colabutilscs4337

Google Colab utilities for CS4337 (computer vision with PyTorch and fastai). One import gives you short, safe helpers for the chores around every assignment:

  • Setup: install extra tools without breaking Colab’s preinstalled PyTorch and fastai, load API keys from Colab Secrets, check the runtime.

  • Runs and data: mount Google Drive and get a fresh, timestamped run folder per experiment.

  • Look at your data: image grids and one-row-per-class views of any dataset.

  • Look inside your model: torchinfo summaries, the Netron graph viewer, loading fastai and PyTorch checkpoints.

  • Explain predictions: Grad-CAM heatmaps, layer activations, feature maps.

%pip install -q --upgrade colabutilscs4337
import colabutilscs4337 as cu

cu.env.load_secrets("WANDB_API_KEY")     # from the Colab Secrets panel
run = cu.drive.new_run("baseline")       # MyDrive/runs/<notebook>_<time>_baseline/
cu.viz.show_by_class(train_ds)           # one row of examples per class
cam = cu.viz.gradcam(model, x)           # where is the model looking?
cu.viz.show_cam(x, cam)

Getting started

User guide

API reference

Project

Links