Model Fine-tuning: Difference between revisions

From CMU -- Language Technologies Institute -- HPC Wiki
Jump to navigation Jump to search
Yifengw2 (talk | contribs)
adding fine tune resources
 
Yifengw2 (talk | contribs)
adding fine tune resources and time estimation suggestions
 
Line 1: Line 1:
===Fine-tune Tutorial===
===Fine-tune Resources===


* [https://huggingface.co/docs/transformers/training Fine-tune a pre-trained model on HuggingFace]
* [https://huggingface.co/docs/transformers/training Fine-tune a pre-trained model on HuggingFace]
* [https://platform.openai.com/docs/guides/fine-tuning Fine-tune documentation on OpenAI]
* [https://platform.openai.com/docs/guides/fine-tuning Fine-tune documentation on OpenAI]
*
*[https://huggingface.co/blog/stackllama#stackllama-a-hands-on-guide-to-train-llama-with-rlhf StackLLaMA: A hands-on guide to train LLaMA with RLHF]
*Parameter-Efficient Fine-Tuning (PEFT) -- An efficient fine-tuning approach to reduce computational resources and time cost
**[https://huggingface.co/blog/peft Blog]
**[https://github.com/huggingface/peft Github]
 
=== Fine-tune time estimation (Based on previous Hackathon feedback) ===
 
* Fine-tuning ESM-2 model with 35M parameter takes ~4.5 hours
* Fine-tuning ESM-2 model with 8M parameter takes ~2.5 hours

Latest revision as of 01:41, 4 August 2023

Fine-tune Resources

[edit | edit source]

Fine-tune time estimation (Based on previous Hackathon feedback)

[edit | edit source]
  • Fine-tuning ESM-2 model with 35M parameter takes ~4.5 hours
  • Fine-tuning ESM-2 model with 8M parameter takes ~2.5 hours