Assignment

Task 1: Tutorial of using Pytorch and Optuna to create a NN classifier for the dataset USPS

Summary: A walkthrough of how to read-in data, split it into training-validation, create and train a NN using pytorch, and fine tune such using Optuna.

Task 2: Fine tune NNs for different datasets

Summary: Applying the techniques from task 1 onto different datasets.

Submission

Upload the zipped folder of your assignment folder, which should contain the work you have done for Task1 and Task2

Within the zipped folder, it should contain:

• jupyter notebook of Task 1 and Task 2 with completed TODOs and your code
• html of your finalised jupyter notebook for Task 1 and Task 2
• pickle files from Task 2

Setup

With Linux/ MacOS

1. Install Conda
2. Create a conda environment from the provided pytorch.yml

With Windows

1. Install Anaconda

2. Open Anaconda Prompt

3. Navigate to the folder where pytorch_windows.yml file is

cd C:\\User\\folder_where_yml_is

4. Install pytorch environment from pytorch_windows.yml

5. Activate pytorch environment

conda activate pytorch

6. Install pytorch

6.1. CPU

conda install pytorch torchvision torchaudio cpuonly -c pytorch

6.2 CUDA

conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch

7. Open Jupyter Notebook

jupyter notebook

Appendix

Installation of Conda

Installation of Python

Installation of Jupyter Notebook

Installation of Pytorch

Installation of Optuna