This repository has been archived by the owner on Nov 17, 2023. It is now read-only.
Faster GPU NMS operator #16542
Merged
Faster GPU NMS operator #16542
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
zhreshold
reviewed
Oct 19, 2019
Kh4L
reviewed
Oct 23, 2019
|
@zhreshold Do you have any other comments? For the record - I think there are a few outstanding items to fully mark GPU implementation of NMS as "done", namely
That said, I will not be able to finish those 2 outstanding items before 1.6 code freeze and I believe performance improvements are enough to merge this PR and address those 2 points as part of the future PR. |
apeforest
reviewed
Nov 1, 2019
|
|
||
| template <typename DType> | ||
| __global__ void FilterAndPrepareAuxDataKernel(const DType* data, DType* out, DType* scores, | ||
| index_t num_elements_per_batch, |
There was a problem hiding this comment.
nit: indentation alignment?
apeforest
reviewed
Nov 1, 2019
|
|
||
| template <bool check_topk, bool check_score, typename DType> | ||
| __global__ void CompactDataKernel(const index_t* indices, const DType* source, | ||
| DType* destination, const index_t topk, |
There was a problem hiding this comment.
nit: indentation alignment?
zhreshold
approved these changes
Nov 2, 2019
yajiedesign
pushed a commit
to yajiedesign/mxnet
that referenced
this pull request
Nov 6, 2019
* Adding second NMS op * NMS kernel * Removing second sort * Optimization * Adding out-of-place ability to SortByKey * Optimization pt2 * Optimizations pt3 * Do not recompute other boxes area every time * Sort only topk results during second sorting * Cleaning * Fixes from rebase * Fix lint and more fixes from rebase * Fix typo * Early exit in Triangle kernel * Fixes * Fix sort * Fix from rebase * Fix for the mixed naming convention * Fix the index_t with int comparisoon
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
This PR significantly improves the performance of
mx.sym.contrib.box_nmson GPU.@zhreshold @Jerryzcn FYI
Test cases:
Checklist
Essentials
Please feel free to remove inapplicable items for your PR.
Comments