Hello, we encountered a test failure on Windows which caused memory corruption. This resulted in spilled over test failures, causing further CUDA kernels to fail to launch [1]. Thanks to Dick Carter who debugged the issue and tracked it down to a bad implementation of SequenceLastKernel, leading to out-of-bound errors. Is there anybody around who's familiar with SequenceLastKernel and able to help us fix this bug? Further details are available in the linked issue.
Best regards, Marco [1]: https://github.com/apache/incubator-mxnet/issues/11395
