Search found 3233 matches

by Daniel Shawul
Wed Jun 26, 2019 2:50 am
Forum: Computer Chess Club: General Topics
Topic: Need Lc0 help
Replies: 22
Views: 363

Re: Need Lc0 help

Second position search stopped after one hour by itself. It searched 72million nodes only so it must be the effect of smart pruning as g2g3 is the overwhelming favourite. For the previous one, I did "go infinite" but for this one i did "go movetime 7200000" which enabled the smart pruning i think in...
by Daniel Shawul
Wed Jun 26, 2019 1:27 am
Forum: Computer Chess Club: General Topics
Topic: Need Lc0 help
Replies: 22
Views: 363

Re: Need Lc0 help

Egad. Nf6. :shock: It stuck with Nf6 after 2 hours using net ID33004 info depth 22 seldepth 67 time 7225636 nodes 195574102 score cp 16 hashfull 1000 nps 27066 tbhits 0 pv g8f6 d2d4 e7e6 e2e3 c7c5 g1f3 b8c6 c2c3 b7b6 b1d2 c8b7 f1d3 d7d5 a4a5 b6a5 d1a4 f6d7 d3b5 d8b6 d2b3 a7a6 b5e2 f8d6 b3a5 c6a5 a4...
by Daniel Shawul
Wed Jun 26, 2019 12:00 am
Forum: Computer Chess Club: General Topics
Topic: Need Lc0 help
Replies: 22
Views: 363

Re: Need Lc0 help

Ok two hours it is. But note that MCTS engines consume lots of memory for storing their trees so it might not be possible to run for that long. I have 256gb ram though on the cpus so it may go for 2 hours... After 30 minutes on the first position info depth 20 seldepth 60 time 2013411 nodes 57094081...
by Daniel Shawul
Tue Jun 25, 2019 11:02 pm
Forum: Computer Chess Club: General Topics
Topic: Need Lc0 help
Replies: 22
Views: 363

Re: Need Lc0 help

I can help. For how long do you want to run each position ? A peculiarity of lc0 is that it needs history to analyze a position. I think the default behaviour is the worst so I switched to filling history with current position. $ ./lc0 -t 2 --backend=cudnn-fp16 --history-fill=always --verbose-move-s...
by Daniel Shawul
Tue Jun 25, 2019 1:30 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Lazy-evaluation of futures for parallel work-efficient Alpha-Beta search
Replies: 26
Views: 996

Re: Lazy-evaluation of futures for parallel work-efficient Alpha-Beta search

You keep on ignoring the multiple elephants in the room and re-post the same thing again !? A few words from people who actually know what they are doing, i.e. Cilkchess authors Parallel search of game-trees is difficult because the most efficient algorithms for game-tree search are inherently seria...
by Daniel Shawul
Mon Jun 24, 2019 6:44 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Lazy-evaluation of futures for parallel work-efficient Alpha-Beta search
Replies: 26
Views: 996

Re: Lazy-evaluation of futures for parallel work-efficient Alpha-Beta search

* Node Type P: Alpha = -infinity and Beta=infinity. All children of Node-type P can be evaluated in parallel. max(score(child1), score(child2), score(child3)...) is returned. Therefore, all children can be searched in parallel. Isn't the returned score from child1 usually used as bound for child2 e...
by Daniel Shawul
Mon Jun 24, 2019 3:00 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Lazy-evaluation of futures for parallel work-efficient Alpha-Beta search
Replies: 26
Views: 996

Re: Lazy-evaluation of futures for parallel work-efficient Alpha-Beta search

From my understanding: a work-efficient parallel algorithm for Alpha-Beta search is already a "unicorn project" never seen before As you point out, Alpha-Beta pruning in parallel seems like an unsolved problem. But I hope to accomplish it in this topic. The algorithm to parallelize alpha-beta maxim...
by Daniel Shawul
Sun Jun 23, 2019 8:55 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Lazy-evaluation of futures for parallel work-efficient Alpha-Beta search
Replies: 26
Views: 996

Re: Lazy-evaluation of futures for parallel work-efficient Alpha-Beta search

From my understanding: a work-efficient parallel algorithm for Alpha-Beta search is already a "unicorn project" never seen before As you point out, Alpha-Beta pruning in parallel seems like an unsolved problem. But I hope to accomplish it in this topic. The algorithm to parallelize alpha-beta maxim...
by Daniel Shawul
Fri Jun 21, 2019 1:58 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Lazy-evaluation of futures for parallel work-efficient Alpha-Beta search
Replies: 26
Views: 996

Re: Lazy-evaluation of futures for parallel work-efficient Alpha-Beta search

Nothing says one should use a recursive form of alpha-beta, so one could use a rollouts version of alpha-beta that is easier to do on the GPU instead. I already did this for my mcts experiments on the cpu . Adding null-move and lmr was kind of complicated but managed to get them working. I had imple...
by Daniel Shawul
Sun Jun 16, 2019 7:37 pm
Forum: Computer Chess Club: General Topics
Topic: How to enter TCEC?
Replies: 2
Views: 341

Re: How to enter TCEC?

Best way is to contact the organizers at tcec discord or at the official website chat.

They probably want to test your engine before the season starts so go to discord and help them set it up.

Daniel