jgrusewski 2d704eaa07 fix: backward cuBLASLt uses heuristic algo — AlgoGetIds selects graph-incompatible algo on H100 ...

AlgoGetIds picks algorithms by ID order. On H100 (SM90), the first
valid algo uses split-K/stream-K with internal workspace allocation
that silently produces zero output when replayed via CUDA Graph.
The heuristic selects graph-safe algorithms by design.

Also removes FOXHUNT_NO_GRAPH temp env var and raw DtoH diagnostic probe.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

2026-04-14 00:01:34 +02:00

..

docker

fix(critical): 16-byte align weight pointers for cublasLtMatmul FAST_TF32

2026-04-10 21:38:14 +02:00

k8s

fix: backward cuBLASLt uses heuristic algo — AlgoGetIds selects graph-incompatible algo on H100

2026-04-14 00:01:34 +02:00

live/production

infra: automate terragrunt via Argo CI, clean up kapsule module, harden PAT rotation

2026-03-10 23:25:33 +01:00

modules

infra: VPC gateway in Terraform + DNS deadlock prevention

2026-03-19 00:38:20 +01:00

scripts

fix: tune log levels across workspace — demote noisy warn to debug/trace

2026-03-14 11:35:15 +01:00