Hey, i´m trying to load and transform a dataset of images with prefect tasks. Below is a code snippet, first i load images with task

load_data_task

. This provides me

images_train

and

images_val

. And now i want to transform

images_train

with task

rescaling_images_task

. I need to map over `images_train`d(for each image and label) and then apply to

recaling_images_task

. Outside of prefect this works with following code:

images_train_rescaled = images_train.map(lambda x, y: ((preprocess.rescaling_images() (x, training=True), y))

But how i can run this in prefect flow?

preprocess = PreprocessingClass_Images(cfg)

with Flow("Preprocess") as flow:

        #1. load images
        images_train, images_val = preprocess.load_data_task()

        # 2. rescaling
        #the code below works outside of prefect flow
        images_train_rescaled = images_train.map(lambda x, y: ((preprocess.rescaling_images() (x, training=True), y))

I think it would be:

with Flow("Preprocess") as flow:\
        images_train, images_val = preprocess.load_data_task()

        images_train_rescaled = rescaling_images_task.map(images_train)

but the

rescaling_images_task

must be defined to handle 1 image.

ok, so if i map over the images IN task

rescaling_images_task

and then "outside" it should work. I´ll try this. Thanks Kevin =)

You need to log it inside a task

logger = prefect.context.get("logger")
<http://logger.info|logger.info>(images_train.class_names)

because

print

is executed during build time but you want it to execute during runtime (which tasks are for)

Great 🎉 thank u very much =)