{
  "architectures": [
    "AsymmetricMagVitV2Pipline"
  ],
  "decoder": {
    "ch": 128,
    "ch_mult": [
      1,
      2,
      4,
      4
    ],
    "double_z": true,
    "in_channels": 3,
    "num_res_blocks": 2,
    "out_ch": 3,
    "temporal_up_layers": [
      2,
      3
    ],
    "type": "VideoDecoder",
    "video_frame_num": 17,
    "z_channels": 16
  },
  "encoder": {
    "ch": 128,
    "ch_mult": [
      1,
      2,
      4,
      4
    ],
    "double_z": true,
    "down_sampling_layer": [
      1,
      2
    ],
    "in_channels": 3,
    "num_res_blocks": 2,
    "out_ch": 3,
    "type": "VideoEncoder",
    "video_frame_num": 17,
    "z_channels": 16
  },
  "model_type": "AsymmetricMagVitV2Pipline",
  "regularizer": {
    "type": "DiagonalGaussianRegularizer"
  },
  "torch_dtype": "bfloat16",
  "transformers_version": "4.19.1"
}