Cuda 3 device binding - #87

Draft

max-models wants to merge 3 commits into

cuda-3-device-binding

Commits on Sep 30, 2026

CUDA step 1: run feectools with the CuPy backend

max-models
and
claude
committed
0a55cbd
CUDA step 2: MPI with device buffers

max-models
and
claude
committed
ccb39f8
CUDA step 3: bind each MPI rank to its own GPU

max-models
and
claude
committed
3a11ad1