This is specifically designed for inference for recommendations models. It’s not for LLM training or inference.