Search found 481 matches

by jhellis3
Sun Sep 13, 2020 10:14 pm
Forum: Computer Chess Club: General Topics
Topic: Asking to people who believe Leela NN is a book, what they think about SF NN now?
Replies: 86
Views: 15989

Re: Asking to people who believe Leela NN is a book, what they think about SF NN now?

Somehow a leela style net that never saw positions with more than 18 pieces was able to “memorize” opening moves. Yet somehow people arguing this tired old argument are unable to memorize this refutation. :D I'm quite sure myself and others have pointed out the logical flaw in this reasoning, yet y...
by jhellis3
Thu Sep 10, 2020 6:11 pm
Forum: Computer Chess Club: General Topics
Topic: Crystal 3.0
Replies: 34
Views: 10735

Crystal 3.0

Decided it was an appropriate time for release. Changes since 2.0: Various bugfixes mostly due to members of the community pointing out usage issues to me. Some minor eval smoothing (simple linear interpolation between previous value). Introduction of NNUE. For now, NNUE is either on or off, there i...
by jhellis3
Fri Sep 04, 2020 9:06 pm
Forum: Computer Chess Club: General Topics
Topic: SF-NN embedded
Replies: 59
Views: 16433

Re: SF-NN embedded

Downloading a network file and installing it is the exact same process as downloading an engine and installing it.

Apparently, the first time is quite easy, but the second time is exceedingly difficult....
by jhellis3
Sun Aug 30, 2020 6:22 pm
Forum: Computer Chess Club: General Topics
Topic: SF-NN embedded
Replies: 59
Views: 16433

Re: SF-NN embedded

I agree that in general it is solution for a problem which doesn't really exist (much like hybrid eval). It wastes bandwidth, storage space, and even makes things more cumbersome in some situations. For example, say you wanted to use an old net with a new compile. You have been using the (old) baked...
by jhellis3
Tue Aug 11, 2020 6:08 pm
Forum: Computer Chess Club: General Topics
Topic: this will be the merge of a lifetime : SF 80 Elo+
Replies: 121
Views: 28579

Re: this will be the merge of a lifetime : SF 80 Elo+

I doubt it would make that big of a difference really. There is one very small change I would make to nodchip's repo for the training data gen. Maybe I will submit a PR.....
by jhellis3
Mon Aug 10, 2020 12:48 am
Forum: Computer Chess Club: General Topics
Topic: this will be the merge of a lifetime : SF 80 Elo+
Replies: 121
Views: 28579

Re: this will be the merge of a lifetime : SF 80 Elo+

I've pushed the upstream merge to Crystal. Instead of a bool for NNUE, it has off, hybrid, and on as choices. I *think* everything is working fine, but haven't really had the time to go over all the changes closely. Planning on doing a 3.0 release in a couple of weeks, but the impatient can just com...
by jhellis3
Sun Aug 09, 2020 7:57 pm
Forum: Computer Chess Club: General Topics
Topic: this will be the merge of a lifetime : SF 80 Elo+
Replies: 121
Views: 28579

Re: this will be the merge of a lifetime : SF 80 Elo+

Lol, well.... there won't be any of that in Crystal, or at the very least there will be an option to just use the NN.
by jhellis3
Thu Jul 16, 2020 1:48 pm
Forum: Computer Chess Club: General Topics
Topic: Lizard-NNUE Experiment NOT bad with NNUE Net Evaluation.........
Replies: 53
Views: 8236

Re: Lizard-NNUE Experiment NOT bad with NNUE Net Evaluation.........

If what people are after is relatively strong low depth play, they are more than welcome to use Crystal. Free of charge....
by jhellis3
Thu Jul 09, 2020 2:28 am
Forum: Computer Chess Club: General Topics
Topic: Stockfish sewing machine
Replies: 28
Views: 2560

Re: Stockfish sewing machine

I think this is likely just a pruning issue deep in the tree, which can happen especially with relatively narrow & deep winning paths.

Reductions are just incredibly strong regardless of root depth.... and when things get reduced beyond a certain point things can go a bit sideways.