Python with Google Colab

Google Colab provides a jupyter environment based on environment and hosted hardware accelerators by google.

Basic features are explained in this notebook.

GPU Lineup

  • Tesla T4: light training, inference and mixed precision
  • NVidia L4: best for modern LLMs, bfloat16 () support and mid-range training
  • A100: best for deep learning, large datasets and math
  • H100: state-of-the-art transformer training and extreme speed.

Interaction with google drive

To mount from google drive that is useful to access datasets:

from google.colab import drive
import sys
import os
 
# Mount your Google Drive
drive_path="/content/drive"
drive.mount(drive_path)
 
file_path = "/content/drive/MyDrive/MyLabs/"
 
if file_path not in sys.path:
    sys.path.insert(0, file_path)
    

Python’s pip can specify a “target” where the packets are downloaded and builded, to avoid to reinstall them later.