dataset.py

# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# TODO: Address all TODOs and remove all explanatory comments
"""TODO: Add a description here."""


import zipfile
import os
import datasets
from PIL import Image
from io import BytesIO

# TODO: Name of the dataset usually match the script name with CamelCase instead of snake_case
class sdbias(datasets.GeneratorBasedBuilder):
    """TODO: Short description of my dataset."""

    VERSION = datasets.Version("1.1.0")

    # This is an example of a dataset with multiple configurations.
    # If you don't want/need to define several sub-sets in your dataset,
    # just remove the BUILDER_CONFIG_CLASS and the BUILDER_CONFIGS attributes.

    # If you need to make complex sub-parts in the datasets with configurable options
    # You can create your own builder configuration class to store attribute, inheriting from datasets.BuilderConfig
    # BUILDER_CONFIG_CLASS = MyBuilderConfig

    # You will be able to load one or the other configurations in the following list with
    # data = datasets.load_dataset('my_dataset', 'first_domain')
    # data = datasets.load_dataset('my_dataset', 'second_domain')
    BUILDER_CONFIGS = [
        datasets.BuilderConfig(name="first_domain", version=VERSION, description="This part of my dataset covers a first domain"),
    ]

    DEFAULT_CONFIG_NAME = "first_domain"  # It's not mandatory to have a default configuration. Just use one if it make sense.

    def _info(self):
        if self.config.name == "first_domain":  # This is the name of the configuration selected in BUILDER_CONFIGS above
            features = datasets.Features(
                {
                    "adjective": datasets.Value("string"),
                    "profession": datasets.Value("string"),
                    "seed": datasets.Value("int32"),
                    "image": datasets.Image()
                    # These are the features of your dataset like images, labels ...
                }
            )
        return datasets.DatasetInfo(
            # This is the description that will appear on the datasets page.
            description="bla",
            # This defines the different columns of the dataset and their types
            features=features,  # Here we define them above because they are different between the two configurations
            # If there's a common (input, target) tuple from the features, uncomment supervised_keys line below and
            # specify them. They'll be used if as_supervised=True in builder.as_dataset.
            # supervised_keys=("sentence", "label"),
            # Homepage of the dataset for documentation
            homepage="bla",
            # License for the dataset if available
            license="bla",
            # Citation for the dataset
            citation="bli",
        )

    def _split_generators(self, dl_manager):
        # TODO: This method is tasked with downloading/extracting the data and defining the splits depending on the configuration
        # If several configurations are possible (listed in BUILDER_CONFIGS), the configuration selected by the user is in self.config.name

        # dl_manager is a datasets.download.DownloadManager that can be used to download and extract URLS
        # It can accept any type or nested list/dict and will give back the same structure with the url replaced with path to local files.
        # By default the archives will be extracted and a path to a cached folder where they are extracted is returned instead of the archive
        data_dir = "/mnt/1da05489-3812-4f15-a6e5-c8d3c57df39e/StableDiffusionBiasExplorer/zipped_images"
        return [
            datasets.SplitGenerator(
                name=datasets.Split.TRAIN,
                # These kwargs will be passed to _generate_examples
                gen_kwargs={
                    "filepath":data_dir,
                    "split": "train",
                },
            ),
        ]

    # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
    def _generate_examples(self, filepath, split):
        zip_files = os.listdir(filepath)
        key = 0
        for zip_file in zip_files:
            with zipfile.ZipFile(filepath + "/" + zip_file, "r") as zf:
                for f in zf.filelist:
                    if ".jpg" in f.filename:
                        jpg_content = BytesIO(zf.read(f))
                        with Image.open(jpg_content) as image:
                            yield key, {
                                "adjective": zip_file.split("_", 1)[0],
                                "profession": zip_file.split("_", 1)[-1].replace(".zip",""),
                                "seed": int(f.filename.split("Seed_")[-1].split("/")[0]),
                                "image": image,
                            }
                        key+=1