kanshan/TensorRT-LLMs
1
0
Fork 0
mirror of https://github.com/NVIDIA/TensorRT-LLM.git synced 2026-01-13 22:18:36 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity
4,607 Commits 25 Branches 59 Tags 2.1 GiB
main
Commit Graph

278 Commits

Author SHA1 Message Date
Iman Tabrizian
48b09e5a25
[https://nvbugs/5689235][fix] Fix cancellation+chunked prefill+disagg (#10111) 
Signed-off-by: Iman Tabrizian <10105175+tabrizian@users.noreply.github.com>
 2026-01-12 18:23:26 -05:00
Min Yu
9cae7277ea
[https://nvbugs/5726962][feat] Apply fusion for W4AFP8_AWQ MoE (#9838) 
Signed-off-by: Min Yu <171526537+yumin066@users.noreply.github.com>
Signed-off-by: Anthony Chang <27950904+rosenrodt@users.noreply.github.com>
Co-authored-by: Anthony Chang <27950904+rosenrodt@users.noreply.github.com>
 2026-01-06 10:16:41 +08:00
Chuang Zhu
536a8f6a9c
[TRTLLM-9527][feat] Add transferAgent binding (step 1) (#10113) 
Signed-off-by: Chuang Zhu <111838961+chuangz0@users.noreply.github.com>
 2026-01-06 08:40:38 +08:00
Balaram Buddharaju
a792c23dcf
[TRTLLM-9465][fix] Swap TP-CP grouping order (#10350) 
Signed-off-by: Balaram Buddharaju <169953907+brb-nv@users.noreply.github.com>
 2026-01-05 20:08:03 +08:00
Yueh-Ting (eop) Chen
9cee32ab39
[https://nvbugs/5625990][fix] Respect VSWA scheme when doing block store for reuse and load block for reuse in KV cache manager (#10183) 
Signed-off-by: eopXD <yuehtingc@nvidia.com>
 2025-12-29 14:29:14 +08:00
Roey Azran
8408c40d8b
[https://nvbugs/5702786][fix] Fix race conditions in KV cache communication during unexpected termination (#10076) 
Signed-off-by: roeya <165803633+RoeyAzran1992@users.noreply.github.com>
 2025-12-23 14:09:51 +02:00
Wangjue Yao
9f283f330b
[None][feat] Support Mooncake transfer engine as a cache transceiver backend (#8309) 
Signed-off-by: wjueyao <wyao123@terpmail.umd.edu>
Signed-off-by: Shunkang <182541032+Shunkangz@users.noreply.github.co>
Co-authored-by: Shunkang <182541032+Shunkangz@users.noreply.github.co>
 2025-12-19 10:09:51 +08:00
ChristinaZ
dff77efa2a
[None][feat] Add routing support for the new model for both cutlass and trtllm moe backend (#9792) 
Signed-off-by: Christina Zhang <83400082+ChristinaZ@users.noreply.github.com>
 2025-12-15 19:59:08 -08:00
Yihan Wang
9df4dad3b6
[None][fix] Introduce inline namespace to avoid symbol collision (#9541) 
Signed-off-by: Yihan Wang <yihwang@nvidia.com>
 2025-12-12 23:32:15 +08:00
Ludwig Schneider
41ce14ab04
[None][feat] Enable NCCL_SYMMETRIC as default fallback for AllReduce (#9314) 
Signed-off-by: Ludwig Schneider <lschneider@nvidia.com>
 2025-12-07 09:43:26 -08:00
Yuan Tong
becd44f9bc
[None][fix] Correct virtual memory allocation alignment (#9491) 
Signed-off-by: Yuan Tong <13075180+tongyuantongyu@users.noreply.github.com>
 2025-12-01 10:59:19 +08:00
cheshirekow
1379cfac3a
[TRTLLM-9197][infra] Move thirdparty stuff to it's own listfile (#8986) 
Signed-off-by: Josh Bialkowski <1309820+cheshirekow@users.noreply.github.com>
Co-authored-by: Josh Bialkowski <1309820+cheshirekow@users.noreply.github.com>
 2025-11-20 16:44:23 -08:00
Kanghwan
41e5870a70
[#8476][chore] Update license (#8807) 
Signed-off-by: Kanghwan Jang <861393+karljang@users.noreply.github.com>
 2025-11-19 15:05:25 -08:00
ChristinaZ
fbf6c16cd2
[None][fix] Update the default invalid value for deepseek mode of routing (#9222) 
Signed-off-by: Christina Zhang <83400082+ChristinaZ@users.noreply.github.com>
 2025-11-19 10:14:06 +08:00
Patrice Castonguay
9b0f45298f
[None][feat] Have ability to cancel disagg request if KV cache resource are exhausted (#9155) 
Signed-off-by: Patrice Castonguay <55748270+pcastonguay@users.noreply.github.com>
 2025-11-18 20:59:17 -05:00
Robin Kobus
9913dc25ae
[None][refactor] decoding inputs, part 2 (#5799) 
Signed-off-by: Robin Kobus <19427718+Funatiq@users.noreply.github.com>
 2025-11-18 14:38:51 +01:00
Anthony Chang
86cfb3ea7e
[None][feat] Update TRTLLM MoE cubins; reduce mxfp4 weight padding requirement; tighten TMA bound (#9025) 
Signed-off-by: Anthony Chang <27950904+rosenrodt@users.noreply.github.com>
 2025-11-17 10:04:29 +08:00
sunnyqgg
7862b15a65
[TRTLLM-8778][feat] Add tree attention support for blackwell arch (#8975) 
Signed-off-by: qgai <qgai@nvidia.com>
 2025-11-17 09:01:53 +08:00
heyuhhh
f07e9977c6
[None] [feat] Use triton kernels for RocketKV prediction module (#8682) 
Signed-off-by: yuhangh <58161490+heyuhhh@users.noreply.github.com>
 2025-11-13 18:51:09 -08:00
Iman Tabrizian
9ef7eb70e0
[None][fix] Fix KV cache manager test warnings (#9103) 2025-11-13 07:23:04 -08:00
Iman Tabrizian
cdde15b275
[TRTLLM-8540][feat] Add support for disagg in DSv3.2 (#8735) 
Signed-off-by: Iman Tabrizian <10105175+tabrizian@users.noreply.github.com>
 2025-11-12 08:21:11 -08:00
brb-nv
d798d66976
[TRTLLM-7731][feat] Avoid over-allocation of KV cache for transmission in disagg with CP (#8145) 
Signed-off-by: Balaram Buddharaju <169953907+brb-nv@users.noreply.github.com>
 2025-10-31 17:32:39 -07:00
ChristinaZ
13cfd70f57
[None][feat] Add unit tests and revision in block_level kernel for invalid input (#8718) 
Signed-off-by: Christina Zhang <83400082+ChristinaZ@users.noreply.github.com>
 2025-10-30 16:42:18 +08:00
Chuang Zhu
b828b6445b
[https://nvbugs/5612529][fix] Fix transferAgent_test (#8710) 
Signed-off-by: Chuang Zhu <111838961+chuangz0@users.noreply.github.com>
 2025-10-29 09:14:34 +08:00
nvxuanyuc
d1398c05e6
[None][feat] Support ignored prompt length for penalties via new sampling config parameter (#8127) 
Signed-off-by: Xuanyu Chen <xuanyuc@nvidia.com>
 2025-10-27 13:12:31 -04:00
Jinyang Yuan
0a0f93d4a8
[None][fix] Fix the performance issue of FP8 blockwise grouped GEMM when using attention DP (#8501) 
Signed-off-by: Jinyang Yuan <154768711+jinyangyuan-nvidia@users.noreply.github.com>
 2025-10-27 10:18:19 +08:00
Chuang Zhu
2420918e5b
[TRTLLM-7078][chore] optimal kvcache transfer for VWSA (#7952) 
Signed-off-by: Chuang Zhu <111838961+chuangz0@users.noreply.github.com>
 2025-10-24 08:58:16 -04:00
Anish Shanbhag
15de45d782
[TRTLLM-8682][chore] Remove auto_parallel module (#8329) 
Signed-off-by: Anish Shanbhag <ashanbhag@nvidia.com>
 2025-10-22 20:53:08 -04:00
Shi Xiaowei
a0024f4d34
[None][doc] Facilitates the integration of the transfer agent (#7867) 
Signed-off-by: Shixiaowei02 <39303645+Shixiaowei02@users.noreply.github.com>
 2025-10-21 20:06:24 +08:00
ChristinaZ
c8b9998acb
[TRTLLM-8637][feat] Optimize the routing kernel for DeepseekV3 (MoE CUTLASS backend); Add support for KimiK2 and Qwen-next (MoE TRTLLM backend) (#7761) 
Signed-off-by: Christina Zhang <83400082+ChristinaZ@users.noreply.github.com>
 2025-10-20 10:08:31 +08:00
ChristinaZ
db1c271bc6
[None][feat] Revise the calculation related to TileN in routing of MOE TRTLLM backend (#8148) 
Signed-off-by: Christina Zhang <83400082+ChristinaZ@users.noreply.github.com>
 2025-10-16 09:15:46 +08:00
Chuang Zhu
40d129a415
[None][fix] Fix cache buffer size for window (#8320) 
Signed-off-by: Chuang Zhu <111838961+chuangz0@users.noreply.github.com>
 2025-10-16 09:01:11 +08:00
Fanrong Li
0d20a8fd61
[TRTLLM-8536][feat] Add the sparse attention framework and one use case--RocketKV support (#8086) 
Signed-off-by: Fanrong Li <23290157+lfr-0531@users.noreply.github.com>
Signed-off-by: yuhangh <58161490+heyuhhh@users.noreply.github.com>
Co-authored-by: yuhangh <58161490+heyuhhh@users.noreply.github.com>
 2025-10-14 08:23:16 -07:00
Yueh-Ting (eop) Chen
4882815fa1
[TLLM-6777][feature] Support SWA KV cache reuse OOW block detach (#7922) 
This MR is a continuation of #6768. In the previous merge request,
OOW (out-of-window) blocks are only detached when reuse is not enabled,
that is, the block movement behavior is identical between SWA and full
attention when reuse is enabled.

This merge request attempts to enable OOW block detach when reuse is
enabled. The required changes are:

- Let KV cache manager keep track of which block is used by which
  sequence
- Remove restriction for the eviction policy to be able to release a
  non-leaf block

Along with the development, bugs inside freeChildren and offload
mechanism under getFreeBlock is resolved because they will affect the
functionality this merge request is trying to achieve.

When a block goes OOW, it is released from the sequence, it will be
available to be reclaimed and the block is held by the eviction policy
for another sequence to acquire upon calling. On the other hand, we
want to potentially store the sequence for reuse. To safely achieve
this, the record of block ownership is done under
WindowBlockManager::getFreeBlock. If the block acquired was originally
owned by another sequence that is live inside the manager, then we
invalidate the sequence for store for reuse.

At the end of a sequence (when removeSequence is called toward it),
the KV cache manager will check if the sequence has all blocks not
reclaimed by another sequence. If so, then the sequence is safe to
be stored for reuse and store for reuse action will be performed.

Signed-off-by: eopXD <yuehtingc@nvidia.com>
 2025-10-13 09:18:12 -07:00
Faraz
27a5091fcb
[None][feat] GPT-OSS Sm120/Sm121 Support (#7937) 
Signed-off-by: Perkz Zheng <67892460+PerkzZheng@users.noreply.github.com>
Signed-off-by: list <58580514+farazkh80@users.noreply.github.com>
Signed-off-by: Vincent Huang <vincenth@nvidia.com>
Co-authored-by: Perkz Zheng <67892460+PerkzZheng@users.noreply.github.com>
Co-authored-by: Vincent Huang <vincenth@nvidia.com>
 2025-10-06 16:59:06 -04:00
Jonas Yang CN
88ea2c4ee9
[TRTLLM-7349][feat] Adding new orchestrator type -- ray (#7520) 
Signed-off-by: Erin Ho <14718778+hchings@users.noreply.github.com>
Co-authored-by: Yuan Tong <13075180+tongyuantongyu@users.noreply.github.com>
Co-authored-by: Erin Ho <14718778+hchings@users.noreply.github.com>
 2025-10-04 08:12:24 +08:00
Robin Kobus
e2f69c5c23
[None] [refactor] Minor cleanup and improvements (#7619) 
Signed-off-by: Robin Kobus <19427718+Funatiq@users.noreply.github.com>
 2025-10-03 11:40:06 +02:00
Patrice Castonguay
fefa7d8fa3
[None][feat] Support for cancelling requests with disaggregation (#8114) 
Signed-off-by: Shunkang <182541032+Shunkangz@users.noreply.github.co>
Signed-off-by: Patrice Castonguay <55748270+pcastonguay@users.noreply.github.com>
Co-authored-by: Shunkang <182541032+Shunkangz@users.noreply.github.co>
 2025-10-02 11:04:26 -07:00
Iman Tabrizian
33282351a2
[TRTLLM-6106][feat] Add support for KVCache transfer from KVCache reuse path (#6348) 
Signed-off-by: Iman Tabrizian <10105175+tabrizian@users.noreply.github.com>
 2025-09-27 19:29:30 -04:00
Chuang Zhu
f98fa0cf8b
[None][feat] Optimize kv cache transfer TEP (#7613) 
Signed-off-by: Chuang Zhu <111838961+chuangz0@users.noreply.github.com>
 2025-09-25 20:20:04 -07:00
Yueh-Ting (eop) Chen
c5012423f5
[None][chore] Remove developer name in comment (#7981) 
Signed-off-by: eopXD <yuehtingc@nvidia.com>
 2025-09-25 06:43:38 -07:00
Guoming Zhang
202bed4574 [None][chroe] Rename TensorRT-LLM to TensorRT LLM for source code. (#7851) 
Signed-off-by: nv-guomingz <137257613+nv-guomingz@users.noreply.github.com>
Signed-off-by: Wangshanshan <30051912+dominicshanshan@users.noreply.github.com>
 2025-09-25 21:02:35 +08:00
Yueh-Ting (eop) Chen
cf100933cc
[TRTLLM-6341][feature] Support SWA KV cache reuse (#6768) 
This merge request attempts to support more SWA KV cache functionality
inside the KV cache manager. Before this merge request, the KV cache for
sliding window attention (SWA) only holds "window size" number of blocks
and reuse them in a cyclic manner. We will not be able to utilize more
GPU memory with this design, leading to a limited max batch size
throughput. Additionally, we will not be able to support KV cache reuse
with this design.

In this MR, we change such behavior to let the manager write blocks in
a linear manner. With a linear block writing behavior, as the attention
window moves on, the out-of-window (OOW) blocks will be detached. Right
now for the sake of a correct feature first, we directly offload the
OOW block from the primary block pool (GPU memory) to the secondary
block pool (host memory). We will improve this in the future by
delegating the block movement to the eviction policy.

KV cache reuse for SWA is not developed in this merge request and will
be amended in a follow-up merge request.

Writing the blocks linearly, the maximum number of blocks allocated for
a sequence(`GenerationRequest`) is the "max sequence length" specified.
The `GenerationRequest` that stores the cache block bookkeeping
structure will now keep "max sequence length" tokens of blocks.

Given the above, main changes are (more context in the MR):
- Remove "cyclic" concept under the kv cache manager, such concept
  originally guards the block reuse under kv cache manager.
- Add detach mechanism and have it under `KVCacheManager::addToken`.
  Please note that detach is still guarded off for SWA when reuse
  is enabled. A follow-up merge request will proceed to improve this.
- Enforce "max sequence length" to be a non-optional parameter to
  the `KVCacheManager`/`BlockManager`
- Let all window size resource pool get identical proportion of memory
- Fix free memory calculation under `resource_manager.py`

Signed-off-by: eopXD <yuehtingc@nvidia.com>
Co-authored-by: Tomer Asida <tasida@nvidia.com>
 2025-09-24 14:28:24 +08:00
ChristinaZ
be576a3152
[None] [feat] Enable run_post_quant_allgather for MoE TRTLLM backend (#6794) 
Signed-off-by: Christina Zhang <83400082+ChristinaZ@users.noreply.github.com>
 2025-09-23 08:24:21 +08:00
brb-nv
8879ec4d35 [https://nvbugs/5501557][fix] Fix out-of-bounds vector access for model with multiple layer types (#7636) 
Signed-off-by: Balaram Buddharaju <169953907+brb-nv@users.noreply.github.com>
Signed-off-by: Wangshanshan <30051912+dominicshanshan@users.noreply.github.com>
 2025-09-22 14:28:38 +08:00
xiweny
822cb0115b
[TRTLLM-6286] [perf] Add NoSmem epilogue schedule and dynamic cluster shape for sm10x group gemm (#7757) 
Signed-off-by: Xiwen Yu <13230610+VALLIS-NERIA@users.noreply.github.com>
Signed-off-by: djns99 <40156487+djns99@users.noreply.github.com>
Co-authored-by: djns99 <40156487+djns99@users.noreply.github.com>
 2025-09-21 11:38:17 +08:00
brb-nv
e10a027a03
[TRTLLM-7731][feat] KV cache transmission in disagg with CP on gen side (#7624) 
Signed-off-by: Balaram Buddharaju <169953907+brb-nv@users.noreply.github.com>
 2025-09-20 06:15:26 -07:00
Iman Tabrizian
6ce0624208
[TRTLLM-8044][refactor] Rename data -> cache for cacheTransceiver (#7659) 2025-09-16 08:43:56 -04:00
xiweny
c076a02b38
[TRTLLM-4629] [feat] Add support of CUDA13 and sm103 devices (#7568) 
Signed-off-by: Xiwen Yu <13230610+VALLIS-NERIA@users.noreply.github.com>
Signed-off-by: Tian Zheng <29906817+Tom-Zheng@users.noreply.github.com>
Signed-off-by: Daniel Stokes <dastokes@nvidia.com>
Signed-off-by: Zhanrui Sun <zhanruis@nvidia.com>
Signed-off-by: Xiwen Yu <xiweny@nvidia.com>
Signed-off-by: Jiagan Cheng <jiaganc@nvidia.com>
Signed-off-by: Yiqing Yan <yiqingy@nvidia.com>
Signed-off-by: Bo Deng <deemod@nvidia.com>
Signed-off-by: ZhanruiSunCh <184402041+ZhanruiSunCh@users.noreply.github.com>
Signed-off-by: xiweny <13230610+VALLIS-NERIA@users.noreply.github.com>
Co-authored-by: Tian Zheng <29906817+Tom-Zheng@users.noreply.github.com>
Co-authored-by: Daniel Stokes <dastokes@nvidia.com>
Co-authored-by: Zhanrui Sun <zhanruis@nvidia.com>
Co-authored-by: Jiagan Cheng <jiaganc@nvidia.com>
Co-authored-by: Yiqing Yan <yiqingy@nvidia.com>
Co-authored-by: Bo Deng <deemod@nvidia.com>
Co-authored-by: Zhanrui Sun <184402041+ZhanruiSunCh@users.noreply.github.com>
 2025-09-16 09:56:18 +08:00
jmydurant
7deefb3d2b
[TRTLLM-7192][feat] optimize MLA chunked prefill && support fp8 mla chunked prefill (#7477) 
Signed-off-by: Mingyang Jiang <13463932+jmydurant@users.noreply.github.com>
 2025-09-15 21:43:49 +08:00