 Henning Thieß
|
c4af6cf263
|
Qwen3.5-MoE: support modelopt_fp4 checkpoints that quantize attention (+ load baked FP8 KV scales) (#31220)
|
2026-07-30 14:30:26 -07:00 |
|
|
saatwiknagpal
|
5339450ed4
|
Support SGLANG_SIMULATE_ACC_LEN for DFLASH (#32595)
|
2026-07-30 14:10:06 -07:00 |
|
|
Broduker
|
b61cb5f9de
|
Fix DeepSeek V4 loading with RunAI Model Streamer. (#30240)
|
2026-07-30 23:03:34 +08:00 |
|
|
Xiaoyu Zhang
|
c5bd3d7dce
|
[diffusion][benchmark] Add reproducible request-manifest offline benchmark (#32917)
|
2026-07-30 22:11:18 +08:00 |
|
|
Xiaoyu Zhang
|
7784ac8f91
|
[diffusion][docs] Fix Cosmos3 model sizes (#32916)
|
2026-07-30 22:10:23 +08:00 |
|
|
Xiaoyu Zhang
|
2e9c82b359
|
[Kernel] Remove unreachable AOT headers (#32842)
|
2026-07-30 22:08:57 +08:00 |
|
|
cen121212
|
b78d3999b5
|
【NPU】fix decode MTP + eagle shape error (#32791)
|
2026-07-30 21:34:16 +08:00 |
|
|
Mick
|
db3da62333
|
[diffusion] feat: unify encoder folding and batch data-parallel encoding (#30211)
|
2026-07-30 20:15:22 +08:00 |
|
|
Yi (Vincent) Zhong
|
1f04eaab6a
|
Fix LFM 2 tool parser. (#27614)
|
2026-07-30 10:15:08 +00:00 |
|
|
YC Yen-Ching Tseng
|
fd86795107
|
[AMD] MiniMax-M3: opt-in custom/quick all-reduce on ROCm (#32230)
|
2026-07-30 02:56:07 -07:00 |
|
|
Liangsheng Yin
|
9f56553408
|
[Perf] Fast-path chain-style draft token organization in multi-layer EAGLE (#32887)
|
2026-07-30 02:55:21 -07:00 |
|
|
YC Yen-Ching Tseng
|
04d6fb4d6c
|
[AMD] Minimax-M3 : unblock mxfp8 block convert on gfx950 (#32036)
|
2026-07-30 02:43:27 -07:00 |
|
|
Zheng Wengang
|
4ba7d5ad93
|
[BugFix][EPD] Early-release mooncake GPU embeddings; fix gpu_id via scheduler.ps (#31591)
|
2026-07-30 17:36:55 +08:00 |
|
|
Liangsheng YinandKaixi
|
6ab3231b97
|
[Perf] Skip the target-verify tree mask fill when the backend never reads it (#32886)
Co-authored-by: Kaixi <kaiximatteoc@nvidia.com>
|
2026-07-30 02:32:38 -07:00 |
|
|
kangwangamd
|
4b52758c76
|
[AMD] Skip test_update_weights_from_disk on ROCm pending reload fix (#31924) (#31925)
|
2026-07-30 02:32:09 -07:00 |
|
|
Ding Yinandyinding
|
fc007e1f00
|
Add SM90 FP8 MegaMoE support for DeepSeek-V4 (#29016)
Co-authored-by: yinding <yinding@bytedance.com>
|
2026-07-30 01:48:10 -07:00 |
|
|
|
f46d5f25b4
|
[4/N][CP] Support interleave strategy for cp v2 (#30482)
Co-authored-by: Xinyuan Tong <xinyuantong.cs@gmail.com>
Co-authored-by: Baizhou Zhang <sobereddiezhang@gmail.com>
|
2026-07-30 01:45:32 -07:00 |
|
|
Liangsheng Yin
|
c192145830
|
[Kernel] Fuse KV-cache writes for asymmetric K/V (head_dim != v_head_dim) (#32813)
|
2026-07-30 00:26:10 -07:00 |
|
|
Liangsheng Yin
|
2625fdfe6b
|
[Fix] Count multi-layer draft-extend replays in the fwd-occupancy device timer (#32867)
|
2026-07-30 00:21:34 -07:00 |
|
|
Yanbin Jiang
|
92b3a51ba6
|
[LoRA] Fix Marlin MoE kernel import (#32884)
|
2026-07-30 15:17:29 +08:00 |
|
|
Ho-Ren (Jack) ChuangandClaude Fable 5
|
e4a40a71f8
|
[DSA] Q8KV8 FP8 Sparse Prefill on GLM-5.2 & DeepSeek-V3.2: Q8-Path & Shared-Path Optimizations (#31888)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
|
2026-07-30 15:15:11 +08:00 |
|
|
Jinwoo Jeong
|
4f51dad1da
|
fix: prevent ReqTimeStats from being dropped during IPC serialization (#31339)
|
2026-07-30 14:42:38 +08:00 |
|
|
Shangming Cai
|
f6ff5e8bb0
|
[PD] Handle abort requests in PP mode (#32797)
|
2026-07-30 14:39:44 +08:00 |
|
|
wenxuewuhdandronnie_zheng
|
36afd442c7
|
[DLLM] vectorized joint/low-confidence decoding and skip redundant attn init (#21094)
Co-authored-by: ronnie_zheng <zl19940307@163.com>
|
2026-07-30 09:13:14 +03:00 |
|
|
Ke Bao
|
07a087bf45
|
Fix Inkling tool-call parsing recovery, content handling, and streaming (#32861)
|
2026-07-30 14:11:09 +08:00 |
|
|
Liangsheng Yin
|
f4e0ac382e
|
[misc] Remove unused multi_layer_draft_forward_cg module (#32881)
|
2026-07-29 21:17:16 -07:00 |
|
|
Jimmy Shong
|
ed361ae7f0
|
Fix attention backends for models with per-layer head counts (num_attention_heads_per_layer) (#32625)
|
2026-07-29 20:03:00 -07:00 |
|
|
Mick
|
22faf9fef8
|
embedding: centralize capabilities and complete OpenAI compatibility (#32481)
|
2026-07-30 10:28:52 +08:00 |
|
|
Liangsheng Yin
|
313a518bee
|
[Spec] Emit step trace span for multi-layer draft-extend graph replays (#32850)
|
2026-07-29 19:22:58 -07:00 |
|
|
Sam ShleiferandClaude Fable 5
|
62dfaaa0e0
|
[Nemotron] Fix decode track-save reading the stale tail of the CUDA-graph track buffer (#32555)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
|
2026-07-29 19:03:44 -07:00 |
|
|
Xuanyi LiandR0CKSTAR
|
8fbf960980
|
[MLX] Size request capacity by attention DP (#32115)
Co-authored-by: R0CKSTAR <yeahdongcn@gmail.com>
|
2026-07-29 18:18:22 -07:00 |
|
|
Xiaoyu Zhang
|
1d9c292547
|
[Kernel] Add inventory guards and clean benchmark layout (#32788)
|
2026-07-30 09:03:24 +08:00 |
|
|
Mohammad Miadh Angkad
|
a55e1764a2
|
Enable GPT-OSS FlashInfer MXFP4 on SM120 (#32668)
|
2026-07-30 00:04:23 +00:00 |
|
|
Liangsheng Yin
|
e5c46ff07d
|
[Fix] Route asymmetric-KV models to fa4 on SM100 and pin MiMoV2 FP8 MoE to flashinfer_trtllm (#32818)
|
2026-07-29 16:37:43 -07:00 |
|
|
cctry
|
3c1717d9b6
|
Follow up on #30157 post-merge review (#32672)
|
2026-07-29 15:03:59 -07:00 |
|
|
Sam (Kesen Li)
|
8fc54d46ef
|
Fix MoE reduce-scatterv eligibility check (#32663)
|
2026-07-29 14:58:55 -07:00 |
|
|
Willow LopezandJiminator
|
ffd4705baa
|
fix(reasoning): honor Poolside template thinking defaults (#32540)
Co-authored-by: Jiminator <69131491+Jiminator@users.noreply.github.com>
|
2026-07-29 21:02:19 +00:00 |
|
|
Sam Shleifer
|
d0e69d3881
|
[feat] Optional base64 encoding for the flat prompt top logprob arrays (#31960)
|
2026-07-29 12:15:56 -07:00 |
|
|
|
e4f7f7b380
|
fix(qwen3.5): restrict MoE weights to local PP layers (#32022)
Co-authored-by: chunjiang.hc <chunjiang.hc@alibaba-inc.com>
Co-authored-by: Yihao Wang <42559837+AgainstEntropy@users.noreply.github.com>
|
2026-07-29 12:01:09 -07:00 |
|
|
ziang663andChao Shi
|
eefb434d17
|
[PD+PP] Honor PP consensus for bootstrap and prealloc (#31869)
Co-authored-by: Chao Shi <chao.shi@alibaba-inc.com>
|
2026-07-30 01:19:17 +08:00 |
|
|
Yoray Zack
|
62d0f81f16
|
[2/N] elastic-ep: Enable EPLB after scale-up (#30553)
|
2026-07-30 01:06:56 +08:00 |
|
|
Piotr Mazurek
|
d19999b755
|
[LFM2] Wire Lfm2MoeForCausalLM into the LFM2 serving override tables (#30780)
|
2026-07-29 17:02:14 +00:00 |
|
|
Hert4andMohammad Miadh Angkad
|
f69af7b7ad
|
[Bugfix] compressed-tensors: mixed-precision checkpoints silently load unquantized (#32736)
Co-authored-by: Mohammad Miadh Angkad <176301910+mmangkad@users.noreply.github.com>
|
2026-07-29 09:55:33 -07:00 |
|
|
YAMY
|
fddfc1fb5e
|
[GDN] Support FlashInfer GDN prefill with extra-buffer radix cache (#29735)
|
2026-07-30 00:47:35 +08:00 |
|
|
huangtingweiandalphabetc1
|
50029f05a3
|
[HiCache] Merge HiCache event checks to reduce decode overhead (#30511)
Co-authored-by: alphabetc1 <47200617+alphabetc1@users.noreply.github.com>
|
2026-07-29 23:48:48 +08:00 |
|
|
Nicolas Castet
|
e1f2f9d1fa
|
Disable extra NCCL CUDA event synchronization with symm mem (#27089)
|
2026-07-29 22:39:36 +08:00 |
|
|
Wu Jiangming
|
1c6a0e91e1
|
fix mqa preshuffle layout issue for deepseek v4 (#31563)
|
2026-07-29 07:36:48 -07:00 |
|
|
Mick
|
22151edca1
|
[diffusion] optimization: accelerate CUDA video output finalization (#32784)
|
2026-07-29 22:04:20 +08:00 |
|
|
Xiaoyu Zhang
|
0ebbe43dbb
|
fix(diffusion): size VSA top-k from padded blocks (#32695)
|
2026-07-29 21:58:41 +08:00 |
|
|
Xiaoyu Zhang
|
4f5b50c576
|
perf(diffusion): decode Wan VAE in BF16 (#32697)
|
2026-07-29 21:57:50 +08:00 |
|