Large Language Models for EDA Cloud Job Resource and Lifetime Prediction

This paper presents a compelling application of Large Language Models (LLMs) to a practical problem in the Electronic Design Automation (EDA) industry: resource and job lifetime prediction in cloud environments. The authors address the limitations of traditional machine learning methods by leveraging the power of LLMs for text-to-text regression. The introduction of scientific notation and prefix filling to constrain the LLM's output is a clever approach to improve reliability. The finding that full-attention finetuning enhances prediction accuracy is also significant. The use of real-world cloud datasets to validate the framework strengthens the paper's credibility and establishes a new performance baseline for the EDA domain. The research is well-motivated and the results are promising.