1 Release published by 1 person

• v2.7.0 Torch-TensorRT v2.7.0

  published May 7, 2025

31 Pull requests merged by 8 people

• Add Linux CI build for aarch64

  #3516 merged May 13, 2025

• Update mutable_torchtrt_module_example.py

  #3519 merged May 13, 2025

• add nspect ignore file

  #3514 merged May 9, 2025

• chore: miscellaneous fixes for handling graph breaks

  #3488 merged May 9, 2025

• upgrade modelopt

  #3511 merged May 8, 2025

• add python version dependency for modelopt

  #3507 merged May 7, 2025

• Fix Windows CI for Release 2.7 (#3505)

  #3506 merged May 7, 2025

• Fix Windows CI for Release 2.7

  #3505 merged May 7, 2025

• chore: remove pre-cxx11 abi references in doc

  #3503 merged May 3, 2025

• disabling dla args for hope igx platform

  #3487 merged May 2, 2025

• chore: remove pre-cxx11 abi

  #3473 merged Apr 30, 2025

• Cherrypick of PR 3462

  #3483 merged Apr 30, 2025

• fix: cherry pick PR of 3445

  #3457 merged Apr 29, 2025

• add python3.13 into the final release artifact

  #3499 merged Apr 29, 2025

• Cherry pick to release 2.7-3499-add python3.13 into the final release artifact

  #3500 merged Apr 29, 2025

• Cherrypick #3481 and #3445

  #3498 merged Apr 29, 2025

• chore: test bf16 fixes in CI

  #3491 merged Apr 29, 2025

• fix: l2_limit_for_tiling (#3479) cherry pick to 2.7

  #3493 merged Apr 27, 2025

• fix: l2_limit_for_tiling

  #3479 merged Apr 27, 2025

• Cherrypick #3424 for release/2.7

  #3484 merged Apr 27, 2025

• Cherrypick #3440 for release/2.7

  #3485 merged Apr 27, 2025

• Enabled refit on Python 3.13

  #3481 merged Apr 25, 2025

• Cherrypick of PR 3472

  #3482 merged Apr 25, 2025

• slight code reorg and bug correction for cross_compile

  #3472 merged Apr 22, 2025

• fix: index_put converter to handle multi-shape slicing with None

  #3475 merged Apr 22, 2025

• fix: support masked_scatter by lowering path and corner case of maske…

  #3476 merged Apr 22, 2025

• feat: Support flashinfer.rmsnorm

  #3424 merged Apr 18, 2025

• feat: rmsnorm lowering

  #3440 merged Apr 18, 2025

• Cherrypick #3455 for release/2.7

  #3480 merged Apr 18, 2025

• bf16 support for elementwise operation

  #3462 merged Apr 17, 2025

• chore: reenable py313

  #3455 merged Apr 17, 2025

11 Pull requests opened by 6 people

• Cross compile guard

  #3486 opened Apr 22, 2025

• Debugging guide for TensorRT

  #3489 opened Apr 23, 2025

• [Docs] Updating title for jetpack installation guide

  #3495 opened Apr 27, 2025

• Add fp4 support(WIP)

  #3496 opened Apr 29, 2025

• chore(deps): bump transformers from 4.48.0 to 4.50.0 in /examples/dynamo

  #3497 opened Apr 29, 2025

• feat: TensorRT AOT Plugin

  #3504 opened May 5, 2025

• TRT to TRT-LLM version map

  #3510 opened May 8, 2025

• chore: update the docstring for llama2 rmsnorm automatic plugin example

  #3512 opened May 8, 2025

• slice scatter support for dynamic cases [WIP]

  #3513 opened May 9, 2025

• reuse the generate binary builds from test-infra

  #3515 opened May 11, 2025

• add feature gate for tensorrt plugin

  #3518 opened May 12, 2025

9 Issues closed by 5 people

• Please Promote torch_tensorrt 2.7 release artifacts

  #3508 closed May 7, 2025

• 🐛 [Bug] Wrong output shape from cross compiled exported program

  #3400 closed Apr 29, 2025

• 🐛 [Bug] torch._subclasses.fake_tensor.MetadataMismatchError: Devices cpu and cuda:0 are not equal! (_scaled_dot_product_flash_attention)

  #3408 closed Apr 23, 2025

• 🐛 [Bug] Error Code 3: API Usage Error on IConvolutionLayer::setPaddingNd (Paligemma2)

  #3409 closed Apr 23, 2025

• 🐛 [Bug] Unsupported ops : torch.ops.aten.index_put.default - validator: False

  #3432 closed Apr 23, 2025

• 🐛 [Bug] Unsupported ops : torch.ops.aten.masked_scatter.default (Paligemma2)

  #3410 closed Apr 23, 2025

• ✨[Feature] Duplicate Subgraph `TRTEngine` Detection, Caching, and Reuse

  #2674 closed Apr 21, 2025

• ✨[Feature] Autogen TRT Plugins

  #2817 closed Apr 21, 2025

• ✨[Feature] Proper bfloat16 support in elementwise converter

  #3458 closed Apr 21, 2025

5 Issues opened by 5 people

• ✨[Feature] Key configuration quick feedback workflows for PRs

  #3501 opened May 1, 2025

• 🐛 [Bug] Difficulties Quantizing FP16 Models to INT8 Using torch_tensorrt (MLP, CNN, Attention, LSTM, Transformer)

  #3494 opened Apr 27, 2025

• ✨[Feature] Add FX tests to CI

  #3492 opened Apr 26, 2025

• ❓ [Question] Is SAM2 supported when compiling with the Dynamo backend on JetPack 6.1 or 6.2?

  #3478 opened Apr 17, 2025

• 🐛 [Bug] Encountered bug when using Torch-TensorRT

  #3477 opened Apr 16, 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.

• TRT-LLM loading mechanism tool

  #3398 commented on May 13, 2025 • 5 new comments

• Added CPU offloading

  #3452 commented on May 12, 2025 • 5 new comments

• 🐛 [Bug] quantized_resnet_test.py failed on no attribute 'EXPLICIT_PRECISION'

  #3362 commented on Apr 15, 2025 • 0 new comments

• ❓ [Question] Why does the speed (fps) of torch-tensorrt perform so badly in `torch.multiprocessing`?

  #3095 commented on Apr 15, 2025 • 0 new comments

• 🐛 [Bug] Assertion error in test_cross_compile_for_windows

  #3430 commented on Apr 15, 2025 • 0 new comments

• ❓ [Question] Running LayerNorm in fp16

  #2730 commented on Apr 25, 2025 • 0 new comments

• 🐛 [Bug] AssertionError: end must be an integer

  #3448 commented on Apr 29, 2025 • 0 new comments

• ✨[Feature] Is there a plan to support to convert quantized PT2 to trt ?

  #3471 commented on Apr 29, 2025 • 0 new comments

• ❓ [Question] How wo you export a triton kernel with model to a serialized engine that can be run in c++?

  #3469 commented on May 7, 2025 • 0 new comments

• 🐛 [Bug] Expected input tensors to have type Half, found type float

  #2113 commented on May 10, 2025 • 0 new comments

• chore(deps): bump transformers from 4.40.2 to 4.48.0 in /tests/py

  #3412 commented on Apr 29, 2025 • 0 new comments

• Added flux demo

  #3418 commented on May 13, 2025 • 0 new comments