Release best Gemini-tuned Whisper Base and Small with code, normalization and evaluation

cd9b2d8 verified 1 day ago

8.05 kB

	{
	"schema_version": "burst-canonical-v1",
	"min_training_events": 200,
	"names": [
	"Other / rare burst",
	"Affirmative Grunt",
	"Ahem",
	"Breath",
	"Breathy Giggle",
	"Cackle",
	"Childlike Giggle",
	"Chuckle",
	"Contented Sigh",
	"Cough",
	"Crying",
	"Displeased Grunt",
	"Effort Grunt",
	"Exasperated Sigh",
	"Exhausted Groan",
	"Fearful Gasp",
	"Frustrated Groan",
	"Growl",
	"Guffaw",
	"Gulp",
	"Hiccup",
	"Hum",
	"Kissing Sound",
	"Laughter",
	"Lip Smack",
	"Low Mumble",
	"Mournful Wail",
	"Nervous Giggle",
	"Nervous Gulp",
	"Purr",
	"Quiet Sob",
	"Relief Sigh",
	"Scream",
	"Sharp Inhale",
	"Sharp Whistle",
	"Shriek",
	"Sigh",
	"Sneeze",
	"Snicker",
	"Sniff",
	"Snort",
	"Snorting Giggle",
	"Soft Whistle",
	"Spitting",
	"Surprised Gasp",
	"Swallows",
	"Throat Clearing",
	"Tongue Click",
	"Trembling Whimper",
	"Tsk",
	"Whispered Mumble",
	"Wistful Sigh",
	"Yawn"
	],
	"class_counts": {
	"Other / rare burst": 1077,
	"Affirmative Grunt": 88976,
	"Ahem": 3908320,
	"Breath": 12721,
	"Breathy Giggle": 386990,
	"Cackle": 3857,
	"Childlike Giggle": 287725,
	"Chuckle": 306023,
	"Contented Sigh": 1709805,
	"Cough": 10431,
	"Crying": 3958,
	"Displeased Grunt": 28165,
	"Effort Grunt": 1300,
	"Exasperated Sigh": 34585,
	"Exhausted Groan": 244785,
	"Fearful Gasp": 21182,
	"Frustrated Groan": 771,
	"Growl": 426,
	"Guffaw": 698,
	"Gulp": 51607,
	"Hiccup": 81196,
	"Hum": 90375,
	"Kissing Sound": 1508,
	"Laughter": 9361,
	"Lip Smack": 27433,
	"Low Mumble": 5272667,
	"Mournful Wail": 1715,
	"Nervous Giggle": 220,
	"Nervous Gulp": 141991,
	"Purr": 29974,
	"Quiet Sob": 17042,
	"Relief Sigh": 87705,
	"Scream": 40056,
	"Sharp Inhale": 216834,
	"Sharp Whistle": 1583,
	"Shriek": 2842,
	"Sigh": 6904,
	"Sneeze": 7353,
	"Snicker": 321,
	"Sniff": 7351,
	"Snort": 16834,
	"Snorting Giggle": 12751,
	"Soft Whistle": 1222,
	"Spitting": 280,
	"Surprised Gasp": 988611,
	"Swallows": 12358,
	"Throat Clearing": 18070,
	"Tongue Click": 1198,
	"Trembling Whimper": 916,
	"Tsk": 43760,
	"Whispered Mumble": 422,
	"Wistful Sigh": 584259,
	"Yawn": 65068
	},
	"class_weights": [
	1.9905116558074951,
	0.25,
	0.25,
	0.5791777968406677,
	0.25,
	1.051835298538208,
	0.25,
	0.25,
	0.25,
	0.6396017074584961,
	1.0383281707763672,
	0.38924017548561096,
	1.8117609024047852,
	0.3512599766254425,
	0.25,
	0.4488374590873718,
	2.3525853157043457,
	3.164956569671631,
	2.472548723220825,
	0.28755322098731995,
	0.25,
	0.25,
	1.6821776628494263,
	0.6751673221588135,
	0.39439910650253296,
	0.25,
	1.5773952007293701,
	4.0,
	0.25,
	0.377311646938324,
	0.5003942251205444,
	0.25,
	0.32639145851135254,
	0.25,
	1.6418447494506836,
	1.225351095199585,
	0.786180317401886,
	0.7617987990379333,
	3.6460282802581787,
	0.761902391910553,
	0.5034761428833008,
	0.5784961581230164,
	1.868688702583313,
	3.9038538932800293,
	0.25,
	0.5876225829124451,
	0.48595210909843445,
	1.887313961982727,
	2.1583645343780518,
	0.3122726082801819,
	3.1799211502075195,
	0.25,
	0.2560877501964569
	],
	"ladder_raw_to_canonical": [
	0,
	25,
	44,
	2,
	8,
	51,
	14,
	21,
	7,
	4,
	6,
	52,
	33,
	5,
	13,
	21,
	32,
	48,
	3,
	35,
	26,
	50,
	49,
	12,
	9,
	39,
	29,
	9,
	17,
	18,
	0,
	16,
	1,
	0,
	40,
	42,
	41,
	31,
	21,
	34,
	11,
	3,
	0,
	15,
	46,
	0,
	0,
	20,
	22,
	3,
	30,
	24,
	47,
	3,
	0,
	3,
	20,
	19,
	0,
	0,
	0,
	3,
	0,
	24,
	0
	],
	"p3_raw_to_canonical": [
	0,
	1,
	2,
	3,
	4,
	5,
	6,
	7,
	46,
	47,
	8,
	9,
	9,
	10,
	3,
	3,
	11,
	12,
	13,
	14,
	3,
	15,
	16,
	17,
	18,
	19,
	0,
	3,
	20,
	20,
	0,
	21,
	22,
	22,
	23,
	0,
	0,
	24,
	24,
	0,
	26,
	27,
	28,
	0,
	0,
	0,
	0,
	0,
	0,
	29,
	30,
	31,
	32,
	33,
	34,
	35,
	36,
	3,
	24,
	37,
	38,
	39,
	40,
	41,
	0,
	21,
	42,
	43,
	0,
	44,
	45,
	0,
	0,
	46,
	47,
	47,
	48,
	49,
	51,
	0,
	52
	],
	"aliases": {
	"Coughing": "Cough",
	"Hiccups": "Hiccup",
	"Clears Throat": "Throat Clearing",
	"Tongue Clicking": "Tongue Click",
	"Clicks Tongue": "Tongue Click",
	"Lip Smacking": "Lip Smack",
	"Smacks Lips": "Lip Smack",
	"Kissing Noises": "Kissing Sound",
	"Kissing Sounds": "Kissing Sound",
	"Gulps": "Gulp",
	"Sobs": "Sob",
	"Panting": "Pant",
	"Soft Hum": "Hum",
	"Resonant Hum": "Hum",
	"Humming": "Hum",
	"Deep Breath": "Breath",
	"Deep Breathing": "Breath",
	"Fast Breathing": "Breath",
	"Heavy Breathing": "Breath",
	"Slow Breathing": "Breath",
	"Breath": "Breath",
	"Normal Breathing": "Breath"
	},
	"raw_p3_classes": [
	"Affirmative Grunt",
	"Ahem",
	"Breath",
	"Breathy Giggle",
	"Cackle",
	"Childlike Giggle",
	"Chuckle",
	"Clears Throat",
	"Clicks Tongue",
	"Contented Sigh",
	"Cough",
	"Coughing",
	"Crying",
	"Deep Breath",
	"Deep Breathing",
	"Displeased Grunt",
	"Effort Grunt",
	"Exasperated Sigh",
	"Exhausted Groan",
	"Fast Breathing",
	"Fearful Gasp",
	"Frustrated Groan",
	"Growl",
	"Guffaw",
	"Gulps",
	"Gurgling",
	"Heavy Breathing",
	"Hiccup",
	"Hiccups",
	"Hiss",
	"Humming",
	"Kissing Noises",
	"Kissing Sounds",
	"Laughter",
	"Licking Sound",
	"Lip Popping",
	"Lip Smack",
	"Lip Smacking",
	"Moan",
	"Mournful Wail",
	"Nervous Giggle",
	"Nervous Gulp",
	"Nose Blowing",
	"Pain Moan",
	"Pant",
	"Panting",
	"Person Whistling Playfully",
	"Pleasure Moan",
	"Purr",
	"Quiet Sob",
	"Relief Sigh",
	"Scream",
	"Sharp Inhale",
	"Sharp Whistle",
	"Shriek",
	"Sigh",
	"Slow Breathing",
	"Smacks Lips",
	"Sneeze",
	"Snicker",
	"Sniff",
	"Snort",
	"Snorting Giggle",
	"Sobs",
	"Soft Hum",
	"Soft Whistle",
	"Spitting",
	"Sucking Noise",
	"Surprised Gasp",
	"Swallows",
	"Teeth Chattering",
	"Teeth Grinding",
	"Throat Clearing",
	"Tongue Click",
	"Tongue Clicking",
	"Trembling Whimper",
	"Tsk",
	"Wistful Sigh",
	"Wolf Whistle",
	"Yawn"
	],
	"raw_ladder_classes": [
	"Low Mumble",
	"Surprised Gasp",
	"Ahem",
	"Contented Sigh",
	"Wistful Sigh",
	"Exhausted Groan",
	"Resonant Hum",
	"Chuckle",
	"Breathy Giggle",
	"Childlike Giggle",
	"Yawn",
	"Sharp Inhale",
	"Cackle",
	"Exasperated Sigh",
	"Soft Hum",
	"Scream",
	"Trembling Whimper",
	"Deep Breath",
	"Shriek",
	"Mournful Wail",
	"Whispered Mumble",
	"Tsk",
	"Effort Grunt",
	"Coughing",
	"Sniff",
	"Purr",
	"Cough",
	"Growl",
	"Guffaw",
	"Person Whistling Playfully",
	"Frustrated Groan",
	"Affirmative Grunt",
	"Convulsive Sob",
	"Snort",
	"Soft Whistle",
	"Snorting Giggle",
	"Relief Sigh",
	"Humming",
	"Sharp Whistle",
	"Displeased Grunt",
	"Deep Breathing",
	"Pleasure Moan",
	"Fearful Gasp",
	"Clears Throat",
	"Pain Moan",
	"Hiss",
	"Hiccups",
	"Kissing Noises",
	"Slow Breathing",
	"Quiet Sob",
	"Lip Smack",
	"Tongue Click",
	"Fast Breathing",
	"Gurgling",
	"Heavy Breathing",
	"Hiccup",
	"Gulps",
	"Licking Sound",
	"Person Whistling to Get Attention",
	"Sobs",
	"Normal Breathing",
	"Wolf Whistle",
	"Smacks Lips",
	"Slurping Noises"
	]
	}