Machine Learning tracking experiment and debugging tools.
| Branch | Build | Coverage | Linting | Release | License |
|---|---|---|---|---|---|
| main |  |
 |
  |
 |
UETAI is a customize PyTorch logger which will able to help users track machine learning experiment, and esily debug raw datasets and trained models.
UETAI provided tools for helping user tracking their experiment, visualizing the dataset, results, and debuging the model (and the raw dataset also) with little effort by integrated the tools into the dashboards which users are using for logging.
In this beta version, we will only focus on integrated Comet ML, which is amazing dashboard with well-writen API and customable panel
Getting started
Firstly, you must sign up for an account from one of these supported MLTE (Machine Learning tracking experiment) tools, each dashboard will give you a unique API key to log in dashboard from any terminal or code:
| Dashboard | Status |
|---|---|
| Comet ML | |
| Weights & Biases | |
| MLFlow |
Install uetai
You install uetai with pip by running:
pip install uetaiOr install from source repository:
git clone [email protected]:UETAILab/uetai.git; cd uetai
pip install -e .Basic usage
Importing and initialize your supported dashboard logger (for example: Comet ML) and start logging your experiment:
from src import CometLogger
logger = CometLogger(project_name="Uetai project")
# training process
logger.log({"loss": loss, "acc": acc})Examples
Coming soon...
The team
UETAI is a non-profit project hosted by AI Laboratory of University of Engineering and Technology.
UETAI is currently maintained by manhdung20112000 with the support from BS. Phi Nguyen Van - gungui98 as an advisor.
License
UETAI has a MIT license, as found in the LICENSE file.