chess-model / chess_mcvs.py
test1978's picture
Add chess_mcvs.py
c2b5eda verified
Raw History Blame
29.7 kB
# chess_mcvs.py
"""
chess_mcvs.py - Full implementation analogous to breakthrough_mcvs.py
Features:
- Chess game logic using python-chess (full rules)
- Fixed 64×64 padded weighted adjacency matrix (Hilbert-ordered pieces)
- Piece-specific kappa values
- Queen-only promotions for flat 4096 policy head compatibility
- Policy and Value CNNs (lightweight with pooling for feasibility)
- HilbertOrderedZoneDatabase with add methods and proper k-NN zone score
- MCVSSearcher (guided λ-PUCT with policy/value/zone)
- UCTSearcher (baseline)
- train_networks function
- Ready for chess_mcvs_vs_uct.py tournament
Requires: pip install chess torch numpy
"""
import numpy as np
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torch.utils.data import DataLoader, TensorDataset
import chess
import random
import time
import os
import bisect
from collections import Counter
from typing import List, Dict, Tuple
# ============================================================================
# HILBERT CURVE UTILITIES
# ============================================================================
def xy2d(n: int, x: int, y: int) -> int:
d = 0
s = n // 2
while s > 0:
rx = 1 if (x & s) > 0 else 0
ry = 1 if (y & s) > 0 else 0
d += s * s * ((3 * rx) ^ ry)
if ry == 0:
if rx == 1:
x = n - 1 - x
y = n - 1 - y
x, y = y, x
s //= 2
return d
def matrix_to_hilbert_index(W: np.ndarray) -> int:
if W.size == 0:
return 0
W_flat = W.flatten()
max_idx = np.argmax(W_flat)
x = max_idx % 64
y = max_idx // 64
return xy2d(64, x, y)
# ============================================================================
# ABC MODEL & WEIGHTED MATRIX
# ============================================================================
from abc_model import ABCModelDynamic
from matrix_model import WeightedMatrixABC
# ============================================================================
# CHESS GAME LOGIC
# ============================================================================
class Chess:
def __init__(self):
self.board = chess.Board()
self.move_count = 0
self.position_history = Counter()
self._update_position_key()
def _update_position_key(self):
key = (self.board.fen(), self.board.turn)
self.position_history[key] += 1
def copy(self):
new_game = Chess()
new_game.board = self.board.copy()
new_game.move_count = self.move_count
new_game.position_history = self.position_history.copy()
return new_game
def undo_move(self) -> None:
"""
Undo the last move (pop) and revert bookkeeping (move_count, position_history).
This complements `apply_move` which increments move_count and updates
the position_history for the new state after a push. `undo_move` decrements
the same position record and pops the board to restore the previous state.
"""
# The current board fen/turn corresponds to the state that was last pushed;
# decrement its count before popping to fully revert the update.
key = (self.board.fen(), self.board.turn)
if key in self.position_history:
# decrement safely (don't remove key to avoid KeyError on repeated ops)
self.position_history[key] = max(0, self.position_history[key] - 1)
# pop the board and update counters
self.board.pop()
self.move_count = max(0, self.move_count - 1)
def get_legal_moves(self) -> List[chess.Move]:
moves = []
for move in self.board.legal_moves:
if move.promotion is None or move.promotion == chess.QUEEN:
moves.append(move)
return moves
def apply_move(self, move: chess.Move) -> None:
self.board.push(move)
self.move_count += 1
self._update_position_key()
def is_terminal(self) -> bool:
return self.board.is_game_over(claim_draw=True)
def check_winner(self) -> int:
if not self.is_terminal():
return 0
outcome = self.board.outcome(claim_draw=True)
if outcome is None or outcome.winner is None:
return 0
return 1 if outcome.winner == chess.WHITE else -1
def get_weighted_adjacency_matrix(self) -> np.ndarray:
occupied = []
for sq in chess.SQUARES:
piece = self.board.piece_at(sq)
if piece:
file = chess.square_file(sq)
rank = chess.square_rank(sq)
x = float(file) - 3.5
y = float(rank) - 3.5
color_val = 1.0 if piece.color == chess.WHITE else 1.1
type_val = {
chess.PAWN: 1.0,
chess.KNIGHT: 3.0,
chess.BISHOP: 3.2,
chess.ROOK: 5.0,
chess.QUEEN: 9.0,
chess.KING: 20.0
}[piece.piece_type]
kappa = np.array([1.0, color_val, type_val])
hilbert_d = xy2d(8, file, rank)
occupied.append((hilbert_d, (x, y), kappa))
occupied.sort(key=lambda t: t[0])
abc = ABCModelDynamic(n=2, t=1.0, T=1.41)
for _, pos, kappa_vec in occupied:
abc.add_piece(pos, delta_values=(1.0, kappa_vec[1], kappa_vec[2]), kappa_vector=kappa_vec)
w_calc = WeightedMatrixABC(abc, sigma=1.0)
W_var = w_calc.compute_weighted_matrix()
num = len(occupied)
W = np.zeros((64, 64))
if num > 0:
W[:num, :num] = W_var
return W
# ============================================================================
# NEURAL NETWORKS
# ============================================================================
class PolicyNetworkCNN(nn.Module):
def __init__(self):
super().__init__()
self.conv = nn.Sequential(
nn.Conv2d(1, 64, kernel_size=5, padding=2, stride=2),
nn.ReLU(),
nn.Conv2d(64, 128, kernel_size=5, padding=2, stride=2),
nn.ReLU(),
nn.Conv2d(128, 128, kernel_size=3, padding=1, stride=2),
nn.ReLU(),
)
self.fc = nn.Linear(128 * 8 * 8, 4096)
def forward(self, x):
x = self.conv(x)
x = x.view(x.size(0), -1)
return self.fc(x)
class ValueNetworkCNN(nn.Module):
def __init__(self):
super().__init__()
self.conv = nn.Sequential(
nn.Conv2d(1, 64, kernel_size=5, padding=2, stride=2),
nn.ReLU(),
nn.Conv2d(64, 128, kernel_size=5, padding=2, stride=2),
nn.ReLU(),
nn.Conv2d(128, 128, kernel_size=3, padding=1, stride=2),
nn.ReLU(),
)
self.fc = nn.Sequential(
nn.Linear(128 * 8 * 8, 512),
nn.ReLU(),
nn.Linear(512, 1),
nn.Tanh()
)
def forward(self, x):
x = self.conv(x)
x = x.view(x.size(0), -1)
return self.fc(x)
# ============================================================================
# ZONE DATABASE
# ============================================================================
class HilbertOrderedZoneDatabase:
def __init__(self, filepath: str = "chess_zone_db.npz", max_size: int = 100000):
self.filepath = filepath
self.max_size = max_size # Maximum matrices per zone
self.winning_matrices = []
self.winning_indices = []
self.losing_matrices = []
self.losing_indices = []
self.draw_matrices = []
self.draw_indices = []
self.load()
def add_winning_matrix(self, W: np.ndarray):
if W.shape != (64, 64):
return
hilbert_idx = matrix_to_hilbert_index(W)
pos = bisect.bisect_left(self.winning_indices, hilbert_idx)
self.winning_indices.insert(pos, hilbert_idx)
self.winning_matrices.insert(pos, W.copy())
def add_losing_matrix(self, W: np.ndarray):
if W.shape != (64, 64):
return
hilbert_idx = matrix_to_hilbert_index(W)
pos = bisect.bisect_left(self.losing_indices, hilbert_idx)
self.losing_indices.insert(pos, hilbert_idx)
self.losing_matrices.insert(pos, W.copy())
def add_draw_matrix(self, W: np.ndarray):
if W.shape != (64, 64):
return
hilbert_idx = matrix_to_hilbert_index(W)
pos = bisect.bisect_left(self.draw_indices, hilbert_idx)
self.draw_indices.insert(pos, hilbert_idx)
self.draw_matrices.insert(pos, W.copy())
def add_game_record(self, trajectory: List[Chess], result: int, sample_rate: float = 0.3):
"""Add game record with sampling to control database size."""
# Sample states: always keep first and last, random subset of middle
sampled_states = []
for i, state in enumerate(trajectory):
if i == 0 or i == len(trajectory) - 1:
# Always keep start and end positions
sampled_states.append(state)
elif random.random() < sample_rate:
# Randomly sample middle positions
sampled_states.append(state)
for state in sampled_states:
W = state.get_weighted_adjacency_matrix()
if result == 1:
self.add_winning_matrix(W)
elif result == -1:
self.add_losing_matrix(W)
else:
self.add_draw_matrix(W)
def save(self):
"""Save database with pruning if too large."""
# Prune if any zone exceeds 80% of max_size
threshold = int(self.max_size * 0.8)
if (len(self.winning_matrices) > threshold or
len(self.losing_matrices) > threshold or
len(self.draw_matrices) > threshold):
print(f"Pruning database before save...")
self.prune_database(target_size=int(self.max_size * 0.7))
try:
np.savez_compressed(
self.filepath,
winning=np.array(self.winning_matrices, dtype=object),
winning_indices=np.array(self.winning_indices),
losing=np.array(self.losing_matrices, dtype=object),
losing_indices=np.array(self.losing_indices),
draw=np.array(self.draw_matrices, dtype=object),
draw_indices=np.array(self.draw_indices),
)
print(f"Zone DB saved: W={len(self.winning_matrices)}, L={len(self.losing_matrices)}, D={len(self.draw_matrices)}")
except MemoryError:
print(f"MemoryError during save! Pruning more aggressively...")
self.prune_database(target_size=1000)
# Try again with much smaller size
np.savez_compressed(
self.filepath,
winning=np.array(self.winning_matrices, dtype=object),
winning_indices=np.array(self.winning_indices),
losing=np.array(self.losing_matrices, dtype=object),
losing_indices=np.array(self.losing_indices),
draw=np.array(self.draw_matrices, dtype=object),
draw_indices=np.array(self.draw_indices),
)
def load(self):
if os.path.exists(self.filepath):
try:
data = np.load(self.filepath, allow_pickle=True)
self.winning_matrices = list(data['winning'])
self.winning_indices = list(data['winning_indices'])
self.losing_matrices = list(data['losing'])
self.losing_indices = list(data['losing_indices'])
self.draw_matrices = list(data['draw'])
self.draw_indices = list(data['draw_indices'])
print(f"Loaded zone DB: W={len(self.winning_matrices)}, L={len(self.losing_matrices)}, D={len(self.draw_matrices)}")
except Exception as e:
print(f"Failed to load zone database: {e}")
def prune_database(self, target_size: int = 5000):
"""Keep only evenly-spaced diverse samples along Hilbert curve."""
def prune_zone(matrices, indices, target):
if len(matrices) <= target:
return matrices, indices
# Keep evenly spaced samples
step = max(1, len(matrices) // target)
keep_idx = list(range(0, len(matrices), step))[:target]
new_matrices = [matrices[i] for i in keep_idx]
new_indices = [indices[i] for i in keep_idx]
return new_matrices, new_indices
old_sizes = (len(self.winning_matrices), len(self.losing_matrices), len(self.draw_matrices))
self.winning_matrices, self.winning_indices = prune_zone(
self.winning_matrices, self.winning_indices, target_size
)
self.losing_matrices, self.losing_indices = prune_zone(
self.losing_matrices, self.losing_indices, target_size
)
self.draw_matrices, self.draw_indices = prune_zone(
self.draw_matrices, self.draw_indices, target_size
)
new_sizes = (len(self.winning_matrices), len(self.losing_matrices), len(self.draw_matrices))
print(f"Pruned: W {old_sizes}→{new_sizes}, L {old_sizes}→{new_sizes}, D {old_sizes}→{new_sizes}")
def compute_zone_score(self, W: np.ndarray, k: int = 5, beta: float = 0.5) -> float:
"""Compute zone guidance score Z(x(t), a) ∈ [-1, 1]."""
if W.shape != (64, 64):
return 0.0
def knn_similarity(matrices, k_val):
if len(matrices) == 0:
return 0.0
k_actual = min(k_val, len(matrices))
distances = []
for mat in matrices[:k_actual]:
dist = np.sum(np.abs(W - mat)) / (64.0 * 64.0)
distances.append(dist)
similarities = [1.0 - d for d in distances]
return np.mean(similarities)
zone_win = knn_similarity(self.winning_matrices, k)
zone_loss = knn_similarity(self.losing_matrices, k)
zone_draw = knn_similarity(self.draw_matrices, k)
Z = zone_win - zone_loss + beta * zone_draw
return float(np.clip(Z, -1.0, 1.0))
# ============================================================================
# TRAINING FUNCTION
# ============================================================================
def train_networks(policy_net, value_net, W_states, policies, values, epochs=5, batch_size=32, device='cpu', lr=0.001):
policy_net.train()
value_net.train()
opt_p = optim.Adam(policy_net.parameters(), lr=lr)
opt_v = optim.Adam(value_net.parameters(), lr=lr)
W_tensor = torch.tensor(np.array(W_states)[:, np.newaxis, :, :], dtype=torch.float32)
pi_tensor = torch.tensor(np.array(policies), dtype=torch.float32)
v_tensor = torch.tensor(np.array(values), dtype=torch.float32).unsqueeze(1)
dataset = TensorDataset(W_tensor, pi_tensor, v_tensor)
loader = DataLoader(dataset, batch_size=batch_size, shuffle=True)
for epoch in range(epochs):
total_loss = 0.0
for batch_W, batch_pi, batch_v in loader:
batch_W = batch_W.to(device)
batch_pi = batch_pi.to(device)
batch_v = batch_v.to(device)
logits = policy_net(batch_W)
# Fixed policy loss: KL divergence between network policy and MCTS visit distribution
policy_loss = F.kl_div(F.log_softmax(logits, dim=1), batch_pi, reduction='batchmean')
pred_v = value_net(batch_W)
value_loss = F.mse_loss(pred_v, batch_v)
loss = policy_loss + value_loss
total_loss += loss.item()
opt_p.zero_grad()
opt_v.zero_grad()
loss.backward()
# Optional but recommended: gradient clipping
torch.nn.utils.clip_grad_norm_(policy_net.parameters(), max_norm=1.0)
torch.nn.utils.clip_grad_norm_(value_net.parameters(), max_norm=1.0)
opt_p.step()
opt_v.step()
# Optional: print epoch loss for monitoring
print(f"Epoch {epoch+1}/{epochs} - Loss: {total_loss/len(loader):.4f}")
# ============================================================================
# MCVS SEARCHER
# ============================================================================
class MCVSSearcher:
def __init__(self, policy_net, value_net, zone_db, device='cpu',
cpuct=1.41, lambda_zone=0.0, k_zone=5, use_nets=True,
dirichlet_alpha: float = 0.3, dirichlet_noise_fraction: float = 0.25):
self.policy_net = policy_net.to(device) if policy_net and use_nets else None
self.value_net = value_net.to(device) if value_net and use_nets else None
self.zone_db = zone_db
self.device = device
self.cpuct = cpuct
self.lambda_zone = lambda_zone
self.k_zone = k_zone
self.use_nets = use_nets
# Dirichlet noise hyperparameters (alpha for the Dirichlet, and fraction)
self.dirichlet_alpha = float(dirichlet_alpha)
self.dirichlet_noise_fraction = float(dirichlet_noise_fraction)
def _rollout(self, game: Chess) -> float:
"""Simple random rollout for leaf evaluation when no value network.
Uses push/pop (apply_move / undo_move) on the provided `game` to avoid
allocating a full copy per rollout. The rollout will undo its own
moves before returning.
"""
depth = 0
max_depth = 60 # reduced depth for faster rollouts (configurable)
moves_played = []
while not game.is_terminal() and depth < max_depth:
moves = game.get_legal_moves()
if not moves:
break
m = random.choice(moves)
game.apply_move(m)
moves_played.append(m)
depth += 1
winner = game.check_winner()
# Undo rollout moves
for _ in range(len(moves_played)):
game.undo_move()
return float(winner) # +1, -1, or 0
class Node:
def __init__(self, prior: float = 0.0, zone: float = 0.0):
self.prior = prior
self.zone = zone
self.visit_count = 0
self.value_sum = 0.0
self.children: Dict[chess.Move, 'MCVSSearcher.Node'] = {}
def search_with_time_budget(self, game: Chess, time_budget: float) -> Tuple[Dict[chess.Move, int], int]:
root = self.Node()
start_time = time.time()
simulations = 0
while time.time() - start_time < time_budget:
current_game = game
node = root
path = []
moves_made = []
# Selection
while node.children:
path.append(node)
total_n = sum(c.visit_count for c in node.children.values()) + 1
best_score = -float('inf')
best_move = None
best_child = None
for move, child in node.children.items():
q = child.value_sum / child.visit_count if child.visit_count > 0 else 0.0
p_lambda = max(child.prior + self.lambda_zone * child.zone, 0.001)
u = self.cpuct * p_lambda * np.sqrt(total_n) / (1 + child.visit_count)
score = q + u
if score > best_score:
best_score = score
best_move = move
best_child = child
current_game.apply_move(best_move)
moves_made.append(best_move)
node = best_child
path.append(node)
# Terminal node?
if current_game.is_terminal():
value = float(current_game.check_winner())
else:
# Expansion & Evaluation
legal_moves = current_game.get_legal_moves()
if self.use_nets and self.policy_net is not None and self.value_net is not None:
# === Neural network branch ===
W = current_game.get_weighted_adjacency_matrix()
W_tensor = torch.from_numpy(W)[None, None, :, :].float().to(self.device)
with torch.no_grad():
logits = self.policy_net(W_tensor)[0]
value = self.value_net(W_tensor)[0, 0].item()
probs = F.softmax(logits, dim=0).cpu().numpy()
prior_dict = {}
prior_sum = 0.0
for m in legal_moves:
idx = m.from_square * 64 + m.to_square
p = float(probs[idx])
prior_dict[m] = p
prior_sum += p
if prior_sum > 0.0:
for m in prior_dict:
prior_dict[m] /= prior_sum
else:
p = 1.0 / len(legal_moves)
prior_dict = {m: p for m in legal_moves}
else:
# === Zone-only / no-nets branch ===
value = self._rollout(current_game)
prior_dict = {m: 1.0 / len(legal_moves) for m in legal_moves} if legal_moves else {}
# Dirichlet noise only at root and only when using nets.
# Use configured hyperparameters so the same alpha/fraction
# can be applied to both priors and zone guidance.
if len(path) == 1 and self.use_nets and legal_moves:
noise = np.random.dirichlet([self.dirichlet_alpha] * len(legal_moves))
for i, m in enumerate(legal_moves):
prior_dict[m] = (1 - self.dirichlet_noise_fraction) * prior_dict[m] + \
self.dirichlet_noise_fraction * noise[i]
# Renormalize
prior_sum = sum(prior_dict.values())
if prior_sum > 0:
for m in prior_dict:
prior_dict[m] /= prior_sum
# Create children with zone scores. Use apply/undo on the shared
# board to avoid allocating a full copy per child. Skip heavy
# weighted-matrix computations when zone guidance is disabled
# (lambda_zone == 0) or when the DB is empty.
zone_db_empty = (len(self.zone_db.winning_matrices) == 0 and
len(self.zone_db.losing_matrices) == 0 and
len(self.zone_db.draw_matrices) == 0)
need_zone = (self.lambda_zone != 0.0) and (not zone_db_empty)
for i, m in enumerate(legal_moves):
if need_zone:
current_game.apply_move(m)
child_W = current_game.get_weighted_adjacency_matrix()
child_Z = self.zone_db.compute_zone_score(child_W, k=self.k_zone)
current_game.undo_move()
else:
child_Z = 0.0
# If Dirichlet noise was added to priors at the root, apply the same
# noise (mapped to [-1,1]) to zone guidance so both signals share
# the same stochastic perturbation controlled by the same alpha.
if len(path) == 1 and self.use_nets and legal_moves:
z_noise = (noise[i] * 2.0) - 1.0 # map [0,1] -> [-1,1]
child_Z = (1 - self.dirichlet_noise_fraction) * child_Z + \
self.dirichlet_noise_fraction * z_noise
child = self.Node(prior=prior_dict.get(m, 0.0), zone=child_Z)
node.children[m] = child
# Backpropagation
v = value
for n in reversed(path):
n.visit_count += 1
n.value_sum += v
v = -v
# Undo moves that were applied during selection (the rollout
# already undid its own moves). This restores `game` to the root
# state for the next simulation.
for _ in range(len(moves_made)):
current_game.undo_move()
simulations += 1
visits = {move: child.visit_count for move, child in root.children.items()}
if not visits:
legal = game.get_legal_moves()
visits = {random.choice(legal): 1} if legal else {}
return visits, simulations
# ============================================================================
# UCT SEARCHER (your original from the provided file)
# ============================================================================
class UCTSearcher:
def __init__(self, cpuct=np.sqrt(2.0)):
self.cpuct = cpuct
class Node:
def __init__(self):
self.visit_count = 0
self.value_sum = 0.0
self.children: Dict[chess.Move, 'UCTSearcher.Node'] = {}
def _rollout(self, game: Chess) -> float:
depth = 0
max_depth = 60
moves_played = []
while not game.is_terminal() and depth < max_depth:
moves = game.get_legal_moves()
if not moves:
break
m = random.choice(moves)
game.apply_move(m)
moves_played.append(m)
depth += 1
if depth >= max_depth:
# undo rollout moves
for _ in range(len(moves_played)):
game.undo_move()
return 0.0
winner = game.check_winner()
# undo rollout moves
for _ in range(len(moves_played)):
game.undo_move()
if winner == 1:
return 1.0
elif winner == -1:
return -1.0
else:
return 0.0
def search_with_time_budget(self, game: Chess, time_budget: float) -> Tuple[Dict[chess.Move, int], int]:
root = self.Node()
start_time = time.time()
simulations = 0
while time.time() - start_time < time_budget:
current_game = game
node = root
path = [node]
moves_made = []
while True:
legal_moves = current_game.get_legal_moves()
if current_game.is_terminal():
winner = current_game.check_winner()
value = 1.0 if winner == 1 else -1.0 if winner == -1 else 0.0
break
if not legal_moves:
value = 0.0
break
unvisited_moves = [m for m in legal_moves if m not in node.children]
if unvisited_moves:
move = random.choice(unvisited_moves)
child = self.Node()
node.children[move] = child
current_game.apply_move(move)
moves_made.append(move)
path.append(child)
value = self._rollout(current_game)
break
best_score = -float('inf')
best_move = None
best_child = None
total_visits = node.visit_count
for move in legal_moves:
child = node.children[move]
Q = child.value_sum / (child.visit_count + 1e-8) if child.visit_count > 0 else 0.0
U = self.cpuct * np.sqrt(np.log(total_visits + 1) / (child.visit_count + 1e-8))
score = Q + U
if score > best_score:
best_score = score
best_move = move
best_child = child
current_game.apply_move(best_move)
moves_made.append(best_move)
node = best_child
path.append(node)
for n in reversed(path):
n.visit_count += 1
n.value_sum += value
value = -value
# Undo selection moves to restore root state
for _ in range(len(moves_made)):
current_game.undo_move()
simulations += 1
visits = {move: child.visit_count for move, child in root.children.items()}
if not visits:
legal = game.get_legal_moves()
return {random.choice(legal): 1} if legal else {}, simulations
return visits, simulations
if __name__ == "__main__":
print("chess_mcvs.py loaded successfully")