我试图在自己的模型上实现迁移学习，但失败了。我的实现遵循了以下指南
https://keras.io/guides/transfer_learning/
tensoflow 2.4.1Keras 2.4.3
旧模型（效果非常好）：
model = Sequential()inputShape = (256, 256, 3)chanDim = -1 # CONV => RELU => POOLmodel.add(Conv2D(32, (3, 3), padding="same", input_shape=inputShape))model.add(Activation("relu"))model.add(BatchNormalization(axis=chanDim))model.add(MaxPooling2D(pool_size=(3, 3)))model.add(Dropout(0.25))# (CONV => RELU) * 2 => POOLmodel.add(Conv2D(64, (3, 3), padding="same"))model.add(Activation("relu"))model.add(BatchNormalization(axis=chanDim))model.add(Conv2D(64, (3, 3), padding="same"))model.add(Activation("relu"))model.add(BatchNormalization(axis=chanDim))model.add(MaxPooling2D(pool_size=(2, 2)))model.add(Dropout(0.25))# (CONV => RELU) * 2 => POOLmodel.add(Conv2D(128, (3, 3), padding="same"))model.add(Activation("relu"))model.add(BatchNormalization(axis=chanDim))model.add(Conv2D(128, (3, 3), padding="same"))model.add(Activation("relu"))model.add(BatchNormalization(axis=chanDim))model.add(MaxPooling2D(pool_size=(2, 2)))model.add(Dropout(0.25))# 第一个（也是唯一一个）全连接层 => RELU 层model.add(Flatten())model.add(Dense(1024))model.add(Activation("relu"))model.add(BatchNormalization())model.add(Dropout(0.5))# softmax 分类器model.add(Dense(classes))model.add(Activation("softmax"))迁移学习：
old_model = load_model('old.model')# 移除顶部的2个激活层for i in range(2): old_model.pop()# 标记加载的层为不可训练for layer in old_model.layers: layer.trainable = False# 初始化新模型in_puts = Input(shape=(256, 256, 3))count = len(old_model.layers)ll = old_model.layers[count - 1].outputclasses = len(lb.classes_)ll = Dense(classes)(ll)ll = Activation("softmax", name="activation3_" + NODE)(ll)model = Model(inputs=in_puts, outputs=ll) # ERRORopt = Adam(lr=INIT_LR, decay=INIT_LR / EPOCHS)model.compile(loss="categorical_crossentropy", optimizer=opt, metrics=["accuracy"])# 训练网络H = model.fit(x_train, y_train, validation_data=(x_test, y_test), steps_per_epoch=len(x_train) // BS, epochs=EPOCHS, verbose=1)# 将模型保存到磁盘model.save("new.model")错误
ValueError: Graph disconnected: cannot obtain value for tensorKerasTensor(type_spec=TensorSpec(shape=(None, 256, 256, 3), dtype=tf.float32,name='conv2d_input'), name='conv2d_input', description="created by layer 'conv2d_input'") atlayer "conv2d". The following previous layers were accessed without issue: []回答：
这里有一个简单的方法来对你的模型进行迁移学习
classes = 10sub_old_model = Model(old_model.input, old_model.layers[-3].output)sub_old_model.trainable = Falsell = Dense(classes)(sub_old_model.output)ll = Activation("softmax")(ll)model = Model(inputs=sub_old_model.input, outputs=ll) 首先，创建一个子模型，其中包含你想冻结的老模型的层（trainable = False）。在我们的例子中，我们选取了所有层，排除了最后的 Dense 和 Softmax 激活层。
然后将子模型的输出传递给新的可训练层。
在这一点上，你只需创建一个新的模型实例来组装所有部分即可。
