aidkit, the first aid kit for AI development, verification & validation. The AI-debugging &-boosting toolkit is both the embodiment of quality standards, as well as the plug & play tool for AI developers who want to put their model through its paces in every section of the AI lifecycle to reduce costs / iterations.
Project description
aidkit is the quality gate between machine learning models and the deployment of those models.
Installation
- Activate your virtual environment with python 3.6, e.g.
source venv/bin/activate
pip install aidkit
Example Usage
Authenticate
The only requirement for using aidkit is having a license for it.
To authenticate, you need to run the following once:
python -m aidkit.authenticate --url <subdomain>.aidkit.ai --token <your auth token>
Model
You can upload a model to aidkit, or list the names of all models uploaded.
For uploading, you need a keras .h5 file, that contains a LSTM architecture. Do the following to upload it:
python -m aidkit.model --file <path to your h5 file>
To list all uploaded models type:
python -m aidkit.model
Data
You can upload a data set to aidkit, or list the names of all datasets uploaded.
For uploading, you need a zip file. We expect a zip, containing a folder, that is named like the dataset should be called. This subfolder contains INPUT and OUTPUT folders that each contain csv files. Do the following to upload it:
python -m aidkit.data --file <path to your zip file>
To list all uploaded datasets type:
python -m aidkit.data
Analysis
You can start a new quality analysis. For doing so, you need a toml file. This file will follow a specified toml standard. Do the following to upload it:
python -m aidkit.analysis --file <path to your toml file>
To list all uploaded datasets type:
python -m aidkit.analysis
Visualization
After running an analysis you can observe the results in our web-GUI. to get the link type:
python -m aidkit.url
Just follow the link and authorize yourself with your credentials.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.