facebookresearch / facebookresearch/SlowFast
Changing Attention.py to be compatible for the pre-trained models
Open
- Dominant language
- Python
- Stars
- 7.4k
- Forks
- 1.3k
- PR merge metrics
- No merged PRs in 30d
Description
Is it correct to do this and change the Pool KV kernel of 14, 15 layers (Since they are of scale 5 as mentioned in the paper) to get the pretrained model working? Or will the authors be uploading new models that are compatible with kqv kernels?
Contributor guide
Assessment
This issue has not been assessed yet.