{
  "vae": {
    "in_channels": 1,
    "out_ch": 1,
    "ch": 128,
    "ch_mult": [
      1,
      2,
      4,
      8
    ],
    "num_res_blocks": 2,
    "attn_resolutions": [],
    "double_z": true,
    "z_channels": 16,
    "embed_dim": 16,
    "dropout": 0.1,
    "resolution": 256,
    "mel_bins": 256,
    "scale_factor_z": 0.3342
  },
  "ldm": {
    "in_channels": 32,
    "model_channels": 128,
    "out_channels": 16,
    "num_res_blocks": 2
  },
  "mel": {
    "n_fft": 2048,
    "hop": 480,
    "sr": 48000,
    "n_mels": 256,
    "fmin": 20,
    "fmax": 24000
  },
  "audio": {
    "sample_rate": 48000,
    "frame_samples": 245760,
    "frame_sec": 5.12
  },
  "format": "int4",
  "quantization": {
    "mode": "mlx_affine_flat",
    "bits": 4,
    "group_size": 64,
    "rule": "every weight with ndim>=2 and (prod(shape[1:]) % group_size == 0); reshape to (O, fan_in) before mx.quantize."
  },
  "quantized_shapes": {
    "ldm.input_blocks.1.0.emb_layers.1.weight": [
      128,
      512
    ],
    "ldm.input_blocks.1.0.in_layers.2.weight": [
      128,
      3,
      3,
      128
    ],
    "ldm.input_blocks.1.0.out_layers.3.weight": [
      128,
      3,
      3,
      128
    ],
    "ldm.input_blocks.10.0.emb_layers.1.weight": [
      640,
      512
    ],
    "ldm.input_blocks.10.0.in_layers.2.weight": [
      640,
      3,
      3,
      384
    ],
    "ldm.input_blocks.10.0.out_layers.3.weight": [
      640,
      3,
      3,
      640
    ],
    "ldm.input_blocks.10.0.skip_connection.weight": [
      640,
      1,
      1,
      384
    ],
    "ldm.input_blocks.10.1.proj_in.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.input_blocks.10.1.proj_out.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.input_blocks.10.1.transformer_blocks.0.attn1.to_k.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.1.transformer_blocks.0.attn1.to_out.0.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.1.transformer_blocks.0.attn1.to_q.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.1.transformer_blocks.0.attn1.to_v.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.1.transformer_blocks.0.attn2.to_k.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.1.transformer_blocks.0.attn2.to_out.0.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.1.transformer_blocks.0.attn2.to_q.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.1.transformer_blocks.0.attn2.to_v.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.1.transformer_blocks.0.ff.net.0.proj.weight": [
      5120,
      640
    ],
    "ldm.input_blocks.10.1.transformer_blocks.0.ff.net.2.weight": [
      640,
      2560
    ],
    "ldm.input_blocks.10.2.proj_in.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.input_blocks.10.2.proj_out.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.input_blocks.10.2.transformer_blocks.0.attn1.to_k.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.2.transformer_blocks.0.attn1.to_out.0.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.2.transformer_blocks.0.attn1.to_q.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.2.transformer_blocks.0.attn1.to_v.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.2.transformer_blocks.0.attn2.to_k.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.2.transformer_blocks.0.attn2.to_out.0.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.2.transformer_blocks.0.attn2.to_q.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.2.transformer_blocks.0.attn2.to_v.weight": [
      640,
      640
    ],
    "ldm.input_blocks.10.2.transformer_blocks.0.ff.net.0.proj.weight": [
      5120,
      640
    ],
    "ldm.input_blocks.10.2.transformer_blocks.0.ff.net.2.weight": [
      640,
      2560
    ],
    "ldm.input_blocks.11.0.emb_layers.1.weight": [
      640,
      512
    ],
    "ldm.input_blocks.11.0.in_layers.2.weight": [
      640,
      3,
      3,
      640
    ],
    "ldm.input_blocks.11.0.out_layers.3.weight": [
      640,
      3,
      3,
      640
    ],
    "ldm.input_blocks.11.1.proj_in.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.input_blocks.11.1.proj_out.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.input_blocks.11.1.transformer_blocks.0.attn1.to_k.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.1.transformer_blocks.0.attn1.to_out.0.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.1.transformer_blocks.0.attn1.to_q.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.1.transformer_blocks.0.attn1.to_v.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.1.transformer_blocks.0.attn2.to_k.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.1.transformer_blocks.0.attn2.to_out.0.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.1.transformer_blocks.0.attn2.to_q.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.1.transformer_blocks.0.attn2.to_v.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.1.transformer_blocks.0.ff.net.0.proj.weight": [
      5120,
      640
    ],
    "ldm.input_blocks.11.1.transformer_blocks.0.ff.net.2.weight": [
      640,
      2560
    ],
    "ldm.input_blocks.11.2.proj_in.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.input_blocks.11.2.proj_out.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.input_blocks.11.2.transformer_blocks.0.attn1.to_k.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.2.transformer_blocks.0.attn1.to_out.0.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.2.transformer_blocks.0.attn1.to_q.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.2.transformer_blocks.0.attn1.to_v.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.2.transformer_blocks.0.attn2.to_k.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.2.transformer_blocks.0.attn2.to_out.0.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.2.transformer_blocks.0.attn2.to_q.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.2.transformer_blocks.0.attn2.to_v.weight": [
      640,
      640
    ],
    "ldm.input_blocks.11.2.transformer_blocks.0.ff.net.0.proj.weight": [
      5120,
      640
    ],
    "ldm.input_blocks.11.2.transformer_blocks.0.ff.net.2.weight": [
      640,
      2560
    ],
    "ldm.input_blocks.2.0.emb_layers.1.weight": [
      128,
      512
    ],
    "ldm.input_blocks.2.0.in_layers.2.weight": [
      128,
      3,
      3,
      128
    ],
    "ldm.input_blocks.2.0.out_layers.3.weight": [
      128,
      3,
      3,
      128
    ],
    "ldm.input_blocks.3.0.op.weight": [
      128,
      3,
      3,
      128
    ],
    "ldm.input_blocks.4.0.emb_layers.1.weight": [
      256,
      512
    ],
    "ldm.input_blocks.4.0.in_layers.2.weight": [
      256,
      3,
      3,
      128
    ],
    "ldm.input_blocks.4.0.out_layers.3.weight": [
      256,
      3,
      3,
      256
    ],
    "ldm.input_blocks.4.0.skip_connection.weight": [
      256,
      1,
      1,
      128
    ],
    "ldm.input_blocks.4.1.proj_in.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.input_blocks.4.1.proj_out.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.input_blocks.4.1.transformer_blocks.0.attn1.to_k.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.1.transformer_blocks.0.attn1.to_out.0.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.1.transformer_blocks.0.attn1.to_q.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.1.transformer_blocks.0.attn1.to_v.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.1.transformer_blocks.0.attn2.to_k.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.1.transformer_blocks.0.attn2.to_out.0.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.1.transformer_blocks.0.attn2.to_q.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.1.transformer_blocks.0.attn2.to_v.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.1.transformer_blocks.0.ff.net.0.proj.weight": [
      2048,
      256
    ],
    "ldm.input_blocks.4.1.transformer_blocks.0.ff.net.2.weight": [
      256,
      1024
    ],
    "ldm.input_blocks.4.2.proj_in.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.input_blocks.4.2.proj_out.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.input_blocks.4.2.transformer_blocks.0.attn1.to_k.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.2.transformer_blocks.0.attn1.to_out.0.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.2.transformer_blocks.0.attn1.to_q.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.2.transformer_blocks.0.attn1.to_v.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.2.transformer_blocks.0.attn2.to_k.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.2.transformer_blocks.0.attn2.to_out.0.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.2.transformer_blocks.0.attn2.to_q.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.2.transformer_blocks.0.attn2.to_v.weight": [
      256,
      256
    ],
    "ldm.input_blocks.4.2.transformer_blocks.0.ff.net.0.proj.weight": [
      2048,
      256
    ],
    "ldm.input_blocks.4.2.transformer_blocks.0.ff.net.2.weight": [
      256,
      1024
    ],
    "ldm.input_blocks.5.0.emb_layers.1.weight": [
      256,
      512
    ],
    "ldm.input_blocks.5.0.in_layers.2.weight": [
      256,
      3,
      3,
      256
    ],
    "ldm.input_blocks.5.0.out_layers.3.weight": [
      256,
      3,
      3,
      256
    ],
    "ldm.input_blocks.5.1.proj_in.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.input_blocks.5.1.proj_out.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.input_blocks.5.1.transformer_blocks.0.attn1.to_k.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.1.transformer_blocks.0.attn1.to_out.0.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.1.transformer_blocks.0.attn1.to_q.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.1.transformer_blocks.0.attn1.to_v.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.1.transformer_blocks.0.attn2.to_k.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.1.transformer_blocks.0.attn2.to_out.0.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.1.transformer_blocks.0.attn2.to_q.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.1.transformer_blocks.0.attn2.to_v.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.1.transformer_blocks.0.ff.net.0.proj.weight": [
      2048,
      256
    ],
    "ldm.input_blocks.5.1.transformer_blocks.0.ff.net.2.weight": [
      256,
      1024
    ],
    "ldm.input_blocks.5.2.proj_in.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.input_blocks.5.2.proj_out.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.input_blocks.5.2.transformer_blocks.0.attn1.to_k.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.2.transformer_blocks.0.attn1.to_out.0.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.2.transformer_blocks.0.attn1.to_q.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.2.transformer_blocks.0.attn1.to_v.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.2.transformer_blocks.0.attn2.to_k.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.2.transformer_blocks.0.attn2.to_out.0.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.2.transformer_blocks.0.attn2.to_q.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.2.transformer_blocks.0.attn2.to_v.weight": [
      256,
      256
    ],
    "ldm.input_blocks.5.2.transformer_blocks.0.ff.net.0.proj.weight": [
      2048,
      256
    ],
    "ldm.input_blocks.5.2.transformer_blocks.0.ff.net.2.weight": [
      256,
      1024
    ],
    "ldm.input_blocks.6.0.op.weight": [
      256,
      3,
      3,
      256
    ],
    "ldm.input_blocks.7.0.emb_layers.1.weight": [
      384,
      512
    ],
    "ldm.input_blocks.7.0.in_layers.2.weight": [
      384,
      3,
      3,
      256
    ],
    "ldm.input_blocks.7.0.out_layers.3.weight": [
      384,
      3,
      3,
      384
    ],
    "ldm.input_blocks.7.0.skip_connection.weight": [
      384,
      1,
      1,
      256
    ],
    "ldm.input_blocks.7.1.proj_in.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.input_blocks.7.1.proj_out.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.input_blocks.7.1.transformer_blocks.0.attn1.to_k.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.1.transformer_blocks.0.attn1.to_out.0.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.1.transformer_blocks.0.attn1.to_q.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.1.transformer_blocks.0.attn1.to_v.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.1.transformer_blocks.0.attn2.to_k.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.1.transformer_blocks.0.attn2.to_out.0.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.1.transformer_blocks.0.attn2.to_q.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.1.transformer_blocks.0.attn2.to_v.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.1.transformer_blocks.0.ff.net.0.proj.weight": [
      3072,
      384
    ],
    "ldm.input_blocks.7.1.transformer_blocks.0.ff.net.2.weight": [
      384,
      1536
    ],
    "ldm.input_blocks.7.2.proj_in.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.input_blocks.7.2.proj_out.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.input_blocks.7.2.transformer_blocks.0.attn1.to_k.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.2.transformer_blocks.0.attn1.to_out.0.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.2.transformer_blocks.0.attn1.to_q.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.2.transformer_blocks.0.attn1.to_v.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.2.transformer_blocks.0.attn2.to_k.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.2.transformer_blocks.0.attn2.to_out.0.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.2.transformer_blocks.0.attn2.to_q.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.2.transformer_blocks.0.attn2.to_v.weight": [
      384,
      384
    ],
    "ldm.input_blocks.7.2.transformer_blocks.0.ff.net.0.proj.weight": [
      3072,
      384
    ],
    "ldm.input_blocks.7.2.transformer_blocks.0.ff.net.2.weight": [
      384,
      1536
    ],
    "ldm.input_blocks.8.0.emb_layers.1.weight": [
      384,
      512
    ],
    "ldm.input_blocks.8.0.in_layers.2.weight": [
      384,
      3,
      3,
      384
    ],
    "ldm.input_blocks.8.0.out_layers.3.weight": [
      384,
      3,
      3,
      384
    ],
    "ldm.input_blocks.8.1.proj_in.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.input_blocks.8.1.proj_out.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.input_blocks.8.1.transformer_blocks.0.attn1.to_k.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.1.transformer_blocks.0.attn1.to_out.0.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.1.transformer_blocks.0.attn1.to_q.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.1.transformer_blocks.0.attn1.to_v.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.1.transformer_blocks.0.attn2.to_k.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.1.transformer_blocks.0.attn2.to_out.0.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.1.transformer_blocks.0.attn2.to_q.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.1.transformer_blocks.0.attn2.to_v.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.1.transformer_blocks.0.ff.net.0.proj.weight": [
      3072,
      384
    ],
    "ldm.input_blocks.8.1.transformer_blocks.0.ff.net.2.weight": [
      384,
      1536
    ],
    "ldm.input_blocks.8.2.proj_in.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.input_blocks.8.2.proj_out.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.input_blocks.8.2.transformer_blocks.0.attn1.to_k.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.2.transformer_blocks.0.attn1.to_out.0.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.2.transformer_blocks.0.attn1.to_q.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.2.transformer_blocks.0.attn1.to_v.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.2.transformer_blocks.0.attn2.to_k.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.2.transformer_blocks.0.attn2.to_out.0.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.2.transformer_blocks.0.attn2.to_q.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.2.transformer_blocks.0.attn2.to_v.weight": [
      384,
      384
    ],
    "ldm.input_blocks.8.2.transformer_blocks.0.ff.net.0.proj.weight": [
      3072,
      384
    ],
    "ldm.input_blocks.8.2.transformer_blocks.0.ff.net.2.weight": [
      384,
      1536
    ],
    "ldm.input_blocks.9.0.op.weight": [
      384,
      3,
      3,
      384
    ],
    "ldm.middle_block.0.emb_layers.1.weight": [
      640,
      512
    ],
    "ldm.middle_block.0.in_layers.2.weight": [
      640,
      3,
      3,
      640
    ],
    "ldm.middle_block.0.out_layers.3.weight": [
      640,
      3,
      3,
      640
    ],
    "ldm.middle_block.1.proj_in.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.middle_block.1.proj_out.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.middle_block.1.transformer_blocks.0.attn1.to_k.weight": [
      640,
      640
    ],
    "ldm.middle_block.1.transformer_blocks.0.attn1.to_out.0.weight": [
      640,
      640
    ],
    "ldm.middle_block.1.transformer_blocks.0.attn1.to_q.weight": [
      640,
      640
    ],
    "ldm.middle_block.1.transformer_blocks.0.attn1.to_v.weight": [
      640,
      640
    ],
    "ldm.middle_block.1.transformer_blocks.0.attn2.to_k.weight": [
      640,
      640
    ],
    "ldm.middle_block.1.transformer_blocks.0.attn2.to_out.0.weight": [
      640,
      640
    ],
    "ldm.middle_block.1.transformer_blocks.0.attn2.to_q.weight": [
      640,
      640
    ],
    "ldm.middle_block.1.transformer_blocks.0.attn2.to_v.weight": [
      640,
      640
    ],
    "ldm.middle_block.1.transformer_blocks.0.ff.net.0.proj.weight": [
      5120,
      640
    ],
    "ldm.middle_block.1.transformer_blocks.0.ff.net.2.weight": [
      640,
      2560
    ],
    "ldm.middle_block.2.proj_in.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.middle_block.2.proj_out.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.middle_block.2.transformer_blocks.0.attn1.to_k.weight": [
      640,
      640
    ],
    "ldm.middle_block.2.transformer_blocks.0.attn1.to_out.0.weight": [
      640,
      640
    ],
    "ldm.middle_block.2.transformer_blocks.0.attn1.to_q.weight": [
      640,
      640
    ],
    "ldm.middle_block.2.transformer_blocks.0.attn1.to_v.weight": [
      640,
      640
    ],
    "ldm.middle_block.2.transformer_blocks.0.attn2.to_k.weight": [
      640,
      640
    ],
    "ldm.middle_block.2.transformer_blocks.0.attn2.to_out.0.weight": [
      640,
      640
    ],
    "ldm.middle_block.2.transformer_blocks.0.attn2.to_q.weight": [
      640,
      640
    ],
    "ldm.middle_block.2.transformer_blocks.0.attn2.to_v.weight": [
      640,
      640
    ],
    "ldm.middle_block.2.transformer_blocks.0.ff.net.0.proj.weight": [
      5120,
      640
    ],
    "ldm.middle_block.2.transformer_blocks.0.ff.net.2.weight": [
      640,
      2560
    ],
    "ldm.middle_block.3.emb_layers.1.weight": [
      640,
      512
    ],
    "ldm.middle_block.3.in_layers.2.weight": [
      640,
      3,
      3,
      640
    ],
    "ldm.middle_block.3.out_layers.3.weight": [
      640,
      3,
      3,
      640
    ],
    "ldm.out.2.weight": [
      16,
      3,
      3,
      128
    ],
    "ldm.output_blocks.0.0.emb_layers.1.weight": [
      640,
      512
    ],
    "ldm.output_blocks.0.0.in_layers.2.weight": [
      640,
      3,
      3,
      1280
    ],
    "ldm.output_blocks.0.0.out_layers.3.weight": [
      640,
      3,
      3,
      640
    ],
    "ldm.output_blocks.0.0.skip_connection.weight": [
      640,
      1,
      1,
      1280
    ],
    "ldm.output_blocks.0.1.proj_in.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.output_blocks.0.1.proj_out.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.output_blocks.0.1.transformer_blocks.0.attn1.to_k.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.1.transformer_blocks.0.attn1.to_out.0.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.1.transformer_blocks.0.attn1.to_q.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.1.transformer_blocks.0.attn1.to_v.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.1.transformer_blocks.0.attn2.to_k.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.1.transformer_blocks.0.attn2.to_out.0.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.1.transformer_blocks.0.attn2.to_q.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.1.transformer_blocks.0.attn2.to_v.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.1.transformer_blocks.0.ff.net.0.proj.weight": [
      5120,
      640
    ],
    "ldm.output_blocks.0.1.transformer_blocks.0.ff.net.2.weight": [
      640,
      2560
    ],
    "ldm.output_blocks.0.2.proj_in.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.output_blocks.0.2.proj_out.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.output_blocks.0.2.transformer_blocks.0.attn1.to_k.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.2.transformer_blocks.0.attn1.to_out.0.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.2.transformer_blocks.0.attn1.to_q.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.2.transformer_blocks.0.attn1.to_v.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.2.transformer_blocks.0.attn2.to_k.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.2.transformer_blocks.0.attn2.to_out.0.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.2.transformer_blocks.0.attn2.to_q.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.2.transformer_blocks.0.attn2.to_v.weight": [
      640,
      640
    ],
    "ldm.output_blocks.0.2.transformer_blocks.0.ff.net.0.proj.weight": [
      5120,
      640
    ],
    "ldm.output_blocks.0.2.transformer_blocks.0.ff.net.2.weight": [
      640,
      2560
    ],
    "ldm.output_blocks.1.0.emb_layers.1.weight": [
      640,
      512
    ],
    "ldm.output_blocks.1.0.in_layers.2.weight": [
      640,
      3,
      3,
      1280
    ],
    "ldm.output_blocks.1.0.out_layers.3.weight": [
      640,
      3,
      3,
      640
    ],
    "ldm.output_blocks.1.0.skip_connection.weight": [
      640,
      1,
      1,
      1280
    ],
    "ldm.output_blocks.1.1.proj_in.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.output_blocks.1.1.proj_out.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.output_blocks.1.1.transformer_blocks.0.attn1.to_k.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.1.transformer_blocks.0.attn1.to_out.0.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.1.transformer_blocks.0.attn1.to_q.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.1.transformer_blocks.0.attn1.to_v.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.1.transformer_blocks.0.attn2.to_k.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.1.transformer_blocks.0.attn2.to_out.0.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.1.transformer_blocks.0.attn2.to_q.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.1.transformer_blocks.0.attn2.to_v.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.1.transformer_blocks.0.ff.net.0.proj.weight": [
      5120,
      640
    ],
    "ldm.output_blocks.1.1.transformer_blocks.0.ff.net.2.weight": [
      640,
      2560
    ],
    "ldm.output_blocks.1.2.proj_in.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.output_blocks.1.2.proj_out.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.output_blocks.1.2.transformer_blocks.0.attn1.to_k.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.2.transformer_blocks.0.attn1.to_out.0.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.2.transformer_blocks.0.attn1.to_q.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.2.transformer_blocks.0.attn1.to_v.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.2.transformer_blocks.0.attn2.to_k.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.2.transformer_blocks.0.attn2.to_out.0.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.2.transformer_blocks.0.attn2.to_q.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.2.transformer_blocks.0.attn2.to_v.weight": [
      640,
      640
    ],
    "ldm.output_blocks.1.2.transformer_blocks.0.ff.net.0.proj.weight": [
      5120,
      640
    ],
    "ldm.output_blocks.1.2.transformer_blocks.0.ff.net.2.weight": [
      640,
      2560
    ],
    "ldm.output_blocks.10.0.emb_layers.1.weight": [
      128,
      512
    ],
    "ldm.output_blocks.10.0.in_layers.2.weight": [
      128,
      3,
      3,
      256
    ],
    "ldm.output_blocks.10.0.out_layers.3.weight": [
      128,
      3,
      3,
      128
    ],
    "ldm.output_blocks.10.0.skip_connection.weight": [
      128,
      1,
      1,
      256
    ],
    "ldm.output_blocks.11.0.emb_layers.1.weight": [
      128,
      512
    ],
    "ldm.output_blocks.11.0.in_layers.2.weight": [
      128,
      3,
      3,
      256
    ],
    "ldm.output_blocks.11.0.out_layers.3.weight": [
      128,
      3,
      3,
      128
    ],
    "ldm.output_blocks.11.0.skip_connection.weight": [
      128,
      1,
      1,
      256
    ],
    "ldm.output_blocks.2.0.emb_layers.1.weight": [
      640,
      512
    ],
    "ldm.output_blocks.2.0.in_layers.2.weight": [
      640,
      3,
      3,
      1024
    ],
    "ldm.output_blocks.2.0.out_layers.3.weight": [
      640,
      3,
      3,
      640
    ],
    "ldm.output_blocks.2.0.skip_connection.weight": [
      640,
      1,
      1,
      1024
    ],
    "ldm.output_blocks.2.1.proj_in.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.output_blocks.2.1.proj_out.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.output_blocks.2.1.transformer_blocks.0.attn1.to_k.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.1.transformer_blocks.0.attn1.to_out.0.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.1.transformer_blocks.0.attn1.to_q.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.1.transformer_blocks.0.attn1.to_v.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.1.transformer_blocks.0.attn2.to_k.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.1.transformer_blocks.0.attn2.to_out.0.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.1.transformer_blocks.0.attn2.to_q.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.1.transformer_blocks.0.attn2.to_v.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.1.transformer_blocks.0.ff.net.0.proj.weight": [
      5120,
      640
    ],
    "ldm.output_blocks.2.1.transformer_blocks.0.ff.net.2.weight": [
      640,
      2560
    ],
    "ldm.output_blocks.2.2.proj_in.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.output_blocks.2.2.proj_out.weight": [
      640,
      1,
      1,
      640
    ],
    "ldm.output_blocks.2.2.transformer_blocks.0.attn1.to_k.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.2.transformer_blocks.0.attn1.to_out.0.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.2.transformer_blocks.0.attn1.to_q.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.2.transformer_blocks.0.attn1.to_v.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.2.transformer_blocks.0.attn2.to_k.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.2.transformer_blocks.0.attn2.to_out.0.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.2.transformer_blocks.0.attn2.to_q.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.2.transformer_blocks.0.attn2.to_v.weight": [
      640,
      640
    ],
    "ldm.output_blocks.2.2.transformer_blocks.0.ff.net.0.proj.weight": [
      5120,
      640
    ],
    "ldm.output_blocks.2.2.transformer_blocks.0.ff.net.2.weight": [
      640,
      2560
    ],
    "ldm.output_blocks.2.3.conv.weight": [
      640,
      3,
      3,
      640
    ],
    "ldm.output_blocks.3.0.emb_layers.1.weight": [
      384,
      512
    ],
    "ldm.output_blocks.3.0.in_layers.2.weight": [
      384,
      3,
      3,
      1024
    ],
    "ldm.output_blocks.3.0.out_layers.3.weight": [
      384,
      3,
      3,
      384
    ],
    "ldm.output_blocks.3.0.skip_connection.weight": [
      384,
      1,
      1,
      1024
    ],
    "ldm.output_blocks.3.1.proj_in.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.output_blocks.3.1.proj_out.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.output_blocks.3.1.transformer_blocks.0.attn1.to_k.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.1.transformer_blocks.0.attn1.to_out.0.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.1.transformer_blocks.0.attn1.to_q.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.1.transformer_blocks.0.attn1.to_v.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.1.transformer_blocks.0.attn2.to_k.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.1.transformer_blocks.0.attn2.to_out.0.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.1.transformer_blocks.0.attn2.to_q.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.1.transformer_blocks.0.attn2.to_v.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.1.transformer_blocks.0.ff.net.0.proj.weight": [
      3072,
      384
    ],
    "ldm.output_blocks.3.1.transformer_blocks.0.ff.net.2.weight": [
      384,
      1536
    ],
    "ldm.output_blocks.3.2.proj_in.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.output_blocks.3.2.proj_out.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.output_blocks.3.2.transformer_blocks.0.attn1.to_k.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.2.transformer_blocks.0.attn1.to_out.0.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.2.transformer_blocks.0.attn1.to_q.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.2.transformer_blocks.0.attn1.to_v.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.2.transformer_blocks.0.attn2.to_k.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.2.transformer_blocks.0.attn2.to_out.0.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.2.transformer_blocks.0.attn2.to_q.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.2.transformer_blocks.0.attn2.to_v.weight": [
      384,
      384
    ],
    "ldm.output_blocks.3.2.transformer_blocks.0.ff.net.0.proj.weight": [
      3072,
      384
    ],
    "ldm.output_blocks.3.2.transformer_blocks.0.ff.net.2.weight": [
      384,
      1536
    ],
    "ldm.output_blocks.4.0.emb_layers.1.weight": [
      384,
      512
    ],
    "ldm.output_blocks.4.0.in_layers.2.weight": [
      384,
      3,
      3,
      768
    ],
    "ldm.output_blocks.4.0.out_layers.3.weight": [
      384,
      3,
      3,
      384
    ],
    "ldm.output_blocks.4.0.skip_connection.weight": [
      384,
      1,
      1,
      768
    ],
    "ldm.output_blocks.4.1.proj_in.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.output_blocks.4.1.proj_out.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.output_blocks.4.1.transformer_blocks.0.attn1.to_k.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.1.transformer_blocks.0.attn1.to_out.0.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.1.transformer_blocks.0.attn1.to_q.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.1.transformer_blocks.0.attn1.to_v.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.1.transformer_blocks.0.attn2.to_k.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.1.transformer_blocks.0.attn2.to_out.0.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.1.transformer_blocks.0.attn2.to_q.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.1.transformer_blocks.0.attn2.to_v.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.1.transformer_blocks.0.ff.net.0.proj.weight": [
      3072,
      384
    ],
    "ldm.output_blocks.4.1.transformer_blocks.0.ff.net.2.weight": [
      384,
      1536
    ],
    "ldm.output_blocks.4.2.proj_in.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.output_blocks.4.2.proj_out.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.output_blocks.4.2.transformer_blocks.0.attn1.to_k.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.2.transformer_blocks.0.attn1.to_out.0.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.2.transformer_blocks.0.attn1.to_q.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.2.transformer_blocks.0.attn1.to_v.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.2.transformer_blocks.0.attn2.to_k.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.2.transformer_blocks.0.attn2.to_out.0.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.2.transformer_blocks.0.attn2.to_q.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.2.transformer_blocks.0.attn2.to_v.weight": [
      384,
      384
    ],
    "ldm.output_blocks.4.2.transformer_blocks.0.ff.net.0.proj.weight": [
      3072,
      384
    ],
    "ldm.output_blocks.4.2.transformer_blocks.0.ff.net.2.weight": [
      384,
      1536
    ],
    "ldm.output_blocks.5.0.emb_layers.1.weight": [
      384,
      512
    ],
    "ldm.output_blocks.5.0.in_layers.2.weight": [
      384,
      3,
      3,
      640
    ],
    "ldm.output_blocks.5.0.out_layers.3.weight": [
      384,
      3,
      3,
      384
    ],
    "ldm.output_blocks.5.0.skip_connection.weight": [
      384,
      1,
      1,
      640
    ],
    "ldm.output_blocks.5.1.proj_in.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.output_blocks.5.1.proj_out.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.output_blocks.5.1.transformer_blocks.0.attn1.to_k.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.1.transformer_blocks.0.attn1.to_out.0.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.1.transformer_blocks.0.attn1.to_q.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.1.transformer_blocks.0.attn1.to_v.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.1.transformer_blocks.0.attn2.to_k.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.1.transformer_blocks.0.attn2.to_out.0.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.1.transformer_blocks.0.attn2.to_q.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.1.transformer_blocks.0.attn2.to_v.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.1.transformer_blocks.0.ff.net.0.proj.weight": [
      3072,
      384
    ],
    "ldm.output_blocks.5.1.transformer_blocks.0.ff.net.2.weight": [
      384,
      1536
    ],
    "ldm.output_blocks.5.2.proj_in.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.output_blocks.5.2.proj_out.weight": [
      384,
      1,
      1,
      384
    ],
    "ldm.output_blocks.5.2.transformer_blocks.0.attn1.to_k.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.2.transformer_blocks.0.attn1.to_out.0.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.2.transformer_blocks.0.attn1.to_q.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.2.transformer_blocks.0.attn1.to_v.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.2.transformer_blocks.0.attn2.to_k.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.2.transformer_blocks.0.attn2.to_out.0.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.2.transformer_blocks.0.attn2.to_q.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.2.transformer_blocks.0.attn2.to_v.weight": [
      384,
      384
    ],
    "ldm.output_blocks.5.2.transformer_blocks.0.ff.net.0.proj.weight": [
      3072,
      384
    ],
    "ldm.output_blocks.5.2.transformer_blocks.0.ff.net.2.weight": [
      384,
      1536
    ],
    "ldm.output_blocks.5.3.conv.weight": [
      384,
      3,
      3,
      384
    ],
    "ldm.output_blocks.6.0.emb_layers.1.weight": [
      256,
      512
    ],
    "ldm.output_blocks.6.0.in_layers.2.weight": [
      256,
      3,
      3,
      640
    ],
    "ldm.output_blocks.6.0.out_layers.3.weight": [
      256,
      3,
      3,
      256
    ],
    "ldm.output_blocks.6.0.skip_connection.weight": [
      256,
      1,
      1,
      640
    ],
    "ldm.output_blocks.6.1.proj_in.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.output_blocks.6.1.proj_out.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.output_blocks.6.1.transformer_blocks.0.attn1.to_k.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.1.transformer_blocks.0.attn1.to_out.0.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.1.transformer_blocks.0.attn1.to_q.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.1.transformer_blocks.0.attn1.to_v.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.1.transformer_blocks.0.attn2.to_k.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.1.transformer_blocks.0.attn2.to_out.0.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.1.transformer_blocks.0.attn2.to_q.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.1.transformer_blocks.0.attn2.to_v.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.1.transformer_blocks.0.ff.net.0.proj.weight": [
      2048,
      256
    ],
    "ldm.output_blocks.6.1.transformer_blocks.0.ff.net.2.weight": [
      256,
      1024
    ],
    "ldm.output_blocks.6.2.proj_in.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.output_blocks.6.2.proj_out.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.output_blocks.6.2.transformer_blocks.0.attn1.to_k.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.2.transformer_blocks.0.attn1.to_out.0.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.2.transformer_blocks.0.attn1.to_q.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.2.transformer_blocks.0.attn1.to_v.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.2.transformer_blocks.0.attn2.to_k.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.2.transformer_blocks.0.attn2.to_out.0.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.2.transformer_blocks.0.attn2.to_q.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.2.transformer_blocks.0.attn2.to_v.weight": [
      256,
      256
    ],
    "ldm.output_blocks.6.2.transformer_blocks.0.ff.net.0.proj.weight": [
      2048,
      256
    ],
    "ldm.output_blocks.6.2.transformer_blocks.0.ff.net.2.weight": [
      256,
      1024
    ],
    "ldm.output_blocks.7.0.emb_layers.1.weight": [
      256,
      512
    ],
    "ldm.output_blocks.7.0.in_layers.2.weight": [
      256,
      3,
      3,
      512
    ],
    "ldm.output_blocks.7.0.out_layers.3.weight": [
      256,
      3,
      3,
      256
    ],
    "ldm.output_blocks.7.0.skip_connection.weight": [
      256,
      1,
      1,
      512
    ],
    "ldm.output_blocks.7.1.proj_in.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.output_blocks.7.1.proj_out.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.output_blocks.7.1.transformer_blocks.0.attn1.to_k.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.1.transformer_blocks.0.attn1.to_out.0.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.1.transformer_blocks.0.attn1.to_q.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.1.transformer_blocks.0.attn1.to_v.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.1.transformer_blocks.0.attn2.to_k.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.1.transformer_blocks.0.attn2.to_out.0.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.1.transformer_blocks.0.attn2.to_q.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.1.transformer_blocks.0.attn2.to_v.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.1.transformer_blocks.0.ff.net.0.proj.weight": [
      2048,
      256
    ],
    "ldm.output_blocks.7.1.transformer_blocks.0.ff.net.2.weight": [
      256,
      1024
    ],
    "ldm.output_blocks.7.2.proj_in.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.output_blocks.7.2.proj_out.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.output_blocks.7.2.transformer_blocks.0.attn1.to_k.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.2.transformer_blocks.0.attn1.to_out.0.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.2.transformer_blocks.0.attn1.to_q.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.2.transformer_blocks.0.attn1.to_v.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.2.transformer_blocks.0.attn2.to_k.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.2.transformer_blocks.0.attn2.to_out.0.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.2.transformer_blocks.0.attn2.to_q.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.2.transformer_blocks.0.attn2.to_v.weight": [
      256,
      256
    ],
    "ldm.output_blocks.7.2.transformer_blocks.0.ff.net.0.proj.weight": [
      2048,
      256
    ],
    "ldm.output_blocks.7.2.transformer_blocks.0.ff.net.2.weight": [
      256,
      1024
    ],
    "ldm.output_blocks.8.0.emb_layers.1.weight": [
      256,
      512
    ],
    "ldm.output_blocks.8.0.in_layers.2.weight": [
      256,
      3,
      3,
      384
    ],
    "ldm.output_blocks.8.0.out_layers.3.weight": [
      256,
      3,
      3,
      256
    ],
    "ldm.output_blocks.8.0.skip_connection.weight": [
      256,
      1,
      1,
      384
    ],
    "ldm.output_blocks.8.1.proj_in.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.output_blocks.8.1.proj_out.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.output_blocks.8.1.transformer_blocks.0.attn1.to_k.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.1.transformer_blocks.0.attn1.to_out.0.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.1.transformer_blocks.0.attn1.to_q.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.1.transformer_blocks.0.attn1.to_v.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.1.transformer_blocks.0.attn2.to_k.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.1.transformer_blocks.0.attn2.to_out.0.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.1.transformer_blocks.0.attn2.to_q.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.1.transformer_blocks.0.attn2.to_v.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.1.transformer_blocks.0.ff.net.0.proj.weight": [
      2048,
      256
    ],
    "ldm.output_blocks.8.1.transformer_blocks.0.ff.net.2.weight": [
      256,
      1024
    ],
    "ldm.output_blocks.8.2.proj_in.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.output_blocks.8.2.proj_out.weight": [
      256,
      1,
      1,
      256
    ],
    "ldm.output_blocks.8.2.transformer_blocks.0.attn1.to_k.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.2.transformer_blocks.0.attn1.to_out.0.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.2.transformer_blocks.0.attn1.to_q.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.2.transformer_blocks.0.attn1.to_v.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.2.transformer_blocks.0.attn2.to_k.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.2.transformer_blocks.0.attn2.to_out.0.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.2.transformer_blocks.0.attn2.to_q.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.2.transformer_blocks.0.attn2.to_v.weight": [
      256,
      256
    ],
    "ldm.output_blocks.8.2.transformer_blocks.0.ff.net.0.proj.weight": [
      2048,
      256
    ],
    "ldm.output_blocks.8.2.transformer_blocks.0.ff.net.2.weight": [
      256,
      1024
    ],
    "ldm.output_blocks.8.3.conv.weight": [
      256,
      3,
      3,
      256
    ],
    "ldm.output_blocks.9.0.emb_layers.1.weight": [
      128,
      512
    ],
    "ldm.output_blocks.9.0.in_layers.2.weight": [
      128,
      3,
      3,
      384
    ],
    "ldm.output_blocks.9.0.out_layers.3.weight": [
      128,
      3,
      3,
      128
    ],
    "ldm.output_blocks.9.0.skip_connection.weight": [
      128,
      1,
      1,
      384
    ],
    "ldm.time_embed.0.weight": [
      512,
      128
    ],
    "ldm.time_embed.2.weight": [
      512,
      512
    ],
    "vae.decoder.conv_out.weight": [
      1,
      3,
      3,
      128
    ],
    "vae.decoder.mid.attn_1.k.weight": [
      1024,
      1,
      1,
      1024
    ],
    "vae.decoder.mid.attn_1.proj_out.weight": [
      1024,
      1,
      1,
      1024
    ],
    "vae.decoder.mid.attn_1.q.weight": [
      1024,
      1,
      1,
      1024
    ],
    "vae.decoder.mid.attn_1.v.weight": [
      1024,
      1,
      1,
      1024
    ],
    "vae.decoder.mid.block_1.conv1.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.decoder.mid.block_1.conv2.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.decoder.mid.block_2.conv1.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.decoder.mid.block_2.conv2.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.decoder.up.0.block.0.conv1.weight": [
      128,
      3,
      3,
      256
    ],
    "vae.decoder.up.0.block.0.conv2.weight": [
      128,
      3,
      3,
      128
    ],
    "vae.decoder.up.0.block.0.nin_shortcut.weight": [
      128,
      1,
      1,
      256
    ],
    "vae.decoder.up.0.block.1.conv1.weight": [
      128,
      3,
      3,
      128
    ],
    "vae.decoder.up.0.block.1.conv2.weight": [
      128,
      3,
      3,
      128
    ],
    "vae.decoder.up.0.block.2.conv1.weight": [
      128,
      3,
      3,
      128
    ],
    "vae.decoder.up.0.block.2.conv2.weight": [
      128,
      3,
      3,
      128
    ],
    "vae.decoder.up.1.block.0.conv1.weight": [
      256,
      3,
      3,
      512
    ],
    "vae.decoder.up.1.block.0.conv2.weight": [
      256,
      3,
      3,
      256
    ],
    "vae.decoder.up.1.block.0.nin_shortcut.weight": [
      256,
      1,
      1,
      512
    ],
    "vae.decoder.up.1.block.1.conv1.weight": [
      256,
      3,
      3,
      256
    ],
    "vae.decoder.up.1.block.1.conv2.weight": [
      256,
      3,
      3,
      256
    ],
    "vae.decoder.up.1.block.2.conv1.weight": [
      256,
      3,
      3,
      256
    ],
    "vae.decoder.up.1.block.2.conv2.weight": [
      256,
      3,
      3,
      256
    ],
    "vae.decoder.up.1.upsample.conv.weight": [
      256,
      3,
      3,
      256
    ],
    "vae.decoder.up.2.block.0.conv1.weight": [
      512,
      3,
      3,
      1024
    ],
    "vae.decoder.up.2.block.0.conv2.weight": [
      512,
      3,
      3,
      512
    ],
    "vae.decoder.up.2.block.0.nin_shortcut.weight": [
      512,
      1,
      1,
      1024
    ],
    "vae.decoder.up.2.block.1.conv1.weight": [
      512,
      3,
      3,
      512
    ],
    "vae.decoder.up.2.block.1.conv2.weight": [
      512,
      3,
      3,
      512
    ],
    "vae.decoder.up.2.block.2.conv1.weight": [
      512,
      3,
      3,
      512
    ],
    "vae.decoder.up.2.block.2.conv2.weight": [
      512,
      3,
      3,
      512
    ],
    "vae.decoder.up.2.upsample.conv.weight": [
      512,
      3,
      3,
      512
    ],
    "vae.decoder.up.3.block.0.conv1.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.decoder.up.3.block.0.conv2.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.decoder.up.3.block.1.conv1.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.decoder.up.3.block.1.conv2.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.decoder.up.3.block.2.conv1.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.decoder.up.3.block.2.conv2.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.decoder.up.3.upsample.conv.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.encoder.conv_out.weight": [
      32,
      3,
      3,
      1024
    ],
    "vae.encoder.down.0.block.0.conv1.weight": [
      128,
      3,
      3,
      128
    ],
    "vae.encoder.down.0.block.0.conv2.weight": [
      128,
      3,
      3,
      128
    ],
    "vae.encoder.down.0.block.1.conv1.weight": [
      128,
      3,
      3,
      128
    ],
    "vae.encoder.down.0.block.1.conv2.weight": [
      128,
      3,
      3,
      128
    ],
    "vae.encoder.down.0.downsample.conv.weight": [
      128,
      3,
      3,
      128
    ],
    "vae.encoder.down.1.block.0.conv1.weight": [
      256,
      3,
      3,
      128
    ],
    "vae.encoder.down.1.block.0.conv2.weight": [
      256,
      3,
      3,
      256
    ],
    "vae.encoder.down.1.block.0.nin_shortcut.weight": [
      256,
      1,
      1,
      128
    ],
    "vae.encoder.down.1.block.1.conv1.weight": [
      256,
      3,
      3,
      256
    ],
    "vae.encoder.down.1.block.1.conv2.weight": [
      256,
      3,
      3,
      256
    ],
    "vae.encoder.down.1.downsample.conv.weight": [
      256,
      3,
      3,
      256
    ],
    "vae.encoder.down.2.block.0.conv1.weight": [
      512,
      3,
      3,
      256
    ],
    "vae.encoder.down.2.block.0.conv2.weight": [
      512,
      3,
      3,
      512
    ],
    "vae.encoder.down.2.block.0.nin_shortcut.weight": [
      512,
      1,
      1,
      256
    ],
    "vae.encoder.down.2.block.1.conv1.weight": [
      512,
      3,
      3,
      512
    ],
    "vae.encoder.down.2.block.1.conv2.weight": [
      512,
      3,
      3,
      512
    ],
    "vae.encoder.down.2.downsample.conv.weight": [
      512,
      3,
      3,
      512
    ],
    "vae.encoder.down.3.block.0.conv1.weight": [
      1024,
      3,
      3,
      512
    ],
    "vae.encoder.down.3.block.0.conv2.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.encoder.down.3.block.0.nin_shortcut.weight": [
      1024,
      1,
      1,
      512
    ],
    "vae.encoder.down.3.block.1.conv1.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.encoder.down.3.block.1.conv2.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.encoder.mid.attn_1.k.weight": [
      1024,
      1,
      1,
      1024
    ],
    "vae.encoder.mid.attn_1.proj_out.weight": [
      1024,
      1,
      1,
      1024
    ],
    "vae.encoder.mid.attn_1.q.weight": [
      1024,
      1,
      1,
      1024
    ],
    "vae.encoder.mid.attn_1.v.weight": [
      1024,
      1,
      1,
      1024
    ],
    "vae.encoder.mid.block_1.conv1.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.encoder.mid.block_1.conv2.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.encoder.mid.block_2.conv1.weight": [
      1024,
      3,
      3,
      1024
    ],
    "vae.encoder.mid.block_2.conv2.weight": [
      1024,
      3,
      3,
      1024
    ],
    "voc.audio_block.downsamples.0.0.weight": [
      96,
      4,
      48
    ],
    "voc.audio_block.downsamples.1.0.weight": [
      192,
      4,
      96
    ],
    "voc.audio_block.downsamples.2.0.weight": [
      384,
      4,
      192
    ],
    "voc.audio_block.downsamples.3.0.weight": [
      768,
      12,
      384
    ],
    "voc.audio_block.downsamples.4.0.weight": [
      1536,
      20,
      768
    ],
    "voc.conv_pre.weight": [
      1536,
      7,
      256
    ],
    "voc.resblocks.0.convs1.0.weight": [
      768,
      3,
      768
    ],
    "voc.resblocks.0.convs1.1.weight": [
      768,
      3,
      768
    ],
    "voc.resblocks.0.convs1.2.weight": [
      768,
      3,
      768
    ],
    "voc.resblocks.0.convs2.0.weight": [
      768,
      3,
      768
    ],
    "voc.resblocks.0.convs2.1.weight": [
      768,
      3,
      768
    ],
    "voc.resblocks.0.convs2.2.weight": [
      768,
      3,
      768
    ],
    "voc.resblocks.1.convs1.0.weight": [
      768,
      7,
      768
    ],
    "voc.resblocks.1.convs1.1.weight": [
      768,
      7,
      768
    ],
    "voc.resblocks.1.convs1.2.weight": [
      768,
      7,
      768
    ],
    "voc.resblocks.1.convs2.0.weight": [
      768,
      7,
      768
    ],
    "voc.resblocks.1.convs2.1.weight": [
      768,
      7,
      768
    ],
    "voc.resblocks.1.convs2.2.weight": [
      768,
      7,
      768
    ],
    "voc.resblocks.2.convs1.0.weight": [
      768,
      11,
      768
    ],
    "voc.resblocks.2.convs1.1.weight": [
      768,
      11,
      768
    ],
    "voc.resblocks.2.convs1.2.weight": [
      768,
      11,
      768
    ],
    "voc.resblocks.2.convs2.0.weight": [
      768,
      11,
      768
    ],
    "voc.resblocks.2.convs2.1.weight": [
      768,
      11,
      768
    ],
    "voc.resblocks.2.convs2.2.weight": [
      768,
      11,
      768
    ],
    "voc.resblocks.3.convs1.0.weight": [
      384,
      3,
      384
    ],
    "voc.resblocks.3.convs1.1.weight": [
      384,
      3,
      384
    ],
    "voc.resblocks.3.convs1.2.weight": [
      384,
      3,
      384
    ],
    "voc.resblocks.3.convs2.0.weight": [
      384,
      3,
      384
    ],
    "voc.resblocks.3.convs2.1.weight": [
      384,
      3,
      384
    ],
    "voc.resblocks.3.convs2.2.weight": [
      384,
      3,
      384
    ],
    "voc.resblocks.4.convs1.0.weight": [
      384,
      7,
      384
    ],
    "voc.resblocks.4.convs1.1.weight": [
      384,
      7,
      384
    ],
    "voc.resblocks.4.convs1.2.weight": [
      384,
      7,
      384
    ],
    "voc.resblocks.4.convs2.0.weight": [
      384,
      7,
      384
    ],
    "voc.resblocks.4.convs2.1.weight": [
      384,
      7,
      384
    ],
    "voc.resblocks.4.convs2.2.weight": [
      384,
      7,
      384
    ],
    "voc.resblocks.5.convs1.0.weight": [
      384,
      11,
      384
    ],
    "voc.resblocks.5.convs1.1.weight": [
      384,
      11,
      384
    ],
    "voc.resblocks.5.convs1.2.weight": [
      384,
      11,
      384
    ],
    "voc.resblocks.5.convs2.0.weight": [
      384,
      11,
      384
    ],
    "voc.resblocks.5.convs2.1.weight": [
      384,
      11,
      384
    ],
    "voc.resblocks.5.convs2.2.weight": [
      384,
      11,
      384
    ],
    "voc.resblocks.6.convs1.0.weight": [
      192,
      3,
      192
    ],
    "voc.resblocks.6.convs1.1.weight": [
      192,
      3,
      192
    ],
    "voc.resblocks.6.convs1.2.weight": [
      192,
      3,
      192
    ],
    "voc.resblocks.6.convs2.0.weight": [
      192,
      3,
      192
    ],
    "voc.resblocks.6.convs2.1.weight": [
      192,
      3,
      192
    ],
    "voc.resblocks.6.convs2.2.weight": [
      192,
      3,
      192
    ],
    "voc.resblocks.7.convs1.0.weight": [
      192,
      7,
      192
    ],
    "voc.resblocks.7.convs1.1.weight": [
      192,
      7,
      192
    ],
    "voc.resblocks.7.convs1.2.weight": [
      192,
      7,
      192
    ],
    "voc.resblocks.7.convs2.0.weight": [
      192,
      7,
      192
    ],
    "voc.resblocks.7.convs2.1.weight": [
      192,
      7,
      192
    ],
    "voc.resblocks.7.convs2.2.weight": [
      192,
      7,
      192
    ],
    "voc.resblocks.8.convs1.0.weight": [
      192,
      11,
      192
    ],
    "voc.resblocks.8.convs1.1.weight": [
      192,
      11,
      192
    ],
    "voc.resblocks.8.convs1.2.weight": [
      192,
      11,
      192
    ],
    "voc.resblocks.8.convs2.0.weight": [
      192,
      11,
      192
    ],
    "voc.resblocks.8.convs2.1.weight": [
      192,
      11,
      192
    ],
    "voc.resblocks.8.convs2.2.weight": [
      192,
      11,
      192
    ],
    "voc.ups.0.0.weight": [
      768,
      20,
      1536
    ],
    "voc.ups.1.0.weight": [
      384,
      12,
      768
    ],
    "voc.ups.2.0.weight": [
      192,
      4,
      384
    ],
    "voc.ups.3.0.weight": [
      96,
      4,
      192
    ],
    "voc.ups.4.0.weight": [
      48,
      4,
      96
    ]
  }
}