 YC Yen-Ching Tseng
|
3b5aae278e
|
Fix EagleDraftExtendInput missing kv_indptr crash with triton/DP attention (#28221)
|
2026-06-17 14:54:33 -07:00 |
|
|
Baizhou ZhangandShijin
|
4b817f5d7f
|
Upgrade fa3 hash (#28394)
Co-authored-by: Shijin <dovis.zhang02@gmail.com>
|
2026-06-17 13:32:11 -07:00 |
|
|
Qiaolin Yu
|
753aa89a83
|
[spec decoding] fix mrope_positions in draft extend (#28464)
|
2026-06-17 13:25:31 -07:00 |
|
|
Lijuan Tang
|
0a28a929dc
|
[MLX] Add Metal profiling hooks to server profiler (#28122)
|
2026-06-17 13:06:20 -07:00 |
|
|
Khoa Pham
|
3c4130c741
|
[Kimi K2.5] Fix eagle3 aux capture for tp>1 when AR fusion is enabled (#28343)
|
2026-06-17 12:59:11 -07:00 |
|
|
Thomas WangandBingxu Chen
|
f5b041622b
|
[AMD] Fix deepseek-v4 mtp accept length issue (#28520)
Co-authored-by: Bingxu Chen <bingxche@amd.com>
|
2026-06-17 11:13:55 -07:00 |
|
|
kousakawangandkousakawang
|
8aaca72c21
|
[FIX]Fix Step3-VL multi-image embedding and local patch splitting (#24970)
Co-authored-by: kousakawang <wanghanpei@bytedance.com>
|
2026-06-17 10:31:32 -07:00 |
|
|
Junlin Wu
|
873196f7fa
|
♻️ [llm][npu][quant] Delegate MXFP8 dense scheme to kernel and use torch.ops.npu (#28505)
|
2026-06-17 10:18:26 -07:00 |
|
|
Mick
|
735a256f98
|
[diffusion] feat: use LocalAttention for mistral3 encoder (#28176)
|
2026-06-17 21:18:41 +08:00 |
|
|
Aleksi VesantoandMick
|
dad890fff1
|
[diffusion] perf: shard text when using sp in flux.1/2 (#27066)
Co-authored-by: Mick <mickjagger19@icloud.com>
|
2026-06-17 21:17:39 +08:00 |
|
|
Ziang Li
|
3fb65ebabd
|
[RL] Fix FlashInfer TRTLLM MXFP8 dense weight layout (#28459)
|
2026-06-17 10:33:35 +00:00 |
|
|
lmyybh
|
2f1390fcb1
|
fix: preserve divisible FP8 block K configs on CUDA (#27553)
|
2026-06-17 10:30:56 +00:00 |
|
|
YC Yen-Ching Tseng
|
9b8c41171a
|
[AMD] Fall back to layer_first layout for kernel write-back on ROCm (#28473)
|
2026-06-17 03:01:43 -07:00 |
|
|
Thomas Wang
|
21a95333d4
|
[AMD] Add transpose_scale arg for o_proj to fix GLM accuracy issue (#27798)
|
2026-06-17 01:01:02 -07:00 |
|
|
Liangsheng Yin
|
f86e9b48e8
|
[Perf] Make spec-decode penalty H2D non-blocking and share decode cumulate path (#28500)
|
2026-06-17 00:59:58 -07:00 |
|
|
Ryan Zzzandzhujunyu
|
8fd1694dd2
|
Deepseek v4: support mixed dtype compression states (#27277)
Co-authored-by: zhujunyu <zhujunyu.666@bytedance.com>
|
2026-06-17 00:56:41 -07:00 |
|
|
Peng Xingchen
|
d0e974f40b
|
[NPU] Use use_dsa to dispatch Ascend DSA attention (#28436)
|
2026-06-17 15:47:12 +08:00 |
|
|
|
b54f8432ad
|
Batch EAGLE draft/draft-extend replay memcpys via grouped foreach copy (#28465)
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Qiaolin Yu <liin1211@outlook.com>
|
2026-06-17 00:34:21 -07:00 |
|
|
Liangsheng Yin
|
3bc618485a
|
[Perf] Make latest_output_ids H2D non-blocking in prepare_for_decode (#28491)
|
2026-06-16 23:46:35 -07:00 |
|
|
Baizhou Zhang
|
27291118b9
|
Upgrade sgl-deep-gemm to 0.1.3 (#28402)
|
2026-06-16 23:06:36 -07:00 |
|
|
Oxana Korzh
|
c01f62e341
|
[bugfix] guard NVIDIA SM-capability checks with is_cuda() for AMD/ROCm (#28486)
|
2026-06-16 22:51:42 -07:00 |
|
|
Mick
|
0f5e14e1d9
|
[diffusion] fix: use Megatron-style tp for native encoders and dits (#28318)
|
2026-06-17 13:07:44 +08:00 |
|
|
Rohit Kumar Singhandgithub-actions[bot]
|
9371062ef3
|
Fix deep seek ocr2 image processing (#27884)
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
|
2026-06-17 12:59:18 +08:00 |
|
|
Chengze Fan
|
66ac385f52
|
fix(moe): MoRI EP init_mori_op missing BF16 dispatch branch (#28469)
Signed-off-by: Chengze Fan <fancz2002@gmail.com>
|
2026-06-16 21:02:06 -07:00 |
|
|
Brayden ZhongandBrayden Zhong
|
b8a73bfba0
|
Call Flashinfer mm_fp8 for per-tensor FP8 GEMMs on SM100 (#28333)
Co-authored-by: Brayden Zhong <brayden@radixark.ai>
|
2026-06-16 20:50:03 -07:00 |
|
|
Yufeng HeandYufeng He
|
6c8fdb5b62
|
[diffusion] fix: fix PicklingError with --backend diffusers on non-T2I models (#21472)
Co-authored-by: Yufeng He <40085740+universeplayer@users.noreply.github.com>
|
2026-06-17 11:14:25 +08:00 |
|
|
Zilin Zhu
|
f06e2d3d1f
|
Support asymmetric compressed-tensors MoE (#27690)
|
2026-06-16 19:53:33 -07:00 |
|
|
zhaozx-cnandgithub-actions[bot]
|
224b1dc775
|
[NPU]Replace ascend vision attn operator (#25768)
Signed-off-by: zhaozx-cn <zhaozx2116@163.com>
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
|
2026-06-17 10:33:37 +08:00 |
|
|
Qeeweew
|
0ae4740bd1
|
fix: add missing clamp_limit for CompressedTensorsWNA16MoE (#27328)
|
2026-06-16 19:29:57 -07:00 |
|
|
Trevor MorrisandClaude Opus 4.7
|
9c53853ea3
|
Use pack topk ids triton kernel for flashinfer_trtllm_routed (#25702)
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
2026-06-16 19:27:14 -07:00 |
|
|
Yanbin Jiang
|
093908d4c0
|
[LoRA] Fix chunked SGMV (csgmv) CUDA graph segment replay (#28371)
|
2026-06-16 19:19:07 -07:00 |
|
|
|
37ef295c78
|
[AMD] Feat/dp moe reduce scatter (#28216)
Co-authored-by: wunhuang <wunhuang@amd.com>
Co-authored-by: Wang, FangYuan <39615225+At1a8@users.noreply.github.com>
|
2026-06-16 18:12:39 -07:00 |
|
|
jasonjk-park
|
d86a7e7018
|
Custom spec algorithm can handle server args (#28162)
|
2026-06-16 17:13:52 -07:00 |
|
|
ashwini rathi
|
4f9b12c5dd
|
[XPU] Guard tvm_ffi import in dsv4 compress modules under TYPE_CHECKING (#28426)
|
2026-06-16 15:52:12 -07:00 |
|
|
Hanming Lu
|
a10eee3d80
|
[Tokenizer] Fix abort racing server crash when large amount of aborts (#28341)
|
2026-06-16 14:49:35 -07:00 |
|
|
huangtingwei
|
9b4432fe18
|
[HiCache]Asymmetric pool support direct backend (#28446)
|
2026-06-16 13:17:57 -07:00 |
|
|
Jason Mancuso
|
c0a6c3ce66
|
Fix circular import when sglang.srt.model_executor.runner_backend is imported first (#28002)
|
2026-06-16 12:04:34 -07:00 |
|
|
|
799584e173
|
fix: get_processor fails when --tokenizer-path lacks model config.json (#25643)
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: ronnie_zheng <zl19940307@163.com>
|
2026-06-16 20:25:05 +03:00 |
|
|
feliang-gitandxutizhou
|
92b42c8d8a
|
LPLB: linear-programming load balancer for MoE expert parallelism (#24515)
Co-authored-by: xutizhou <xutingz@nvidia.com>
|
2026-06-16 10:19:42 -07:00 |
|
|
Zhangheng
|
78b6a4fabf
|
[UnifiedTree]: Clean up some unused dead code. (#28389)
|
2026-06-16 21:53:35 +08:00 |
|
|
|
265202cda2
|
fix(openai): validate assistant tool call arguments before chat template (#28035)
Co-authored-by: Xinyuan Tong <115166877+JustinTong0323@users.noreply.github.com>
Co-authored-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
|
2026-06-16 12:21:23 +00:00 |
|
|
Yinghai LuandLianmin Zheng
|
fcca4611fa
|
[CAR] Let custom allreduce support VMM based allocation (#27593)
Co-authored-by: Lianmin Zheng <lianminzheng@gmail.com>
|
2026-06-16 04:49:05 -07:00 |
|
|
Wang, FangYuanandThomas Wang
|
a362ba9da3
|
[AMD] Feat: Add prefill context parallel support for deepseek v4 unified kv attention (#27928)
Co-authored-by: Thomas Wang <thomawan@amd.com>
|
2026-06-16 02:00:51 -07:00 |
|
|
jacky.cheng
|
149fabcca7
|
[AMD] Fuse sigmoid + mul into single Triton kernel for shared expert gating (#27636)
|
2026-06-16 01:16:58 -07:00 |
|
|
YC Yen-Ching Tseng
|
102392df5b
|
[AMD][Fix] Skip EPLB topk remap when global server args are unset (#28404)
|
2026-06-16 01:13:22 -07:00 |
|
|
Xiaoyu Zhang
|
c5b9106c1a
|
[perf] Use default torch compile mode for Wan2.2 T2V A14B (#28304)
|
2026-06-16 15:45:00 +08:00 |
|
|
Baizhou Zhang
|
77f327cb6e
|
[2/n] [CP] Add context parallel strategy abstractions (#27313)
|
2026-06-16 00:20:04 -07:00 |
|
|
Zhangheng
|
6c908b3a3a
|
[UnifiedTree]: Replace anonymous tuples with NamedTuples in UnifiedRadixCache (#28375)
|
2026-06-16 14:54:04 +08:00 |
|
|
Junlin Wuandronnie_zheng
|
2a8ea70059
|
✨ [llm][npu][quant] Add W8A8 MXFP8 quantization support for Qwen3 Dense on Ascend NPU (#22352)
Co-authored-by: ronnie_zheng <zl19940307@163.com>
|
2026-06-16 09:45:18 +03:00 |
|
|
Michael
|
72d962be88
|
[AMD] Fix jit-kernel-unit-test-amd: activation.cuh ROCm build + per_token CUDA-only (R165) (#27947)
|
2026-06-15 23:44:26 -07:00 |
|