GPU CI #334
GPU CI #334
Conversation
meta-cla
bot
added
the
CLA Signed
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #334 +/- ##
=======================================
Coverage ? 64.09%
=======================================
Files ? 79
Lines ? 7760
Branches ? 0
=======================================
Hits ? 4974
Misses ? 2786
Partials ? 0 ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
|
hey |
|
| - name: Download and install vLLM and its dependencies | ||
| # TODO: this honestly could not be hackier if I tried | ||
| run: | | ||
| python -m pip install -r .github/packaging/vllm_reqs.txt |
|
OK at least in the run https://github.com/meta-pytorch/forge/actions/runs/18421419301?pr=334 all unit tests but 3 pass. Two look like pretty normal failures, the only reasonably suspicious one is the linker-related error in |
| # TODO: this honestly could not be hackier if I tried | ||
| run: | | ||
| python -m pip install -r .github/packaging/vllm_reqs.txt | ||
| python -m pip install vllm==0.10.1.dev0+g6d8d0a24c.d20251009.cu129 --no-cache-dir --index-url https://download.pytorch.org/whl/preview/forge |
There was a problem hiding this comment.
if we were to put these into folders by day and have a latest, then could we just do something like this?
python -m pip install vllm --no-cache-dir --index-url https://download.pytorch.org/whl/preview/forge/latest
There was a problem hiding this comment.
Ideally yes but need to check on the ability to add new subdirectories from devinfra's point of view. cc @atalman on the best approach here
| # TODO: this honestly could not be hackier if I tried | ||
| run: | | ||
| python -m pip install -r .github/packaging/vllm_reqs.txt | ||
| python -m pip install vllm==0.10.1.dev0+g6d8d0a24c.d20251009.cu129 --no-cache-dir --index-url https://download.pytorch.org/whl/preview/forge |
There was a problem hiding this comment.
Could these dependencies go in the pyproject toml?
There was a problem hiding this comment.
leaving cleanup as a follow-up. most important thing here is to get this working imo. after that i will open some other smaller refactor PRs (e.g. @allenwang28 was talking about consolidation with the install script as well).
5 participants
this PR is the culmination of a long and sad story. but we can now run forge tests on GPU runners