| version: 1.0 | |
| system: "cross" | |
| diffwrap: | |
| cls_embedding: | |
| content_dim: 768 | |
| content_hidden: 256 | |
| unet: | |
| sample_size: [1, 1] | |
| in_channels: 256 | |
| out_channels: 256 | |
| layers_per_block: 2 | |
| block_out_channels: [256] | |
| down_block_types: | |
| [ | |
| "CrossAttnDownBlock2D", | |
| ] | |
| up_block_types: | |
| [ | |
| "CrossAttnUpBlock2D", | |
| ] | |
| attention_head_dim: 32 | |
| cross_attention_dim: 768 | |