Issues

• Engine build failure of TensorRT 10.14.1 and 11.3.0 when running S2M2 stereo matching above ~1 MPix input on Jetson AGX Thor (sm_110)

  Module:Engine Build
  NVIDIA/TensorRT#4866 · Filter by author george-auradine opened on Oct 4, 2026
  1

• Polygraphy parse_meta accepts a shape with a missing bracket

  Module:Polygraphy
  NVIDIA/TensorRT#4864 · Filter by author SashaMIT opened on Sep 30, 2026

• 邀请 tensorrt 加入 GithubStarMate，让更多人发现你的作品

  NVIDIA/TensorRT#4863 · Filter by author gecko9336 opened on Sep 30, 2026

• ScatterND in a dynamic-shape engine silently drops updates beyond ~N when the update count M exceeds the output size N (static engine correct)

  NVIDIA/TensorRT#4861 · Filter by author braindevices opened on Sep 29, 2026

• TRT 11.3: Attention does not compile inside If, even with decomposable=True

  Module:ONNX
  NVIDIA/TensorRT#4860 · Filter by author vient opened on Sep 28, 2026
  1

• [Bug]: trtllm-serve renders a client-supplied chat_template server-side, a string-blow-up expression OOM-kills whole server (unauthenticated crash DoS)

  NVIDIA/TensorRT#4859 · Filter by author Yunzez opened on Sep 27, 2026

• onnx_graphsurgeon: Constant with float8_e4m3fnuz values is exported as FLOAT8E4M3FN

  Module:GraphSurgeon
  NVIDIA/TensorRT#4856 · Filter by author luicarus opened on Sep 21, 2026

• trtllm-serve accepts response_format and never enforces it

  NVIDIA/TensorRT#4855 · Filter by author Yunzez opened on Sep 17, 2026

• Wrong FP16 results when a fused convolution, activation and residual add reads a strided concat view

  Module:Accuracy
  NVIDIA/TensorRT#4854 · Filter by author Y-T-G opened on Sep 17, 2026
  1

• TensorRT silently produces wrong results for batch > 1 on SegFormer (b0/b1/b2 verified, fp32, static and dynamic engines, TRT 10.11 & 11.1, sm89 + sm120)

  Module:Accuracy
  NVIDIA/TensorRT#4853 · Filter by author Daniel595 opened on Sep 15, 2026
  1

• [CMake] Windows package config references missing bin/tensorrt_player.exe (tensorrt_capture_player)

  Module:OSS Build
  NVIDIA/TensorRT#4851 · Filter by author patrikhuber opened on Sep 14, 2026

• AOT plugins leak CUDA modules on dynamic-shape changes in TensorRT 11.1/11.2.1

  Module:Plugins
  NVIDIA/TensorRT#4849 · Filter by author lsj574 opened on Sep 12, 2026
  1

• Question: How much large-scale scaling behavior can be predicted before benchmarking?

  NVIDIA/TensorRT#4848 · Filter by author alexbuiko-sketch opened on Sep 11, 2026

• Incorrect output failure of TensorRT 11.1 and 11.2 when running dynamic If → Transpose → Resize on GPU GTX 1660 SUPER and A10G

  Module:Accuracy
  NVIDIA/TensorRT#4846 · Filter by author lsj574 opened on Sep 6, 2026

• TensorRT 10.14+ multi-GPU build regression with Cask-backed ops

  Module:Engine Build
  NVIDIA/TensorRT#4845 · Filter by author braindevices opened on Sep 5, 2026

• BF16 strongly-typed engine build SIGSEGVs in Myelin on RTX 5090 (sm_120) for DETR-family transformer — identical graph builds fine as FP16 (reproduced on 11.0.0.114 / 11.1.0.106 / 11.2.1.2)

  Module:Engine Build
  NVIDIA/TensorRT#4842 · Filter by author viktorsroze-sketch opened on Aug 27, 2026

• Serialized engine size failure of TensorRT 11.2 when building an unfused pointwise op on a large activation on GPU H100

  Module:Engine Build
  NVIDIA/TensorRT#4841 · Filter by author himmelroman opened on Aug 26, 2026
  2

• Segmentation fault in TensorRT 10.3 during Myelin compilation of PointCNN-style model on Jetson Orin NX

  Module:Engine Build
  NVIDIA/TensorRT#4840 · Filter by author trodrbr opened on Aug 26, 2026

• FP16 model compilation failure with TensorRT 11.2.1.2 on 4090 RTX

  Module:ONNX
  NVIDIA/TensorRT#4838 · Filter by author h6197627 opened on Aug 25, 2026
  2

• Silently incorrect fp16/bf16 outputs from fused MHA on consumer Blackwell (sm_120) for ViT/DETR model (SAM 3 detector) — fp32 exact, persists in 10.16.1, breaking the MHA fusion restores accuracy

  Module:Accuracy
  NVIDIA/TensorRT#4837 · Filter by author crisreimberg opened on Aug 24, 2026

• W4A16 FP4 build succeeds while silently realizing FP32 (no warning, no refusal)

  Module:Quantization
  NVIDIA/TensorRT#4833 · Filter by author geoffrey-delhomme opened on Aug 14, 2026
  2

• Myelin "Could not infer output types for operation: dequantize" on SM 8.7 (Orin); identical engine builds on SM 11.0 (Thor) with the same TensorRT 10.16.2.10

  Module:Quantization
  NVIDIA/TensorRT#4832 · Filter by author geoffrey-delhomme opened on Aug 14, 2026
  2

• [Architecture Discussion] Overhead and safety of asynchronous soft-stop signaling for KV-cache reclamation during generation loops

  Feature Request
  NVIDIA/TensorRT#4831 · Filter by author alexbuiko-sketch opened on Aug 14, 2026
  1

• Myelin NVRTC compilation failure on Jetson Thor (SM110) building Fast-FoundationStereo - reproduces on TRT 10.14.1 and 11.2.1.2

  Module:EmbeddedModule:Engine Build
  NVIDIA/TensorRT#4830 · Filter by author juliusekela opened on Aug 13, 2026
  1

• TensorRT 11.0.0.114 builder finds no valid tactic for BF16 ConvTranspose2d (works fine in FP16)

  Module:Engine Build
  NVIDIA/TensorRT#4829 · Filter by author robbysun opened on Aug 12, 2026
  2