ValueError:输入张量必须至少具有秩5 (depthwise_conv2d)

import tensorflow as tf

# 假设原始输入张量维度为 (batch_size, height, width, channels)
original_input = tf.random.uniform((32, 64, 64, 3))

# 调整维度以匹配深度可分离卷积的要求
# 添加一个虚拟的 depth 维度
input_with_depth = tf.expand_dims(original_input, axis=1)

# 现在 input_with_depth 的维度为 (batch_size, 1, height, width, channels)
print(input_with_depth.shape)  # 输出: (32, 1, 64, 64, 3)

# 可以继续进行深度可分离卷积操作

import tensorflow as tf
from tensorflow.keras.layers import DepthwiseConv2D, Conv2D

# 创建一个简单的模型
model = tf.keras.Sequential([
    DepthwiseConv2D(kernel_size=(3, 3), padding='same', input_shape=(64, 64, 3)),
    Conv2D(filters=32, kernel_size=(1, 1), activation='relu')
])

# 编译模型
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])

# 生成虚拟数据
x_train = tf.random.uniform((100, 64, 64, 3))
y_train = tf.random.uniform((100,), maxval=10, dtype=tf.int32)

# 训练模型
model.fit(x_train, y_train, epochs=5)