transformer/config.json

{
  "_class_name": "CogView4Transformer2DModel",
  "_diffusers_version": "0.33.0.dev0",
  "attention_head_dim": 128,
  "condition_dim": 256,
  "in_channels": 16,
  "num_attention_heads": 32,
  "num_layers": 28,
  "out_channels": 16,
  "patch_size": 2,
  "pos_embed_max_size": 128,
  "rope_axes_dim": [
    256,
    256
  ],
  "sample_size": 128,
  "text_embed_dim": 4096,
  "time_embed_dim": 512
}