TRAINING.md

LFAI Documentation - Training

This is the official documentation for the LFAI lstm model.

Training is easy!

Youtube Tutorial

Dataset

The first step to start training your model is to find a suitable dataset for your model. A good place to find datasets for free is Kaggle in this example we will be going searching for datasets here Kaggle Text Datasets.

In this example I will use the Shakespeare text because it is only 1M in size which is very small and perfect as a demo dataset.

I will now extract the dataset to the dataset folder in my git clone of LFAI. My folder structure looks like.

dataset/
└── text.txt
0 directories, 1 file

Now that my data is downloaded we need to train a new model on that data. Open a Command Prompt or Terminal in the top level of the git clone/download of LFAI.

Training

We will now execute a command to start training on that text data!

Windows 10/11

python train.py --name="Shakespeare" --dataset="dataset/text.txt" --batchsize=16 --contextsize=128

Linux/Unix

python3 train.py --name="Shakespeare" --dataset="dataset/text.txt" --batchsize=16 --contextsize=128

Final Chapter

This is the end you did it! Now you can move on to the Next Section!