m_qalm.py

import json
import os
import datasets
import ast

_DESCRIPTION = """\
The M-QALM Dataset Repository contains Multiple-Choice and Abstractive Questions for evaluating the performance of LLMs in the clinical and biomedical domain.
"""

_HOMEPAGE = "https://huggingface.co/datasets/anand-s/m_qalm"

_LICENSE = "Apache License 2.0"

# Define URLs for different parts of the dataset if applicable
_URLS = {
    "train_normal_mcqa": "https://huggingface.co/datasets/anand-s/m_qalm/resolve/main/train_normal_mcqa.zip",
    "val_normal_mcqa": "https://huggingface.co/datasets/anand-s/m_qalm/resolve/main/val_normal_mcqa.zip",
    "test_normal_mcqa": "https://huggingface.co/datasets/anand-s/m_qalm/resolve/main/test_normal_mcqa.zip",
    "train_context_mcqa": "https://huggingface.co/datasets/anand-s/m_qalm/resolve/main/train_context_mcqa.zip",
    "val_context_mcqa": "https://huggingface.co/datasets/anand-s/m_qalm/resolve/main/val_context_mcqa.zip",
    "test_context_mcqa": "https://huggingface.co/datasets/anand-s/m_qalm/resolve/main/test_context_mcqa.zip",
    "train_aqa": "https://huggingface.co/datasets/anand-s/m_qalm/resolve/main/train_aqa.zip",
    "val_aqa": "https://huggingface.co/datasets/anand-s/m_qalm/resolve/main/val_aqa.zip",
    "test_aqa": "https://huggingface.co/datasets/anand-s/m_qalm/resolve/main/test_aqa.zip",    

}

class MQalm(datasets.GeneratorBasedBuilder):
    """Dataset for multiple choice questions from Test Repo."""

    VERSION = datasets.Version("1.0.0")

    BUILDER_CONFIGS = [
        datasets.BuilderConfig(name="train_normal_mcqa", version=VERSION, description="Train set MCQA"),
        datasets.BuilderConfig(name="val_normal_mcqa", version=VERSION, description="Val set MCQA"),
        datasets.BuilderConfig(name="test_normal_mcqa", version=VERSION, description="Test set MCQA"),
        datasets.BuilderConfig(name="train_context_mcqa", version=VERSION, description="Train set context MCQA"),
        datasets.BuilderConfig(name="val_context_mcqa", version=VERSION, description="Val set context MCQA"),
        datasets.BuilderConfig(name="test_context_mcqa", version=VERSION, description="Test set context MCQA"),
        datasets.BuilderConfig(name="train_aqa", version=VERSION, description="Train set AQA"),
        datasets.BuilderConfig(name="val_aqa", version=VERSION, description="Val set AQA"),
        datasets.BuilderConfig(name="test_aqa", version=VERSION, description="Test set AQA"),
    ]


    def _info(self):
        features_dict = {
            "train_normal_mcqa": datasets.Features({
                "prompt": datasets.Value("string"),
                "question": datasets.Value("string"),
                "options": datasets.Sequence(datasets.Value("string")),
                "answer": datasets.Value("string"),
                "num_options": datasets.Value("string"),
                "question_type": datasets.Value("string"),
                "dataset_name": datasets.Value("string"),
            }),
            "val_normal_mcqa": datasets.Features({
                "prompt": datasets.Value("string"),
                "question": datasets.Value("string"),
                "options": datasets.Sequence(datasets.Value("string")),
                "answer": datasets.Value("string"),
                "num_options": datasets.Value("string"),
                "question_type": datasets.Value("string"),
                "dataset_name": datasets.Value("string"),
                "few_shot_prompt": datasets.Sequence(datasets.Features({
                    "question": datasets.Value("string"),
                    "answer": datasets.Value("string"),
                    "options": datasets.Sequence(datasets.Value("string")),
                })),                
            }),
            "test_normal_mcqa": datasets.Features({
                "prompt": datasets.Value("string"),
                "question": datasets.Value("string"),
                "options": datasets.Sequence(datasets.Value("string")),
                "answer": datasets.Value("string"),
                "num_options": datasets.Value("string"),
                "question_type": datasets.Value("string"),
                "dataset_name": datasets.Value("string"),
                "few_shot_prompt": datasets.Sequence(datasets.Features({
                    "question": datasets.Value("string"),
                    "answer": datasets.Value("string"),
                    "options": datasets.Sequence(datasets.Value("string")),
                })),                
            }),
            "train_context_mcqa": datasets.Features({
                "prompt": datasets.Value("string"),
                "question": datasets.Value("string"),
                "options": datasets.Sequence(datasets.Value("string")),
                "answer": datasets.Value("string"),
                "num_options": datasets.Value("string"),
                "question_type": datasets.Value("string"),
                "dataset_name": datasets.Value("string"),
                "context": datasets.Sequence(datasets.Value("string"))
            }),
            "val_context_mcqa": datasets.Features({
                "prompt": datasets.Value("string"),
                "question": datasets.Value("string"),
                "options": datasets.Sequence(datasets.Value("string")),
                "answer": datasets.Value("string"),
                "num_options": datasets.Value("string"),
                "question_type": datasets.Value("string"),
                "dataset_name": datasets.Value("string"),
                "context": datasets.Sequence(datasets.Value("string")),
                "few_shot_prompt": [{
                    "question": datasets.Value("string"),
                    "answer": datasets.Value("string"),
                    "options": datasets.Sequence(datasets.Value("string")),
                    "context": datasets.Sequence(datasets.Value("string")),
                }],                
            }),
            "test_context_mcqa": datasets.Features({
                "prompt": datasets.Value("string"),
                "question": datasets.Value("string"),
                "options": datasets.Sequence(datasets.Value("string")),
                "answer": datasets.Value("string"),
                "num_options": datasets.Value("string"),
                "question_type": datasets.Value("string"),
                "dataset_name": datasets.Value("string"),
                "context": datasets.Sequence(datasets.Value("string")),
                "few_shot_prompt": datasets.Sequence(datasets.Features({
                    "question": datasets.Value("string"),
                    "answer": datasets.Value("string"),
                    "options": datasets.Sequence(datasets.Value("string")),
                    "context": datasets.Sequence(datasets.Value("string")),               
                })),                
            }),
            "train_aqa": datasets.Features({
                "question": datasets.Value("string"), 
                "answer": datasets.Value("string"),  
                "prompt": datasets.Value("string"),
                "dataset_name": datasets.Value("string"),
            }),
            "val_aqa": datasets.Features({
                "question": datasets.Value("string"),  
                "answer": datasets.Sequence(datasets.Value("string")),  
                "prompt": datasets.Value("string"),
                "dataset_name": datasets.Value("string"),
            }), 
            "test_aqa": datasets.Features({
                "question": datasets.Value("string"),  
                "answer": datasets.Sequence(datasets.Value("string")),  
                "prompt": datasets.Value("string"),
                "dataset_name": datasets.Value("string"),
            }),            
        }
        
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            features=features_dict[self.config.name],
            homepage=_HOMEPAGE,
            license=_LICENSE
        )

    def _split_generators(self, dl_manager):
        # Download and extract all the files in the directory
        data_dir = dl_manager.download_and_extract(_URLS[self.config.name])
        return [
            datasets.SplitGenerator(
                name=self.config.name,
                gen_kwargs={"directory": data_dir, "split": self.config.name},
            )
        ]

    def _generate_examples(self, directory, split):
        # Iterate over each file in the directory
        key_idx = 0
        for filename in os.listdir(directory):
            filepath = os.path.join(directory, filename)
            if filepath.endswith(".jsonl"):
                with open(filepath, encoding="utf-8") as f:
                    for key, row in enumerate(f):
                        data = json.loads(row)
                        if split == "train_normal_mcqa":
                            yield key_idx, {
                                "prompt": data["prompt"],
                                "question": data["question"],
                                "options": data["options"],
                                "answer": data["answer"],
                                "num_options": data["num_options"],
                                "question_type": data["question_type"],
                                "dataset_name": os.path.split(filepath)[-1].replace(".jsonl","")
                            }
                            key_idx +=1
                        elif split in ["val_normal_mcqa", "test_normal_mcqa"]:
                            yield key_idx, {
                                "prompt": data["prompt"],
                                "question": data["question"],
                                "options": data["options"],
                                "answer": data["answer"],
                                "num_options": data["num_options"],
                                "question_type": data["question_type"],
                                "dataset_name": os.path.split(filepath)[-1].replace(".jsonl",""),
                                "few_shot_prompt": [{
                                "question": item["question"],
                                "answer": item["answer"],
                                "options": item["options"],
                            } for item in data["few_shot_prompt"]]
                            }
                            key_idx +=1

                        elif split == "train_context_mcqa":
                            yield key_idx, {
                                "prompt": data["prompt"],
                                "question": data["question"],
                                "options": data["options"],
                                "answer": data["answer"],
                                "num_options": data["num_options"],
                                "context": data["context"],
                                "question_type": data["question_type"],
                                "dataset_name": os.path.split(filepath)[-1].replace(".jsonl","")
                            }
                            key_idx +=1
                        elif split in ["val_context_mcqa", "test_context_mcqa"]:

                            yield key_idx, {
                                "prompt": data["prompt"],
                                "question": data["question"],
                                "options": data["options"],
                                "answer": data["answer"],
                                "num_options": data["num_options"],
                                "context": data["context"],                                
                                "question_type": data["question_type"],
                                "dataset_name": os.path.split(filepath)[-1].replace(".jsonl",""),
                                "few_shot_prompt": data["few_shot_prompt"]}                                                        
                            key_idx +=1

                        elif split in ["train_aqa", "val_aqa", "test_aqa"]:
                            yield key_idx, {
                                "prompt": data["prompt"],
                                "question": data["question"],
                                "answer": data["answer"],
                                "dataset_name": os.path.split(filepath)[-1].replace(".jsonl","")
                            }
                            key_idx +=1