CityLearn.py

import pickle
import datasets
import numpy as np

_DESCRIPTION = """The dataset consists of tuples of (observations, actions, rewards, dones) sampled by agents
    interacting with the CityLearn 2022 Phase 1 environment (only first 5 buildings)"""

_BASE_URL = "https://huggingface.co/datasets/TobiTob/CityLearn/resolve/main"
_URLS = {
    "random_230": f"{_BASE_URL}/random_230x5x38.pkl",
    "f_230": f"{_BASE_URL}/f_230x5x38.pkl",
    "f_24": f"{_BASE_URL}/f_24x5x364.pkl",
    "fr_24": f"{_BASE_URL}/fr_24x5x364.pkl",
    "fn_24": f"{_BASE_URL}/fn_24x5x3649.pkl",
    "fn_230": f"{_BASE_URL}/fnn_230x5x380.pkl",
    "rb_24": f"{_BASE_URL}/rb_24x5x364.pkl",
    "rb_50": f"{_BASE_URL}/rb_50x5x175.pkl",
    "rb_108": f"{_BASE_URL}/rb_108x5x81.pkl",
    "rb_230": f"{_BASE_URL}/rb_230x5x38.pkl",
    "rb_461": f"{_BASE_URL}/rb_461x5x19.pkl",
    "rb_973": f"{_BASE_URL}/rb_973x5x9.pkl",
    "rb_2189": f"{_BASE_URL}/rb_2189x5x4.pkl",
    "rbn_24": f"{_BASE_URL}/rb_24x5x18247.pkl",
}


class DecisionTransformerCityLearnDataset(datasets.GeneratorBasedBuilder):
    
    # You will be able to load one configuration in the following list with
    # data = datasets.load_dataset('TobiTob/CityLearn', 'data_name')
    BUILDER_CONFIGS = [
        datasets.BuilderConfig(
            name="random_230",
            description="Random environment interactions. Sequence length = 230, Buildings = 5, Episodes = 1 ",
        ),
        datasets.BuilderConfig(
            name="f_230",
            description="Data sampled from an expert LSTM policy. Sequence length = 230, Buildings = 5, Episodes = 1 ",
        ),
        datasets.BuilderConfig(
            name="f_24",
            description="Data sampled from an expert LSTM policy. Used the old reward function. Sequence length = 24, Buildings = 5, Episodes = 1 ",
        ),
        datasets.BuilderConfig(
            name="fr_24",
            description="Data sampled from an expert LSTM policy. Used the new reward function. Sequence length = 24, Buildings = 5, Episodes = 1 ",
        ),
        datasets.BuilderConfig(
            name="fn_24",
            description="Data sampled from an expert LSTM policy, extended with noise. Sequence length = 24, Buildings = 5, Episodes = 10 ",
        ),
        datasets.BuilderConfig(
            name="fn_230",
            description="Data sampled from an expert LSTM policy, extended with noise. Sequence length = 230, Buildings = 5, Episodes = 10 ",
        ),
        datasets.BuilderConfig(
            name="rb_24",
            description="Data sampled from a simple rule based policy. Used the new reward function. Sequence length = 24, Buildings = 5, Episodes = 1 ",
        ),
        datasets.BuilderConfig(
            name="rb_50",
            description="Data sampled from a simple rule based policy. Used the new reward function. Sequence length = 50, Buildings = 5, Episodes = 1 ",
        ),
        datasets.BuilderConfig(
            name="rb_108",
            description="Data sampled from a simple rule based policy. Used the new reward function. Sequence length = 108, Buildings = 5, Episodes = 1 ",
        ),
        datasets.BuilderConfig(
            name="rb_230",
            description="Data sampled from a simple rule based policy. Used the new reward function. Sequence length = 230, Buildings = 5, Episodes = 1 ",
        ),
        datasets.BuilderConfig(
            name="rb_461",
            description="Data sampled from a simple rule based policy. Used the new reward function. Sequence length = 461, Buildings = 5, Episodes = 1 ",
        ),
        datasets.BuilderConfig(
            name="rb_973",
            description="Data sampled from a simple rule based policy. Used the new reward function. Sequence length = 973, Buildings = 5, Episodes = 1 ",
        ),
        datasets.BuilderConfig(
            name="rb_2189",
            description="Data sampled from a simple rule based policy. Used the new reward function. Sequence length = 2189, Buildings = 5, Episodes = 1 ",
        ),
        datasets.BuilderConfig(
            name="rbn_24",
            description="Data sampled from a simple rule based policy. Used the new reward function and changed some interactions with noise. Sequence length = 24, Buildings = 5, Episodes = 50 ",
        ),
    ]

    def _info(self):

        features = datasets.Features(
            {
                "observations": datasets.Sequence(datasets.Sequence(datasets.Value("float32"))),
                "actions": datasets.Sequence(datasets.Sequence(datasets.Value("float32"))),
                "rewards": datasets.Sequence(datasets.Value("float32")),
                "dones": datasets.Sequence(datasets.Value("bool")),
            }
        )

        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features,
        )

    def _split_generators(self, dl_manager):
        urls = _URLS[self.config.name]
        data_dir = dl_manager.download_and_extract(urls)
        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):
        with open(filepath, "rb") as f:
            trajectories = pickle.load(f)

            for idx, traj in enumerate(trajectories):
                yield idx, {
                    "observations": traj["observations"],
                    "actions": traj["actions"],
                    "rewards": np.expand_dims(traj["rewards"], axis=1),
                    "dones": np.expand_dims(traj.get("dones", traj.get("terminals")), axis=1),
                }