temp pr for test - #1771
Draft
shepark wants to merge 5 commits into
Draft
Conversation
…r upstream fused-kernel inlining (#50400) vllm-project#1754 Signed-off-by: Iryna Boiko <iboiko@habana.ai>
Every HPU run (offline LLM, server, or pytest) printed a segfault block at
process teardown, right after "request processing complete; starting resource
teardown", once per EngineCore:
WARNING [interface.py] Current platform hpu does not have 'empty_cache' attribute.
!!!!!!! Segfault encountered !!!!!!!
... at::accelerator::emptyHostCache()
... torch::accelerator::initModule(...)::{lambda()vllm-project#5}
... Py_RunMain / Py_BytesMain / _start
Root cause: vLLM PR #51107 swapped cleanup_dist_env_and_memory() from
torch._C._host_emptyCache() to torch.accelerator.empty_host_cache(). On
torch>=2.9 (docker has 2.12.0a0) the attribute exists, so vLLM's
`except AttributeError` guard no longer catches it; the call dispatches to
at::accelerator::emptyHostCache(), which has no HPU host-cache hook and
segfaults. patches.py already patched empty_cache (PR vllm-project#1430) and the now-dead
torch._C._host_emptyCache, but not the live torch.accelerator.empty_host_cache.
Fix: add an HPU-safe no-op _hpu_accelerator_empty_host_cache() shim (sibling of
_hpu_accelerator_empty_cache) and install it in apply(). Harmless-but-noisy:
the crash fires after all work completes, so results/exit codes were never
affected; this just removes the crash from logs/CI.
Tested (container jha_pytorch_124-482-new, torch 2.12.0a0, gaudi3):
- After apply(): torch.accelerator.empty_host_cache resolves to the no-op shim.
- Same minimal LLM.generate() repro, patched vs. stashed-unpatched control:
unpatched: 2 "Segfault encountered / emptyHostCache()" blocks at teardown
patched: 0 segfault lines, clean teardown, exit 0
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Signed-off-by: Jimin Ha <jimin.ha@intel.com>
…-bias overflow- vllm-project#1749 Signed-off-by: Iryna Boiko <iboiko@habana.ai>
shepark
had a problem deploying
to
pre-merge-approval
August 28, 2026 21:00 — with 
GitHub Actions
Error
GitHub Actions
Error
shepark
had a problem deploying
to
pre-merge-approval
August 28, 2026 23:39 — with
GitHub Actions
Error
GitHub Actions
Error
shepark
force-pushed
the
shepark/for_ci_28
branch
from
August 28, 2026 23:42
8da05c4 to
22246e7
Compare
shepark
had a problem deploying
to
pre-merge-approval
August 28, 2026 23:42 — with
GitHub Actions
Error
GitHub Actions
Error
…ash after upstream fused-kernel inlining (#50400) vllm-project#1754" This reverts commit 9f40e6f.
shepark
had a problem deploying
to
pre-merge-approval
August 28, 2026 23:42 — with
GitHub Actions
Error
GitHub Actions
Error
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
4 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.
No description provided.