1 Release published by 1 person

• v2.8.0 Torch-TensorRT v2.8.0

  published Aug 9, 2025

14 Pull requests merged by 4 people

• Breaking Change: Remove the deprecated int8 calibrator related

  #3759 merged Aug 10, 2025

• Cherry pick jetson enablement from 2.8 release branch to main

  #3765 merged Aug 10, 2025

• chore(deps): bump transformers from 4.51.3 to 4.53.0 in /tools/perf

  #3754 merged Aug 8, 2025

• add nvshmem in aarch64

  #3769 merged Aug 8, 2025

• fix build cancellation issue

  #3768 merged Aug 8, 2025

• Fix Jetson FP4 gate issue

  #3764 merged Aug 8, 2025

• fix typing-extensions issue

  #3761 merged Aug 7, 2025

• broadcast_remove - cherry pick 3700

  #3757 merged Aug 7, 2025

• add typing_extensions as test dependencies which is required by modelopt

  #3743 merged Aug 7, 2025

• remove breakpoint()

  #3750 merged Aug 6, 2025

• Fixed SDPA slow down and linear slow down

  #3700 merged Aug 4, 2025

• enable back jetpack build

  #3720 merged Aug 4, 2025

• Upgrade perf_run script to support TRT 10 and fix some issues

  #3650 merged Aug 4, 2025

• add test cases for strong typing

  #3739 merged Aug 4, 2025

11 Pull requests opened by 5 people

• fix: Inferred dimensions at build time in reshape

  #3746 opened Aug 5, 2025

• cherry pick 3700 to 2.8 release: Broadcast removal

  #3747 opened Aug 5, 2025

• add strong typing fix

  #3749 opened Aug 5, 2025

• fix: atan2 strong type support & bug fix for integer dynamic shape

  #3751 opened Aug 6, 2025

• Add support for TensorRT-RTX

  #3753 opened Aug 6, 2025

• Revised the lowering pass according to Bo's suggestion

  #3756 opened Aug 7, 2025

• fix: batch norm issue encountered in RAFT

  #3758 opened Aug 7, 2025

• Please do not review: perf test rtx only

  #3763 opened Aug 7, 2025

• fix: set example models to eval mode and follow the convention

  #3770 opened Aug 9, 2025

• upgrade torchvision from 0.23.0 to 0.24.0

  #3772 opened Aug 10, 2025

• fix the typo

  #3773 opened Aug 10, 2025

5 Issues closed by 2 people

• 🐛 [Bug] SDPA decomposition causing TorchTRT to be 2x slower than ONNX on SD3.5

  #3682 closed Aug 4, 2025

• 🐛 [Bug] SDPA in Torch-TRT is slower than SDPA in ONNX as batch size (num_frames) grow for Wan2.1

  #3695 closed Aug 4, 2025

• [Bug] MHA Kernels and Linear Kernels are slower in FLUX

  #3707 closed Aug 4, 2025

• 🐛 [Bug] Changing input size would affect the TRT engine size, testing on BERT

  #3634 closed Aug 4, 2025

• 🐛 [Bug] perf_run.py script doesn't support TRT10 and there are some known bugs

  #3709 closed Aug 4, 2025

7 Issues opened by 5 people

• Please promote torch_tensorrt 2.8 release artifacts to pytorch index

  #3771 opened Aug 9, 2025

• ❓ [Question] C++ Windows runtime error

  #3766 opened Aug 8, 2025

• 🐛 [Bug] CudaGraph cannot work with module with graph breaks

  #3755 opened Aug 6, 2025

• 🐛 [Bug] Refitter test failed when constant fold is disabled

  #3752 opened Aug 6, 2025

• 🐛 [Bug] Llama2_flashinfer_rmsnorm example is broken

  #3748 opened Aug 5, 2025

• 🐛 [Bug] Compilation failure with NVFP4 Quantization with dynamic shapes

  #3745 opened Aug 5, 2025

• ✨[Feature] Pre-compiled C++ Binaries for Windows

  #3744 opened Aug 5, 2025

12 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• integrated vlm code for benchmark for Eagle2

  #3698 commented on Aug 8, 2025 • 17 new comments

• TRT-LLM loading mechanism tool

  #3398 commented on Aug 8, 2025 • 3 new comments

• Removal of BAZEL build files from python package and changes to make cpp tests work

  #3641 commented on Aug 7, 2025 • 1 new comment

• 🐛 [Bug] Large Accuracy Issue

  #3626 commented on Aug 4, 2025 • 0 new comments

• 🐛 [Bug] TensorRT-RTX BatchNorm constant fold got nan

  #3699 commented on Aug 6, 2025 • 0 new comments

• 🐛 [Bug] perf gap reduce on RAFT

  #3731 commented on Aug 7, 2025 • 0 new comments

• ✨[Feature] Add FX tests to CI

  #3492 commented on Aug 10, 2025 • 0 new comments

• chore: move external dep installation into a separate script

  #3672 commented on Aug 8, 2025 • 0 new comments

• fix: prelu perf gap on Unet

  #3717 commented on Aug 8, 2025 • 0 new comments

• feat: Add support for Groot N1.5 model

  #3736 commented on Aug 6, 2025 • 0 new comments

• Feat: Pre-quantized LLM model support

  #3740 commented on Aug 8, 2025 • 0 new comments

• Tentatively eliminate graph break overhead

  #3741 commented on Aug 7, 2025 • 0 new comments