|
--- |
|
license: cc-by-4.0 |
|
tags: |
|
- code |
|
--- |
|
# Dataset Card for Pipeline2Code |
|
|
|
## Dataset origin |
|
[Code4ML: a Large-scale Dataset of annotated Machine Learning Code](https://zenodo.org/record/7733823) |
|
|
|
## Dataset Summary |
|
|
|
This dataset is designed for the iterative generation of Machine Learning (ML) code based on high-level ML pipeline descriptions. |
|
It consists of code snippets extracted from Kaggle kernels, organized as Jupyter Notebook snippets. |
|
Each kernel includes a set of prompts and completions. |
|
The initial prompt contains an <SOS> token, meta-information about the task the notebook aims |
|
to solve, and the semantic type of the code snippet. The final prompt of each kernel consists of the semantic |
|
type of the code snippet followed by an <EOS> token. Each prompt is associated with a code snippet completion. |
|
Subsequent prompts include previously generated completions and the semantic type of the snippet. |
