Hugging Face's logo

seba
/
VoxCPM2ANE-Preview

Text-to-Speech
Core ML
CoreML
ANE
AppleNeuralEngine
NeuralEngine
TTS
Model card Files
xet
 Community
seba commited on
Commit
3c292ce
·
verified ·
1 Parent(s): 31d8f64

Upload audio_vae_decoder_lf4.mlmodelc

Browse files
Files changed (5) hide show
  1. audio_vae_decoder_lf4.mlmodelc/analytics/coremldata.bin +3 -0
  2. audio_vae_decoder_lf4.mlmodelc/coremldata.bin +3 -0
  3. audio_vae_decoder_lf4.mlmodelc/metadata.json +356 -0
  4. audio_vae_decoder_lf4.mlmodelc/model.mil +0 -0
  5. audio_vae_decoder_lf4.mlmodelc/weights/weight.bin +3 -0
audio_vae_decoder_lf4.mlmodelc/analytics/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:80433465ac6586223d87c0f1949bb8afda666ade71a6533f790b66aab512fa4a
3
+ size 243
audio_vae_decoder_lf4.mlmodelc/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c54e37d43090e73b2aa38d7b67fd50f9a088c1352ac46702b0968c4a7b2e59d6
3
+ size 1162
audio_vae_decoder_lf4.mlmodelc/metadata.json ADDED
@@ -0,0 +1,356 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "metadataOutputVersion" : "3.0",
4
+ "storagePrecision" : "Float16",
5
+ "outputSchema" : [
6
+ {
7
+ "hasShapeFlexibility" : "0",
8
+ "isOptional" : "0",
9
+ "dataType" : "Float32",
10
+ "formattedType" : "MultiArray (Float32)",
11
+ "shortDescription" : "",
12
+ "shape" : "[]",
13
+ "name" : "audio",
14
+ "type" : "MultiArray"
15
+ },
16
+ {
17
+ "hasShapeFlexibility" : "0",
18
+ "isOptional" : "0",
19
+ "dataType" : "Float32",
20
+ "formattedType" : "MultiArray (Float32)",
21
+ "shortDescription" : "",
22
+ "shape" : "[]",
23
+ "name" : "new_cache_0",
24
+ "type" : "MultiArray"
25
+ },
26
+ {
27
+ "hasShapeFlexibility" : "0",
28
+ "isOptional" : "0",
29
+ "dataType" : "Float32",
30
+ "formattedType" : "MultiArray (Float32 1 × 2048 × 1)",
31
+ "shortDescription" : "",
32
+ "shape" : "[1, 2048, 1]",
33
+ "name" : "new_cache_1",
34
+ "type" : "MultiArray"
35
+ },
36
+ {
37
+ "hasShapeFlexibility" : "0",
38
+ "isOptional" : "0",
39
+ "dataType" : "Float32",
40
+ "formattedType" : "MultiArray (Float32)",
41
+ "shortDescription" : "",
42
+ "shape" : "[]",
43
+ "name" : "new_cache_2",
44
+ "type" : "MultiArray"
45
+ },
46
+ {
47
+ "hasShapeFlexibility" : "0",
48
+ "isOptional" : "0",
49
+ "dataType" : "Float32",
50
+ "formattedType" : "MultiArray (Float32 1 × 1024 × 1)",
51
+ "shortDescription" : "",
52
+ "shape" : "[1, 1024, 1]",
53
+ "name" : "new_cache_3",
54
+ "type" : "MultiArray"
55
+ },
56
+ {
57
+ "hasShapeFlexibility" : "0",
58
+ "isOptional" : "0",
59
+ "dataType" : "Float32",
60
+ "formattedType" : "MultiArray (Float32)",
61
+ "shortDescription" : "",
62
+ "shape" : "[]",
63
+ "name" : "new_cache_4",
64
+ "type" : "MultiArray"
65
+ },
66
+ {
67
+ "hasShapeFlexibility" : "0",
68
+ "isOptional" : "0",
69
+ "dataType" : "Float32",
70
+ "formattedType" : "MultiArray (Float32 1 × 512 × 1)",
71
+ "shortDescription" : "",
72
+ "shape" : "[1, 512, 1]",
73
+ "name" : "new_cache_5",
74
+ "type" : "MultiArray"
75
+ },
76
+ {
77
+ "hasShapeFlexibility" : "0",
78
+ "isOptional" : "0",
79
+ "dataType" : "Float32",
80
+ "formattedType" : "MultiArray (Float32)",
81
+ "shortDescription" : "",
82
+ "shape" : "[]",
83
+ "name" : "new_cache_6",
84
+ "type" : "MultiArray"
85
+ },
86
+ {
87
+ "hasShapeFlexibility" : "0",
88
+ "isOptional" : "0",
89
+ "dataType" : "Float32",
90
+ "formattedType" : "MultiArray (Float32 1 × 256 × 1)",
91
+ "shortDescription" : "",
92
+ "shape" : "[1, 256, 1]",
93
+ "name" : "new_cache_7",
94
+ "type" : "MultiArray"
95
+ },
96
+ {
97
+ "hasShapeFlexibility" : "0",
98
+ "isOptional" : "0",
99
+ "dataType" : "Float32",
100
+ "formattedType" : "MultiArray (Float32)",
101
+ "shortDescription" : "",
102
+ "shape" : "[]",
103
+ "name" : "new_cache_8",
104
+ "type" : "MultiArray"
105
+ },
106
+ {
107
+ "hasShapeFlexibility" : "0",
108
+ "isOptional" : "0",
109
+ "dataType" : "Float32",
110
+ "formattedType" : "MultiArray (Float32 1 × 128 × 1)",
111
+ "shortDescription" : "",
112
+ "shape" : "[1, 128, 1]",
113
+ "name" : "new_cache_9",
114
+ "type" : "MultiArray"
115
+ },
116
+ {
117
+ "hasShapeFlexibility" : "0",
118
+ "isOptional" : "0",
119
+ "dataType" : "Float32",
120
+ "formattedType" : "MultiArray (Float32)",
121
+ "shortDescription" : "",
122
+ "shape" : "[]",
123
+ "name" : "new_cache_10",
124
+ "type" : "MultiArray"
125
+ },
126
+ {
127
+ "hasShapeFlexibility" : "0",
128
+ "isOptional" : "0",
129
+ "dataType" : "Float32",
130
+ "formattedType" : "MultiArray (Float32 1 × 64 × 1)",
131
+ "shortDescription" : "",
132
+ "shape" : "[1, 64, 1]",
133
+ "name" : "new_cache_11",
134
+ "type" : "MultiArray"
135
+ },
136
+ {
137
+ "hasShapeFlexibility" : "0",
138
+ "isOptional" : "0",
139
+ "dataType" : "Float32",
140
+ "formattedType" : "MultiArray (Float32)",
141
+ "shortDescription" : "",
142
+ "shape" : "[]",
143
+ "name" : "new_cache_12",
144
+ "type" : "MultiArray"
145
+ },
146
+ {
147
+ "hasShapeFlexibility" : "0",
148
+ "isOptional" : "0",
149
+ "dataType" : "Float32",
150
+ "formattedType" : "MultiArray (Float32)",
151
+ "shortDescription" : "",
152
+ "shape" : "[]",
153
+ "name" : "new_cache_13",
154
+ "type" : "MultiArray"
155
+ }
156
+ ],
157
+ "modelParameters" : [
158
+
159
+ ],
160
+ "specificationVersion" : 9,
161
+ "mlProgramOperationTypeHistogram" : {
162
+ "Ios18.sliceByIndex" : 32,
163
+ "Ios18.add" : 66,
164
+ "Ios18.convTranspose" : 6,
165
+ "Ios18.tanh" : 1,
166
+ "Ios18.sin" : 43,
167
+ "Ios18.concat" : 32,
168
+ "Split" : 6,
169
+ "Ios18.pow" : 43,
170
+ "Ios18.cast" : 30,
171
+ "Ios18.mul" : 91,
172
+ "Ios18.conv" : 39
173
+ },
174
+ "computePrecision" : "Mixed (Float16, Float32, Int32)",
175
+ "isUpdatable" : "0",
176
+ "stateSchema" : [
177
+
178
+ ],
179
+ "availability" : {
180
+ "macOS" : "15.0",
181
+ "tvOS" : "18.0",
182
+ "visionOS" : "2.0",
183
+ "watchOS" : "11.0",
184
+ "iOS" : "18.0",
185
+ "macCatalyst" : "18.0"
186
+ },
187
+ "modelType" : {
188
+ "name" : "MLModelType_mlProgram"
189
+ },
190
+ "userDefinedMetadata" : {
191
+ "com.github.apple.coremltools.version" : "9.0",
192
+ "com.github.apple.coremltools.source_dialect" : "TorchScript",
193
+ "com.github.apple.coremltools.conversion_date" : "2026-05-15",
194
+ "max_latent_frames" : "16",
195
+ "com.github.apple.coremltools.source" : "torch==2.11.0",
196
+ "sr_cond" : "48000",
197
+ "min_latent_frames" : "1"
198
+ },
199
+ "inputSchema" : [
200
+ {
201
+ "dataType" : "Float32",
202
+ "hasShapeFlexibility" : "1",
203
+ "isOptional" : "0",
204
+ "shapeFlexibility" : "1 × 64 × 1...16",
205
+ "shapeRange" : "[[1, 1], [64, 64], [1, 16]]",
206
+ "formattedType" : "MultiArray (Float32 1 × 64 × 4)",
207
+ "type" : "MultiArray",
208
+ "shape" : "[1, 64, 4]",
209
+ "name" : "z",
210
+ "shortDescription" : ""
211
+ },
212
+ {
213
+ "hasShapeFlexibility" : "0",
214
+ "isOptional" : "0",
215
+ "dataType" : "Float32",
216
+ "formattedType" : "MultiArray (Float32 1 × 64 × 6)",
217
+ "shortDescription" : "",
218
+ "shape" : "[1, 64, 6]",
219
+ "name" : "cache_0",
220
+ "type" : "MultiArray"
221
+ },
222
+ {
223
+ "hasShapeFlexibility" : "0",
224
+ "isOptional" : "0",
225
+ "dataType" : "Float32",
226
+ "formattedType" : "MultiArray (Float32 1 × 2048 × 1)",
227
+ "shortDescription" : "",
228
+ "shape" : "[1, 2048, 1]",
229
+ "name" : "cache_1",
230
+ "type" : "MultiArray"
231
+ },
232
+ {
233
+ "hasShapeFlexibility" : "0",
234
+ "isOptional" : "0",
235
+ "dataType" : "Float32",
236
+ "formattedType" : "MultiArray (Float32 1 × 1024 × 78)",
237
+ "shortDescription" : "",
238
+ "shape" : "[1, 1024, 78]",
239
+ "name" : "cache_2",
240
+ "type" : "MultiArray"
241
+ },
242
+ {
243
+ "hasShapeFlexibility" : "0",
244
+ "isOptional" : "0",
245
+ "dataType" : "Float32",
246
+ "formattedType" : "MultiArray (Float32 1 × 1024 × 1)",
247
+ "shortDescription" : "",
248
+ "shape" : "[1, 1024, 1]",
249
+ "name" : "cache_3",
250
+ "type" : "MultiArray"
251
+ },
252
+ {
253
+ "hasShapeFlexibility" : "0",
254
+ "isOptional" : "0",
255
+ "dataType" : "Float32",
256
+ "formattedType" : "MultiArray (Float32 1 × 512 × 78)",
257
+ "shortDescription" : "",
258
+ "shape" : "[1, 512, 78]",
259
+ "name" : "cache_4",
260
+ "type" : "MultiArray"
261
+ },
262
+ {
263
+ "hasShapeFlexibility" : "0",
264
+ "isOptional" : "0",
265
+ "dataType" : "Float32",
266
+ "formattedType" : "MultiArray (Float32 1 × 512 × 1)",
267
+ "shortDescription" : "",
268
+ "shape" : "[1, 512, 1]",
269
+ "name" : "cache_5",
270
+ "type" : "MultiArray"
271
+ },
272
+ {
273
+ "hasShapeFlexibility" : "0",
274
+ "isOptional" : "0",
275
+ "dataType" : "Float32",
276
+ "formattedType" : "MultiArray (Float32 1 × 256 × 78)",
277
+ "shortDescription" : "",
278
+ "shape" : "[1, 256, 78]",
279
+ "name" : "cache_6",
280
+ "type" : "MultiArray"
281
+ },
282
+ {
283
+ "hasShapeFlexibility" : "0",
284
+ "isOptional" : "0",
285
+ "dataType" : "Float32",
286
+ "formattedType" : "MultiArray (Float32 1 × 256 × 1)",
287
+ "shortDescription" : "",
288
+ "shape" : "[1, 256, 1]",
289
+ "name" : "cache_7",
290
+ "type" : "MultiArray"
291
+ },
292
+ {
293
+ "hasShapeFlexibility" : "0",
294
+ "isOptional" : "0",
295
+ "dataType" : "Float32",
296
+ "formattedType" : "MultiArray (Float32 1 × 128 × 78)",
297
+ "shortDescription" : "",
298
+ "shape" : "[1, 128, 78]",
299
+ "name" : "cache_8",
300
+ "type" : "MultiArray"
301
+ },
302
+ {
303
+ "hasShapeFlexibility" : "0",
304
+ "isOptional" : "0",
305
+ "dataType" : "Float32",
306
+ "formattedType" : "MultiArray (Float32 1 × 128 × 1)",
307
+ "shortDescription" : "",
308
+ "shape" : "[1, 128, 1]",
309
+ "name" : "cache_9",
310
+ "type" : "MultiArray"
311
+ },
312
+ {
313
+ "hasShapeFlexibility" : "0",
314
+ "isOptional" : "0",
315
+ "dataType" : "Float32",
316
+ "formattedType" : "MultiArray (Float32 1 × 64 × 78)",
317
+ "shortDescription" : "",
318
+ "shape" : "[1, 64, 78]",
319
+ "name" : "cache_10",
320
+ "type" : "MultiArray"
321
+ },
322
+ {
323
+ "hasShapeFlexibility" : "0",
324
+ "isOptional" : "0",
325
+ "dataType" : "Float32",
326
+ "formattedType" : "MultiArray (Float32 1 × 64 × 1)",
327
+ "shortDescription" : "",
328
+ "shape" : "[1, 64, 1]",
329
+ "name" : "cache_11",
330
+ "type" : "MultiArray"
331
+ },
332
+ {
333
+ "hasShapeFlexibility" : "0",
334
+ "isOptional" : "0",
335
+ "dataType" : "Float32",
336
+ "formattedType" : "MultiArray (Float32 1 × 32 × 78)",
337
+ "shortDescription" : "",
338
+ "shape" : "[1, 32, 78]",
339
+ "name" : "cache_12",
340
+ "type" : "MultiArray"
341
+ },
342
+ {
343
+ "hasShapeFlexibility" : "0",
344
+ "isOptional" : "0",
345
+ "dataType" : "Float32",
346
+ "formattedType" : "MultiArray (Float32 1 × 32 × 6)",
347
+ "shortDescription" : "",
348
+ "shape" : "[1, 32, 6]",
349
+ "name" : "cache_13",
350
+ "type" : "MultiArray"
351
+ }
352
+ ],
353
+ "generatedClassName" : "audio_vae_decoder_rangedim_1_16",
354
+ "method" : "predict"
355
+ }
356
+ ]
audio_vae_decoder_lf4.mlmodelc/model.mil ADDED
The diff for this file is too large to render. See raw diff
 
audio_vae_decoder_lf4.mlmodelc/weights/weight.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ae6c42806a29cb06485b388ddce932dbe4c708540b7e31f018df543d2797196
3
+ size 91508928