If it was a GPT wrapper, it wouldn't require an A100/H100 GPU; the container has a model wrapper, sure, but also it has the wrapped, standalone model, as well;

If it was a GPT wrapper, it wouldn't require an A100/H100 GPU; the container has a model wrapper, sure, but also it has the wrapped, standalone model, as well; its not calling OpenAI's model.