The notebook titled 'Assignment_2A' contains code for part A, including training a CNN model from scratch in Tensorflow and Keras and running wandb sweep. To test the code, run each cell one by one. wandb commands have been commented to do a trial run without logging in wandb. The model has been trained after the Train function with the best configuration to evaluate on the test set.
In the notebook titled 'Assignment_2A_guided_backprop.ipynb' contains code for training the model with best set of hyperparameters and implementing guided backpropagation and visualizing filters in the 1st convolution layer.
The notebook titled 'Assignment_2B_Wandb.ipynb' contains code for part B. It contains a function to run sweep for the pretrained models. Run each cell one by one to do a trial run and the wandb commands to avoid logging in wandb. The model has been trained once after the Train function with the best paramters to evaluate on the test set.