google-deepmind / google-deepmind/alphageometry
Enabling multithreading
Open
- Dominant language
- Python
- Stars
- 4.9k
- Forks
- 572
- PR merge metrics
- No merged PRs in 30d
Description
In the AlphaGeometry Nature paper, the authors use up to 250 CPU workers to run AlphaGeometry. However, when I run AlphaGeometry, only one thread is used.
How can I use multiple threads? It seems like DD-AR is often a bottleneck, rather than the LLM, when a GPU is used.
Contributor guide
Assessment
This issue has not been assessed yet.