Closed mstribitaka closed 6 years ago
Hi,
thanks for your question. I'm running on a TitanX and it takes 50 minutes per epoch. There's currently a bug with DataParallel which prevents multi-gpu training, but once that's solved training time should go down!
Update : @mstribitaka I had a small memory overlap in the testing loop, you should now be able to double the batch size
I'm gonna close this issue, as training time is reasonable on multi gpu. I tested on pytorch 0.2 and speed seems proportional to the amount of gpus available.
Hey there,
Been trying out your code and just wondering what GPU card you're using and how long it takes? I'm using a GTX 1080 and it's taking around 1 hour for 1 epoch. Does that sound reasonable to you?
We had to reduce the batch size to 64 otherwise it ran out of memory.
Thanks!