Command line and Python library interface to Cleanlab Studio. Analyze datasets and produce cleansets (cleaned datasets) with Cleanlab Studio in a single line of code!
You can install the Cleanlab Studio client from PyPI with:
pip install cleanlab-studio
If you already have the client installed and wish to upgrade to the latest version, run:
pip install --upgrade cleanlab-studio
from cleanlab_studio import Studio
# create your Cleanlab Studio API client with your API key, found here: https://app.cleanlab.ai/account
studio = Studio(<your api key>)
# upload your dataset via a filepath, Pandas DataFrame, or PySpark DataFrame!
dataset_id: str = studio.upload_dataset(<your dataset>, <your dataset name>)
# navigate to Cleanlab Studio, create a project, and improve your labels
# download your cleanset or apply corrections to your local Pandas or PySpark dataset!
# you can find your cleanset ID by clicking on the Export Cleanset button in your project
cleanset = studio.download_cleanlab_columns(<your cleanset id>)
corrected_dataset = studio.apply_corrections(<your cleanset id>, <your dataset>)
The Python API offers significantly more functionality than is available in the Cleanlab Studio Web Application. Learn more via the tutorials and quickstart guide.
cleanlab login. You can find your API key at https://app.cleanlab.ai/account.cleanlab dataset upload.cleanlab cleanset download.Learn more about the Cleanlab Studio CLI from the quickstart guide.
Cleanlab Studio supports the following data formats:
Information on how to format your dataset can be found by clicking the tutorial on https://app.cleanlab.ai/upload or in the Datasets concept guide. We also provide tutorials for converting certain common Python image or text datasets.
All schema information will be inferred by default when uploading a dataset through the Python API. We provide some options to override the inferred schema if necessary:
modality kwarg to studio.upload_dataset(). Supported modalities include "text", "tabular", and "image"id_column kwarg to studio.upload_dataset()schema_overrides kwarg to studio.upload_dataset() in the following format:{
<name_of_column_to_override>: {
"data_type": <desired_data_type>,
"feature_type": <desired_feature_type>,
},
...
}
To specify the column types in your dataset, create a JSON file named schema.json. If you would like to edit an inferred schema (rather than starting from scratch) follow these steps:
cleanlab dataset uploadn to decliney to accept. Then enter the filename you'd like to save to (schema.json by default)cleanlab dataset upload --schema_path [path to schema file]Your schema file should be formatted as follows:
{
"metadata": {
"id_column": "tweet_id",
"modality": "text",
"name": "Tweets.csv"
},
"fields": {
"tweet_id": {
"data_type": "string",
"feature_type": "identifier"
},
"sentiment": {
"data_type": "string",
"feature_type": "categorical"
},
"sentiment_confidence": {
"data_type": "float",
"feature_type": "numeric"
},
"retweet_count": {
"data_type": "integer",
"feature_type": "numeric"
},
"text": {
"data_type": "string",
"feature_type": "text"
},
"tweet_created": {
"data_type": "boolean",
"feature_type": "boolean"
},
"tweet_created": {
"data_type": "string",
"feature_type": "datetime"
}
},
"version": "0.1.12"
}
This is the schema of a hypothetical dataset Tweets.csv that contains tweets, where the column tweet_id contains a
unique identifier for each record. Each column in the dataset is specified under fields with its data type and feature
type.
Data type refers to the type of the field's values: string, integer, float, or boolean.
Note that the integer type is partially strict, meaning floats that are equal to integers (e.g. 1.0, 2.0, etc)
will be accepted, but floats like 0.8 and 1.5 will not. In contrast, the float type is lenient, meaning integers
are accepted. Users should select the float type if the field may include float values. Note too that integers can have
categorical and identifier feature types, whereas floats cannot.
For booleans, the list of accepted values are: true/false, t/f, yes/no, 1/0, 1.0/0.0.
Feature type refers to the secondary type of the field, relating to how it is used in a machine learning model, such as whether it is:
Some feature types can only correspond to specific data types. The list of possible feature types for each data type is shown below
| Data type | Feature type |
|---|---|
| string | text, categorical, datetime, identifier, filepath |
| integer | categorical, datetime, identifier, numeric |
| float | datetime, numeric |
| boolean | boolean |
The datetime type should be used for datetime strings, e.g. "2015-02-24 11:35:52 -0800", and Unix timestamps (which
will be integers or floats). Datetime values must be parseable
by polars.from_epoch for integer/floats or polars.Expr.str.strptime for strings.
version indicates the version of the Cleanlab CLI package version used to generate the schema.
FAQs
Client interface for all things Cleanlab Studio
We found that cleanlab-studio demonstrated a healthy version release cadence and project activity because the last version was released less than a year ago. It has 4 open source maintainers collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Security News
PyPI confirms no security flaws were exploited in the Ultralytics supply chain attack and highlights improvements for safer package publishing.
Security News
Research
The Socket Research Team breaks down a malicious wrapper package that uses obfuscation to harvest credentials and exfiltrate sensitive data.
Research
Security News
Attackers used a malicious npm package typosquatting a popular ESLint plugin to steal sensitive data, execute commands, and exploit developer systems.