Example usage for erase_diff algorithm (CompVis model)

Note: Currently we only have support for unlearn canvas dataset. I2p and generic dataset support needs to be added.

The default configuration for training is provided by erase_diff_train_mu. You can run the training with the default settings as follows:

Using the Default Configuration

from mu.algorithms.erase_diff.algorithm import EraseDiffAlgorithm
from mu.algorithms.erase_diff.configs import erase_diff_train_mu

algorithm = EraseDiffAlgorithm(
    erase_diff_train_mu
)
algorithm.run()

Overriding the Default Configuration

If you need to override the existing configuration settings, you can specify your custom parameters (such as ckpt_path and raw_dataset_dir) directly when initializing the algorithm. For example:

Machine unlearning using unlearn canvas dataset:

from mu.algorithms.erase_diff.algorithm import EraseDiffAlgorithm
from mu.algorithms.erase_diff.configs import erase_diff_train_mu

algorithm = EraseDiffAlgorithm(
    erase_diff_train_mu,
    ckpt_path="/home/ubuntu/Projects/UnlearnCanvas/UnlearnCanvas/machine_unlearning/models/compvis/style50/compvis.ckpt", #replace it with your ckpt path
    raw_dataset_dir="data/quick-canvas-dataset/sample",
    use_sample = True, #uses sample dataset
    template_name = "Abstractionism",
    templete="class",
    dataset_type = "unlearncanvas",
    devices = "0"
)
algorithm.run()

Note: If you want to use a sample dataset for testing purposes, set use_sample=True (default).Otherwise, set use_sample=False to use the full dataset.

Note You can choose from a set of predefined template_name options to erase specific concept when working with the unlearncanvas dataset to perform unlearning. For instance, in the unlearncanvas dataset context, the available choices include:

"Abstractionism", "Artist_Sketch", "Blossom_Season", "Bricks", "Byzantine", "Cartoon",
 "Cold_Warm", "Color_Fantasy", "Comic_Etch", "Crayon", "Cubism", "Dadaism", "Dapple",
 "Defoliation", "Early_Autumn", "Expressionism", "Fauvism", "French", "Glowing_Sunset",
 "Gorgeous_Love", "Greenfield", "Impressionism", "Ink_Art", "Joy", "Liquid_Dreams",
 "Magic_Cube", "Meta_Physics", "Meteor_Shower", "Monet", "Mosaic", "Neon_Lines", "On_Fire",
 "Pastel", "Pencil_Drawing", "Picasso", "Pop_Art", "Red_Blue_Ink", "Rust", "Seed_Images",
 "Sketch", "Sponge_Dabbed", "Structuralism", "Superstring", "Surrealism", "Ukiyoe",
 "Van_Gogh", "Vibrant_Flow", "Warm_Love", "Warm_Smear", "Watercolor", "Winter"

For example if you want to train using Abstractionism then you can pass template_name = "Abstractionism".

Also you can choose type of template to use during training. The available choices are style and class.

Add your own unlearning algorithms:

For detailed instructions on adding your own algorithm, please see the Contribution section.