pymc-devs / pymc-devs/pytensor
Transform graph to make large constants symbolic inputs
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 644
- Forks
- 208
- Avg merge
- 2d 14h
- Merged PRs (30d)
- 16
Description
Description
JAX jitting can be insanely slow when there are large constants in the graph. We could add a helper to convert any large constants to symbolic inputs (we already did some constant folding work on our end anyway), so JAX can't get hang up on those.
See related discussion on their side: https://github.com/jax-ml/jax/issues/21300
The idea is to have a pytensor.graph.replace.replace_large_constants_by_inputs that returns the graph with constants replaced by PyTensor input variables and the respective values
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 at the proposed pytensor.graph.replace.replace_large_constants_by_inputs entry point and review the related JAX discussion to understand the large-constant problem. Define completion as returning a graph where large constants become PyTensor input variables together with their corresponding values; the issue does not name a specific test file.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100