Conversation
…ization eval Optimization-type eval for AMD_OSS issue #4. Agent must fix W4A16 MoE dispatch (model crashes on ROCm) then optimize decode latency for Kimi-K2.5 (1T MoE, TP=8, batch=1, input=8192, output=2048). - Docker image pins sglang at pre-fix commit (before PR #17863) - Base image sglang replaced with pre-fix code to prevent data leak - Binary test harness (crash=0, runs=100), continuous metric via orchestrator - Benchmark uses local model checkout (no HF download needed) - Editable install at /sgl-workspace/sglang/ (base image packaging) - immutable_files protects bench script and harness from replacement Cross-validated by @john: pre-fix 0.0, post-fix 100.0 (24.0ms decode). Reviewed by @alex (no static blockers).
Collaborator
Author
|
Duplicate of #11 |
4 tasks
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
sglang-kimi-w4a16-moe-optimizeeval instance for AMD_OSS issue Add issue-level eval bundles for vLLM #35925 and #33123 #4 (Kimi-K2.5 model support)/root/.cache/huggingface/Kimi-K2.5Cross-validation results
gptq_marlin_repacknot available on ROCm)immutable_filesFiles
Dockerfile: Pre-fix sglang checkout, data leak prevention, fork remotesbench_kimi_w4a16.sh: Fixed benchmark params, agent-configurable backends viabench_config.envtest_harness.py: Binary scoring (crash=0, runs=100)task_description.md: Symptom-only description, no solution hintstask.yaml: Optimization task with decode_median_ms metricmetadata.json: Hard difficulty, 8 GPUs required