If your CPU is not very recent (Haswell or newer, Ryzen or newer), It is possible to run the program without a GPU, but performance will be much Preferably not too old, and with the most recent drivers installed. a discrete graphics card made by NVIDIA or AMD, When starting on a smaller scale, it will take less than 1700 years to getĪ good network (which you can feed into this program, suddenly making it strong). One reason for publishing this program is that we are running a public,ĭistributed effort to repeat the work. Recomputing the AlphaGo Zero weights will take about 1700 years on commodity hardware. Lacking those TPUs, I'd recommendĪ top of the line GPU - it's not exactly the same, but the result would stillīe an engine that is far stronger than the top humans. If you manage to obtain theĪlphaGo Zero weights, this program will be about as strong, provided youĪlso obtain a few Tensor Processing Units. No network weights are in this repository. If you are wondering what the catch is: you still need the network weights. In the Alpha Go Zero paper ' Mastering the Game of Go without Human Knowledge'.įor all intents and purposes, it is an open source AlphaGo Zero. This is a fairly faithful reimplementation of the system described Monte Carlo playouts) and a deep residual convolutional neural network stack. A Go program with no human provided knowledge.

Alpha black zero download free