How to calculate macro, weighted Precision Recall values in tensorflow?

Question

I am trying to get macro and weighted precision, recall values while training ResNet/ResNeXt models. How I compile my model:

f1 =tfa.metrics.F1Score(num_classes=2, average='weighted')
model.compile (
    loss = 'categorical_crossentropy',
    optimizer="adam" ,metrics=[ f1, "Precision", "Recall"]
)

When I fit the model after compiling like that my precision and recall scores are always same in each step. In online I saw that that might be because I am calculating micro scores. But, in the documentation of tf.metrics there is no 'average' parameter for recall or precision. What can I do or why do I get same values?

I tried to use precision and recall function from torchmetrics library.

f1 =tfa.metrics.F1Score(num_classes=2, average='weighted')
binaryloss = tf.keras.losses.BinaryCrossentropy()
prec = torchmetrics.Precision(num_classes=2, threshold=0.5, average='weighted', task='binary')
recall = torchmetrics.Recall(task="binary", average='weighted', num_classes=2)
model.compile (
    loss = 'categorical_crossentropy',
    optimizer="adam" ,metrics=[ f1, prec, recall]
)

But when I compile the model like that, I get error after fitting part. The error is the following: ** RuntimeError: Predictions and targets are expected to have the same shape, but got (None, None) and (None, 2).**