Update selfchess.py

selfchess.py

@@ -5,7 +5,6 @@ import chess
 import os
 import chess.engine as eng
 import torch.multiprocessing as mp
-import random
 
 # CONFIGURATION
 CONFIG = {
@@ -149,13 +148,17 @@ def game_gen(engine_side):
             evaling = {}
             for move in board.legal_moves:
                 board.push(move)
-                evaling[move] = -search(board, depth=CONFIG["search_depth"], alpha=float('-inf'), beta=float('inf')) * random.uniform(1-CONFIG["epsilon"],1+CONFIG["epsilon"])
+                evaling[move] = -search(board, depth=CONFIG["search_depth"], alpha=float('-inf'), beta=float('inf'))
                 board.pop()
             
             if not evaling:
                 break
             
-            move = max(evaling, key=evaling.get)
+            keys = list(evaling.keys())
+            logits = torch.tensor(list(evaling.values())).to(device)
+            probs = torch.softmax(logits,dim=0)
+            idx = torch.multinomial(probs,num_samples=1)
+            move = keys[idx]
         else:
             result = engine.play(board, lim)
             move = result.move