ETT time series dataset (#4213)

* initial ETT dataset

* make univariate with covariates

* added factor for 15T dataset

* comments

* added readme

* removed unused imports

* added dummy dataset

* added dataset_info.json

* black

* remove whitespace

* fix readme

* added kashif

* fixed typos in readme

* fixed dataset name

Commit from https://github.com/huggingface/datasets/commit/15c98031c333eb50cb168da8ec88e2c59e2ba7fe

README.md

@@ -0,0 +1,214 @@
+---
+annotations_creators:
+- no-annotation
+language_creators:
+- found
+languages:
+- unknown
+licenses:
+- cc-by-4-0
+multilinguality:
+- monolingual
+pretty_name: Electricity Transformer Temperature
+size_categories:
+- 1K<n<10K
+source_datasets:
+- original
+task_categories:
+- time-series-forecasting
+task_ids:
+- univariate-time-series-forecasting
+- multivariate-time-series-forecasting
+---
+
+# Dataset Card for [Electricity Transformer Temperature](https://github.com/zhouhaoyi/ETDataset)
+
+## Table of Contents
+- [Table of Contents](#table-of-contents)
+- [Dataset Description](#dataset-description)
+  - [Dataset Summary](#dataset-summary)
+  - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
+  - [Languages](#languages)
+- [Dataset Structure](#dataset-structure)
+  - [Data Instances](#data-instances)
+  - [Data Fields](#data-fields)
+  - [Data Splits](#data-splits)
+- [Dataset Creation](#dataset-creation)
+  - [Curation Rationale](#curation-rationale)
+  - [Source Data](#source-data)
+  - [Annotations](#annotations)
+  - [Personal and Sensitive Information](#personal-and-sensitive-information)
+- [Considerations for Using the Data](#considerations-for-using-the-data)
+  - [Social Impact of Dataset](#social-impact-of-dataset)
+  - [Discussion of Biases](#discussion-of-biases)
+  - [Other Known Limitations](#other-known-limitations)
+- [Additional Information](#additional-information)
+  - [Dataset Curators](#dataset-curators)
+  - [Licensing Information](#licensing-information)
+  - [Citation Information](#citation-information)
+  - [Contributions](#contributions)
+
+## Dataset Description
+
+- **Homepage:** [Electricity Transformer Dataset](https://github.com/zhouhaoyi/ETDataset)
+- **Repository:** https://github.com/zhouhaoyi/ETDataset
+- **Paper:** [Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting](https://arxiv.org/abs/2012.07436)
+- **Point of Contact:** [Haoyi Zhou](mailto:zhouhy@act.buaa.edu.cn)
+
+### Dataset Summary
+
+The electric power distribution problem is the distribution of electricity to different areas depending on its sequential usage. But predicting the future demand of a specific area is difficult, as it varies with weekdays, holidays, seasons, weather, temperatures, etc. However, no existing method can perform a long-term prediction based on super long-term real-world data with high precision. Any false predictions may damage the electrical transformer. So currently, without an efficient method to predict future electric usage, managers have to make decisions based on the empirical number, which is much higher than the real-world demands. It causes unnecessary waste of electric and equipment depreciation. On the other hand, the oil temperatures can reflect the condition of the  Transformer. One of the most efficient strategies is to predict how the electrical transformers' oil temperature is safe and avoid unnecessary waste. As a result, to address this problem, the authors and Beijing Guowang Fuda Science & Technology Development Company have provided 2-years worth of data.
+
+Specifically, the dataset combines short-term periodical patterns, long-term periodical patterns, long-term trends, and many irregular patterns. The dataset are obtained from  2 Electricity Transformers at 2 stations  and come in an `1H` (hourly) or `15T` (15-minute) frequency containing 2 year * 365 days * 24 hours * (4 for 15T) times = 17,520 (70,080 for 15T) data points.
+
+The target time series is the  **O**il **T**emperature and the dataset comes with the following 6 covariates in the univariate setup:
+* **H**igh **U**se**F**ul **L**oad 
+* **H**igh **U**se**L**ess **L**oad
+* **M**iddle **U**se**F**ul **L**oad 
+* **M**iddle **U**se**L**ess **L**oad 
+* **L**ow **U**se**F**ul **L**oad 
+* **L**ow **U**se**L**ess **L**oad
+
+
+### Dataset Usage
+
+To load a particular variant of the dataset just specify its name e.g:
+
+```python
+load_dataset("ett", "m1", multivariate=False) # univariate 15-min frequency dataset from first transformer
+```
+
+or to specify a prediction length:
+
+```python
+load_dataset("ett", "h2", prediction_length=48) # multivariate dataset from second transformer with prediction length of 48 (hours)
+```
+
+
+### Supported Tasks and Leaderboards
+
+The time series data is split into  train/val/test  set of 12/4/4 months respectively.  Given the prediction length (default: 1 day (24 hours or 24*4 15T)) we create rolling windows of this size for the val/test sets. 
+
+#### `time-series-forecasting`
+
+##### `univariate-time-series-forecasting`
+
+The univariate time series forecasting tasks involves learning the future one dimensional `target` values of a time series in a dataset for some `prediction_length` time steps. The performance of the forecast models can then be validated via the ground truth in the `validation` split and tested via the `test` split. The covriates are stored in the `feat_dynamic_real` key of each time series.
+
+##### `multivariate-time-series-forecasting`
+
+The multivariate time series forecasting task involves learning the future vector of `target` values of a time series in a dataset for some `prediction_length` time steps. Similar to the univariate setting the performance of a multivariate model can be validated via the ground truth in the `validation` split and tested via the `test` split.
+
+
+### Languages
+
+## Dataset Structure
+
+### Data Instances
+
+A sample from the training set is provided below:
+
+```python
+{
+  'start': datetime.datetime(2012, 1, 1, 0, 0),
+  'target': [14.0, 18.0, 21.0, 20.0, 22.0, 20.0, ...],
+  'feat_static_cat': [0], 
+  'feat_dynamic_real': [[0.3, 0.4], [0.1, 0.6], ...],
+  'item_id': 'OT'
+}
+```
+
+### Data Fields
+
+For the univariate regular time series each series has the following keys:
+
+* `start`: a datetime of the first entry of each time series in the dataset
+* `target`: an array[float32] of the actual target values
+* `feat_static_cat`: an array[uint64] which contains a categorical identifier of each time series in the dataset
+* `feat_dynamic_real`: optional array of covariate features
+* `item_id`: a string identifier of each time series in a dataset for reference
+
+For the multivariate time series the `target` is a vector of the multivariate dimension for each time point.
+
+### Data Splits
+
+The time series data is split into  train/val/test  set of 12/4/4 months respectively.
+
+## Dataset Creation
+
+### Curation Rationale
+
+Develop time series methods  that can perform a long-term prediction based on super long-term real-world data with high precision.
+
+### Source Data
+
+#### Initial Data Collection and Normalization
+
+[More Information Needed]
+
+#### Who are the source language producers?
+
+[More Information Needed]
+
+### Annotations
+
+#### Annotation process
+
+[More Information Needed]
+
+#### Who are the annotators?
+
+[More Information Needed]
+
+### Personal and Sensitive Information
+
+[More Information Needed]
+
+## Considerations for Using the Data
+
+### Social Impact of Dataset
+
+[More Information Needed]
+
+### Discussion of Biases
+
+[More Information Needed]
+
+### Other Known Limitations
+
+[More Information Needed]
+
+## Additional Information
+
+### Dataset Curators
+
+* [Haoyi Zhou](mailto:zhouhy@act.buaa.edu.cn)
+
+### Licensing Information
+
+[Creative Commons Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/legalcode)
+
+### Citation Information
+
+```tex
+@inproceedings{haoyietal-informer-2021,
+  author    = {Haoyi Zhou and
+               Shanghang Zhang and
+               Jieqi Peng and
+               Shuai Zhang and
+               Jianxin Li and
+               Hui Xiong and
+               Wancai Zhang},
+  title     = {Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting},
+  booktitle = {The Thirty-Fifth {AAAI} Conference on Artificial Intelligence, {AAAI} 2021, Virtual Conference},
+  volume    = {35},
+  number    = {12},
+  pages     = {11106--11115},
+  publisher = {{AAAI} Press},
+  year      = {2021},
+}
+```
+
+### Contributions
+
+Thanks to [@kashif](https://github.com/kashif) for adding this dataset.

dataset_infos.json

@@ -0,0 +1 @@
+{"h1": {"description": "The data of Electricity Transformers from two separated counties\nin China collected for two years at hourly and 15-min frequencies.\nEach data point consists of the target value \"oil temperature\" and\n6 power load features. The train/val/test is 12/4/4 months.\n", "citation": "@inproceedings{haoyietal-informer-2021,\n  author    = {Haoyi Zhou and\n               Shanghang Zhang and\n               Jieqi Peng and\n               Shuai Zhang and\n               Jianxin Li and\n               Hui Xiong and\n               Wancai Zhang},\n  title     = {Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting},\n  booktitle = {The Thirty-Fifth {AAAI} Conference on Artificial Intelligence, {AAAI} 2021, Virtual Conference},\n  volume    = {35},\n  number    = {12},\n  pages     = {11106--11115},\n  publisher = {{AAAI} Press},\n  year      = {2021},\n}\n", "homepage": "https://github.com/zhouhaoyi/ETDataset", "license": "The Creative Commons Attribution 4.0 International License. https://creativecommons.org/licenses/by/4.0/", "features": {"start": {"dtype": "timestamp[s]", "id": null, "_type": "Value"}, "target": {"feature": {"dtype": "float32", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "feat_static_cat": {"feature": {"dtype": "uint64", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "feat_dynamic_real": {"feature": {"feature": {"dtype": "float32", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "length": -1, "id": null, "_type": "Sequence"}, "item_id": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "ett", "config_name": "h1", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 249540, "num_examples": 1, "dataset_name": "ett"}, "test": {"name": "test", "num_bytes": 79930740, "num_examples": 240, "dataset_name": "ett"}, "validation": {"name": "validation", "num_bytes": 34975770, "num_examples": 120, "dataset_name": "ett"}}, "download_checksums": {"https://raw.githubusercontent.com/zhouhaoyi/ETDataset/main/ETT-small/ETTh1.csv": {"num_bytes": 2589657, "checksum": "f18de3ad269cef59bb07b5438d79bb3042d3be49bdeecf01c1cd6d29695ee066"}}, "download_size": 2589657, "post_processing_size": null, "dataset_size": 115156050, "size_in_bytes": 117745707}, "h2": {"description": "The data of Electricity Transformers from two separated counties\nin China collected for two years at hourly and 15-min frequencies.\nEach data point consists of the target value \"oil temperature\" and\n6 power load features. The train/val/test is 12/4/4 months.\n", "citation": "@inproceedings{haoyietal-informer-2021,\n  author    = {Haoyi Zhou and\n               Shanghang Zhang and\n               Jieqi Peng and\n               Shuai Zhang and\n               Jianxin Li and\n               Hui Xiong and\n               Wancai Zhang},\n  title     = {Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting},\n  booktitle = {The Thirty-Fifth {AAAI} Conference on Artificial Intelligence, {AAAI} 2021, Virtual Conference},\n  volume    = {35},\n  number    = {12},\n  pages     = {11106--11115},\n  publisher = {{AAAI} Press},\n  year      = {2021},\n}\n", "homepage": "https://github.com/zhouhaoyi/ETDataset", "license": "The Creative Commons Attribution 4.0 International License. https://creativecommons.org/licenses/by/4.0/", "features": {"start": {"dtype": "timestamp[s]", "id": null, "_type": "Value"}, "target": {"feature": {"dtype": "float32", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "feat_static_cat": {"feature": {"dtype": "uint64", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "feat_dynamic_real": {"feature": {"feature": {"dtype": "float32", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "length": -1, "id": null, "_type": "Sequence"}, "item_id": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "ett", "config_name": "h2", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 249540, "num_examples": 1, "dataset_name": "ett"}, "test": {"name": "test", "num_bytes": 79930740, "num_examples": 240, "dataset_name": "ett"}, "validation": {"name": "validation", "num_bytes": 34975770, "num_examples": 120, "dataset_name": "ett"}}, "download_checksums": {"https://raw.githubusercontent.com/zhouhaoyi/ETDataset/main/ETT-small/ETTh2.csv": {"num_bytes": 2417960, "checksum": "a3dc2c597b9218c7ce1cd55eb77b283fd459a1d09d753063f944967dd6b9218b"}}, "download_size": 2417960, "post_processing_size": null, "dataset_size": 115156050, "size_in_bytes": 117574010}, "m1": {"description": "The data of Electricity Transformers from two separated counties\nin China collected for two years at hourly and 15-min frequencies.\nEach data point consists of the target value \"oil temperature\" and\n6 power load features. The train/val/test is 12/4/4 months.\n", "citation": "@inproceedings{haoyietal-informer-2021,\n  author    = {Haoyi Zhou and\n               Shanghang Zhang and\n               Jieqi Peng and\n               Shuai Zhang and\n               Jianxin Li and\n               Hui Xiong and\n               Wancai Zhang},\n  title     = {Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting},\n  booktitle = {The Thirty-Fifth {AAAI} Conference on Artificial Intelligence, {AAAI} 2021, Virtual Conference},\n  volume    = {35},\n  number    = {12},\n  pages     = {11106--11115},\n  publisher = {{AAAI} Press},\n  year      = {2021},\n}\n", "homepage": "https://github.com/zhouhaoyi/ETDataset", "license": "The Creative Commons Attribution 4.0 International License. https://creativecommons.org/licenses/by/4.0/", "features": {"start": {"dtype": "timestamp[s]", "id": null, "_type": "Value"}, "target": {"feature": {"dtype": "float32", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "feat_static_cat": {"feature": {"dtype": "uint64", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "feat_dynamic_real": {"feature": {"feature": {"dtype": "float32", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "length": -1, "id": null, "_type": "Sequence"}, "item_id": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "ett", "config_name": "m1", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 997980, "num_examples": 1, "dataset_name": "ett"}, "test": {"name": "test", "num_bytes": 0, "num_examples": 0, "dataset_name": "ett"}, "validation": {"name": "validation", "num_bytes": 0, "num_examples": 0, "dataset_name": "ett"}}, "download_checksums": {"https://raw.githubusercontent.com/zhouhaoyi/ETDataset/main/ETT-small/ETTm1.csv": {"num_bytes": 10360719, "checksum": "6ce1759b1a18e3328421d5d75fadcb316c449fcd7cec32820c8dafda71986c9e"}}, "download_size": 10360719, "post_processing_size": null, "dataset_size": 997980, "size_in_bytes": 11358699}, "m2": {"description": "The data of Electricity Transformers from two separated counties\nin China collected for two years at hourly and 15-min frequencies.\nEach data point consists of the target value \"oil temperature\" and\n6 power load features. The train/val/test is 12/4/4 months.\n", "citation": "@inproceedings{haoyietal-informer-2021,\n  author    = {Haoyi Zhou and\n               Shanghang Zhang and\n               Jieqi Peng and\n               Shuai Zhang and\n               Jianxin Li and\n               Hui Xiong and\n               Wancai Zhang},\n  title     = {Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting},\n  booktitle = {The Thirty-Fifth {AAAI} Conference on Artificial Intelligence, {AAAI} 2021, Virtual Conference},\n  volume    = {35},\n  number    = {12},\n  pages     = {11106--11115},\n  publisher = {{AAAI} Press},\n  year      = {2021},\n}\n", "homepage": "https://github.com/zhouhaoyi/ETDataset", "license": "The Creative Commons Attribution 4.0 International License. https://creativecommons.org/licenses/by/4.0/", "features": {"start": {"dtype": "timestamp[s]", "id": null, "_type": "Value"}, "target": {"feature": {"dtype": "float32", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "feat_static_cat": {"feature": {"dtype": "uint64", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "feat_dynamic_real": {"feature": {"feature": {"dtype": "float32", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "length": -1, "id": null, "_type": "Sequence"}, "item_id": {"dtype": "string", "id": null, "_type": "Value"}}, "post_processed": null, "supervised_keys": null, "task_templates": null, "builder_name": "ett", "config_name": "m2", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 997980, "num_examples": 1, "dataset_name": "ett"}, "test": {"name": "test", "num_bytes": 0, "num_examples": 0, "dataset_name": "ett"}, "validation": {"name": "validation", "num_bytes": 0, "num_examples": 0, "dataset_name": "ett"}}, "download_checksums": {"https://raw.githubusercontent.com/zhouhaoyi/ETDataset/main/ETT-small/ETTm2.csv": {"num_bytes": 9677236, "checksum": "db973ca252c6410a30d0469b13d696cf919648d0f3fd588c60f03fdbdbadd1fd"}}, "download_size": 9677236, "post_processing_size": null, "dataset_size": 997980, "size_in_bytes": 10675216}}

dummy/h1/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff7007e9aa6ef9d12714df177250528f3899730166a7441f47342c4b0aa6d18e
+size 535

dummy/h2/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6e27267dc2d291c54e45d03b7f00a16672eb89771e4c462d321bf175976c3be
+size 549

dummy/m1/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02a6a92a71f0d19d9ec4d5f96f1a681a4a849ab1f87edbd53747e7dc6358726d
+size 499

dummy/m2/1.0.0/dummy_data.zip

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a388f1645cf5d38e3625295ecb0f6a31f9d66ce1a1f028a258ac54358306f4d0
+size 551

ett.py

@@ -0,0 +1,242 @@
+# 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.
+"""Electricity Transformer Temperature (ETT) dataset."""
+from dataclasses import dataclass
+
+import pandas as pd
+
+import datasets
+
+
+_CITATION = """\
+@inproceedings{haoyietal-informer-2021,
+  author    = {Haoyi Zhou and
+               Shanghang Zhang and
+               Jieqi Peng and
+               Shuai Zhang and
+               Jianxin Li and
+               Hui Xiong and
+               Wancai Zhang},
+  title     = {Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting},
+  booktitle = {The Thirty-Fifth {AAAI} Conference on Artificial Intelligence, {AAAI} 2021, Virtual Conference},
+  volume    = {35},
+  number    = {12},
+  pages     = {11106--11115},
+  publisher = {{AAAI} Press},
+  year      = {2021},
+}
+"""
+
+_DESCRIPTION = """\
+The data of Electricity Transformers from two separated counties
+in China collected for two years at hourly and 15-min frequencies.
+Each data point consists of the target value "oil temperature" and
+6 power load features. The train/val/test is 12/4/4 months.
+"""
+
+_HOMEPAGE = "https://github.com/zhouhaoyi/ETDataset"
+
+_LICENSE = "The Creative Commons Attribution 4.0 International License. https://creativecommons.org/licenses/by/4.0/"
+
+# The HuggingFace Datasets library doesn't host the datasets but only points to the original files.
+# This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
+_URLS = {
+    "h1": "https://raw.githubusercontent.com/zhouhaoyi/ETDataset/main/ETT-small/ETTh1.csv",
+    "h2": "https://raw.githubusercontent.com/zhouhaoyi/ETDataset/main/ETT-small/ETTh2.csv",
+    "m1": "https://raw.githubusercontent.com/zhouhaoyi/ETDataset/main/ETT-small/ETTm1.csv",
+    "m2": "https://raw.githubusercontent.com/zhouhaoyi/ETDataset/main/ETT-small/ETTm2.csv",
+}
+
+
+@dataclass
+class ETTBuilderConfig(datasets.BuilderConfig):
+    """ETT builder config."""
+
+    prediction_length: int = 24
+    multivariate: bool = False
+
+
+class ETT(datasets.GeneratorBasedBuilder):
+    """Electricity Transformer Temperature (ETT) dataset"""
+
+    VERSION = datasets.Version("1.0.0")
+
+    # You will be able to load one or the other configurations in the following list with
+    # data = datasets.load_dataset('ett', 'h1')
+    # data = datasets.load_dataset('ett', 'm2')
+    BUILDER_CONFIGS = [
+        ETTBuilderConfig(
+            name="h1",
+            version=VERSION,
+            description="Time series from first county at hourly frequency.",
+        ),
+        ETTBuilderConfig(
+            name="h2",
+            version=VERSION,
+            description="Time series from second county at hourly frequency.",
+        ),
+        ETTBuilderConfig(
+            name="m1",
+            version=VERSION,
+            description="Time series from first county at 15-min frequency.",
+        ),
+        ETTBuilderConfig(
+            name="m2",
+            version=VERSION,
+            description="Time series from second county at 15-min frequency.",
+        ),
+    ]
+
+    DEFAULT_CONFIG_NAME = "h1"  # It's not mandatory to have a default configuration. Just use one if it make sense.
+
+    def _info(self):
+        if self.config.multivariate:
+            features = datasets.Features(
+                {
+                    "start": datasets.Value("timestamp[s]"),
+                    "target": datasets.Sequence(datasets.Sequence(datasets.Value("float32"))),
+                    "feat_static_cat": datasets.Sequence(datasets.Value("uint64")),
+                    "item_id": datasets.Value("string"),
+                }
+            )
+        else:
+            features = datasets.Features(
+                {
+                    "start": datasets.Value("timestamp[s]"),
+                    "target": datasets.Sequence(datasets.Value("float32")),
+                    "feat_static_cat": datasets.Sequence(datasets.Value("uint64")),
+                    "feat_dynamic_real": datasets.Sequence(datasets.Sequence(datasets.Value("float32"))),
+                    "item_id": datasets.Value("string"),
+                }
+            )
+
+        return datasets.DatasetInfo(
+            # This is the description that will appear on the datasets page.
+            description=_DESCRIPTION,
+            # 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=_HOMEPAGE,
+            # License for the dataset if available
+            license=_LICENSE,
+            # Citation for the dataset
+            citation=_CITATION,
+        )
+
+    def _split_generators(self, dl_manager):
+        urls = _URLS[self.config.name]
+        filepath = dl_manager.download_and_extract(urls)
+
+        return [
+            datasets.SplitGenerator(
+                name=datasets.Split.TRAIN,
+                # These kwargs will be passed to _generate_examples
+                gen_kwargs={
+                    "filepath": filepath,
+                    "split": "train",
+                },
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.TEST,
+                # These kwargs will be passed to _generate_examples
+                gen_kwargs={
+                    "filepath": filepath,
+                    "split": "test",
+                },
+            ),
+            datasets.SplitGenerator(
+                name=datasets.Split.VALIDATION,
+                # These kwargs will be passed to _generate_examples
+                gen_kwargs={
+                    "filepath": filepath,
+                    "split": "dev",
+                },
+            ),
+        ]
+
+    # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
+    def _generate_examples(self, filepath, split):
+        data = pd.read_csv(filepath, parse_dates=True, index_col=0)
+        start_date = data.index.min()
+
+        if self.config.name in ["m1", "m2"]:
+            factor = 4  # 15-min frequency
+        else:
+            factor = 1  # hourly frequency
+        train_end_date_index = 12 * 30 * 24 * factor  # 1 year
+
+        if split == "dev":
+            end_date_index = 12 * 30 * 24 + 4 * 30 * 24 * factor  # 1 year + 4 months
+        else:
+            end_date_index = 12 * 30 * 24 + 8 * 30 * 24 * factor  # 1 year + 8 months
+
+        if self.config.multivariate:
+            if split in ["test", "dev"]:
+                # rolling windows of prediction_length for dev and test
+                for i, index in enumerate(
+                    range(
+                        train_end_date_index,
+                        end_date_index,
+                        self.config.prediction_length,
+                    )
+                ):
+                    yield i, {
+                        "start": start_date,
+                        "target": data[: index + self.config.prediction_length].values.astype("float32").T,
+                        "feat_static_cat": [0],
+                        "item_id": "0",
+                    }
+            else:
+                yield 0, {
+                    "start": start_date,
+                    "target": data[:train_end_date_index].values.astype("float32").T,
+                    "feat_static_cat": [0],
+                    "item_id": "0",
+                }
+        else:
+            if split in ["test", "dev"]:
+                # rolling windows of prediction_length for dev and test
+                for i, index in enumerate(
+                    range(
+                        train_end_date_index,
+                        end_date_index,
+                        self.config.prediction_length,
+                    )
+                ):
+                    target = data["OT"][: index + self.config.prediction_length].values.astype("float32")
+                    feat_dynamic_real = data[["HUFL", "HULL", "MUFL", "MULL", "LUFL", "LULL"]][
+                        : index + self.config.prediction_length
+                    ].values.T.astype("float32")
+                    yield i, {
+                        "start": start_date,
+                        "target": target,
+                        "feat_dynamic_real": feat_dynamic_real,
+                        "feat_static_cat": [0],
+                        "item_id": "OT",
+                    }
+            else:
+                target = data["OT"][:train_end_date_index].values.astype("float32")
+                feat_dynamic_real = data[["HUFL", "HULL", "MUFL", "MULL", "LUFL", "LULL"]][
+                    :train_end_date_index
+                ].values.T.astype("float32")
+                yield 0, {
+                    "start": start_date,
+                    "target": target,
+                    "feat_dynamic_real": feat_dynamic_real,
+                    "feat_static_cat": [0],
+                    "item_id": "OT",
+                }