models/SparseVideoNav-Models/svn_ckpt/config.json

{
  "_class_name": "SVNModel",
  "_diffusers_version": "0.35.1",
  "vgm_dim": 1536,
  "vgm_ffn_dim": 8960,
  "vgm_num_heads": 12,
  "vgm_num_layers": 30,
  "vgm_qformer_config": {
    "depth": 4,
    "dim": 512,
    "heads": 8,
    "language_dim": 4096,
    "num_latents": 10240,
    "video_feature_dim": 16
  },
  "vgm_video_former_config": {
    "depth": 6,
    "dim": 512,
    "heads": 8,
    "language_dim": 4096,
    "num_frame": 10,
    "num_latents": 2560,
    "video_feature_dim": 512
  }
}