google-deepmind / google-deepmind/kinetics-i3d
about batch size
Open
- Dominant language
- Python
- Stars
- 1.8k
- Forks
- 467
- PR merge metrics
- No merged PRs in 30d
Description
Hi,
In the paper _"Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset"_, it is said you used 64 GPUs to train 3D ConvNets.
But in readme, you said minibatch size is 6.
Does it mean that during multi-gpu training, the batch size is 6*64 = 384 ?
Thanks！
Contributor guide
Assessment
This issue has not been assessed yet.