Search found 3376 matches

by Daniel Shawul
Sun Nov 10, 2019 1:40 pm
Forum: Computer Chess Club: General Topics
Topic: Winter NN Training Script
Replies: 5
Views: 486

Re: Winter NN Training Script

Rather high up on my list of priorities is to add saving the netweights in an open format in my training script. If the nets get much larger then I think it makes sense to have the nets be loaded externally. Especially as I would eventually like to support networks from other engines in the long ru...
by Daniel Shawul
Sat Nov 09, 2019 6:15 pm
Forum: Computer Chess Club: General Topics
Topic: Winter NN Training Script
Replies: 5
Views: 486

Re: Winter NN Training Script

Where can i download your network ? If it just weights that you have, could you produce a frozen tensorflow protobuf graph (pb), or since you are using keras the HDF5 graph works too. I should be able to use it in Scorpio after that once I figure out your input and outputs. I want to see how it does...
by Daniel Shawul
Fri Nov 08, 2019 12:05 am
Forum: Computer Chess Club: General Topics
Topic: Fritz 17
Replies: 253
Views: 15917

Re: Fritz 17

Leela wave I was doing MCTS after A0 news before Leela was born so you got it backwards. Sure lc0 did use NN first but i was mostly focused on understanding how MCTS can cope with AB -- for which you don't need NN. Then came Komodo MCTS after I am done with all my investigation, that then came leel...
by Daniel Shawul
Thu Nov 07, 2019 4:42 pm
Forum: Computer Chess Club: General Topics
Topic: Fritz 17
Replies: 253
Views: 15917

Re: Fritz 17

It is interesting that you consider Scorpio to be a derivative of Lc0. I strongly suspect that Daniel Shawul would take exception to this description. He is not alone in his thoughts. Daniel or the original poster? If the latter, then you’ve produced a real head scratcher. Is leela chess, as the fi...
by Daniel Shawul
Sat Oct 26, 2019 12:38 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: UCI Win/Draw/Loss reporting
Replies: 60
Views: 2780

Re: UCI Win/Draw/Loss reporting

I missed this discussion.
So what is agreed up on: info wdl 200 600 200 format for uci, and what for xboard?
Thanks.
by Daniel Shawul
Sat Oct 26, 2019 12:09 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Another take on DTS?
Replies: 20
Views: 1335

Re: Another take on DTS?

I am pretty sure DTS was tested upto atleast 16 cores when it was first published and YBW upto a thousand core with its distributed scheme. Also 8-core testers were here since 10 years ago I believe. A lot of programmers were using YBW for SMP, a few on DTS since it requires iterative search and is ...
by Daniel Shawul
Wed Oct 23, 2019 2:17 pm
Forum: Computer Chess Club: General Topics
Topic: google quantum computer
Replies: 23
Views: 1261

Re: google quantum computer

Yikes, just learned IBM was talking about the "Summit supercomputer" when they said it takes 2.5 days... salty much?!
Ofcourse since they built Summit, they want to use this opportunity to divert undeserved attention to Summit instead of the real
breakthrough -- if that indeed what it is.
by Daniel Shawul
Wed Oct 23, 2019 2:01 pm
Forum: Computer Chess Club: General Topics
Topic: google quantum computer
Replies: 23
Views: 1261

google quantum computer

https://www.theverge.com/2019/10/23/20928294/google-quantum-supremacy-sycamore-computer-qubit-milestone Google has announced its quantum processor that is able to do what used to take 10,000 year in just 200 seconds. I am not knowledgeable enough to judge the validity of this claim, but IBM, for ins...
by Daniel Shawul
Tue Oct 15, 2019 2:39 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Understanding and Pushing the Limits of the Elo Rating Algorithm
Replies: 7
Views: 1134

Re: Understanding and Pushing the Limits of the Elo Rating Algorithm

How did they miss our work Paired Comparisons with Ties ?
We compared three draw models and found out Davidson fits chess data best -- which they reframed as their own K-elo if I am not mistaken.
Maybe I should alert the authors to our paper and makes them spell out what their unique contribution.