Releases: google-deepmind/alphagenome_research
Releases · google-deepmind/alphagenome_research
v0.2.0
Added
- Non-zero mean computation to fine-tuning notebook.
- Gene LFC loss for GenomeTracksHead.
- Align alternate predictions for splice junctions when making indel
predictions.
Changed
- Fix splice-site and splice-site usage scoring to use the correct gene mask
extractor. - Only consider splice sites that are within a gene body, and overlaps with
the provided variant. - Pass requested outputs to the model prediction call, significantly reducing
roofline memory consumption by eliding redundant computation. - Pad gene masks in gene scoring. This reduces the number of jit
recompilations for different numbers of genes.
v0.1.0
Added
- Add code and notebook example for fine-tuning AlphaGenome.
- Better support for loading pre-trained model with different heads and
metadata.
Changed
- Various fixes to better align predictions with the AlphaGenome API (e.g.
metadata column names). - Fixed variant extraction to correctly insert variants on the negative
strand.