Tutoring on fine-tuning skill with Unsloth/lm-evaluation-harness

---

---

I'm part of a open source AI research team. Current focus is this project: https://github.com/thehunmonkgroup/raspberry-paper-to-cot-pipeline In essence, trying to create a pipeline that extracts high quality CoTs from academic papers. The pipeline itself is solid, but I'm stuck on assessing the quality of the data. In particular, my goal is to fine-tune some 7/8B models via Unsloth and compare their performance to vanilla models using lm-evaluation-harness -- I am, however, quite inexperienced with fine-tuning. I can run a fine-tune, but have no idea how to make any kind of adjustments to the process or the data to improve the quality of the fine-tuned model. So, I'm looking for a consultant to help tutor me on this skill.

Keyword: Project Consultant

Price: $65.0

Machine Learning AI Model Training AI Model Development Hyperparameter Tuning PyTorch