openpi_v2.py

# coding=utf-8
# Copyright 2022 The TensorFlow Datasets Authors and the HuggingFace Datasets Authors.
#
# 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.

# Lint as: python3
"""Openpi V2: A Dataset for tracking state changes in prcedural text by using an unrestricted library"""

import json
import os
import sys
import textwrap 

import numpy as np
import datasets

logger = datasets.logging.get_logger(__name__)

_OPENPI_V2_CITATION = """\
@inproceedings{
    title={{OPENPI V2}: }
	author={}
	note={}
	year={2022}
}
"""

_OPENPI_V2_DESCRIPTION = """\
TEMPORARY DESCRIPTION
"""

_LICENSE = "CC BY 4.0"
_VERSION = "1.0.0"
_HOMEPAGE = "https://allenai.org/data/openpi"
_URL = "https://huggingface.co/datasets/abhinavk/openpi_v2/resolve/main/data/"
_URLS = {"train": _URL + "train-data.json",
            "dev": _URL + "dev-data.json",
            "test": _URL + "test-data.json"}


class OpenpiConfig(datasets.BuilderConfig):
    """BuilderConfig for Openpi V2."""

    def __init__(
        self, 
        features,
        data_url,
        citation,
        url,
        process_label = lambda x: x,
        **kwargs,
    ):
        
        super(OpenpiConfig, self).__init__(version = datasets.Version(_VERSION), **kwargs)
        self.features = features
        self.data_url = data_url
        self.citation = citation
        self.url = url
        self.process_label = process_label


class OpenpiV2(datasets.GeneratorBasedBuilder):

    BUILDER_CONFIGS = [
        OpenpiConfig(
            name = "openpi_text",
            description = textwrap.dedent(
                """\
            """
            ),
            features = datasets.Features({
                "goal": datasets.Value("string"),
                "steps": [datasets.Value("string")],
                "topics": datasets.Value("string"),
                "image_urls": [datasets.Value("string")],
                "states": [{
                    "answers_openpiv1_metadata": {
                        "entity": datasets.Value("string"),
                        "attribute": datasets.Value("string"),
                        "answers": [datasets.Value("string")],
                        "modality": [datasets.Value("string")]
                    },
                    "entity": datasets.Value("string"),
                    "attribute": datasets.Value("string"),
                    "answers": [datasets.Value("string")],
                    "saliency": datasets.Value("float32")
                }]
            }),
            data_url = _URLS,
            citation = textwrap.dedent(
                """\
                @inproceedings{
                    title={},
                    author={},
                    booktitle={},
                    year={}
                }"""
            ),
            url = _HOMEPAGE
        ),
        OpenpiConfig(
            name = "Task 1",
            description = textwrap.dedent(
                """\
                Given paragraph (e.g., with 5 steps), identify entities that change 
                (challenge: implicit entities, some explicit entities that don’t change)."""
            ),
            features = datasets.Features({
                "steps": [datasets.Value("string")],
                "entity_changes": [[datasets.Value("string")]]
            }),
            data_url = _URLS,
            citation = textwrap.dedent(
                """\
                @inproceedings{
                    title={},
                    author={},
                    booktitle={},
                    year={}
                }"""
            ),
            url = _HOMEPAGE
        ),
        OpenpiConfig(
            name = "Task 3",
            description = textwrap.dedent(
                """\
                Given paragraph (e.g., with 5 steps), identify the attributes of entity that change 
                (challenge: implicit entities, attributes & many combinations)."""
            ),
            features = datasets.Features({
                "steps": [datasets.Value("string")],
                "attr_entity_changes": [datasets.Value("string")]
            }),
            data_url = _URLS,
            citation = textwrap.dedent(
                """\
                @inproceedings{
                    title={},
                    author={},
                    booktitle={},
                    year={}
                }"""
            ),
            url = _HOMEPAGE
        ),
        OpenpiConfig(
            name = "Task 4",
            description = textwrap.dedent(
                """\
                Task 4: Given paragraph & an entity, identify the sequence of attribute value changes 
                (challenge: implicit attributes)."""
            ),
            features = datasets.Features({
                "steps": [datasets.Value("string")],
                "entity": datasets.Value("string"),
                "attribute_changes": [[datasets.Value("string")]]
            }),
            data_url = _URLS,
            citation = textwrap.dedent(
                """\
                @inproceedings{
                    title={},
                    author={},
                    booktitle={},
                    year={}
                }"""
            ),
            url = _HOMEPAGE
        ),
        OpenpiConfig(
            name = "Task 7",
            description = textwrap.dedent(
                """\
                Task 7: Given image url, identify the visual attributes of entity and 
                non-visual attributes of entity that change."""
            ),
            features = datasets.Features({
                "image_url": datasets.Value("string"),
                "visual_attr": [datasets.Value("string")],
                "non_visual_attr": [datasets.Value("string")]
            }),
            data_url = _URLS,
            citation = textwrap.dedent(
                """\
                @inproceedings{
                    title={},
                    author={},
                    booktitle={},
                    year={}
                }"""
            ),
            url = _HOMEPAGE
        ),
    ]

    def _info(self):
        return datasets.DatasetInfo(
            description = _OPENPI_V2_DESCRIPTION,
            features = self.config.features,
            supervised_keys = None,
            homepage = self.config.url,
            citation = self.config.citation + "\n" + _OPENPI_V2_CITATION
        )

    def _split_generators(self, dl_manager):
        downloaded_files = dl_manager.download_and_extract(self.config.data_url)

        return [
            datasets.SplitGenerator(
                name = datasets.Split.TRAIN,
                gen_kwargs = {
                    "filepath": downloaded_files["train"]
                },
            ),
            datasets.SplitGenerator(
                name = datasets.Split.VALIDATION,
                gen_kwargs = {
                    "filepath": downloaded_files["dev"]
                },
            ),
            datasets.SplitGenerator(
                name = datasets.Split.TEST,
                gen_kwargs = {
                    "filepath": downloaded_files["test"]
                }
            ),
        ]


    @staticmethod
    def change_occur(dataset):
        for step in dataset:
            if len(step) > 0:
                return True

        return False

    @staticmethod
    def find_change(key, dataset):
        res = []

        for state in dataset:
            if OpenpiV2.change_occur(state["answers"]):
                list_key = state[key].split(" | ")
                res.append([el for el in list_key])

        return (res)

    @staticmethod
    def find_attr_entity_change(dataset):
        attr_change = []

        for state in dataset:
            if OpenpiV2.change_occur(state["answers"]):
                change_str = "[" + state["attribute"] + "] of [" + state["entity"] + "] changed"
                attr_change.append(change_str)
        
        return attr_change

    def _generate_examples(self, filepath):
        logger.info("generating examples from = %s", filepath)

        if self.config.name == "openpi_text":
            with open(filepath) as f:
                dataset = json.load(f)
    
                for id_ in dataset:
                    yield int(id_), {
                        "goal": dataset[id_]["goal"],
                        "steps": dataset[id_]["steps"],
                        "topics": dataset[id_]["topics"],
                        "image_urls": dataset[id_]["image_urls"],
                        "states": dataset[id_]["states"],
                    }
        
        elif self.config.name == "Task 1":
            with open(filepath) as f:
                dataset = json.load(f)

                for id_ in dataset:
                    steps_ar = dataset[id_]["steps"]
                    entity_changes = OpenpiV2.find_change("entity", dataset[id_]["states"])

                    yield int(id_), {
                        "steps": steps_ar,
                        "entity_changes": entity_changes
                    }

        elif self.config.name == "Task 3":
            with open(filepath) as f:
                dataset = json.load(f)

                for id_ in dataset:
                    steps_ar = dataset[id_]["steps"]
                    attr_entity_changes = OpenpiV2.find_attr_entity_change(dataset[id_]["states"])
        
                    yield int(id_), {
                        "steps": steps_ar,
                        "attr_entity_changes": attr_entity_changes
                    }

        elif self.config.name == "Task 4":
            with open(filepath) as f:
                dataset = json.load(f)

                for id_ in dataset:
                    steps_ar = dataset[id_]["steps"]

                    for state in dataset[id_]["states"]:
                        for el in state["entity"].split(" | "):
                            entity = el
                            attribute_changes = []

                            for state2 in dataset[id_]["states"]:
                                flag = False

                                for el2 in state2["entity"].split(" | "):
                                    if entity == el2:
                                        flag = True
                                        break
                            
                                if flag == False:
                                    continue

                                if OpenpiV2.change_occur(state2["answers"]):
                                    list_attribute = state2["attribute"].split(" | ")
                                    attribute_changes.append([el for el in list_attribute])

                            yield int(id_), {
                                "steps": steps_ar, 
                                "entity": entity,
                                "attribute_changes": attribute_changes
                            }

        elif self.config.name == "Task 7":
            with open(filepath) as f:
                dataset = json.load(f)

                for id_ in dataset:
                    N = len(dataset[id_]["image_urls"])

                    for i in range(N):
                        image_url = dataset[id_]["image_urls"][i]
                        visual_attr = []
                        non_visual_attr = []

                        for state in dataset[id_]["states"]:
                            if len(state["answers"][i]) > 0:
                                visual = False
                                non_visual = False
                                
                                for el in state["answers_openpiv1_metadata"]["modality"][i].split(" | "):
                                    visual = (visual or (el == "with_image"))
                                    non_visual = (non_visual or (el == "without_image"))

                                change_str = "[" + state["attribute"] + "] of [" + state["entity"] + "] changed"
                                
                                if visual:
                                    visual_attr.append(change_str)
                                if non_visual:
                                    non_visual_attr.append(change_str)
                       
                        yield int(id_), {
                            "image_url": image_url,
                            "visual_attr": visual_attr,
                            "non_visual_attr": non_visual_attr
                        }