Sign in
Migration Mapping Assistant Your Saved List Partners Sell in AWS Marketplace Amazon Web Services Home Help

ML Workbench for TensorFlow

Starting from $0.03/hr or from $150.00/yr (43% savings) for software + AWS usage fees

Ubuntu 18.04 with Jupyter, JupyterLab, TensorBoard and preconfigured conda environments for Tensorflow 1.13.1 and TensorFlow 2 Alpha including the latest matching versions of CUDA 10.0 and cuDNN 7.5.0 for GPU-accelerated computing. There is no need for an SSH session to launch a notebook. All services,... See more

Customer Reviews

  • 1
  • 4 star
  • 3 star
  • 2 star
  • 1 star
Create Your Own Review

It's a snap ...

  • By Bigvulcandeal
  • on 01/25/2019

This really is a zero-admin solution. I had to develop a way to give a bunch of architecture and design students a hands-on demo of how to use AI., and have them run the lab in under an hour.

The basic idea is that creating something like a rudimentary image classifier is so conventionalized that they can do it, first time, in under an hour .. even if their technical skills are pretty modest.

Of course, this only holds true if there are no hitches setting up the machine and installing the requisite software etc. In fact, setup would probably take well over an hour for someone who is highly skilled.

Then I foresaw issues having the students see what is going on .. even if I set up VNC or some such, some of them were apt to struggle .. basically, I was teaching Tensorflow for poets to folks who are almost literally poets. I needed something that was ready to rock, and simple.

Enter this netcubed deeplearning AMI. It's perfect. I bought a subscription, created a development machine to test my approach, and I'm planning to mint a bunch of machines for the class. Then I will just hand out URLs and passwords, and they'll be off and running.

I did need to one thing to get tensorboard to run. Apparently the machine wants the log files to go in a predetermined folder in order to have the tensorboard client work properly. The tech support folks had me do the following in the terminal:

# move to the home directory
cd ~
# remove tensorlogs folder
rm -rf ~/tensorlogs
# make tensorlogs folder a symlink where tf for poets log files actually go
ln -s ~/tensorflow-for-poets-2/tf_files/training_summaries tensorlogs

In any case, great product, great service. The email tech support was extremely responsive. The template works beautifully when instantiated using a t3.medium instance .. "TF for Poets" ran right out of the box..

showing 1 - 1