config.json

{
  "activation_dropout": 0.0,
  "activation_function": "gelu",
  "architectures": [
    "VoxtralEncoder"
  ],
  "attention_dropout": 0.0,
  "dropout": 0.0,
  "dtype": "float32",
  "hidden_size": 32,
  "initializer_range": 0.02,
  "intermediate_size": 32,
  "layerdrop": 0.0,
  "max_source_positions": 200,
  "model_type": "voxtral_encoder",
  "num_attention_heads": 2,
  "num_hidden_layers": 2,
  "num_mel_bins": 80,
  "scale_embedding": false,
  "transformers_version": "5.16.0.dev0",
  "vocab_size": 99
}