Repository files navigation

Easily build a trainer for your Depp Neural Network model and experiment as many as you want to find optimal combination of components(model, optimizer, scheduler) and hyper-parameters in a well-organized manner.

• No more boilerplate code to train and evaluate your DNN model. just focus on your model.
• Simply swap your dataset, model, optimizer and scheduler in the configuration file to find optimal combination. Your code doesn't need to be changed!!!.
• Support Cross Validation, OOF(Out of Fold) Prediction
• Support WandB(https://wandb.ai/) or tensorboard logging.
• Support checkpoint management(Save and load a model. Resume the previous training)
• BuilT easily integrates with Kaggle(https://www.kaggle.com/) notebook. (todo: add notebook link)

Please follow the instruction below to install BuilT.

git clone https://github.com/UoA-CARES/BuilT.gitcd BuilTpython setup.py install

BuilT can be installed using pip(https://pypi.org/project/BuilT/).

pip install built

(todo)

(todo)

(todo)

conda create -n conda_BuilT python=3.7conda activate conda_BuilTpip install -r requirements.txt

https://packaging.python.org/tutorials/packaging-projects/