facebookresearch / facebookresearch/SlowFast
Confusion related to precise_bn step
Open
- Dominant language
- Python
- Stars
- 7.4k
- Forks
- 1.3k
- PR merge metrics
- No merged PRs in 30d
Description
Hi,
First of all, thanks for this wonderful repo.
I am taking inspirations from your repo and building my own on custom dataset and models. I have a doubt related the the `precise_bn` step. Lets say, I am not using `distributed dataparallel` and using `dataparallel`. So, in this case do I need to call `misc.aggregate_sub_bn_stats(model)`? As, I don't need to aggregate explicitly when using 'dataparallel'.
Thanks in advance.
Contributor guide
Assessment
This issue has not been assessed yet.