Hugging Face's logo

webxos
/
microd_v2

English
micro-distillery
custom
bigram
grpo
vae
distilled
smollm2
qwen2.5
olmo
tinyllama
Model card Files
xet
 Community
webxos commited on
Commit
55fdc4a
·
verified ·
1 Parent(s): bb3afee

Delete pytorch_model.bin

Browse files
Files changed (1) hide show
  1. pytorch_model.bin +0 -222
pytorch_model.bin DELETED
@@ -1,222 +0,0 @@
1
- {
2
- "metadata": {
3
- "format": "torch",
4
- "version": "1.0",
5
- "model": "micro-distill-grpo-vae",
6
- "hidden_size": 512,
7
- "num_layers": 8,
8
- "num_heads": 8,
9
- "vocab_size": 50257,
10
- "training_steps": 100
11
- },
12
- "tensors": {
13
- "transformer.wte.weight": {
14
- "shape": [
15
- 50257,
16
- 512
17
- ],
18
- "dtype": "float32",
19
- "size": "98.2 MB"
20
- },
21
- "transformer.wpe.weight": {
22
- "shape": [
23
- 1024,
24
- 512
25
- ],
26
- "dtype": "float32",
27
- "size": "2.0 MB"
28
- },
29
- "transformer.h.0.ln_1.weight": {
30
- "shape": [
31
- 512
32
- ],
33
- "dtype": "float32",
34
- "size": "2.0 KB"
35
- },
36
- "transformer.h.0.attn.c_attn.weight": {
37
- "shape": [
38
- 512,
39
- 1536
40
- ],
41
- "dtype": "float32",
42
- "size": "3.0 MB"
43
- },
44
- "transformer.h.0.mlp.c_fc.weight": {
45
- "shape": [
46
- 512,
47
- 2048
48
- ],
49
- "dtype": "float32",
50
- "size": "4.0 MB"
51
- },
52
- "transformer.h.1.ln_1.weight": {
53
- "shape": [
54
- 512
55
- ],
56
- "dtype": "float32",
57
- "size": "2.0 KB"
58
- },
59
- "transformer.h.1.attn.c_attn.weight": {
60
- "shape": [
61
- 512,
62
- 1536
63
- ],
64
- "dtype": "float32",
65
- "size": "3.0 MB"
66
- },
67
- "transformer.h.1.mlp.c_fc.weight": {
68
- "shape": [
69
- 512,
70
- 2048
71
- ],
72
- "dtype": "float32",
73
- "size": "4.0 MB"
74
- },
75
- "transformer.h.2.ln_1.weight": {
76
- "shape": [
77
- 512
78
- ],
79
- "dtype": "float32",
80
- "size": "2.0 KB"
81
- },
82
- "transformer.h.2.attn.c_attn.weight": {
83
- "shape": [
84
- 512,
85
- 1536
86
- ],
87
- "dtype": "float32",
88
- "size": "3.0 MB"
89
- },
90
- "transformer.h.2.mlp.c_fc.weight": {
91
- "shape": [
92
- 512,
93
- 2048
94
- ],
95
- "dtype": "float32",
96
- "size": "4.0 MB"
97
- },
98
- "transformer.h.3.ln_1.weight": {
99
- "shape": [
100
- 512
101
- ],
102
- "dtype": "float32",
103
- "size": "2.0 KB"
104
- },
105
- "transformer.h.3.attn.c_attn.weight": {
106
- "shape": [
107
- 512,
108
- 1536
109
- ],
110
- "dtype": "float32",
111
- "size": "3.0 MB"
112
- },
113
- "transformer.h.3.mlp.c_fc.weight": {
114
- "shape": [
115
- 512,
116
- 2048
117
- ],
118
- "dtype": "float32",
119
- "size": "4.0 MB"
120
- },
121
- "transformer.h.4.ln_1.weight": {
122
- "shape": [
123
- 512
124
- ],
125
- "dtype": "float32",
126
- "size": "2.0 KB"
127
- },
128
- "transformer.h.4.attn.c_attn.weight": {
129
- "shape": [
130
- 512,
131
- 1536
132
- ],
133
- "dtype": "float32",
134
- "size": "3.0 MB"
135
- },
136
- "transformer.h.4.mlp.c_fc.weight": {
137
- "shape": [
138
- 512,
139
- 2048
140
- ],
141
- "dtype": "float32",
142
- "size": "4.0 MB"
143
- },
144
- "transformer.h.5.ln_1.weight": {
145
- "shape": [
146
- 512
147
- ],
148
- "dtype": "float32",
149
- "size": "2.0 KB"
150
- },
151
- "transformer.h.5.attn.c_attn.weight": {
152
- "shape": [
153
- 512,
154
- 1536
155
- ],
156
- "dtype": "float32",
157
- "size": "3.0 MB"
158
- },
159
- "transformer.h.5.mlp.c_fc.weight": {
160
- "shape": [
161
- 512,
162
- 2048
163
- ],
164
- "dtype": "float32",
165
- "size": "4.0 MB"
166
- },
167
- "transformer.h.6.ln_1.weight": {
168
- "shape": [
169
- 512
170
- ],
171
- "dtype": "float32",
172
- "size": "2.0 KB"
173
- },
174
- "transformer.h.6.attn.c_attn.weight": {
175
- "shape": [
176
- 512,
177
- 1536
178
- ],
179
- "dtype": "float32",
180
- "size": "3.0 MB"
181
- },
182
- "transformer.h.6.mlp.c_fc.weight": {
183
- "shape": [
184
- 512,
185
- 2048
186
- ],
187
- "dtype": "float32",
188
- "size": "4.0 MB"
189
- },
190
- "transformer.h.7.ln_1.weight": {
191
- "shape": [
192
- 512
193
- ],
194
- "dtype": "float32",
195
- "size": "2.0 KB"
196
- },
197
- "transformer.h.7.attn.c_attn.weight": {
198
- "shape": [
199
- 512,
200
- 1536
201
- ],
202
- "dtype": "float32",
203
- "size": "3.0 MB"
204
- },
205
- "transformer.h.7.mlp.c_fc.weight": {
206
- "shape": [
207
- 512,
208
- 2048
209
- ],
210
- "dtype": "float32",
211
- "size": "4.0 MB"
212
- },
213
- "lm_head.weight": {
214
- "shape": [
215
- 50257,
216
- 512
217
- ],
218
- "dtype": "float32",
219
- "size": "98.2 MB"
220
- }
221
- }
222
- }