jimbozhang commited on
Commit
9682d44
1 Parent(s): 2b1dbec

Upload CedForAudioClassification

Browse files
Files changed (2) hide show
  1. config.json +1086 -0
  2. model.safetensors +3 -0
config.json ADDED
@@ -0,0 +1,1086 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "CedForAudioClassification"
4
+ ],
5
+ "attn_drop_rate": 0.0,
6
+ "center": true,
7
+ "depth": 12,
8
+ "drop_path_rate": 0.0,
9
+ "drop_rate": 0.0,
10
+ "embed_dim": 192,
11
+ "eval_avg": "mean",
12
+ "f_max": 8000,
13
+ "f_min": 0,
14
+ "hop_size": 160,
15
+ "id2label": {
16
+ "0": "Speech",
17
+ "1": "Male speech",
18
+ "2": "Female speech",
19
+ "3": "Child speech",
20
+ "4": "Conversation",
21
+ "5": "Narration",
22
+ "6": "Babbling",
23
+ "7": "Speech synthesizer",
24
+ "8": "Shout",
25
+ "9": "Bellow",
26
+ "10": "Whoop",
27
+ "11": "Yell",
28
+ "12": "Battle cry",
29
+ "13": "Children shouting",
30
+ "14": "Screaming",
31
+ "15": "Whispering",
32
+ "16": "Laughter",
33
+ "17": "Baby laughter",
34
+ "18": "Giggle",
35
+ "19": "Snicker",
36
+ "20": "Belly laugh",
37
+ "21": "Chuckle",
38
+ "22": "Crying",
39
+ "23": "Baby cry",
40
+ "24": "Whimper",
41
+ "25": "Wail",
42
+ "26": "Sigh",
43
+ "27": "Singing",
44
+ "28": "Choir",
45
+ "29": "Yodeling",
46
+ "30": "Chant",
47
+ "31": "Mantra",
48
+ "32": "Male singing",
49
+ "33": "Female singing",
50
+ "34": "Child singing",
51
+ "35": "Synthetic singing",
52
+ "36": "Rapping",
53
+ "37": "Humming",
54
+ "38": "Groan",
55
+ "39": "Grunt",
56
+ "40": "Whistling",
57
+ "41": "Breathing",
58
+ "42": "Wheeze",
59
+ "43": "Snoring",
60
+ "44": "Gasp",
61
+ "45": "Pant",
62
+ "46": "Snort",
63
+ "47": "Cough",
64
+ "48": "Throat clearing",
65
+ "49": "Sneeze",
66
+ "50": "Sniff",
67
+ "51": "Run",
68
+ "52": "Shuffle",
69
+ "53": "Walk",
70
+ "54": "Chewing",
71
+ "55": "Biting",
72
+ "56": "Gargling",
73
+ "57": "Stomach rumble",
74
+ "58": "Burping",
75
+ "59": "Hiccup",
76
+ "60": "Fart",
77
+ "61": "Hands",
78
+ "62": "Finger snapping",
79
+ "63": "Clapping",
80
+ "64": "Heart sounds",
81
+ "65": "Heart murmur",
82
+ "66": "Cheering",
83
+ "67": "Applause",
84
+ "68": "Chatter",
85
+ "69": "Crowd",
86
+ "70": "Hubbub",
87
+ "71": "Children playing",
88
+ "72": "Animal",
89
+ "73": "Domestic animals",
90
+ "74": "Dog",
91
+ "75": "Bark",
92
+ "76": "Yip",
93
+ "77": "Howl",
94
+ "78": "Bow-wow",
95
+ "79": "Growling",
96
+ "80": "Whimper (dog)",
97
+ "81": "Cat",
98
+ "82": "Purr",
99
+ "83": "Meow",
100
+ "84": "Hiss",
101
+ "85": "Caterwaul",
102
+ "86": "Livestock",
103
+ "87": "Horse",
104
+ "88": "Clip-clop",
105
+ "89": "Neigh",
106
+ "90": "Cattle",
107
+ "91": "Moo",
108
+ "92": "Cowbell",
109
+ "93": "Pig",
110
+ "94": "Oink",
111
+ "95": "Goat",
112
+ "96": "Bleat",
113
+ "97": "Sheep",
114
+ "98": "Fowl",
115
+ "99": "Chicken",
116
+ "100": "Cluck",
117
+ "101": "Crowing",
118
+ "102": "Turkey",
119
+ "103": "Gobble",
120
+ "104": "Duck",
121
+ "105": "Quack",
122
+ "106": "Goose",
123
+ "107": "Honk",
124
+ "108": "Wild animals",
125
+ "109": "Roaring cats (lions",
126
+ "110": "Roar",
127
+ "111": "Bird",
128
+ "112": "Bird vocalization",
129
+ "113": "Chirp",
130
+ "114": "Squawk",
131
+ "115": "Pigeon",
132
+ "116": "Coo",
133
+ "117": "Crow",
134
+ "118": "Caw",
135
+ "119": "Owl",
136
+ "120": "Hoot",
137
+ "121": "Bird flight",
138
+ "122": "Canidae",
139
+ "123": "Rodents",
140
+ "124": "Mouse",
141
+ "125": "Patter",
142
+ "126": "Insect",
143
+ "127": "Cricket",
144
+ "128": "Mosquito",
145
+ "129": "Fly",
146
+ "130": "Buzz",
147
+ "131": "Bee",
148
+ "132": "Frog",
149
+ "133": "Croak",
150
+ "134": "Snake",
151
+ "135": "Rattle",
152
+ "136": "Whale vocalization",
153
+ "137": "Music",
154
+ "138": "Musical instrument",
155
+ "139": "Plucked string instrument",
156
+ "140": "Guitar",
157
+ "141": "Electric guitar",
158
+ "142": "Bass guitar",
159
+ "143": "Acoustic guitar",
160
+ "144": "Steel guitar",
161
+ "145": "Tapping (guitar technique)",
162
+ "146": "Strum",
163
+ "147": "Banjo",
164
+ "148": "Sitar",
165
+ "149": "Mandolin",
166
+ "150": "Zither",
167
+ "151": "Ukulele",
168
+ "152": "Keyboard (musical)",
169
+ "153": "Piano",
170
+ "154": "Electric piano",
171
+ "155": "Organ",
172
+ "156": "Electronic organ",
173
+ "157": "Hammond organ",
174
+ "158": "Synthesizer",
175
+ "159": "Sampler",
176
+ "160": "Harpsichord",
177
+ "161": "Percussion",
178
+ "162": "Drum kit",
179
+ "163": "Drum machine",
180
+ "164": "Drum",
181
+ "165": "Snare drum",
182
+ "166": "Rimshot",
183
+ "167": "Drum roll",
184
+ "168": "Bass drum",
185
+ "169": "Timpani",
186
+ "170": "Tabla",
187
+ "171": "Cymbal",
188
+ "172": "Hi-hat",
189
+ "173": "Wood block",
190
+ "174": "Tambourine",
191
+ "175": "Rattle (instrument)",
192
+ "176": "Maraca",
193
+ "177": "Gong",
194
+ "178": "Tubular bells",
195
+ "179": "Mallet percussion",
196
+ "180": "Marimba",
197
+ "181": "Glockenspiel",
198
+ "182": "Vibraphone",
199
+ "183": "Steelpan",
200
+ "184": "Orchestra",
201
+ "185": "Brass instrument",
202
+ "186": "French horn",
203
+ "187": "Trumpet",
204
+ "188": "Trombone",
205
+ "189": "Bowed string instrument",
206
+ "190": "String section",
207
+ "191": "Violin",
208
+ "192": "Pizzicato",
209
+ "193": "Cello",
210
+ "194": "Double bass",
211
+ "195": "Wind instrument",
212
+ "196": "Flute",
213
+ "197": "Saxophone",
214
+ "198": "Clarinet",
215
+ "199": "Harp",
216
+ "200": "Bell",
217
+ "201": "Church bell",
218
+ "202": "Jingle bell",
219
+ "203": "Bicycle bell",
220
+ "204": "Tuning fork",
221
+ "205": "Chime",
222
+ "206": "Wind chime",
223
+ "207": "Change ringing (campanology)",
224
+ "208": "Harmonica",
225
+ "209": "Accordion",
226
+ "210": "Bagpipes",
227
+ "211": "Didgeridoo",
228
+ "212": "Shofar",
229
+ "213": "Theremin",
230
+ "214": "Singing bowl",
231
+ "215": "Scratching (performance technique)",
232
+ "216": "Pop music",
233
+ "217": "Hip hop music",
234
+ "218": "Beatboxing",
235
+ "219": "Rock music",
236
+ "220": "Heavy metal",
237
+ "221": "Punk rock",
238
+ "222": "Grunge",
239
+ "223": "Progressive rock",
240
+ "224": "Rock and roll",
241
+ "225": "Psychedelic rock",
242
+ "226": "Rhythm and blues",
243
+ "227": "Soul music",
244
+ "228": "Reggae",
245
+ "229": "Country",
246
+ "230": "Swing music",
247
+ "231": "Bluegrass",
248
+ "232": "Funk",
249
+ "233": "Folk music",
250
+ "234": "Middle Eastern music",
251
+ "235": "Jazz",
252
+ "236": "Disco",
253
+ "237": "Classical music",
254
+ "238": "Opera",
255
+ "239": "Electronic music",
256
+ "240": "House music",
257
+ "241": "Techno",
258
+ "242": "Dubstep",
259
+ "243": "Drum and bass",
260
+ "244": "Electronica",
261
+ "245": "Electronic dance music",
262
+ "246": "Ambient music",
263
+ "247": "Trance music",
264
+ "248": "Music of Latin America",
265
+ "249": "Salsa music",
266
+ "250": "Flamenco",
267
+ "251": "Blues",
268
+ "252": "Music for children",
269
+ "253": "New-age music",
270
+ "254": "Vocal music",
271
+ "255": "A capella",
272
+ "256": "Music of Africa",
273
+ "257": "Afrobeat",
274
+ "258": "Christian music",
275
+ "259": "Gospel music",
276
+ "260": "Music of Asia",
277
+ "261": "Carnatic music",
278
+ "262": "Music of Bollywood",
279
+ "263": "Ska",
280
+ "264": "Traditional music",
281
+ "265": "Independent music",
282
+ "266": "Song",
283
+ "267": "Background music",
284
+ "268": "Theme music",
285
+ "269": "Jingle (music)",
286
+ "270": "Soundtrack music",
287
+ "271": "Lullaby",
288
+ "272": "Video game music",
289
+ "273": "Christmas music",
290
+ "274": "Dance music",
291
+ "275": "Wedding music",
292
+ "276": "Happy music",
293
+ "277": "Funny music",
294
+ "278": "Sad music",
295
+ "279": "Tender music",
296
+ "280": "Exciting music",
297
+ "281": "Angry music",
298
+ "282": "Scary music",
299
+ "283": "Wind",
300
+ "284": "Rustling leaves",
301
+ "285": "Wind noise (microphone)",
302
+ "286": "Thunderstorm",
303
+ "287": "Thunder",
304
+ "288": "Water",
305
+ "289": "Rain",
306
+ "290": "Raindrop",
307
+ "291": "Rain on surface",
308
+ "292": "Stream",
309
+ "293": "Waterfall",
310
+ "294": "Ocean",
311
+ "295": "Waves",
312
+ "296": "Steam",
313
+ "297": "Gurgling",
314
+ "298": "Fire",
315
+ "299": "Crackle",
316
+ "300": "Vehicle",
317
+ "301": "Boat",
318
+ "302": "Sailboat",
319
+ "303": "Rowboat",
320
+ "304": "Motorboat",
321
+ "305": "Ship",
322
+ "306": "Motor vehicle (road)",
323
+ "307": "Car",
324
+ "308": "Vehicle horn",
325
+ "309": "Toot",
326
+ "310": "Car alarm",
327
+ "311": "Power windows",
328
+ "312": "Skidding",
329
+ "313": "Tire squeal",
330
+ "314": "Car passing by",
331
+ "315": "Race car",
332
+ "316": "Truck",
333
+ "317": "Air brake",
334
+ "318": "Air horn",
335
+ "319": "Reversing beeps",
336
+ "320": "Ice cream truck",
337
+ "321": "Bus",
338
+ "322": "Emergency vehicle",
339
+ "323": "Police car (siren)",
340
+ "324": "Ambulance (siren)",
341
+ "325": "Fire engine",
342
+ "326": "Motorcycle",
343
+ "327": "Traffic noise",
344
+ "328": "Rail transport",
345
+ "329": "Train",
346
+ "330": "Train whistle",
347
+ "331": "Train horn",
348
+ "332": "Railroad car",
349
+ "333": "Train wheels squealing",
350
+ "334": "Subway",
351
+ "335": "Aircraft",
352
+ "336": "Aircraft engine",
353
+ "337": "Jet engine",
354
+ "338": "Propeller",
355
+ "339": "Helicopter",
356
+ "340": "Fixed-wing aircraft",
357
+ "341": "Bicycle",
358
+ "342": "Skateboard",
359
+ "343": "Engine",
360
+ "344": "Light engine (high frequency)",
361
+ "345": "Dental drill",
362
+ "346": "Lawn mower",
363
+ "347": "Chainsaw",
364
+ "348": "Medium engine (mid frequency)",
365
+ "349": "Heavy engine (low frequency)",
366
+ "350": "Engine knocking",
367
+ "351": "Engine starting",
368
+ "352": "Idling",
369
+ "353": "Accelerating",
370
+ "354": "Door",
371
+ "355": "Doorbell",
372
+ "356": "Ding-dong",
373
+ "357": "Sliding door",
374
+ "358": "Slam",
375
+ "359": "Knock",
376
+ "360": "Tap",
377
+ "361": "Squeak",
378
+ "362": "Cupboard open or close",
379
+ "363": "Drawer open or close",
380
+ "364": "Dishes",
381
+ "365": "Cutlery",
382
+ "366": "Chopping (food)",
383
+ "367": "Frying (food)",
384
+ "368": "Microwave oven",
385
+ "369": "Blender",
386
+ "370": "Water tap",
387
+ "371": "Sink (filling or washing)",
388
+ "372": "Bathtub (filling or washing)",
389
+ "373": "Hair dryer",
390
+ "374": "Toilet flush",
391
+ "375": "Toothbrush",
392
+ "376": "Electric toothbrush",
393
+ "377": "Vacuum cleaner",
394
+ "378": "Zipper (clothing)",
395
+ "379": "Keys jangling",
396
+ "380": "Coin (dropping)",
397
+ "381": "Scissors",
398
+ "382": "Electric shaver",
399
+ "383": "Shuffling cards",
400
+ "384": "Typing",
401
+ "385": "Typewriter",
402
+ "386": "Computer keyboard",
403
+ "387": "Writing",
404
+ "388": "Alarm",
405
+ "389": "Telephone",
406
+ "390": "Telephone bell ringing",
407
+ "391": "Ringtone",
408
+ "392": "Telephone dialing",
409
+ "393": "Dial tone",
410
+ "394": "Busy signal",
411
+ "395": "Alarm clock",
412
+ "396": "Siren",
413
+ "397": "Civil defense siren",
414
+ "398": "Buzzer",
415
+ "399": "Smoke detector",
416
+ "400": "Fire alarm",
417
+ "401": "Foghorn",
418
+ "402": "Whistle",
419
+ "403": "Steam whistle",
420
+ "404": "Mechanisms",
421
+ "405": "Ratchet",
422
+ "406": "Clock",
423
+ "407": "Tick",
424
+ "408": "Tick-tock",
425
+ "409": "Gears",
426
+ "410": "Pulleys",
427
+ "411": "Sewing machine",
428
+ "412": "Mechanical fan",
429
+ "413": "Air conditioning",
430
+ "414": "Cash register",
431
+ "415": "Printer",
432
+ "416": "Camera",
433
+ "417": "Single-lens reflex camera",
434
+ "418": "Tools",
435
+ "419": "Hammer",
436
+ "420": "Jackhammer",
437
+ "421": "Sawing",
438
+ "422": "Filing (rasp)",
439
+ "423": "Sanding",
440
+ "424": "Power tool",
441
+ "425": "Drill",
442
+ "426": "Explosion",
443
+ "427": "Gunshot",
444
+ "428": "Machine gun",
445
+ "429": "Fusillade",
446
+ "430": "Artillery fire",
447
+ "431": "Cap gun",
448
+ "432": "Fireworks",
449
+ "433": "Firecracker",
450
+ "434": "Burst",
451
+ "435": "Eruption",
452
+ "436": "Boom",
453
+ "437": "Wood",
454
+ "438": "Chop",
455
+ "439": "Splinter",
456
+ "440": "Crack",
457
+ "441": "Glass",
458
+ "442": "Chink",
459
+ "443": "Shatter",
460
+ "444": "Liquid",
461
+ "445": "Splash",
462
+ "446": "Slosh",
463
+ "447": "Squish",
464
+ "448": "Drip",
465
+ "449": "Pour",
466
+ "450": "Trickle",
467
+ "451": "Gush",
468
+ "452": "Fill (with liquid)",
469
+ "453": "Spray",
470
+ "454": "Pump (liquid)",
471
+ "455": "Stir",
472
+ "456": "Boiling",
473
+ "457": "Sonar",
474
+ "458": "Arrow",
475
+ "459": "Whoosh",
476
+ "460": "Thump",
477
+ "461": "Thunk",
478
+ "462": "Electronic tuner",
479
+ "463": "Effects unit",
480
+ "464": "Chorus effect",
481
+ "465": "Basketball bounce",
482
+ "466": "Bang",
483
+ "467": "Slap",
484
+ "468": "Whack",
485
+ "469": "Smash",
486
+ "470": "Breaking",
487
+ "471": "Bouncing",
488
+ "472": "Whip",
489
+ "473": "Flap",
490
+ "474": "Scratch",
491
+ "475": "Scrape",
492
+ "476": "Rub",
493
+ "477": "Roll",
494
+ "478": "Crushing",
495
+ "479": "Crumpling",
496
+ "480": "Tearing",
497
+ "481": "Beep",
498
+ "482": "Ping",
499
+ "483": "Ding",
500
+ "484": "Clang",
501
+ "485": "Squeal",
502
+ "486": "Creak",
503
+ "487": "Rustle",
504
+ "488": "Whir",
505
+ "489": "Clatter",
506
+ "490": "Sizzle",
507
+ "491": "Clicking",
508
+ "492": "Clickety-clack",
509
+ "493": "Rumble",
510
+ "494": "Plop",
511
+ "495": "Jingle",
512
+ "496": "Hum",
513
+ "497": "Zing",
514
+ "498": "Boing",
515
+ "499": "Crunch",
516
+ "500": "Silence",
517
+ "501": "Sine wave",
518
+ "502": "Harmonic",
519
+ "503": "Chirp tone",
520
+ "504": "Sound effect",
521
+ "505": "Pulse",
522
+ "506": "Inside",
523
+ "507": "Inside",
524
+ "508": "Inside",
525
+ "509": "Outside",
526
+ "510": "Outside",
527
+ "511": "Reverberation",
528
+ "512": "Echo",
529
+ "513": "Noise",
530
+ "514": "Environmental noise",
531
+ "515": "Static",
532
+ "516": "Mains hum",
533
+ "517": "Distortion",
534
+ "518": "Sidetone",
535
+ "519": "Cacophony",
536
+ "520": "White noise",
537
+ "521": "Pink noise",
538
+ "522": "Throbbing",
539
+ "523": "Vibration",
540
+ "524": "Television",
541
+ "525": "Radio",
542
+ "526": "Field recording"
543
+ },
544
+ "label2id": {
545
+ "A capella": 255,
546
+ "Accelerating": 353,
547
+ "Accordion": 209,
548
+ "Acoustic guitar": 143,
549
+ "Afrobeat": 257,
550
+ "Air brake": 317,
551
+ "Air conditioning": 413,
552
+ "Air horn": 318,
553
+ "Aircraft": 335,
554
+ "Aircraft engine": 336,
555
+ "Alarm": 388,
556
+ "Alarm clock": 395,
557
+ "Ambient music": 246,
558
+ "Ambulance (siren)": 324,
559
+ "Angry music": 281,
560
+ "Animal": 72,
561
+ "Applause": 67,
562
+ "Arrow": 458,
563
+ "Artillery fire": 430,
564
+ "Babbling": 6,
565
+ "Baby cry": 23,
566
+ "Baby laughter": 17,
567
+ "Background music": 267,
568
+ "Bagpipes": 210,
569
+ "Bang": 466,
570
+ "Banjo": 147,
571
+ "Bark": 75,
572
+ "Basketball bounce": 465,
573
+ "Bass drum": 168,
574
+ "Bass guitar": 142,
575
+ "Bathtub (filling or washing)": 372,
576
+ "Battle cry": 12,
577
+ "Beatboxing": 218,
578
+ "Bee": 131,
579
+ "Beep": 481,
580
+ "Bell": 200,
581
+ "Bellow": 9,
582
+ "Belly laugh": 20,
583
+ "Bicycle": 341,
584
+ "Bicycle bell": 203,
585
+ "Bird": 111,
586
+ "Bird flight": 121,
587
+ "Bird vocalization": 112,
588
+ "Biting": 55,
589
+ "Bleat": 96,
590
+ "Blender": 369,
591
+ "Bluegrass": 231,
592
+ "Blues": 251,
593
+ "Boat": 301,
594
+ "Boiling": 456,
595
+ "Boing": 498,
596
+ "Boom": 436,
597
+ "Bouncing": 471,
598
+ "Bow-wow": 78,
599
+ "Bowed string instrument": 189,
600
+ "Brass instrument": 185,
601
+ "Breaking": 470,
602
+ "Breathing": 41,
603
+ "Burping": 58,
604
+ "Burst": 434,
605
+ "Bus": 321,
606
+ "Busy signal": 394,
607
+ "Buzz": 130,
608
+ "Buzzer": 398,
609
+ "Cacophony": 519,
610
+ "Camera": 416,
611
+ "Canidae": 122,
612
+ "Cap gun": 431,
613
+ "Car": 307,
614
+ "Car alarm": 310,
615
+ "Car passing by": 314,
616
+ "Carnatic music": 261,
617
+ "Cash register": 414,
618
+ "Cat": 81,
619
+ "Caterwaul": 85,
620
+ "Cattle": 90,
621
+ "Caw": 118,
622
+ "Cello": 193,
623
+ "Chainsaw": 347,
624
+ "Change ringing (campanology)": 207,
625
+ "Chant": 30,
626
+ "Chatter": 68,
627
+ "Cheering": 66,
628
+ "Chewing": 54,
629
+ "Chicken": 99,
630
+ "Child singing": 34,
631
+ "Child speech": 3,
632
+ "Children playing": 71,
633
+ "Children shouting": 13,
634
+ "Chime": 205,
635
+ "Chink": 442,
636
+ "Chirp": 113,
637
+ "Chirp tone": 503,
638
+ "Choir": 28,
639
+ "Chop": 438,
640
+ "Chopping (food)": 366,
641
+ "Chorus effect": 464,
642
+ "Christian music": 258,
643
+ "Christmas music": 273,
644
+ "Chuckle": 21,
645
+ "Church bell": 201,
646
+ "Civil defense siren": 397,
647
+ "Clang": 484,
648
+ "Clapping": 63,
649
+ "Clarinet": 198,
650
+ "Classical music": 237,
651
+ "Clatter": 489,
652
+ "Clickety-clack": 492,
653
+ "Clicking": 491,
654
+ "Clip-clop": 88,
655
+ "Clock": 406,
656
+ "Cluck": 100,
657
+ "Coin (dropping)": 380,
658
+ "Computer keyboard": 386,
659
+ "Conversation": 4,
660
+ "Coo": 116,
661
+ "Cough": 47,
662
+ "Country": 229,
663
+ "Cowbell": 92,
664
+ "Crack": 440,
665
+ "Crackle": 299,
666
+ "Creak": 486,
667
+ "Cricket": 127,
668
+ "Croak": 133,
669
+ "Crow": 117,
670
+ "Crowd": 69,
671
+ "Crowing": 101,
672
+ "Crumpling": 479,
673
+ "Crunch": 499,
674
+ "Crushing": 478,
675
+ "Crying": 22,
676
+ "Cupboard open or close": 362,
677
+ "Cutlery": 365,
678
+ "Cymbal": 171,
679
+ "Dance music": 274,
680
+ "Dental drill": 345,
681
+ "Dial tone": 393,
682
+ "Didgeridoo": 211,
683
+ "Ding": 483,
684
+ "Ding-dong": 356,
685
+ "Disco": 236,
686
+ "Dishes": 364,
687
+ "Distortion": 517,
688
+ "Dog": 74,
689
+ "Domestic animals": 73,
690
+ "Door": 354,
691
+ "Doorbell": 355,
692
+ "Double bass": 194,
693
+ "Drawer open or close": 363,
694
+ "Drill": 425,
695
+ "Drip": 448,
696
+ "Drum": 164,
697
+ "Drum and bass": 243,
698
+ "Drum kit": 162,
699
+ "Drum machine": 163,
700
+ "Drum roll": 167,
701
+ "Dubstep": 242,
702
+ "Duck": 104,
703
+ "Echo": 512,
704
+ "Effects unit": 463,
705
+ "Electric guitar": 141,
706
+ "Electric piano": 154,
707
+ "Electric shaver": 382,
708
+ "Electric toothbrush": 376,
709
+ "Electronic dance music": 245,
710
+ "Electronic music": 239,
711
+ "Electronic organ": 156,
712
+ "Electronic tuner": 462,
713
+ "Electronica": 244,
714
+ "Emergency vehicle": 322,
715
+ "Engine": 343,
716
+ "Engine knocking": 350,
717
+ "Engine starting": 351,
718
+ "Environmental noise": 514,
719
+ "Eruption": 435,
720
+ "Exciting music": 280,
721
+ "Explosion": 426,
722
+ "Fart": 60,
723
+ "Female singing": 33,
724
+ "Female speech": 2,
725
+ "Field recording": 526,
726
+ "Filing (rasp)": 422,
727
+ "Fill (with liquid)": 452,
728
+ "Finger snapping": 62,
729
+ "Fire": 298,
730
+ "Fire alarm": 400,
731
+ "Fire engine": 325,
732
+ "Firecracker": 433,
733
+ "Fireworks": 432,
734
+ "Fixed-wing aircraft": 340,
735
+ "Flamenco": 250,
736
+ "Flap": 473,
737
+ "Flute": 196,
738
+ "Fly": 129,
739
+ "Foghorn": 401,
740
+ "Folk music": 233,
741
+ "Fowl": 98,
742
+ "French horn": 186,
743
+ "Frog": 132,
744
+ "Frying (food)": 367,
745
+ "Funk": 232,
746
+ "Funny music": 277,
747
+ "Fusillade": 429,
748
+ "Gargling": 56,
749
+ "Gasp": 44,
750
+ "Gears": 409,
751
+ "Giggle": 18,
752
+ "Glass": 441,
753
+ "Glockenspiel": 181,
754
+ "Goat": 95,
755
+ "Gobble": 103,
756
+ "Gong": 177,
757
+ "Goose": 106,
758
+ "Gospel music": 259,
759
+ "Groan": 38,
760
+ "Growling": 79,
761
+ "Grunge": 222,
762
+ "Grunt": 39,
763
+ "Guitar": 140,
764
+ "Gunshot": 427,
765
+ "Gurgling": 297,
766
+ "Gush": 451,
767
+ "Hair dryer": 373,
768
+ "Hammer": 419,
769
+ "Hammond organ": 157,
770
+ "Hands": 61,
771
+ "Happy music": 276,
772
+ "Harmonic": 502,
773
+ "Harmonica": 208,
774
+ "Harp": 199,
775
+ "Harpsichord": 160,
776
+ "Heart murmur": 65,
777
+ "Heart sounds": 64,
778
+ "Heavy engine (low frequency)": 349,
779
+ "Heavy metal": 220,
780
+ "Helicopter": 339,
781
+ "Hi-hat": 172,
782
+ "Hiccup": 59,
783
+ "Hip hop music": 217,
784
+ "Hiss": 84,
785
+ "Honk": 107,
786
+ "Hoot": 120,
787
+ "Horse": 87,
788
+ "House music": 240,
789
+ "Howl": 77,
790
+ "Hubbub": 70,
791
+ "Hum": 496,
792
+ "Humming": 37,
793
+ "Ice cream truck": 320,
794
+ "Idling": 352,
795
+ "Independent music": 265,
796
+ "Insect": 126,
797
+ "Inside": 508,
798
+ "Jackhammer": 420,
799
+ "Jazz": 235,
800
+ "Jet engine": 337,
801
+ "Jingle": 495,
802
+ "Jingle (music)": 269,
803
+ "Jingle bell": 202,
804
+ "Keyboard (musical)": 152,
805
+ "Keys jangling": 379,
806
+ "Knock": 359,
807
+ "Laughter": 16,
808
+ "Lawn mower": 346,
809
+ "Light engine (high frequency)": 344,
810
+ "Liquid": 444,
811
+ "Livestock": 86,
812
+ "Lullaby": 271,
813
+ "Machine gun": 428,
814
+ "Mains hum": 516,
815
+ "Male singing": 32,
816
+ "Male speech": 1,
817
+ "Mallet percussion": 179,
818
+ "Mandolin": 149,
819
+ "Mantra": 31,
820
+ "Maraca": 176,
821
+ "Marimba": 180,
822
+ "Mechanical fan": 412,
823
+ "Mechanisms": 404,
824
+ "Medium engine (mid frequency)": 348,
825
+ "Meow": 83,
826
+ "Microwave oven": 368,
827
+ "Middle Eastern music": 234,
828
+ "Moo": 91,
829
+ "Mosquito": 128,
830
+ "Motor vehicle (road)": 306,
831
+ "Motorboat": 304,
832
+ "Motorcycle": 326,
833
+ "Mouse": 124,
834
+ "Music": 137,
835
+ "Music for children": 252,
836
+ "Music of Africa": 256,
837
+ "Music of Asia": 260,
838
+ "Music of Bollywood": 262,
839
+ "Music of Latin America": 248,
840
+ "Musical instrument": 138,
841
+ "Narration": 5,
842
+ "Neigh": 89,
843
+ "New-age music": 253,
844
+ "Noise": 513,
845
+ "Ocean": 294,
846
+ "Oink": 94,
847
+ "Opera": 238,
848
+ "Orchestra": 184,
849
+ "Organ": 155,
850
+ "Outside": 510,
851
+ "Owl": 119,
852
+ "Pant": 45,
853
+ "Patter": 125,
854
+ "Percussion": 161,
855
+ "Piano": 153,
856
+ "Pig": 93,
857
+ "Pigeon": 115,
858
+ "Ping": 482,
859
+ "Pink noise": 521,
860
+ "Pizzicato": 192,
861
+ "Plop": 494,
862
+ "Plucked string instrument": 139,
863
+ "Police car (siren)": 323,
864
+ "Pop music": 216,
865
+ "Pour": 449,
866
+ "Power tool": 424,
867
+ "Power windows": 311,
868
+ "Printer": 415,
869
+ "Progressive rock": 223,
870
+ "Propeller": 338,
871
+ "Psychedelic rock": 225,
872
+ "Pulleys": 410,
873
+ "Pulse": 505,
874
+ "Pump (liquid)": 454,
875
+ "Punk rock": 221,
876
+ "Purr": 82,
877
+ "Quack": 105,
878
+ "Race car": 315,
879
+ "Radio": 525,
880
+ "Rail transport": 328,
881
+ "Railroad car": 332,
882
+ "Rain": 289,
883
+ "Rain on surface": 291,
884
+ "Raindrop": 290,
885
+ "Rapping": 36,
886
+ "Ratchet": 405,
887
+ "Rattle": 135,
888
+ "Rattle (instrument)": 175,
889
+ "Reggae": 228,
890
+ "Reverberation": 511,
891
+ "Reversing beeps": 319,
892
+ "Rhythm and blues": 226,
893
+ "Rimshot": 166,
894
+ "Ringtone": 391,
895
+ "Roar": 110,
896
+ "Roaring cats (lions": 109,
897
+ "Rock and roll": 224,
898
+ "Rock music": 219,
899
+ "Rodents": 123,
900
+ "Roll": 477,
901
+ "Rowboat": 303,
902
+ "Rub": 476,
903
+ "Rumble": 493,
904
+ "Run": 51,
905
+ "Rustle": 487,
906
+ "Rustling leaves": 284,
907
+ "Sad music": 278,
908
+ "Sailboat": 302,
909
+ "Salsa music": 249,
910
+ "Sampler": 159,
911
+ "Sanding": 423,
912
+ "Sawing": 421,
913
+ "Saxophone": 197,
914
+ "Scary music": 282,
915
+ "Scissors": 381,
916
+ "Scrape": 475,
917
+ "Scratch": 474,
918
+ "Scratching (performance technique)": 215,
919
+ "Screaming": 14,
920
+ "Sewing machine": 411,
921
+ "Shatter": 443,
922
+ "Sheep": 97,
923
+ "Ship": 305,
924
+ "Shofar": 212,
925
+ "Shout": 8,
926
+ "Shuffle": 52,
927
+ "Shuffling cards": 383,
928
+ "Sidetone": 518,
929
+ "Sigh": 26,
930
+ "Silence": 500,
931
+ "Sine wave": 501,
932
+ "Singing": 27,
933
+ "Singing bowl": 214,
934
+ "Single-lens reflex camera": 417,
935
+ "Sink (filling or washing)": 371,
936
+ "Siren": 396,
937
+ "Sitar": 148,
938
+ "Sizzle": 490,
939
+ "Ska": 263,
940
+ "Skateboard": 342,
941
+ "Skidding": 312,
942
+ "Slam": 358,
943
+ "Slap": 467,
944
+ "Sliding door": 357,
945
+ "Slosh": 446,
946
+ "Smash": 469,
947
+ "Smoke detector": 399,
948
+ "Snake": 134,
949
+ "Snare drum": 165,
950
+ "Sneeze": 49,
951
+ "Snicker": 19,
952
+ "Sniff": 50,
953
+ "Snoring": 43,
954
+ "Snort": 46,
955
+ "Sonar": 457,
956
+ "Song": 266,
957
+ "Soul music": 227,
958
+ "Sound effect": 504,
959
+ "Soundtrack music": 270,
960
+ "Speech": 0,
961
+ "Speech synthesizer": 7,
962
+ "Splash": 445,
963
+ "Splinter": 439,
964
+ "Spray": 453,
965
+ "Squawk": 114,
966
+ "Squeak": 361,
967
+ "Squeal": 485,
968
+ "Squish": 447,
969
+ "Static": 515,
970
+ "Steam": 296,
971
+ "Steam whistle": 403,
972
+ "Steel guitar": 144,
973
+ "Steelpan": 183,
974
+ "Stir": 455,
975
+ "Stomach rumble": 57,
976
+ "Stream": 292,
977
+ "String section": 190,
978
+ "Strum": 146,
979
+ "Subway": 334,
980
+ "Swing music": 230,
981
+ "Synthesizer": 158,
982
+ "Synthetic singing": 35,
983
+ "Tabla": 170,
984
+ "Tambourine": 174,
985
+ "Tap": 360,
986
+ "Tapping (guitar technique)": 145,
987
+ "Tearing": 480,
988
+ "Techno": 241,
989
+ "Telephone": 389,
990
+ "Telephone bell ringing": 390,
991
+ "Telephone dialing": 392,
992
+ "Television": 524,
993
+ "Tender music": 279,
994
+ "Theme music": 268,
995
+ "Theremin": 213,
996
+ "Throat clearing": 48,
997
+ "Throbbing": 522,
998
+ "Thump": 460,
999
+ "Thunder": 287,
1000
+ "Thunderstorm": 286,
1001
+ "Thunk": 461,
1002
+ "Tick": 407,
1003
+ "Tick-tock": 408,
1004
+ "Timpani": 169,
1005
+ "Tire squeal": 313,
1006
+ "Toilet flush": 374,
1007
+ "Tools": 418,
1008
+ "Toot": 309,
1009
+ "Toothbrush": 375,
1010
+ "Traditional music": 264,
1011
+ "Traffic noise": 327,
1012
+ "Train": 329,
1013
+ "Train horn": 331,
1014
+ "Train wheels squealing": 333,
1015
+ "Train whistle": 330,
1016
+ "Trance music": 247,
1017
+ "Trickle": 450,
1018
+ "Trombone": 188,
1019
+ "Truck": 316,
1020
+ "Trumpet": 187,
1021
+ "Tubular bells": 178,
1022
+ "Tuning fork": 204,
1023
+ "Turkey": 102,
1024
+ "Typewriter": 385,
1025
+ "Typing": 384,
1026
+ "Ukulele": 151,
1027
+ "Vacuum cleaner": 377,
1028
+ "Vehicle": 300,
1029
+ "Vehicle horn": 308,
1030
+ "Vibraphone": 182,
1031
+ "Vibration": 523,
1032
+ "Video game music": 272,
1033
+ "Violin": 191,
1034
+ "Vocal music": 254,
1035
+ "Wail": 25,
1036
+ "Walk": 53,
1037
+ "Water": 288,
1038
+ "Water tap": 370,
1039
+ "Waterfall": 293,
1040
+ "Waves": 295,
1041
+ "Wedding music": 275,
1042
+ "Whack": 468,
1043
+ "Whale vocalization": 136,
1044
+ "Wheeze": 42,
1045
+ "Whimper": 24,
1046
+ "Whimper (dog)": 80,
1047
+ "Whip": 472,
1048
+ "Whir": 488,
1049
+ "Whispering": 15,
1050
+ "Whistle": 402,
1051
+ "Whistling": 40,
1052
+ "White noise": 520,
1053
+ "Whoop": 10,
1054
+ "Whoosh": 459,
1055
+ "Wild animals": 108,
1056
+ "Wind": 283,
1057
+ "Wind chime": 206,
1058
+ "Wind instrument": 195,
1059
+ "Wind noise (microphone)": 285,
1060
+ "Wood": 437,
1061
+ "Wood block": 173,
1062
+ "Writing": 387,
1063
+ "Yell": 11,
1064
+ "Yip": 76,
1065
+ "Yodeling": 29,
1066
+ "Zing": 497,
1067
+ "Zipper (clothing)": 378,
1068
+ "Zither": 150
1069
+ },
1070
+ "mlp_ratio": 4.0,
1071
+ "model_type": "ced",
1072
+ "n_fft": 512,
1073
+ "n_mels": 64,
1074
+ "name": "ced-tiny",
1075
+ "num_heads": 3,
1076
+ "outputdim": 527,
1077
+ "pad_last": true,
1078
+ "patch_size": 16,
1079
+ "patch_stride": 16,
1080
+ "pooling": "mean",
1081
+ "qkv_bias": true,
1082
+ "target_length": 1012,
1083
+ "torch_dtype": "float32",
1084
+ "transformers_version": "4.35.2",
1085
+ "win_size": 512
1086
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0e086f0cd62814c6def89001f3f25193f75955696f6975ef6800af31d00d6dd7
3
+ size 22028820