altavir opened this issue 3 years ago · 0 comments
Default tensor dot product operation is too slow. Like 100 times slower than other implementations.