Work

MyTorch

A foundational deep learning library built from first principles in Python

Status
Active Development
Period
September 2025 - Present
MyTorch preview

Overview

Built neural network library from scratch in Python and C++ replicating PyTorch functionality.

Implemented core elements: activation functions (ReLU, GeLU, Sigmoid, Tanh), linear layers, batch normalization, optimizers (Adam, AdamW, SGD), allowing a 100% complete setup of MLPs.

Implemented essential deep learning models including CNN, RNN, LSTM, GRU and Transformers, and conducted three models training, confirming understanding of deep learning architectures.

Methods and Tools

  • Neural Networks
  • Deep Learning
  • PyTorch
  • Python
  • C++
  • MLPs
  • CNNs
  • RNNs
  • Transformers