feldberlin / wavenet

An unconditioned Wavenet implementation with fast generation.
3 stars 0 forks source link

Channels Last #30

Closed purzelrakete closed 3 years ago

purzelrakete commented 3 years ago

What

Switch to channels last as described here.

Hypothesis

I expect a 20% speed increase, because I am on half precision tensor cores.

Results

Write up the results of your experiment once it has completed and has been analysed. Include links to the treatment run, and also to the baseline if appropriate.

Acceptance Criteria

purzelrakete commented 3 years ago

Channels last is not implemented for conv1d: https://github.com/pytorch/pytorch/wiki/Operators-with-Channels-Last-support. Closing ticket.