{ "_class_name": "IdolTransformerModel", "_diffusers_version": "0.42.0.dev0", "attention_head_dim": 64, "camera_channels": 5, "hidden_size": 1024, "image_embed_dim": 1280, "image_vae_embed_dim": 128, "in_channels": 16, "mlp_ratio": 4.0, "num_attention_heads": 16, "num_latent_tokens": 8192, "num_layers": 24, "num_refiner_layers": 2 }