Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Strength comparison to LeelaZero or Facebook ELF #1

Open
bood opened this issue May 11, 2018 · 5 comments
Open

Strength comparison to LeelaZero or Facebook ELF #1

bood opened this issue May 11, 2018 · 5 comments

Comments

@bood
Copy link

bood commented May 11, 2018

Did you run any test for a strength comparison with previous published Go programs?
If so, could you publish your result too? Thanks.

BTW: is this published network the same as the one used on the World AI Go Tournament 2018?

@godmoves
Copy link
Contributor

It seems that they use a 40b net on the World AI Go Tournament.

@Nefeste
Copy link

Nefeste commented May 12, 2018

LZ-ELF (62b5417b) is stronger than this version of BensonDarr "off-the-shelf" without TensorRT. Probably need some tuning with pondering/time settings/config tunung/better network to get this bot play at the level of LZ-ELF or higher.

@godmoves
Copy link
Contributor

FYI, some tests are mentioned here. leela-zero/leela-zero#1405

@baduk1
Copy link

baduk1 commented Jun 16, 2018

On my PC (2x 1080ti ASUS Strix SLI, Core i9 64 Gb) LZ-ELF is always winning PhoenixGo
Here is 60 sec game (150-650k visits average) https://www.youtube.com/watch?v=WY4tV5w_g0w
Here is a game when as a handicap ELF starts from 10-10 and still wins on 30 sec https://www.youtube.com/watch?v=M-zwD2855Us

@Godady
Copy link

Godady commented Jun 19, 2018

The point is the released PhoenixGo weight is too weak, not good enough for play against Elf weight.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

5 participants