Running fast.ai notebooks with Amazon SageMaker - blog post

Hi all,

If anyone is interested in running fast.ai with Amazon SageMaker notebook instances I have a blog post here detailing the setup steps: https://aws.amazon.com/blogs/machine-learning/running-fast-ai-notebooks-with-amazon-sagemaker/

Regards,
Matt McClean

12 Likes

Hi Matt!

Thank you for spearheading the solutions flow for making fast.ai notebooks setup on AWS SageMaker easier for those who are starting on the path!

WHY I WROTE THIS

Recently, a few people faced issues with getting set up on SageMaker, presumably due to an bug in the FastaiSageMakerStack template dependencies on conda env ver updates (resulting in conda env inconsistency manifested in unavailability of fastai kernel for running notebooks).

RELEVANT LINKS

The current stack template is located here: https://fastai-cfn.s3.amazonaws.com/sagemaker-cfn-course-v4.yml

The list of CloudFormation stack templates with links is here: Amazon SageMaker | Practical Deep Learning for Coders

The issue and the solution that has worked is documented here: Sagemaker Notebook deployment problem - no FastAi kernel - #10 by deep_learner

QUESTIONS AND ACTIONS

If this, indeed, is a bug, I would like to properly document it as an issue and ultimately submit a pull request to adjust the stack template, which should save a few people headaches.

The most relevant/latest points I could find on this issue on github were your code commits below, yet they seem outdated (ver 3 of the course):

Did I miss something? If you could kindly point me to the right direction to start (including the right spot on github to work on this and general advice re documenting the issue and submitting the PR without wasting anybody’s time) it would be great!

Thank you!

Best
PO