FaST-ER-LMM – Gauthier Brach

Skip to content

~/gbrach

misc

A GPU port of FaST-LMM. Around 1,000–10,000× faster in benchmarks.

Written in PyTorch.

Side-by-side progress of the same scan on one and two GPUs — 1 GPU / 2 GPUs

Runtime per phenotype and speedup over FaST-LMM on one and two GPUs, from 500 to 100,000 strains — Runtime / speedup

Close agreement between FaST-ER-LMM and FaST-LMM across 26,414 yeast phenotypes and 95,804 variants — Same results