decoder/vision/transformer/config.json

{
  "_class_name": "VisionTransformer",
  "_diffusers_version": "0.35.2",
  "axes_dim": [
    8,
    36,
    36
  ],
  "context_in_dim": 1536,
  "depth": 0,
  "depth_single_blocks": 35,
  "guidance_embed": false,
  "hidden_size": 1920,
  "in_channels": 16,
  "mlp_ratio": 4.0,
  "num_heads": 24,
  "qkv_bias": true,
  "theta": 10000,
  "use_patchify": false,
  "vec_in_dim": 1536
}