ast-esc50 / config.json
jmcmeen's picture
Upload model
6fb947c
{
"architectures": [
"ASTForAudioClassification"
],
"attention_probs_dropout_prob": 0.0,
"dtype": "float32",
"frequency_stride": 10,
"hidden_act": "gelu",
"hidden_dropout_prob": 0.0,
"hidden_size": 768,
"id2label": {
"0": "airplane",
"1": "breathing",
"2": "brushing_teeth",
"3": "can_opening",
"4": "car_horn",
"5": "cat",
"6": "chainsaw",
"7": "chirping_birds",
"8": "church_bells",
"9": "clapping",
"10": "clock_alarm",
"11": "clock_tick",
"12": "coughing",
"13": "cow",
"14": "crackling_fire",
"15": "crickets",
"16": "crow",
"17": "crying_baby",
"18": "dog",
"19": "door_wood_creaks",
"20": "door_wood_knock",
"21": "drinking_sipping",
"22": "engine",
"23": "fireworks",
"24": "footsteps",
"25": "frog",
"26": "glass_breaking",
"27": "hand_saw",
"28": "helicopter",
"29": "hen",
"30": "insects",
"31": "keyboard_typing",
"32": "laughing",
"33": "mouse_click",
"34": "pig",
"35": "pouring_water",
"36": "rain",
"37": "rooster",
"38": "sea_waves",
"39": "sheep",
"40": "siren",
"41": "sneezing",
"42": "snoring",
"43": "thunderstorm",
"44": "toilet_flush",
"45": "train",
"46": "vacuum_cleaner",
"47": "washing_machine",
"48": "water_drops",
"49": "wind"
},
"initializer_range": 0.02,
"intermediate_size": 3072,
"label2id": {
"airplane": 0,
"breathing": 1,
"brushing_teeth": 2,
"can_opening": 3,
"car_horn": 4,
"cat": 5,
"chainsaw": 6,
"chirping_birds": 7,
"church_bells": 8,
"clapping": 9,
"clock_alarm": 10,
"clock_tick": 11,
"coughing": 12,
"cow": 13,
"crackling_fire": 14,
"crickets": 15,
"crow": 16,
"crying_baby": 17,
"dog": 18,
"door_wood_creaks": 19,
"door_wood_knock": 20,
"drinking_sipping": 21,
"engine": 22,
"fireworks": 23,
"footsteps": 24,
"frog": 25,
"glass_breaking": 26,
"hand_saw": 27,
"helicopter": 28,
"hen": 29,
"insects": 30,
"keyboard_typing": 31,
"laughing": 32,
"mouse_click": 33,
"pig": 34,
"pouring_water": 35,
"rain": 36,
"rooster": 37,
"sea_waves": 38,
"sheep": 39,
"siren": 40,
"sneezing": 41,
"snoring": 42,
"thunderstorm": 43,
"toilet_flush": 44,
"train": 45,
"vacuum_cleaner": 46,
"washing_machine": 47,
"water_drops": 48,
"wind": 49
},
"layer_norm_eps": 1e-12,
"max_length": 1024,
"model_type": "audio-spectrogram-transformer",
"num_attention_heads": 12,
"num_hidden_layers": 12,
"num_mel_bins": 128,
"patch_size": 16,
"problem_type": "single_label_classification",
"qkv_bias": true,
"time_stride": 10,
"transformers_version": "4.57.3"
}