text_encoder_2/config.json

{
  "_name_or_path": "laion/clap-htsat-fused",
  "architectures": ["CLIPTextModelWithProjection"],
  "hidden_act": "gelu",
  "hidden_size": 1024,
  "initializer_range": 0.02,
  "intermediate_size": 4096,
  "layer_norm_eps": 1e-05,
  "model_type": "clip",
  "num_attention_heads": 16,
  "num_hidden_layers": 24,
  "pad_token_id": 1,
  "projection_dim": 1280,
  "vocab_size": 49408
}