|
--- |
|
license: mit |
|
tags: |
|
- genomics |
|
- single-cell |
|
- biology |
|
--- |
|
After the installation of FlashAttention2, the model can be loaded using: |
|
```python |
|
from scooby.modeling import Scooby |
|
flooby_neurips = Scooby.from_pretrained( |
|
'johahi/neurips-scooby-flash', |
|
cell_emb_dim=14, |
|
embedding_dim=1920, |
|
n_tracks=3, |
|
return_center_bins_only=True, |
|
disable_cache=False, |
|
use_transform_borzoi_emb=False, |
|
) |
|
``` |
|
Since the transform_borzoi_emb is not used, the models performance is slightly (0.52 vs 0.54 across cell types) worse than neurips-scooby. |
