Episode 8: Autopilot

Scheduling Compute in the Cloud.
This example revisits 'Episode 06-statistics-redux: Computing in the Cloud'. With Metaflow, you don't need to make any code changes to schedule your flow in the cloud. In this example, we will schedule the stats.py workflow using the step-functions create command-line argument. This instructs Metaflow to schedule your flow on AWS Step Functions without changing any code. You can execute your flow on AWS Step Functions by using the step-functions trigger command-line argument. You can use a notebook to set up a simple dashboard to monitor all of your Metaflow flows.
You can find the tutorial code on GitHub.
Showcasing:
step-functions create command-line option
step-functions trigger command-line option
Accessing data locally or remotely through the Metaflow Client API​
Before playing this episode:
1.
python -m pip install notebook
2.
python -m pip install plotly
3.
This tutorial requires access to compute and storage resources on AWS, which can be configured by 
1.
Following the instructions here or 
2.
Requesting a sandbox.
To play this episode:
1.
cd metaflow-tutorials
2.
python 02-statistics/stats.py --environment=conda --with conda:python=3.7,libraries="{pandas:0.24.2}" step-functions create --max-workers 4
3.
python 02-statistics/stats.py step-functions trigger
4.
jupyter-notebook 08-autopilot/autopilot.ipynb
5.
Open autopilot.ipynb in your remote Sagemaker notebook.

Episode 7: Worldview

Next - Metaflow on AWS

Metaflow on AWS

Last modified 3mo ago

Export as PDF

Copy link