| input_channels: 768 | |
| output_channels: 768 | |
| encode_channels: 768 | |
| decode_channels: 768 | |
| code_dim: 768 | |
| codebook_num: 1 | |
| codebook_size: 1024 | |
| bias: true | |
| enc_ratios: [ 1, 1 ] | |
| dec_ratios: [ 1, 1 ] | |
| enc_strides: [ 1, 1 ] # no downsampling | |
| dec_strides: [ 1, 1 ] | |
| enc_kernel_size: 3 | |
| dec_kernel_size: 3 | |
| enc_block_dilations: [ 1, 1 ] | |
| enc_block_kernel_size: 3 | |
| dec_block_dilations: [ 1, 1 ] | |
| dec_block_kernel_size: 3 |