Fault-Tolerance for LLM InferenceWhat happens if the GPU crashes while answering your ChatGPT request.