transformer/config.json

{
  "_class_name": "StableAudioDiT",
  "io_channels": 64,
  "embed_dim": 1536,
  "depth": 24,
  "cond_token_dim": 768,
  "global_cond_dim": 1536,
  "project_cond_tokens": false,
  "num_attention_heads": 24,
  "project_global_cond": false
}