Hugging Face's logo

argmaxinc
/
ttskit-coreml

WhisperKit
Core ML
qwen
qwen3
qwen-tts
qwen3-tts
tts
apple
apple silicon
Model card Files
xet
 Community
3

Add W8A16-stream-multifunction SpeechDecoder

#3
by EduardoPacheco - opened
base: refs/heads/main
←
from: refs/pr/3
Discussion Files changed
+12092
-0
Files changed (10) hide show
  1. qwen3_tts/speech_decoder/12hz-0.6b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/analytics/coremldata.bin +3 -0
  2. qwen3_tts/speech_decoder/12hz-0.6b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/coremldata.bin +3 -0
  3. qwen3_tts/speech_decoder/12hz-0.6b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/metadata.json +579 -0
  4. qwen3_tts/speech_decoder/12hz-0.6b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/model.mil +0 -0
  5. qwen3_tts/speech_decoder/12hz-0.6b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/weights/weight.bin +3 -0
  6. qwen3_tts/speech_decoder/12hz-1.7b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/analytics/coremldata.bin +3 -0
  7. qwen3_tts/speech_decoder/12hz-1.7b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/coremldata.bin +3 -0
  8. qwen3_tts/speech_decoder/12hz-1.7b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/metadata.json +579 -0
  9. qwen3_tts/speech_decoder/12hz-1.7b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/model.mil +0 -0
  10. qwen3_tts/speech_decoder/12hz-1.7b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/weights/weight.bin +3 -0
qwen3_tts/speech_decoder/12hz-0.6b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/analytics/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3fd77fbdb6571721611661deb3bf3184aef6cb12b789235f6c04053b50f50f0f
3
+ size 243
qwen3_tts/speech_decoder/12hz-0.6b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8f40ab9923fffd07b7906bcfd55e79d5d85f2be729d6ea347826f2768b20e921
3
+ size 1115
qwen3_tts/speech_decoder/12hz-0.6b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/metadata.json ADDED
@@ -0,0 +1,579 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "metadataOutputVersion" : "3.0",
4
+ "storagePrecision" : "Mixed (Float16, Palettized (8 bits), UInt8)",
5
+ "outputSchema" : [
6
+ {
7
+ "hasShapeFlexibility" : "0",
8
+ "isOptional" : "0",
9
+ "dataType" : "Float16",
10
+ "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 1920)",
11
+ "shortDescription" : "",
12
+ "shape" : "[1, 1, 1, 1920]",
13
+ "name" : "audio",
14
+ "type" : "MultiArray"
15
+ },
16
+ {
17
+ "hasShapeFlexibility" : "0",
18
+ "isOptional" : "0",
19
+ "dataType" : "Float16",
20
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 1)",
21
+ "shortDescription" : "",
22
+ "shape" : "[1, 8192, 1, 1]",
23
+ "name" : "key_cache_updates",
24
+ "type" : "MultiArray"
25
+ },
26
+ {
27
+ "hasShapeFlexibility" : "0",
28
+ "isOptional" : "0",
29
+ "dataType" : "Float16",
30
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 1)",
31
+ "shortDescription" : "",
32
+ "shape" : "[1, 8192, 1, 1]",
33
+ "name" : "value_cache_updates",
34
+ "type" : "MultiArray"
35
+ },
36
+ {
37
+ "hasShapeFlexibility" : "0",
38
+ "isOptional" : "0",
39
+ "dataType" : "Float16",
40
+ "formattedType" : "MultiArray (Float16 1 × 1024 × 1 × 1)",
41
+ "shortDescription" : "",
42
+ "shape" : "[1, 1024, 1, 1]",
43
+ "name" : "hidden_context_update",
44
+ "type" : "MultiArray"
45
+ },
46
+ {
47
+ "hasShapeFlexibility" : "0",
48
+ "isOptional" : "0",
49
+ "dataType" : "Float16",
50
+ "formattedType" : "MultiArray (Float16 1 × 512 × 1 × 1)",
51
+ "shortDescription" : "",
52
+ "shape" : "[1, 512, 1, 1]",
53
+ "name" : "pre_conv_context_update",
54
+ "type" : "MultiArray"
55
+ }
56
+ ],
57
+ "modelParameters" : [
58
+
59
+ ],
60
+ "specificationVersion" : 9,
61
+ "functions" : [
62
+ {
63
+ "inputSchema" : [
64
+ {
65
+ "hasShapeFlexibility" : "0",
66
+ "isOptional" : "0",
67
+ "dataType" : "Int32",
68
+ "formattedType" : "MultiArray (Int32 1 × 16 × 1)",
69
+ "shortDescription" : "",
70
+ "shape" : "[1, 16, 1]",
71
+ "name" : "audio_codes",
72
+ "type" : "MultiArray"
73
+ },
74
+ {
75
+ "hasShapeFlexibility" : "0",
76
+ "isOptional" : "0",
77
+ "dataType" : "Int32",
78
+ "formattedType" : "MultiArray (Int32 1)",
79
+ "shortDescription" : "",
80
+ "shape" : "[1]",
81
+ "name" : "cache_length",
82
+ "type" : "MultiArray"
83
+ },
84
+ {
85
+ "hasShapeFlexibility" : "0",
86
+ "isOptional" : "0",
87
+ "dataType" : "Float16",
88
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 80)",
89
+ "shortDescription" : "",
90
+ "shape" : "[1, 8192, 1, 80]",
91
+ "name" : "key_cache",
92
+ "type" : "MultiArray"
93
+ },
94
+ {
95
+ "hasShapeFlexibility" : "0",
96
+ "isOptional" : "0",
97
+ "dataType" : "Float16",
98
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 80)",
99
+ "shortDescription" : "",
100
+ "shape" : "[1, 8192, 1, 80]",
101
+ "name" : "value_cache",
102
+ "type" : "MultiArray"
103
+ },
104
+ {
105
+ "hasShapeFlexibility" : "0",
106
+ "isOptional" : "0",
107
+ "dataType" : "Float16",
108
+ "formattedType" : "MultiArray (Float16 1 × 1 × 80)",
109
+ "shortDescription" : "",
110
+ "shape" : "[1, 1, 80]",
111
+ "name" : "kv_cache_update_mask",
112
+ "type" : "MultiArray"
113
+ },
114
+ {
115
+ "hasShapeFlexibility" : "0",
116
+ "isOptional" : "0",
117
+ "dataType" : "Float16",
118
+ "formattedType" : "MultiArray (Float16 1 × 80)",
119
+ "shortDescription" : "",
120
+ "shape" : "[1, 80]",
121
+ "name" : "key_padding_mask",
122
+ "type" : "MultiArray"
123
+ },
124
+ {
125
+ "hasShapeFlexibility" : "0",
126
+ "isOptional" : "0",
127
+ "dataType" : "Float16",
128
+ "formattedType" : "MultiArray (Float16 1 × 1024 × 1 × 8)",
129
+ "shortDescription" : "",
130
+ "shape" : "[1, 1024, 1, 8]",
131
+ "name" : "hidden_context",
132
+ "type" : "MultiArray"
133
+ },
134
+ {
135
+ "hasShapeFlexibility" : "0",
136
+ "isOptional" : "0",
137
+ "dataType" : "Float16",
138
+ "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 8)",
139
+ "shortDescription" : "",
140
+ "shape" : "[1, 1, 1, 8]",
141
+ "name" : "hidden_context_mask",
142
+ "type" : "MultiArray"
143
+ },
144
+ {
145
+ "hasShapeFlexibility" : "0",
146
+ "isOptional" : "0",
147
+ "dataType" : "Float16",
148
+ "formattedType" : "MultiArray (Float16 1 × 512 × 1 × 2)",
149
+ "shortDescription" : "",
150
+ "shape" : "[1, 512, 1, 2]",
151
+ "name" : "pre_conv_context",
152
+ "type" : "MultiArray"
153
+ }
154
+ ],
155
+ "computePrecision" : "Mixed (Float16, Float32, Int32, UInt16)",
156
+ "storagePrecision" : "Mixed (Float16, Palettized (8 bits), UInt8)",
157
+ "stateSchema" : [
158
+
159
+ ],
160
+ "outputSchema" : [
161
+ {
162
+ "hasShapeFlexibility" : "0",
163
+ "isOptional" : "0",
164
+ "dataType" : "Float16",
165
+ "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 1920)",
166
+ "shortDescription" : "",
167
+ "shape" : "[1, 1, 1, 1920]",
168
+ "name" : "audio",
169
+ "type" : "MultiArray"
170
+ },
171
+ {
172
+ "hasShapeFlexibility" : "0",
173
+ "isOptional" : "0",
174
+ "dataType" : "Float16",
175
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 1)",
176
+ "shortDescription" : "",
177
+ "shape" : "[1, 8192, 1, 1]",
178
+ "name" : "key_cache_updates",
179
+ "type" : "MultiArray"
180
+ },
181
+ {
182
+ "hasShapeFlexibility" : "0",
183
+ "isOptional" : "0",
184
+ "dataType" : "Float16",
185
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 1)",
186
+ "shortDescription" : "",
187
+ "shape" : "[1, 8192, 1, 1]",
188
+ "name" : "value_cache_updates",
189
+ "type" : "MultiArray"
190
+ },
191
+ {
192
+ "hasShapeFlexibility" : "0",
193
+ "isOptional" : "0",
194
+ "dataType" : "Float16",
195
+ "formattedType" : "MultiArray (Float16 1 × 1024 × 1 × 1)",
196
+ "shortDescription" : "",
197
+ "shape" : "[1, 1024, 1, 1]",
198
+ "name" : "hidden_context_update",
199
+ "type" : "MultiArray"
200
+ },
201
+ {
202
+ "hasShapeFlexibility" : "0",
203
+ "isOptional" : "0",
204
+ "dataType" : "Float16",
205
+ "formattedType" : "MultiArray (Float16 1 × 512 × 1 × 1)",
206
+ "shortDescription" : "",
207
+ "shape" : "[1, 512, 1, 1]",
208
+ "name" : "pre_conv_context_update",
209
+ "type" : "MultiArray"
210
+ }
211
+ ],
212
+ "name" : "latency",
213
+ "mlProgramOperationTypeHistogram" : {
214
+ "Ios18.expandDims" : 25,
215
+ "Ios18.softmax" : 8,
216
+ "Ios18.mul" : 312,
217
+ "Ios18.matmul" : 16,
218
+ "Ios18.rsqrt" : 17,
219
+ "Ios18.batchNorm" : 29,
220
+ "Ios16.reduceMean" : 17,
221
+ "Split" : 2,
222
+ "Tile" : 6,
223
+ "Ios18.gather" : 18,
224
+ "Ios18.add" : 189,
225
+ "Ios18.layerNorm" : 2,
226
+ "Ios18.reshape" : 72,
227
+ "Pad" : 1,
228
+ "Ios18.round" : 29,
229
+ "Ios18.constexprLutToDense" : 114,
230
+ "Ios18.conv" : 99,
231
+ "Ios18.concat" : 40,
232
+ "Ios18.transpose" : 28,
233
+ "Ios18.sub" : 30,
234
+ "Ios18.cast" : 17,
235
+ "Ios18.silu" : 8,
236
+ "Ios18.gelu" : 2,
237
+ "Ios18.clip" : 1,
238
+ "Ios18.sliceByIndex" : 89,
239
+ "Ios18.squeeze" : 2
240
+ }
241
+ },
242
+ {
243
+ "inputSchema" : [
244
+ {
245
+ "hasShapeFlexibility" : "0",
246
+ "isOptional" : "0",
247
+ "dataType" : "Int32",
248
+ "formattedType" : "MultiArray (Int32 1 × 16 × 4)",
249
+ "shortDescription" : "",
250
+ "shape" : "[1, 16, 4]",
251
+ "name" : "audio_codes",
252
+ "type" : "MultiArray"
253
+ },
254
+ {
255
+ "hasShapeFlexibility" : "0",
256
+ "isOptional" : "0",
257
+ "dataType" : "Int32",
258
+ "formattedType" : "MultiArray (Int32 1)",
259
+ "shortDescription" : "",
260
+ "shape" : "[1]",
261
+ "name" : "cache_length",
262
+ "type" : "MultiArray"
263
+ },
264
+ {
265
+ "hasShapeFlexibility" : "0",
266
+ "isOptional" : "0",
267
+ "dataType" : "Float16",
268
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 80)",
269
+ "shortDescription" : "",
270
+ "shape" : "[1, 8192, 1, 80]",
271
+ "name" : "key_cache",
272
+ "type" : "MultiArray"
273
+ },
274
+ {
275
+ "hasShapeFlexibility" : "0",
276
+ "isOptional" : "0",
277
+ "dataType" : "Float16",
278
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 80)",
279
+ "shortDescription" : "",
280
+ "shape" : "[1, 8192, 1, 80]",
281
+ "name" : "value_cache",
282
+ "type" : "MultiArray"
283
+ },
284
+ {
285
+ "hasShapeFlexibility" : "0",
286
+ "isOptional" : "0",
287
+ "dataType" : "Float16",
288
+ "formattedType" : "MultiArray (Float16 1 × 4 × 80)",
289
+ "shortDescription" : "",
290
+ "shape" : "[1, 4, 80]",
291
+ "name" : "kv_cache_update_mask",
292
+ "type" : "MultiArray"
293
+ },
294
+ {
295
+ "hasShapeFlexibility" : "0",
296
+ "isOptional" : "0",
297
+ "dataType" : "Float16",
298
+ "formattedType" : "MultiArray (Float16 1 × 80)",
299
+ "shortDescription" : "",
300
+ "shape" : "[1, 80]",
301
+ "name" : "key_padding_mask",
302
+ "type" : "MultiArray"
303
+ },
304
+ {
305
+ "hasShapeFlexibility" : "0",
306
+ "isOptional" : "0",
307
+ "dataType" : "Float16",
308
+ "formattedType" : "MultiArray (Float16 1 × 1024 × 1 × 8)",
309
+ "shortDescription" : "",
310
+ "shape" : "[1, 1024, 1, 8]",
311
+ "name" : "hidden_context",
312
+ "type" : "MultiArray"
313
+ },
314
+ {
315
+ "hasShapeFlexibility" : "0",
316
+ "isOptional" : "0",
317
+ "dataType" : "Float16",
318
+ "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 8)",
319
+ "shortDescription" : "",
320
+ "shape" : "[1, 1, 1, 8]",
321
+ "name" : "hidden_context_mask",
322
+ "type" : "MultiArray"
323
+ },
324
+ {
325
+ "hasShapeFlexibility" : "0",
326
+ "isOptional" : "0",
327
+ "dataType" : "Float16",
328
+ "formattedType" : "MultiArray (Float16 1 × 512 × 1 × 2)",
329
+ "shortDescription" : "",
330
+ "shape" : "[1, 512, 1, 2]",
331
+ "name" : "pre_conv_context",
332
+ "type" : "MultiArray"
333
+ },
334
+ {
335
+ "hasShapeFlexibility" : "0",
336
+ "isOptional" : "0",
337
+ "dataType" : "Float16",
338
+ "formattedType" : "MultiArray (Float16 1 × 4 × 80)",
339
+ "shortDescription" : "",
340
+ "shape" : "[1, 4, 80]",
341
+ "name" : "qk_mask",
342
+ "type" : "MultiArray"
343
+ }
344
+ ],
345
+ "computePrecision" : "Mixed (Float16, Float32, Int16, Int32, UInt16)",
346
+ "storagePrecision" : "Mixed (Float16, Palettized (8 bits), UInt8)",
347
+ "stateSchema" : [
348
+
349
+ ],
350
+ "outputSchema" : [
351
+ {
352
+ "hasShapeFlexibility" : "0",
353
+ "isOptional" : "0",
354
+ "dataType" : "Float16",
355
+ "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 7680)",
356
+ "shortDescription" : "",
357
+ "shape" : "[1, 1, 1, 7680]",
358
+ "name" : "audio",
359
+ "type" : "MultiArray"
360
+ },
361
+ {
362
+ "hasShapeFlexibility" : "0",
363
+ "isOptional" : "0",
364
+ "dataType" : "Float16",
365
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 4)",
366
+ "shortDescription" : "",
367
+ "shape" : "[1, 8192, 1, 4]",
368
+ "name" : "key_cache_updates",
369
+ "type" : "MultiArray"
370
+ },
371
+ {
372
+ "hasShapeFlexibility" : "0",
373
+ "isOptional" : "0",
374
+ "dataType" : "Float16",
375
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 4)",
376
+ "shortDescription" : "",
377
+ "shape" : "[1, 8192, 1, 4]",
378
+ "name" : "value_cache_updates",
379
+ "type" : "MultiArray"
380
+ },
381
+ {
382
+ "hasShapeFlexibility" : "0",
383
+ "isOptional" : "0",
384
+ "dataType" : "Float16",
385
+ "formattedType" : "MultiArray (Float16 1 × 1024 × 1 × 4)",
386
+ "shortDescription" : "",
387
+ "shape" : "[1, 1024, 1, 4]",
388
+ "name" : "hidden_context_update",
389
+ "type" : "MultiArray"
390
+ },
391
+ {
392
+ "hasShapeFlexibility" : "0",
393
+ "isOptional" : "0",
394
+ "dataType" : "Float16",
395
+ "formattedType" : "MultiArray (Float16 1 × 512 × 1 × 4)",
396
+ "shortDescription" : "",
397
+ "shape" : "[1, 512, 1, 4]",
398
+ "name" : "pre_conv_context_update",
399
+ "type" : "MultiArray"
400
+ }
401
+ ],
402
+ "name" : "throughput",
403
+ "mlProgramOperationTypeHistogram" : {
404
+ "Ios18.expandDims" : 26,
405
+ "Ios18.softmax" : 8,
406
+ "Ios18.mul" : 296,
407
+ "Ios18.matmul" : 32,
408
+ "Ios18.rsqrt" : 17,
409
+ "Ios18.batchNorm" : 29,
410
+ "Ios16.reduceMean" : 17,
411
+ "Split" : 2,
412
+ "Ios18.greaterEqual" : 2,
413
+ "Select" : 2,
414
+ "Tile" : 6,
415
+ "Ios18.gather" : 18,
416
+ "Ios18.add" : 200,
417
+ "Ios16.reduceSum" : 1,
418
+ "Ios18.layerNorm" : 2,
419
+ "Ios18.reshape" : 104,
420
+ "Pad" : 1,
421
+ "Ios18.round" : 29,
422
+ "Ios18.constexprLutToDense" : 116,
423
+ "Ios18.conv" : 99,
424
+ "Ios18.concat" : 40,
425
+ "Ios18.transpose" : 62,
426
+ "Ios18.sub" : 30,
427
+ "Ios18.cast" : 20,
428
+ "Ios18.silu" : 8,
429
+ "Ios18.gelu" : 2,
430
+ "Ios18.clip" : 1,
431
+ "Ios18.sliceByIndex" : 89,
432
+ "Ios18.squeeze" : 2
433
+ }
434
+ }
435
+ ],
436
+ "mlProgramOperationTypeHistogram" : {
437
+ "Ios18.expandDims" : 25,
438
+ "Ios18.softmax" : 8,
439
+ "Ios18.mul" : 312,
440
+ "Ios18.matmul" : 16,
441
+ "Ios18.rsqrt" : 17,
442
+ "Ios18.batchNorm" : 29,
443
+ "Ios16.reduceMean" : 17,
444
+ "Split" : 2,
445
+ "Tile" : 6,
446
+ "Ios18.gather" : 18,
447
+ "Ios18.add" : 189,
448
+ "Ios18.layerNorm" : 2,
449
+ "Ios18.reshape" : 72,
450
+ "Pad" : 1,
451
+ "Ios18.round" : 29,
452
+ "Ios18.constexprLutToDense" : 114,
453
+ "Ios18.conv" : 99,
454
+ "Ios18.concat" : 40,
455
+ "Ios18.transpose" : 28,
456
+ "Ios18.sub" : 30,
457
+ "Ios18.cast" : 17,
458
+ "Ios18.silu" : 8,
459
+ "Ios18.gelu" : 2,
460
+ "Ios18.clip" : 1,
461
+ "Ios18.sliceByIndex" : 89,
462
+ "Ios18.squeeze" : 2
463
+ },
464
+ "isUpdatable" : "0",
465
+ "stateSchema" : [
466
+
467
+ ],
468
+ "availability" : {
469
+ "macOS" : "15.0",
470
+ "tvOS" : "18.0",
471
+ "visionOS" : "2.0",
472
+ "watchOS" : "11.0",
473
+ "iOS" : "18.0",
474
+ "macCatalyst" : "18.0"
475
+ },
476
+ "computePrecision" : "Mixed (Float16, Float32, Int32, UInt16)",
477
+ "modelType" : {
478
+ "name" : "MLModelType_mlProgram"
479
+ },
480
+ "inputSchema" : [
481
+ {
482
+ "hasShapeFlexibility" : "0",
483
+ "isOptional" : "0",
484
+ "dataType" : "Int32",
485
+ "formattedType" : "MultiArray (Int32 1 × 16 × 1)",
486
+ "shortDescription" : "",
487
+ "shape" : "[1, 16, 1]",
488
+ "name" : "audio_codes",
489
+ "type" : "MultiArray"
490
+ },
491
+ {
492
+ "hasShapeFlexibility" : "0",
493
+ "isOptional" : "0",
494
+ "dataType" : "Int32",
495
+ "formattedType" : "MultiArray (Int32 1)",
496
+ "shortDescription" : "",
497
+ "shape" : "[1]",
498
+ "name" : "cache_length",
499
+ "type" : "MultiArray"
500
+ },
501
+ {
502
+ "hasShapeFlexibility" : "0",
503
+ "isOptional" : "0",
504
+ "dataType" : "Float16",
505
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 80)",
506
+ "shortDescription" : "",
507
+ "shape" : "[1, 8192, 1, 80]",
508
+ "name" : "key_cache",
509
+ "type" : "MultiArray"
510
+ },
511
+ {
512
+ "hasShapeFlexibility" : "0",
513
+ "isOptional" : "0",
514
+ "dataType" : "Float16",
515
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 80)",
516
+ "shortDescription" : "",
517
+ "shape" : "[1, 8192, 1, 80]",
518
+ "name" : "value_cache",
519
+ "type" : "MultiArray"
520
+ },
521
+ {
522
+ "hasShapeFlexibility" : "0",
523
+ "isOptional" : "0",
524
+ "dataType" : "Float16",
525
+ "formattedType" : "MultiArray (Float16 1 × 1 × 80)",
526
+ "shortDescription" : "",
527
+ "shape" : "[1, 1, 80]",
528
+ "name" : "kv_cache_update_mask",
529
+ "type" : "MultiArray"
530
+ },
531
+ {
532
+ "hasShapeFlexibility" : "0",
533
+ "isOptional" : "0",
534
+ "dataType" : "Float16",
535
+ "formattedType" : "MultiArray (Float16 1 × 80)",
536
+ "shortDescription" : "",
537
+ "shape" : "[1, 80]",
538
+ "name" : "key_padding_mask",
539
+ "type" : "MultiArray"
540
+ },
541
+ {
542
+ "hasShapeFlexibility" : "0",
543
+ "isOptional" : "0",
544
+ "dataType" : "Float16",
545
+ "formattedType" : "MultiArray (Float16 1 × 1024 × 1 × 8)",
546
+ "shortDescription" : "",
547
+ "shape" : "[1, 1024, 1, 8]",
548
+ "name" : "hidden_context",
549
+ "type" : "MultiArray"
550
+ },
551
+ {
552
+ "hasShapeFlexibility" : "0",
553
+ "isOptional" : "0",
554
+ "dataType" : "Float16",
555
+ "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 8)",
556
+ "shortDescription" : "",
557
+ "shape" : "[1, 1, 1, 8]",
558
+ "name" : "hidden_context_mask",
559
+ "type" : "MultiArray"
560
+ },
561
+ {
562
+ "hasShapeFlexibility" : "0",
563
+ "isOptional" : "0",
564
+ "dataType" : "Float16",
565
+ "formattedType" : "MultiArray (Float16 1 × 512 × 1 × 2)",
566
+ "shortDescription" : "",
567
+ "shape" : "[1, 512, 1, 2]",
568
+ "name" : "pre_conv_context",
569
+ "type" : "MultiArray"
570
+ }
571
+ ],
572
+ "defaultFunctionName" : "latency",
573
+ "generatedClassName" : "SpeechDecoder",
574
+ "userDefinedMetadata" : {
575
+
576
+ },
577
+ "method" : "predict"
578
+ }
579
+ ]
qwen3_tts/speech_decoder/12hz-0.6b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/model.mil ADDED
The diff for this file is too large to render. See raw diff
 
qwen3_tts/speech_decoder/12hz-0.6b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/weights/weight.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:827114b7e98262297d55ed977405a56e3a26f431c22968e31604a3ce7d7dec50
3
+ size 115358272
qwen3_tts/speech_decoder/12hz-1.7b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/analytics/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3fd77fbdb6571721611661deb3bf3184aef6cb12b789235f6c04053b50f50f0f
3
+ size 243
qwen3_tts/speech_decoder/12hz-1.7b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/coremldata.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8f40ab9923fffd07b7906bcfd55e79d5d85f2be729d6ea347826f2768b20e921
3
+ size 1115
qwen3_tts/speech_decoder/12hz-1.7b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/metadata.json ADDED
@@ -0,0 +1,579 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ {
3
+ "metadataOutputVersion" : "3.0",
4
+ "storagePrecision" : "Mixed (Float16, Palettized (8 bits), UInt8)",
5
+ "outputSchema" : [
6
+ {
7
+ "hasShapeFlexibility" : "0",
8
+ "isOptional" : "0",
9
+ "dataType" : "Float16",
10
+ "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 1920)",
11
+ "shortDescription" : "",
12
+ "shape" : "[1, 1, 1, 1920]",
13
+ "name" : "audio",
14
+ "type" : "MultiArray"
15
+ },
16
+ {
17
+ "hasShapeFlexibility" : "0",
18
+ "isOptional" : "0",
19
+ "dataType" : "Float16",
20
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 1)",
21
+ "shortDescription" : "",
22
+ "shape" : "[1, 8192, 1, 1]",
23
+ "name" : "key_cache_updates",
24
+ "type" : "MultiArray"
25
+ },
26
+ {
27
+ "hasShapeFlexibility" : "0",
28
+ "isOptional" : "0",
29
+ "dataType" : "Float16",
30
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 1)",
31
+ "shortDescription" : "",
32
+ "shape" : "[1, 8192, 1, 1]",
33
+ "name" : "value_cache_updates",
34
+ "type" : "MultiArray"
35
+ },
36
+ {
37
+ "hasShapeFlexibility" : "0",
38
+ "isOptional" : "0",
39
+ "dataType" : "Float16",
40
+ "formattedType" : "MultiArray (Float16 1 × 1024 × 1 × 1)",
41
+ "shortDescription" : "",
42
+ "shape" : "[1, 1024, 1, 1]",
43
+ "name" : "hidden_context_update",
44
+ "type" : "MultiArray"
45
+ },
46
+ {
47
+ "hasShapeFlexibility" : "0",
48
+ "isOptional" : "0",
49
+ "dataType" : "Float16",
50
+ "formattedType" : "MultiArray (Float16 1 × 512 × 1 × 1)",
51
+ "shortDescription" : "",
52
+ "shape" : "[1, 512, 1, 1]",
53
+ "name" : "pre_conv_context_update",
54
+ "type" : "MultiArray"
55
+ }
56
+ ],
57
+ "modelParameters" : [
58
+
59
+ ],
60
+ "specificationVersion" : 9,
61
+ "functions" : [
62
+ {
63
+ "inputSchema" : [
64
+ {
65
+ "hasShapeFlexibility" : "0",
66
+ "isOptional" : "0",
67
+ "dataType" : "Int32",
68
+ "formattedType" : "MultiArray (Int32 1 × 16 × 1)",
69
+ "shortDescription" : "",
70
+ "shape" : "[1, 16, 1]",
71
+ "name" : "audio_codes",
72
+ "type" : "MultiArray"
73
+ },
74
+ {
75
+ "hasShapeFlexibility" : "0",
76
+ "isOptional" : "0",
77
+ "dataType" : "Int32",
78
+ "formattedType" : "MultiArray (Int32 1)",
79
+ "shortDescription" : "",
80
+ "shape" : "[1]",
81
+ "name" : "cache_length",
82
+ "type" : "MultiArray"
83
+ },
84
+ {
85
+ "hasShapeFlexibility" : "0",
86
+ "isOptional" : "0",
87
+ "dataType" : "Float16",
88
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 80)",
89
+ "shortDescription" : "",
90
+ "shape" : "[1, 8192, 1, 80]",
91
+ "name" : "key_cache",
92
+ "type" : "MultiArray"
93
+ },
94
+ {
95
+ "hasShapeFlexibility" : "0",
96
+ "isOptional" : "0",
97
+ "dataType" : "Float16",
98
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 80)",
99
+ "shortDescription" : "",
100
+ "shape" : "[1, 8192, 1, 80]",
101
+ "name" : "value_cache",
102
+ "type" : "MultiArray"
103
+ },
104
+ {
105
+ "hasShapeFlexibility" : "0",
106
+ "isOptional" : "0",
107
+ "dataType" : "Float16",
108
+ "formattedType" : "MultiArray (Float16 1 × 1 × 80)",
109
+ "shortDescription" : "",
110
+ "shape" : "[1, 1, 80]",
111
+ "name" : "kv_cache_update_mask",
112
+ "type" : "MultiArray"
113
+ },
114
+ {
115
+ "hasShapeFlexibility" : "0",
116
+ "isOptional" : "0",
117
+ "dataType" : "Float16",
118
+ "formattedType" : "MultiArray (Float16 1 × 80)",
119
+ "shortDescription" : "",
120
+ "shape" : "[1, 80]",
121
+ "name" : "key_padding_mask",
122
+ "type" : "MultiArray"
123
+ },
124
+ {
125
+ "hasShapeFlexibility" : "0",
126
+ "isOptional" : "0",
127
+ "dataType" : "Float16",
128
+ "formattedType" : "MultiArray (Float16 1 × 1024 × 1 × 8)",
129
+ "shortDescription" : "",
130
+ "shape" : "[1, 1024, 1, 8]",
131
+ "name" : "hidden_context",
132
+ "type" : "MultiArray"
133
+ },
134
+ {
135
+ "hasShapeFlexibility" : "0",
136
+ "isOptional" : "0",
137
+ "dataType" : "Float16",
138
+ "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 8)",
139
+ "shortDescription" : "",
140
+ "shape" : "[1, 1, 1, 8]",
141
+ "name" : "hidden_context_mask",
142
+ "type" : "MultiArray"
143
+ },
144
+ {
145
+ "hasShapeFlexibility" : "0",
146
+ "isOptional" : "0",
147
+ "dataType" : "Float16",
148
+ "formattedType" : "MultiArray (Float16 1 × 512 × 1 × 2)",
149
+ "shortDescription" : "",
150
+ "shape" : "[1, 512, 1, 2]",
151
+ "name" : "pre_conv_context",
152
+ "type" : "MultiArray"
153
+ }
154
+ ],
155
+ "computePrecision" : "Mixed (Float16, Float32, Int32, UInt16)",
156
+ "storagePrecision" : "Mixed (Float16, Palettized (8 bits), UInt8)",
157
+ "stateSchema" : [
158
+
159
+ ],
160
+ "outputSchema" : [
161
+ {
162
+ "hasShapeFlexibility" : "0",
163
+ "isOptional" : "0",
164
+ "dataType" : "Float16",
165
+ "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 1920)",
166
+ "shortDescription" : "",
167
+ "shape" : "[1, 1, 1, 1920]",
168
+ "name" : "audio",
169
+ "type" : "MultiArray"
170
+ },
171
+ {
172
+ "hasShapeFlexibility" : "0",
173
+ "isOptional" : "0",
174
+ "dataType" : "Float16",
175
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 1)",
176
+ "shortDescription" : "",
177
+ "shape" : "[1, 8192, 1, 1]",
178
+ "name" : "key_cache_updates",
179
+ "type" : "MultiArray"
180
+ },
181
+ {
182
+ "hasShapeFlexibility" : "0",
183
+ "isOptional" : "0",
184
+ "dataType" : "Float16",
185
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 1)",
186
+ "shortDescription" : "",
187
+ "shape" : "[1, 8192, 1, 1]",
188
+ "name" : "value_cache_updates",
189
+ "type" : "MultiArray"
190
+ },
191
+ {
192
+ "hasShapeFlexibility" : "0",
193
+ "isOptional" : "0",
194
+ "dataType" : "Float16",
195
+ "formattedType" : "MultiArray (Float16 1 × 1024 × 1 × 1)",
196
+ "shortDescription" : "",
197
+ "shape" : "[1, 1024, 1, 1]",
198
+ "name" : "hidden_context_update",
199
+ "type" : "MultiArray"
200
+ },
201
+ {
202
+ "hasShapeFlexibility" : "0",
203
+ "isOptional" : "0",
204
+ "dataType" : "Float16",
205
+ "formattedType" : "MultiArray (Float16 1 × 512 × 1 × 1)",
206
+ "shortDescription" : "",
207
+ "shape" : "[1, 512, 1, 1]",
208
+ "name" : "pre_conv_context_update",
209
+ "type" : "MultiArray"
210
+ }
211
+ ],
212
+ "name" : "latency",
213
+ "mlProgramOperationTypeHistogram" : {
214
+ "Ios18.expandDims" : 25,
215
+ "Ios18.softmax" : 8,
216
+ "Ios18.mul" : 312,
217
+ "Ios18.matmul" : 16,
218
+ "Ios18.rsqrt" : 17,
219
+ "Ios18.batchNorm" : 29,
220
+ "Ios16.reduceMean" : 17,
221
+ "Split" : 2,
222
+ "Tile" : 6,
223
+ "Ios18.gather" : 18,
224
+ "Ios18.add" : 189,
225
+ "Ios18.layerNorm" : 2,
226
+ "Ios18.reshape" : 72,
227
+ "Pad" : 1,
228
+ "Ios18.round" : 29,
229
+ "Ios18.constexprLutToDense" : 114,
230
+ "Ios18.conv" : 99,
231
+ "Ios18.concat" : 40,
232
+ "Ios18.transpose" : 28,
233
+ "Ios18.sub" : 30,
234
+ "Ios18.cast" : 17,
235
+ "Ios18.silu" : 8,
236
+ "Ios18.gelu" : 2,
237
+ "Ios18.clip" : 1,
238
+ "Ios18.sliceByIndex" : 89,
239
+ "Ios18.squeeze" : 2
240
+ }
241
+ },
242
+ {
243
+ "inputSchema" : [
244
+ {
245
+ "hasShapeFlexibility" : "0",
246
+ "isOptional" : "0",
247
+ "dataType" : "Int32",
248
+ "formattedType" : "MultiArray (Int32 1 × 16 × 4)",
249
+ "shortDescription" : "",
250
+ "shape" : "[1, 16, 4]",
251
+ "name" : "audio_codes",
252
+ "type" : "MultiArray"
253
+ },
254
+ {
255
+ "hasShapeFlexibility" : "0",
256
+ "isOptional" : "0",
257
+ "dataType" : "Int32",
258
+ "formattedType" : "MultiArray (Int32 1)",
259
+ "shortDescription" : "",
260
+ "shape" : "[1]",
261
+ "name" : "cache_length",
262
+ "type" : "MultiArray"
263
+ },
264
+ {
265
+ "hasShapeFlexibility" : "0",
266
+ "isOptional" : "0",
267
+ "dataType" : "Float16",
268
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 80)",
269
+ "shortDescription" : "",
270
+ "shape" : "[1, 8192, 1, 80]",
271
+ "name" : "key_cache",
272
+ "type" : "MultiArray"
273
+ },
274
+ {
275
+ "hasShapeFlexibility" : "0",
276
+ "isOptional" : "0",
277
+ "dataType" : "Float16",
278
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 80)",
279
+ "shortDescription" : "",
280
+ "shape" : "[1, 8192, 1, 80]",
281
+ "name" : "value_cache",
282
+ "type" : "MultiArray"
283
+ },
284
+ {
285
+ "hasShapeFlexibility" : "0",
286
+ "isOptional" : "0",
287
+ "dataType" : "Float16",
288
+ "formattedType" : "MultiArray (Float16 1 × 4 × 80)",
289
+ "shortDescription" : "",
290
+ "shape" : "[1, 4, 80]",
291
+ "name" : "kv_cache_update_mask",
292
+ "type" : "MultiArray"
293
+ },
294
+ {
295
+ "hasShapeFlexibility" : "0",
296
+ "isOptional" : "0",
297
+ "dataType" : "Float16",
298
+ "formattedType" : "MultiArray (Float16 1 × 80)",
299
+ "shortDescription" : "",
300
+ "shape" : "[1, 80]",
301
+ "name" : "key_padding_mask",
302
+ "type" : "MultiArray"
303
+ },
304
+ {
305
+ "hasShapeFlexibility" : "0",
306
+ "isOptional" : "0",
307
+ "dataType" : "Float16",
308
+ "formattedType" : "MultiArray (Float16 1 × 1024 × 1 × 8)",
309
+ "shortDescription" : "",
310
+ "shape" : "[1, 1024, 1, 8]",
311
+ "name" : "hidden_context",
312
+ "type" : "MultiArray"
313
+ },
314
+ {
315
+ "hasShapeFlexibility" : "0",
316
+ "isOptional" : "0",
317
+ "dataType" : "Float16",
318
+ "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 8)",
319
+ "shortDescription" : "",
320
+ "shape" : "[1, 1, 1, 8]",
321
+ "name" : "hidden_context_mask",
322
+ "type" : "MultiArray"
323
+ },
324
+ {
325
+ "hasShapeFlexibility" : "0",
326
+ "isOptional" : "0",
327
+ "dataType" : "Float16",
328
+ "formattedType" : "MultiArray (Float16 1 × 512 × 1 × 2)",
329
+ "shortDescription" : "",
330
+ "shape" : "[1, 512, 1, 2]",
331
+ "name" : "pre_conv_context",
332
+ "type" : "MultiArray"
333
+ },
334
+ {
335
+ "hasShapeFlexibility" : "0",
336
+ "isOptional" : "0",
337
+ "dataType" : "Float16",
338
+ "formattedType" : "MultiArray (Float16 1 × 4 × 80)",
339
+ "shortDescription" : "",
340
+ "shape" : "[1, 4, 80]",
341
+ "name" : "qk_mask",
342
+ "type" : "MultiArray"
343
+ }
344
+ ],
345
+ "computePrecision" : "Mixed (Float16, Float32, Int16, Int32, UInt16)",
346
+ "storagePrecision" : "Mixed (Float16, Palettized (8 bits), UInt8)",
347
+ "stateSchema" : [
348
+
349
+ ],
350
+ "outputSchema" : [
351
+ {
352
+ "hasShapeFlexibility" : "0",
353
+ "isOptional" : "0",
354
+ "dataType" : "Float16",
355
+ "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 7680)",
356
+ "shortDescription" : "",
357
+ "shape" : "[1, 1, 1, 7680]",
358
+ "name" : "audio",
359
+ "type" : "MultiArray"
360
+ },
361
+ {
362
+ "hasShapeFlexibility" : "0",
363
+ "isOptional" : "0",
364
+ "dataType" : "Float16",
365
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 4)",
366
+ "shortDescription" : "",
367
+ "shape" : "[1, 8192, 1, 4]",
368
+ "name" : "key_cache_updates",
369
+ "type" : "MultiArray"
370
+ },
371
+ {
372
+ "hasShapeFlexibility" : "0",
373
+ "isOptional" : "0",
374
+ "dataType" : "Float16",
375
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 4)",
376
+ "shortDescription" : "",
377
+ "shape" : "[1, 8192, 1, 4]",
378
+ "name" : "value_cache_updates",
379
+ "type" : "MultiArray"
380
+ },
381
+ {
382
+ "hasShapeFlexibility" : "0",
383
+ "isOptional" : "0",
384
+ "dataType" : "Float16",
385
+ "formattedType" : "MultiArray (Float16 1 × 1024 × 1 × 4)",
386
+ "shortDescription" : "",
387
+ "shape" : "[1, 1024, 1, 4]",
388
+ "name" : "hidden_context_update",
389
+ "type" : "MultiArray"
390
+ },
391
+ {
392
+ "hasShapeFlexibility" : "0",
393
+ "isOptional" : "0",
394
+ "dataType" : "Float16",
395
+ "formattedType" : "MultiArray (Float16 1 × 512 × 1 × 4)",
396
+ "shortDescription" : "",
397
+ "shape" : "[1, 512, 1, 4]",
398
+ "name" : "pre_conv_context_update",
399
+ "type" : "MultiArray"
400
+ }
401
+ ],
402
+ "name" : "throughput",
403
+ "mlProgramOperationTypeHistogram" : {
404
+ "Ios18.expandDims" : 26,
405
+ "Ios18.softmax" : 8,
406
+ "Ios18.mul" : 296,
407
+ "Ios18.matmul" : 32,
408
+ "Ios18.rsqrt" : 17,
409
+ "Ios18.batchNorm" : 29,
410
+ "Ios16.reduceMean" : 17,
411
+ "Split" : 2,
412
+ "Ios18.greaterEqual" : 2,
413
+ "Select" : 2,
414
+ "Tile" : 6,
415
+ "Ios18.gather" : 18,
416
+ "Ios18.add" : 200,
417
+ "Ios16.reduceSum" : 1,
418
+ "Ios18.layerNorm" : 2,
419
+ "Ios18.reshape" : 104,
420
+ "Pad" : 1,
421
+ "Ios18.round" : 29,
422
+ "Ios18.constexprLutToDense" : 116,
423
+ "Ios18.conv" : 99,
424
+ "Ios18.concat" : 40,
425
+ "Ios18.transpose" : 62,
426
+ "Ios18.sub" : 30,
427
+ "Ios18.cast" : 20,
428
+ "Ios18.silu" : 8,
429
+ "Ios18.gelu" : 2,
430
+ "Ios18.clip" : 1,
431
+ "Ios18.sliceByIndex" : 89,
432
+ "Ios18.squeeze" : 2
433
+ }
434
+ }
435
+ ],
436
+ "mlProgramOperationTypeHistogram" : {
437
+ "Ios18.expandDims" : 25,
438
+ "Ios18.softmax" : 8,
439
+ "Ios18.mul" : 312,
440
+ "Ios18.matmul" : 16,
441
+ "Ios18.rsqrt" : 17,
442
+ "Ios18.batchNorm" : 29,
443
+ "Ios16.reduceMean" : 17,
444
+ "Split" : 2,
445
+ "Tile" : 6,
446
+ "Ios18.gather" : 18,
447
+ "Ios18.add" : 189,
448
+ "Ios18.layerNorm" : 2,
449
+ "Ios18.reshape" : 72,
450
+ "Pad" : 1,
451
+ "Ios18.round" : 29,
452
+ "Ios18.constexprLutToDense" : 114,
453
+ "Ios18.conv" : 99,
454
+ "Ios18.concat" : 40,
455
+ "Ios18.transpose" : 28,
456
+ "Ios18.sub" : 30,
457
+ "Ios18.cast" : 17,
458
+ "Ios18.silu" : 8,
459
+ "Ios18.gelu" : 2,
460
+ "Ios18.clip" : 1,
461
+ "Ios18.sliceByIndex" : 89,
462
+ "Ios18.squeeze" : 2
463
+ },
464
+ "isUpdatable" : "0",
465
+ "stateSchema" : [
466
+
467
+ ],
468
+ "availability" : {
469
+ "macOS" : "15.0",
470
+ "tvOS" : "18.0",
471
+ "visionOS" : "2.0",
472
+ "watchOS" : "11.0",
473
+ "iOS" : "18.0",
474
+ "macCatalyst" : "18.0"
475
+ },
476
+ "computePrecision" : "Mixed (Float16, Float32, Int32, UInt16)",
477
+ "modelType" : {
478
+ "name" : "MLModelType_mlProgram"
479
+ },
480
+ "inputSchema" : [
481
+ {
482
+ "hasShapeFlexibility" : "0",
483
+ "isOptional" : "0",
484
+ "dataType" : "Int32",
485
+ "formattedType" : "MultiArray (Int32 1 × 16 × 1)",
486
+ "shortDescription" : "",
487
+ "shape" : "[1, 16, 1]",
488
+ "name" : "audio_codes",
489
+ "type" : "MultiArray"
490
+ },
491
+ {
492
+ "hasShapeFlexibility" : "0",
493
+ "isOptional" : "0",
494
+ "dataType" : "Int32",
495
+ "formattedType" : "MultiArray (Int32 1)",
496
+ "shortDescription" : "",
497
+ "shape" : "[1]",
498
+ "name" : "cache_length",
499
+ "type" : "MultiArray"
500
+ },
501
+ {
502
+ "hasShapeFlexibility" : "0",
503
+ "isOptional" : "0",
504
+ "dataType" : "Float16",
505
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 80)",
506
+ "shortDescription" : "",
507
+ "shape" : "[1, 8192, 1, 80]",
508
+ "name" : "key_cache",
509
+ "type" : "MultiArray"
510
+ },
511
+ {
512
+ "hasShapeFlexibility" : "0",
513
+ "isOptional" : "0",
514
+ "dataType" : "Float16",
515
+ "formattedType" : "MultiArray (Float16 1 × 8192 × 1 × 80)",
516
+ "shortDescription" : "",
517
+ "shape" : "[1, 8192, 1, 80]",
518
+ "name" : "value_cache",
519
+ "type" : "MultiArray"
520
+ },
521
+ {
522
+ "hasShapeFlexibility" : "0",
523
+ "isOptional" : "0",
524
+ "dataType" : "Float16",
525
+ "formattedType" : "MultiArray (Float16 1 × 1 × 80)",
526
+ "shortDescription" : "",
527
+ "shape" : "[1, 1, 80]",
528
+ "name" : "kv_cache_update_mask",
529
+ "type" : "MultiArray"
530
+ },
531
+ {
532
+ "hasShapeFlexibility" : "0",
533
+ "isOptional" : "0",
534
+ "dataType" : "Float16",
535
+ "formattedType" : "MultiArray (Float16 1 × 80)",
536
+ "shortDescription" : "",
537
+ "shape" : "[1, 80]",
538
+ "name" : "key_padding_mask",
539
+ "type" : "MultiArray"
540
+ },
541
+ {
542
+ "hasShapeFlexibility" : "0",
543
+ "isOptional" : "0",
544
+ "dataType" : "Float16",
545
+ "formattedType" : "MultiArray (Float16 1 × 1024 × 1 × 8)",
546
+ "shortDescription" : "",
547
+ "shape" : "[1, 1024, 1, 8]",
548
+ "name" : "hidden_context",
549
+ "type" : "MultiArray"
550
+ },
551
+ {
552
+ "hasShapeFlexibility" : "0",
553
+ "isOptional" : "0",
554
+ "dataType" : "Float16",
555
+ "formattedType" : "MultiArray (Float16 1 × 1 × 1 × 8)",
556
+ "shortDescription" : "",
557
+ "shape" : "[1, 1, 1, 8]",
558
+ "name" : "hidden_context_mask",
559
+ "type" : "MultiArray"
560
+ },
561
+ {
562
+ "hasShapeFlexibility" : "0",
563
+ "isOptional" : "0",
564
+ "dataType" : "Float16",
565
+ "formattedType" : "MultiArray (Float16 1 × 512 × 1 × 2)",
566
+ "shortDescription" : "",
567
+ "shape" : "[1, 512, 1, 2]",
568
+ "name" : "pre_conv_context",
569
+ "type" : "MultiArray"
570
+ }
571
+ ],
572
+ "defaultFunctionName" : "latency",
573
+ "generatedClassName" : "SpeechDecoder",
574
+ "userDefinedMetadata" : {
575
+
576
+ },
577
+ "method" : "predict"
578
+ }
579
+ ]
qwen3_tts/speech_decoder/12hz-1.7b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/model.mil ADDED
The diff for this file is too large to render. See raw diff
 
qwen3_tts/speech_decoder/12hz-1.7b-customvoice/W8A16-stream-multifunction/SpeechDecoder.mlmodelc/weights/weight.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:827114b7e98262297d55ed977405a56e3a26f431c22968e31604a3ce7d7dec50
3
+ size 115358272