Fast.ai is one of the most popular DeepLearning MOOC. Course students need access to a GPU machine with NVIDIA drivers, anaconda environment with lots of pre-requisite packages. Most lessons are based on interactive Jupyter Notebooks.
Clouderizer team has come up with a community project template to automate the setup, download latest code and dataset, and enable secure access to Jupyter Notebooks and SSH terminal from Clouderizer console.
This project template can be started on any Ubuntu machine. We can run it on a local machine or any cloud machine (AWS, GCP or Paperspace). Since entire project data (code + dataset + output) is synced in real time with Clouderizer Drive, we can also, at any point, seamlessly switch between local and cloud machine as well.
Clouderizer account can be linked with your AWS account, which enables us to spin GPU Spot instance (p2.xlarge – Tesla K80 GPU – ~0.35USD / hr) with fast.ai project template in one click.
Similar integration with GCP and Paperspace is in works and will be available soon.
Without further ado, here are the steps to launch fast.ai project template.
- Sign in to your Clouderizer account. (Sign up here if you don’t have one already)
- Go to Community Project from left menu.
- Search fastai template and press Clone to copy this project in your account.
- This will start New Project Wizard with all fields pre-populated. You can give this a name of your choice.
- Press Next through the Name, Setup and Workspace screens (all setup and workspace instructions are pre-populated based on latest instructions from fast.ai site)
- On Machine screen, you will be asked configure the machine type. By default, this project template uses AWS p2.xlarge machine (with Tesla K80 GPU) spot instance. You can press refresh icon next to Spot Recommendation to get best SPOT bidding price recommendation across AWS regions.
In case you have not already linked your AWS account with Clouderizer, please follow instructions here to do so.
- Press Next and then Close to complete the project creation.
- You should now see FastAI project listed under Projects page. Press Start to launch the machine. This will submit the spot request to AWS and wait for machine to get allocated. Once machine is successfully allocated, you should see project status as Starting. All pre-requisite packages, drivers, latest course material, datasets will be downloaded and installed during this time. This should take around 20 mins.
- Once your machine is ready, project status should change to Running.
- You can now press Jupyter on project card, to launch Jupyter Notebook. On Jupyter Notebook, navigate to courses -> dl1. You can see all Fast AI course Part I course notebooks here.
- Click on lesson1.ipynb to view lesson1 notebook.
- Step through each block by pressing Run button on top toolbar to verify that environment is setup properly and you are able to execute this notebook without any error.
- Thats it. You are all set for an awesome DL experience. Happy DeepLearning!
In case you have any questions, feel free to ask in our community here. We will be more than happy to help.