keras.json is in .keras in your home directory %USERPROFILE%
if that’s not the issue, i’m not sure why it’s not using your GPU, so yeah, I would reinstall everything besides visual studio. Going through it might refresh to you how it stacks.
Shawn,
I want to give a little bit of try before going for full reinstall. I find that keras.json is available in my home dir but .theanorc is missing.I think this is why the GPU is not being used. Is it ok if I create a .theanorc as a JSON file with the content shown in the lesson 1 video.
I don’t have theanorc in my filesystem I’m fairly sure, and neither did you previously. So, not sure if it would help, but can’t hurt to try. The other thing I would be then suspicious of in your setup is the environment variables. Make sure you’re passing the right flags to theano. You could also try docker for simpler startup, think someone on another thread here has made an image.
This is from some guide I didn’t use:
"After installing the python libraries you need to tell Theano to use the GPU instead of the CPU. A lot of older posts would have you set this in the system environment, but it is possible to make a config file in your home directory named “.theanorc.txt” instead. This also makes it easy to switch out config files. Inside the file put the following:
[global]
device = gpu
floatX = float32
[nvcc]
compiler_bindir=C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\bin
I redid the installation. As you said it was an issue with env variables. However I find now I get No module named bcolz on running the note book. Last time I had resolved it by conda install bcloz. Same is not able to install bcloz. What am I missing?
pretty sure it’s bcolz 
Shawn,
Sorry for bothering you again. Now the NB is working fine. The CPU to GPU switch through keras and Theano from bash is successfully happening.
But I want to know how we can be sure that GPU is being used when we run the notebook?
For each epoch, with sample data the time taken is approx 25 sec. With the main data it is taking forever, the ETA shown is around 7 Hours. Is it reasonable if GPU is being used?
If you type import theano at a Python prompt the output will tell you if the GPU will be utilised.