lsy643 opened a new pull request #6108:
URL: https://github.com/apache/incubator-tvm/pull/6108
In this PR, the CUDA compute funtions of `get_valid_counts` and `nms` are
changed to make them work as expected.
1. For `get_valid_counts`, only one thread is used for one image. I am not
sure whether this is a good way
2. For `nms`, there are two changes
2.1 make `box_indices` to map back to the original data indices
2.2 create `rearrange_indices_out` for `nms` when `return_indices == True`
3. Test cases for gpu version of`get_valid_counts` and `nms` are enabled now
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
