These are chat archives for arita37/tensorflow

5th
Apr 2017
Michael Phillips
@rndapology_twitter
Apr 05 2017 03:36
Hi guys, I'm just trying to write a quick script to check my jpegs for size/validity before I feed them into a convnet, but .shape returns rubbish - I can't see what I'm doing wrong
https://pastebin.com/vPxyxX1e
the shape is wrong pretty much always, returns dimensions that none of my images even has
on different runs of the script the same file appears to return different dimensions via .shape
Fei
@XuefeiW
Apr 05 2017 09:22
Does anyone use Google Cloud ML Engine for training Deep Learning using GPUs? When I set the scale tier (BASIC_GPU) and region (europe-west1) I keep getting error on GPUs(no details of what's the error), and the training finished successfully on CPU. I am using gcloud beta and runtime version of Tensorflow is 1.0. Apart from setting scale tier and region to enable GPU service, is there any other settings we need to do before using GPUs? Thanks.