Skip to content

Latest commit

 

History

History
337 lines (227 loc) · 9.23 KB

INSTALL.md

File metadata and controls

337 lines (227 loc) · 9.23 KB
Raw

INSTALL file for Faiss (Fair AI Similarity Search)

Install via Conda

The easiest way to install FAISS is from Anaconda. We regularly push stable releases to the pytorch conda channel.

Currently we support faiss-cpu both on Linux and OSX. We also provide faiss-gpu compiled with CUDA8/CUDA9/CUDA10 on Linux systems.

You can easily install it by

# CPU version only
conda install faiss-cpu -c pytorch

# GPU version
conda install faiss-gpu cudatoolkit=8.0 -c pytorch # For CUDA8
conda install faiss-gpu cudatoolkit=9.0 -c pytorch # For CUDA9
conda install faiss-gpu cudatoolkit=10.0 -c pytorch # For CUDA10

Compile from source

The Faiss compilation works in 2 steps:

  1. compile the C++ core and examples

  2. compile the Python interface

Steps 2 depends on 1.

It is also possible to build a pure C interface. This optional process is described separately (please see the C interface installation file)

General compilation instructions

TL;DR:

cmake -B build .
make -C build

# For the Python interface:
cd build/faiss/python && python setup.py install
  1. cmake

This generates the system-dependent configuration/build files.

A few useful options:

  • -DFAISS_ENABLE_GPU=OFF in order to build the CPU part only.
  • -DCUDAToolkit_ROOT=/path/to/cuda-10.1 in order to hint to the path of the cudatoolkit.
  • -DCMAKE_CUDA_ARCHITECTURES="75;72" for specifying which GPU architectures to build against.
  • -DPython_EXECUTABLE=/path/to/python3.7 in order to build a python interface for a different python than the default one.
  1. make

This builds the C++ library (the whole library if a suitable cuda toolkit was found, or the CPU part only otherwise).

  1. make install (optional)

This installs the headers and libraries.

Faiss is supported on x86_64 machines on Linux and Mac OS. It has been found to run on other platforms as well, see other platforms

Faiss requires a C++ compiler that understands:

  • the Intel intrinsics for SSE instructions,
  • the GCC intrinsic for the popcount instruction,
  • basic OpenMP.

There are indications for specific configurations in the troubleshooting section of the wiki.

https://github.com/facebookresearch/faiss/wiki/Troubleshooting

Faiss comes as a .a archive, that can be linked with executables or dynamic libraries (useful for the Python wrapper).

BLAS/Lapack

The only variables that need to be configured for the C++ Faiss are the BLAS/Lapack flags (a linear aglebra software package). It needs a flag telling whether BLAS/Lapack uses 32 or 64 bit integers and the linking flags. Faiss uses the Fortran 77 interface of BLAS/Lapack and thus does not need an include path.

There are several BLAS implementations, depending on the OS and machine. To have reasonable performance, the BLAS library should be multithreaded.

To check that the link flags are correct, and verify whether the implementation uses 32 or 64 bit integers, you can

make test_blas

and run

./misc/test_blas

Testing Faiss

A basic usage example is in

demos/demo_ivfpq_indexing

which you can build by calling make demo_ivfpq_indexing

It makes a small index, stores it and performs some searches. A normal runtime is around 20s. With a fast machine and Intel MKL's BLAS it runs in 2.5s.

To run the whole test suite:

make test (for the CPU part)

A real-life benchmark

A bit longer example runs and evaluates Faiss on the SIFT1M dataset. To run it, please download the ANN_SIFT1M dataset from

http://corpus-texmex.irisa.fr/

and unzip it to the subdirectory sift1M at the root of the source directory for this repository.

Then compile and run the following (after ensuring you have installed faiss):

make demos
./demos/demo_sift1M

This is a demonstration of the high-level auto-tuning API. You can try setting a different index_key to find the indexing structure that gives the best performance.

Python Interface

The Python interface is provided via SWIG (Simple Wrapper and Interface Generator) and an additional level of manual wrappers (in faiss/python/faiss.py).

SWIG generates two wrapper files: a Python file (faiss/python/swigfaiss.py) and a C++ file that must be compiled to a dynamic library (faiss/python/_swigfaiss.so).

Testing the Python wrapper

Often, a successful compile does not mean that the library works, because missing symbols are detected only at runtime. You should be able to load the Faiss dynamic library:

python -c "import faiss"

In case of failure, it reports the first missing symbol. To see all missing symbols (on Linux), use

ldd -r _swigfaiss.so

Sometimes, problems (eg with BLAS libraries) appear only when actually calling a BLAS function. A simple way to check this

python -c "import faiss, numpy
faiss.Kmeans(10, 20).train(numpy.random.rand(1000, 10).astype('float32'))"

Real-life test

The following script extends the demo_sift1M test to several types of indexes. This must be run from the root of the source directory for this repository:

mkdir tmp             # graphs of the output will be written here
PYTHONPATH=. python demos/demo_auto_tune.py

It will cycle through a few types of indexes and find optimal operating points. You can play around with the types of indexes.

Step 3: Compiling the GPU implementation

The GPU version is a superset of the CPU version. In addition it requires the cuda compiler and related libraries (Cublas)

The nvcc-specific flags to pass to the compiler, based on your desired compute capability can be customized by providing the CMAKE_CUDA_ARCHITECTURES flag to CMake. Only compute capability 3.5+ is supported. For example, we enable by default:

-gencode=arch=compute_35,code=compute_35
-gencode=arch=compute_52,code=compute_52
-gencode=arch=compute_60,code=compute_60
-gencode=arch=compute_61,code=compute_61
-gencode=arch=compute_70,code=compute_70
-gencode=arch=compute_75,code=compute_75

However, look at https://developer.nvidia.com/cuda-gpus to determine what compute capability you need to use, and replace our gencode specifications with the one(s) you need.

Most other flags are related to the C++11 compiler used by nvcc to complile the actual C++ code. They are normally just transmitted by nvcc, except some of them that are not recognized and that should be escaped by prefixing them with -Xcompiler. Also link flags that are prefixed with -Wl, should be passed with -Xlinker.

You may want to add -j 10 to use 10 threads during compile.

Testing the GPU implementation

Compile the example with

make demo_ivfpq_indexing_gpu

This produce the GPU code equivalent to the CPU demo_ivfpq_indexing. It also shows how to translate indexed from/to the GPU.

Python example with GPU support

The auto-tuning example above also runs on the GPU. Edit demos/demo_auto_tune.py at line 100 with the values

keys_to_test = keys_gpu
use_gpu = True

and you can run

export PYTHONPATH=.
python demos/demo_auto_tune.py

to test the GPU code.

Docker instructions

For using GPU capabilities of Faiss, you'll need to run "nvidia-docker" rather than "docker". Make sure that docker (https://docs.docker.com/engine/installation/) and nvidia-docker (https://github.com/NVIDIA/nvidia-docker) are installed on your system

To build the "faiss" image, run

nvidia-docker build -t faiss .

or if you don't want/need to clone the sources, just run

nvidia-docker build -t faiss github.com/facebookresearch/faiss

If you want to run the tests during the docker build, uncomment the last 3 "RUN" steps in the Dockerfile. But you might want to run the tests by yourself, so just run

nvidia-docker run -ti --name faiss faiss bash

and run what you want. If you need a dataset (like sift1M), download it inside the created container, or better, mount a directory from the host

nvidia-docker run -ti --name faiss -v /my/host/data/folder/ann_dataset/sift/:/opt/faiss/sift1M faiss bash

How to use Faiss in your own projects

C++

The makefile generates a static and a dynamic library

libfaiss.a
libfaiss.so (or libfaiss.dylib)

the executable should be linked to one of these. If you use the static version (.a), add the LDFLAGS used in the Makefile.

For binary-only distributions, the headers should be under a faiss/ directory, so that they can be included as

#include <faiss/IndexIVFPQ.h>
#include <faiss/gpu/GpuIndexFlat.h>

Python

To import Faiss in your own Python project, you need the files

__init__.py
swigfaiss.py
_swigfaiss.so

to be present in a faiss/ directory visible in the PYTHONPATH or in the current directory. Then Faiss can be used in python with

import faiss