config.json

{
  "architectures": [
    "MSClapModel"
  ],
  "audio_config": {
    "model_type": "msclap_audio_model"
  },
  "hidden_size": 768,
  "initializer_factor": 1.0,
  "logit_scale_init_value": 14.285714285714285,
  "model_type": "clap",
  "num_hidden_layers": 8,
  "projection_dim": 1024,
  "text_config": {
    "model_type": "msclap_text_model",
    "projection_dim": 1024
  },
  "torch_dtype": "float32",
  "transformers_version": "4.43.0.dev0"
}