pshashid commited on
Commit
2db2cef
·
verified ·
1 Parent(s): a18d0ca

Add/update moonshine-tiny

Browse files
moonshine-tiny/config.json ADDED
@@ -0,0 +1,33 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "MoonshineForConditionalGeneration"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 1,
8
+ "decoder_hidden_act": "silu",
9
+ "decoder_num_attention_heads": 8,
10
+ "decoder_num_hidden_layers": 6,
11
+ "decoder_num_key_value_heads": 8,
12
+ "decoder_start_token_id": 1,
13
+ "dtype": "float32",
14
+ "encoder_hidden_act": "gelu",
15
+ "encoder_num_attention_heads": 8,
16
+ "encoder_num_hidden_layers": 6,
17
+ "encoder_num_key_value_heads": 8,
18
+ "eos_token_id": 2,
19
+ "hidden_size": 288,
20
+ "initializer_range": 0.02,
21
+ "intermediate_size": 1152,
22
+ "is_encoder_decoder": true,
23
+ "max_position_embeddings": 194,
24
+ "model_type": "moonshine",
25
+ "pad_head_dim_to_multiple_of": 8,
26
+ "pad_token_id": 2,
27
+ "partial_rotary_factor": 0.9,
28
+ "rope_scaling": null,
29
+ "rope_theta": 10000.0,
30
+ "transformers_version": "4.57.6",
31
+ "use_cache": true,
32
+ "vocab_size": 32768
33
+ }
moonshine-tiny/decoder_model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c3994d01187c27f966be8f7b52806b5da270793235c3ff7cd6fffff83904ea93
3
+ size 78110938
moonshine-tiny/decoder_model_merged.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28ae600e002f6e877f96174a7cd43128a3fd8b9545b1e662266a40c6ded8d3fd
3
+ size 78563921
moonshine-tiny/decoder_with_past_model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0fcb7d0d3921098415e2652c84d3d9364e3acb4f9253877713ac52e56e4dff37
3
+ size 74113712
moonshine-tiny/encoder_model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a3ea57a0ce80bf2515b6ce218076956c09a7dec307089d43f34491c46c44e60f
3
+ size 30962848
moonshine-tiny/generation_config.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 1,
4
+ "decoder_start_token_id": 1,
5
+ "eos_token_id": 2,
6
+ "max_length": 194,
7
+ "pad_token_id": 2,
8
+ "transformers_version": "4.57.6"
9
+ }
moonshine-tiny/meta.json ADDED
@@ -0,0 +1,766 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "moonshine-tiny",
3
+ "source": "moonshine-ai/moonshine-tiny",
4
+ "method": "optimum",
5
+ "note": "needs a recent optimum; if export fails, the moonshine-ai repos also publish ONNX files",
6
+ "revision": "390624ed33d594443aa4aa221f5b9f283b545b5a",
7
+ "validation": "optimum (atol default)",
8
+ "opset_requested": 17,
9
+ "files": {
10
+ "decoder_model.onnx": {
11
+ "sha256": "c3994d01187c27f966be8f7b52806b5da270793235c3ff7cd6fffff83904ea93",
12
+ "bytes": 78110938,
13
+ "inputs": {
14
+ "encoder_attention_mask": [
15
+ "batch_size",
16
+ "encoder_sequence_length"
17
+ ],
18
+ "input_ids": [
19
+ "batch_size",
20
+ "decoder_sequence_length"
21
+ ],
22
+ "encoder_hidden_states": [
23
+ "batch_size",
24
+ "encoder_sequence_length",
25
+ 288
26
+ ]
27
+ },
28
+ "outputs": {
29
+ "logits": [
30
+ "batch_size",
31
+ "decoder_sequence_length",
32
+ 32768
33
+ ],
34
+ "present.0.decoder.key": [
35
+ "batch_size",
36
+ 8,
37
+ "past_decoder_sequence_length + decoder_sequence_length",
38
+ 36
39
+ ],
40
+ "present.0.decoder.value": [
41
+ "batch_size",
42
+ 8,
43
+ "past_decoder_sequence_length + decoder_sequence_length",
44
+ 36
45
+ ],
46
+ "present.0.encoder.key": [
47
+ "batch_size",
48
+ 8,
49
+ "encoder_sequence_length",
50
+ 36
51
+ ],
52
+ "present.0.encoder.value": [
53
+ "batch_size",
54
+ 8,
55
+ "encoder_sequence_length",
56
+ 36
57
+ ],
58
+ "present.1.decoder.key": [
59
+ "batch_size",
60
+ 8,
61
+ "past_decoder_sequence_length + decoder_sequence_length",
62
+ 36
63
+ ],
64
+ "present.1.decoder.value": [
65
+ "batch_size",
66
+ 8,
67
+ "past_decoder_sequence_length + decoder_sequence_length",
68
+ 36
69
+ ],
70
+ "present.1.encoder.key": [
71
+ "batch_size",
72
+ 8,
73
+ "encoder_sequence_length",
74
+ 36
75
+ ],
76
+ "present.1.encoder.value": [
77
+ "batch_size",
78
+ 8,
79
+ "encoder_sequence_length",
80
+ 36
81
+ ],
82
+ "present.2.decoder.key": [
83
+ "batch_size",
84
+ 8,
85
+ "past_decoder_sequence_length + decoder_sequence_length",
86
+ 36
87
+ ],
88
+ "present.2.decoder.value": [
89
+ "batch_size",
90
+ 8,
91
+ "past_decoder_sequence_length + decoder_sequence_length",
92
+ 36
93
+ ],
94
+ "present.2.encoder.key": [
95
+ "batch_size",
96
+ 8,
97
+ "encoder_sequence_length",
98
+ 36
99
+ ],
100
+ "present.2.encoder.value": [
101
+ "batch_size",
102
+ 8,
103
+ "encoder_sequence_length",
104
+ 36
105
+ ],
106
+ "present.3.decoder.key": [
107
+ "batch_size",
108
+ 8,
109
+ "past_decoder_sequence_length + decoder_sequence_length",
110
+ 36
111
+ ],
112
+ "present.3.decoder.value": [
113
+ "batch_size",
114
+ 8,
115
+ "past_decoder_sequence_length + decoder_sequence_length",
116
+ 36
117
+ ],
118
+ "present.3.encoder.key": [
119
+ "batch_size",
120
+ 8,
121
+ "encoder_sequence_length",
122
+ 36
123
+ ],
124
+ "present.3.encoder.value": [
125
+ "batch_size",
126
+ 8,
127
+ "encoder_sequence_length",
128
+ 36
129
+ ],
130
+ "present.4.decoder.key": [
131
+ "batch_size",
132
+ 8,
133
+ "past_decoder_sequence_length + decoder_sequence_length",
134
+ 36
135
+ ],
136
+ "present.4.decoder.value": [
137
+ "batch_size",
138
+ 8,
139
+ "past_decoder_sequence_length + decoder_sequence_length",
140
+ 36
141
+ ],
142
+ "present.4.encoder.key": [
143
+ "batch_size",
144
+ 8,
145
+ "encoder_sequence_length",
146
+ 36
147
+ ],
148
+ "present.4.encoder.value": [
149
+ "batch_size",
150
+ 8,
151
+ "encoder_sequence_length",
152
+ 36
153
+ ],
154
+ "present.5.decoder.key": [
155
+ "batch_size",
156
+ 8,
157
+ "past_decoder_sequence_length + decoder_sequence_length",
158
+ 36
159
+ ],
160
+ "present.5.decoder.value": [
161
+ "batch_size",
162
+ 8,
163
+ "past_decoder_sequence_length + decoder_sequence_length",
164
+ 36
165
+ ],
166
+ "present.5.encoder.key": [
167
+ "batch_size",
168
+ 8,
169
+ "encoder_sequence_length",
170
+ 36
171
+ ],
172
+ "present.5.encoder.value": [
173
+ "batch_size",
174
+ 8,
175
+ "encoder_sequence_length",
176
+ 36
177
+ ]
178
+ },
179
+ "opset": 17
180
+ },
181
+ "decoder_model_merged.onnx": {
182
+ "sha256": "28ae600e002f6e877f96174a7cd43128a3fd8b9545b1e662266a40c6ded8d3fd",
183
+ "bytes": 78563921,
184
+ "inputs": {
185
+ "encoder_attention_mask": [
186
+ "batch_size",
187
+ "encoder_sequence_length"
188
+ ],
189
+ "input_ids": [
190
+ "batch_size",
191
+ "decoder_sequence_length"
192
+ ],
193
+ "encoder_hidden_states": [
194
+ "batch_size",
195
+ "encoder_sequence_length",
196
+ 288
197
+ ],
198
+ "past_key_values.0.decoder.key": [
199
+ "batch_size",
200
+ 8,
201
+ "past_decoder_sequence_length",
202
+ 36
203
+ ],
204
+ "past_key_values.0.decoder.value": [
205
+ "batch_size",
206
+ 8,
207
+ "past_decoder_sequence_length",
208
+ 36
209
+ ],
210
+ "past_key_values.0.encoder.key": [
211
+ "batch_size",
212
+ 8,
213
+ "encoder_sequence_length",
214
+ 36
215
+ ],
216
+ "past_key_values.0.encoder.value": [
217
+ "batch_size",
218
+ 8,
219
+ "encoder_sequence_length",
220
+ 36
221
+ ],
222
+ "past_key_values.1.decoder.key": [
223
+ "batch_size",
224
+ 8,
225
+ "past_decoder_sequence_length",
226
+ 36
227
+ ],
228
+ "past_key_values.1.decoder.value": [
229
+ "batch_size",
230
+ 8,
231
+ "past_decoder_sequence_length",
232
+ 36
233
+ ],
234
+ "past_key_values.1.encoder.key": [
235
+ "batch_size",
236
+ 8,
237
+ "encoder_sequence_length",
238
+ 36
239
+ ],
240
+ "past_key_values.1.encoder.value": [
241
+ "batch_size",
242
+ 8,
243
+ "encoder_sequence_length",
244
+ 36
245
+ ],
246
+ "past_key_values.2.decoder.key": [
247
+ "batch_size",
248
+ 8,
249
+ "past_decoder_sequence_length",
250
+ 36
251
+ ],
252
+ "past_key_values.2.decoder.value": [
253
+ "batch_size",
254
+ 8,
255
+ "past_decoder_sequence_length",
256
+ 36
257
+ ],
258
+ "past_key_values.2.encoder.key": [
259
+ "batch_size",
260
+ 8,
261
+ "encoder_sequence_length",
262
+ 36
263
+ ],
264
+ "past_key_values.2.encoder.value": [
265
+ "batch_size",
266
+ 8,
267
+ "encoder_sequence_length",
268
+ 36
269
+ ],
270
+ "past_key_values.3.decoder.key": [
271
+ "batch_size",
272
+ 8,
273
+ "past_decoder_sequence_length",
274
+ 36
275
+ ],
276
+ "past_key_values.3.decoder.value": [
277
+ "batch_size",
278
+ 8,
279
+ "past_decoder_sequence_length",
280
+ 36
281
+ ],
282
+ "past_key_values.3.encoder.key": [
283
+ "batch_size",
284
+ 8,
285
+ "encoder_sequence_length",
286
+ 36
287
+ ],
288
+ "past_key_values.3.encoder.value": [
289
+ "batch_size",
290
+ 8,
291
+ "encoder_sequence_length",
292
+ 36
293
+ ],
294
+ "past_key_values.4.decoder.key": [
295
+ "batch_size",
296
+ 8,
297
+ "past_decoder_sequence_length",
298
+ 36
299
+ ],
300
+ "past_key_values.4.decoder.value": [
301
+ "batch_size",
302
+ 8,
303
+ "past_decoder_sequence_length",
304
+ 36
305
+ ],
306
+ "past_key_values.4.encoder.key": [
307
+ "batch_size",
308
+ 8,
309
+ "encoder_sequence_length",
310
+ 36
311
+ ],
312
+ "past_key_values.4.encoder.value": [
313
+ "batch_size",
314
+ 8,
315
+ "encoder_sequence_length",
316
+ 36
317
+ ],
318
+ "past_key_values.5.decoder.key": [
319
+ "batch_size",
320
+ 8,
321
+ "past_decoder_sequence_length",
322
+ 36
323
+ ],
324
+ "past_key_values.5.decoder.value": [
325
+ "batch_size",
326
+ 8,
327
+ "past_decoder_sequence_length",
328
+ 36
329
+ ],
330
+ "past_key_values.5.encoder.key": [
331
+ "batch_size",
332
+ 8,
333
+ "encoder_sequence_length",
334
+ 36
335
+ ],
336
+ "past_key_values.5.encoder.value": [
337
+ "batch_size",
338
+ 8,
339
+ "encoder_sequence_length",
340
+ 36
341
+ ],
342
+ "use_cache_branch": [
343
+ 1
344
+ ]
345
+ },
346
+ "outputs": {
347
+ "logits": [
348
+ "batch_size",
349
+ "decoder_sequence_length",
350
+ 32768
351
+ ],
352
+ "present.0.decoder.key": [
353
+ "batch_size",
354
+ 8,
355
+ "past_decoder_sequence_length + decoder_sequence_length",
356
+ 36
357
+ ],
358
+ "present.0.decoder.value": [
359
+ "batch_size",
360
+ 8,
361
+ "past_decoder_sequence_length + decoder_sequence_length",
362
+ 36
363
+ ],
364
+ "present.0.encoder.key": [
365
+ "batch_size",
366
+ 8,
367
+ "encoder_sequence_length",
368
+ 36
369
+ ],
370
+ "present.0.encoder.value": [
371
+ "batch_size",
372
+ 8,
373
+ "encoder_sequence_length",
374
+ 36
375
+ ],
376
+ "present.1.decoder.key": [
377
+ "batch_size",
378
+ 8,
379
+ "past_decoder_sequence_length + decoder_sequence_length",
380
+ 36
381
+ ],
382
+ "present.1.decoder.value": [
383
+ "batch_size",
384
+ 8,
385
+ "past_decoder_sequence_length + decoder_sequence_length",
386
+ 36
387
+ ],
388
+ "present.1.encoder.key": [
389
+ "batch_size",
390
+ 8,
391
+ "encoder_sequence_length",
392
+ 36
393
+ ],
394
+ "present.1.encoder.value": [
395
+ "batch_size",
396
+ 8,
397
+ "encoder_sequence_length",
398
+ 36
399
+ ],
400
+ "present.2.decoder.key": [
401
+ "batch_size",
402
+ 8,
403
+ "past_decoder_sequence_length + decoder_sequence_length",
404
+ 36
405
+ ],
406
+ "present.2.decoder.value": [
407
+ "batch_size",
408
+ 8,
409
+ "past_decoder_sequence_length + decoder_sequence_length",
410
+ 36
411
+ ],
412
+ "present.2.encoder.key": [
413
+ "batch_size",
414
+ 8,
415
+ "encoder_sequence_length",
416
+ 36
417
+ ],
418
+ "present.2.encoder.value": [
419
+ "batch_size",
420
+ 8,
421
+ "encoder_sequence_length",
422
+ 36
423
+ ],
424
+ "present.3.decoder.key": [
425
+ "batch_size",
426
+ 8,
427
+ "past_decoder_sequence_length + decoder_sequence_length",
428
+ 36
429
+ ],
430
+ "present.3.decoder.value": [
431
+ "batch_size",
432
+ 8,
433
+ "past_decoder_sequence_length + decoder_sequence_length",
434
+ 36
435
+ ],
436
+ "present.3.encoder.key": [
437
+ "batch_size",
438
+ 8,
439
+ "encoder_sequence_length",
440
+ 36
441
+ ],
442
+ "present.3.encoder.value": [
443
+ "batch_size",
444
+ 8,
445
+ "encoder_sequence_length",
446
+ 36
447
+ ],
448
+ "present.4.decoder.key": [
449
+ "batch_size",
450
+ 8,
451
+ "past_decoder_sequence_length + decoder_sequence_length",
452
+ 36
453
+ ],
454
+ "present.4.decoder.value": [
455
+ "batch_size",
456
+ 8,
457
+ "past_decoder_sequence_length + decoder_sequence_length",
458
+ 36
459
+ ],
460
+ "present.4.encoder.key": [
461
+ "batch_size",
462
+ 8,
463
+ "encoder_sequence_length",
464
+ 36
465
+ ],
466
+ "present.4.encoder.value": [
467
+ "batch_size",
468
+ 8,
469
+ "encoder_sequence_length",
470
+ 36
471
+ ],
472
+ "present.5.decoder.key": [
473
+ "batch_size",
474
+ 8,
475
+ "past_decoder_sequence_length + decoder_sequence_length",
476
+ 36
477
+ ],
478
+ "present.5.decoder.value": [
479
+ "batch_size",
480
+ 8,
481
+ "past_decoder_sequence_length + decoder_sequence_length",
482
+ 36
483
+ ],
484
+ "present.5.encoder.key": [
485
+ "batch_size",
486
+ 8,
487
+ "encoder_sequence_length",
488
+ 36
489
+ ],
490
+ "present.5.encoder.value": [
491
+ "batch_size",
492
+ 8,
493
+ "encoder_sequence_length",
494
+ 36
495
+ ]
496
+ },
497
+ "opset": 17
498
+ },
499
+ "decoder_with_past_model.onnx": {
500
+ "sha256": "0fcb7d0d3921098415e2652c84d3d9364e3acb4f9253877713ac52e56e4dff37",
501
+ "bytes": 74113712,
502
+ "inputs": {
503
+ "encoder_attention_mask": [
504
+ "batch_size",
505
+ "encoder_sequence_length"
506
+ ],
507
+ "input_ids": [
508
+ "batch_size",
509
+ "decoder_sequence_length"
510
+ ],
511
+ "encoder_hidden_states": [
512
+ "batch_size",
513
+ "encoder_sequence_length",
514
+ 288
515
+ ],
516
+ "past_key_values.0.decoder.key": [
517
+ "batch_size",
518
+ 8,
519
+ "past_decoder_sequence_length",
520
+ 36
521
+ ],
522
+ "past_key_values.0.decoder.value": [
523
+ "batch_size",
524
+ 8,
525
+ "past_decoder_sequence_length",
526
+ 36
527
+ ],
528
+ "past_key_values.0.encoder.key": [
529
+ "batch_size",
530
+ 8,
531
+ "encoder_sequence_length",
532
+ 36
533
+ ],
534
+ "past_key_values.0.encoder.value": [
535
+ "batch_size",
536
+ 8,
537
+ "encoder_sequence_length",
538
+ 36
539
+ ],
540
+ "past_key_values.1.decoder.key": [
541
+ "batch_size",
542
+ 8,
543
+ "past_decoder_sequence_length",
544
+ 36
545
+ ],
546
+ "past_key_values.1.decoder.value": [
547
+ "batch_size",
548
+ 8,
549
+ "past_decoder_sequence_length",
550
+ 36
551
+ ],
552
+ "past_key_values.1.encoder.key": [
553
+ "batch_size",
554
+ 8,
555
+ "encoder_sequence_length",
556
+ 36
557
+ ],
558
+ "past_key_values.1.encoder.value": [
559
+ "batch_size",
560
+ 8,
561
+ "encoder_sequence_length",
562
+ 36
563
+ ],
564
+ "past_key_values.2.decoder.key": [
565
+ "batch_size",
566
+ 8,
567
+ "past_decoder_sequence_length",
568
+ 36
569
+ ],
570
+ "past_key_values.2.decoder.value": [
571
+ "batch_size",
572
+ 8,
573
+ "past_decoder_sequence_length",
574
+ 36
575
+ ],
576
+ "past_key_values.2.encoder.key": [
577
+ "batch_size",
578
+ 8,
579
+ "encoder_sequence_length",
580
+ 36
581
+ ],
582
+ "past_key_values.2.encoder.value": [
583
+ "batch_size",
584
+ 8,
585
+ "encoder_sequence_length",
586
+ 36
587
+ ],
588
+ "past_key_values.3.decoder.key": [
589
+ "batch_size",
590
+ 8,
591
+ "past_decoder_sequence_length",
592
+ 36
593
+ ],
594
+ "past_key_values.3.decoder.value": [
595
+ "batch_size",
596
+ 8,
597
+ "past_decoder_sequence_length",
598
+ 36
599
+ ],
600
+ "past_key_values.3.encoder.key": [
601
+ "batch_size",
602
+ 8,
603
+ "encoder_sequence_length",
604
+ 36
605
+ ],
606
+ "past_key_values.3.encoder.value": [
607
+ "batch_size",
608
+ 8,
609
+ "encoder_sequence_length",
610
+ 36
611
+ ],
612
+ "past_key_values.4.decoder.key": [
613
+ "batch_size",
614
+ 8,
615
+ "past_decoder_sequence_length",
616
+ 36
617
+ ],
618
+ "past_key_values.4.decoder.value": [
619
+ "batch_size",
620
+ 8,
621
+ "past_decoder_sequence_length",
622
+ 36
623
+ ],
624
+ "past_key_values.4.encoder.key": [
625
+ "batch_size",
626
+ 8,
627
+ "encoder_sequence_length",
628
+ 36
629
+ ],
630
+ "past_key_values.4.encoder.value": [
631
+ "batch_size",
632
+ 8,
633
+ "encoder_sequence_length",
634
+ 36
635
+ ],
636
+ "past_key_values.5.decoder.key": [
637
+ "batch_size",
638
+ 8,
639
+ "past_decoder_sequence_length",
640
+ 36
641
+ ],
642
+ "past_key_values.5.decoder.value": [
643
+ "batch_size",
644
+ 8,
645
+ "past_decoder_sequence_length",
646
+ 36
647
+ ],
648
+ "past_key_values.5.encoder.key": [
649
+ "batch_size",
650
+ 8,
651
+ "encoder_sequence_length",
652
+ 36
653
+ ],
654
+ "past_key_values.5.encoder.value": [
655
+ "batch_size",
656
+ 8,
657
+ "encoder_sequence_length",
658
+ 36
659
+ ]
660
+ },
661
+ "outputs": {
662
+ "logits": [
663
+ "batch_size",
664
+ "1",
665
+ 32768
666
+ ],
667
+ "present.0.decoder.key": [
668
+ "batch_size",
669
+ 8,
670
+ "past_decoder_sequence_length + decoder_sequence_length",
671
+ 36
672
+ ],
673
+ "present.0.decoder.value": [
674
+ "batch_size",
675
+ 8,
676
+ "past_decoder_sequence_length + decoder_sequence_length",
677
+ 36
678
+ ],
679
+ "present.1.decoder.key": [
680
+ "batch_size",
681
+ 8,
682
+ "past_decoder_sequence_length + decoder_sequence_length",
683
+ 36
684
+ ],
685
+ "present.1.decoder.value": [
686
+ "batch_size",
687
+ 8,
688
+ "past_decoder_sequence_length + decoder_sequence_length",
689
+ 36
690
+ ],
691
+ "present.2.decoder.key": [
692
+ "batch_size",
693
+ 8,
694
+ "past_decoder_sequence_length + decoder_sequence_length",
695
+ 36
696
+ ],
697
+ "present.2.decoder.value": [
698
+ "batch_size",
699
+ 8,
700
+ "past_decoder_sequence_length + decoder_sequence_length",
701
+ 36
702
+ ],
703
+ "present.3.decoder.key": [
704
+ "batch_size",
705
+ 8,
706
+ "past_decoder_sequence_length + decoder_sequence_length",
707
+ 36
708
+ ],
709
+ "present.3.decoder.value": [
710
+ "batch_size",
711
+ 8,
712
+ "past_decoder_sequence_length + decoder_sequence_length",
713
+ 36
714
+ ],
715
+ "present.4.decoder.key": [
716
+ "batch_size",
717
+ 8,
718
+ "past_decoder_sequence_length + decoder_sequence_length",
719
+ 36
720
+ ],
721
+ "present.4.decoder.value": [
722
+ "batch_size",
723
+ 8,
724
+ "past_decoder_sequence_length + decoder_sequence_length",
725
+ 36
726
+ ],
727
+ "present.5.decoder.key": [
728
+ "batch_size",
729
+ 8,
730
+ "past_decoder_sequence_length + decoder_sequence_length",
731
+ 36
732
+ ],
733
+ "present.5.decoder.value": [
734
+ "batch_size",
735
+ 8,
736
+ "past_decoder_sequence_length + decoder_sequence_length",
737
+ 36
738
+ ]
739
+ },
740
+ "opset": 17
741
+ },
742
+ "encoder_model.onnx": {
743
+ "sha256": "a3ea57a0ce80bf2515b6ce218076956c09a7dec307089d43f34491c46c44e60f",
744
+ "bytes": 30962848,
745
+ "inputs": {
746
+ "input_values": [
747
+ "batch_size",
748
+ "encoder_sequence_length"
749
+ ],
750
+ "attention_mask": [
751
+ "batch_size",
752
+ "encoder_sequence_length"
753
+ ]
754
+ },
755
+ "outputs": {
756
+ "last_hidden_state": [
757
+ "batch_size",
758
+ "( ( ( encoder_sequence_length - 127 ) // 64 + 1 - 7 ) // 3 + 1 - 3 ) // 2 + 1",
759
+ 288
760
+ ]
761
+ },
762
+ "opset": 17
763
+ }
764
+ },
765
+ "exported_at": "2026-10-19T18:45:24.363061+00:00"
766
+ }
moonshine-tiny/preprocessor_config.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_normalize": false,
3
+ "feature_extractor_type": "Wav2Vec2FeatureExtractor",
4
+ "feature_size": 1,
5
+ "padding_side": "right",
6
+ "padding_value": 0.0,
7
+ "return_attention_mask": true,
8
+ "sampling_rate": 16000
9
+ }
moonshine-tiny/special_tokens_map.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {}
moonshine-tiny/tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
moonshine-tiny/tokenizer_config.json ADDED
The diff for this file is too large to render. See raw diff