Summary: 37 instances, 31 unique

Text	Count
// TODO remove the dependency on input and use instead its sizes -> save memory	2
# TODO use size instead of scale to make it robust to different sizes	1
# TODO Ideally, just remove this and let me model handle arbitrary	1
# TODO think about a representation of batch of boxes	1
# TODO maybe add an API for querying the ws / hs	1
# TODO resolve this difference and make it consistent. It should be per image,	1
// TODO raise error if not compiled with CUDA	1
# TODO maybe push this to nn?	1
TO_REMOVE = 1  # TODO remove	1
# FIXME ideally this would be achieved with a CombinedLRScheduler,	1
THCState *state = at::globalContext().lazyInitCUDA(); // TODO replace with getTHCState	1
# TODO make it pretty	1
# TODO add squeeze?	1
# TODO might be better to add an extra field	1
// TODO add more checks	1
# TODO need to make the use_07_metric format available	1
# TODO: specify init for the above	2
//TODO: larger threads-per-block might be better here, because each CTA uses 32 KB of shmem,	1
# TODO check if want to return a single BoxList or a composite	1
# TODO should I filter empty boxes here?	1
TO_REMOVE = 1  # TODO remove	1
# TODO rename x to roi_box_features, if it doesn't increase memory consumption	1
# TODO redundant, remove	1
# TODO chck if necessary	1
# TODO:  Is this JIT compatible?	1
// TODO make it in a common file	2
# TODO replace with get_img_info?	3
torch.cuda.empty_cache()  # TODO check if it helps	2
# TODO maybe remove this and make it explicit in the documentation	1
# FIXME remove this once c10d fixes the bug it has	1
// TODO improve this part	1