Deep Learning AMI with Source Code (CUDA 9, Amazon Linux)
Amazon Web Services | 5.0Linux/Unix, Amazon Linux 2017.09 - 64-bit Amazon Machine Image (AMI)
gpu is not supported in screen sessions
If you activate the tensorflow_p36 environment, and then run a python script that uses tensorflow, you see that the gpu is utilized. However, if you start a screen session, then activate the tensorflow_p36 environment, running the same script will only run on the CPU. Very aggravating when you need to train large scale models that take days (requiring the use of screen/tmux)
- Leave a Comment |
- Mark review as helpful
Looks promising
Excited to start working with P3 instances with their increased GPU performance. Eagerly awaiting Keras support! Thanks for developing these AMIs.
Thx, still not yet fully operational
Launching went smoothly, still:
> We definitely miss Keras 2.
> Mentioned README.md is not where you mentioned. It is located here: /home/ec2-user/AmazonLinuxCuda9.README.md
> Met some problems updating python2 packages (force to manually remove pip from /usr/lib/python2.7/dist-packages/ and reinstall it with the traditional sudo python2.7 get-pip.py
You rock, waiting for AMI updates!