Using Google Colaboratory

Google Colaboratory (or Google Colab) is an online computing service provided by Google. In essence, Google Colab is an online Jupyter notebook with GPU/CUDA capability.

Its free notebook can provide CUDA computing with 15 GB VRAM equivalent to an NVIDIA RTX 3050 or RTX 3060. The performance is decent if you don’t have a discrete GPU at hand.

Let’s look at the advantages and disadvantages of using Google Colab:

• Advantages:
  • No need to manually install CUDA and Python
  • Everything is online; you can save a link and reopen it anywhere
  • The installation of pip and downloading of resources are fast
• Disadvantages:
  • There is a disk limitation for each notebook
  • You don't have full control of the backend server; terminal access requires a Colab Pro subscription
  • The performance is not guaranteed so you may experience slow GPU inference during peak time and could be disconnected for long-time computing
  • The Colab notebook compute environment...