All pull requests

Pull requests

• feat: Add blockwise FP8 GEMM support for Qwen3-VL

  #220·Sunk011 opened Sep 23, 2026·

  ·
  1

• feat: add native continuous batching and chunked prefill for LLM runtime

  #215·ajmalrasi opened Sep 15, 2026·

  ·

• Fix Qwen3 ONNX export bug related to 0/1 specialization

  #213·victoroliv2 opened Sep 14, 2026·

  ·

• fix https://github.com/NVIDIA/TensorRT-Edge-LLM/issues/211: advance the Philox offset per sampled token

  #212·filipemartinsubrobotics opened Sep 12, 2026·

  ·
  1

• fix: say why Int4GroupwiseGemmPluginV2 cannot execute when its kernels are absent

  #209·filipemartinsubrobotics opened Sep 11, 2026·

  ·
  1

• feat: enable SM103 (B300 / Blackwell Ultra) across the attention SM allowlists

  #207·filipemartinsubrobotics opened Sep 11, 2026·

  ·
  1

• fix https://github.com/NVIDIA/TensorRT-Edge-LLM/issues/204: pass the CUDA include path to the XQA NVRTC JIT

  #205·filipemartinsubrobotics opened Sep 11, 2026·

  ·
  1

• fix: Share the target lm_head for DSpark drafts without their own head

  #200·rain-sicoreai opened Sep 9, 2026·

  ·

• feat: Add request batching and spec tree-size options to the experimental server, fix FP8-KV and unquantized MTP draft builds

  #199·rain-sicoreai opened Sep 8, 2026·

  ·
  1

• feat: add InternVLA-N1-DualVLN vision-language navigation support

  #193·hungho77 opened Aug 24, 2026·

  ·
  1

  2

• fix(qwen3_5): skip GDN input-projection fusion on mixed projection dt…

  #182·potoior opened Aug 17, 2026·

  ·

• fix #161: XQA cubin generation ignores resolved output dir under spawn/forkserver

  #177·Hi5808 opened Aug 15, 2026·

  ·

• fix(qwen3_moe): export W4A16 NVFP4 checkpoints via NVFP4 MoE path

  #176·potoior opened Aug 13, 2026·

  ·
  3

• fix(checkpoint): Preserve destination tensor dtype

  #149·suharvest opened Jul 25, 2026·

  ·
  1

• feat(builder): Set workspace RAM limits and enhance log level controls

  #126·cjac opened Jul 6, 2026·

  ·

• Strip /usr/include from apt pybind11's interface include dirs

  #121·whitesscott opened Jul 2, 2026·

  ·

• Use <cmath> instead of <math.h> in context attention FMHA headers

  #120·whitesscott opened Jul 1, 2026·

  ·

• fix: Respect native Orin CUDA architecture and CuTe link requirements

  #118·suharvest opened Jun 28, 2026·

  ·
  1

  1

• fix: propagate CuTe DSL runtime link requirements for static libraries

  #103·francismelon opened Jun 7, 2026·

  ·

• fix https://github.com/NVIDIA/TensorRT-Edge-LLM/issues/87: hard-error instead of silent return when CuTe DSL GEMM is not compiled

  #97·suharvest opened May 27, 2026·

  ·

• imageUtilKernels.cu: optimize initAttentionMaskKernel

  #28·vitamin-chaos opened Jan 26, 2026·

  ·
  3