import csv
import json
import os
import uuid
from typing import Literal
import subprocess
from adalflow.utils.data import Dataset
from adalflow.datasets.types import Example
from adalflow.utils.file_io import save_csv
from adalflow.datasets.utils import prepare_dataset_path
# TODO: here users clean adalflow created files
[docs]
class BigBenchHard(Dataset):
__doc__ = """Big Bench Hard dataset for object counting task.
You can find the task name from the following link:
https://github.com/suzgunmirac/BIG-Bench-Hard/tree/main/bbh
Data will be saved to ~/.adalflow/cache_datasets/BBH_object_counting/{split}.csv
if root is not specified.
Size for each split:
- train: 50 examples
- val: 50 examples
- test: 100 examples
Args:
task_name (str): The name of the task. "BHH_{task_name}" is the task name in the dataset.
root (str, optional): Root directory of the dataset to save the data. Defaults to ~/.adalflow/cache_datasets/task_name.
split (str, optional): The dataset split, supports ``"train"`` (default), ``"val"`` and ``"test"``.
"""
def __init__(
self,
task_name: Literal["BBH_object_counting"] = "BBH_object_counting",
root: str = None,
split: Literal["train", "val", "test"] = "train",
*args,
**kwargs,
):
if split not in ["train", "val", "test"]:
raise ValueError("Split must be one of 'train', 'val', 'test'")
# if root is None:
# root = get_adalflow_default_root_path()
# print(f"Saving dataset to {root}")
self.root = root
self.split = split
self.task_name = "_".join(task_name.split("_")[1:])
data_path = prepare_dataset_path(self.root, self.task_name)
self._check_or_download_dataset(data_path, split)
self.data = []
split_csv_path = os.path.join(data_path, f"{split}.csv")
with open(split_csv_path, newline="") as csvfile:
reader = csv.DictReader(csvfile)
for row in reader:
self.data.append(
Example(question=row["x"], answer=row["y"], id=row["id"])
) # dont use a tuple, use a dict {"x": ..., "y": ...}
def _check_or_download_dataset(self, data_path: str = None, split: str = "train"):
if data_path is None:
raise ValueError("data_path must be specified")
json_path = os.path.join(data_path, f"{self.task_name}.json")
split_csv_path = os.path.join(data_path, f"{split}.csv")
if os.path.exists(split_csv_path):
return
print(f"Downloading dataset to {json_path}")
subprocess.call(
[
"wget",
f"https://raw.githubusercontent.com/suzgunmirac/BIG-Bench-Hard/main/bbh/{self.task_name}.json",
"-O",
json_path,
]
)
with open(json_path) as json_file:
data = json.load(json_file)
examples = data["examples"]
# NOTE: better to shuffle the examples before splitting.
# We do this splitting in order to be consistent with text-grad paper.
train_examples = [
{"x": ex["input"], "y": ex["target"], "id": str(uuid.uuid4())}
for ex in examples[:50]
]
val_examples = [
{"x": ex["input"], "y": ex["target"], "id": str(uuid.uuid4())}
for ex in examples[50:100]
]
test_examples = [
{"x": ex["input"], "y": ex["target"], "id": str(uuid.uuid4())}
for ex in examples[150:250]
]
# ensure the
for split, examples in zip(
["train", "val", "test"], [train_examples, val_examples, test_examples]
):
target_path = os.path.join(data_path, f"{split}.csv")
print(f"Saving {split} split to {target_path}")
save_csv(examples, f=target_path, fieldnames=["x", "y", "id"])
def __getitem__(self, index) -> Example:
return self.data[index]
def __len__(self):
return len(self.data)
[docs]
@staticmethod
def get_default_task_instruction():
_task_description = "You will answer a reasoning question. Think step by step. The last line of your response should be of the following format: 'Answer: $VALUE' where VALUE is a numerical value."
return _task_description
if __name__ == "__main__":
from adalflow.datasets.big_bench_hard import BigBenchHard
dataset = BigBenchHard("BBH_word_sorting", split="train")
print(dataset[0:10])
print(len(dataset))
print(dataset.get_default_task_instruction())
