Skip to content

Crzax/CMU10-714-DeepLearningSystems

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

CMU 10-714: Deep Learning Systems

This repository contains my complete implementations for the assignments of the CMU 10-714: Deep Learning Systems course. All tasks across all homework assignments have been fully implemented and tested.

Project Overview

The goal of this course is to build a deep learning library called needle (Necessary Elements of Deep LEarning) from scratch, covering everything from automatic differentiation to hardware backends and high-level model architectures.

Assignments Summary

  • Implemented softmax regression for MNIST digit classification.
  • Developed a C++ extension for matrix operations to improve performance.
  • Built the core needle autograd engine.
  • Implemented forward and backward passes for basic mathematical operations.
  • Developed the computational graph system for automatic gradient calculation.
  • Implemented high-level neural network modules (nn.Module, nn.Linear, nn.ReLU, nn.LayerNorm, etc.).
  • Developed optimizers including SGD with momentum and Adam.
  • Built a data loading pipeline with Dataset and DataLoader classes.
  • Implemented and trained MLP and ResNet architectures.
  • Developed a custom ndarray library to handle memory management and broadcasting.
  • Implemented efficient CPU (C++) and GPU (CUDA) backends for tensor operations.
  • Integrated the backends with the needle autograd system.
  • Implemented Convolutional layers and pooling operations.
  • Developed sequence models including Recurrent Neural Networks (RNN) and Long Short-Term Memory (LSTM) networks.
  • Trained models on CIFAR-10 (image classification) and Penn Treebank (language modeling) datasets.
  • Implemented the Transformer architecture from scratch.
  • Developed Multi-Head Attention, Transformer layers, and positional embeddings.
  • Built a decoder-only Transformer for language modeling tasks.

Implementation Status

All assignments (HW0 - HW4 Extra) are fully implemented. This includes:

  • Core autograd engine.
  • CPU and CUDA backends.
  • Neural network modules and optimizers.
  • Advanced architectures (ResNet, LSTM, Transformers).

About

the implementation of the assignments of CMU10-714: Deep Learning System

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published