Datasets:
Tasks:
Text2Text Generation
Modalities:
Text
Formats:
json
Languages:
code
Size:
1K - 10K
Tags:
code-generation
License:
File size: 1,080 Bytes
c34fa1f d92ccbc c34fa1f d92ccbc 1cc585a d92ccbc e338500 d92ccbc |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 |
---
license: cc-by-sa-4.0
language:
- code
task_categories:
- text2text-generation
tags:
- code-generation
pretty_name: DS-1000
---
<h1 align="center"> DS-1000 in simplied format </h1>
This is a reformatting of the DS-1000 dataset with improved ease of use.
* the test process is cleaned up and all contained in the field `code_context`, with the unnecessary pickle and multiple file IOs removed.
* `code_context` defines `test_execution(code: str)` and `test_string(code: str)` which takes the generated code and evalutes their correctness.
* Python is upgraded to 3.10 and libraries upgraded to the latest versions (`environment.yaml`)
* checked that the performance is not overly sensitive to library versions (`enviroment-old-lib.yaml`).
* Removed obsolete `parser` module and simplified string testing
* a few examples were manually fixed; very minor inconsistency due to `import` side-effects when generating tests.
Testing code and examples are in the [original repo](https://github.com/xlang-ai/DS-1000/simplified).
Reformatting credits: Yuhang Lai, Sida Wang
|