A Django web app developed to make the process of training and testing a Machine Learning model easier.
Currently supports the following functionalities-
- Train and test a Text Classification ULMFiT model (Fastai).
- Train an Image Classifier by fine-tuning ResNet18 pre-trained model (Pytorch).
- Live tracking of model metrics like- training loss, gradients etc. using Weights and Biases
Future work-
- Add hyper-parameter tuning to enable the user to configure his model.
- Improve front-end.
The working of web app is as-
- Choose your application.
- Choose if you want to train or test the model.
- Upload your training data in a zip file.
- Follow the instuctions as shown.
- Open your Weights and Biases account and track model training.