transformer/config.json

{
  "_class_name": "HunyuanVideoTransformer3DModel",
  "_diffusers_version": "0.32.0.dev0",
  "attention_head_dim": 128,
  "guidance_embeds": true,
  "in_channels": 16,
  "mlp_ratio": 4.0,
  "num_attention_heads": 24,
  "num_layers": 20,
  "num_refiner_layers": 2,
  "num_single_layers": 40,
  "out_channels": 16,
  "patch_size": 2,
  "patch_size_t": 1,
  "pooled_projection_dim": 768,
  "qk_norm": "rms_norm",
  "rope_axes_dim": [
    16,
    56,
    56
  ],
  "rope_theta": 256.0,
  "text_embed_dim": 4096
}