Pytorch

Resources

Courses

#COURSE Intro to Deep Learning with PyTorch (Udacity)
#COURSE Learn PyTorch for Deep Learning: Zero to Mastery book
- https://zerotomastery.io/courses/learn-pytorch/

Code

#CODE PyTorch (Facebook)
- http://pytorch.org
- Ecosystem tools
- Tensors and Dynamic neural networks in Python with strong GPU acceleration
- https://sagivtech.com/2017/09/19/optimizing-pytorch-training-code/
#CODE PADL - Pipeline Abtractions for Deep Learning
- https://padl.ai/
- https://medium.com/pytorch/padl-portable-pytorch-pipelines-facilitating-deep-learning-model-use-c73b7ed95bcc
#CODE TorchStudio
- IDE for PyTorch and its ecosystem
- https://www.assemblyai.com/blog/beginners-guide-to-torchstudio-pytorch-only-ide/
#CODE Pytorch-lightning
- https://medium.com/pytorch/introducing-lightning-flash-the-fastest-way-to-get-started-with-deep-learning-202f196b3b98
#CODE Fastai
- fastai simplifies training fast and accurate neural nets using modern best practices
- https://docs.fast.ai/
- #PAPER Fastai: A Layered API for Deep Learning (Howard 2020)
#CODE kornia
- Open Source Differentiable Computer Vision Library
- https://kornia.github.io/
#CODE NMF
- A modular framework for vision & language multimodal research from Facebook AI Research (FAIR)
- https://mmf.sh/
#CODE Pytext (Facebook) - A natural language modeling framework based on PyTorch
- https://fb.me/pytextdocs
- PyText is a deep-learning based NLP modeling framework built on PyTorch
#CODE Clearml
- ClearML - Auto-Magical CI/CD to streamline your ML workflow. Experiment Manager, AI/Data Engineering/ML Ops and Data-Management
- https://clear.ml/docs
#CODE Composer
- A PyTorch Library for Efficient Neural Network Training
#CODE ColossalAI
- Colossal-AI: A Unified Deep Learning System for Large-Scale Parallel Training
- https://analyticsindiamag.com/a-guide-to-parallel-deep-learning-with-colossal-ai/

References

#PAPER PyTorch: An Imperative Style, High-Performance Deep Learning Library (Paszke 2019)
- Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it provides an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs