(
data_file: str = None,
dataset_type: str = "humaneval",
num_shot=None,
)
| 66 | |
| 67 | |
| 68 | def read_dataset( |
| 69 | data_file: str = None, |
| 70 | dataset_type: str = "humaneval", |
| 71 | num_shot=None, |
| 72 | ) -> Dict: |
| 73 | if num_shot is not None: |
| 74 | print(f"{num_shot}-shot setting...") |
| 75 | if "humaneval" in dataset_type.lower(): |
| 76 | if data_file is None: |
| 77 | current_path = os.path.dirname(os.path.abspath(__file__)) |
| 78 | data_file = os.path.join(current_path, "..", "humaneval-x", "python", "data", "humaneval_python.jsonl.gz") |
| 79 | dataset = {task["task_id"]: task for task in stream_jsonl(data_file)} |
| 80 | else: |
| 81 | raise f"Dataset: {dataset_type} not supported." |
| 82 | |
| 83 | return dataset |
| 84 | |
| 85 | def estimate_pass_at_k( |
| 86 | num_samples: Union[int, List[int], np.ndarray], |
no test coverage detected