microsoft / microsoft/onnxruntime

Why CUDAExecutionProvider doesn't support ConvInteger Op?

Open
#5,127 5 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

ep:CUDA feature request
Dominant language
C++
Stars
21.9k
Forks
4.2k
Avg merge
4d 11h
Merged PRs (30d)
184

Description

**Is your feature request related to a problem? Please describe.**
When I want to execute ConvInteger quant comptation on GPU, I can't find support operation.

**System information**
- ONNX Runtime version (you are using): 1.4

**Describe the solution you'd like**
Support cuda version ConvInteger Op.

**Describe alternatives you've considered**
1.ConvInteger is a very common operation on cpu, gpu, arm and so on, I am little confused why CUDAExecutionProvider doesn't support ConvInteger Op.
2.Although I can use MatMulInteger and Img2col to complish the same work as ConvInteger can do, the former method is too slow.

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 tracing ConvInteger support in the CUDAExecutionProvider and compare it with the existing CPU, MatMulInteger, and im2col paths described in the issue. Done means CUDAExecutionProvider can execute ConvInteger correctly and without the slow workaround; the issue names no files or tests, so locating suitable coverage is part of the investigation.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.