PyTorch101

A hands-on, notebook-based course covering PyTorch from tensor fundamentals to transfer learning. All notebooks run on standard benchmarks and are developed in Kaggle before syncing to this repository. Target audience: computational biologists and bioinformatics students who want a practical deep learning foundation.

Curriculum

Notebook	Topic	Standard Dataset	Open in Kaggle
01_fundamentals.ipynb	Tensors, Autograd, nn.Module, Training Loops	Synthetic
02_regression.ipynb	Tabular Regression with MLPs	California Housing
03_classification.ipynb	Multi-class Classification and Decision Boundaries	MNIST, make_moons
04_cnn.ipynb	Convolutional Neural Networks	CIFAR-10
05_rnn_lstm_gru.ipynb	Sequence Modeling: RNN vs LSTM vs GRU	Sequential MNIST
06_transfer_learning.ipynb	Fine-tuning Pretrained Networks	CIFAR-10

Why this matters for Biology

The same patterns apply directly to biological data:

• Regression: Predicting drug IC50 from molecular fingerprints.
• Classification: Cell-type identification from flow cytometry or single-cell profiles.
• CNN: Automated analysis of microscopy and histopathology images.
• RNN/LSTM/GRU: Modeling DNA, RNA, and protein sequences.
• Transfer Learning: Adapting foundation vision models to bioimaging, or protein language models to downstream structure prediction.

Repository Structure

PyTorch101/
├── README.md
├── requirements.txt
├── .gitignore
├── utils.py
├── 01_fundamentals.ipynb
├── 02_regression.ipynb
├── 03_classification.ipynb
├── 04_cnn.ipynb
├── 05_rnn_lstm_gru.ipynb
└── 06_transfer_learning.ipynb

Quick Start

Local

git clone https://github.com/Suhaila-Samir-El-behairy/PyTorch101.git
cd PyTorch101
pip install -r requirements.txt
jupyter lab

Kaggle

Each notebook is synced from Kaggle. To run directly:

1. Click the Open in Kaggle badge at the top of this file.
2. In Kaggle, enable GPU under Settings → Accelerator.
3. Enable Internet under Settings if you need to download MNIST or CIFAR-10.
4. Run all cells.

Requirements

Core dependencies are listed below. Seerequirements.txt for the full list.

• torch
• torchvision
• scikit-learn
• matplotlib
• numpy

Notes on Kaggle Sync

These notebooks are authored in Kaggle and pushed to GitHub via Kaggle's native GitHub integration. Notebook outputs (plots, prints, and cell metadata) are often committed automatically. If you clone locally, you should clear outputs before making manual commits to keep the repository size small:

jupyter nbconvert --clear-output --inplace *.ipynb

License

MIT