Cache load_inline Torch headers in a Modal Volume - #527

Merged

msaroufim merged 8 commits into

mainfrom

codex/modal-pch-volume

Sep 7, 2026

Commits on Sep 7, 2026

Cache PyTorch inline headers on Modal volumes
msaroufim
committed
b16acc3
Warm header profiles before the existing Modal rollout
msaroufim
committed
38b9892
Record fresh T4 PCH benchmarks on current main
msaroufim
committed
6c502bd
Test PCH hits, misses, changed headers and non-inline builds on Modal
msaroufim
committed
3f9cd13
Explain automatic PCH usage for submission authors
msaroufim
committed
3b8fcb8
Benchmark instantiated CUTLASS GEMMs and trim PCH experiment artifacts
msaroufim
committed
8232215
Measure CUTLASS with default and minimal CUDA headers
msaroufim
committed
e226b86
Focus PCH results on load_inline submissions
msaroufim
committed
034fda3