CNNs, RNNs, and Transformers implemented and trained from scratch in PyTorch.
This repo tracks my progression through deep learning architectures — starting from basic feedforward networks up to full Transformer blocks — building each one to understand it rather than only using pretrained versions.
- Feedforward neural network with manual backpropagation walkthrough
- CNN for image classification
- RNN / LSTM for sequence modeling
- Transformer encoder block (multi-head attention, positional encoding, layer norm)
Each model has its own folder with the implementation, the dataset used, and a results table (accuracy/loss).
Python, PyTorch, NumPy, Matplotlib
Salik Hussain — github.com/salikhussain71-code
MIT