{
  "model_type": "whisper",
  "architectures": [
    "WhisperForConditionalGeneration"
  ],
  "d_model": 768,
  "d_ff": 3072,
  "num_layers": 12,
  "num_heads": 12,
  "vocab_size": 50265,
  "dropout": 0.1,
  "attention_dropout": 0.1,
  "activation_function": "gelu",
  "max_position_embeddings": 1024,
  "layer_norm_eps": 1e-5,
  "use_cache": true
}