MathCodeBench/code-execution

---
license: apache-2.0
---

# Dataset Card for MathCodeBench Code Execution

This dataset is the MathCodeBench/code-execution dataset converted to Parquet format for efficient processing.

## Dataset Description

- **Repository:** [MathCodeBench/code-execution](https://huggingface.co/datasets/MathCodeBench/code-execution)
- **Point of Contact:** MathCodeBench Community

## Dataset Structure

The dataset is provided as a single Parquet file: `combined_data.parquet`

## Data Fields

The dataset contains a single field:
- json_data: string (JSON-encoded data containing code, input, and output)

## Data Files

This dataset combines two JSONL files:
- demo.jsonl
- cruxeval.jsonl

## Dataset Creation

### Source Data

This is the original MathCodeBench/code-execution dataset on Hugging Face.

### Processing

The original dataset JSONL files were loaded directly from Hugging Face, combined, and then stored in a Parquet file using PyArrow for improved performance and storage efficiency.

## Considerations for Using the Data

Users should be aware that the dataset contains JSON-encoded data with code snippets, inputs, and expected outputs. It's primarily intended for benchmarking and evaluating code understanding and generation models. To use the data, you'll need to parse the JSON strings.

## Additional Information

For more information, please refer to the original dataset page: [MathCodeBench/code-execution](https://huggingface.co/datasets/MathCodeBench/code-execution)