Summary of Stay Tuned: An Empirical Study Of the Impact Of Hyperparameters on Llm Tuning in Real-world Applications, by Alon Halfon et al.
Stay Tuned: An Empirical Study of the Impact of Hyperparameters on LLM Tuning in Real-World Applications
by Alon Halfon, Shai Gretz, Ofir Arviv, Artem Spector, Orith Toledo-Ronen, Yoav Katz, Liat Ein-Dor, Michal Shmueli-Scheuer, Noam Slonim
First submitted to arxiv on: 25 Jul 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This research paper presents recommended hyperparameter configurations for fine-tuning large language models (LLMs) on downstream tasks. The authors describe Coverage-based Search (CBS), a method for ranking hyperparameter configurations based on an offline grid search, providing a practical starting point for practitioners. They focus on two SOTA LLMs, Llama-3-8B and Mistral-7B, and two commonly used tuning methods, full fine-tuning and LoRa, conducting over 10,000 tuning experiments. The results suggest that Llama-3-8B and LoRA are generally preferred models, and exploring a few hyperparameter configurations can provide excellent results in practice. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps make language models better at understanding text by suggesting the best ways to adjust their settings. It looks at two top-performing models, Llama-3-8B and Mistral-7B, and finds that they usually work well when fine-tuned just right. The researchers even tested many different settings and found that a few simple ones can give great results. |
Keywords
» Artificial intelligence » Fine tuning » Grid search » Hyperparameter » Llama » Lora
