ae_metadata.json

{
  "decoder.conv_in.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.conv_in.weight": {
    "shape": [
      512,
      16,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.conv_out.bias": {
    "shape": [
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.conv_out.weight": {
    "shape": [
      3,
      128,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.attn_1.k.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.attn_1.k.weight": {
    "shape": [
      512,
      512,
      1,
      1
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.attn_1.norm.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.attn_1.norm.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.attn_1.proj_out.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.attn_1.proj_out.weight": {
    "shape": [
      512,
      512,
      1,
      1
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.attn_1.q.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.attn_1.q.weight": {
    "shape": [
      512,
      512,
      1,
      1
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.attn_1.v.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.attn_1.v.weight": {
    "shape": [
      512,
      512,
      1,
      1
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_1.conv1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_1.conv1.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_1.conv2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_1.conv2.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_1.norm1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_1.norm1.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_1.norm2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_1.norm2.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_2.conv1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_2.conv1.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_2.conv2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_2.conv2.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_2.norm1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_2.norm1.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_2.norm2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.mid.block_2.norm2.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.norm_out.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.norm_out.weight": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.0.conv1.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.0.conv1.weight": {
    "shape": [
      128,
      256,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.0.conv2.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.0.conv2.weight": {
    "shape": [
      128,
      128,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.0.nin_shortcut.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.0.nin_shortcut.weight": {
    "shape": [
      128,
      256,
      1,
      1
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.0.norm1.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.0.norm1.weight": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.0.norm2.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.0.norm2.weight": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.1.conv1.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.1.conv1.weight": {
    "shape": [
      128,
      128,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.1.conv2.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.1.conv2.weight": {
    "shape": [
      128,
      128,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.1.norm1.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.1.norm1.weight": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.1.norm2.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.1.norm2.weight": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.2.conv1.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.2.conv1.weight": {
    "shape": [
      128,
      128,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.2.conv2.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.2.conv2.weight": {
    "shape": [
      128,
      128,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.2.norm1.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.2.norm1.weight": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.2.norm2.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.0.block.2.norm2.weight": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.0.conv1.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.0.conv1.weight": {
    "shape": [
      256,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.0.conv2.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.0.conv2.weight": {
    "shape": [
      256,
      256,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.0.nin_shortcut.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.0.nin_shortcut.weight": {
    "shape": [
      256,
      512,
      1,
      1
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.0.norm1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.0.norm1.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.0.norm2.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.0.norm2.weight": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.1.conv1.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.1.conv1.weight": {
    "shape": [
      256,
      256,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.1.conv2.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.1.conv2.weight": {
    "shape": [
      256,
      256,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.1.norm1.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.1.norm1.weight": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.1.norm2.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.1.norm2.weight": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.2.conv1.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.2.conv1.weight": {
    "shape": [
      256,
      256,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.2.conv2.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.2.conv2.weight": {
    "shape": [
      256,
      256,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.2.norm1.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.2.norm1.weight": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.2.norm2.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.block.2.norm2.weight": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.upsample.conv.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.1.upsample.conv.weight": {
    "shape": [
      256,
      256,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.0.conv1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.0.conv1.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.0.conv2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.0.conv2.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.0.norm1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.0.norm1.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.0.norm2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.0.norm2.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.1.conv1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.1.conv1.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.1.conv2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.1.conv2.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.1.norm1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.1.norm1.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.1.norm2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.1.norm2.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.2.conv1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.2.conv1.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.2.conv2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.2.conv2.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.2.norm1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.2.norm1.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.2.norm2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.block.2.norm2.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.upsample.conv.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.2.upsample.conv.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.0.conv1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.0.conv1.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.0.conv2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.0.conv2.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.0.norm1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.0.norm1.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.0.norm2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.0.norm2.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.1.conv1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.1.conv1.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.1.conv2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.1.conv2.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.1.norm1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.1.norm1.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.1.norm2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.1.norm2.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.2.conv1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.2.conv1.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.2.conv2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.2.conv2.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.2.norm1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.2.norm1.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.2.norm2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.block.2.norm2.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.upsample.conv.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "decoder.up.3.upsample.conv.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.conv_in.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.conv_in.weight": {
    "shape": [
      128,
      3,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.conv_out.bias": {
    "shape": [
      32
    ],
    "dtype": "torch.float32"
  },
  "encoder.conv_out.weight": {
    "shape": [
      32,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.0.conv1.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.0.conv1.weight": {
    "shape": [
      128,
      128,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.0.conv2.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.0.conv2.weight": {
    "shape": [
      128,
      128,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.0.norm1.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.0.norm1.weight": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.0.norm2.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.0.norm2.weight": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.1.conv1.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.1.conv1.weight": {
    "shape": [
      128,
      128,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.1.conv2.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.1.conv2.weight": {
    "shape": [
      128,
      128,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.1.norm1.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.1.norm1.weight": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.1.norm2.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.block.1.norm2.weight": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.downsample.conv.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.0.downsample.conv.weight": {
    "shape": [
      128,
      128,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.0.conv1.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.0.conv1.weight": {
    "shape": [
      256,
      128,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.0.conv2.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.0.conv2.weight": {
    "shape": [
      256,
      256,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.0.nin_shortcut.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.0.nin_shortcut.weight": {
    "shape": [
      256,
      128,
      1,
      1
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.0.norm1.bias": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.0.norm1.weight": {
    "shape": [
      128
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.0.norm2.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.0.norm2.weight": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.1.conv1.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.1.conv1.weight": {
    "shape": [
      256,
      256,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.1.conv2.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.1.conv2.weight": {
    "shape": [
      256,
      256,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.1.norm1.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.1.norm1.weight": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.1.norm2.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.block.1.norm2.weight": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.downsample.conv.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.1.downsample.conv.weight": {
    "shape": [
      256,
      256,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.0.conv1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.0.conv1.weight": {
    "shape": [
      512,
      256,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.0.conv2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.0.conv2.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.0.nin_shortcut.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.0.nin_shortcut.weight": {
    "shape": [
      512,
      256,
      1,
      1
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.0.norm1.bias": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.0.norm1.weight": {
    "shape": [
      256
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.0.norm2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.0.norm2.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.1.conv1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.1.conv1.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.1.conv2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.1.conv2.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.1.norm1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.1.norm1.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.1.norm2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.block.1.norm2.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.downsample.conv.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.2.downsample.conv.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.0.conv1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.0.conv1.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.0.conv2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.0.conv2.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.0.norm1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.0.norm1.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.0.norm2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.0.norm2.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.1.conv1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.1.conv1.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.1.conv2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.1.conv2.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.1.norm1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.1.norm1.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.1.norm2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.down.3.block.1.norm2.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.attn_1.k.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.attn_1.k.weight": {
    "shape": [
      512,
      512,
      1,
      1
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.attn_1.norm.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.attn_1.norm.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.attn_1.proj_out.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.attn_1.proj_out.weight": {
    "shape": [
      512,
      512,
      1,
      1
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.attn_1.q.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.attn_1.q.weight": {
    "shape": [
      512,
      512,
      1,
      1
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.attn_1.v.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.attn_1.v.weight": {
    "shape": [
      512,
      512,
      1,
      1
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_1.conv1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_1.conv1.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_1.conv2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_1.conv2.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_1.norm1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_1.norm1.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_1.norm2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_1.norm2.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_2.conv1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_2.conv1.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_2.conv2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_2.conv2.weight": {
    "shape": [
      512,
      512,
      3,
      3
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_2.norm1.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_2.norm1.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_2.norm2.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.mid.block_2.norm2.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.norm_out.bias": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  },
  "encoder.norm_out.weight": {
    "shape": [
      512
    ],
    "dtype": "torch.float32"
  }
}