NVIDIA / NVIDIA/TensorRT

ONNX to TensorRT conversion fails due to GridSample

Open
#4,666 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Module:ONNX
Dominant language
C++
Stars
13.4k
Forks
2.4k
Avg merge
5d 3h
Merged PRs (30d)
2

Description

Description

I am trying to convert a model from PyTorch to ONNX, which was successful. However, when I convert the ONNX to TensorRT, an error occurs because the older version of TensorRT doesn't support GridSample. Newer versions of TensorRT solve the above issue, but I am limited to using the current version, 8.2.1.8.

I tried installing mmdeploy and adding the libmmdeploy_tensorrt_ops.so operator library file to the trtexec --plugin command (https://github.com/NVIDIA/TensorRT/issues/2756), but I'm still getting the same error.
Is there a solution to this problem? Or a guide on how to add the GridSample operator to the aforementioned version of TensorRT?

Environment

TensorRT Version: 8.2.1.8

CUDA Version: 10.2

Operating System: Jetpack 4.6.1 [L4T 32.7.1]

Python Version: 3.6.15

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reproducing the ONNX-to-TensorRT conversion with TensorRT 8.2.1.8 on JetPack 4.6.1, then inspect the trtexec --plugin invocation and the libmmdeploy_tensorrt_ops.so library. A useful resolution would document whether GridSample can be supported on this version or identify a compatible workaround.

Written by the indexing model from the issue text.

Assessment

Tech stack
python, pytorch
Domain
machine-learning
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.