Hugging Face's logo

Datasets:
Banana-Leopard
/
TASTY

Tasks:
Text Classification
Zero-Shot Classification
Modalities:
Text
Formats:
csv
Languages:
English
Size:
1K - 10K
ArXiv:
Tags:
code
Libraries:
Datasets
Dask
License:
Dataset card Data Studio Files Community
1
TASTY / README.md
Banana-Leopard's picture
Banana-Leopard
Update README.md
c5db289 verified
preview code
|
raw
history blame contribute delete
5.67 kB
metadata
license: cc-by-3.0
task_categories:
  - text-classification
  - zero-shot-classification
language:
  - en
tags:
  - code
pretty_name: C++ and Python Code with Complexities from the TASTY paper
size_categories:
  - 1K<n<10K

Dataset Card for TASTY C++ and Python Codes with Complexities

This is a dataset of code snippets with their complexities, both space and time.

As part of this initial release, we cover C++ and Python.

This data was collected as part of our work on the paper called TASTY, published at the ICLR DL4Code workshop, a few years back.

We scraped the data from the popular coding website called GeeksForGeeks (GFG). It is under the CCBY license.

We published this paper before the advent of ChatGPT, so this is not recent by any means and things on the GFG website might have changed since we scraped the data. There was also a lot of manual work that went into correcting the scraped data and putting it into the format that we needed it to be, so some of the complexities might not be exactly what you see on the GFG website. For instance, what is given as "O(n)" on the GFG website is instead written as "linear" in our dataset. In the same way, linear equivalents, "O(n+m)" are also cast to "linear". The same applies to all the classes in our dataset.

There is work being undertaken on the next version of the paper, where we plan to expand to dataset to several more languages and a lot more examples. Probably 10x more samples for 4-5 more languages. There will be a fundamental change in the tone of the work as well since TASTY was published before the ChatGPT, LLMs were not as popular then.

Dataset Details

There are more than 1000 but less than 2000 codes and their space + time complexity.

Dataset Description

  • Curated by: Banana-Leopard
  • Funded by [Banana-Leopard]:
  • Shared by [Banana-Leopard]:
  • Language(s) (NLP): C++ and Python
  • License: CCBY

Dataset Sources

Uses

  • Classification of space and time complexity
  • Eventual Auto Regressive prediciton of the same
  • Cross Language Transfer

Direct Use

Read the paper above for direct uses.