| name: unet | |
| type: unet | |
| ch: 128 | |
| num_res_blocks: 2 | |
| num_scales: 4 | |
| ch_mult: [1, 2, 2, 2] | |
| input_channels: 3 | |
| output_channels: -1 # determined by vocab_size | |
| scale_count_to_put_attn: 1 # at 16 res | |
| data_min_max: [0, 255] # No need currently | |
| dropout: 0.1 | |
| skip_rescale: True | |
| time_conditioning: True # Whether to add in time embeddings | |
| time_scale_factor: 1000 | |
| time_embed_dim: ${.ch} | |
| fix_logistic: False | |
| size: ${data.size} | |
| cond_dim: ${.ch} | |
| length: ${data.length} |