jax
Cap xdist workers per GPU to avoid oversubscription
#805

Open

Commits

[ROCm] Add TheRock 7.13 and TheRock latest to CI test matrices

mminutoli committed 117 days ago
[ROCm] Embed TheRock-aware wheel RUNPATH at link time

mminutoli committed 117 days ago
[ROCm] Put TheRock SDK bin on PATH via so rocminfo/rocm-smi resolve in /opt/rocm-less images

mminutoli committed 117 days ago
[ROCm] Update CI runner labels to amd-do-linux.jax.gpu.gfx950.{1,4,8} (#807)

mminutoli committed 117 days ago
[ROCm] Put TheRock SDK bin on PATH via so rocminfo/rocm-smi resolve in /opt/rocm-less images

psanal35 committed 117 days ago
Resolve ROCm workflow conflict markers

psanal35 committed 117 days ago
Isolate MIOpen cache per xdist worker to avoid lock contention

psanal35 committed 117 days ago
Cap xdist workers per GPU to avoid oversubscription

psanal35 committed 117 days ago
Temp: Sweep per-GPU worker count to find the best value

psanal35 committed 117 days ago