 Jonny Kongandjonnykong
|
4d78d59e51
|
[MoE] Make simulated expert routing support DP>1, and fuse into one triton kernel (#29718)
Co-authored-by: jonnykong <jonnykong@fb.com>
|
2026-08-28 14:22:45 -07:00 |
|
|
Liangsheng Yin
|
f65b2b2b15
|
[Fix] Lazy-import aiter in DSv4 paged_decode to unbreak CPU CI (#36914)
|
2026-08-28 14:16:53 -07:00 |
|
|
karverma-amd
|
2a96ebf648
|
[AMD][DSV4] perf: retune decode split-K heuristic for MI355X (#36094)
|
2026-08-28 13:48:41 -07:00 |
|
|
weireweireandweireweire
|
3254f9b47c
|
[Blackwell] Reserve SMs for DeepGEMM MegaMoE grid barriers (#36657)
Co-authored-by: weireweire <20922698+weireweire@users.noreply.github.com>
|
2026-08-28 13:37:36 -07:00 |
|
|
Mohammad Miadh Angkad
|
70088aa5db
|
Fix KeyError on batch requests whose state is freed before it is read (#36638)
|
2026-08-28 12:57:07 -07:00 |
|
|
Zhiqiang Xie
|
f611e0c073
|
[HiCache] Fence load-back behind the forward stream (#36738)
|
2026-08-28 11:12:51 -07:00 |
|
|
Zhiqiang Xie
|
c9bba091f8
|
[HiCache] Key storage prefetch by the request namespace (#36382)
|
2026-08-28 11:12:10 -07:00 |
|
|
Cheng WanandClaude Opus 5
|
ef20fab38a
|
config: the forwarding slots go; the dispatcher calls the family directly (#36792)
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
|
2026-08-28 10:26:29 -07:00 |
|
|
Cheng WanandClaude Opus 5
|
7bc3204117
|
config: three cache and pool readers take the bags (#36791)
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
|
2026-08-28 10:21:32 -07:00 |
|
|
Cheng WanandClaude Opus 5
|
43c63a22ff
|
config: the derived parallel widths are computed from the leaves (#36790)
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
|
2026-08-28 10:18:46 -07:00 |
|
|
Cheng WanandClaude Opus 5
|
c2928e86d7
|
config: the resolution pipeline moves out of the record (#36789)
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
|
2026-08-28 10:17:24 -07:00 |
|
|
pllimax
|
726665e08e
|
[CI] Read subprocess stdout on a background thread to avoid EOF deadlock (#36673)
|
2026-08-29 01:11:02 +08:00 |
|
|
Chao Shi
|
8e04f66a70
|
HiCache: avoid unnecessary all-reduce in check_prefetch_progress (#36425)
|
2026-08-29 00:52:53 +08:00 |
|
|
Alison Shao
|
23cb11ae01
|
Fix KV cache pool sized far too small when weight-loading memory is still referenced (#36583)
|
2026-08-28 09:47:23 -07:00 |
|
|
匡胤鑫
|
b644771e07
|
[Fix] Route the Mooncake MoE A2A backend through Kimi K3's EP-A2A / SP-MoE fast path (#36862)
|
2026-08-28 21:46:35 +08:00 |
|
|
Артем СавкинandXiaoyu Zhang
|
ecbadf0b4b
|
[NPU] [Diffusion] support distributed inference pipeline for GLM-Image (#31320)
Co-authored-by: Xiaoyu Zhang <1182563586@qq.com>
|
2026-08-28 15:39:05 +03:00 |
|
|
Mick
|
803b4fb31c
|
[diffusion] refactor: scope model-specific API parameters (#35613)
|
2026-08-28 19:08:30 +08:00 |
|
|
ming_wang
|
d56706459c
|
bugfix for index_fill_ on NPU (#36759)
|
2026-08-28 17:41:49 +08:00 |
|
|
Xinguo ZhuandMa Mingfei
|
74df026877
|
fix(cpu): skip GPU JIT MoE top-k on CPU (#35677)
Co-authored-by: Ma Mingfei <mingfei.ma@intel.com>
|
2026-08-28 02:31:26 -07:00 |
|
|
K Chiranjeevi
|
9cee0a31d1
|
[XPU] Lazily import tvm_ffi-dependent all_reduce kernel in minimax_m2 (#35290)
Signed-off-by: Chiranjeevi Koppula <ckoppula@habana.ai>
|
2026-08-28 02:18:03 -07:00 |
|
|
McZyWu
|
69a49fede8
|
fix(kimi-k3): preserve dense ModelSlim MLA weights (#36603)
|
2026-08-28 02:15:52 -07:00 |
|
|
Zheng Wengang
|
1e6d041f78
|
[BugFix][VLM] keep Qwen3-VL MoE inference deepstack order (#34690)
|
2026-08-28 17:02:16 +08:00 |
|
|
Zhiqiang Xie
|
3785b2d20f
|
[HiCache] Reject load-back specs that claim nodes pinned by an in-flight load-back (#35931)
|
2026-08-28 01:59:06 -07:00 |
|
|
Xiaoyu Zhang
|
eebb99c049
|
[diffusion][kernel] avoid 4D scale-shift autotuning (#36521)
|
2026-08-28 16:58:19 +08:00 |
|
|
Xiaoyu Zhang
|
45424d8434
|
[diffusion][kernel] support transposed residual-gate add (#36504)
|
2026-08-28 16:54:37 +08:00 |
|
|
Zhiqiang Xie
|
1061e34785
|
[HiCache] Retry L3 storage prefetch after a missed attempt (#36227)
|
2026-08-28 01:35:21 -07:00 |
|
|
Zhiqiang Xie
|
3f1031d697
|
[HiCache] Heal the storage existence cache on a hybrid prefetch discard (#36386)
|
2026-08-28 01:33:40 -07:00 |
|
|
Yikai Zhangandamdpilot-upstream-sync
|
d3b972cbf0
|
fix(lora): build the MoE LoRA align JIT kernel on ROCm (#36379)
Co-authored-by: amdpilot-upstream-sync <amdpilot-upstream-sync@users.noreply.github.com>
|
2026-08-28 01:18:42 -07:00 |
|
|
Bingxu Chen
|
0c7d017dbb
|
[AMD][Fix] Qwen3.5: make empty-batch guard tuple-aware on fused AR+quant path (#35341)
|
2026-08-28 01:11:53 -07:00 |
|
|
|
acc918b3ec
|
[AMD] Qwen3.5 ASM FMHA chunked-prefill context attention (#36758)
Co-authored-by: Zijie Chen <300606707+zijiecode@users.noreply.github.com>
Co-authored-by: jacky.cheng <yichiche@amd.com>
|
2026-08-27 23:37:45 -07:00 |
|
|
|
9fc60a8afd
|
[PD] Fix MORI-IO ABORT bootstrap message handling (#29133)
Co-authored-by: billishyahao <bill.he@amd.com>
Co-authored-by: TianDi101 <tiandi950321@163.com>
Co-authored-by: Duyi-Wang <duyi.wang@amd.com>
Co-authored-by: Shangming Cai <csmthu@gmail.com>
|
2026-08-27 23:02:35 -07:00 |
|
|
YAMY
|
137fab2f48
|
fix(deps): pin compressed-tensors to 0.18.0 (#36794)
|
2026-08-27 22:52:47 -07:00 |
|
|
Thomas Wang
|
baf09f3954
|
[AMD] Enable deepseek-v4 topk_transform v2 kernel (#36684)
|
2026-08-27 22:36:11 -07:00 |
|
|
wufann
|
cce0b53466
|
[AMD] Increase gfx950 DSA model indexer topk_transform kernel occupancy (#35628)
|
2026-08-27 22:20:52 -07:00 |
|
|
Xinyuan Tong
|
0665102ce5
|
[Fix] Resolve tool argument types through top-level anyOf/oneOf/allOf (#36626)
|
2026-08-27 21:50:30 -07:00 |
|
|
Yihao WangandClaude Fable 5
|
7088f21922
|
[diffusion] fix: make tail_attn_meta cuda-graph capturable (#36658)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
|
2026-08-28 12:22:07 +08:00 |
|
|
Kevin MiandClaude Fable 5
|
ad5a105a4a
|
[Diffusion] Fix the five unit tests failing on main (#36726)
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
|
2026-08-27 21:18:32 -07:00 |
|
|
|
2380121e9b
|
[HiCache] Stop populating host-pool mmaps twice (-13% allocation time) (#36705)
Co-authored-by: DellCurry <30748980+DellCurry@users.noreply.github.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
|
2026-08-28 11:54:11 +08:00 |
|
|
karverma-amd
|
2b209711d8
|
[AMD][DSV4] perf: MXFP8 MoRI dispatch to match the w4a8 MoE input format (#36119)
|
2026-08-27 20:45:40 -07:00 |
|
|
karverma-amd
|
aa0a0aa3c3
|
[AMD][DSV4] perf: bound the MoRI receive buffer during decode (#36130)
|
2026-08-27 20:27:52 -07:00 |
|
|
YAMY
|
6480cce9bc
|
perf: skip redundant scheduler metadata gather for DP1 (#36568)
|
2026-08-27 20:04:54 -07:00 |
|
|
Liangsheng Yin
|
daf6317196
|
[mem_cache] Add free_full to release the full side of a tombstoned SWA node (#36637)
|
2026-08-27 19:34:48 -07:00 |
|
|
Kurt Shuster
|
b7686e17d6
|
[k3] declare packed_modules_mapping on KimiK3ForConditionalGeneration (#36211)
|
2026-08-28 10:22:36 +08:00 |
|
|
Aurick Qiao
|
84a6f51bd6
|
Fix DeepSeek V4 multistream QKV buffer lifetime (#36547)
|
2026-08-28 10:10:36 +08:00 |
|
|
|
cbd0271574
|
[diffusion] model: add cosmos3 transfer capability (#34747)
Co-authored-by: Kedi Wu <kediw@nvidia.com>
Co-authored-by: Kedi Wu <31940276+kediwu0331@users.noreply.github.com>
|
2026-08-28 09:37:22 +08:00 |
|
|
billishyahao
|
de2fb50120
|
[AMD] Enable aiter mla asm path through padding attn heads for Kimi K3 (#36356)
|
2026-08-27 18:30:29 -07:00 |
|
|
Xiaoyu Zhang
|
96b31770f9
|
[diffusion] fuse Helios paired transposed RoPE (#36502)
|
2026-08-28 08:57:37 +08:00 |
|
|
Aurick QiaoandYuwei An
|
26fd7fdaa2
|
[Feature] Support PP in full prefill CUDA graphs (#35451)
Co-authored-by: Yuwei An <ayw.sirius19@gmail.com>
|
2026-08-27 17:32:00 -07:00 |
|
|
ashwini rathi
|
7cbe564829
|
[Fix][XPU/ROCm/NPU] Defer sgl_kernel.quantization import in expert_pack (#36529)
|
2026-08-27 16:17:24 -07:00 |
|
|
Baizhou Zhang
|
76217f6603
|
Revert "[NPU] [bugfix] Fix import of ggml_moe_a8_vec and Fix NPU MLA HiCache backup accessing missing data_ptrs" (#36747)
|
2026-08-27 16:16:30 -07:00 |
|