atul-g / linux_terminal_assistant

A terminal assistant which can answer and help out with linux related queries.

Linux Terminal Assistant

A terminal assistant which can answer and help out with linux related queries.

Model:

1. A transformer model which uses attention mechanism was trained for 30 epochs.

2. The dataset used was the Stack Exchange Data Dump of the askubuntu and unix.stackexchange websites.

3. The dataset was processed into question-answer pairs. Sqlite was also used in the process.

Usage:

1. First install the requirements using pip install -r requirements.txt.

2. For directly using the chatbot, you will also need to have trained weights of the model. Here link of the weights of a model which I have trained for 30 epochs (takes around 30 hours in total). Download the folder and make sure the downloaded checkpoints directory stays in the root of the project directory.

3. Now run python model_predict.py.

Preview:

As you can see, the answer is far from perfect but the fact that the model is able to make proper sentences including proper command syntax is a big positive considering the size of the dataset and the smaller dimensions of model due to less available computing resources. The model could improve greatly by increasing the hyper-parameter values, a larger dataset and also by training it for more steps.