Skip to content

Pull requests: NVIDIA/TensorRT-LLM

New pull request New
215 Open 1,224 Closed
215 Open 1,224 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Draft: test: [CI] remove closed bugs
#4046 opened May 4, 2025 by xinhe-nv • Draft
fix: apply rope twice in Qwen3.
#4040 opened May 3, 2025 by yuxianq Loading…
4
chore: cleanup llmapi for 1.0
#4039 opened May 2, 2025 by hchings Loading…
@hchings
3
[Deepseek] Add fp8 kvcache test
#4038 opened May 2, 2025 by hlu1 Loading…
4
Add disaggregated serving accuracy tests
#4036 opened May 2, 2025 by Tabrizian Loading…
8
test: Test OOB access issue in penaltyKernel for endId=-1
#4035 opened May 2, 2025 by brb-nv Loading…
6
Draft: feat: non-invasive pipeline parallelism
#4034 opened May 2, 2025 by yuxianq Loading…
13
[nvbug/5248986][fix] Skip debugCheckSemaphores in stream capture mode
#4032 opened May 2, 2025 by mikeiovine Loading…
@mikeiovine
4
[DRAFT] setting attention prior
#4031 opened May 2, 2025 by vklimkov-nvidia Loading…
3
[DRAFT] Introducing multi-vocab token sampling for audio generation
#4030 opened May 2, 2025 by vklimkov-nvidia Loading…
10
fix: instantiate decoder early in pytorch
#4029 opened May 2, 2025 by dcampora Loading…
3
feat:enable kvcache to be reused during request generation Community Engagement Community want to contribute
#4028 opened May 2, 2025 by narutolhy Loading…
1
Refactor: Restructure C++ tests for better modularisation of non-shared code
#4027 opened May 2, 2025 by DomBrown Loading…
@DomBrown
14
fix: Properly get decoding mode according to same logic as cpp.
#4026 opened May 2, 2025 by dcampora Loading…
6
[AutoDeploy][perf] Further optimize flashinfer backend in AutoDeploy
#4024 opened May 2, 2025 by suyoggupta Loading…
8
feat: Enable AutoDeploy to llm-eval example
#4020 opened May 2, 2025 by meenchen Loading…
feat:add slurm support and add b40 to test-db
#4019 opened May 2, 2025 by yuanjingx87 Loading…
14
TorchLLM: Pass local dir to processor creation
#4018 opened May 1, 2025 by milesial Loading…
1
Draft: Support long context LLama 4 (flashinfer backend)
#4017 opened May 1, 2025 by vanshilshah97 • Draft
[Deepseek] Refactor Deepseek Decoder layer
#4016 opened May 1, 2025 by hlu1 Loading…
12
[Draft][AutoDeploy] Split prefill and decode in AD's flashinfer backend
#4015 opened May 1, 2025 by suyoggupta • Draft
[fix] support llama + eagle head checkpoint conversion
#4013 opened May 1, 2025 by jhaotingc Loading…
bench: Port benchmark_serving.py
#4011 opened May 1, 2025 by kaiyux • Draft
1
Put the computation of Q and K norm (in attn) into a single CUDA stream, and get a 5% - 8% throughput improvement on Qwen3 4B and Qwen3 - moe 30B - A3B.
#4005 opened May 1, 2025 by shaonvidia Loading…
experiments: set self.max_position_embedding=None in Attention module
#4002 opened Apr 30, 2025 by qixiang-99 • Draft
10
ProTip! no:milestone will show everything without a milestone.