{
  "_class_name": "MOSSImageTransformer",
  "_diffusers_version": "0.40.0.dev0",
  "dim": 64,
  "ffn_dim": 128,
  "freq_dim": 32,
  "latent_channels": 128,
  "num_heads": 2,
  "num_layers": 4,
  "rope_base": 10000.0,
  "text_dim": 32
}