config.json

{
  "_name_or_path": "MIT/ast-finetuned-speech-commands-v2",
  "architectures": [
    "ASTForAudioClassification"
  ],
  "attention_probs_dropout_prob": 0.0,
  "frequency_stride": 10,
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.0,
  "hidden_size": 768,
  "id2label": {
    "0": "backward",
    "1": "follow",
    "2": "five",
    "3": "bed",
    "4": "zero",
    "5": "on",
    "6": "learn",
    "7": "two",
    "8": "house",
    "9": "tree",
    "10": "dog",
    "11": "stop",
    "12": "seven",
    "13": "eight",
    "14": "down",
    "15": "six",
    "16": "forward",
    "17": "cat",
    "18": "right",
    "19": "visual",
    "20": "four",
    "21": "wow",
    "22": "no",
    "23": "nine",
    "24": "off",
    "25": "three",
    "26": "left",
    "27": "marvin",
    "28": "yes",
    "29": "up",
    "30": "sheila",
    "31": "happy",
    "32": "bird",
    "33": "go",
    "34": "one"
  },
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "label2id": {
    "backward": 0,
    "bed": 3,
    "bird": 32,
    "cat": 17,
    "dog": 10,
    "down": 14,
    "eight": 13,
    "five": 2,
    "follow": 1,
    "forward": 16,
    "four": 20,
    "go": 33,
    "happy": 31,
    "house": 8,
    "learn": 6,
    "left": 26,
    "marvin": 27,
    "nine": 23,
    "no": 22,
    "off": 24,
    "on": 5,
    "one": 34,
    "right": 18,
    "seven": 12,
    "sheila": 30,
    "six": 15,
    "stop": 11,
    "three": 25,
    "tree": 9,
    "two": 7,
    "up": 29,
    "visual": 19,
    "wow": 21,
    "yes": 28,
    "zero": 4
  },
  "layer_norm_eps": 1e-12,
  "max_length": 128,
  "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,
  "torch_dtype": "float32",
  "transformers_version": "4.37.0"
}