Repository navigation
Yield in busy loop to fix the high latency issue we are seeing in DI. - #151
Merged
Merged
Conversation
Signed-off-by: Fenghui Zhang <fhzhang@google.com>
yarongmu-google
approved these changes
Jul 9, 2025
QiliangCui2023
pushed a commit
that referenced
this pull request
Apr 27, 2026
Build #151: abstract methods create_weights and get_fused_moe_quant_config not implemented on VllmNvfp4MoEMethod. Add full create_weights registering NVFP4 MoE weight params (packed uint8 weights, E4M3 block scales, FP32 global scales). Signed-off-by: Qiliang Cui <derrhein@gmail.com>
wenxindongwork
added a commit
that referenced
this pull request
Oct 29, 2026
…n engines on failure - Bump .buildkite/vllm_lkg.version 2fb320d8 -> 4541fcdd (vLLM main as of 2026-10-29 17:07 UTC). The fp8 MoE `per_act_token_quant` and continue-decode scheduler incompatibilities reported from tpu-vllm-integration build #151 were already fixed on main by #3924; today's integration build #157 (main + vLLM HEAD) passes the JAX unit tests and all Runai Model Streamer correctness tests, so this moves the pin forward to the current head. - tests/e2e/test_runai_model_streamer_loader.py: wrap generate() in try/finally so the engine core is always shut down. Previously a failure in the first LLM left its EngineCore process alive, and the second LLM in the same test then died with EngineDeadError, masking the real error. Signed-off-by: wenxindongwork <wenxindong@google.com>
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
The issue is that the busy loop is hogging CPUs. Adding a sleep here will release the cpu for other threads.
Tests
Manual test.
Checklist
Before submitting this PR, please make sure (put X in square brackets):