在不影响尺度训练的情况下添加图层

import tensorflow as tf
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Conv2D, Input

# 假设我们有一个现有的模型
def create_base_model(input_shape):
    inputs = Input(shape=input_shape)
    x = Conv2D(32, (3, 3), activation='relu')(inputs)
    x = Conv2D(64, (3, 3), activation='relu')(x)
    model = Model(inputs, x)
    return model

# 创建基础模型
base_model = create_base_model((64, 64, 3))

# 添加新的卷积层
new_layer = Conv2D(128, (3, 3), activation='relu')(base_model.output)

# 构建新模型
new_model = Model(base_model.input, new_layer)

# 查看新模型结构
new_model.summary()

# 编译和训练新模型
new_model.compile(optimizer='adam', loss='sparse_categorical_crossentropy', metrics=['accuracy'])

# 假设我们有训练数据 X_train 和 y_train
# new_model.fit(X_train, y_train, epochs=10, batch_size=32)