 inkcherryandDuyi-Wang
|
a3bf25dc62
|
[AMD] Clamp MORI intranode grid GPUs (#39763)
Co-authored-by: Duyi-Wang <duyi.wang@amd.com>
|
2026-09-16 02:13:59 -07:00 |
|
|
+4
|
7d5696b3a1
|
dsv4.1: communication kernels and wrappers (#39653)
Co-authored-by: Cheng Wan <54331508+ch-wan@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: DarkSharpness <2040703891@qq.com>
Co-authored-by: DarkSharpness <76582120+DarkSharpness@users.noreply.github.com>
Co-authored-by: Ke Bao <ispobaoke@gmail.com>
Co-authored-by: Khoa Pham <khoa.pham@radixark.ai>
Co-authored-by: Xiaoyu Zhang <1182563586@qq.com>
Co-authored-by: Xiaoyu Zhang <xiaoyu.zhang@radixark.ai>
Co-authored-by: Yuhao Yang <47235274+yhyang201@users.noreply.github.com>
Co-authored-by: Yuwei An <ayw.sirius19@gmail.com>
Co-authored-by: Zhichen Zeng <zczeng@uw.edu>
Co-authored-by: Ziyi Xu <ziyi.xu@radixark.ai>
|
2026-09-16 02:09:40 -07:00 |
|
|
Bingxu Chen
|
5a0c1e21e9
|
[AMD] Preserve deterministic inference when Lean Attention is enabled (#37740)
|
2026-09-16 02:09:08 -07:00 |
|
|
Bingxu Chen
|
2cbfaefbf9
|
[AMD] Avoid the FP8 wo_a path when the weight is BF16 (#38453)
Signed-off-by: bingxche <bingxche@amd.com>
|
2026-09-16 01:57:15 -07:00 |
|
|
Jensenandronnie_zheng
|
5212797880
|
[NPU] fix npu docker workspace directory (#39732)
Signed-off-by: xiedeyantu <czjourney@163.com>
Co-authored-by: ronnie_zheng <zl19940307@163.com>
|
2026-09-16 11:51:38 +03:00 |
|
|
Liangsheng YinandBBuf
|
fc5a979f21
|
[misc] Merge FlashInfer autotune caches across spec workers, pad MXFP4 TP shards, drop dead ngram attrs (#39678)
Co-authored-by: BBuf <1182563586@qq.com>
|
2026-09-16 01:50:29 -07:00 |
|
|
vstone-w
|
0e528dc9ff
|
bugfix:fix unifiedcache c128 radix cache management (#39426)
|
2026-09-16 16:47:42 +08:00 |
|
|
|
8baeded6f3
|
[Router] Pin to the prefix owner when the whole fleet is queueing (--saturation-queue-floor) (#39169)
Co-authored-by: Kangyan Zhou <kangyan.zhou@radixark.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
2026-09-16 16:16:24 +08:00 |
|
|
Alison Shao
|
ad28b91fae
|
ci: fix always-failing coverage job, add by-GPU-count view (#39697)
|
2026-09-16 01:13:08 -07:00 |
|
|
Void
|
9c8d4641ff
|
[Fix] Fix GLM5 mHC PP forward (#39720)
|
2026-09-16 00:32:53 -07:00 |
|
|
|
54f0d72adf
|
[Router] Fleet-wide sampling contract 2/3: enforce and inject per request (#39001)
Co-authored-by: Kangyan Zhou <kangyan.zhou@radixark.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
2026-09-16 00:29:33 -07:00 |
|
|
|
7835f1de9a
|
[Router] Add the shutdown-drain configuration surface (#39015)
Co-authored-by: Kangyan Zhou <kangyan.zhou@radixark.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
2026-09-16 00:26:50 -07:00 |
|
|
Jensen
|
5beb2fd552
|
[NPU] Avoid device synchronization in Ascend sampling (#39404)
|
2026-09-16 10:15:16 +03:00 |
|
|
|
e2d56bbbfc
|
[Router] Bound the e2e worker memory budget so prefill graph capture stops OOMing (#39713)
Co-authored-by: Kangyan Zhou <kangyan.zhou@radixark.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
2026-09-16 00:13:09 -07:00 |
|
|
|
d634320e48
|
[Router] Count open HTTP exchanges until their response body finishes (#39014)
Co-authored-by: Kangyan Zhou <kangyan.zhou@radixark.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
2026-09-16 00:04:38 -07:00 |
|
|
+4
|
faaff1eca8
|
dsv4.1: Top-k kernels and candidate selection (#39648)
Co-authored-by: BBuf <1182563586@qq.com>
Co-authored-by: Cheng Wan <54331508+ch-wan@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: DarkSharpness <2040703891@qq.com>
Co-authored-by: DarkSharpness <76582120+DarkSharpness@users.noreply.github.com>
Co-authored-by: Ke Bao <ispobaoke@gmail.com>
Co-authored-by: Khoa Pham <khoa.pham@radixark.ai>
Co-authored-by: Xiaoyu Zhang <1182563586@qq.com>
Co-authored-by: Xiaoyu Zhang <xiaoyu.zhang@radixark.ai>
Co-authored-by: Yuhao Yang <47235274+yhyang201@users.noreply.github.com>
Co-authored-by: Yuwei An <ayw.sirius19@gmail.com>
Co-authored-by: Zhichen Zeng <zczeng@uw.edu>
Co-authored-by: Ziyi Xu <ziyi.xu@radixark.ai>
|
2026-09-15 23:54:35 -07:00 |
|
|
Hank Han
|
a64be2e430
|
[Kernel] Add H20 block-FP8 MoE configs for GLM-5.3-Flash EP4/EP8 (#38913)
|
2026-09-15 23:42:18 -07:00 |
|
|
Shuwen WangandZhangheng
|
6b33338242
|
[HiCache] Forward prefix metadata to v2 storage calls (#39567)
Co-authored-by: Zhangheng <hzh0425@apache.org>
|
2026-09-16 14:05:42 +08:00 |
|
|
Liangsheng Yin
|
91f691c490
|
dsv4.1: standalone kernels and Python wrappers (#39646)
|
2026-09-15 22:31:18 -07:00 |
|
|
|
afde31a2f5
|
[router] Speak cleartext h2c on both edges: serve it inbound, forward it outbound (#39006)
Co-authored-by: Kangyan Zhou <kangyan.zhou@radixark.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
2026-09-15 22:30:31 -07:00 |
|
|
|
c1f5b4736a
|
[Router] Add --worker-queue-limit: stop sending cache-affinity traffic to a queueing worker (#39168)
Co-authored-by: Kangyan Zhou <kangyan.zhou@radixark.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
2026-09-16 13:24:48 +08:00 |
|
|
Shangming CaiandClaude Opus 5
|
3f8eb35ead
|
[PD] Do not admit intake-rejected requests to a PD handoff (#38935)
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
2026-09-16 12:44:49 +08:00 |
|
|
Dayananda VandClaude Opus 5
|
100e1cd0d9
|
[Fix] Spawn, don't fork, the benchmark server process (#34712)
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
|
2026-09-16 12:44:09 +08:00 |
|
|
|
5f17e3a75f
|
[Router] Fleet-wide sampling contract 1/3: the config surface (#39000)
Co-authored-by: Kangyan Zhou <kangyan.zhou@radixark.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
2026-09-15 21:39:06 -07:00 |
|
|
Xinguo ZhuandAKatydid
|
4678536df6
|
[CPU] Add fp8_per_tensor_scaled_mm_cpu kernel (#32618)
Co-authored-by: AKatydid <xinguojoe@gmail.com>
|
2026-09-16 12:31:05 +08:00 |
|
|
Thomas Wang
|
f60652a43e
|
[AMD] Update deepseek-v4 PDI and cache policy setting for agentic workload (#39702)
|
2026-09-15 21:16:30 -07:00 |
|
|
Xiaoyu Zhang
|
444b29c932
|
[Diffusion] Clean up obsolete worker plumbing, dead helpers, and tests (#39293)
|
2026-09-16 12:11:45 +08:00 |
|
|
Xinyi Song
|
954bb6804a
|
[AMD][bugfix] Fix DSV4 MTP crash (#39547)
|
2026-09-15 21:10:23 -07:00 |
|
|
Xiaoyu Zhang
|
537ac52c5b
|
[SRT] Clean up no-op compiler pass, dead helpers, and migration tests (#39295)
|
2026-09-16 12:04:21 +08:00 |
|
|
Zhang, Jiejing
|
6331e43081
|
[AMD] Prefer HIP Top-K for GLM-5.x on ROCm (#39631)
|
2026-09-15 20:55:57 -07:00 |
|
|
jacky.chengandChangLiu0709
|
a9bb4d7d45
|
[AMD] Align Qwen3.5 MI355X HiCache cookbook with kernel / page_first (#39572)
Co-authored-by: ChangLiu0709 <cliu1004@amd.com>
|
2026-09-16 11:22:28 +08:00 |
|
|
chx96642264
|
5f6dd44edc
|
[NPU] [bugfix] Fix undefined require_attn_tp_allgather (#39561)
|
2026-09-16 11:02:40 +08:00 |
|
|
Mohammad Miadh AngkadandMohammad Angkad
|
935cbf24ec
|
Accept MXFP8 dispatch in FlashInfer A2A TRT-LLM MoE (#39613)
Co-authored-by: Mohammad Angkad <mohammad.angkad@radixark.ai>
|
2026-09-15 19:23:15 -07:00 |
|
|
Rahul Vijayaraghavan
|
d4ad368ed9
|
[Intel XPU] Enable fused_moe_triton tuning on XPU and add tuned DeepSeek-OCR-2 configs (#28723)
|
2026-09-16 10:20:22 +08:00 |
|
|
Cao E
|
f11cd8ab0e
|
[XPU] Use torch scaled_mm for XPU block FP8 linear (#35605)
|
2026-09-16 10:10:31 +08:00 |
|
|
Yuxingwang-intelandMa Mingfei
|
2cb51f5d22
|
[CPU] [Diffusion] Add fused scale-shift and norm kernels for CPU (#33452)
Co-authored-by: Ma Mingfei <mingfei.ma@intel.com>
|
2026-09-16 10:01:29 +08:00 |
|
|
Zhang, Jiejingandfanxingran
|
f920be4b09
|
[AMD] GLM-5.2 NextN: cast draft fused MoE to per-channel FP8 (#39155)
Co-authored-by: fanxingran <xingran.fan@amd.com>
|
2026-09-15 18:56:48 -07:00 |
|
|
|
7eedd57ab0
|
[ROCm] Fix EAGLE spec-decode verify silently sampling greedy on HIP (#37134)
Co-authored-by: JohnQinAMD <yanyuan.qin@amd.com>
Co-authored-by: Jiejing Zhang <jiejing.zhang@amd.com>
|
2026-09-15 18:49:49 -07:00 |
|
|
Liangsheng Yin
|
2f7d04da5d
|
dsv4.1: Rust extension modules for image preprocessing, KV pool names, and PD bootstrap (#39677)
|
2026-09-15 18:11:48 -07:00 |
|
|
Jinyan Yi
|
dbd7281b06
|
fix(npu): fix hybrid KV transfer with PP prefill in PD disaggregation (#38402)
|
2026-09-16 09:10:17 +08:00 |
|
|
Zhiqiang Xie
|
4e9e407d37
|
[HiCache] Label radix-cache metrics per rank and split the "shrunk" prefetch reason (#39280)
|
2026-09-15 17:54:04 -07:00 |
|
|
Siyuan ChenandYuwei An
|
c9a8fba991
|
[DSV4][BCG] Optimize the heavy memory use of C4 Indexer when BCG is enabled (#36534)
Co-authored-by: Yuwei An <ayw.sirius19@gmail.com>
|
2026-09-15 17:23:11 -07:00 |
|
|
|
bbef93e22a
|
[Router] Add a KV storage-tier coverage row to the Grafana dashboard (4/4) (#39111)
Co-authored-by: Kangyan Zhou <kangyan.zhou@radixark.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
2026-09-15 17:09:11 -07:00 |
|
|
|
63845a1bb2
|
[router] Resolve a wire protocol per worker at registration (#39004)
Co-authored-by: Kangyan Zhou <kangyan.zhou@radixark.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
2026-09-15 16:54:37 -07:00 |
|
|
Liangsheng Yin
|
0dabef3d30
|
[misc] Fix tool-call index, graph padded-row count, and prefill-graph input_embeds refresh (#39574)
|
2026-09-15 16:37:22 -07:00 |
|
|
|
fb91baedab
|
[Router] Real-GPU e2e coverage for storage-tier-aware cache routing (3/4) (#39110)
Co-authored-by: Kangyan Zhou <kangyan.zhou@radixark.ai>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
|
2026-09-15 15:48:51 -07:00 |
|
|
Lianmin Zheng
|
b803cfa0c4
|
Add external multimodal processors to the Rust frontend (#39329)
|
2026-09-15 15:38:47 -07:00 |
|
|
|
2929a39927
|
Use a shared byte budget for unified hybrid-SWA memory (#36729)
Co-authored-by: yhzhuang <yhzhuang@fb.com>
Co-authored-by: Cheng Wan <cheng.wan@radixark.ai>
|
2026-09-15 15:27:00 -07:00 |
|
|
|
4da5599e93
|
Fix first-token metadata and reused attention-layer indexing (#39328)
Co-authored-by: Jinghui Zhang <jinghui@meta.com>
Co-authored-by: Lucia Fang <116399278+luccafong@users.noreply.github.com>
|
2026-09-15 15:22:26 -07:00 |
|
|
|
08b1922405
|
fix(gemma4): set lm_head_is_tied for Gemma4UnifiedForConditionalGeneration (#35809)
Co-authored-by: kpham-sgl <khoa.pham@radixark.ai>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-09-15 15:16:39 -07:00 |
|