issues
search
cosmobobak
/
viridithas
A superhuman chess engine.
MIT License
94
stars
8
forks
source link
issues
Newest
Newest
Most commented
Recently updated
Oldest
Least commented
Least recently updated
make tuning much more SPSA-compatible
#101
cosmobobak
closed
9 months ago
0
Bugfix for OOB in singular verification
#100
cosmobobak
closed
10 months ago
0
Update network history with `newcomb`.
#99
cosmobobak
closed
10 months ago
0
Update network history with test `jupiter`.
#98
cosmobobak
closed
10 months ago
0
Add support for custom bucketing.
#97
cosmobobak
closed
10 months ago
0
New network, ID `artemis`.
#96
cosmobobak
closed
11 months ago
0
Use calloc for hashtable initialisation.
#95
cosmobobak
closed
11 months ago
0
Remove EPD test-suite functionality.
#94
cosmobobak
closed
11 months ago
0
Boost history adjustment when static eval is low.
#93
cosmobobak
closed
11 months ago
0
Reduce only quiets using !PV
#92
cosmobobak
closed
11 months ago
0
Do TT-reductions in predicted Cut Nodes.
#91
cosmobobak
closed
11 months ago
0
Introduce `cut_node` flag, use in LMR.
#90
cosmobobak
closed
11 months ago
0
Noisy evaluation
#89
lucasart
opened
1 year ago
3
Modify LMR re-search depth using value.
#88
cosmobobak
closed
1 year ago
1
Code quality improvements.
#87
cosmobobak
closed
1 year ago
0
Allow LMR to reduce winning captures.
#86
cosmobobak
closed
1 year ago
0
Extract aspiration loop into separate function
#85
cosmobobak
closed
1 year ago
0
Make aspiration windows move more smoothly.
#84
cosmobobak
closed
1 year ago
0
compilation errors : Rust version and env. var.
#83
tissatussa
closed
1 year ago
4
Increase hidden layer size from 1024x2 to 1536x2.
#82
cosmobobak
closed
1 year ago
0
Use `include_bytes!` for attack SquareSets
#81
cosmobobak
closed
1 year ago
0
index main history by whether the from- and to-squares are attacked by the opponent
#80
Ciekce
closed
1 year ago
2
New non-bucketed 1024x2 network.
#79
cosmobobak
closed
1 year ago
0
Decrease Singular Extension Margin.
#78
cosmobobak
closed
1 year ago
0
Support bucketless relnets more robustly.
#77
cosmobobak
closed
1 year ago
0
Extend singular moves more often.
#76
cosmobobak
closed
1 year ago
0
Refactor TT Probing
#75
cosmobobak
closed
1 year ago
1
Parallel initialisation of TT memory.
#74
cosmobobak
closed
1 year ago
0
Fix for SMP nodecount reporting
#73
cosmobobak
closed
1 year ago
0
Generate training data in binary format.
#72
cosmobobak
closed
1 year ago
0
cache static evaluation in TT
#71
cosmobobak
closed
1 year ago
1
Switch to HalfKA feature set.
#70
cosmobobak
closed
1 year ago
1
Squash changes.
#69
cosmobobak
closed
1 year ago
0
Use more time.
#68
cosmobobak
closed
1 year ago
0
Improve code quality.
#67
cosmobobak
closed
1 year ago
0
Squash all of the HalfKA work into single commit.
#66
cosmobobak
closed
1 year ago
0
faster inference
#65
cosmobobak
closed
1 year ago
0
Fix data generation.
#64
cosmobobak
closed
1 year ago
0
Switch to an i8 output layer.
#63
cosmobobak
closed
1 year ago
0
switch to much larger network
#62
cosmobobak
closed
1 year ago
0
TTMove bugfix, better code.
#61
cosmobobak
closed
1 year ago
0
Prevent NMP if TT indicates it will fail.
#60
cosmobobak
closed
1 year ago
0
Refactor to type-safe SquareSet type.
#59
cosmobobak
closed
1 year ago
0
Update history to be move iteration depdent.
#58
cosmobobak
closed
1 year ago
0
Material scaling for NNUE eval.
#57
cosmobobak
closed
1 year ago
0
Use a greater fraction of the clock increment.
#56
cosmobobak
closed
1 year ago
0
Smaller net trained on FRC + cleaned-up old data, ID: `gemstone`
#55
cosmobobak
closed
1 year ago
0
Subtree-Size Time Management
#54
cosmobobak
closed
1 year ago
0
Return `tt_value` rather than alpha/beta on ttcut
#53
cosmobobak
closed
1 year ago
1
Only update best-move when alpha is raised.
#52
cosmobobak
closed
1 year ago
0
Previous
Next