MENU

Fun & Interesting

Understanding 4bit Quantization: QLoRA explained (w/ Colab)

Discover AI 45,368 2 years ago
Video Not Working? Fix It Now

QLoRA 4bit Quantization for memory efficient fine-tuning of LLMs explained in detailed. 4-bit quantization QLoRA for beginners, theory and code. PEFT - parameter efficient fine-tuning methods. Based on my first videos on the theory of LoRA and other PEFT methods (https://youtu.be/YVU5wAA6Txo) and the detailed code implementation of LoRA in my video (https://youtu.be/A-a-l_sFtYM) now my third video on 4-bit quantization and QLoRA. An additional Colab NB with code to fine-tune FALCON 7B with QLoRA 4-bit quantization and Transformer Reinforcement Learning (TLR). Huggingface Accelerate now supports 4-bit QLoRA LLM models. https://github.com/huggingface/accelerate QLoRA 4-bit Colab NB: (all rights with Author Artidoro Pagnoni) https://colab.research.google.com/drive/1BiQiw31DT7-cDp1-0ySXvvhzqomTdI-o?usp=sharing #4bit #4bits #quantization #languagemodel #largelanguagemodels

Comment