
+ "use_cache": false,
34
+ "vocab_size": 4096
35
+ }
out/glu-situglu-9L_run/checkpoint-500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d6ad1b2f9292270cbe02adffb215fb9bd003529264527bd5ce08b5c8ee8ff1c3
3
+ size 4011552
out/glu-situglu-9L_run/checkpoint-500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5fdb70a1e14ba341da1780bcb83dbf02cbbb70494c93963ab1065f7d8107a116
3
+ size 8074746
out/glu-situglu-9L_run/checkpoint-500/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3cf9097d4513154245c48236b6ec5137b7ee2a21c9f58f2cba798ea275c6026f
3
+ size 14244
out/glu-situglu-9L_run/checkpoint-500/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e328dcb5672eefa48692126747c226aede5ae2e575011f8f5a2b20326a61f850
3
+ size 1064
out/glu-situglu-9L_run/checkpoint-500/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
out/glu-situglu-9L_run/checkpoint-500/tokenizer_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_prefix_space": false,
3
+ "backend": "tokenizers",
4
+ "bos_token": "<|endoftext|>",
5
+ "eos_token": "<|endoftext|>",
6
+ "errors": "replace",
7
+ "is_local": false,
8
+ "local_files_only": false,
9
+ "model_max_length": 1024,
10
+ "pad_token": "<|endoftext|>",
11
+ "tokenizer_class": "GPT2Tokenizer",
12
+ "unk_token": "<|endoftext|>"
13
+ }
out/glu-situglu-9L_run/checkpoint-500/trainer_state.json ADDED
@@ -0,0 +1,225 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 0.03370407819346141,
6
+ "eval_steps": 200,
7
+ "global_step": 500,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.0013481631277384564,
14
+ "grad_norm": 1.2578125,
15
+ "learning_rate": 0.0003,
16
+ "loss": 7.793804931640625,
17
+ "step": 20
18
+ },
19
+ {
20
+ "epoch": 0.002696326255476913,
21
+ "grad_norm": 1.171875,
22
+ "learning_rate": 0.0003,
23
+ "loss": 7.022835540771484,
24
+ "step": 40
25
+ },
26
+ {
27
+ "epoch": 0.004044489383215369,
28
+ "grad_norm": 1.0,
29
+ "learning_rate": 0.0003,
30
+ "loss": 6.450706481933594,
31
+ "step": 60
32
+ },
33
+ {
34
+ "epoch": 0.005392652510953826,
35
+ "grad_norm": 1.9296875,
36
+ "learning_rate": 0.0003,
37
+ "loss": 6.003890609741211,
38
+ "step": 80
39
+ },
40
+ {
41
+ "epoch": 0.006740815638692282,
42
+ "grad_norm": 3.53125,
43
+ "learning_rate": 0.0003,
44
+ "loss": 5.4485828399658205,
45
+ "step": 100
46
+ },
47
+ {
48
+ "epoch": 0.008088978766430738,
49
+ "grad_norm": 2.9375,
50
+ "learning_rate": 0.0003,
51
+ "loss": 4.944174575805664,
52
+ "step": 120
53
+ },
54
+ {
55
+ "epoch": 0.009437141894169195,
56
+ "grad_norm": 3.3125,
57
+ "learning_rate": 0.0003,
58
+ "loss": 4.507948684692383,
59
+ "step": 140
60
+ },
61
+ {
62
+ "epoch": 0.010785305021907651,
63
+ "grad_norm": 2.859375,
64
+ "learning_rate": 0.0003,
65
+ "loss": 4.094369125366211,
66
+ "step": 160
67
+ },
68
+ {
69
+ "epoch": 0.012133468149646108,
70
+ "grad_norm": 2.703125,
71
+ "learning_rate": 0.0003,
72
+ "loss": 3.715929412841797,
73
+ "step": 180
74
+ },
75
+ {
76
+ "epoch": 0.013481631277384564,
77
+ "grad_norm": 2.203125,
78
+ "learning_rate": 0.0003,
79
+ "loss": 3.352855682373047,
80
+ "step": 200
81
+ },
82
+ {
83
+ "epoch": 0.013481631277384564,
84
+ "eval_loss": 3.1672160625457764,
85
+ "eval_runtime": 8.5158,
86
+ "eval_samples_per_second": 1118.739,
87
+ "eval_steps_per_second": 17.497,
88
+ "step": 200
89
+ },
90
+ {
91
+ "epoch": 0.01482979440512302,
92
+ "grad_norm": 2.203125,
93
+ "learning_rate": 0.0003,
94
+ "loss": 2.9979101181030274,
95
+ "step": 220
96
+ },
97
+ {
98
+ "epoch": 0.016177957532861477,
99
+ "grad_norm": 1.9140625,
100
+ "learning_rate": 0.0003,
101
+ "loss": 2.638140106201172,
102
+ "step": 240
103
+ },
104
+ {
105
+ "epoch": 0.01752612066059993,
106
+ "grad_norm": 2.140625,
107
+ "learning_rate": 0.0003,
108
+ "loss": 2.2962114334106447,
109
+ "step": 260
110
+ },
111
+ {
112
+ "epoch": 0.01887428378833839,
113
+ "grad_norm": 1.84375,
114
+ "learning_rate": 0.0003,
115
+ "loss": 1.9714105606079102,
116
+ "step": 280
117
+ },
118
+ {
119
+ "epoch": 0.020222446916076844,
120
+ "grad_norm": 2.0625,
121
+ "learning_rate": 0.0003,
122
+ "loss": 1.6777704238891602,
123
+ "step": 300
124
+ },
125
+ {
126
+ "epoch": 0.021570610043815303,
127
+ "grad_norm": 1.4765625,
128
+ "learning_rate": 0.0003,
129
+ "loss": 1.4184170722961427,
130
+ "step": 320
131
+ },
132
+ {
133
+ "epoch": 0.022918773171553757,
134
+ "grad_norm": 1.4375,
135
+ "learning_rate": 0.0003,
136
+ "loss": 1.2252904891967773,
137
+ "step": 340
138
+ },
139
+ {
140
+ "epoch": 0.024266936299292215,
141
+ "grad_norm": 1.0859375,
142
+ "learning_rate": 0.0003,
143
+ "loss": 1.0854849815368652,
144
+ "step": 360
145
+ },
146
+ {
147
+ "epoch": 0.02561509942703067,
148
+ "grad_norm": 1.109375,
149
+ "learning_rate": 0.0003,
150
+ "loss": 0.9601390838623047,
151
+ "step": 380
152
+ },
153
+ {
154
+ "epoch": 0.026963262554769128,
155
+ "grad_norm": 0.60546875,
156
+ "learning_rate": 0.0003,
157
+ "loss": 0.8285150527954102,
158
+ "step": 400
159
+ },
160
+ {
161
+ "epoch": 0.026963262554769128,
162
+ "eval_loss": 0.77406907081604,
163
+ "eval_runtime": 8.4577,
164
+ "eval_samples_per_second": 1126.423,
165
+ "eval_steps_per_second": 17.617,
166
+ "step": 400
167
+ },
168
+ {
169
+ "epoch": 0.028311425682507583,
170
+ "grad_norm": 0.5234375,
171
+ "learning_rate": 0.0003,
172
+ "loss": 0.7167386531829834,
173
+ "step": 420
174
+ },
175
+ {
176
+ "epoch": 0.02965958881024604,
177
+ "grad_norm": 0.498046875,
178
+ "learning_rate": 0.0003,
179
+ "loss": 0.6465096950531006,
180
+ "step": 440
181
+ },
182
+ {
183
+ "epoch": 0.031007751937984496,
184
+ "grad_norm": 0.51953125,
185
+ "learning_rate": 0.0003,
186
+ "loss": 0.5791397094726562,
187
+ "step": 460
188
+ },
189
+ {
190
+ "epoch": 0.032355915065722954,
191
+ "grad_norm": 0.4375,
192
+ "learning_rate": 0.0003,
193
+ "loss": 0.5380198001861572,
194
+ "step": 480
195
+ },
196
+ {
197
+ "epoch": 0.03370407819346141,
198
+ "grad_norm": 0.447265625,
199
+ "learning_rate": 0.0003,
200
+ "loss": 0.49995784759521483,
201
+ "step": 500
202
+ }
203
+ ],
204
+ "logging_steps": 20,
205
+ "max_steps": 2000,
206
+ "num_input_tokens_seen": 0,
207
+ "num_train_epochs": 1,
208
+ "save_steps": 500,
209
+ "stateful_callbacks": {
210
+ "TrainerControl": {
211
+ "args": {
212
+ "should_epoch_stop": false,
213
+ "should_evaluate": false,
214
+ "should_log": false,
215
+ "should_save": true,
216
+ "should_training_stop": false
217
+ },
218
+ "attributes": {}
219
+ }
220
+ },
221
+ "total_flos": 145194221568000.0,
222
+ "train_batch_size": 64,
223
+ "trial_name": null,
224
+ "trial_params": null
225
+ }
out/glu-situglu-9L_run/checkpoint-500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f2d9b411968661617d32794fc4981dd80c49ba782deaa6cc247aefafa8768d2e
3
+ size 4920