This PR adds a random time-based sampler, and also updates our benchmark.
It must take a num_clips parameter instead of seconds_between_clip_starts like in clips_at_regular_timestamps().
This means the _generic_time_based_sampler takes both num_clips and seconds_between_clip_starts as parameter, and they're mutually exclusive. That is not user-facing.
The _implem.py file is growing large. When this PR is merged, I think it'll be time to re-organize and split the implementation in different files. Typically the policies could be kept separate.
This PR adds a random time-based sampler, and also updates our benchmark.
It must take a
num_clips
parameter instead ofseconds_between_clip_starts
like inclips_at_regular_timestamps()
.This means the
_generic_time_based_sampler
takes bothnum_clips
andseconds_between_clip_starts
as parameter, and they're mutually exclusive. That is not user-facing.The
_implem.py
file is growing large. When this PR is merged, I think it'll be time to re-organize and split the implementation in different files. Typically the policies could be kept separate.