Google Colab : A Poor Person’s GPU
So you’ve already done few Machine Learning projects & courses etc., and are just breaking into deep learning. One of the biggest problem you will face is that of training your deep learning models on GPU.
You laptop however awesome it may be, most likely will not have the required GPU capability, so then you will start looking into cloud GPU services like Amazon Web Services, Google Cloud Platform, FloydHub, etc., etc.
These are great options, but the two main challenges you’ll run into are:
- Configuration: Configuring your project / data with cloud GPU is a pain in the rear.
- Cost: Cost associated with training your model on cloud GPU racks up soon and is heavy on your wallet.
Google Colaboratory is the answer for your personal project’s deep learning models. Its FREE and requires almost no configuration at all.
Step 1: Adding the Colaboratory App to your Google Drive
Login to your Google Drive, and
- Click on NEW → MORE → CONNECT MORE APPS
- Type Colaboratory in search bar, and click CONNECT
Step 2: Launching a Jupyter Notebook With Python 3 & GPU
Assuming you followed all the steps above, you should see a Jupyter Notebook open up. Then follow the steps below.
- Click on EDIT → NOTEBOOK SETTINGS
- Select PYTHON 3 as Runtime Type & GPU for Hardware Accelerator
- Click CONNECT
Step 3: BE AWESOME, and share the knowledge with others
Following all the steps, you should now have a decent yet powerful GPU for training your deep learning models.
Also, here is a link to Google Colab FAQs: https://research.google.com/colaboratory/faq.html
Please do follow my publication Life @ Digital Health for more stories.