Skip to content

Latest commit

 

History

History
298 lines (265 loc) · 32 KB

README.md

File metadata and controls

298 lines (265 loc) · 32 KB

Awesome Torch

A curated list of awesome Torch tutorials, projects and communities.

Table of Contents

Tutorials

Model Zoo

Codes and related articles. (#) means authors of code and paper are different.

Recurrent Networks

Convolutional Networks

Reinforcement Learning

  • Deep Q-network, DeepMind-Atari-Deep-Q-Learner
    • Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Andrei A. Rusu, Joel Veness, Marc G. Bellemare, Alex Graves, Martin Riedmiller, Andreas K. Fidjeland, Georg Ostrovski, Stig Petersen, Charles Beattie, Amir Sadik, Ioannis Antonoglou, Helen King, Dharshan Kumaran, Daan Wierstra, Shane Legg, Demis Hassabis, Human-Level Control through Deep Reinforcement Learning, Nature, [Paper]
  • Deep Attention Recurrent Q-Network
    • (#) Ivan Sorokin, Alexey Seleznev, Mikhail Pavlov, Aleksandr Fedorov, Anastasiia Ignateva, Deep Attention Recurrent Q-Network, NIPS 2015, [Paper]
  • Grid World DQN using torch7
    • (#) Marc G. Bellemare, Georg Ostrovski, Arthur Guez, Philip S. Thomas, Rémi Munos, Increasing the Action Gap: New Operators for Reinforcement Learning, arXiv:1512.04860, [Paper]
  • Deep Q-Networks for Accelerating the Training of Deep Neural Networks
    • Jie Fu, Zichuan Lin, Miao Liu, Nicholas Leonard, Jiashi Feng, Tat-Seng Chua, Deep Q-Networks for Accelerating the Training of Deep Neural Networks, arXiv:1606.01467, [Paper]
  • ActorMimic
    • Emilio Parisotto, Jimmy Lei Ba, Ruslan Salakhutdinov, Actor-Mimic: Deep Multitask and Transfer Reinforcement Learning, ICLR 2016, [Paper]
  • MazeBase: a sandbox for learning from games
    • Sainbayar Sukhbaatar, Arthur Szlam, Gabriel Synnaeve, Soumith Chintala, Rob Fergus, MazeBase: A Sandbox for Learning from Games, arXiv:1511.07401, [Paper]
  • mario-ai
    • This project contains code to train a model that automatically plays the first level of Super Mario World using only raw pixels as the input (no hand-engineered features).The used technique is deep Q-learning, as described in the Atari paper (Summary), combined with a Spatial Transformer.
  • Deep Successor Reinforcement Learning (DSR)
    • Tejas D. Kulkarni, Ardavan Saeedi, Simanta Gautam, Samuel J. Gershman, Deep Successor Reinforcement Learning, arXiv:1606.02396, [Paper]
  • ViZDoom
    • ViZDoom allows developing AI bots that play Doom using only the visual information (the screen buffer). It is primarily intended for research in machine visual learning, and deep reinforcement learning, in particular.
  • MIXER - Sequence Level Training with Recurrent Neural Networks
    • Marc'Aurelio Ranzato, Sumit Chopra, Michael Auli, Wojciech Zaremba, Sequence Level Training with Recurrent Neural Networks, ICLR 2016, [Paper]
  • TorchQLearning
    • Implementation of a simple example of Q learning in Torch.
  • rltorch
    • This package is a Reinforcement Learning package written in LUA for Torch.
  • Opponent Modeling in Deep Reinforcement Learning
    • He He, Jordan Boyd-Graber, Kevin Kwok, Hal Daumé III, Opponent Modeling in Deep Reinforcement Learning, ICML 2016, [Paper]
### ETC

Libraries

Model related

  • nn : an easy and modular way to build and train simple or complex neural networks [Code] [Documentation]
  • dpnn : extensions to the nn lib, more modules [Code]
  • nnx : extension to the nn lib, experimental neural network modules and criterions [Code]
  • nninit : weight initialisation schemes [Code]
  • rnn : Recurrent Neural Network library [Code]
  • optim : A numeric optimization package for Torch [Code]
  • dp : a deep learning library designed for streamlining research and development [Code] [Documentation]
  • nngraph : provides graphical computation for nn library [Code] [Oxford Introduction]
  • nnlr : Add layer-wise learning rate schemes to Torch [Code]
  • optnet: Memory optimizations for torch neural networks. [Code]
  • autograd : Autograd automatically differentiates native Torch code. [Code]
  • torchnet: framework for torch which provides a set of abstractions aiming at encouraging code re-use as well as encouraging modular programming [Code] [Paper]

GPU related

  • distro-cl: An OpenCL distribution for Torch [Code]
  • cutorch : A CUDA backend for Torch [Code]
  • cudnn : Torch FFI bindings for NVIDIA CuDNN [Code]
  • fbcunn : Facebook's extensions to torch/cunn [Code] [Documentation]

IDE related

  • iTorch : IPython kernel for Torch with visualization and plotting [Code]
  • Lua Development Tools (LDT) : based on Eclipse [Code]
  • zbs-torch : A lightweight Lua-based IDE for Lua with code completion, syntax highlighting, live coding, remote debugger, and code analyzer [Code]

ETC

  • fblualib : Facebook libraries and utilities for Lua [Code]
  • loadcaffe : Load Caffe networks in Torch [Code]
  • Purdue e-lab lib : A collection of snippets and libraries [Code]
  • torch-android : Torch for Android [Code]
  • torch-models : Implementation of state-of-art models in Torch. [Code]
  • lutorpy : Lutorpy is a libray built for deep learning with torch in python. [Code]
  • CoreNLP.lua : Lua client for Stanford CoreNLP. [Code]
  • Torchlib: Data structures and libraries for Torch. [Code]
  • THFFmpeg: Torch bindings for FFmpeg (reading videos only) [Code]
  • tunnel: Data Driven Framework for Distributed Computing in Torch 7, [Code]
  • pytorch: Python wrappers for torch and lua, [Code]
  • lutorpy: Use torch in python for deep learning., [Code]
  • torch-pcl: Point Cloud Library (PCL) bindings for Torch, [Code]
  • Moses: A Lua utility-belt library for functional programming. It complements the built-in Lua table library, making easier operations on arrays, lists, collections. [Cpde]

Links