Hugging Face's logo

fromziro
/
64D-vs-32D

Safetensors
English
llama
ablation
heads
slms
dillionv2-tokenizer
llms
Model card Files
xet
 Community
Harley-ml commited on
Commit
3ca312f
·
verified ·
1 Parent(s): 4de987d

Upload 3 files

Browse files
Files changed (3) hide show
  1. config_a/config.json +32 -0
  2. config_a/model.safetensors +3 -0
  3. config_a/trainer_state.json +242 -0
config_a/config.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "LlamaForCausalLM"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 1,
8
+ "dtype": "float32",
9
+ "eos_token_id": 2,
10
+ "head_dim": 32,
11
+ "hidden_act": "silu",
12
+ "hidden_size": 256,
13
+ "initializer_range": 0.02,
14
+ "intermediate_size": 512,
15
+ "max_position_embeddings": 256,
16
+ "mlp_bias": false,
17
+ "model_type": "llama",
18
+ "num_attention_heads": 8,
19
+ "num_hidden_layers": 6,
20
+ "num_key_value_heads": 2,
21
+ "pad_token_id": null,
22
+ "pretraining_tp": 1,
23
+ "rms_norm_eps": 1e-06,
24
+ "rope_parameters": {
25
+ "rope_theta": 2500.0,
26
+ "rope_type": "default"
27
+ },
28
+ "tie_word_embeddings": true,
29
+ "transformers_version": "5.13.1",
30
+ "use_cache": false,
31
+ "vocab_size": 2564
32
+ }
config_a/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fad200394759863c02699fbe58929f4bec18f7504c2a97e58faa13edf907b6fc
3
+ size 16014208
config_a/trainer_state.json ADDED
@@ -0,0 +1,242 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_global_step": null,
3
+ "best_metric": null,
4
+ "best_model_checkpoint": null,
5
+ "epoch": 1.0,
6
+ "eval_steps": 500,
7
+ "global_step": 2428,
8
+ "is_hyper_param_search": false,
9
+ "is_local_process_zero": true,
10
+ "is_world_process_zero": true,
11
+ "log_history": [
12
+ {
13
+ "epoch": 0.0411946446961895,
14
+ "grad_norm": 0.25275173783302307,
15
+ "learning_rate": 0.003,
16
+ "loss": 6.336950073242187,
17
+ "step": 100
18
+ },
19
+ {
20
+ "epoch": 0.082389289392379,
21
+ "grad_norm": 0.3259478509426117,
22
+ "learning_rate": 0.003,
23
+ "loss": 5.181984252929688,
24
+ "step": 200
25
+ },
26
+ {
27
+ "epoch": 0.12358393408856849,
28
+ "grad_norm": 0.6726269721984863,
29
+ "learning_rate": 0.003,
30
+ "loss": 4.444449768066407,
31
+ "step": 300
32
+ },
33
+ {
34
+ "epoch": 0.164778578784758,
35
+ "grad_norm": 0.319759726524353,
36
+ "learning_rate": 0.003,
37
+ "loss": 3.9950070190429687,
38
+ "step": 400
39
+ },
40
+ {
41
+ "epoch": 0.2059732234809475,
42
+ "grad_norm": 0.32863080501556396,
43
+ "learning_rate": 0.003,
44
+ "loss": 3.753446350097656,
45
+ "step": 500
46
+ },
47
+ {
48
+ "epoch": 0.2059732234809475,
49
+ "eval_loss": 3.6475272178649902,
50
+ "eval_runtime": 6.6256,
51
+ "eval_samples_per_second": 484.482,
52
+ "eval_steps_per_second": 60.673,
53
+ "step": 500
54
+ },
55
+ {
56
+ "epoch": 0.24716786817713698,
57
+ "grad_norm": 0.29980480670928955,
58
+ "learning_rate": 0.003,
59
+ "loss": 3.5713235473632814,
60
+ "step": 600
61
+ },
62
+ {
63
+ "epoch": 0.2883625128733265,
64
+ "grad_norm": 0.3114969730377197,
65
+ "learning_rate": 0.003,
66
+ "loss": 3.4531216430664062,
67
+ "step": 700
68
+ },
69
+ {
70
+ "epoch": 0.329557157569516,
71
+ "grad_norm": 0.27909472584724426,
72
+ "learning_rate": 0.003,
73
+ "loss": 3.3734033203125,
74
+ "step": 800
75
+ },
76
+ {
77
+ "epoch": 0.3707518022657055,
78
+ "grad_norm": 0.25308936834335327,
79
+ "learning_rate": 0.003,
80
+ "loss": 3.31923095703125,
81
+ "step": 900
82
+ },
83
+ {
84
+ "epoch": 0.411946446961895,
85
+ "grad_norm": 0.4033835828304291,
86
+ "learning_rate": 0.003,
87
+ "loss": 3.2776348876953123,
88
+ "step": 1000
89
+ },
90
+ {
91
+ "epoch": 0.411946446961895,
92
+ "eval_loss": 3.2628297805786133,
93
+ "eval_runtime": 7.2919,
94
+ "eval_samples_per_second": 440.216,
95
+ "eval_steps_per_second": 55.13,
96
+ "step": 1000
97
+ },
98
+ {
99
+ "epoch": 0.45314109165808447,
100
+ "grad_norm": 0.2694193720817566,
101
+ "learning_rate": 0.003,
102
+ "loss": 3.24587158203125,
103
+ "step": 1100
104
+ },
105
+ {
106
+ "epoch": 0.49433573635427397,
107
+ "grad_norm": 0.384750097990036,
108
+ "learning_rate": 0.003,
109
+ "loss": 3.2172216796875,
110
+ "step": 1200
111
+ },
112
+ {
113
+ "epoch": 0.5355303810504635,
114
+ "grad_norm": 0.26695516705513,
115
+ "learning_rate": 0.003,
116
+ "loss": 3.1926251220703126,
117
+ "step": 1300
118
+ },
119
+ {
120
+ "epoch": 0.576725025746653,
121
+ "grad_norm": 0.30025020241737366,
122
+ "learning_rate": 0.003,
123
+ "loss": 3.176413269042969,
124
+ "step": 1400
125
+ },
126
+ {
127
+ "epoch": 0.6179196704428425,
128
+ "grad_norm": 0.25509148836135864,
129
+ "learning_rate": 0.003,
130
+ "loss": 3.1571221923828126,
131
+ "step": 1500
132
+ },
133
+ {
134
+ "epoch": 0.6179196704428425,
135
+ "eval_loss": 3.144334077835083,
136
+ "eval_runtime": 6.7143,
137
+ "eval_samples_per_second": 478.084,
138
+ "eval_steps_per_second": 59.872,
139
+ "step": 1500
140
+ },
141
+ {
142
+ "epoch": 0.659114315139032,
143
+ "grad_norm": 0.399799108505249,
144
+ "learning_rate": 0.003,
145
+ "loss": 3.1416632080078126,
146
+ "step": 1600
147
+ },
148
+ {
149
+ "epoch": 0.7003089598352215,
150
+ "grad_norm": 0.25501111149787903,
151
+ "learning_rate": 0.003,
152
+ "loss": 3.128678894042969,
153
+ "step": 1700
154
+ },
155
+ {
156
+ "epoch": 0.741503604531411,
157
+ "grad_norm": 0.26887112855911255,
158
+ "learning_rate": 0.003,
159
+ "loss": 3.1167843627929686,
160
+ "step": 1800
161
+ },
162
+ {
163
+ "epoch": 0.7826982492276005,
164
+ "grad_norm": 0.26506391167640686,
165
+ "learning_rate": 0.003,
166
+ "loss": 3.101407775878906,
167
+ "step": 1900
168
+ },
169
+ {
170
+ "epoch": 0.82389289392379,
171
+ "grad_norm": 0.2081276923418045,
172
+ "learning_rate": 0.002899325557098001,
173
+ "loss": 3.0907388305664063,
174
+ "step": 2000
175
+ },
176
+ {
177
+ "epoch": 0.82389289392379,
178
+ "eval_loss": 3.0802929401397705,
179
+ "eval_runtime": 6.4594,
180
+ "eval_samples_per_second": 496.952,
181
+ "eval_steps_per_second": 62.235,
182
+ "step": 2000
183
+ },
184
+ {
185
+ "epoch": 0.8650875386199794,
186
+ "grad_norm": 0.18226471543312073,
187
+ "learning_rate": 0.0022915870802995226,
188
+ "loss": 3.0740560913085937,
189
+ "step": 2100
190
+ },
191
+ {
192
+ "epoch": 0.9062821833161689,
193
+ "grad_norm": 0.16565178334712982,
194
+ "learning_rate": 0.0013644373889485982,
195
+ "loss": 3.050361328125,
196
+ "step": 2200
197
+ },
198
+ {
199
+ "epoch": 0.9474768280123584,
200
+ "grad_norm": 0.1562734991312027,
201
+ "learning_rate": 0.0004919882091667163,
202
+ "loss": 3.012582092285156,
203
+ "step": 2300
204
+ },
205
+ {
206
+ "epoch": 0.9886714727085479,
207
+ "grad_norm": 0.1223239079117775,
208
+ "learning_rate": 2.6279207952956684e-05,
209
+ "loss": 2.992548522949219,
210
+ "step": 2400
211
+ },
212
+ {
213
+ "epoch": 1.0,
214
+ "eval_loss": 2.982126474380493,
215
+ "eval_runtime": 6.3389,
216
+ "eval_samples_per_second": 506.4,
217
+ "eval_steps_per_second": 63.418,
218
+ "step": 2428
219
+ }
220
+ ],
221
+ "logging_steps": 100,
222
+ "max_steps": 2428,
223
+ "num_input_tokens_seen": 0,
224
+ "num_train_epochs": 1,
225
+ "save_steps": 500,
226
+ "stateful_callbacks": {
227
+ "TrainerControl": {
228
+ "args": {
229
+ "should_epoch_stop": false,
230
+ "should_evaluate": false,
231
+ "should_log": false,
232
+ "should_save": true,
233
+ "should_training_stop": true
234
+ },
235
+ "attributes": {}
236
+ }
237
+ },
238
+ "total_flos": 9979821293568000.0,
239
+ "train_batch_size": 400,
240
+ "trial_name": null,
241
+ "trial_params": null
242
+ }