For inference, how do we control the clip size

Hi,

I have a long video I want to do inference on. I want pyslowfast to do inference every N seconds, or every M frames. How do I control that in the configs yaml file? And how is this related to the config options, if it is? What do the config options really mean for the input video/dataset?

```
# Number of overlapping frames between 2 consecutive clips.
# Increase this number for more frequent action predictions.
# The number of overlapping frames cannot be larger than
# half of the sequence length `cfg.DATA.NUM_FRAMES * cfg.DATA.SAMPLING_RATE`
```
_C.DEMO.BUFFER_SIZE = 0

`# The number of frames of the input clip.
`_C.DATA.NUM_FRAMES = 8

`# The video sampling rate of the input clip.
`_C.DATA.SAMPLING_RATE = 8

Does it mean that with a buffer_size of 16, the demo waits until it has 16 frames, makes a person prediction on the center frame, then uses the predicted bounding boxes from center frame, and the 16 frames for action prediction?