v1.0.0 + Format (#96)

Bench: 184439

Author: jw1912

.cargo/config.toml

@@ -1,4 +1,5 @@
 /target
+/.cargo
 *-policy-*
 *.data
 *.binpack

.gitignore

@@ -1,6 +1,6 @@
 [package]
 name = "monty"
-version = "0.1.0"
+version = "1.0.0"
 edition = "2021"
 authors = ["Jamie Whiting"]
 

Cargo.lock

@@ -4,22 +4,37 @@
 #### "MCTS is cool."
 
 ![License](https://img.shields.io/github/license/jw1912/monty?style=for-the-badge)
-[![GitHub release (latest by date)](https://img.shields.io/github/v/release/jw1912/monty?style=for-the-badge)](https://github.com/jw1912/akimbo/releases/latest)
-[![Commits](https://img.shields.io/github/commits-since/jw1912/monty/latest?style=for-the-badge)](https://github.com/jw1912/akimbo/commits/main)
+[![GitHub release (latest by date)](https://img.shields.io/github/v/release/jw1912/monty?style=for-the-badge)](https://github.com/jw1912/monty/releases/latest)
+[![Commits](https://img.shields.io/github/commits-since/jw1912/monty/latest?style=for-the-badge)](https://github.com/jw1912/monty/commits/main)
 
 </div>
 
 ## Compiling
 You need a network file, and run the following command
 ```
-make <name of game> EVALFILE=<relative path to network file>
+make <name of game (lowercase)> EXE=<output path> EVALFILE=<path to network file>
 ```
-to compile monty for the given game.
+to compile monty for the given game (chess, ataxx, shatranj).
 
-## Supported Games
-- Ataxx
-- Chess
-- Shatranj
+## Originality Status
+
+The first version (0.1.0) used external data for value networks and self-generated policy data. The networks were then reset
+completely, and all future versions are trained exclusively on monty's own data, generated from scratch with uniform policy
+and material counting value.
+
+## Credits
+Thanks to everyone at SWE as usual, in particular Cosmo (Viridithas) and Zuppa (Alexandria), for helping with data generation, and Plutie, for running an LTC tune.
+
+## ELO
+
+<div align="center">
+
+| Version | Release Date | CCRL 40/15 | CCRL Blitz | CCRL FRC | Notes |
+| :-: | :-: | :-: | :-: | :-: | :-: |
+| [1.0.0](https://github.com/jw1912/monty/releases/tag/v1.0.0) | 28th May 2024 | TBD | TBD | TBD | Fully Original Data |
+| [0.1.0](https://github.com/jw1912/monty/releases/tag/v0.1.0) | 26th March 2024 | - | - | 2974 | First Release |
+
+</div>
 
 ## How it works
 

Cargo.toml

@@ -1,10 +1,10 @@
-#[cfg(not(feature="nonet"))]
+#[cfg(not(feature = "nonet"))]
 use std::env;
 
-#[cfg(not(feature="nonet"))]
+#[cfg(not(feature = "nonet"))]
 const DEFAULT_PATH: &str = "resources/net.network";
 
-#[cfg(not(feature="nonet"))]
+#[cfg(not(feature = "nonet"))]
 fn main() {
     println!("cargo:rerun-if-env-changed=EVALFILE");
     println!("cargo:rerun-if-changed=resources/chess.network");
@@ -15,5 +15,5 @@ fn main() {
     }
 }
 
-#[cfg(feature="nonet")]
+#[cfg(feature = "nonet")]
 fn main() {}

README.md

@@ -1,5 +1,8 @@
 use datagen::{parse_args, run_datagen};
-use monty::{chess::{PolicyNetwork, ValueNetwork}, UciLike};
+use monty::{
+    chess::{PolicyNetwork, ValueNetwork},
+    UciLike,
+};
 
 #[repr(C)]
 struct Nets(ValueNetwork, PolicyNetwork);

build.rs

@@ -1,5 +1,8 @@
 use datagen::{parse_args, run_datagen};
-use monty::{shatranj::{PolicyNetwork, Shatranj}, ValueNetwork};
+use monty::{
+    shatranj::{PolicyNetwork, Shatranj},
+    ValueNetwork,
+};
 
 #[repr(C)]
 struct Nets(ValueNetwork<768, 8>, PolicyNetwork);
@@ -15,4 +18,4 @@ fn main() {
     let (threads, book, policy) = parse_args(args);
 
     run_datagen::<Shatranj, 112>(1_000, threads, policy, "Shatranj", &POLICY, &VALUE, book);
-}
+}

datagen/src/bin/monty.rs

@@ -1,5 +1,8 @@
 use datagen::{parse_args, run_datagen};
-use monty::{ataxx::{Ataxx, PolicyNetwork}, ValueNetwork};
+use monty::{
+    ataxx::{Ataxx, PolicyNetwork},
+    ValueNetwork,
+};
 
 #[repr(C)]
 struct Nets(ValueNetwork<2916, 256>, PolicyNetwork);
@@ -15,4 +18,4 @@ fn main() {
     let (threads, book, policy) = parse_args(args);
 
     run_datagen::<Ataxx, 114>(1_000, threads, policy, "Ataxx", &POLICY, &VALUE, book);
-}
+}

datagen/src/bin/montyj.rs

@@ -197,7 +197,7 @@ impl Binpack {
             let mov = u16::from_le_bytes([buf[0], buf[1]]);
             let score = i16::from_le_bytes([buf[2], buf[3]]);
 
-            f(&mut board, &castling,mov.into(), score, result);
+            f(&mut board, &castling, mov.into(), score, result);
         }
 
         Ok(())

datagen/src/bin/montyxx.rs

@@ -191,4 +191,4 @@ impl Binpack {
 
         Ok(())
     }
-}
+}

datagen/src/impls/chess.rs

@@ -8,7 +8,11 @@ pub use thread::{write, DatagenThread};
 use monty::{ataxx::Ataxx, chess::Chess, shatranj::Shatranj, GameRep};
 
 use std::{
-    env::Args, fs::File, io::Read, sync::atomic::{AtomicBool, Ordering}, time::Duration
+    env::Args,
+    fs::File,
+    io::Read,
+    sync::atomic::{AtomicBool, Ordering},
+    time::Duration,
 };
 
 pub type AtaxxPolicyData = PolicyData<Ataxx, 114>;
@@ -145,11 +149,11 @@ pub fn parse_args(mut args: Args) -> (usize, Option<String>, bool) {
                 1 => {
                     threads = Some(arg.parse().expect("can't parse"));
                     mode = 0;
-                },
+                }
                 2 => {
                     book = Some(arg);
                     mode = 0;
-                },
+                }
                 _ => println!("unrecognised argument {arg}"),
             },
         }

datagen/src/impls/shatranj.rs

@@ -22,7 +22,12 @@ pub struct DatagenThread<'a, T: DatagenSupport> {
 }
 
 impl<'a, T: DatagenSupport> DatagenThread<'a, T> {
-    pub fn new(id: u32, params: MctsParams, stop: &'a AtomicBool, book: Option<Vec<&'a str>>) -> Self {
+    pub fn new(
+        id: u32,
+        params: MctsParams,
+        stop: &'a AtomicBool,
+        book: Option<Vec<&'a str>>,
+    ) -> Self {
         Self {
             id,
             rng: Rand::with_seed(),
@@ -92,7 +97,6 @@ impl<'a, T: DatagenSupport> DatagenThread<'a, T> {
             T::from_fen(T::STARTPOS)
         };
 
-
         // play 8 or 9 random moves
         for _ in 0..(8 + (self.rng.rand_int() % 2)) {
             let mut moves = Vec::new();
@@ -129,7 +133,8 @@ impl<'a, T: DatagenSupport> DatagenThread<'a, T> {
 
         // play out game
         loop {
-            let mut searcher = Searcher::new(position.clone(), tree, self.params.clone(), policy, value);
+            let mut searcher =
+                Searcher::new(position.clone(), tree, self.params.clone(), policy, value);
 
             let (bm, score) = searcher.search(limits, false, &mut 0, &None);
 

datagen/src/lib.rs

@@ -1,10 +1,12 @@
-use monty::{chess::{ValueNetwork, PolicyNetwork, Uci}, UciLike};
+use monty::{
+    chess::{PolicyNetwork, Uci, ValueNetwork},
+    UciLike,
+};
 
 #[repr(C)]
 struct Nets(ValueNetwork, PolicyNetwork);
 
-const NETS: Nets =
-    unsafe { std::mem::transmute(*include_bytes!("../../resources/net.network")) };
+const NETS: Nets = unsafe { std::mem::transmute(*include_bytes!("../../resources/net.network")) };
 
 static VALUE: ValueNetwork = NETS.0;
 static POLICY: PolicyNetwork = NETS.1;
@@ -19,4 +21,4 @@ fn main() {
     }
 
     Uci::run(&POLICY, &VALUE);
-}
+}

datagen/src/thread.rs

@@ -1,10 +1,12 @@
-use monty::{shatranj::{Uci, PolicyNetwork}, UciLike, ValueNetwork};
+use monty::{
+    shatranj::{PolicyNetwork, Uci},
+    UciLike, ValueNetwork,
+};
 
 #[repr(C)]
 struct Nets(ValueNetwork<768, 8>, PolicyNetwork);
 
-const NETS: Nets =
-    unsafe { std::mem::transmute(*include_bytes!("../../resources/net.network")) };
+const NETS: Nets = unsafe { std::mem::transmute(*include_bytes!("../../resources/net.network")) };
 
 static VALUE: ValueNetwork<768, 8> = NETS.0;
 static POLICY: PolicyNetwork = NETS.1;
@@ -19,4 +21,4 @@ fn main() {
     }
 
     Uci::run(&POLICY, &VALUE);
-}
+}

src/bin/monty.rs

@@ -1,11 +1,12 @@
-use monty::{ataxx::{Uai, PolicyNetwork}, UciLike, ValueNetwork};
-
+use monty::{
+    ataxx::{PolicyNetwork, Uai},
+    UciLike, ValueNetwork,
+};
 
 #[repr(C)]
 struct Nets(ValueNetwork<2916, 256>, PolicyNetwork);
 
-const NETS: Nets =
-    unsafe { std::mem::transmute(*include_bytes!("../../resources/net.network")) };
+const NETS: Nets = unsafe { std::mem::transmute(*include_bytes!("../../resources/net.network")) };
 
 static VALUE: ValueNetwork<2916, 256> = NETS.0;
 static POLICY: PolicyNetwork = NETS.1;
@@ -20,4 +21,4 @@ fn main() {
     }
 
     Uai::run(&POLICY, &VALUE);
-}
+}

src/bin/montyj.rs

@@ -15,10 +15,7 @@ pub trait UciLike: Sized {
 
     fn options();
 
-    fn run(
-        policy: &<Self::Game as GameRep>::Policy,
-        value: &<Self::Game as GameRep>::Value,
-    ) {
+    fn run(policy: &<Self::Game as GameRep>::Policy, value: &<Self::Game as GameRep>::Value) {
         let mut prev = None;
         let mut pos = Self::Game::default();
         let mut params = Self::Game::default_mcts_params();
@@ -41,7 +38,16 @@ pub trait UciLike: Sized {
                 "setoption" => setoption(&commands, &mut params, &mut report_moves, &mut tree),
                 "position" => position(commands, &mut pos, &mut prev, &mut tree),
                 "go" => {
-                    let res = go(&commands, tree, prev, &pos, &params, report_moves, policy, value);
+                    let res = go(
+                        &commands,
+                        tree,
+                        prev,
+                        &pos,
+                        &params,
+                        report_moves,
+                        policy,
+                        value,
+                    );
 
                     tree = res.0;
                     prev = Some(res.1);

src/bin/montyxx.rs

@@ -54,7 +54,8 @@ pub trait GameRep: Clone + Default + Send + Sync {
 
     fn get_policy_feats(&self) -> Self::PolicyInputs;
 
-    fn get_policy(&self, mov: Self::Move, feats: &Self::PolicyInputs, policy: &Self::Policy) -> f32;
+    fn get_policy(&self, mov: Self::Move, feats: &Self::PolicyInputs, policy: &Self::Policy)
+        -> f32;
 
     fn get_value(&self, value: &Self::Value) -> i32;
 

src/comm.rs

@@ -147,7 +147,7 @@ impl GameRep for Chess {
         &self,
         mov: Self::Move,
         (feats, threats): &Self::PolicyInputs,
-        policy: &Self::Policy
+        policy: &Self::Policy,
     ) -> f32 {
         policy.get(&self.board, &mov, feats, *threats)
     }

src/games.rs

@@ -160,11 +160,7 @@ impl Board {
 
     pub fn map_value_features<F: FnMut(usize)>(&self, mut f: F) {
         let flip = self.stm() == Side::BLACK;
-        let hm = if self.king_index() % 8 > 3 {
-            7
-        } else {
-            0
-        };
+        let hm = if self.king_index() % 8 > 3 { 7 } else { 0 };
 
         let mut threats = self.threats_by(self.stm() ^ 1);
         let mut defences = self.threats_by(self.stm());

src/games/chess.rs

@@ -42,7 +42,10 @@ impl Castling {
             rook_files = [[0, 7]; 2];
         }
 
-        let mut ret = Self { rook_files, ..Default::default() };
+        let mut ret = Self {
+            rook_files,
+            ..Default::default()
+        };
 
         ret.castle_mask[usize::from(rook_files[0][0])] = 7;
         ret.castle_mask[usize::from(rook_files[0][1])] = 11;

src/games/chess/board.rs

@@ -4,7 +4,7 @@ const SCALE: i32 = 400;
 
 #[repr(C)]
 pub struct ValueNetwork {
-    l1: Layer<{768 * 4}, 512>,
+    l1: Layer<{ 768 * 4 }, 512>,
     l2: Layer<512, 16>,
     l3: Layer<16, 16>,
     l4: Layer<16, 16>,
@@ -72,7 +72,7 @@ struct Accumulator<const HIDDEN: usize> {
     vals: [f32; HIDDEN],
 }
 
-impl<const HIDDEN: usize>  Accumulator<HIDDEN> {
+impl<const HIDDEN: usize> Accumulator<HIDDEN> {
     fn madd(&mut self, mul: f32, other: &Self) {
         for (i, &j) in self.vals.iter_mut().zip(other.vals.iter()) {
             *i += mul * j;

src/games/chess/frc.rs

@@ -129,7 +129,12 @@ impl GameRep for Shatranj {
         feats
     }
 
-    fn get_policy(&self, mov: Self::Move, feats: &goober::SparseVector, policy: &Self::Policy) -> f32 {
+    fn get_policy(
+        &self,
+        mov: Self::Move,
+        feats: &goober::SparseVector,
+        policy: &Self::Policy,
+    ) -> f32 {
         policy.get(&self.board, &mov, feats)
     }
 

src/games/chess/value.rs

@@ -6,7 +6,7 @@ mod tree;
 mod value;
 
 pub use comm::UciLike;
-pub use games::{GameRep, GameState, ataxx, chess, shatranj};
+pub use games::{ataxx, chess, shatranj, GameRep, GameState};
 pub use mcts::{Limits, Searcher};
 pub use params::MctsParams;
 pub use tree::Tree;