why use the `width_per_group` as the `dim_in`

Hi,
Thanks for the open-resource work.

I wonder why use the `width_per_group` as the `dim_in`?
Do you think it is more reasonable to use `width_per_group * num_groups` as `dim_in`, although this two are alway same when `num_groups=1`.

Or do I miss something?

Thanks in advance.

https://github.com/facebookresearch/SlowFast/blob/bda005363ff589405e19b0ed0ba063eb0752ec97/slowfast/models/video_model_builder.py#L463-L465