Update selfchess.py

selfchess.py

@@ -157,11 +157,11 @@ def game_gen(engine_side):
             keys = list(evaling.keys())
             logits = torch.tensor(list(evaling.values())).to(device)
             probs = torch.softmax(logits,dim=0)
-            bests = torch.multinomial(probs,num_samples=CONFIG["epsilon"])
-            bestl = []
-            for idx in bests:
-                bestl += [logits[idx].item()]
-            move = keys[torch.argmax(bestl).item()]
+            epsilon = min(CONFIG["epsilon"],len(keys))
+            bests = torch.multinomial(probs,num_samples=epsilon,replacement=False)
+            best_idx = bests[torch.argmax(logits[bests])]
+            move = keys[best_idx.item()]
+            
         else:
             result = engine.play(board, lim)
             move = result.move