Felladrin commited on
Commit
2fe6704
·
verified ·
1 Parent(s): e628bc5

Upload folder using huggingface_hub

Browse files
README.md ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: transformers.js
3
+ base_model:
4
+ - MIT/ast-finetuned-audioset-10-10-0.4593
5
+ ---
6
+
7
+ # ast-finetuned-audioset-10-10-0.4593 (ONNX)
8
+
9
+ This is an ONNX version of [MIT/ast-finetuned-audioset-10-10-0.4593](https://huggingface.co/MIT/ast-finetuned-audioset-10-10-0.4593). It was automatically converted and uploaded using [this space](https://huggingface.co/spaces/onnx-community/convert-to-onnx).
config.json ADDED
@@ -0,0 +1,1083 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_attn_implementation_autoset": true,
3
+ "_name_or_path": "MIT/ast-finetuned-audioset-10-10-0.4593",
4
+ "architectures": [
5
+ "ASTForAudioClassification"
6
+ ],
7
+ "attention_probs_dropout_prob": 0.0,
8
+ "frequency_stride": 10,
9
+ "hidden_act": "gelu",
10
+ "hidden_dropout_prob": 0.0,
11
+ "hidden_size": 768,
12
+ "id2label": {
13
+ "0": "Speech",
14
+ "1": "Male speech, man speaking",
15
+ "2": "Female speech, woman speaking",
16
+ "3": "Child speech, kid speaking",
17
+ "4": "Conversation",
18
+ "5": "Narration, monologue",
19
+ "6": "Babbling",
20
+ "7": "Speech synthesizer",
21
+ "8": "Shout",
22
+ "9": "Bellow",
23
+ "10": "Whoop",
24
+ "11": "Yell",
25
+ "12": "Battle cry",
26
+ "13": "Children shouting",
27
+ "14": "Screaming",
28
+ "15": "Whispering",
29
+ "16": "Laughter",
30
+ "17": "Baby laughter",
31
+ "18": "Giggle",
32
+ "19": "Snicker",
33
+ "20": "Belly laugh",
34
+ "21": "Chuckle, chortle",
35
+ "22": "Crying, sobbing",
36
+ "23": "Baby cry, infant cry",
37
+ "24": "Whimper",
38
+ "25": "Wail, moan",
39
+ "26": "Sigh",
40
+ "27": "Singing",
41
+ "28": "Choir",
42
+ "29": "Yodeling",
43
+ "30": "Chant",
44
+ "31": "Mantra",
45
+ "32": "Male singing",
46
+ "33": "Female singing",
47
+ "34": "Child singing",
48
+ "35": "Synthetic singing",
49
+ "36": "Rapping",
50
+ "37": "Humming",
51
+ "38": "Groan",
52
+ "39": "Grunt",
53
+ "40": "Whistling",
54
+ "41": "Breathing",
55
+ "42": "Wheeze",
56
+ "43": "Snoring",
57
+ "44": "Gasp",
58
+ "45": "Pant",
59
+ "46": "Snort",
60
+ "47": "Cough",
61
+ "48": "Throat clearing",
62
+ "49": "Sneeze",
63
+ "50": "Sniff",
64
+ "51": "Run",
65
+ "52": "Shuffle",
66
+ "53": "Walk, footsteps",
67
+ "54": "Chewing, mastication",
68
+ "55": "Biting",
69
+ "56": "Gargling",
70
+ "57": "Stomach rumble",
71
+ "58": "Burping, eructation",
72
+ "59": "Hiccup",
73
+ "60": "Fart",
74
+ "61": "Hands",
75
+ "62": "Finger snapping",
76
+ "63": "Clapping",
77
+ "64": "Heart sounds, heartbeat",
78
+ "65": "Heart murmur",
79
+ "66": "Cheering",
80
+ "67": "Applause",
81
+ "68": "Chatter",
82
+ "69": "Crowd",
83
+ "70": "Hubbub, speech noise, speech babble",
84
+ "71": "Children playing",
85
+ "72": "Animal",
86
+ "73": "Domestic animals, pets",
87
+ "74": "Dog",
88
+ "75": "Bark",
89
+ "76": "Yip",
90
+ "77": "Howl",
91
+ "78": "Bow-wow",
92
+ "79": "Growling",
93
+ "80": "Whimper (dog)",
94
+ "81": "Cat",
95
+ "82": "Purr",
96
+ "83": "Meow",
97
+ "84": "Hiss",
98
+ "85": "Caterwaul",
99
+ "86": "Livestock, farm animals, working animals",
100
+ "87": "Horse",
101
+ "88": "Clip-clop",
102
+ "89": "Neigh, whinny",
103
+ "90": "Cattle, bovinae",
104
+ "91": "Moo",
105
+ "92": "Cowbell",
106
+ "93": "Pig",
107
+ "94": "Oink",
108
+ "95": "Goat",
109
+ "96": "Bleat",
110
+ "97": "Sheep",
111
+ "98": "Fowl",
112
+ "99": "Chicken, rooster",
113
+ "100": "Cluck",
114
+ "101": "Crowing, cock-a-doodle-doo",
115
+ "102": "Turkey",
116
+ "103": "Gobble",
117
+ "104": "Duck",
118
+ "105": "Quack",
119
+ "106": "Goose",
120
+ "107": "Honk",
121
+ "108": "Wild animals",
122
+ "109": "Roaring cats (lions, tigers)",
123
+ "110": "Roar",
124
+ "111": "Bird",
125
+ "112": "Bird vocalization, bird call, bird song",
126
+ "113": "Chirp, tweet",
127
+ "114": "Squawk",
128
+ "115": "Pigeon, dove",
129
+ "116": "Coo",
130
+ "117": "Crow",
131
+ "118": "Caw",
132
+ "119": "Owl",
133
+ "120": "Hoot",
134
+ "121": "Bird flight, flapping wings",
135
+ "122": "Canidae, dogs, wolves",
136
+ "123": "Rodents, rats, mice",
137
+ "124": "Mouse",
138
+ "125": "Patter",
139
+ "126": "Insect",
140
+ "127": "Cricket",
141
+ "128": "Mosquito",
142
+ "129": "Fly, housefly",
143
+ "130": "Buzz",
144
+ "131": "Bee, wasp, etc.",
145
+ "132": "Frog",
146
+ "133": "Croak",
147
+ "134": "Snake",
148
+ "135": "Rattle",
149
+ "136": "Whale vocalization",
150
+ "137": "Music",
151
+ "138": "Musical instrument",
152
+ "139": "Plucked string instrument",
153
+ "140": "Guitar",
154
+ "141": "Electric guitar",
155
+ "142": "Bass guitar",
156
+ "143": "Acoustic guitar",
157
+ "144": "Steel guitar, slide guitar",
158
+ "145": "Tapping (guitar technique)",
159
+ "146": "Strum",
160
+ "147": "Banjo",
161
+ "148": "Sitar",
162
+ "149": "Mandolin",
163
+ "150": "Zither",
164
+ "151": "Ukulele",
165
+ "152": "Keyboard (musical)",
166
+ "153": "Piano",
167
+ "154": "Electric piano",
168
+ "155": "Organ",
169
+ "156": "Electronic organ",
170
+ "157": "Hammond organ",
171
+ "158": "Synthesizer",
172
+ "159": "Sampler",
173
+ "160": "Harpsichord",
174
+ "161": "Percussion",
175
+ "162": "Drum kit",
176
+ "163": "Drum machine",
177
+ "164": "Drum",
178
+ "165": "Snare drum",
179
+ "166": "Rimshot",
180
+ "167": "Drum roll",
181
+ "168": "Bass drum",
182
+ "169": "Timpani",
183
+ "170": "Tabla",
184
+ "171": "Cymbal",
185
+ "172": "Hi-hat",
186
+ "173": "Wood block",
187
+ "174": "Tambourine",
188
+ "175": "Rattle (instrument)",
189
+ "176": "Maraca",
190
+ "177": "Gong",
191
+ "178": "Tubular bells",
192
+ "179": "Mallet percussion",
193
+ "180": "Marimba, xylophone",
194
+ "181": "Glockenspiel",
195
+ "182": "Vibraphone",
196
+ "183": "Steelpan",
197
+ "184": "Orchestra",
198
+ "185": "Brass instrument",
199
+ "186": "French horn",
200
+ "187": "Trumpet",
201
+ "188": "Trombone",
202
+ "189": "Bowed string instrument",
203
+ "190": "String section",
204
+ "191": "Violin, fiddle",
205
+ "192": "Pizzicato",
206
+ "193": "Cello",
207
+ "194": "Double bass",
208
+ "195": "Wind instrument, woodwind instrument",
209
+ "196": "Flute",
210
+ "197": "Saxophone",
211
+ "198": "Clarinet",
212
+ "199": "Harp",
213
+ "200": "Bell",
214
+ "201": "Church bell",
215
+ "202": "Jingle bell",
216
+ "203": "Bicycle bell",
217
+ "204": "Tuning fork",
218
+ "205": "Chime",
219
+ "206": "Wind chime",
220
+ "207": "Change ringing (campanology)",
221
+ "208": "Harmonica",
222
+ "209": "Accordion",
223
+ "210": "Bagpipes",
224
+ "211": "Didgeridoo",
225
+ "212": "Shofar",
226
+ "213": "Theremin",
227
+ "214": "Singing bowl",
228
+ "215": "Scratching (performance technique)",
229
+ "216": "Pop music",
230
+ "217": "Hip hop music",
231
+ "218": "Beatboxing",
232
+ "219": "Rock music",
233
+ "220": "Heavy metal",
234
+ "221": "Punk rock",
235
+ "222": "Grunge",
236
+ "223": "Progressive rock",
237
+ "224": "Rock and roll",
238
+ "225": "Psychedelic rock",
239
+ "226": "Rhythm and blues",
240
+ "227": "Soul music",
241
+ "228": "Reggae",
242
+ "229": "Country",
243
+ "230": "Swing music",
244
+ "231": "Bluegrass",
245
+ "232": "Funk",
246
+ "233": "Folk music",
247
+ "234": "Middle Eastern music",
248
+ "235": "Jazz",
249
+ "236": "Disco",
250
+ "237": "Classical music",
251
+ "238": "Opera",
252
+ "239": "Electronic music",
253
+ "240": "House music",
254
+ "241": "Techno",
255
+ "242": "Dubstep",
256
+ "243": "Drum and bass",
257
+ "244": "Electronica",
258
+ "245": "Electronic dance music",
259
+ "246": "Ambient music",
260
+ "247": "Trance music",
261
+ "248": "Music of Latin America",
262
+ "249": "Salsa music",
263
+ "250": "Flamenco",
264
+ "251": "Blues",
265
+ "252": "Music for children",
266
+ "253": "New-age music",
267
+ "254": "Vocal music",
268
+ "255": "A capella",
269
+ "256": "Music of Africa",
270
+ "257": "Afrobeat",
271
+ "258": "Christian music",
272
+ "259": "Gospel music",
273
+ "260": "Music of Asia",
274
+ "261": "Carnatic music",
275
+ "262": "Music of Bollywood",
276
+ "263": "Ska",
277
+ "264": "Traditional music",
278
+ "265": "Independent music",
279
+ "266": "Song",
280
+ "267": "Background music",
281
+ "268": "Theme music",
282
+ "269": "Jingle (music)",
283
+ "270": "Soundtrack music",
284
+ "271": "Lullaby",
285
+ "272": "Video game music",
286
+ "273": "Christmas music",
287
+ "274": "Dance music",
288
+ "275": "Wedding music",
289
+ "276": "Happy music",
290
+ "277": "Funny music",
291
+ "278": "Sad music",
292
+ "279": "Tender music",
293
+ "280": "Exciting music",
294
+ "281": "Angry music",
295
+ "282": "Scary music",
296
+ "283": "Wind",
297
+ "284": "Rustling leaves",
298
+ "285": "Wind noise (microphone)",
299
+ "286": "Thunderstorm",
300
+ "287": "Thunder",
301
+ "288": "Water",
302
+ "289": "Rain",
303
+ "290": "Raindrop",
304
+ "291": "Rain on surface",
305
+ "292": "Stream",
306
+ "293": "Waterfall",
307
+ "294": "Ocean",
308
+ "295": "Waves, surf",
309
+ "296": "Steam",
310
+ "297": "Gurgling",
311
+ "298": "Fire",
312
+ "299": "Crackle",
313
+ "300": "Vehicle",
314
+ "301": "Boat, Water vehicle",
315
+ "302": "Sailboat, sailing ship",
316
+ "303": "Rowboat, canoe, kayak",
317
+ "304": "Motorboat, speedboat",
318
+ "305": "Ship",
319
+ "306": "Motor vehicle (road)",
320
+ "307": "Car",
321
+ "308": "Vehicle horn, car horn, honking",
322
+ "309": "Toot",
323
+ "310": "Car alarm",
324
+ "311": "Power windows, electric windows",
325
+ "312": "Skidding",
326
+ "313": "Tire squeal",
327
+ "314": "Car passing by",
328
+ "315": "Race car, auto racing",
329
+ "316": "Truck",
330
+ "317": "Air brake",
331
+ "318": "Air horn, truck horn",
332
+ "319": "Reversing beeps",
333
+ "320": "Ice cream truck, ice cream van",
334
+ "321": "Bus",
335
+ "322": "Emergency vehicle",
336
+ "323": "Police car (siren)",
337
+ "324": "Ambulance (siren)",
338
+ "325": "Fire engine, fire truck (siren)",
339
+ "326": "Motorcycle",
340
+ "327": "Traffic noise, roadway noise",
341
+ "328": "Rail transport",
342
+ "329": "Train",
343
+ "330": "Train whistle",
344
+ "331": "Train horn",
345
+ "332": "Railroad car, train wagon",
346
+ "333": "Train wheels squealing",
347
+ "334": "Subway, metro, underground",
348
+ "335": "Aircraft",
349
+ "336": "Aircraft engine",
350
+ "337": "Jet engine",
351
+ "338": "Propeller, airscrew",
352
+ "339": "Helicopter",
353
+ "340": "Fixed-wing aircraft, airplane",
354
+ "341": "Bicycle",
355
+ "342": "Skateboard",
356
+ "343": "Engine",
357
+ "344": "Light engine (high frequency)",
358
+ "345": "Dental drill, dentist's drill",
359
+ "346": "Lawn mower",
360
+ "347": "Chainsaw",
361
+ "348": "Medium engine (mid frequency)",
362
+ "349": "Heavy engine (low frequency)",
363
+ "350": "Engine knocking",
364
+ "351": "Engine starting",
365
+ "352": "Idling",
366
+ "353": "Accelerating, revving, vroom",
367
+ "354": "Door",
368
+ "355": "Doorbell",
369
+ "356": "Ding-dong",
370
+ "357": "Sliding door",
371
+ "358": "Slam",
372
+ "359": "Knock",
373
+ "360": "Tap",
374
+ "361": "Squeak",
375
+ "362": "Cupboard open or close",
376
+ "363": "Drawer open or close",
377
+ "364": "Dishes, pots, and pans",
378
+ "365": "Cutlery, silverware",
379
+ "366": "Chopping (food)",
380
+ "367": "Frying (food)",
381
+ "368": "Microwave oven",
382
+ "369": "Blender",
383
+ "370": "Water tap, faucet",
384
+ "371": "Sink (filling or washing)",
385
+ "372": "Bathtub (filling or washing)",
386
+ "373": "Hair dryer",
387
+ "374": "Toilet flush",
388
+ "375": "Toothbrush",
389
+ "376": "Electric toothbrush",
390
+ "377": "Vacuum cleaner",
391
+ "378": "Zipper (clothing)",
392
+ "379": "Keys jangling",
393
+ "380": "Coin (dropping)",
394
+ "381": "Scissors",
395
+ "382": "Electric shaver, electric razor",
396
+ "383": "Shuffling cards",
397
+ "384": "Typing",
398
+ "385": "Typewriter",
399
+ "386": "Computer keyboard",
400
+ "387": "Writing",
401
+ "388": "Alarm",
402
+ "389": "Telephone",
403
+ "390": "Telephone bell ringing",
404
+ "391": "Ringtone",
405
+ "392": "Telephone dialing, DTMF",
406
+ "393": "Dial tone",
407
+ "394": "Busy signal",
408
+ "395": "Alarm clock",
409
+ "396": "Siren",
410
+ "397": "Civil defense siren",
411
+ "398": "Buzzer",
412
+ "399": "Smoke detector, smoke alarm",
413
+ "400": "Fire alarm",
414
+ "401": "Foghorn",
415
+ "402": "Whistle",
416
+ "403": "Steam whistle",
417
+ "404": "Mechanisms",
418
+ "405": "Ratchet, pawl",
419
+ "406": "Clock",
420
+ "407": "Tick",
421
+ "408": "Tick-tock",
422
+ "409": "Gears",
423
+ "410": "Pulleys",
424
+ "411": "Sewing machine",
425
+ "412": "Mechanical fan",
426
+ "413": "Air conditioning",
427
+ "414": "Cash register",
428
+ "415": "Printer",
429
+ "416": "Camera",
430
+ "417": "Single-lens reflex camera",
431
+ "418": "Tools",
432
+ "419": "Hammer",
433
+ "420": "Jackhammer",
434
+ "421": "Sawing",
435
+ "422": "Filing (rasp)",
436
+ "423": "Sanding",
437
+ "424": "Power tool",
438
+ "425": "Drill",
439
+ "426": "Explosion",
440
+ "427": "Gunshot, gunfire",
441
+ "428": "Machine gun",
442
+ "429": "Fusillade",
443
+ "430": "Artillery fire",
444
+ "431": "Cap gun",
445
+ "432": "Fireworks",
446
+ "433": "Firecracker",
447
+ "434": "Burst, pop",
448
+ "435": "Eruption",
449
+ "436": "Boom",
450
+ "437": "Wood",
451
+ "438": "Chop",
452
+ "439": "Splinter",
453
+ "440": "Crack",
454
+ "441": "Glass",
455
+ "442": "Chink, clink",
456
+ "443": "Shatter",
457
+ "444": "Liquid",
458
+ "445": "Splash, splatter",
459
+ "446": "Slosh",
460
+ "447": "Squish",
461
+ "448": "Drip",
462
+ "449": "Pour",
463
+ "450": "Trickle, dribble",
464
+ "451": "Gush",
465
+ "452": "Fill (with liquid)",
466
+ "453": "Spray",
467
+ "454": "Pump (liquid)",
468
+ "455": "Stir",
469
+ "456": "Boiling",
470
+ "457": "Sonar",
471
+ "458": "Arrow",
472
+ "459": "Whoosh, swoosh, swish",
473
+ "460": "Thump, thud",
474
+ "461": "Thunk",
475
+ "462": "Electronic tuner",
476
+ "463": "Effects unit",
477
+ "464": "Chorus effect",
478
+ "465": "Basketball bounce",
479
+ "466": "Bang",
480
+ "467": "Slap, smack",
481
+ "468": "Whack, thwack",
482
+ "469": "Smash, crash",
483
+ "470": "Breaking",
484
+ "471": "Bouncing",
485
+ "472": "Whip",
486
+ "473": "Flap",
487
+ "474": "Scratch",
488
+ "475": "Scrape",
489
+ "476": "Rub",
490
+ "477": "Roll",
491
+ "478": "Crushing",
492
+ "479": "Crumpling, crinkling",
493
+ "480": "Tearing",
494
+ "481": "Beep, bleep",
495
+ "482": "Ping",
496
+ "483": "Ding",
497
+ "484": "Clang",
498
+ "485": "Squeal",
499
+ "486": "Creak",
500
+ "487": "Rustle",
501
+ "488": "Whir",
502
+ "489": "Clatter",
503
+ "490": "Sizzle",
504
+ "491": "Clicking",
505
+ "492": "Clickety-clack",
506
+ "493": "Rumble",
507
+ "494": "Plop",
508
+ "495": "Jingle, tinkle",
509
+ "496": "Hum",
510
+ "497": "Zing",
511
+ "498": "Boing",
512
+ "499": "Crunch",
513
+ "500": "Silence",
514
+ "501": "Sine wave",
515
+ "502": "Harmonic",
516
+ "503": "Chirp tone",
517
+ "504": "Sound effect",
518
+ "505": "Pulse",
519
+ "506": "Inside, small room",
520
+ "507": "Inside, large room or hall",
521
+ "508": "Inside, public space",
522
+ "509": "Outside, urban or manmade",
523
+ "510": "Outside, rural or natural",
524
+ "511": "Reverberation",
525
+ "512": "Echo",
526
+ "513": "Noise",
527
+ "514": "Environmental noise",
528
+ "515": "Static",
529
+ "516": "Mains hum",
530
+ "517": "Distortion",
531
+ "518": "Sidetone",
532
+ "519": "Cacophony",
533
+ "520": "White noise",
534
+ "521": "Pink noise",
535
+ "522": "Throbbing",
536
+ "523": "Vibration",
537
+ "524": "Television",
538
+ "525": "Radio",
539
+ "526": "Field recording"
540
+ },
541
+ "initializer_range": 0.02,
542
+ "intermediate_size": 3072,
543
+ "label2id": {
544
+ "A capella": 255,
545
+ "Accelerating, revving, vroom": 353,
546
+ "Accordion": 209,
547
+ "Acoustic guitar": 143,
548
+ "Afrobeat": 257,
549
+ "Air brake": 317,
550
+ "Air conditioning": 413,
551
+ "Air horn, truck horn": 318,
552
+ "Aircraft": 335,
553
+ "Aircraft engine": 336,
554
+ "Alarm": 388,
555
+ "Alarm clock": 395,
556
+ "Ambient music": 246,
557
+ "Ambulance (siren)": 324,
558
+ "Angry music": 281,
559
+ "Animal": 72,
560
+ "Applause": 67,
561
+ "Arrow": 458,
562
+ "Artillery fire": 430,
563
+ "Babbling": 6,
564
+ "Baby cry, infant cry": 23,
565
+ "Baby laughter": 17,
566
+ "Background music": 267,
567
+ "Bagpipes": 210,
568
+ "Bang": 466,
569
+ "Banjo": 147,
570
+ "Bark": 75,
571
+ "Basketball bounce": 465,
572
+ "Bass drum": 168,
573
+ "Bass guitar": 142,
574
+ "Bathtub (filling or washing)": 372,
575
+ "Battle cry": 12,
576
+ "Beatboxing": 218,
577
+ "Bee, wasp, etc.": 131,
578
+ "Beep, bleep": 481,
579
+ "Bell": 200,
580
+ "Bellow": 9,
581
+ "Belly laugh": 20,
582
+ "Bicycle": 341,
583
+ "Bicycle bell": 203,
584
+ "Bird": 111,
585
+ "Bird flight, flapping wings": 121,
586
+ "Bird vocalization, bird call, bird song": 112,
587
+ "Biting": 55,
588
+ "Bleat": 96,
589
+ "Blender": 369,
590
+ "Bluegrass": 231,
591
+ "Blues": 251,
592
+ "Boat, Water vehicle": 301,
593
+ "Boiling": 456,
594
+ "Boing": 498,
595
+ "Boom": 436,
596
+ "Bouncing": 471,
597
+ "Bow-wow": 78,
598
+ "Bowed string instrument": 189,
599
+ "Brass instrument": 185,
600
+ "Breaking": 470,
601
+ "Breathing": 41,
602
+ "Burping, eructation": 58,
603
+ "Burst, pop": 434,
604
+ "Bus": 321,
605
+ "Busy signal": 394,
606
+ "Buzz": 130,
607
+ "Buzzer": 398,
608
+ "Cacophony": 519,
609
+ "Camera": 416,
610
+ "Canidae, dogs, wolves": 122,
611
+ "Cap gun": 431,
612
+ "Car": 307,
613
+ "Car alarm": 310,
614
+ "Car passing by": 314,
615
+ "Carnatic music": 261,
616
+ "Cash register": 414,
617
+ "Cat": 81,
618
+ "Caterwaul": 85,
619
+ "Cattle, bovinae": 90,
620
+ "Caw": 118,
621
+ "Cello": 193,
622
+ "Chainsaw": 347,
623
+ "Change ringing (campanology)": 207,
624
+ "Chant": 30,
625
+ "Chatter": 68,
626
+ "Cheering": 66,
627
+ "Chewing, mastication": 54,
628
+ "Chicken, rooster": 99,
629
+ "Child singing": 34,
630
+ "Child speech, kid speaking": 3,
631
+ "Children playing": 71,
632
+ "Children shouting": 13,
633
+ "Chime": 205,
634
+ "Chink, clink": 442,
635
+ "Chirp tone": 503,
636
+ "Chirp, tweet": 113,
637
+ "Choir": 28,
638
+ "Chop": 438,
639
+ "Chopping (food)": 366,
640
+ "Chorus effect": 464,
641
+ "Christian music": 258,
642
+ "Christmas music": 273,
643
+ "Chuckle, chortle": 21,
644
+ "Church bell": 201,
645
+ "Civil defense siren": 397,
646
+ "Clang": 484,
647
+ "Clapping": 63,
648
+ "Clarinet": 198,
649
+ "Classical music": 237,
650
+ "Clatter": 489,
651
+ "Clickety-clack": 492,
652
+ "Clicking": 491,
653
+ "Clip-clop": 88,
654
+ "Clock": 406,
655
+ "Cluck": 100,
656
+ "Coin (dropping)": 380,
657
+ "Computer keyboard": 386,
658
+ "Conversation": 4,
659
+ "Coo": 116,
660
+ "Cough": 47,
661
+ "Country": 229,
662
+ "Cowbell": 92,
663
+ "Crack": 440,
664
+ "Crackle": 299,
665
+ "Creak": 486,
666
+ "Cricket": 127,
667
+ "Croak": 133,
668
+ "Crow": 117,
669
+ "Crowd": 69,
670
+ "Crowing, cock-a-doodle-doo": 101,
671
+ "Crumpling, crinkling": 479,
672
+ "Crunch": 499,
673
+ "Crushing": 478,
674
+ "Crying, sobbing": 22,
675
+ "Cupboard open or close": 362,
676
+ "Cutlery, silverware": 365,
677
+ "Cymbal": 171,
678
+ "Dance music": 274,
679
+ "Dental drill, dentist's drill": 345,
680
+ "Dial tone": 393,
681
+ "Didgeridoo": 211,
682
+ "Ding": 483,
683
+ "Ding-dong": 356,
684
+ "Disco": 236,
685
+ "Dishes, pots, and pans": 364,
686
+ "Distortion": 517,
687
+ "Dog": 74,
688
+ "Domestic animals, pets": 73,
689
+ "Door": 354,
690
+ "Doorbell": 355,
691
+ "Double bass": 194,
692
+ "Drawer open or close": 363,
693
+ "Drill": 425,
694
+ "Drip": 448,
695
+ "Drum": 164,
696
+ "Drum and bass": 243,
697
+ "Drum kit": 162,
698
+ "Drum machine": 163,
699
+ "Drum roll": 167,
700
+ "Dubstep": 242,
701
+ "Duck": 104,
702
+ "Echo": 512,
703
+ "Effects unit": 463,
704
+ "Electric guitar": 141,
705
+ "Electric piano": 154,
706
+ "Electric shaver, electric razor": 382,
707
+ "Electric toothbrush": 376,
708
+ "Electronic dance music": 245,
709
+ "Electronic music": 239,
710
+ "Electronic organ": 156,
711
+ "Electronic tuner": 462,
712
+ "Electronica": 244,
713
+ "Emergency vehicle": 322,
714
+ "Engine": 343,
715
+ "Engine knocking": 350,
716
+ "Engine starting": 351,
717
+ "Environmental noise": 514,
718
+ "Eruption": 435,
719
+ "Exciting music": 280,
720
+ "Explosion": 426,
721
+ "Fart": 60,
722
+ "Female singing": 33,
723
+ "Female speech, woman speaking": 2,
724
+ "Field recording": 526,
725
+ "Filing (rasp)": 422,
726
+ "Fill (with liquid)": 452,
727
+ "Finger snapping": 62,
728
+ "Fire": 298,
729
+ "Fire alarm": 400,
730
+ "Fire engine, fire truck (siren)": 325,
731
+ "Firecracker": 433,
732
+ "Fireworks": 432,
733
+ "Fixed-wing aircraft, airplane": 340,
734
+ "Flamenco": 250,
735
+ "Flap": 473,
736
+ "Flute": 196,
737
+ "Fly, housefly": 129,
738
+ "Foghorn": 401,
739
+ "Folk music": 233,
740
+ "Fowl": 98,
741
+ "French horn": 186,
742
+ "Frog": 132,
743
+ "Frying (food)": 367,
744
+ "Funk": 232,
745
+ "Funny music": 277,
746
+ "Fusillade": 429,
747
+ "Gargling": 56,
748
+ "Gasp": 44,
749
+ "Gears": 409,
750
+ "Giggle": 18,
751
+ "Glass": 441,
752
+ "Glockenspiel": 181,
753
+ "Goat": 95,
754
+ "Gobble": 103,
755
+ "Gong": 177,
756
+ "Goose": 106,
757
+ "Gospel music": 259,
758
+ "Groan": 38,
759
+ "Growling": 79,
760
+ "Grunge": 222,
761
+ "Grunt": 39,
762
+ "Guitar": 140,
763
+ "Gunshot, gunfire": 427,
764
+ "Gurgling": 297,
765
+ "Gush": 451,
766
+ "Hair dryer": 373,
767
+ "Hammer": 419,
768
+ "Hammond organ": 157,
769
+ "Hands": 61,
770
+ "Happy music": 276,
771
+ "Harmonic": 502,
772
+ "Harmonica": 208,
773
+ "Harp": 199,
774
+ "Harpsichord": 160,
775
+ "Heart murmur": 65,
776
+ "Heart sounds, heartbeat": 64,
777
+ "Heavy engine (low frequency)": 349,
778
+ "Heavy metal": 220,
779
+ "Helicopter": 339,
780
+ "Hi-hat": 172,
781
+ "Hiccup": 59,
782
+ "Hip hop music": 217,
783
+ "Hiss": 84,
784
+ "Honk": 107,
785
+ "Hoot": 120,
786
+ "Horse": 87,
787
+ "House music": 240,
788
+ "Howl": 77,
789
+ "Hubbub, speech noise, speech babble": 70,
790
+ "Hum": 496,
791
+ "Humming": 37,
792
+ "Ice cream truck, ice cream van": 320,
793
+ "Idling": 352,
794
+ "Independent music": 265,
795
+ "Insect": 126,
796
+ "Inside, large room or hall": 507,
797
+ "Inside, public space": 508,
798
+ "Inside, small room": 506,
799
+ "Jackhammer": 420,
800
+ "Jazz": 235,
801
+ "Jet engine": 337,
802
+ "Jingle (music)": 269,
803
+ "Jingle bell": 202,
804
+ "Jingle, tinkle": 495,
805
+ "Keyboard (musical)": 152,
806
+ "Keys jangling": 379,
807
+ "Knock": 359,
808
+ "Laughter": 16,
809
+ "Lawn mower": 346,
810
+ "Light engine (high frequency)": 344,
811
+ "Liquid": 444,
812
+ "Livestock, farm animals, working animals": 86,
813
+ "Lullaby": 271,
814
+ "Machine gun": 428,
815
+ "Mains hum": 516,
816
+ "Male singing": 32,
817
+ "Male speech, man speaking": 1,
818
+ "Mallet percussion": 179,
819
+ "Mandolin": 149,
820
+ "Mantra": 31,
821
+ "Maraca": 176,
822
+ "Marimba, xylophone": 180,
823
+ "Mechanical fan": 412,
824
+ "Mechanisms": 404,
825
+ "Medium engine (mid frequency)": 348,
826
+ "Meow": 83,
827
+ "Microwave oven": 368,
828
+ "Middle Eastern music": 234,
829
+ "Moo": 91,
830
+ "Mosquito": 128,
831
+ "Motor vehicle (road)": 306,
832
+ "Motorboat, speedboat": 304,
833
+ "Motorcycle": 326,
834
+ "Mouse": 124,
835
+ "Music": 137,
836
+ "Music for children": 252,
837
+ "Music of Africa": 256,
838
+ "Music of Asia": 260,
839
+ "Music of Bollywood": 262,
840
+ "Music of Latin America": 248,
841
+ "Musical instrument": 138,
842
+ "Narration, monologue": 5,
843
+ "Neigh, whinny": 89,
844
+ "New-age music": 253,
845
+ "Noise": 513,
846
+ "Ocean": 294,
847
+ "Oink": 94,
848
+ "Opera": 238,
849
+ "Orchestra": 184,
850
+ "Organ": 155,
851
+ "Outside, rural or natural": 510,
852
+ "Outside, urban or manmade": 509,
853
+ "Owl": 119,
854
+ "Pant": 45,
855
+ "Patter": 125,
856
+ "Percussion": 161,
857
+ "Piano": 153,
858
+ "Pig": 93,
859
+ "Pigeon, dove": 115,
860
+ "Ping": 482,
861
+ "Pink noise": 521,
862
+ "Pizzicato": 192,
863
+ "Plop": 494,
864
+ "Plucked string instrument": 139,
865
+ "Police car (siren)": 323,
866
+ "Pop music": 216,
867
+ "Pour": 449,
868
+ "Power tool": 424,
869
+ "Power windows, electric windows": 311,
870
+ "Printer": 415,
871
+ "Progressive rock": 223,
872
+ "Propeller, airscrew": 338,
873
+ "Psychedelic rock": 225,
874
+ "Pulleys": 410,
875
+ "Pulse": 505,
876
+ "Pump (liquid)": 454,
877
+ "Punk rock": 221,
878
+ "Purr": 82,
879
+ "Quack": 105,
880
+ "Race car, auto racing": 315,
881
+ "Radio": 525,
882
+ "Rail transport": 328,
883
+ "Railroad car, train wagon": 332,
884
+ "Rain": 289,
885
+ "Rain on surface": 291,
886
+ "Raindrop": 290,
887
+ "Rapping": 36,
888
+ "Ratchet, pawl": 405,
889
+ "Rattle": 135,
890
+ "Rattle (instrument)": 175,
891
+ "Reggae": 228,
892
+ "Reverberation": 511,
893
+ "Reversing beeps": 319,
894
+ "Rhythm and blues": 226,
895
+ "Rimshot": 166,
896
+ "Ringtone": 391,
897
+ "Roar": 110,
898
+ "Roaring cats (lions, tigers)": 109,
899
+ "Rock and roll": 224,
900
+ "Rock music": 219,
901
+ "Rodents, rats, mice": 123,
902
+ "Roll": 477,
903
+ "Rowboat, canoe, kayak": 303,
904
+ "Rub": 476,
905
+ "Rumble": 493,
906
+ "Run": 51,
907
+ "Rustle": 487,
908
+ "Rustling leaves": 284,
909
+ "Sad music": 278,
910
+ "Sailboat, sailing ship": 302,
911
+ "Salsa music": 249,
912
+ "Sampler": 159,
913
+ "Sanding": 423,
914
+ "Sawing": 421,
915
+ "Saxophone": 197,
916
+ "Scary music": 282,
917
+ "Scissors": 381,
918
+ "Scrape": 475,
919
+ "Scratch": 474,
920
+ "Scratching (performance technique)": 215,
921
+ "Screaming": 14,
922
+ "Sewing machine": 411,
923
+ "Shatter": 443,
924
+ "Sheep": 97,
925
+ "Ship": 305,
926
+ "Shofar": 212,
927
+ "Shout": 8,
928
+ "Shuffle": 52,
929
+ "Shuffling cards": 383,
930
+ "Sidetone": 518,
931
+ "Sigh": 26,
932
+ "Silence": 500,
933
+ "Sine wave": 501,
934
+ "Singing": 27,
935
+ "Singing bowl": 214,
936
+ "Single-lens reflex camera": 417,
937
+ "Sink (filling or washing)": 371,
938
+ "Siren": 396,
939
+ "Sitar": 148,
940
+ "Sizzle": 490,
941
+ "Ska": 263,
942
+ "Skateboard": 342,
943
+ "Skidding": 312,
944
+ "Slam": 358,
945
+ "Slap, smack": 467,
946
+ "Sliding door": 357,
947
+ "Slosh": 446,
948
+ "Smash, crash": 469,
949
+ "Smoke detector, smoke alarm": 399,
950
+ "Snake": 134,
951
+ "Snare drum": 165,
952
+ "Sneeze": 49,
953
+ "Snicker": 19,
954
+ "Sniff": 50,
955
+ "Snoring": 43,
956
+ "Snort": 46,
957
+ "Sonar": 457,
958
+ "Song": 266,
959
+ "Soul music": 227,
960
+ "Sound effect": 504,
961
+ "Soundtrack music": 270,
962
+ "Speech": 0,
963
+ "Speech synthesizer": 7,
964
+ "Splash, splatter": 445,
965
+ "Splinter": 439,
966
+ "Spray": 453,
967
+ "Squawk": 114,
968
+ "Squeak": 361,
969
+ "Squeal": 485,
970
+ "Squish": 447,
971
+ "Static": 515,
972
+ "Steam": 296,
973
+ "Steam whistle": 403,
974
+ "Steel guitar, slide guitar": 144,
975
+ "Steelpan": 183,
976
+ "Stir": 455,
977
+ "Stomach rumble": 57,
978
+ "Stream": 292,
979
+ "String section": 190,
980
+ "Strum": 146,
981
+ "Subway, metro, underground": 334,
982
+ "Swing music": 230,
983
+ "Synthesizer": 158,
984
+ "Synthetic singing": 35,
985
+ "Tabla": 170,
986
+ "Tambourine": 174,
987
+ "Tap": 360,
988
+ "Tapping (guitar technique)": 145,
989
+ "Tearing": 480,
990
+ "Techno": 241,
991
+ "Telephone": 389,
992
+ "Telephone bell ringing": 390,
993
+ "Telephone dialing, DTMF": 392,
994
+ "Television": 524,
995
+ "Tender music": 279,
996
+ "Theme music": 268,
997
+ "Theremin": 213,
998
+ "Throat clearing": 48,
999
+ "Throbbing": 522,
1000
+ "Thump, thud": 460,
1001
+ "Thunder": 287,
1002
+ "Thunderstorm": 286,
1003
+ "Thunk": 461,
1004
+ "Tick": 407,
1005
+ "Tick-tock": 408,
1006
+ "Timpani": 169,
1007
+ "Tire squeal": 313,
1008
+ "Toilet flush": 374,
1009
+ "Tools": 418,
1010
+ "Toot": 309,
1011
+ "Toothbrush": 375,
1012
+ "Traditional music": 264,
1013
+ "Traffic noise, roadway noise": 327,
1014
+ "Train": 329,
1015
+ "Train horn": 331,
1016
+ "Train wheels squealing": 333,
1017
+ "Train whistle": 330,
1018
+ "Trance music": 247,
1019
+ "Trickle, dribble": 450,
1020
+ "Trombone": 188,
1021
+ "Truck": 316,
1022
+ "Trumpet": 187,
1023
+ "Tubular bells": 178,
1024
+ "Tuning fork": 204,
1025
+ "Turkey": 102,
1026
+ "Typewriter": 385,
1027
+ "Typing": 384,
1028
+ "Ukulele": 151,
1029
+ "Vacuum cleaner": 377,
1030
+ "Vehicle": 300,
1031
+ "Vehicle horn, car horn, honking": 308,
1032
+ "Vibraphone": 182,
1033
+ "Vibration": 523,
1034
+ "Video game music": 272,
1035
+ "Violin, fiddle": 191,
1036
+ "Vocal music": 254,
1037
+ "Wail, moan": 25,
1038
+ "Walk, footsteps": 53,
1039
+ "Water": 288,
1040
+ "Water tap, faucet": 370,
1041
+ "Waterfall": 293,
1042
+ "Waves, surf": 295,
1043
+ "Wedding music": 275,
1044
+ "Whack, thwack": 468,
1045
+ "Whale vocalization": 136,
1046
+ "Wheeze": 42,
1047
+ "Whimper": 24,
1048
+ "Whimper (dog)": 80,
1049
+ "Whip": 472,
1050
+ "Whir": 488,
1051
+ "Whispering": 15,
1052
+ "Whistle": 402,
1053
+ "Whistling": 40,
1054
+ "White noise": 520,
1055
+ "Whoop": 10,
1056
+ "Whoosh, swoosh, swish": 459,
1057
+ "Wild animals": 108,
1058
+ "Wind": 283,
1059
+ "Wind chime": 206,
1060
+ "Wind instrument, woodwind instrument": 195,
1061
+ "Wind noise (microphone)": 285,
1062
+ "Wood": 437,
1063
+ "Wood block": 173,
1064
+ "Writing": 387,
1065
+ "Yell": 11,
1066
+ "Yip": 76,
1067
+ "Yodeling": 29,
1068
+ "Zing": 497,
1069
+ "Zipper (clothing)": 378,
1070
+ "Zither": 150
1071
+ },
1072
+ "layer_norm_eps": 1e-12,
1073
+ "max_length": 1024,
1074
+ "model_type": "audio-spectrogram-transformer",
1075
+ "num_attention_heads": 12,
1076
+ "num_hidden_layers": 12,
1077
+ "num_mel_bins": 128,
1078
+ "patch_size": 16,
1079
+ "qkv_bias": true,
1080
+ "time_stride": 10,
1081
+ "torch_dtype": "float32",
1082
+ "transformers_version": "4.49.0"
1083
+ }
onnx/model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:249b57721b7752442ec10dc3ace6f7dcd3ff0460d2c98a8709925ce5b72b69e0
3
+ size 346659433
onnx/model_bnb4.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f8d365922e23f890c9ceea323c05a6251f123524bbc66f260a0c12a2c3808721
3
+ size 54707755
onnx/model_fp16.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:910808cad2642170d7606326e0615c02a86fdce047767b8cd7106f11c5e9c6b9
3
+ size 173471361
onnx/model_int8.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6739e480a4faf4651f60d2238b0eb8cd58bead2a07ae1da5e2fbc72e3f7c98a6
3
+ size 90592065
onnx/model_q4.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f50b4abd54d454ef092236e485f09ed85e16f49a2fe8ba81189fb5ff8a25510
3
+ size 60015643
onnx/model_q4f16.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:484d2feb90e636a1c449eb16b2fc13d54f703cdd01700fad6ac63c82df33e8d4
3
+ size 51388564
onnx/model_quantized.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:65116d63d3a34eae4dbb9c403e0f2dd48ef5c576aa38eb4e298e15c5c3d5f741
3
+ size 90592065
onnx/model_uint8.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:65116d63d3a34eae4dbb9c403e0f2dd48ef5c576aa38eb4e298e15c5c3d5f741
3
+ size 90592065
preprocessor_config.json ADDED
@@ -0,0 +1,13 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_normalize": true,
3
+ "feature_extractor_type": "ASTFeatureExtractor",
4
+ "feature_size": 1,
5
+ "max_length": 1024,
6
+ "mean": -4.2677393,
7
+ "num_mel_bins": 128,
8
+ "padding_side": "right",
9
+ "padding_value": 0.0,
10
+ "return_attention_mask": false,
11
+ "sampling_rate": 16000,
12
+ "std": 4.5689974
13
+ }
quantize_config.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "modes": [
3
+ "fp16",
4
+ "q8",
5
+ "int8",
6
+ "uint8",
7
+ "q4",
8
+ "q4f16",
9
+ "bnb4"
10
+ ],
11
+ "per_channel": true,
12
+ "reduce_range": true,
13
+ "block_size": null,
14
+ "is_symmetric": true,
15
+ "accuracy_level": null,
16
+ "quant_type": 1,
17
+ "op_block_list": null
18
+ }