transformer/config.json

{
  "_class_name": "SanaWMTransformer3DModel",
  "_diffusers_version": "0.39.0.dev0",
  "cam_attn_compress": 1,
  "caption_channels": 2304,
  "chunk_plucker_channels": 48,
  "chunk_plucker_post_attn_blocks": 20,
  "conv_kernel_size": 4,
  "cross_norm": true,
  "ffn_type": "GLUMBConvTemp",
  "fp32_attention": true,
  "image_size": 720,
  "in_channels": 128,
  "init_cam_from_base": true,
  "k_conv_only": true,
  "learn_sigma": false,
  "linear_head_dim": 112,
  "mixed_precision": "bf16",
  "mlp_acts": [
    "silu",
    "silu",
    null
  ],
  "mlp_ratio": 3.0,
  "pos_embed_type": "wan_rope",
  "pred_sigma": false,
  "qk_norm": true,
  "softmax_every_n": 4,
  "t_kernel_size": 3,
  "use_chunk_plucker_post_attn": true,
  "use_pe": true,
  "y_norm": true
}