google-deepmind / google-deepmind/alphafold
differences between multimer v1/v2 and v3
Open
usage question
- Dominant language
- Python
- Stars
- 14.9k
- Forks
- 2.9k
- PR merge metrics
- No merged PRs in 30d
Description
Another thing that got introduced in v3 is bfloat16 (which I think changes the results when used in v1/v2)
To be consistent, I think this should be included in:
https://github.com/deepmind/alphafold/blob/main/alphafold/model/config.py#L132
```python
common_updates = {"model.global_config.bfloat16":False}
```
Contributor guide
Assessment
This issue has not been assessed yet.