config.json

{
  "architectures": [
    "ClapModel"
  ],
  "audio_config": {
    "_attn_implementation_autoset": true,
    "aff_block_r": 4,
    "attention_probs_dropout_prob": 0.0,
    "depths": [
      2,
      2,
      6,
      2
    ],
    "drop_path_rate": 0.0,
    "enable_fusion": true,
    "enable_patch_fusion": true,
    "enable_patch_layer_norm": true,
    "flatten_patch_embeds": true,
    "fusion_num_hidden_layers": 2,
    "fusion_type": null,
    "hidden_act": "gelu",
    "hidden_dropout_prob": 0.1,
    "hidden_size": 768,
    "initializer_factor": 1.0,
    "layer_norm_eps": 1e-05,
    "mlp_ratio": 4.0,
    "model_type": "clap_audio_model",
    "num_attention_heads": [
      4,
      8,
      16,
      32
    ],
    "num_classes": 527,
    "num_hidden_layers": 4,
    "num_mel_bins": 64,
    "patch_embed_input_channels": 1,
    "patch_embeds_hidden_size": 96,
    "patch_size": 4,
    "patch_stride": [
      4,
      4
    ],
    "projection_dim": 512,
    "projection_hidden_act": "relu",
    "projection_hidden_size": 768,
    "qkv_bias": true,
    "spec_size": 256,
    "torch_dtype": "float32",
    "window_size": 8
  },
  "hidden_size": 768,
  "initializer_factor": 1.0,
  "logit_scale_init_value": 14.285714285714285,
  "model_type": "clap",
  "num_hidden_layers": 16,
  "projection_dim": 512,
  "projection_hidden_act": "relu",
  "text_config": {
    "_attn_implementation_autoset": true,
    "attention_probs_dropout_prob": 0.1,
    "classifier_dropout": null,
    "fusion_hidden_size": 768,
    "fusion_num_hidden_layers": 2,
    "hidden_act": "gelu",
    "hidden_dropout_prob": 0.1,
    "hidden_size": 768,
    "initializer_factor": 1.0,
    "initializer_range": 0.02,
    "intermediate_size": 3072,
    "layer_norm_eps": 1e-12,
    "max_position_embeddings": 514,
    "model_type": "clap_text_model",
    "num_attention_heads": 12,
    "num_hidden_layers": 12,
    "position_embedding_type": "absolute",
    "projection_dim": 512,
    "projection_hidden_act": "relu",
    "projection_hidden_size": 768,
    "torch_dtype": "float32",
    "type_vocab_size": 1,
    "use_cache": true,
    "vocab_size": 50265
  },
  "torch_dtype": "float32",
  "transformers_version": "4.50.0"
}