linux_job_v3: refresh the HF cache unattended, and honour the label off a PR - #8683
Open
huydhn wants to merge 1 commit into
Open
linux_job_v3: refresh the HF cache unattended, and honour the label off a PR#8683huydhn wants to merge 1 commit into
huydhn wants to merge 1 commit into
Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub. |
meta-cla
Bot
added
the
CLA Signed
huydhn
marked this pull request as draft
August 31, 2026 23:53
huydhn
force-pushed
the
osdc/hf-cache-writable
branch
from
August 31, 2026 23:56
45207f9 to
b71aa4e
Compare
huydhn
changed the title
huydhn
force-pushed
the
osdc/hf-cache-writable
branch
from
September 1, 2026 01:10
b71aa4e to
9133b90
Compare
huydhn
changed the title
huydhn
force-pushed
the
osdc/hf-cache-writable
branch
5 times, most recently
from
September 1, 2026 07:07
c0b5cd1 to
ab6c869
Compare
…ff a PR The refresh label is read from github.event.pull_request.labels, which only exists on a pull_request event, so the ciflow tag run sitting next to a labelled pull request ignores it, writes at the read-only /mnt/hf_cache mount, and dies on any model the seeded cache does not carry: OSError: [Errno 30] Read-only file system: '/mnt/hf_cache/hub/models--...' Two changes, both following pytorch's _linux-test.yml. A ciflow tag names its pull request, so resolve the label off that, the same resolution filter_test_configs.py does. And a schedule or workflow_dispatch run refreshes unconditionally, which is what keeps the shared cache seeded -- otherwise the only way to add a model is to label a pull request, which is also the only way to notice one was needed. Anything else, a plain push included, does not refresh, so the read-only mount stays the default. test-hf-cache-mode already asserted whichever mode it was given; it now also requires that a workflow_dispatch run took the refresh branch, which covers the unattended path. The ciflow tag path has no coverage: test-infra has no ciflow tags to run against. Authored with Claude Code.
huydhn
force-pushed
the
osdc/hf-cache-writable
branch
from
September 1, 2026 07:11
ab6c869 to
b89475e
Compare
huydhn
marked this pull request as ready for review
September 1, 2026 07:12
This was referenced Sep 1, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The refresh label is read from
github.event.pull_request.labels, which only exists on apull_requestevent. So theciflowtag run sitting next to a labelled PR ignores it, writes at the read-only/mnt/hf_cachemount, and dies on any model the seeded cache does not carry:Example: a tag run reporting
HF_CACHE_REFRESH: 0on a labelled PR's own commit.Two changes, both following pytorch's
_linux-test.yml:filter_test_configs.pydoes.scheduleandworkflow_dispatchrefresh unconditionally. This is what keeps the cache seeded; otherwise the only way to add a model is to label a PR, which is also the only way to find out one was needed.Anything else, a plain push included, does not refresh, so the read-only mount stays the default.
Tests
test-hf-cache-modeintest_linux_job_v3.ymlalready asserted whichever mode it was given; it now also requires that aworkflow_dispatchrun took the refresh branch, covering the unattended path.The ciflow tag path has no coverage — test-infra has no ciflow tags to run against. I verified it by hand against pytorch/executorch:
ciflow/trunk/22247(labelled) resolves to refresh,ciflow/trunk/22106(unlabelled) does not.Authored with Claude Code.