What

A Go program with no human provided knowledge. Using MCTS (but without Monte Carlo playouts) and a deep residual convolutional neural network stack.

This is a fairly faithful reimplementation of the system described in the Alpha Go Zero paper "Mastering the Game of Go without Human Knowledge". For all intents and purposes, it is an open source AlphaGo Zero.

Wait, what?

If you are wondering what the catch is: you still need the network weights. No network weights are in this repository. If you manage to obtain the AlphaGo Zero weights, this program will be about as strong, provided you also obtain a few Tensor Processing Units. Lacking those TPUs, I'd recommend a top of the line GPU - it's not exactly the same, but the result would still be an engine that is far stronger than the top humans.

Gimme the weights

Recomputing the AlphaGo Zero weights will take about 1700 years on commodity hardware.

One reason for publishing this program is that we are running a public, distributed effort to repeat the work. Working together, and especially when starting on a smaller scale, it will take less than 1700 years to get a good network (which you can feed into this program, suddenly making it strong).

I want to help

Using your own hardware

You need a PC with a GPU, i.e. a discrete graphics card made by NVIDIA or AMD, preferably not too old, and with the most recent drivers installed.

It is possible to run the program without a GPU, but performance will be much lower. If your CPU is not very recent (Haswell or newer, Ryzen or newer), performance will be outright bad, and it's probably of no use trying to join the distributed effort. But you can still play, especially if you are patient.

Windows

Head to the Github releases page at https://github.com/leela-zero/leela-zero/releases, download the latest release, unzip, and launch autogtp.exe. It will connect to the server automatically and do its work in the background, uploading results after each game. You can just close the autogtp window to stop it.

macOS and Linux

Follow the instructions below to compile the leelaz and autogtp binaries in the build subdirectory. Then run autogtp as explained in the contributing instructions below. Contributing will start when you run autogtp.

Using a Cloud provider

Many cloud companies offer free trials (or paid solutions, not discussed here) that are usable for helping the leela-zero project.

There are community maintained instructions available here:

I just want to play with Leela Zero right now

Download the best known network weights file from here, or, if you prefer a more human style, a (weaker) network trained from human games here.

If you are on Windows, download an official release from here and head to the Usage section of this README.

If you are on macOS, Leela Zero is available through Homebrew, the de facto standard package manager. You can install it with:

brew install leela-zero

If you are on Unix, you have to compile the program yourself. Follow the compilation instructions below and then read the Usage section.

Compiling AutoGTP and/or Leela Zero

Requirements

GCC, Clang or MSVC, any C++14 compiler
Boost 1.58.x or later, headers and program_options, filesystem and system libraries (libboost-dev, libboost-program-options-dev and libboost-filesystem-dev on Debian/Ubuntu)
zlib library (zlib1g & zlib1g-dev on Debian/Ubuntu)
Standard OpenCL C headers (opencl-headers on Debian/Ubuntu, or at https://github.com/KhronosGroup/OpenCL-Headers/tree/master/CL)
OpenCL ICD loader (ocl-icd-libopencl1 on Debian/Ubuntu, or reference implementation at https://github.com/KhronosGroup/OpenCL-ICD-Loader)
An OpenCL capable device, preferably a very, very fast GPU, with recent drivers is strongly recommended (OpenCL 1.1 support is enough). Don't forget to install the OpenCL driver if this part is packaged seperately by the Linux distribution (e.g. nvidia-opencl-icd). If you do not have a GPU, add the define "USE_CPU_ONLY", for example by adding -DUSE_CPU_ONLY=1 to the cmake command line.
Optional: BLAS Library: OpenBLAS (libopenblas-dev) or Intel MKL
The program has been tested on Windows, Linux and macOS.

Example of compiling - Ubuntu & similar

# Test for OpenCL support & compatibility
sudo apt install clinfo && clinfo

# Clone github repo
git clone https://github.com/leela-zero/leela-zero
cd leela-zero
git submodule update --init --recursive

# Install build depedencies
sudo apt install cmake g++ libboost-dev libboost-program-options-dev libboost-filesystem-dev opencl-headers ocl-icd-libopencl1 ocl-icd-opencl-dev zlib1g-dev

# Use a stand alone build directory to keep source dir clean
mkdir build && cd build

# Compile leelaz and autogtp in build subdirectory with cmake
cmake ..
cmake --build .

# Optional: test if your build works correctly
./tests

Example of compiling - macOS

# Clone github repo
git clone https://github.com/leela-zero/leela-zero
cd leela-zero
git submodule update --init --recursive

# Install build depedencies
brew install boost cmake zlib

# Use a stand alone build directory to keep source dir clean
mkdir build && cd build

# Compile leelaz and autogtp in build subdirectory with cmake
cmake ..
cmake --build .

# Optional: test if your build works correctly
./tests

Example of compiling - Windows

# Clone github repo
git clone https://github.com/leela-zero/leela-zero
cd leela-zero
git submodule update --init --recursive

cd msvc
Double-click the leela-zero2015.sln or leela-zero2017.sln corresponding
to the Visual Studio version you have.
# Build from Visual Studio 2015 or 2017

Contributing

For Windows, you can use a release package, see "I want to help".

Unix and macOS, after finishing the compile and while in the build directory:

# Copy leelaz binary to autogtp subdirectory
cp leelaz autogtp

# Run AutoGTP to start contributing
./autogtp/autogtp

Usage for playing or analyzing games

Leela Zero is not meant to be used directly. You need a graphical interface for it, which will interface with Leela Zero through the GTP protocol.

The engine supports the GTP protocol, version 2.

Lizzie is a client specifically for Leela Zero which shows live search probilities, a win rate graph, and has an automatic game analysis mode. Has binaries for Windows, Mac, and Linux.

Sabaki is a very nice looking GUI with GTP 2 capability.

LeelaSabaki is modified to show variations and winning statistics in the game tree, as well as a heatmap on the game board.

GoReviewPartner is a tool for automated review and analysis of games using bots (saved as .rsgf files), Leela Zero is supported.

A lot of go software can interface to an engine via GTP, so look around.

Add the --gtp commandline option on the engine command line to enable Leela Zero's GTP support. You will need a weights file, specify that with the -w option.

All required commands are supported, as well as the tournament subset, and "loadsgf". The full set can be seen with "list_commands". The time control can be specified over GTP via the time_settings command. The kgs-time_settings extension is also supported. These have to be supplied by the GTP 2 interface, not via the command line!

Weights format

The weights file is a text file with each line containing a row of coefficients. The layout of the network is as in the AlphaGo Zero paper, but any number of residual blocks is allowed, and any number of outputs (filters) per layer, as long as the latter is the same for all layers. The program will autodetect the amounts on startup. The first line contains a version number.

Convolutional layers have 2 weight rows:
1. convolution weights
2. channel biases
Batchnorm layers have 2 weight rows:
1. batchnorm means
2. batchnorm variances
Innerproduct (fully connected) layers have 2 weight rows:
1. layer weights
2. output biases

The convolution weights are in [output, input, filter_size, filter_size] order, the fully connected layer weights are in [output, input] order. The residual tower is first, followed by the policy head, and then the value head. All convolution filters are 3x3 except for the ones at the start of the policy and value head, which are 1x1 (as in the paper).

There are 18 inputs to the first layer, instead of 17 as in the paper. The original AlphaGo Zero design has a slight imbalance in that it is easier for the black player to see the board edge (due to how padding works in neural networks). This has been fixed in Leela Zero. The inputs are:

1) Side to move stones at time T=0
2) Side to move stones at time T=-1  (0 if T=0)
...
8) Side to move stones at time T=-7  (0 if T<=6)
9) Other side stones at time T=0
10) Other side stones at time T=-1   (0 if T=0)
...
16) Other side stones at time T=-7   (0 if T<=6)
17) All 1 if black is to move, 0 otherwise
18) All 1 if white is to move, 0 otherwise

Each of these forms a 19 x 19 bit plane.

In the training/caffe directory there is a zero.prototxt file which contains a description of the full 40 residual block design, in (NVIDIA)-Caffe protobuff format. It can be used to set up nv-caffe for training a suitable network. The zero_mini.prototxt file describes a smaller 12 residual block case. The training/tf directory contains the network construction in TensorFlow format, in the tfprocess.py file.

Expert note: the channel biases seem redundant in the network topology because they are followed by a batchnorm layer, which is supposed to normalize the mean. In reality, they encode "beta" parameters from a center/scale operation in the batchnorm layer, corrected for the effect of the batchnorm mean/variance adjustment. At inference time, Leela Zero will fuse the channel bias into the batchnorm mean, thereby offsetting it and performing the center operation. This roundabout construction exists solely for backwards compatibility. If this paragraph does not make any sense to you, ignore its existence and just add the channel bias layer as you normally would, output will be correct.

Training

Getting the data

At the end of the game, you can send Leela Zero a "dump_training" command, followed by the winner of the game (either "white" or "black") and a filename, e.g:

dump_training white train.txt

This will save (append) the training data to disk, in the format described below, and compressed with gzip.

Training data is reset on a new game.

Supervised learning

Leela can convert a database of concatenated SGF games into a datafile suitable for learning:

dump_supervised sgffile.sgf train.txt

This will cause a sequence of gzip compressed files to be generated, starting with the name train.txt and containing training data generated from the specified SGF, suitable for use in a Deep Learning framework.

Training data format

The training data consists of files with the following data, all in text format:

16 lines of hexadecimal strings, each 361 bits longs, corresponding to the first 16 input planes from the previous section
1 line with 1 number indicating who is to move, 0=black, 1=white, from which the last 2 input planes can be reconstructed
1 line with 362 (19x19 + 1) floating point numbers, indicating the search probabilities (visit counts) at the end of the search for the move in question. The last number is the probability of passing.
1 line with either 1 or -1, corresponding to the outcome of the game for the player to move

Running the training

For training a new network, you can use an existing framework (Caffe, TensorFlow, PyTorch, Theano), with a set of training data as described above. You still need to contruct a model description (2 examples are provided for Caffe), parse the input file format, and outputs weights in the proper format.

There is a complete implementation for TensorFlow in the training/tf directory.

Supervised learning with TensorFlow

This requires a working installation of TensorFlow 1.4 or later:

src/leelaz -w weights.txt
dump_supervised bigsgf.sgf train.out
exit
training/tf/parse.py 6 128 train.out

This will run and regularly dump Leela Zero weight files (of networks with 6 blocks and 128 filters) to disk, as well as snapshots of the learning state numbered by the batch number. If interrupted, training can be resumed with:

training/tf/parse.py 6 128 train.out leelaz-model-batchnumber

Todo

Further optimize Winograd transformations.
Improve GPU batching in the search.
Root filtering for handicap play.
More backends:
MKL-DNN based backend.
CUDA specific version using cuDNN or cuBLAS.
AMD specific version using MIOpen/ROCm.

License

The code is released under the GPLv3 or later, except for ThreadPool.h, cl2.hpp, half.hpp and the eigen and clblast_level3 subdirs, which have specific licenses (compatible with GPLv3) mentioned in those files.

Additional permission under GNU GPL version 3 section 7

If you modify this Program, or any covered work, by linking or combining it with NVIDIA Corporation's libraries from the NVIDIA CUDA Toolkit and/or the NVIDIA CUDA Deep Neural Network library and/or the NVIDIA TensorRT inference library (or a modified version of those libraries), containing parts covered by the terms of the respective license agreement, the licensors of this Program grant you additional permission to convey the resulting work.

cuilongji / leela-zero