| name: dimamba | |
| type: dimamba | |
| hidden_size: 256 | |
| cond_dim: 128 | |
| length: ${model.length} # Same length as diffusion model | |
| n_blocks: 8 | |
| scale_by_sigma: True | |
| dropout: 0.1 | |
| tie_word_embeddings: False | |
| bidirectional: True, | |
| bidirectional_strategy: add | |
| bidirectional_weight_tie: True | |
| num_classes: ${data.num_classes} | |
| pooling: mean | |