---
dataset_info:
  features:
  - name: code
    dtype: string
  - name: package
    dtype: string
  - name: path
    dtype: string
  - name: filename
    dtype: string
  - name: parsed_code
    dtype: string
  - name: quality_prob
    dtype: float64
  - name: learning_prob
    dtype: float64
  splits:
  - name: train
    num_bytes: 40005369487
    num_examples: 1902405
  download_size: 11174800633
  dataset_size: 40005369487
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
---
# Dataset Card for "pypi_labeled"

All of the latest package versions from pypi. The original data came from [here](https://py-code.org/datasets). I pulled the latest versions of each package, then extracted only `md`, `rst`, `ipynb`, and `py` files.

I then applied some cleaning:

- rendering notebooks
- removing leading comments/licenses

Then filtered out some low-quality code, and labeled the rest according to learning value and quality.  Subset by those columns to get higher quality code.