Remove unnecessary cuda sync for better perf · pytorch/executorch@e76fba9

Triggered via pull request February 20, 2026 07:42

Gasoonjia

synchronize #17315

gh/gasoonjia/116/head

Status Success

Total duration 29m 28s

Artifacts 4

cuda-windows.yml

on: pull_request

Matrix: export-model-cuda-windows-artifact

Produced during runtime

Name	Size	Digest
mistralai-Voxtral-Mini-3B-2507-cuda-windows-non-quantized Expired	6.82 GB	`sha256:e0de28728399058bb764dfd6609202b9f2e4e0292c415235ad371e9676562d1d`
mistralai-Voxtral-Mini-3B-2507-cuda-windows-quantized-int4-weight-only Expired	6.15 GB	`sha256:f139b29a947c7f80523e1b2d9ad58907aa0ed1c81cdc7e2ecff87dc1ad99e941`
nvidia-parakeet-tdt-cuda-windows-non-quantized Expired	954 MB	`sha256:0f68f7d3daac5d1334b00d82567e12de4768c1ef1474d7e50c23697b551f98cf`
nvidia-parakeet-tdt-cuda-windows-quantized-int4-weight-only Expired	432 MB	`sha256:673fe3b63f71f2fcd5c79cca8e5f0608df479dea926d3cb29de303a4b6f9e755`