Question

我正在实施一个 ConvNet 来预测游戏小地图中的胜局。但是我在训练网络时遇到了问题。

当我运行以下代码时，出现错误：

ValueError: Input 0 is incompatible with layer model: expected shape=(None, 312, 312, 3), found shape=(312, 312, 3)

代码：

import tensorflow as tf
from tensorflow import keras
from tensorflow.python.keras.backend import relu
import numpy as np

#input layers
minimapInput = keras.Input(shape = (312, 312, 3), name="minimap_image")

#layers for mini input
mini = keras.layers.Conv2D(filters=32, kernel_size=3, padding='same', activation='relu')(minimapInput)
mini = keras.layers.Conv2D(filters=32, kernel_size=3, padding='same', activation='relu')(mini)
mini = keras.layers.AveragePooling2D(pool_size=(2,2), strides=2)(mini)
mini = keras.layers.Flatten()(mini)
mini = keras.layers.Dense(250, activation='relu')(mini)
mini = keras.layers.Dense(200, activation='relu')(mini)
mini = keras.layers.Dense(100, activation='relu')(mini)

#output
ctRoundWin = keras.layers.Dense(1, activation=keras.activations.softmax)(mini)

#model
model = keras.Model(inputs=minimapInput, outputs=ctRoundWin)

#creating/reading train data
def readFrames(file, frames, width, height, depth=3):
    output = np.zeros((frames, height, width, depth))
    for frame in range(frames):
        for i in range(depth):
            for j in range(height):
                for k in range(width):
                    try:
                        output[frame, j, k, i] = ord(file.read(1)[0])  
                    except IndexError:
                        output[frame, j, k, i] = 0    
    return output

y_pred = np.zeros(167)
frames = 167
minifile = open("C:\\Users\\s-wel\\OneDrive\\Desktop\\CSAI\\test\\mini.txt", "r")
mini = readFrames(minifile, frames, 312, 312)
minifile.close

#creating tf Dataset
data = tf.data.Dataset.from_tensor_slices((mini, y_pred))

#running it
model.compile(optimizer = keras.optimizers.Adam(learning_rate=0.001), loss = keras.losses.categorical_crossentropy, metrics = keras.metrics.binary_accuracy)
model.fit(data, epochs=1)

我发现的所有其他解决方案都没有真正解决这个问题。

Answer 1

我自己找到了解决方案，问题出在我创建的数据集中。我通过删除该行来修复它

data = tf.data.Dataset.from_tensor_slices((mini, y_pred))

并将最后一行更改为

model.fit([mini, left, right, score], y_pred)

感谢所有帮助过的人

ValueError：输入 0 与层模型不兼容：预期形状=（无，x），找到形状=（x）

1 个答案: