Support exponential moving averages for the covariance statistics on EigenReporter
Open
enhancement
- Dominant language
- Python
- Stars
- 225
- Forks
- 30
- PR merge metrics
- No merged PRs in 30d
Description
Currently `EigenReporter` supports streaming updates to its covariance statistics via the `update` method, but it does an _unweighted_ average. For the training use-case, we'll want to use an exponential moving average since the representations of the model will change over time.
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.