microsoft / microsoft/onnxruntime
Improve the way of specifying used operators when using "--minimal_build" option
Nobody has claimed this yet.
- Dominant language
- C++
- Stars
- 21.9k
- Forks
- 4.2k
- Avg merge
- 4d 8h
- Merged PRs (30d)
- 179
Description
Is your feature request related to a problem? Please describe.
For building with --minimal_build and --include_ops_by_config required_operators.config, if the model is updated, the whole onnxruntime library should be re-build from scratch, which takes more than 1 hour in my case. This isn't very pleasant and makes the build/debug cycle too long.
System information
- ONNX Runtime version (you are using): 1.6.0
Describe the solution you'd like
We need another way to specify which operators are used. For example, one way to do that would be to specify the user's code, not while building the library itself.
Describe alternatives you've considered
Tensorflow Lite has this kind of way to specify which operators are used.
Additional context
Add any other context or screenshots about the feature request here.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by tracing how --minimal_build and --include_ops_by_config required_operators.config are handled during the ONNX Runtime build. Define the user-code-based operator specification approach and verify that updating a model no longer requires rebuilding the entire library from scratch.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- cpp
- Domain
- build-system
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100