MODEL_NAME, ARCH vs different backbones

Hi guys!
First of all, thank you for this amazing repo!
I have questions regarding terminology, cause I got slightly confused by the names `MODEL.ARCH` and `MODEL.MODEL_NAME` and this part of the code:
```
# Supported model types
_MODEL_TYPES = {
"slowfast": SlowFast,
"slow": ResNet,
"c2d": ResNet,
"i3d": ResNet,
"slow_c2d": ResNet,
"x3d": X3D,
"mvit": MViT,
}
```
Based on the description I thought that I could choose the backbone from any of the mentioned above in SlowFast architecture and then e.g. I can have SlowFast architecture with an X3D backbone or I have SlowFast architecture with MViT backbone to extract features?

But what if I choose SlowFast as a backbone? Then I have SlowFast architecture with SlowFast backbone, but what is the backbone of my backbone then - is it ResNet as described in the article or does it work in a different way? I don't really get this part... Or I just choose MViT and it's in no way connected to SlowFast and if I want I can compare it to SlowFast, but why the backbone then?

Second of all, what is the classifier at the end? Is it e.g. a sigmoid-based classifier layer, as it was written in the paper? Is it specified by HEAD_ACT?

Lastly, I wanted to ask why the `MODEL_NAME` for `ARCH`s I3D, C2D is ResNet, and for SlowFast is SlowFast, but maybe the answer to the first question clarifies it all.

I really appreciate any help!