transformer/config.json

{
  "_class_name": "HunyuanTransformer3DModel",
  "_diffusers_version": "0.30.1",
  "activation_fn": "gelu-approximate",
  "add_noise_in_inpaint_model": false,
  "after_norm": false,
  "attention_head_dim": 88,
  "basic_block_type": "basic",
  "cross_attention_dim": 1024,
  "cross_attention_dim_t5": 2048,
  "enable_clip_in_inpaint": true,
  "enable_text_attention_mask": true,
  "hidden_size": 1408,
  "in_channels": 12,
  "learn_sigma": true,
  "mlp_ratio": 4.3637,
  "motion_module_kwargs": null,
  "motion_module_kwargs_even": null,
  "motion_module_kwargs_odd": null,
  "motion_module_type": "VanillaGrid",
  "n_query": 16,
  "norm_type": "layer_norm",
  "num_attention_heads": 16,
  "num_layers": 40,
  "out_channels": 8,
  "patch_size": 2,
  "pooled_projection_dim": 1024,
  "projection_dim": 1024,
  "resize_inpaint_mask_directly": false,
  "sample_size": 128,
  "text_len": 77,
  "text_len_t5": 256,
  "time_position_encoding": true,
  "time_position_encoding_type": "2d_rope"
}