URSA-1.7B/vae/config.json

{
  "_class_name": "AutoencoderVQCosmos3D",
  "_quantizer_name": "FSQuantizer",
  "in_channels": 3,
  "latent_channels": 256,
  "layers_per_block": 2,
  "norm_num_groups": 1,
  "out_channels": 3,
  "sample_size": 1024,
  "sample_frames": 49,
  "num_vq_embeddings": 64000,
  "vq_embed_dim": 6,
  "patch_size": 2,
  "temporal_stride": 4,
  "spatial_stride": 8,
  "block_out_channels": [
    128,
    256,
    512,
    512
  ]
}