{
  "format": "openvision_video_vit_reconstruction_export",
  "source_model": "models/video_vit.py",
  "source_unified_model": "models/unified_tokenizer.py",
  "source_config": "configs/unified_video_vit.py",
  "source_workdir": "gs://data_us_central1_a/experiment/zlt/openvision4//unified_3d_vit/pre_unified_B16_decvit_B1_textdec_L_bz_2_recon_1_lowdim_True_routeakl_False_1e-4_lpips_1_gram_5_contrastive_False_caption_True_capw_1__unified_3d_vit_fp32_20000_res_128_v159/ft/ft_unified_B16_decvit_B1_textdec_L_bz_2_recon_1_lowdim_True_routeakl_True_1e-4_lpips_1_gram_20_contrastive_False_caption_True_capw_1__unified_3d_vit_fp32_2000_res_256_v159/mix/mix_unified_B16_decvit_B1_textdec_L_bz_1_recon_1_lowdim_True_routeakl_True_1e-4_lpips_1_gram_20_contrastive_False_caption_True_capw_1_mixvideo_imgbs_4096_vidbs_256__unified_3d_vit_fp32_1000_res_256_v159",
  "restored_step": 294386,
  "init_image_shape": [
    256,
    1,
    256,
    256,
    3
  ],
  "encoder": {
    "variant": "B/16",
    "patch_size": [
      4,
      16,
      16
    ],
    "width": 768,
    "depth": 12,
    "mlp_dim": 3072,
    "num_heads": 12,
    "pool_type": "gap",
    "post_norm": false,
    "emb_head_bias": false,
    "pad_value": 0.0,
    "in_channels": 3,
    "patch_dim": 3072,
    "posemb": "rope3d",
    "use_dense_general": false,
    "naflex_supported": false
  },
  "decoder": {
    "variant": "B/1",
    "patch_size": [
      1,
      1,
      1
    ],
    "width": 768,
    "depth": 12,
    "mlp_dim": 3072,
    "num_heads": 12,
    "pool_type": "gap",
    "post_norm": false,
    "emb_head_bias": false,
    "pad_value": 0.0,
    "in_channels": 768,
    "patch_dim": 768,
    "posemb": "rope3d",
    "use_dense_general": false,
    "naflex_supported": false
  },
  "lowdim": {
    "use_lowdim_image": true,
    "use_image_lowdim_kl_regularization": true,
    "image_token_bottleneck_dim": 32,
    "paths": {
      "Dense_0": "/home/letianzhang/jax2hf/unified_3d_vit_v159_mix_256/dense/top_dense0.pt",
      "image_lowdim_down": "/home/letianzhang/jax2hf/unified_3d_vit_v159_mix_256/dense/image_lowdim_down.pt",
      "image_lowdim_down_logvar": "/home/letianzhang/jax2hf/unified_3d_vit_v159_mix_256/dense/image_lowdim_down_logvar.pt",
      "image_lowdim_up": "/home/letianzhang/jax2hf/unified_3d_vit_v159_mix_256/dense/image_lowdim_up.pt"
    }
  },
  "naflex_supported": false,
  "weights": {
    "encoder": "encoder/video_vit_pytorch_model.bin",
    "decoder": "decoder/video_vit_pytorch_model.bin",
    "dense_dir": "dense"
  }
}