leela-zero

Description

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.

https://deepmind.com/documents/119/agz_unformatted_nature.pdf

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, the author recommends 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.

Recomputing the AlphaGo Zero weights will take about 1700 years on commodity hardware. Upstream is running a public, distributed effort to repeat this 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). To help with this effort, run the leelaz-autogtp binary provided in this package. The best-known network weights file is at http://zero.sjeng.org/best-network

---

Upload more screenshots

Please help extend the collection of screenshots. Just make a screenshot and upload it here. You don't need to register or anything.

Upload a screenshot

Hint: upload an image here from your clipboard with Ctrl-V

---

Homepage

https://github.com/gcp/leela-zero

---

Install this software package

If the package is available for the distribution you are currently using on your computer then install the software by clicking on…

Install leela-zero