在张量流中,我制作了一个将哈希作为输入的常规网络。作为一个例子,我使用了内置的 python hash()函数(是的,它在每个会话中都改变了盐,但这是一个例子)代码是这样的:from time import timest = time()import tensorflow as tfprint(time() - st)import numpy as npimport chessimport atexitfrom numpy import shapedata = open("data.data", "r").readlines()[:10000]targets = open("targets.data", "r").readlines()[:10000]boards_data = []new_targets = []for i in data: boards_data.append(hash(i))for i in targets: new_targets.append(float(i))print(len(new_targets))print(len(boards_data))print(np.array(new_targets))print(np.array(boards_data))def create_model(): model = tf.keras.models.Sequential() model.add(tf.keras.layers.Reshape((1,1,1))) model.add(tf.keras.layers.Dense(1000, activation="tanh")) model.add(tf.keras.layers.Flatten()) model.add(tf.keras.layers.Dense(1, activation='tanh')) model.compile(loss="mse", optimizer="adam", metrics=['accuracy']) return modelmodel = create_model()model.fit(np.array(boards_data), np.array(new_targets), epochs=10)model.predict(np.array(hash("8/6P1/5k1K/6r1/8/8/8/8 b - - 0 83")))错误在预测中。我在如何修复 Tensorflow 中的“IndexError:列表索引超出范围”中看到了 conv2d 示例 ,但事实并非如此......和回溯:Traceback (most recent call last): File "/Volumes/POOPOO USB/lichess-bot/engines/engine2/nn_evaluation/nn_evaluation2.py", line 36, in <module> model.predict(np.array(hash("8/6P1/5k1K/6r1/8/8/8/8 b - - 0 83"))) File "/Users/ofek/Library/Python/3.8/lib/python/site-packages/tensorflow/python/keras/engine/training.py", line 130, in _method_wrapper return method(self, *args, **kwargs) File "/Users/ofek/Library/Python/3.8/lib/python/site-packages/tensorflow/python/keras/engine/training.py", line 1569, in predict data_handler = data_adapter.DataHandler( File "/Users/ofek/Library/Python/3.8/lib/python/site-packages/tensorflow/python/keras/engine/data_adapter.py", line 1105, in __init__ self._adapter = adapter_cls(
1 回答
至尊宝的传说
TA贡献1789条经验 获得超10个赞
问题是您正在从哈希值创建一个 0d numpy 字符串。预测只能在至少具有一维的数组上运行。您可以检查您的散列值是否为 0d:
print(np.array(hash("8/6P1/5k1K/6r1/8/8/8/8 b - - 0 83")).shape) # outputs: ()
与将哈希值放入列表相比:
print(np.array([hash("8/6P1/5k1K/6r1/8/8/8/8 b - - 0 83")]).shape) # outputs: (1,)
第二个np.array
预测运行没有错误。
添加回答
举报
0/150
提交
取消