yujiepan commited on
Commit
f9d7d40
·
verified ·
1 Parent(s): fca9437

Upload folder using huggingface_hub

Browse files
Files changed (1) hide show
  1. .meta/model.log +274 -0
.meta/model.log CHANGED
@@ -0,0 +1,274 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ HaruModel(
2
+ (lm): Qwen3_5TextModel(
3
+ (embed_tokens): Embedding(248320, 32)
4
+ (layers): ModuleList(
5
+ (0-2): 3 x Qwen3_5DecoderLayer(
6
+ (linear_attn): Qwen3_5GatedDeltaNet(
7
+ (conv1d): Conv1d(128, 128, kernel_size=(4,), stride=(1,), padding=(3,), groups=128, bias=False)
8
+ (norm): Qwen3_5RMSNormGated()
9
+ (out_proj): Linear(in_features=64, out_features=32, bias=False)
10
+ (in_proj_qkv): Linear(in_features=32, out_features=128, bias=False)
11
+ (in_proj_z): Linear(in_features=32, out_features=64, bias=False)
12
+ (in_proj_b): Linear(in_features=32, out_features=4, bias=False)
13
+ (in_proj_a): Linear(in_features=32, out_features=4, bias=False)
14
+ )
15
+ (mlp): Qwen3_5MLP(
16
+ (gate_proj): Linear(in_features=32, out_features=64, bias=False)
17
+ (up_proj): Linear(in_features=32, out_features=64, bias=False)
18
+ (down_proj): Linear(in_features=64, out_features=32, bias=False)
19
+ (act_fn): SiLUActivation()
20
+ )
21
+ (input_layernorm): Qwen3_5RMSNorm((32,), eps=1e-06)
22
+ (post_attention_layernorm): Qwen3_5RMSNorm((32,), eps=1e-06)
23
+ )
24
+ (3): Qwen3_5DecoderLayer(
25
+ (self_attn): Qwen3_5Attention(
26
+ (q_proj): Linear(in_features=32, out_features=128, bias=False)
27
+ (k_proj): Linear(in_features=32, out_features=16, bias=False)
28
+ (v_proj): Linear(in_features=32, out_features=16, bias=False)
29
+ (o_proj): Linear(in_features=64, out_features=32, bias=False)
30
+ (q_norm): Qwen3_5RMSNorm((16,), eps=1e-06)
31
+ (k_norm): Qwen3_5RMSNorm((16,), eps=1e-06)
32
+ )
33
+ (mlp): Qwen3_5MLP(
34
+ (gate_proj): Linear(in_features=32, out_features=64, bias=False)
35
+ (up_proj): Linear(in_features=32, out_features=64, bias=False)
36
+ (down_proj): Linear(in_features=64, out_features=32, bias=False)
37
+ (act_fn): SiLUActivation()
38
+ )
39
+ (input_layernorm): Qwen3_5RMSNorm((32,), eps=1e-06)
40
+ (post_attention_layernorm): Qwen3_5RMSNorm((32,), eps=1e-06)
41
+ )
42
+ (4-6): 3 x Qwen3_5DecoderLayer(
43
+ (linear_attn): Qwen3_5GatedDeltaNet(
44
+ (conv1d): Conv1d(128, 128, kernel_size=(4,), stride=(1,), padding=(3,), groups=128, bias=False)
45
+ (norm): Qwen3_5RMSNormGated()
46
+ (out_proj): Linear(in_features=64, out_features=32, bias=False)
47
+ (in_proj_qkv): Linear(in_features=32, out_features=128, bias=False)
48
+ (in_proj_z): Linear(in_features=32, out_features=64, bias=False)
49
+ (in_proj_b): Linear(in_features=32, out_features=4, bias=False)
50
+ (in_proj_a): Linear(in_features=32, out_features=4, bias=False)
51
+ )
52
+ (mlp): Qwen3_5MLP(
53
+ (gate_proj): Linear(in_features=32, out_features=64, bias=False)
54
+ (up_proj): Linear(in_features=32, out_features=64, bias=False)
55
+ (down_proj): Linear(in_features=64, out_features=32, bias=False)
56
+ (act_fn): SiLUActivation()
57
+ )
58
+ (input_layernorm): Qwen3_5RMSNorm((32,), eps=1e-06)
59
+ (post_attention_layernorm): Qwen3_5RMSNorm((32,), eps=1e-06)
60
+ )
61
+ (7): Qwen3_5DecoderLayer(
62
+ (self_attn): Qwen3_5Attention(
63
+ (q_proj): Linear(in_features=32, out_features=128, bias=False)
64
+ (k_proj): Linear(in_features=32, out_features=16, bias=False)
65
+ (v_proj): Linear(in_features=32, out_features=16, bias=False)
66
+ (o_proj): Linear(in_features=64, out_features=32, bias=False)
67
+ (q_norm): Qwen3_5RMSNorm((16,), eps=1e-06)
68
+ (k_norm): Qwen3_5RMSNorm((16,), eps=1e-06)
69
+ )
70
+ (mlp): Qwen3_5MLP(
71
+ (gate_proj): Linear(in_features=32, out_features=64, bias=False)
72
+ (up_proj): Linear(in_features=32, out_features=64, bias=False)
73
+ (down_proj): Linear(in_features=64, out_features=32, bias=False)
74
+ (act_fn): SiLUActivation()
75
+ )
76
+ (input_layernorm): Qwen3_5RMSNorm((32,), eps=1e-06)
77
+ (post_attention_layernorm): Qwen3_5RMSNorm((32,), eps=1e-06)
78
+ )
79
+ )
80
+ (norm): Qwen3_5RMSNorm((32,), eps=1e-06)
81
+ (rotary_emb): Qwen3_5TextRotaryEmbedding()
82
+ )
83
+ (model_audio_proj): AudioProjector(
84
+ (fc1): Linear(in_features=64, out_features=32, bias=True)
85
+ (norm): RMSNorm((32,), eps=1e-06, elementwise_affine=True)
86
+ (fc2): Linear(in_features=32, out_features=32, bias=True)
87
+ )
88
+ (user_audio_proj): AudioProjector(
89
+ (fc1): Linear(in_features=64, out_features=32, bias=True)
90
+ (norm): RMSNorm((32,), eps=1e-06, elementwise_affine=True)
91
+ (fc2): Linear(in_features=32, out_features=32, bias=True)
92
+ )
93
+ (head): MeanFlowHead(
94
+ (latent_proj): Linear(in_features=64, out_features=32, bias=False)
95
+ (cond_proj): Linear(in_features=32, out_features=32, bias=False)
96
+ (t_embed): ScalarEmbed(
97
+ (fc1): Linear(in_features=256, out_features=32, bias=False)
98
+ (fc2): Linear(in_features=32, out_features=32, bias=False)
99
+ )
100
+ (dt_embed): ScalarEmbed(
101
+ (fc1): Linear(in_features=256, out_features=32, bias=False)
102
+ (fc2): Linear(in_features=32, out_features=32, bias=False)
103
+ )
104
+ (h_embed): ScalarEmbed(
105
+ (fc1): Linear(in_features=256, out_features=32, bias=False)
106
+ (fc2): Linear(in_features=32, out_features=32, bias=False)
107
+ )
108
+ (omega_embed): ScalarEmbed(
109
+ (fc1): Linear(in_features=256, out_features=32, bias=False)
110
+ (fc2): Linear(in_features=32, out_features=32, bias=False)
111
+ )
112
+ (layers): ModuleList(
113
+ (0): HeadLayer(
114
+ (norm): Module()
115
+ (ada): Linear(in_features=32, out_features=96, bias=False)
116
+ (ffn): FeedForward(
117
+ (gate_proj): Linear(in_features=32, out_features=96, bias=False)
118
+ (up_proj): Linear(in_features=32, out_features=96, bias=False)
119
+ (down_proj): Linear(in_features=96, out_features=32, bias=False)
120
+ )
121
+ )
122
+ )
123
+ (final): FinalLayer(
124
+ (ada): Linear(in_features=32, out_features=64, bias=False)
125
+ (proj): Linear(in_features=32, out_features=64, bias=False)
126
+ )
127
+ )
128
+ (codec): VibeVoiceCodec(
129
+ (encoder): Encoder(
130
+ (stem): CausalConv1d()
131
+ (down): ModuleList(
132
+ (0-5): 6 x CausalConv1d()
133
+ )
134
+ (stages): ModuleList(
135
+ (0): ModuleList(
136
+ (0): Block(
137
+ (norm): RMSNorm()
138
+ (conv): CausalConv1d()
139
+ (ffn_norm): RMSNorm()
140
+ (fc1): Linear(in_features=1, out_features=4, bias=True)
141
+ (fc2): Linear(in_features=4, out_features=1, bias=True)
142
+ )
143
+ )
144
+ (1): ModuleList(
145
+ (0): Block(
146
+ (norm): RMSNorm()
147
+ (conv): CausalConv1d()
148
+ (ffn_norm): RMSNorm()
149
+ (fc1): Linear(in_features=2, out_features=8, bias=True)
150
+ (fc2): Linear(in_features=8, out_features=2, bias=True)
151
+ )
152
+ )
153
+ (2): ModuleList(
154
+ (0): Block(
155
+ (norm): RMSNorm()
156
+ (conv): CausalConv1d()
157
+ (ffn_norm): RMSNorm()
158
+ (fc1): Linear(in_features=4, out_features=16, bias=True)
159
+ (fc2): Linear(in_features=16, out_features=4, bias=True)
160
+ )
161
+ )
162
+ (3): ModuleList(
163
+ (0): Block(
164
+ (norm): RMSNorm()
165
+ (conv): CausalConv1d()
166
+ (ffn_norm): RMSNorm()
167
+ (fc1): Linear(in_features=8, out_features=32, bias=True)
168
+ (fc2): Linear(in_features=32, out_features=8, bias=True)
169
+ )
170
+ )
171
+ (4): ModuleList(
172
+ (0): Block(
173
+ (norm): RMSNorm()
174
+ (conv): CausalConv1d()
175
+ (ffn_norm): RMSNorm()
176
+ (fc1): Linear(in_features=16, out_features=64, bias=True)
177
+ (fc2): Linear(in_features=64, out_features=16, bias=True)
178
+ )
179
+ )
180
+ (5): ModuleList(
181
+ (0): Block(
182
+ (norm): RMSNorm()
183
+ (conv): CausalConv1d()
184
+ (ffn_norm): RMSNorm()
185
+ (fc1): Linear(in_features=32, out_features=128, bias=True)
186
+ (fc2): Linear(in_features=128, out_features=32, bias=True)
187
+ )
188
+ )
189
+ (6): ModuleList(
190
+ (0): Block(
191
+ (norm): RMSNorm()
192
+ (conv): CausalConv1d()
193
+ (ffn_norm): RMSNorm()
194
+ (fc1): Linear(in_features=64, out_features=256, bias=True)
195
+ (fc2): Linear(in_features=256, out_features=64, bias=True)
196
+ )
197
+ )
198
+ )
199
+ (head): CausalConv1d()
200
+ )
201
+ (decoder): Decoder(
202
+ (stem): CausalConv1d()
203
+ (up): ModuleList(
204
+ (0-5): 6 x CausalConvTranspose1d()
205
+ )
206
+ (stages): ModuleList(
207
+ (0): ModuleList(
208
+ (0): Block(
209
+ (norm): RMSNorm()
210
+ (conv): CausalConv1d()
211
+ (ffn_norm): RMSNorm()
212
+ (fc1): Linear(in_features=64, out_features=256, bias=True)
213
+ (fc2): Linear(in_features=256, out_features=64, bias=True)
214
+ )
215
+ )
216
+ (1): ModuleList(
217
+ (0): Block(
218
+ (norm): RMSNorm()
219
+ (conv): CausalConv1d()
220
+ (ffn_norm): RMSNorm()
221
+ (fc1): Linear(in_features=32, out_features=128, bias=True)
222
+ (fc2): Linear(in_features=128, out_features=32, bias=True)
223
+ )
224
+ )
225
+ (2): ModuleList(
226
+ (0): Block(
227
+ (norm): RMSNorm()
228
+ (conv): CausalConv1d()
229
+ (ffn_norm): RMSNorm()
230
+ (fc1): Linear(in_features=16, out_features=64, bias=True)
231
+ (fc2): Linear(in_features=64, out_features=16, bias=True)
232
+ )
233
+ )
234
+ (3): ModuleList(
235
+ (0): Block(
236
+ (norm): RMSNorm()
237
+ (conv): CausalConv1d()
238
+ (ffn_norm): RMSNorm()
239
+ (fc1): Linear(in_features=8, out_features=32, bias=True)
240
+ (fc2): Linear(in_features=32, out_features=8, bias=True)
241
+ )
242
+ )
243
+ (4): ModuleList(
244
+ (0): Block(
245
+ (norm): RMSNorm()
246
+ (conv): CausalConv1d()
247
+ (ffn_norm): RMSNorm()
248
+ (fc1): Linear(in_features=4, out_features=16, bias=True)
249
+ (fc2): Linear(in_features=16, out_features=4, bias=True)
250
+ )
251
+ )
252
+ (5): ModuleList(
253
+ (0): Block(
254
+ (norm): RMSNorm()
255
+ (conv): CausalConv1d()
256
+ (ffn_norm): RMSNorm()
257
+ (fc1): Linear(in_features=2, out_features=8, bias=True)
258
+ (fc2): Linear(in_features=8, out_features=2, bias=True)
259
+ )
260
+ )
261
+ (6): ModuleList(
262
+ (0): Block(
263
+ (norm): RMSNorm()
264
+ (conv): CausalConv1d()
265
+ (ffn_norm): RMSNorm()
266
+ (fc1): Linear(in_features=1, out_features=4, bias=True)
267
+ (fc2): Linear(in_features=4, out_features=1, bias=True)
268
+ )
269
+ )
270
+ )
271
+ (head): CausalConv1d()
272
+ )
273
+ )
274
+ )