Improve GPU performance about tensorflow-wavenet HOT 6 OPEN

tensorflow/tensorflow#1136

from tensorflow-wavenet.

Just compared GPU and CPU again, and this time training is significantly faster on the GPU.

from tensorflow-wavenet.

Is this solved then?

from tensorflow-wavenet.

We would probably be able to train larger networks (and with larger sample sizes) if we had a single C++ op that does causal 1d convolution, as we would be able to avoid all the extra copying that's currently being done.
But writing and maintaining the op might be a pain :/

from tensorflow-wavenet.

I've done it before, but not with gpu support...

On Thu, Oct 6, 2016, 18:58 Igor Babuschkin [email protected] wrote:

We would probably be able to train larger networks (and with larger sample
sizes) if we had a single C++ op that does causal 1d convolution, as we
would be able to avoid all the extra copying that's currently being done.
But writing and maintaining the op might be a pain :/

—
You are receiving this because you commented.

Reply to this email directly, view it on GitHub
#25 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/ADCF5gay3mfBQ0wKyF2mamL0xFbu5kq9ks5qxX0mgaJpZM4J_k-r
.

from tensorflow-wavenet.

@ibab I've made a pull request for mixed precision training
#384

and would like to encourage the authors and community to give it a try :)

from tensorflow-wavenet.