fastai

The fastai library simplifies training fast and accurate neural nets using modern best practices. See the fastai website to get started. The library is based on research in to deep learning best practices undertaken at fast.ai, and includes "out of the box" support for vision, text, tabular, and collab (collaborative filtering) models. For brief examples, see the examples folder; detailed examples are provided in the full documentation. For instance, here's how to train an MNIST model using resnet18 (from the vision example):

untar_data(MNIST_PATH)
data = image_data_from_folder(MNIST_PATH)
learn = ConvLearner(data, tvm.resnet18, metrics=accuracy)
learn.fit(1)

Note for course.fast.ai students

If you are using fastai for any course.fast.ai course, you need to use fastai 0.7.x. Please ignore the rest of this document, which is written for fastai 1.0.x, and instead follow the installation instructions here.

Note: If you want to learn how to use fastai v1 from its lead developer, Jeremy Howard, he will be teaching it in the Deep Learning Part I course at the University of San Francisco from Oct 22nd, 2018.

Installation

fastai-1.x can be installed with either conda or pip package managers and also from source. At the moment you can't just run install, since you first need to get the correct pytorch version installed - thus to get fastai-1.x installed choose one of the installation recipes below using your favourite python package manager.

If your system has a recent NVIDIA card with the correctly configured NVIDIA driver please follow the GPU installation instructions. Otherwise, the CPU-ones.

If you experience installation problems, please read about installation issues.

Conda Install

GPU

conda install -c pytorch pytorch-nightly cuda92
conda install -c fastai torchvision-nightly
conda install -c fastai fastai

CPU

conda install -c pytorch pytorch-nightly-cpu
conda install -c fastai torchvision-nightly-cpu
conda install -c fastai fastai

PyPI Install

GPU

pip install torch_nightly -f https://download.pytorch.org/whl/nightly/cu92/torch_nightly.html
pip install fastai

CPU

pip install torch_nightly -f https://download.pytorch.org/whl/nightly/cpu/torch_nightly.html
pip install fastai

NB: this set will also fetch torchvision-nightly, which supports torch-1.x.

Developer Install

First, follow the instructions above for either PyPi or Conda. Then uninstall the fastai package using the same package manager you used to install it, i.e. pip uninstall fastai or conda uninstall fastai, and then, replace it with a pip editable install.

git clone https://github.com/fastai/fastai
cd fastai
tools/run-after-git-clone
pip install -e .[dev]

You can test that the build works by starting the jupyter notebook:

jupyter notebook

and executing an example notebook. For example load examples/tabular.ipynb and run it.

Alternatively, you can do a quick CLI test:

jupyter nbconvert --execute --ExecutePreprocessor.timeout=600 --to notebook examples/tabular.ipynb

If anything goes wrong please read and report installation issues.

Please refer to CONTRIBUTING.md and develop.md for more details on how to contribute to the fastai project.

Installation Issues

If the installation process fails, first make sure your system is supported. And if the problem is still not addressed, please see this installation issues thread.

If you encounter installation problems with conda, make sure you have the latest conda client:

conda update conda

Sometimes you have to run the following instead:

conda install conda

Is My System Supported?

Python: You need to have python 3.6 or higher
CPU or GPU

The pytorch binary package comes with its own CUDA, CuDNN, NCCL, MKL, and other libraries so you don't have to install system-wide NVIDIA's CUDA and related libraries if you don't need them for something else. If you have them installed already it doesn't matter which NVIDIA's CUDA version library you have installed system-wide. Your system could have CUDA 9.0 libraries, and you can still use pytorch build with cuda9.2 libraries without any problem, since the pytorch binary package is self-contained.

The only requirement is that you have installed and configured the NVIDIA driver correctly. Usually you can test that by running nvidia-smi. While it's possible that this application is not available on your system, it's very likely that if it doesn't work, than your don't have your NVIDIA drivers configured properly. And remember that a reboot is always required after installing NVIDIA drivers.

If you build pytorch from source then you will need to first install CUDA, CuDNN, and other required libraries. See pytorch.org.
Operating System:

Since fastai-1.0 relies on pytorch-1.0, you need to be able to install pytorch-1.0 first.

As of this moment pytorch.org's pre-1.0.0 version (torch-nightly) supports:

Platform GPU CPU

linux binary binary

mac source binary

windows source source

Legend: binary = can be installed directly, source = needs to be built from source.

This will change once pytorch 1.0.0 is released and installable packages made available for your system, which could take some time after the official release is made. Please watch for updates here.

If there is no pytorch preview conda or pip package available for your system, you may still be able to build it from source.

Alternatively, please consider installing and using the very solid "0.7.x" version of fastai. Please see the instructions.

Platform	GPU	CPU
linux	binary	binary
mac	source	binary
windows	source	source

Copyright

Copyright 2017 onwards, fast.ai, Inc. Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. A copy of the License is provided in the LICENSE file in this repository.

kimfalk/fastai