Testcases 261015 coverage test by chenyang08056032 · Pull Request #1355 · Ascend/sglang

chenyang08056032 · 2026-10-22T10:11:34Z

Motivation

Modifications

Accuracy Tests

Speed Tests and Profiling

Checklist

Format your code according to the Format code with pre-commit.
Add unit tests according to the Run and add unit tests.
Update documentation according to Write documentations.
Provide accuracy and speed benchmark results according to Test the accuracy and Benchmark the speed.
Follow the SGLang code style guidance.

Review and Merge Process

Ping Merge Oncalls to start the process. See the PR Merge Process.
Get approvals from CODEOWNERS and other reviewers.
Trigger CI tests with comments or contact authorized users to do so.
- Common commands include /tag-and-rerun-ci, /tag-run-ci-label, /rerun-failed-ci
After green CI and required approvals, ask Merge Oncalls or people with Write permission to merge the PR.

CI States

Latest PR Test (Base): ❌ Run #39761248714
Latest PR Test (Extra): ❌ Run #39761248715
Latest PR Test (AMD ROCm 10): ❌ Run #39761248625

…project#45232)

…-project#44393) Co-authored-by: Mohammad Angkad <mohammad.angkad@radixark.ai>

…project#44663) Co-authored-by: hnyls2002 <lsyincs@gmail.com>

…project#45234)

…ix reuse (sgl-project#39592) Signed-off-by: 0z5a <dezhen.lu@student.uni-tuebingen.de> Co-authored-by: Richard Wang <wangrichard08@gmail.com>

…int32 overflow at 37,451 tokens) (sgl-project#43438)

…oject#33007)

…g dump (sgl-project#44681) Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: hnyls2002 <lsyincs@gmail.com>

…ject#45246)

…gl-project#45247)

…KV cache and QSA indexer (sgl-project#45245) Co-authored-by: zijiexia <37504505+zijiexia@users.noreply.github.com>

…project#44901)

…ts only (sgl-project#44904)

…gl-project#44912)

…project#44913)

…es and a shared kit (sgl-project#45236)

…/muse_spark (sgl-project#45112) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

…multimodal input (sgl-project#40899) Co-authored-by: apex-mochen <265981728+apex-mochen@users.noreply.github.com> Co-authored-by: Po-Han Huang (NVIDIA) <53919306+nvpohanh@users.noreply.github.com> Co-authored-by: apex-mochen <2756823972@qq.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: Po-Han Huang <pohanh@nvidia.com>

…roject#40535) Co-authored-by: sunguozhong <sun_gz2772@163.com>

…gl-project#44514)

…ct#41233) Co-authored-by: Po-Han Huang (NVIDIA) <53919306+nvpohanh@users.noreply.github.com> Co-authored-by: Xinyuan Tong <115166877+JustinTong0323@users.noreply.github.com> Co-authored-by: Po-Han Huang <pohanh@nvidia.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

…riant checks are off (sgl-project#45276)

… for DeepGEMM (sgl-project#43556) Signed-off-by: Zihua Wu <13583761+lucifer1004@users.noreply.github.com> Co-authored-by: number-eleven11 <1635219023@qq.com>

…sgl-project#40908) Co-authored-by: ronnie_zheng <zl19940307@163.com>

…l-project#45212) Co-authored-by: Mohammad Angkad <mohammad.angkad@radixark.ai> Co-authored-by: Mohammad Angkad <mohammad.angkad@radixark.ai>

… tail metadata (sgl-project#45275)

)

…tly (sgl-project#45252)

… for MI300 runners (sgl-project#34241)

…r the default prefill context limit and the DSpark draft ring (sgl-project#45507)

…gl-project#45725)

…#41894 (sgl-project#45775)

…all (sgl-project#45778) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

…-coverage-test # Conflicts: # .github/workflows/_npu-pr-test-stage.yml # .github/workflows/_npu-single-node-test-stage.yml # .github/workflows/nightly-test-npu.yml # test/registered/npu/accuracy/qwen3_6_27b/test_npu_qwen3_6_27b_w8a8_1p_in3k5_out1k5_50ms_gpqa.py

Fridge003 and others added 30 commits October 19, 2026 13:40

[CI] tests: Re-enable muse spark B300 PD coverage (sgl-project#44975)

642a671

[CI] radix-cache: Drop redundant unified radix cache unit tests (sgl-…

b20cb36

…project#45232)

[CI] python: Move the remaining CUDA build paths off Python 3.10 (sgl…

7656ea0

…-project#44393) Co-authored-by: Mohammad Angkad <mohammad.angkad@radixark.ai>

[Fix] metrics: record HTTP request metrics in tokenizer workers (sgl-…

81ac153

…project#44663) Co-authored-by: hnyls2002 <lsyincs@gmail.com>

[Perf] dspark: Make the prefill SWA valid-from update sync-free (sgl-…

fdc2dd0

…project#45234)

[Fix] dLLM: Preserve prompt tails and literal mask tokens across pref…

80ea5ee

…ix reuse (sgl-project#39592) Signed-off-by: 0z5a <dezhen.lu@student.uni-tuebingen.de> Co-authored-by: Richard Wang <wangrichard08@gmail.com>

[Fix] attn_residual: int64 program ids in the score/combine kernels (…

49766a7

…int32 overflow at 37,451 tokens) (sgl-project#43438)

[Fix] attention: Honor selected FlashAttention implementation (sgl-pr…

05d7f7a

…oject#33007)

[Fix] symm-mem: Name the window registration in the scheduler watchdo…

ab258b6

…g dump (sgl-project#44681) Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: hnyls2002 <lsyincs@gmail.com>

[Doc] tpu: Introduce SGLang-TorchTPU in README and TPU guide (sgl-pro…

9a94940

…ject#45246)

[Perf] deepseek-v4.1: Reset request-window tags without a host sync (s…

1238edd

…gl-project#45247)

[Doc] cookbook: Default Qwen3.8-Flash-Next datacenter recipes to FP8 …

28448a8

…KV cache and QSA indexer (sgl-project#45245) Co-authored-by: zijiexia <37504505+zijiexia@users.noreply.github.com>

[Fix] moe: Count routed_scaling_factor once when top-k folds it (sgl-…

3a059de

…project#44901)

[Fix] moe: Renormalize per-rank shared-slot routing over routed exper…

48ea486

…ts only (sgl-project#44904)

[Refactor] moe: Split the two meanings of num_fused_shared_experts (s…

964ab2b

…gl-project#44912)

[Refactor] moe: Scope shared-expert configuration to each model (sgl-…

6b90e57

…project#44913)

[CI] radix-cache: Split unified radix cache unit tests into topic fil…

ab580b2

…es and a shared kit (sgl-project#45236)

[Refactor] kernels: Regroup operators by computation and dissolve ops…

693d333

…/muse_spark (sgl-project#45112) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

[Perf] [NPU] mem_cache: Add optional ScatterPa KV cache writes (sgl-p…

ebc71e6

…roject#40535) Co-authored-by: sunguozhong <sun_gz2772@163.com>

[Fix] runtime: Version FlashInfer caches for default-tactic tuning. (s…

1897575

…gl-project#44514)

[Perf] deepseek-v4.1: Skip the request-window history check when inva…

ce329c8

…riant checks are off (sgl-project#45276)

[Fix] [SM120] quantization: Requantize block-FP8 MoE weights to UE8M0…

67d9ee1

… for DeepGEMM (sgl-project#43556) Signed-off-by: Zihua Wu <13583761+lucifer1004@users.noreply.github.com> Co-authored-by: number-eleven11 <1635219023@qq.com>

[Test] constrained: Cover JSON whitespace limits (sgl-project#43981)

2c00b5d

[Fix] [NPU] LoRA: Update for follow-up with changes in sgl-kernel-npu (…

55b2aef

…sgl-project#40908) Co-authored-by: ronnie_zheng <zl19940307@163.com>

[Fix] mamba: Skip zero-width conv pools in the MTP verify scatter (sg…

af85671

…l-project#45212) Co-authored-by: Mohammad Angkad <mohammad.angkad@radixark.ai> Co-authored-by: Mohammad Angkad <mohammad.angkad@radixark.ai>

[Perf] deepseek-v4.1: Drop host syncs from encoder-replay folding and…

5e630fa

… tail metadata (sgl-project#45275)

[Feat] [Intel] hicache: Enable L1<->L2 support on XPU (sgl-project#32503

8683f8a

)

[Fix] [NVIDIA] docker: Fix CUDA 13.4 container by updating torch nigh…

ec1fdc2

…tly (sgl-project#45252)

kangwangamd and others added 7 commits October 22, 2026 01:08

[Fix] [AMD] test: Recalibrate MiniMax-M2.7 GSM8K threshold 0.93->0.91…

1c967f5

… for MI300 runners (sgl-project#34241)

[Test] [AMD] deepseek-v4.1: Fix the HIP breakable-graph unit tests fo…

0a87f16

…r the default prefill context limit and the DSpark draft ring (sgl-project#45507)

[Fix] [Intel] diffusion: Restore copy offload for FLUX.2-dev nightly (s…

5c2cb60

…gl-project#45725)

[Fix] diffusion: Restore the kernel registry import after sgl-project…

3f6c313

…#41894 (sgl-project#45775)

[Fix] diffusion: Drop stale mismatch_msg from FLUX 3 video VAE gate c…

d0a7565

…all (sgl-project#45778) Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

Align qwen3_6_27b w8a8 gpqa accuracy threshold with sgl-project/main

ec989ff

chenyang08056032 requested review from iforgetmyname and ping1jing2 as code owners October 22, 2026 10:11

chenyang08056032 closed this Oct 22, 2026

Navigation Menu

Repository navigation

Testcases 261015 coverage test - #1355

Testcases 261015 coverage test#1355
chenyang08056032 wants to merge 1519 commits into
Ascend:testcases-261015from
chenyang08056032:testcases-261015-coverage-test

chenyang08056032 commented Oct 22, 2026 •

edited by github-actions Bot

Loading

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

20 participants

Conversation

chenyang08056032 commented Oct 22, 2026 • edited by github-actions Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Motivation

Modifications

Accuracy Tests

Speed Tests and Profiling

Checklist

Review and Merge Process

CI States

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

20 participants

chenyang08056032 commented Oct 22, 2026 •

edited by github-actions Bot

Loading