我想用知识蒸馏keras模型预测不平衡数据，y标签值计数是这样的，

y_train.value_counts()
0    9024
1     842
Name: Y_LABEL, dtype: int64`

为了预测不平衡数据，我尝试使用class_weight，但我不知道如何在自定义模型中应用class_weight。如何在知识蒸馏自定义模型中应用class_weight？
我使用keras平台构建知识蒸馏自定义模型，首先，如您所见，y标签不平衡，因此我想使用class_weight

class Distiller(keras.Model):
    def __init__(self, student, teacher):
        super(Distiller, self).__init__()
        self.teacher = teacher
        self.student = student

    def compile(
        self,
        optimizer,
        metrics,
        student_loss_fn,
        distillation_loss_fn,
        alpha=0.1,
        temperature=3,
    ):
        """ Configure the distiller.
        Args:
            optimizer: Keras optimizer for the student weights
            metrics: Keras metrics for evaluation
            student_loss_fn: Loss function of difference between student
                predictions and ground-truth
            distillation_loss_fn: Loss function of difference between soft
                student predictions and soft teacher predictions
            alpha: weight to student_loss_fn and 1-alpha to distillation_loss_fn
            temperature: Temperature for softening probability distributions.
                Larger temperature gives softer distributions.
        """
        super(Distiller, self).compile(optimizer=optimizer, metrics=metrics)
        self.student_loss_fn = student_loss_fn
        self.distillation_loss_fn = distillation_loss_fn
        self.alpha = alpha
        self.temperature = temperature

    def train_step(self, data):
        # Unpack data
        x, y = data

        # Forward pass of teacher
        teacher_predictions = self.teacher(x[:,:52], training=False)

        with tf.GradientTape() as tape:
            # Forward pass of student
            student_predictions = self.student(x[:,52:], training=True)

            # Compute losses
            student_loss = self.student_loss_fn(y, student_predictions)
            distillation_loss = self.distillation_loss_fn(
                tf.nn.softmax(teacher_predictions / self.temperature, axis=1),
                tf.nn.softmax(student_predictions / self.temperature, axis=1),
            )
            loss = self.alpha * student_loss + (1 - self.alpha) * distillation_loss

        # Compute gradients
        trainable_vars = self.student.trainable_variables
        gradients = tape.gradient(loss, trainable_vars)

        # Update weights
        self.optimizer.apply_gradients(zip(gradients, trainable_vars))

        # Update the metrics
        self.compiled_metrics.update_state(y, student_predictions, sample_weight = sample_weight)

        # Return a dict of performance
        results = {m.name: m.result() for m in self.metrics}
        results.update(
            {"student_loss": student_loss, "distillation_loss": distillation_loss}
        )
        return results

    def test_step(self, data):
        # Unpack the data
        x, y = data

        # Compute predictions
        y_prediction = self.student(x, training=False) # validation_data
        # y_prediction = self.student(x[:,52:], training=False) # validation_split

        # Calculate the loss
        student_loss = self.student_loss_fn(y, y_prediction)

        # Update the metrics.
        self.compiled_metrics.update_state(y, y_prediction)

        # Return a dict of performance
        results = {m.name: m.result() for m in self.metrics}
        results.update({"student_loss": student_loss})
        return results

这是类权重

class_weight
{0: 0.5466568414520633, 1: 5.858270989193683}

我编译和拟合模型就像这样

distiller.compile(
        student_loss_fn= tf.keras.losses.BinaryCrossentropy(from_logits = True),
        optimizer=keras.optimizers.Adam(learning_rate = 0.001),
        metrics=[tf.keras.metrics.BinaryAccuracy()], # , f1
        distillation_loss_fn=keras.losses.KLDivergence(),
        alpha=0.1,
        temperature=10)

distillarHistory = distiller.fit(
                             df_out, 
                             y_train, 
                             epochs=50,
                             batch_size = 256,
                             validation_data = (X_test_s, y_test),
                             verbose = 0,
                             class_weight = class_weight,
                             )

但是发生了这样的错误，我如何在train_step中应用class_weight？？

TypeError: in user code:

File "/home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages/keras/engine/training.py", line 1160, in train_function  *
    return step_function(self, iterator)
File "/home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages/keras/engine/training.py", line 1146, in step_function  **
    outputs = model.distribute_strategy.run(run_step, args=(data,))
File "/home/studio-lab-user/.conda/envs/default/lib/python3.9/site-packages/keras/engine/training.py", line 1135, in run_step  **
    outputs = model.train_step(data)
File "/tmp/ipykernel_22361/144175027.py", line 61, in train_step
    self.compiled_metrics.update_state(y, student_predictions, class_weight = class_weight)

TypeError: update_state() got an unexpected keyword argument 'class_weight'

我尝试通过修改自定义模型来使用sample_weight，如下所示：

def train_step(self, data):
        # Unpack data
        x, y, sample_weight = data

        # Forward pass of teacher
        teacher_predictions = self.teacher(x[:,:52], training=False)

        with tf.GradientTape() as tape:
            # Forward pass of student
            student_predictions = self.student(x[:,52:], training=True)

            # Compute losses
            student_loss = self.student_loss_fn(y, student_predictions)
            distillation_loss = self.distillation_loss_fn(
                tf.nn.softmax(teacher_predictions / self.temperature, axis=1),
                tf.nn.softmax(student_predictions / self.temperature, axis=1),
            )
            loss = self.alpha * student_loss + (1 - self.alpha) * distillation_loss

        # Compute gradients
        trainable_vars = self.student.trainable_variables
        gradients = tape.gradient(loss, trainable_vars)

        # Update weights
        self.optimizer.apply_gradients(zip(gradients, trainable_vars))

        # Update the metrics
        self.compiled_metrics.update_state(y, student_predictions, sample_weight = sample_weight)
        # self.compiled_metrics.update_state(y, student_predictions, sample_weight = sample_weight)

        # Return a dict of performance
        results = {m.name: m.result() for m in self.metrics}
        results.update(
            {"student_loss": student_loss, "distillation_loss": distillation_loss}
        )
        return results

distillarHistory = distiller.fit(
                                 df_out, 
                                 y_train, 
                                 epochs=50,
                                 batch_size = 256,
                                 validation_data = (X_test_s, y_test),
                                 verbose = 0,
                                 sample_weight = sample_weight,
                                 )

但是输出与不使用sample_weight而使用分层的5k倍完全相同，

• 不使用sample_weight

Alpha - 0.10 / Temperature 10 
62/62 [==============================] - 0s 2ms/step
KFold F1 scores :  0.561608878542233
62/62 [==============================] - 0s 2ms/step
KFold F1 scores :  0.5664106062792742
62/62 [==============================] - 0s 2ms/step
KFold F1 scores :  0.5908350815131695
62/62 [==============================] - 0s 2ms/step
KFold F1 scores :  0.5793267313367816
62/62 [==============================] - 0s 2ms/step
KFold F1 scores :  0.5918020295603292

--------------------------------------------------------
Alpha=0.1, Temperature=10, F-Score=0.578
--------------------------------------------------------

• 使用样品重量

Alpha - 0.10 / Temperature 10 
62/62 [==============================] - 1s 5ms/step
KFold F1 scores :  0.561608878542233
62/62 [==============================] - 0s 2ms/step
KFold F1 scores :  0.5664106062792742
62/62 [==============================] - 1s 8ms/step
KFold F1 scores :  0.5908350815131695
62/62 [==============================] - 1s 8ms/step
KFold F1 scores :  0.5793267313367816
62/62 [==============================] - 1s 7ms/step
KFold F1 scores :  0.5918020295603292

--------------------------------------------------------
Alpha=0.1, Temperature=10, F-Score=0.578
--------------------------------------------------------

“我想知道对积极数据的采样是消极数据的10倍左右是否相同