Repository navigation
fix(tools): drop unread launcher exports and chassis GPU pins - #100
Merged
Merged
Conversation
This was referenced Oct 16, 2026
VLLM_RDNA_FORCE_FP16, VLLM_ROCM_MOE_PREFILL, and VLLM_ROCM_SKIP_LIVE_TAIL_HASH have no readers in vllm/ or csrc/. rdna_kill_stale sleeps only after a matched pid is signaled. EXL3 debug stays off. GPU ids stay caller-overridable. Co-authored-by: Cursor Agent <cursoragent@cursor.com>
cursor
Bot
force-pushed
the
cursor/rdna-launcher-subtraction-a57e
branch
2 times, most recently
from
October 16, 2026 15:24
17e07f1 to
6cd5b6b
Compare
BlivionIaG
marked this pull request as ready for review
October 16, 2026 15:25
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Purpose
Subtraction-only cleanup of the
serve_rdna.shlauncher. No kernel or numeric path changes.VLLM_RDNA_FORCE_FP16,VLLM_ROCM_MOE_PREFILL, andVLLM_ROCM_SKIP_LIVE_TAIL_HASHfrom the launcher, recipes, andPRINTlist.VLLM_GDN_HIP_PREFILLstays (read invllm/model_executor/layers/mamba/gdn/qwen_gdn_linear_attn.py).rdna_kill_stalestill filters withpgrepplusVLLM_CACHE_ROOT, and sleeps 8 / 3 only after TERM / KILL is actually delivered.VLLM_EXL3_DEBUG=1from27b-exl3.env(diagnostics stay off; the print key remains).full.env2,3,27b-exl3.env4,5,6,7,unit.shHIP 0/1 refusal).HIP_VISIBLE_DEVICES/GPU/GPUIDSstill override.Not a duplicate: no open PR removes these exports. #87 is
tools/rdna2_028deletions (untouched). #99 only inserts two rows neartuning.md~102; this deletes the unread-export note at the old lines 207-209. #50 isAGENTS.md/README.md(untouched). Historical docs,tools/rdna2/serve_v620_*.sh, andtools/rdna2_028still mention the vars and are left alone.No GPU required. Optional follow-up: one EXL3 boot log to confirm
VLLM_EXL3_DEBUGstays off. No model eval: nothing that changes output, accuracy, or serving numerics.AI assistance was used. A human submitter needs to review every line.
Test Plan
Test Result
rgonvllm/andcsrc/: no matches.bash -n: all edited scripts and recipes ok.Empty
pgrep, stubbedsleep/kill,set -e(finished under a 2s cap, noSLEEPline):PRINT=1checks:RECIPE=full HIP_VISIBLE_DEVICES=2,3prints2,3;RECIPE=27b-exl3falls back to0,1,2,3and does not exportVLLM_EXL3_DEBUG;HIP_VISIBLE_DEVICES=4,5,6,7still overrides.