The Bojai applet provides an interactive visual interface to help you prepare, train, and deploy your machine learning models through clearly defined stages. This guide walks you through the full workflow using the UI, including:

• Initializing your model
• Moving from one stage to the other
• Viewing and updating your data
• Training and evaluating your model
• Downloading or using your model

Initialize Your Pipeline

This is the first screen you see when launching the UI. Follow the steps below to begin:

1. Under Enter Model Name, type the name you’d like to assign to this session.
2. Under Enter Data Address, click Browse to select your dataset. Make sure your data matches one of the supported formats.
3. Under Training and Evaluation, enter fractions of the data to be used for each (must add up to 1.0).
4. Click START → to initialize your pipeline and proceed to the Prepare stage.

Moving from one stage to the other

If you want to move from one stage to the other, you can click on the stage name in the upper navigation bar. The following restrictions apply:

1. If you are in prepare, you cannot directly go to deploy stage. First, you should go to train stage and train your model.

2. If you are in train stage, you cannot go to deploy before training your model. Once it is trained, you can go back.

All other movements are supported and can work.

View and Update Your Data

Once you’re in the Prepare stage, you can inspect your dataset and model setup.

To view a specific datapoint:

1. In the input box under View tokenized or untokenized data from your dataset, enter a number from 1 to the number of data points.
2. Click View Raw or View Tokenized to preview the sample.

To view a random datapoint:

1. Leave the input box under View tokenized or untokenized data from your dataset empty.
2. Click View Raw or View Tokenized.

To update your dataset:

1. Under Enter Data Address, click Browse and select your new dataset.
2. Click Update Data.
3. A success or error message will appear to confirm the update.

Train Your Model

In the Train stage, you can configure hyperparameters, start training, and evaluate model performance.

To start training:

1. Click Start Training.
2. A progress window will open. Loss will be shown for each epoch.

⚠️ Warning: Training cannot be stopped once started. Closing the window will not stop training.

To update hyperparameters:

1. Under Update Hyperparameters section, enter the new value for each hyperparameter you want to update. Leave the ones you don’t want to update empty.
2. Click Update Hyperparameters. You should see the new values updated under the Hyperparameters section.

To replace your model:

1. Click Replace Model. Your new model is ready to be trained again.

⚠️ Warning: Replacing deletes the current model. Download it first if you want to keep it.

To evaluate your model in the training stage:

Click Evaluate Model Output to view the model’s performance on the validation data.

Deploy Your Model

In the Deploy stage, you can evaluate your model on new data or on validation data you uploaded earlier, save your model, or use it for inference.

To evaluate on new data:

1. Under New Evaluation Data section, click Browse to select a new dataset.
2. Click Update Data. The Evaluate Model With New Data button appears.
3. Click Evaluate Model With New Data.

To evaluate on original data:

Click Evaluate Model With Original Data to compute metrics on your default validation split.

To use the model for inference:

1. Under Use Model section and under Enter [prompt depending on model], insert the appropriate input.
2. Click Use Model to generate and view a response.

To download the model:

1. Under Download Model section, click Browse to choose the save directory.
2. Click Download Model.
3. A .bin file will be saved. Rename it to avoid overwriting in future sessions.

This completes the end-to-end workflow for using Bojai’s Visual Interface to manage machine learning pipelines.

If you have suggestions on adding more features, submit a GitHub issue describing the new model in detail.