{
  "accum_steps": 1,
  "architectures": [
    "FWKVVisionModel"
  ],
  "batch_size": 32,
  "clip_id": "openai/clip-vit-base-patch32",
  "data_source": "t2i2m",
  "depth": 12,
  "dtype": "float32",
  "epochs": 5,
  "ffn_mult": 4,
  "image_size": 256,
  "latent_cache_path": "fwkv_vision_latent_cache.pt",
  "latent_ch": 4,
  "latent_size": 32,
  "lr": 0.0001,
  "model_type": "fwkv_vision",
  "n_examples": 100000,
  "n_heads": 6,
  "patch": 2,
  "save_dir": "fwkv_vision_ckpt",
  "text_dim": 512,
  "transformers_version": "5.0.0",
  "vae_id": "stabilityai/sd-vae-ft-mse",
  "vae_scaling": 0.18215,
  "width": 384
}