Debugging LLM Inference Performance: Best Practices and Techniques

An automated script diagnoses LLM inference bottlenecks and suggests concrete fixes like lowering max_tokens, reducing temperature, and switching models.