There is a parameter in TrainModel.py called "per_device_train_batch_size". If you run out of memory during training.
Lower that number. If you have extra memory, you can increase it.
The higher it is the faster the training unless you run out of memory, then it will be very slow.

Some Texts were provided in the "Texts" folder to test training. You can add your own texts to that folder and train on them.
You may have to restart the program to see the new model.

Run the code by running UI.py

The code's purpose is to use GPT-2 to generate completions. Right now it supports normal completions, the absolute worst completions
and the worst coherent completions. The code can also train a new model using any text file.