Install mkl for PyTorch and libjpeg/libpng for TorchVision #8248
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/8248
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit c5dcd2d with merge base b1d76c9 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
huydhn
changed the title
Let me look more into that conda issue. This is rather unexpected, I have done a rebase, so the failure is legit. |
Just FYI, the linking error is from this hack https://github.com/pytorch/executorch/blob/main/.ci/docker/common/install_conda.sh#L46-L58 |
|
The run time is around 1h20m now https://github.com/pytorch/executorch/actions/runs/13229090225/job/36927181503 v.s 2h30m https://github.com/pytorch/executorch/actions/runs/13213708877/job/36890417988 |
|
eval_llama-mmlu seems to have gone from 1.9h to under 9 minutes: https://hud.pytorch.org/hud/pytorch/executorch/main/1?per_page=50&name_filter=mmlu&mergeLF=true |
#8173 raised these timeouts. Now that #8248 has landed to fix #8180, we should be able to lower them again. (I'm sending this early so I don't forget; double-check llava-runner running time) ghstack-source-id: cb4c1691907b8bb46a504a2d8cbc00d12b1ef4a4 ghstack-comment-id: 2648474106 Pull Request resolved: #8339
Fixes #8180
I should have caught this earlier that we didn't have mkl here, so PyTorch performance on x86 CPU suffered.