PyTorch-like Deep Learning Framework & NeuralODE

Python, C++, OOP, CUDA

Last updated on Jan 3, 2022

Instructors: Tianqi Chen and Zico Kolter

Course Website

Project Report

Project summary

  • Built completely from scratch a PyTorch-like deep learning library, including automatic differentiation, CPU/GPU backends (in C++), and all the necessary modules to support parameterized layers, loss functions, data loaders, and optimizers.
  • Designed and implemented a NeuralODE layer (operators and backends, solver, AD) within the Pytorch-like library and demonstrated its capability of approximating ODE dynamics with NNs.
Deep Learning
Xiangyu Yin
Xiangyu Yin

Postdoc @ ANL | AI4science, Physics4ML, scientific discovery acceleration & automation

Related