虽然xor问题还是跑不出来，但至少已经确定不是distance的错了

examples/xor.py

@@ -1,7 +1,6 @@
 from typing import Callable, List
 from functools import partial
 
-import jax
 import numpy as np
 
 from utils import Configer
@@ -18,8 +17,7 @@ def evaluate(forward_func: Callable) -> List[float]:
     :return:
     """
     outs = forward_func(xor_inputs)
-    outs = jax.device_get(outs)
-    fitnesses = -np.mean((outs - xor_outputs) ** 2, axis=(1, 2))
+    fitnesses = np.mean((outs - xor_outputs) ** 2, axis=(1, 2))
     # print(fitnesses)
     return fitnesses.tolist()  # returns a list