here is my python code implementing compute_weights function #5
Conversation
| a = [0.0 , 0.0] | ||
| i = pos | ||
| nu = 0.0 | ||
| while i < n and nu < 1e-5: |
There was a problem hiding this comment.
Oh, the main reason for implement compute_weight in C++ is that python's loops are too slow and this loop will be a too slow for large batch size. Can you provide performance report which show that python code isn't slower than c++ implementation?
There was a problem hiding this comment.
Good point. However i haven't compared yet. Thanks for your patience again.
There was a problem hiding this comment.
I cann't run build.py. So i try the python version , but it doesn't work. Have you succeed?
There was a problem hiding this comment.
@01lin currently this implementation is not compatible with pytorch 0.5+
There was a problem hiding this comment.
@belbes With pytorch 0.4.0 and python3, I also cann't run build.py. And the error is " cffi.error.VerificationError: CompileError: command 'gcc' failed with exit status 1". Can you help me?
3 participants
No description provided.