saving train checkpoint

[LorenzoF6]

Hi if i want to train a fastflow model and i want to insert a checkpoint that every x time save a different file of the weights ho can i do it ? thanks

[abc-125]

Hello, you can use ModelCheckpoint; for Fastflow it can look like this if you want to save weights every epoch:

from anomalib.data import MVTec
from anomalib.models import Fastflow
from anomalib.engine import Engine
from anomalib.callbacks.checkpoint import ModelCheckpoint


if __name__ == '__main__':
    # Checkpoint
    checkpoint_callback = ModelCheckpoint(
        filename="{epoch}",
        save_top_k=-1,
    )

    # Initialize components
    datamodule = MVTec()
    model = Fastflow()
    engine = Engine(check_val_every_n_epoch=1, callbacks=[checkpoint_callback])

    # Train the model
    engine.fit(datamodule=datamodule, model=model)