w-ahmad commited on
Commit
a016ea2
·
verified ·
1 Parent(s): bd23566

Auto upload 2026-08-06T21:56:15.080927

Browse files
Files changed (36) hide show
  1. out/glu-gelu-46L_run/README.md +65 -0
  2. out/glu-gelu-46L_run/checkpoint-1500/config.json +35 -0
  3. out/glu-gelu-46L_run/checkpoint-1500/model.safetensors +3 -0
  4. out/glu-gelu-46L_run/checkpoint-1500/optimizer.pt +3 -0
  5. out/glu-gelu-46L_run/checkpoint-1500/rng_state.pth +3 -0
  6. out/glu-gelu-46L_run/checkpoint-1500/scheduler.pt +3 -0
  7. out/glu-gelu-46L_run/checkpoint-1500/tokenizer.json +0 -0
  8. out/glu-gelu-46L_run/checkpoint-1500/tokenizer_config.json +13 -0
  9. out/glu-gelu-46L_run/checkpoint-1500/trainer_state.json +615 -0
  10. out/glu-gelu-46L_run/checkpoint-1500/training_args.bin +3 -0
  11. out/glu-gelu-46L_run/checkpoint-2000/config.json +35 -0
  12. out/glu-gelu-46L_run/checkpoint-2000/model.safetensors +3 -0
  13. out/glu-gelu-46L_run/checkpoint-2000/optimizer.pt +3 -0
  14. out/glu-gelu-46L_run/checkpoint-2000/rng_state.pth +3 -0
  15. out/glu-gelu-46L_run/checkpoint-2000/scheduler.pt +3 -0
  16. out/glu-gelu-46L_run/checkpoint-2000/tokenizer.json +0 -0
  17. out/glu-gelu-46L_run/checkpoint-2000/tokenizer_config.json +13 -0
  18. out/glu-gelu-46L_run/checkpoint-2000/trainer_state.json +814 -0
  19. out/glu-gelu-46L_run/checkpoint-2000/training_args.bin +3 -0
  20. out/glu-gelu-46L_run/model.safetensors +1 -1
  21. out/glu-gelu-46L_run/tokenizer.json +0 -0
  22. out/glu-gelu-46L_run/tokenizer_config.json +13 -0
  23. out/glu-gelu-46L_run/training_log.jsonl +0 -0
  24. out/glu-linear-46L_run/checkpoint-500/config.json +35 -0
  25. out/glu-linear-46L_run/checkpoint-500/model.safetensors +3 -0
  26. out/glu-linear-46L_run/checkpoint-500/optimizer.pt +3 -0
  27. out/glu-linear-46L_run/checkpoint-500/rng_state.pth +3 -0
  28. out/glu-linear-46L_run/checkpoint-500/scheduler.pt +3 -0
  29. out/glu-linear-46L_run/checkpoint-500/tokenizer.json +0 -0
  30. out/glu-linear-46L_run/checkpoint-500/tokenizer_config.json +13 -0
  31. out/glu-linear-46L_run/checkpoint-500/trainer_state.json +225 -0
  32. out/glu-linear-46L_run/checkpoint-500/training_args.bin +3 -0
  33. out/glu-linear-46L_run/config.json +35 -0
  34. out/glu-linear-46L_run/model.safetensors +3 -0
  35. out/glu-linear-46L_run/training_args.bin +3 -0
  36. out/glu-linear-46L_run/training_log.jsonl +0 -0
out/glu-gelu-46L_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-gelu-46L
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-gelu-46L
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: 0.2512
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
+ | 4.0532 | 0.0135 | 200 | 3.8488 |
49
+ | 1.4764 | 0.0270 | 400 | 1.3991 |
50
+ | 0.5977 | 0.0404 | 600 | 0.5825 |
51
+ | 0.4267 | 0.0539 | 800 | 0.4185 |
52
+ | 0.3458 | 0.0674 | 1000 | 0.3437 |
53
+ | 0.3267 | 0.0809 | 1200 | 0.3165 |
54
+ | 0.2840 | 0.0944 | 1400 | 0.2859 |
55
+ | 0.3036 | 0.1079 | 1600 | 0.2998 |
56
+ | 0.2605 | 0.1213 | 1800 | 0.2594 |
57
+ | 0.2497 | 0.1348 | 2000 | 0.2512 |
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-gelu-46L_run/checkpoint-1500/config.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation": "gelu",
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": 46,
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-gelu-46L_run/checkpoint-1500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:199d953c4664614fc1c98a9fc594da5e71475d152d890f864dc14cda9d9e6317
3
+ size 16191120
out/glu-gelu-46L_run/checkpoint-1500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7fc54414d453a8d25efa6b8e82f448ba8cd32150eb53615028de5cad6176b46e
3
+ size 32639090
out/glu-gelu-46L_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-gelu-46L_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-gelu-46L_run/checkpoint-1500/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
out/glu-gelu-46L_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-gelu-46L_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": 7.776710510253906,
17
+ "step": 20
18
+ },
19
+ {
20
+ "epoch": 0.002696326255476913,
21
+ "grad_norm": 1.171875,
22
+ "learning_rate": 0.0003,
23
+ "loss": 7.017825317382813,
24
+ "step": 40
25
+ },
26
+ {
27
+ "epoch": 0.004044489383215369,
28
+ "grad_norm": 0.9296875,
29
+ "learning_rate": 0.0003,
30
+ "loss": 6.486733245849609,
31
+ "step": 60
32
+ },
33
+ {
34
+ "epoch": 0.005392652510953826,
35
+ "grad_norm": 0.6875,
36
+ "learning_rate": 0.0003,
37
+ "loss": 6.160807418823242,
38
+ "step": 80
39
+ },
40
+ {
41
+ "epoch": 0.006740815638692282,
42
+ "grad_norm": 0.640625,
43
+ "learning_rate": 0.0003,
44
+ "loss": 5.966274261474609,
45
+ "step": 100
46
+ },
47
+ {
48
+ "epoch": 0.008088978766430738,
49
+ "grad_norm": 0.9765625,
50
+ "learning_rate": 0.0003,
51
+ "loss": 5.7247802734375,
52
+ "step": 120
53
+ },
54
+ {
55
+ "epoch": 0.009437141894169195,
56
+ "grad_norm": 1.046875,
57
+ "learning_rate": 0.0003,
58
+ "loss": 5.335699462890625,
59
+ "step": 140
60
+ },
61
+ {
62
+ "epoch": 0.010785305021907651,
63
+ "grad_norm": 1.0234375,
64
+ "learning_rate": 0.0003,
65
+ "loss": 4.942352294921875,
66
+ "step": 160
67
+ },
68
+ {
69
+ "epoch": 0.012133468149646108,
70
+ "grad_norm": 1.15625,
71
+ "learning_rate": 0.0003,
72
+ "loss": 4.513675308227539,
73
+ "step": 180
74
+ },
75
+ {
76
+ "epoch": 0.013481631277384564,
77
+ "grad_norm": 1.2421875,
78
+ "learning_rate": 0.0003,
79
+ "loss": 4.05323371887207,
80
+ "step": 200
81
+ },
82
+ {
83
+ "epoch": 0.013481631277384564,
84
+ "eval_loss": 3.848839521408081,
85
+ "eval_runtime": 12.1844,
86
+ "eval_samples_per_second": 781.901,
87
+ "eval_steps_per_second": 12.229,
88
+ "step": 200
89
+ },
90
+ {
91
+ "epoch": 0.01482979440512302,
92
+ "grad_norm": 1.515625,
93
+ "learning_rate": 0.0003,
94
+ "loss": 3.664620208740234,
95
+ "step": 220
96
+ },
97
+ {
98
+ "epoch": 0.016177957532861477,
99
+ "grad_norm": 1.265625,
100
+ "learning_rate": 0.0003,
101
+ "loss": 3.32559814453125,
102
+ "step": 240
103
+ },
104
+ {
105
+ "epoch": 0.01752612066059993,
106
+ "grad_norm": 1.8671875,
107
+ "learning_rate": 0.0003,
108
+ "loss": 3.0320526123046876,
109
+ "step": 260
110
+ },
111
+ {
112
+ "epoch": 0.01887428378833839,
113
+ "grad_norm": 1.984375,
114
+ "learning_rate": 0.0003,
115
+ "loss": 2.754303741455078,
116
+ "step": 280
117
+ },
118
+ {
119
+ "epoch": 0.020222446916076844,
120
+ "grad_norm": 2.171875,
121
+ "learning_rate": 0.0003,
122
+ "loss": 2.4877477645874024,
123
+ "step": 300
124
+ },
125
+ {
126
+ "epoch": 0.021570610043815303,
127
+ "grad_norm": 1.640625,
128
+ "learning_rate": 0.0003,
129
+ "loss": 2.2490768432617188,
130
+ "step": 320
131
+ },
132
+ {
133
+ "epoch": 0.022918773171553757,
134
+ "grad_norm": 1.59375,
135
+ "learning_rate": 0.0003,
136
+ "loss": 2.0304283142089843,
137
+ "step": 340
138
+ },
139
+ {
140
+ "epoch": 0.024266936299292215,
141
+ "grad_norm": 1.484375,
142
+ "learning_rate": 0.0003,
143
+ "loss": 1.8307512283325196,
144
+ "step": 360
145
+ },
146
+ {
147
+ "epoch": 0.02561509942703067,
148
+ "grad_norm": 1.78125,
149
+ "learning_rate": 0.0003,
150
+ "loss": 1.6494441986083985,
151
+ "step": 380
152
+ },
153
+ {
154
+ "epoch": 0.026963262554769128,
155
+ "grad_norm": 1.8203125,
156
+ "learning_rate": 0.0003,
157
+ "loss": 1.4764435768127442,
158
+ "step": 400
159
+ },
160
+ {
161
+ "epoch": 0.026963262554769128,
162
+ "eval_loss": 1.3990651369094849,
163
+ "eval_runtime": 12.2447,
164
+ "eval_samples_per_second": 778.051,
165
+ "eval_steps_per_second": 12.169,
166
+ "step": 400
167
+ },
168
+ {
169
+ "epoch": 0.028311425682507583,
170
+ "grad_norm": 1.5625,
171
+ "learning_rate": 0.0003,
172
+ "loss": 1.3182246208190918,
173
+ "step": 420
174
+ },
175
+ {
176
+ "epoch": 0.02965958881024604,
177
+ "grad_norm": 1.0625,
178
+ "learning_rate": 0.0003,
179
+ "loss": 1.1943863868713378,
180
+ "step": 440
181
+ },
182
+ {
183
+ "epoch": 0.031007751937984496,
184
+ "grad_norm": 1.4609375,
185
+ "learning_rate": 0.0003,
186
+ "loss": 1.0682686805725097,
187
+ "step": 460
188
+ },
189
+ {
190
+ "epoch": 0.032355915065722954,
191
+ "grad_norm": 1.1875,
192
+ "learning_rate": 0.0003,
193
+ "loss": 0.9639272689819336,
194
+ "step": 480
195
+ },
196
+ {
197
+ "epoch": 0.03370407819346141,
198
+ "grad_norm": 0.69921875,
199
+ "learning_rate": 0.0003,
200
+ "loss": 0.8594159126281739,
201
+ "step": 500
202
+ },
203
+ {
204
+ "epoch": 0.03505224132119986,
205
+ "grad_norm": 0.4140625,
206
+ "learning_rate": 0.0003,
207
+ "loss": 0.7737322807312011,
208
+ "step": 520
209
+ },
210
+ {
211
+ "epoch": 0.03640040444893832,
212
+ "grad_norm": 0.419921875,
213
+ "learning_rate": 0.0003,
214
+ "loss": 0.6988646507263183,
215
+ "step": 540
216
+ },
217
+ {
218
+ "epoch": 0.03774856757667678,
219
+ "grad_norm": 0.40234375,
220
+ "learning_rate": 0.0003,
221
+ "loss": 0.6476529598236084,
222
+ "step": 560
223
+ },
224
+ {
225
+ "epoch": 0.03909673070441524,
226
+ "grad_norm": 1.7421875,
227
+ "learning_rate": 0.0003,
228
+ "loss": 0.6122144222259521,
229
+ "step": 580
230
+ },
231
+ {
232
+ "epoch": 0.04044489383215369,
233
+ "grad_norm": 1.3046875,
234
+ "learning_rate": 0.0003,
235
+ "loss": 0.5976991176605224,
236
+ "step": 600
237
+ },
238
+ {
239
+ "epoch": 0.04044489383215369,
240
+ "eval_loss": 0.5825157165527344,
241
+ "eval_runtime": 12.2182,
242
+ "eval_samples_per_second": 779.74,
243
+ "eval_steps_per_second": 12.195,
244
+ "step": 600
245
+ },
246
+ {
247
+ "epoch": 0.04179305695989215,
248
+ "grad_norm": 0.40234375,
249
+ "learning_rate": 0.0003,
250
+ "loss": 0.5630900382995605,
251
+ "step": 620
252
+ },
253
+ {
254
+ "epoch": 0.043141220087630605,
255
+ "grad_norm": 0.3359375,
256
+ "learning_rate": 0.0003,
257
+ "loss": 0.5361714363098145,
258
+ "step": 640
259
+ },
260
+ {
261
+ "epoch": 0.044489383215369056,
262
+ "grad_norm": 0.3125,
263
+ "learning_rate": 0.0003,
264
+ "loss": 0.5064755439758301,
265
+ "step": 660
266
+ },
267
+ {
268
+ "epoch": 0.045837546343107514,
269
+ "grad_norm": 0.322265625,
270
+ "learning_rate": 0.0003,
271
+ "loss": 0.4883098602294922,
272
+ "step": 680
273
+ },
274
+ {
275
+ "epoch": 0.04718570947084597,
276
+ "grad_norm": 0.3125,
277
+ "learning_rate": 0.0003,
278
+ "loss": 0.4687828540802002,
279
+ "step": 700
280
+ },
281
+ {
282
+ "epoch": 0.04853387259858443,
283
+ "grad_norm": 0.333984375,
284
+ "learning_rate": 0.0003,
285
+ "loss": 0.4550455093383789,
286
+ "step": 720
287
+ },
288
+ {
289
+ "epoch": 0.04988203572632288,
290
+ "grad_norm": 0.3515625,
291
+ "learning_rate": 0.0003,
292
+ "loss": 0.4373329639434814,
293
+ "step": 740
294
+ },
295
+ {
296
+ "epoch": 0.05123019885406134,
297
+ "grad_norm": 0.365234375,
298
+ "learning_rate": 0.0003,
299
+ "loss": 0.42957515716552735,
300
+ "step": 760
301
+ },
302
+ {
303
+ "epoch": 0.0525783619817998,
304
+ "grad_norm": 1.9296875,
305
+ "learning_rate": 0.0003,
306
+ "loss": 0.45300583839416503,
307
+ "step": 780
308
+ },
309
+ {
310
+ "epoch": 0.053926525109538256,
311
+ "grad_norm": 0.5390625,
312
+ "learning_rate": 0.0003,
313
+ "loss": 0.42671890258789064,
314
+ "step": 800
315
+ },
316
+ {
317
+ "epoch": 0.053926525109538256,
318
+ "eval_loss": 0.41847455501556396,
319
+ "eval_runtime": 12.2303,
320
+ "eval_samples_per_second": 778.967,
321
+ "eval_steps_per_second": 12.183,
322
+ "step": 800
323
+ },
324
+ {
325
+ "epoch": 0.05527468823727671,
326
+ "grad_norm": 0.34765625,
327
+ "learning_rate": 0.0003,
328
+ "loss": 0.4097603797912598,
329
+ "step": 820
330
+ },
331
+ {
332
+ "epoch": 0.056622851365015166,
333
+ "grad_norm": 0.291015625,
334
+ "learning_rate": 0.0003,
335
+ "loss": 0.39820222854614257,
336
+ "step": 840
337
+ },
338
+ {
339
+ "epoch": 0.057971014492753624,
340
+ "grad_norm": 0.27734375,
341
+ "learning_rate": 0.0003,
342
+ "loss": 0.3863217830657959,
343
+ "step": 860
344
+ },
345
+ {
346
+ "epoch": 0.05931917762049208,
347
+ "grad_norm": 0.27734375,
348
+ "learning_rate": 0.0003,
349
+ "loss": 0.3764716386795044,
350
+ "step": 880
351
+ },
352
+ {
353
+ "epoch": 0.06066734074823053,
354
+ "grad_norm": 0.275390625,
355
+ "learning_rate": 0.0003,
356
+ "loss": 0.373111891746521,
357
+ "step": 900
358
+ },
359
+ {
360
+ "epoch": 0.06201550387596899,
361
+ "grad_norm": 0.283203125,
362
+ "learning_rate": 0.0003,
363
+ "loss": 0.36699652671813965,
364
+ "step": 920
365
+ },
366
+ {
367
+ "epoch": 0.06336366700370745,
368
+ "grad_norm": 0.279296875,
369
+ "learning_rate": 0.0003,
370
+ "loss": 0.35993266105651855,
371
+ "step": 940
372
+ },
373
+ {
374
+ "epoch": 0.06471183013144591,
375
+ "grad_norm": 0.28125,
376
+ "learning_rate": 0.0003,
377
+ "loss": 0.3524770736694336,
378
+ "step": 960
379
+ },
380
+ {
381
+ "epoch": 0.06605999325918437,
382
+ "grad_norm": 0.271484375,
383
+ "learning_rate": 0.0003,
384
+ "loss": 0.34836206436157224,
385
+ "step": 980
386
+ },
387
+ {
388
+ "epoch": 0.06740815638692282,
389
+ "grad_norm": 0.3046875,
390
+ "learning_rate": 0.0003,
391
+ "loss": 0.34576056003570554,
392
+ "step": 1000
393
+ },
394
+ {
395
+ "epoch": 0.06740815638692282,
396
+ "eval_loss": 0.34370264410972595,
397
+ "eval_runtime": 12.2146,
398
+ "eval_samples_per_second": 779.966,
399
+ "eval_steps_per_second": 12.198,
400
+ "step": 1000
401
+ },
402
+ {
403
+ "epoch": 0.06875631951466127,
404
+ "grad_norm": 0.283203125,
405
+ "learning_rate": 0.0003,
406
+ "loss": 0.3385118007659912,
407
+ "step": 1020
408
+ },
409
+ {
410
+ "epoch": 0.07010448264239973,
411
+ "grad_norm": 0.263671875,
412
+ "learning_rate": 0.0003,
413
+ "loss": 0.33219237327575685,
414
+ "step": 1040
415
+ },
416
+ {
417
+ "epoch": 0.07145264577013818,
418
+ "grad_norm": 0.271484375,
419
+ "learning_rate": 0.0003,
420
+ "loss": 0.3300537109375,
421
+ "step": 1060
422
+ },
423
+ {
424
+ "epoch": 0.07280080889787664,
425
+ "grad_norm": 0.283203125,
426
+ "learning_rate": 0.0003,
427
+ "loss": 0.32398457527160646,
428
+ "step": 1080
429
+ },
430
+ {
431
+ "epoch": 0.0741489720256151,
432
+ "grad_norm": 0.287109375,
433
+ "learning_rate": 0.0003,
434
+ "loss": 0.32252542972564696,
435
+ "step": 1100
436
+ },
437
+ {
438
+ "epoch": 0.07549713515335356,
439
+ "grad_norm": 0.310546875,
440
+ "learning_rate": 0.0003,
441
+ "loss": 0.31789841651916506,
442
+ "step": 1120
443
+ },
444
+ {
445
+ "epoch": 0.07684529828109202,
446
+ "grad_norm": 0.3125,
447
+ "learning_rate": 0.0003,
448
+ "loss": 0.31227171421051025,
449
+ "step": 1140
450
+ },
451
+ {
452
+ "epoch": 0.07819346140883048,
453
+ "grad_norm": 0.33984375,
454
+ "learning_rate": 0.0003,
455
+ "loss": 0.3114597320556641,
456
+ "step": 1160
457
+ },
458
+ {
459
+ "epoch": 0.07954162453656892,
460
+ "grad_norm": 1.7734375,
461
+ "learning_rate": 0.0003,
462
+ "loss": 0.3397975921630859,
463
+ "step": 1180
464
+ },
465
+ {
466
+ "epoch": 0.08088978766430738,
467
+ "grad_norm": 0.443359375,
468
+ "learning_rate": 0.0003,
469
+ "loss": 0.32673091888427735,
470
+ "step": 1200
471
+ },
472
+ {
473
+ "epoch": 0.08088978766430738,
474
+ "eval_loss": 0.31654924154281616,
475
+ "eval_runtime": 12.226,
476
+ "eval_samples_per_second": 779.241,
477
+ "eval_steps_per_second": 12.187,
478
+ "step": 1200
479
+ },
480
+ {
481
+ "epoch": 0.08223795079204584,
482
+ "grad_norm": 0.27734375,
483
+ "learning_rate": 0.0003,
484
+ "loss": 0.30962119102478025,
485
+ "step": 1220
486
+ },
487
+ {
488
+ "epoch": 0.0835861139197843,
489
+ "grad_norm": 0.25390625,
490
+ "learning_rate": 0.0003,
491
+ "loss": 0.3049886465072632,
492
+ "step": 1240
493
+ },
494
+ {
495
+ "epoch": 0.08493427704752275,
496
+ "grad_norm": 0.25,
497
+ "learning_rate": 0.0003,
498
+ "loss": 0.3021548748016357,
499
+ "step": 1260
500
+ },
501
+ {
502
+ "epoch": 0.08628244017526121,
503
+ "grad_norm": 0.259765625,
504
+ "learning_rate": 0.0003,
505
+ "loss": 0.29709596633911134,
506
+ "step": 1280
507
+ },
508
+ {
509
+ "epoch": 0.08763060330299967,
510
+ "grad_norm": 0.255859375,
511
+ "learning_rate": 0.0003,
512
+ "loss": 0.29625716209411623,
513
+ "step": 1300
514
+ },
515
+ {
516
+ "epoch": 0.08897876643073811,
517
+ "grad_norm": 0.24609375,
518
+ "learning_rate": 0.0003,
519
+ "loss": 0.29299511909484866,
520
+ "step": 1320
521
+ },
522
+ {
523
+ "epoch": 0.09032692955847657,
524
+ "grad_norm": 0.24609375,
525
+ "learning_rate": 0.0003,
526
+ "loss": 0.2907891273498535,
527
+ "step": 1340
528
+ },
529
+ {
530
+ "epoch": 0.09167509268621503,
531
+ "grad_norm": 0.25,
532
+ "learning_rate": 0.0003,
533
+ "loss": 0.2875590562820435,
534
+ "step": 1360
535
+ },
536
+ {
537
+ "epoch": 0.09302325581395349,
538
+ "grad_norm": 0.25,
539
+ "learning_rate": 0.0003,
540
+ "loss": 0.2870134115219116,
541
+ "step": 1380
542
+ },
543
+ {
544
+ "epoch": 0.09437141894169195,
545
+ "grad_norm": 0.25390625,
546
+ "learning_rate": 0.0003,
547
+ "loss": 0.28395581245422363,
548
+ "step": 1400
549
+ },
550
+ {
551
+ "epoch": 0.09437141894169195,
552
+ "eval_loss": 0.2858739495277405,
553
+ "eval_runtime": 12.1075,
554
+ "eval_samples_per_second": 786.87,
555
+ "eval_steps_per_second": 12.306,
556
+ "step": 1400
557
+ },
558
+ {
559
+ "epoch": 0.0957195820694304,
560
+ "grad_norm": 0.2890625,
561
+ "learning_rate": 0.0003,
562
+ "loss": 0.283489727973938,
563
+ "step": 1420
564
+ },
565
+ {
566
+ "epoch": 0.09706774519716886,
567
+ "grad_norm": 0.248046875,
568
+ "learning_rate": 0.0003,
569
+ "loss": 0.2808219909667969,
570
+ "step": 1440
571
+ },
572
+ {
573
+ "epoch": 0.09841590832490732,
574
+ "grad_norm": 0.248046875,
575
+ "learning_rate": 0.0003,
576
+ "loss": 0.2781242847442627,
577
+ "step": 1460
578
+ },
579
+ {
580
+ "epoch": 0.09976407145264576,
581
+ "grad_norm": 0.2490234375,
582
+ "learning_rate": 0.0003,
583
+ "loss": 0.27568469047546384,
584
+ "step": 1480
585
+ },
586
+ {
587
+ "epoch": 0.10111223458038422,
588
+ "grad_norm": 0.24609375,
589
+ "learning_rate": 0.0003,
590
+ "loss": 0.27444336414337156,
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": 2226156208128000.0,
612
+ "train_batch_size": 64,
613
+ "trial_name": null,
614
+ "trial_params": null
615
+ }
out/glu-gelu-46L_run/checkpoint-1500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ca8220f6bfcbccdaf71518f7684986b0dcb6d39577064a9547bf51ad36ca3f7
3
+ size 4920
out/glu-gelu-46L_run/checkpoint-2000/config.json ADDED
@@ -0,0 +1,35 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation": "gelu",
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": 46,
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-gelu-46L_run/checkpoint-2000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0f163431bc539b0222ec09f957b13c37610b2586d6ba9b8029b52b3c09a6d9d
3
+ size 16191120
out/glu-gelu-46L_run/checkpoint-2000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4763178af85e7cb95e2e866ef1b983696a3e5c2a2b55b9eae753e63c96ce3d87
3
+ size 32639090
out/glu-gelu-46L_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-gelu-46L_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-gelu-46L_run/checkpoint-2000/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
out/glu-gelu-46L_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-gelu-46L_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": 1.2578125,
15
+ "learning_rate": 0.0003,
16
+ "loss": 7.776710510253906,
17
+ "step": 20
18
+ },
19
+ {
20
+ "epoch": 0.002696326255476913,
21
+ "grad_norm": 1.171875,
22
+ "learning_rate": 0.0003,
23
+ "loss": 7.017825317382813,
24
+ "step": 40
25
+ },
26
+ {
27
+ "epoch": 0.004044489383215369,
28
+ "grad_norm": 0.9296875,
29
+ "learning_rate": 0.0003,
30
+ "loss": 6.486733245849609,
31
+ "step": 60
32
+ },
33
+ {
34
+ "epoch": 0.005392652510953826,
35
+ "grad_norm": 0.6875,
36
+ "learning_rate": 0.0003,
37
+ "loss": 6.160807418823242,
38
+ "step": 80
39
+ },
40
+ {
41
+ "epoch": 0.006740815638692282,
42
+ "grad_norm": 0.640625,
43
+ "learning_rate": 0.0003,
44
+ "loss": 5.966274261474609,
45
+ "step": 100
46
+ },
47
+ {
48
+ "epoch": 0.008088978766430738,
49
+ "grad_norm": 0.9765625,
50
+ "learning_rate": 0.0003,
51
+ "loss": 5.7247802734375,
52
+ "step": 120
53
+ },
54
+ {
55
+ "epoch": 0.009437141894169195,
56
+ "grad_norm": 1.046875,
57
+ "learning_rate": 0.0003,
58
+ "loss": 5.335699462890625,
59
+ "step": 140
60
+ },
61
+ {
62
+ "epoch": 0.010785305021907651,
63
+ "grad_norm": 1.0234375,
64
+ "learning_rate": 0.0003,
65
+ "loss": 4.942352294921875,
66
+ "step": 160
67
+ },
68
+ {
69
+ "epoch": 0.012133468149646108,
70
+ "grad_norm": 1.15625,
71
+ "learning_rate": 0.0003,
72
+ "loss": 4.513675308227539,
73
+ "step": 180
74
+ },
75
+ {
76
+ "epoch": 0.013481631277384564,
77
+ "grad_norm": 1.2421875,
78
+ "learning_rate": 0.0003,
79
+ "loss": 4.05323371887207,
80
+ "step": 200
81
+ },
82
+ {
83
+ "epoch": 0.013481631277384564,
84
+ "eval_loss": 3.848839521408081,
85
+ "eval_runtime": 12.1844,
86
+ "eval_samples_per_second": 781.901,
87
+ "eval_steps_per_second": 12.229,
88
+ "step": 200
89
+ },
90
+ {
91
+ "epoch": 0.01482979440512302,
92
+ "grad_norm": 1.515625,
93
+ "learning_rate": 0.0003,
94
+ "loss": 3.664620208740234,
95
+ "step": 220
96
+ },
97
+ {
98
+ "epoch": 0.016177957532861477,
99
+ "grad_norm": 1.265625,
100
+ "learning_rate": 0.0003,
101
+ "loss": 3.32559814453125,
102
+ "step": 240
103
+ },
104
+ {
105
+ "epoch": 0.01752612066059993,
106
+ "grad_norm": 1.8671875,
107
+ "learning_rate": 0.0003,
108
+ "loss": 3.0320526123046876,
109
+ "step": 260
110
+ },
111
+ {
112
+ "epoch": 0.01887428378833839,
113
+ "grad_norm": 1.984375,
114
+ "learning_rate": 0.0003,
115
+ "loss": 2.754303741455078,
116
+ "step": 280
117
+ },
118
+ {
119
+ "epoch": 0.020222446916076844,
120
+ "grad_norm": 2.171875,
121
+ "learning_rate": 0.0003,
122
+ "loss": 2.4877477645874024,
123
+ "step": 300
124
+ },
125
+ {
126
+ "epoch": 0.021570610043815303,
127
+ "grad_norm": 1.640625,
128
+ "learning_rate": 0.0003,
129
+ "loss": 2.2490768432617188,
130
+ "step": 320
131
+ },
132
+ {
133
+ "epoch": 0.022918773171553757,
134
+ "grad_norm": 1.59375,
135
+ "learning_rate": 0.0003,
136
+ "loss": 2.0304283142089843,
137
+ "step": 340
138
+ },
139
+ {
140
+ "epoch": 0.024266936299292215,
141
+ "grad_norm": 1.484375,
142
+ "learning_rate": 0.0003,
143
+ "loss": 1.8307512283325196,
144
+ "step": 360
145
+ },
146
+ {
147
+ "epoch": 0.02561509942703067,
148
+ "grad_norm": 1.78125,
149
+ "learning_rate": 0.0003,
150
+ "loss": 1.6494441986083985,
151
+ "step": 380
152
+ },
153
+ {
154
+ "epoch": 0.026963262554769128,
155
+ "grad_norm": 1.8203125,
156
+ "learning_rate": 0.0003,
157
+ "loss": 1.4764435768127442,
158
+ "step": 400
159
+ },
160
+ {
161
+ "epoch": 0.026963262554769128,
162
+ "eval_loss": 1.3990651369094849,
163
+ "eval_runtime": 12.2447,
164
+ "eval_samples_per_second": 778.051,
165
+ "eval_steps_per_second": 12.169,
166
+ "step": 400
167
+ },
168
+ {
169
+ "epoch": 0.028311425682507583,
170
+ "grad_norm": 1.5625,
171
+ "learning_rate": 0.0003,
172
+ "loss": 1.3182246208190918,
173
+ "step": 420
174
+ },
175
+ {
176
+ "epoch": 0.02965958881024604,
177
+ "grad_norm": 1.0625,
178
+ "learning_rate": 0.0003,
179
+ "loss": 1.1943863868713378,
180
+ "step": 440
181
+ },
182
+ {
183
+ "epoch": 0.031007751937984496,
184
+ "grad_norm": 1.4609375,
185
+ "learning_rate": 0.0003,
186
+ "loss": 1.0682686805725097,
187
+ "step": 460
188
+ },
189
+ {
190
+ "epoch": 0.032355915065722954,
191
+ "grad_norm": 1.1875,
192
+ "learning_rate": 0.0003,
193
+ "loss": 0.9639272689819336,
194
+ "step": 480
195
+ },
196
+ {
197
+ "epoch": 0.03370407819346141,
198
+ "grad_norm": 0.69921875,
199
+ "learning_rate": 0.0003,
200
+ "loss": 0.8594159126281739,
201
+ "step": 500
202
+ },
203
+ {
204
+ "epoch": 0.03505224132119986,
205
+ "grad_norm": 0.4140625,
206
+ "learning_rate": 0.0003,
207
+ "loss": 0.7737322807312011,
208
+ "step": 520
209
+ },
210
+ {
211
+ "epoch": 0.03640040444893832,
212
+ "grad_norm": 0.419921875,
213
+ "learning_rate": 0.0003,
214
+ "loss": 0.6988646507263183,
215
+ "step": 540
216
+ },
217
+ {
218
+ "epoch": 0.03774856757667678,
219
+ "grad_norm": 0.40234375,
220
+ "learning_rate": 0.0003,
221
+ "loss": 0.6476529598236084,
222
+ "step": 560
223
+ },
224
+ {
225
+ "epoch": 0.03909673070441524,
226
+ "grad_norm": 1.7421875,
227
+ "learning_rate": 0.0003,
228
+ "loss": 0.6122144222259521,
229
+ "step": 580
230
+ },
231
+ {
232
+ "epoch": 0.04044489383215369,
233
+ "grad_norm": 1.3046875,
234
+ "learning_rate": 0.0003,
235
+ "loss": 0.5976991176605224,
236
+ "step": 600
237
+ },
238
+ {
239
+ "epoch": 0.04044489383215369,
240
+ "eval_loss": 0.5825157165527344,
241
+ "eval_runtime": 12.2182,
242
+ "eval_samples_per_second": 779.74,
243
+ "eval_steps_per_second": 12.195,
244
+ "step": 600
245
+ },
246
+ {
247
+ "epoch": 0.04179305695989215,
248
+ "grad_norm": 0.40234375,
249
+ "learning_rate": 0.0003,
250
+ "loss": 0.5630900382995605,
251
+ "step": 620
252
+ },
253
+ {
254
+ "epoch": 0.043141220087630605,
255
+ "grad_norm": 0.3359375,
256
+ "learning_rate": 0.0003,
257
+ "loss": 0.5361714363098145,
258
+ "step": 640
259
+ },
260
+ {
261
+ "epoch": 0.044489383215369056,
262
+ "grad_norm": 0.3125,
263
+ "learning_rate": 0.0003,
264
+ "loss": 0.5064755439758301,
265
+ "step": 660
266
+ },
267
+ {
268
+ "epoch": 0.045837546343107514,
269
+ "grad_norm": 0.322265625,
270
+ "learning_rate": 0.0003,
271
+ "loss": 0.4883098602294922,
272
+ "step": 680
273
+ },
274
+ {
275
+ "epoch": 0.04718570947084597,
276
+ "grad_norm": 0.3125,
277
+ "learning_rate": 0.0003,
278
+ "loss": 0.4687828540802002,
279
+ "step": 700
280
+ },
281
+ {
282
+ "epoch": 0.04853387259858443,
283
+ "grad_norm": 0.333984375,
284
+ "learning_rate": 0.0003,
285
+ "loss": 0.4550455093383789,
286
+ "step": 720
287
+ },
288
+ {
289
+ "epoch": 0.04988203572632288,
290
+ "grad_norm": 0.3515625,
291
+ "learning_rate": 0.0003,
292
+ "loss": 0.4373329639434814,
293
+ "step": 740
294
+ },
295
+ {
296
+ "epoch": 0.05123019885406134,
297
+ "grad_norm": 0.365234375,
298
+ "learning_rate": 0.0003,
299
+ "loss": 0.42957515716552735,
300
+ "step": 760
301
+ },
302
+ {
303
+ "epoch": 0.0525783619817998,
304
+ "grad_norm": 1.9296875,
305
+ "learning_rate": 0.0003,
306
+ "loss": 0.45300583839416503,
307
+ "step": 780
308
+ },
309
+ {
310
+ "epoch": 0.053926525109538256,
311
+ "grad_norm": 0.5390625,
312
+ "learning_rate": 0.0003,
313
+ "loss": 0.42671890258789064,
314
+ "step": 800
315
+ },
316
+ {
317
+ "epoch": 0.053926525109538256,
318
+ "eval_loss": 0.41847455501556396,
319
+ "eval_runtime": 12.2303,
320
+ "eval_samples_per_second": 778.967,
321
+ "eval_steps_per_second": 12.183,
322
+ "step": 800
323
+ },
324
+ {
325
+ "epoch": 0.05527468823727671,
326
+ "grad_norm": 0.34765625,
327
+ "learning_rate": 0.0003,
328
+ "loss": 0.4097603797912598,
329
+ "step": 820
330
+ },
331
+ {
332
+ "epoch": 0.056622851365015166,
333
+ "grad_norm": 0.291015625,
334
+ "learning_rate": 0.0003,
335
+ "loss": 0.39820222854614257,
336
+ "step": 840
337
+ },
338
+ {
339
+ "epoch": 0.057971014492753624,
340
+ "grad_norm": 0.27734375,
341
+ "learning_rate": 0.0003,
342
+ "loss": 0.3863217830657959,
343
+ "step": 860
344
+ },
345
+ {
346
+ "epoch": 0.05931917762049208,
347
+ "grad_norm": 0.27734375,
348
+ "learning_rate": 0.0003,
349
+ "loss": 0.3764716386795044,
350
+ "step": 880
351
+ },
352
+ {
353
+ "epoch": 0.06066734074823053,
354
+ "grad_norm": 0.275390625,
355
+ "learning_rate": 0.0003,
356
+ "loss": 0.373111891746521,
357
+ "step": 900
358
+ },
359
+ {
360
+ "epoch": 0.06201550387596899,
361
+ "grad_norm": 0.283203125,
362
+ "learning_rate": 0.0003,
363
+ "loss": 0.36699652671813965,
364
+ "step": 920
365
+ },
366
+ {
367
+ "epoch": 0.06336366700370745,
368
+ "grad_norm": 0.279296875,
369
+ "learning_rate": 0.0003,
370
+ "loss": 0.35993266105651855,
371
+ "step": 940
372
+ },
373
+ {
374
+ "epoch": 0.06471183013144591,
375
+ "grad_norm": 0.28125,
376
+ "learning_rate": 0.0003,
377
+ "loss": 0.3524770736694336,
378
+ "step": 960
379
+ },
380
+ {
381
+ "epoch": 0.06605999325918437,
382
+ "grad_norm": 0.271484375,
383
+ "learning_rate": 0.0003,
384
+ "loss": 0.34836206436157224,
385
+ "step": 980
386
+ },
387
+ {
388
+ "epoch": 0.06740815638692282,
389
+ "grad_norm": 0.3046875,
390
+ "learning_rate": 0.0003,
391
+ "loss": 0.34576056003570554,
392
+ "step": 1000
393
+ },
394
+ {
395
+ "epoch": 0.06740815638692282,
396
+ "eval_loss": 0.34370264410972595,
397
+ "eval_runtime": 12.2146,
398
+ "eval_samples_per_second": 779.966,
399
+ "eval_steps_per_second": 12.198,
400
+ "step": 1000
401
+ },
402
+ {
403
+ "epoch": 0.06875631951466127,
404
+ "grad_norm": 0.283203125,
405
+ "learning_rate": 0.0003,
406
+ "loss": 0.3385118007659912,
407
+ "step": 1020
408
+ },
409
+ {
410
+ "epoch": 0.07010448264239973,
411
+ "grad_norm": 0.263671875,
412
+ "learning_rate": 0.0003,
413
+ "loss": 0.33219237327575685,
414
+ "step": 1040
415
+ },
416
+ {
417
+ "epoch": 0.07145264577013818,
418
+ "grad_norm": 0.271484375,
419
+ "learning_rate": 0.0003,
420
+ "loss": 0.3300537109375,
421
+ "step": 1060
422
+ },
423
+ {
424
+ "epoch": 0.07280080889787664,
425
+ "grad_norm": 0.283203125,
426
+ "learning_rate": 0.0003,
427
+ "loss": 0.32398457527160646,
428
+ "step": 1080
429
+ },
430
+ {
431
+ "epoch": 0.0741489720256151,
432
+ "grad_norm": 0.287109375,
433
+ "learning_rate": 0.0003,
434
+ "loss": 0.32252542972564696,
435
+ "step": 1100
436
+ },
437
+ {
438
+ "epoch": 0.07549713515335356,
439
+ "grad_norm": 0.310546875,
440
+ "learning_rate": 0.0003,
441
+ "loss": 0.31789841651916506,
442
+ "step": 1120
443
+ },
444
+ {
445
+ "epoch": 0.07684529828109202,
446
+ "grad_norm": 0.3125,
447
+ "learning_rate": 0.0003,
448
+ "loss": 0.31227171421051025,
449
+ "step": 1140
450
+ },
451
+ {
452
+ "epoch": 0.07819346140883048,
453
+ "grad_norm": 0.33984375,
454
+ "learning_rate": 0.0003,
455
+ "loss": 0.3114597320556641,
456
+ "step": 1160
457
+ },
458
+ {
459
+ "epoch": 0.07954162453656892,
460
+ "grad_norm": 1.7734375,
461
+ "learning_rate": 0.0003,
462
+ "loss": 0.3397975921630859,
463
+ "step": 1180
464
+ },
465
+ {
466
+ "epoch": 0.08088978766430738,
467
+ "grad_norm": 0.443359375,
468
+ "learning_rate": 0.0003,
469
+ "loss": 0.32673091888427735,
470
+ "step": 1200
471
+ },
472
+ {
473
+ "epoch": 0.08088978766430738,
474
+ "eval_loss": 0.31654924154281616,
475
+ "eval_runtime": 12.226,
476
+ "eval_samples_per_second": 779.241,
477
+ "eval_steps_per_second": 12.187,
478
+ "step": 1200
479
+ },
480
+ {
481
+ "epoch": 0.08223795079204584,
482
+ "grad_norm": 0.27734375,
483
+ "learning_rate": 0.0003,
484
+ "loss": 0.30962119102478025,
485
+ "step": 1220
486
+ },
487
+ {
488
+ "epoch": 0.0835861139197843,
489
+ "grad_norm": 0.25390625,
490
+ "learning_rate": 0.0003,
491
+ "loss": 0.3049886465072632,
492
+ "step": 1240
493
+ },
494
+ {
495
+ "epoch": 0.08493427704752275,
496
+ "grad_norm": 0.25,
497
+ "learning_rate": 0.0003,
498
+ "loss": 0.3021548748016357,
499
+ "step": 1260
500
+ },
501
+ {
502
+ "epoch": 0.08628244017526121,
503
+ "grad_norm": 0.259765625,
504
+ "learning_rate": 0.0003,
505
+ "loss": 0.29709596633911134,
506
+ "step": 1280
507
+ },
508
+ {
509
+ "epoch": 0.08763060330299967,
510
+ "grad_norm": 0.255859375,
511
+ "learning_rate": 0.0003,
512
+ "loss": 0.29625716209411623,
513
+ "step": 1300
514
+ },
515
+ {
516
+ "epoch": 0.08897876643073811,
517
+ "grad_norm": 0.24609375,
518
+ "learning_rate": 0.0003,
519
+ "loss": 0.29299511909484866,
520
+ "step": 1320
521
+ },
522
+ {
523
+ "epoch": 0.09032692955847657,
524
+ "grad_norm": 0.24609375,
525
+ "learning_rate": 0.0003,
526
+ "loss": 0.2907891273498535,
527
+ "step": 1340
528
+ },
529
+ {
530
+ "epoch": 0.09167509268621503,
531
+ "grad_norm": 0.25,
532
+ "learning_rate": 0.0003,
533
+ "loss": 0.2875590562820435,
534
+ "step": 1360
535
+ },
536
+ {
537
+ "epoch": 0.09302325581395349,
538
+ "grad_norm": 0.25,
539
+ "learning_rate": 0.0003,
540
+ "loss": 0.2870134115219116,
541
+ "step": 1380
542
+ },
543
+ {
544
+ "epoch": 0.09437141894169195,
545
+ "grad_norm": 0.25390625,
546
+ "learning_rate": 0.0003,
547
+ "loss": 0.28395581245422363,
548
+ "step": 1400
549
+ },
550
+ {
551
+ "epoch": 0.09437141894169195,
552
+ "eval_loss": 0.2858739495277405,
553
+ "eval_runtime": 12.1075,
554
+ "eval_samples_per_second": 786.87,
555
+ "eval_steps_per_second": 12.306,
556
+ "step": 1400
557
+ },
558
+ {
559
+ "epoch": 0.0957195820694304,
560
+ "grad_norm": 0.2890625,
561
+ "learning_rate": 0.0003,
562
+ "loss": 0.283489727973938,
563
+ "step": 1420
564
+ },
565
+ {
566
+ "epoch": 0.09706774519716886,
567
+ "grad_norm": 0.248046875,
568
+ "learning_rate": 0.0003,
569
+ "loss": 0.2808219909667969,
570
+ "step": 1440
571
+ },
572
+ {
573
+ "epoch": 0.09841590832490732,
574
+ "grad_norm": 0.248046875,
575
+ "learning_rate": 0.0003,
576
+ "loss": 0.2781242847442627,
577
+ "step": 1460
578
+ },
579
+ {
580
+ "epoch": 0.09976407145264576,
581
+ "grad_norm": 0.2490234375,
582
+ "learning_rate": 0.0003,
583
+ "loss": 0.27568469047546384,
584
+ "step": 1480
585
+ },
586
+ {
587
+ "epoch": 0.10111223458038422,
588
+ "grad_norm": 0.24609375,
589
+ "learning_rate": 0.0003,
590
+ "loss": 0.27444336414337156,
591
+ "step": 1500
592
+ },
593
+ {
594
+ "epoch": 0.10246039770812268,
595
+ "grad_norm": 0.2431640625,
596
+ "learning_rate": 0.0003,
597
+ "loss": 0.27230570316314695,
598
+ "step": 1520
599
+ },
600
+ {
601
+ "epoch": 0.10380856083586114,
602
+ "grad_norm": 0.26953125,
603
+ "learning_rate": 0.0003,
604
+ "loss": 0.2712153673171997,
605
+ "step": 1540
606
+ },
607
+ {
608
+ "epoch": 0.1051567239635996,
609
+ "grad_norm": 0.2734375,
610
+ "learning_rate": 0.0003,
611
+ "loss": 0.26991493701934816,
612
+ "step": 1560
613
+ },
614
+ {
615
+ "epoch": 0.10650488709133805,
616
+ "grad_norm": 0.267578125,
617
+ "learning_rate": 0.0003,
618
+ "loss": 0.2705997467041016,
619
+ "step": 1580
620
+ },
621
+ {
622
+ "epoch": 0.10785305021907651,
623
+ "grad_norm": 1.4140625,
624
+ "learning_rate": 0.0003,
625
+ "loss": 0.30360138416290283,
626
+ "step": 1600
627
+ },
628
+ {
629
+ "epoch": 0.10785305021907651,
630
+ "eval_loss": 0.29976317286491394,
631
+ "eval_runtime": 12.1313,
632
+ "eval_samples_per_second": 785.326,
633
+ "eval_steps_per_second": 12.282,
634
+ "step": 1600
635
+ },
636
+ {
637
+ "epoch": 0.10920121334681497,
638
+ "grad_norm": 0.4921875,
639
+ "learning_rate": 0.0003,
640
+ "loss": 0.2871302366256714,
641
+ "step": 1620
642
+ },
643
+ {
644
+ "epoch": 0.11054937647455342,
645
+ "grad_norm": 0.263671875,
646
+ "learning_rate": 0.0003,
647
+ "loss": 0.2725695610046387,
648
+ "step": 1640
649
+ },
650
+ {
651
+ "epoch": 0.11189753960229187,
652
+ "grad_norm": 0.2392578125,
653
+ "learning_rate": 0.0003,
654
+ "loss": 0.26655561923980714,
655
+ "step": 1660
656
+ },
657
+ {
658
+ "epoch": 0.11324570273003033,
659
+ "grad_norm": 0.23828125,
660
+ "learning_rate": 0.0003,
661
+ "loss": 0.2647746801376343,
662
+ "step": 1680
663
+ },
664
+ {
665
+ "epoch": 0.11459386585776879,
666
+ "grad_norm": 0.23046875,
667
+ "learning_rate": 0.0003,
668
+ "loss": 0.26475510597229,
669
+ "step": 1700
670
+ },
671
+ {
672
+ "epoch": 0.11594202898550725,
673
+ "grad_norm": 0.232421875,
674
+ "learning_rate": 0.0003,
675
+ "loss": 0.26238973140716554,
676
+ "step": 1720
677
+ },
678
+ {
679
+ "epoch": 0.1172901921132457,
680
+ "grad_norm": 0.234375,
681
+ "learning_rate": 0.0003,
682
+ "loss": 0.2619847297668457,
683
+ "step": 1740
684
+ },
685
+ {
686
+ "epoch": 0.11863835524098416,
687
+ "grad_norm": 0.236328125,
688
+ "learning_rate": 0.0003,
689
+ "loss": 0.25990753173828124,
690
+ "step": 1760
691
+ },
692
+ {
693
+ "epoch": 0.11998651836872262,
694
+ "grad_norm": 0.228515625,
695
+ "learning_rate": 0.0003,
696
+ "loss": 0.2608365535736084,
697
+ "step": 1780
698
+ },
699
+ {
700
+ "epoch": 0.12133468149646107,
701
+ "grad_norm": 0.2314453125,
702
+ "learning_rate": 0.0003,
703
+ "loss": 0.26053175926208494,
704
+ "step": 1800
705
+ },
706
+ {
707
+ "epoch": 0.12133468149646107,
708
+ "eval_loss": 0.2594231069087982,
709
+ "eval_runtime": 12.1218,
710
+ "eval_samples_per_second": 785.941,
711
+ "eval_steps_per_second": 12.292,
712
+ "step": 1800
713
+ },
714
+ {
715
+ "epoch": 0.12268284462419952,
716
+ "grad_norm": 0.2333984375,
717
+ "learning_rate": 0.0003,
718
+ "loss": 0.2600924015045166,
719
+ "step": 1820
720
+ },
721
+ {
722
+ "epoch": 0.12403100775193798,
723
+ "grad_norm": 0.232421875,
724
+ "learning_rate": 0.0003,
725
+ "loss": 0.25479280948638916,
726
+ "step": 1840
727
+ },
728
+ {
729
+ "epoch": 0.12537917087967643,
730
+ "grad_norm": 0.23828125,
731
+ "learning_rate": 0.0003,
732
+ "loss": 0.2544747829437256,
733
+ "step": 1860
734
+ },
735
+ {
736
+ "epoch": 0.1267273340074149,
737
+ "grad_norm": 0.232421875,
738
+ "learning_rate": 0.0003,
739
+ "loss": 0.2535433769226074,
740
+ "step": 1880
741
+ },
742
+ {
743
+ "epoch": 0.12807549713515334,
744
+ "grad_norm": 0.2314453125,
745
+ "learning_rate": 0.0003,
746
+ "loss": 0.25360000133514404,
747
+ "step": 1900
748
+ },
749
+ {
750
+ "epoch": 0.12942366026289182,
751
+ "grad_norm": 0.2314453125,
752
+ "learning_rate": 0.0003,
753
+ "loss": 0.2511674165725708,
754
+ "step": 1920
755
+ },
756
+ {
757
+ "epoch": 0.13077182339063026,
758
+ "grad_norm": 0.2353515625,
759
+ "learning_rate": 0.0003,
760
+ "loss": 0.2518420457839966,
761
+ "step": 1940
762
+ },
763
+ {
764
+ "epoch": 0.13211998651836873,
765
+ "grad_norm": 0.2333984375,
766
+ "learning_rate": 0.0003,
767
+ "loss": 0.24904108047485352,
768
+ "step": 1960
769
+ },
770
+ {
771
+ "epoch": 0.13346814964610718,
772
+ "grad_norm": 0.2333984375,
773
+ "learning_rate": 0.0003,
774
+ "loss": 0.2502410411834717,
775
+ "step": 1980
776
+ },
777
+ {
778
+ "epoch": 0.13481631277384565,
779
+ "grad_norm": 0.2392578125,
780
+ "learning_rate": 0.0003,
781
+ "loss": 0.24965012073516846,
782
+ "step": 2000
783
+ },
784
+ {
785
+ "epoch": 0.13481631277384565,
786
+ "eval_loss": 0.2511940002441406,
787
+ "eval_runtime": 12.2354,
788
+ "eval_samples_per_second": 778.645,
789
+ "eval_steps_per_second": 12.178,
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": 2968208277504000.0,
811
+ "train_batch_size": 64,
812
+ "trial_name": null,
813
+ "trial_params": null
814
+ }
out/glu-gelu-46L_run/checkpoint-2000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ca8220f6bfcbccdaf71518f7684986b0dcb6d39577064a9547bf51ad36ca3f7
3
+ size 4920
out/glu-gelu-46L_run/model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:62bf52d1672ec3570b36a4918ca41734067b632af0cad695c848b77da699b0f2
3
  size 16191120
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0f163431bc539b0222ec09f957b13c37610b2586d6ba9b8029b52b3c09a6d9d
3
  size 16191120
out/glu-gelu-46L_run/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
out/glu-gelu-46L_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-gelu-46L_run/training_log.jsonl CHANGED
The diff for this file is too large to render. See raw diff
 
out/glu-linear-46L_run/checkpoint-500/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": 46,
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-46L_run/checkpoint-500/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c90e729ffd3245deab2bd41745382113f242b5347eedf495638ce4b4a14167b
3
+ size 16191120
out/glu-linear-46L_run/checkpoint-500/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3229323c2137f0ef6263680d01a3f4e65a3f60679fc2fd3e058577f49d7024a7
3
+ size 32639090
out/glu-linear-46L_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-linear-46L_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-linear-46L_run/checkpoint-500/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
out/glu-linear-46L_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-linear-46L_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": 1.25,
15
+ "learning_rate": 0.0003,
16
+ "loss": 7.778260040283203,
17
+ "step": 20
18
+ },
19
+ {
20
+ "epoch": 0.002696326255476913,
21
+ "grad_norm": 1.1796875,
22
+ "learning_rate": 0.0003,
23
+ "loss": 7.014875793457032,
24
+ "step": 40
25
+ },
26
+ {
27
+ "epoch": 0.004044489383215369,
28
+ "grad_norm": 0.9921875,
29
+ "learning_rate": 0.0003,
30
+ "loss": 6.453499603271484,
31
+ "step": 60
32
+ },
33
+ {
34
+ "epoch": 0.005392652510953826,
35
+ "grad_norm": 0.94140625,
36
+ "learning_rate": 0.0003,
37
+ "loss": 6.045098876953125,
38
+ "step": 80
39
+ },
40
+ {
41
+ "epoch": 0.006740815638692282,
42
+ "grad_norm": 1.34375,
43
+ "learning_rate": 0.0003,
44
+ "loss": 5.617780685424805,
45
+ "step": 100
46
+ },
47
+ {
48
+ "epoch": 0.008088978766430738,
49
+ "grad_norm": 1.9296875,
50
+ "learning_rate": 0.0003,
51
+ "loss": 5.101103973388672,
52
+ "step": 120
53
+ },
54
+ {
55
+ "epoch": 0.009437141894169195,
56
+ "grad_norm": 1.96875,
57
+ "learning_rate": 0.0003,
58
+ "loss": 4.574287033081054,
59
+ "step": 140
60
+ },
61
+ {
62
+ "epoch": 0.010785305021907651,
63
+ "grad_norm": 1.875,
64
+ "learning_rate": 0.0003,
65
+ "loss": 4.0741230010986325,
66
+ "step": 160
67
+ },
68
+ {
69
+ "epoch": 0.012133468149646108,
70
+ "grad_norm": 1.8359375,
71
+ "learning_rate": 0.0003,
72
+ "loss": 3.617207336425781,
73
+ "step": 180
74
+ },
75
+ {
76
+ "epoch": 0.013481631277384564,
77
+ "grad_norm": 2.21875,
78
+ "learning_rate": 0.0003,
79
+ "loss": 3.190230369567871,
80
+ "step": 200
81
+ },
82
+ {
83
+ "epoch": 0.013481631277384564,
84
+ "eval_loss": 2.9775030612945557,
85
+ "eval_runtime": 12.1012,
86
+ "eval_samples_per_second": 787.278,
87
+ "eval_steps_per_second": 12.313,
88
+ "step": 200
89
+ },
90
+ {
91
+ "epoch": 0.01482979440512302,
92
+ "grad_norm": 1.5,
93
+ "learning_rate": 0.0003,
94
+ "loss": 2.8015439987182615,
95
+ "step": 220
96
+ },
97
+ {
98
+ "epoch": 0.016177957532861477,
99
+ "grad_norm": 1.875,
100
+ "learning_rate": 0.0003,
101
+ "loss": 2.4547601699829102,
102
+ "step": 240
103
+ },
104
+ {
105
+ "epoch": 0.01752612066059993,
106
+ "grad_norm": 1.7578125,
107
+ "learning_rate": 0.0003,
108
+ "loss": 2.14495849609375,
109
+ "step": 260
110
+ },
111
+ {
112
+ "epoch": 0.01887428378833839,
113
+ "grad_norm": 1.7890625,
114
+ "learning_rate": 0.0003,
115
+ "loss": 1.8642923355102539,
116
+ "step": 280
117
+ },
118
+ {
119
+ "epoch": 0.020222446916076844,
120
+ "grad_norm": 1.171875,
121
+ "learning_rate": 0.0003,
122
+ "loss": 1.6218135833740235,
123
+ "step": 300
124
+ },
125
+ {
126
+ "epoch": 0.021570610043815303,
127
+ "grad_norm": 1.2890625,
128
+ "learning_rate": 0.0003,
129
+ "loss": 1.4220643997192384,
130
+ "step": 320
131
+ },
132
+ {
133
+ "epoch": 0.022918773171553757,
134
+ "grad_norm": 1.125,
135
+ "learning_rate": 0.0003,
136
+ "loss": 1.2538734436035157,
137
+ "step": 340
138
+ },
139
+ {
140
+ "epoch": 0.024266936299292215,
141
+ "grad_norm": 0.9453125,
142
+ "learning_rate": 0.0003,
143
+ "loss": 1.1154563903808594,
144
+ "step": 360
145
+ },
146
+ {
147
+ "epoch": 0.02561509942703067,
148
+ "grad_norm": 1.0625,
149
+ "learning_rate": 0.0003,
150
+ "loss": 0.987572193145752,
151
+ "step": 380
152
+ },
153
+ {
154
+ "epoch": 0.026963262554769128,
155
+ "grad_norm": 0.65234375,
156
+ "learning_rate": 0.0003,
157
+ "loss": 0.8721179962158203,
158
+ "step": 400
159
+ },
160
+ {
161
+ "epoch": 0.026963262554769128,
162
+ "eval_loss": 0.8266251683235168,
163
+ "eval_runtime": 12.0698,
164
+ "eval_samples_per_second": 789.322,
165
+ "eval_steps_per_second": 12.345,
166
+ "step": 400
167
+ },
168
+ {
169
+ "epoch": 0.028311425682507583,
170
+ "grad_norm": 0.45703125,
171
+ "learning_rate": 0.0003,
172
+ "loss": 0.7757875919342041,
173
+ "step": 420
174
+ },
175
+ {
176
+ "epoch": 0.02965958881024604,
177
+ "grad_norm": 1.3359375,
178
+ "learning_rate": 0.0003,
179
+ "loss": 0.714703893661499,
180
+ "step": 440
181
+ },
182
+ {
183
+ "epoch": 0.031007751937984496,
184
+ "grad_norm": 1.40625,
185
+ "learning_rate": 0.0003,
186
+ "loss": 0.6639005184173584,
187
+ "step": 460
188
+ },
189
+ {
190
+ "epoch": 0.032355915065722954,
191
+ "grad_norm": 0.6796875,
192
+ "learning_rate": 0.0003,
193
+ "loss": 0.6394022464752197,
194
+ "step": 480
195
+ },
196
+ {
197
+ "epoch": 0.03370407819346141,
198
+ "grad_norm": 0.369140625,
199
+ "learning_rate": 0.0003,
200
+ "loss": 0.5926106929779053,
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": 742052069376000.0,
222
+ "train_batch_size": 64,
223
+ "trial_name": null,
224
+ "trial_params": null
225
+ }
out/glu-linear-46L_run/checkpoint-500/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9ff6d85897327bc65a2f864e422bef7e72f4da37412e8cf169a0c3be166bec15
3
+ size 4920
out/glu-linear-46L_run/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": 46,
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-46L_run/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c90e729ffd3245deab2bd41745382113f242b5347eedf495638ce4b4a14167b
3
+ size 16191120
out/glu-linear-46L_run/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9ff6d85897327bc65a2f864e422bef7e72f4da37412e8cf169a0c3be166bec15
3
+ size 4920
out/glu-linear-46L_run/training_log.jsonl ADDED
The diff for this file is too large to render. See raw diff