Question

'我想通过keras.model解决分类问题，但是在运行model.fit之后，我遇到了尺寸错误。我已经运行以下代码：'

class Program
{
    static void Main(string[] args)
    {
        var input = @"[1 (832, 315, title1)]
[2 (232, 115, title2)]
[3 (432, 215, title3)]
[4 (132, 315, title4)]
[5 (932, 415, title5)]
";
        TextToDictionary(input.AsSpan());
    }

    private static Dictionary<int, Tuple<double, double, string>> TextToDictionary(ReadOnlySpan<char> span)
    {
        var output = new Dictionary<int, Tuple<double, double, string>>();

        while (span.Length > 0)
        {
            span = EatUntilToken(span, "[");
            span = ConsumeToAndEatToken(span, "(", out var int1);
            span = ConsumeToAndEatToken(span, ",", out var double1);
            span = ConsumeToAndEatToken(span, ",", out var double2);
            span = ConsumeToAndEatToken(span, ")", out var string1);
            span = EatUntilToken(span, Environment.NewLine);

            output.Add(int.Parse(int1), Tuple.Create(double.Parse(double1), double.Parse(double2), string1.ToString()));
        }

        return output;
    }

    private static ReadOnlySpan<char> ConsumeToAndEatToken(ReadOnlySpan<char> instruction, string token, out ReadOnlySpan<char> value)
    {
        var indexOf = instruction.IndexOf(token);
        if (indexOf < 0)
            throw new Exception($"String {instruction.ToString()} does not contain the expected token {token}");

        value = instruction[..(indexOf)];
        return instruction[(indexOf + token.Length)..];
    }

    private static ReadOnlySpan<char> EatUntilToken(ReadOnlySpan<char> instruction, string token)
    {
        var indexOf = instruction.IndexOf(token);
        if (indexOf < 0)
            throw new Exception($"String {instruction.ToString()} does not contain the expected token {token}");

        return instruction[(indexOf + token.Length)..];
    }
}

'输出：'

print(X_train.shape)
print(y_train.shape)

'运行以下代码后，'

(2588, 39436)
(2588, 6)

model = keras.Sequential(
    [
        keras.Input(shape=(39436,1)),
        layers.Conv1D(32, kernel_size=3, strides=5, activation="relu"),
        layers.MaxPooling1D(pool_size=10),
        layers.Conv1D(64, kernel_size=3, strides=5, activation="relu"),
        layers.MaxPooling1D(pool_size=10),
        layers.Flatten(),
        layers.Dropout(0.5),
        layers.Dense(num_classes, activation="softmax"),
    ]
)

model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])

“我给这个错误：”

ValueError：用户代码中：

model.fit(X_train, y_train, batch_size=128, epochs=15, validation_split=0.3)

“如果您能指导我，那将会是什么问题？”

Answer 1

根据错误消息，您的输入数组的形状为[None, 39436]。但是，在Input层中，您传递的形状为[39436, 1]，该形状与[None, 39436, 1]相匹配，其中None代表样本维。这就是抛出的错误。

您需要通过以下方式匹配形状：

1。重塑输入数据的形状，使其形状为[samples, 39436, 1]，而模型结构保持不变。

这可以通过以下操作来完成（假设train_X是您的输入功能）：

train_X = np.expand_dims(train_X, axis=2)

np.expand_dims在数组形状的索引2处向数组添加新维度。因此，这里将[samples, 39436]重塑为[samples, 39436, 1]。

引用：NumPy docs for expand_dims

OR

2。在input_shape层中更改 Input 参数，以接受[39436,]的形状，以匹配您的数据。

keras分类问题，model.fit命令中的错误

1 个答案: