GPT stands for Generative Pre-trained Transformer.

The "main" training step using huge amounts of inputs is called pre-training. The idea is that after that pre-training, you might fine tune the model for your specific use case.

As opposed to fine-tuning or in-context learning. It really is normal training.