Full textual content turorial (requires MLExpert Professional): https://www.mlexpert.io/prompt-engineering/fine-tuning-llm-on-custom-dataset-with-qlora
On this video, you may discover ways to of fine-tuning the Falcon 7b LLM (40b model is #1 on the Open LLM Leaderboard) on a customized dataset utilizing QLoRA. The Falcon mannequin is free for analysis and industrial use. We’ll use a dataset consisting of Chatbot buyer help FAQs from an ecommerce web site.
All through the video, we’ll cowl the steps of loading the mannequin, implementing a LoRA adapter, and conducting the fine-tuning course of. We’ll additionally monitor the coaching progress utilizing TensorBoard. To conclude, we’ll examine the efficiency of the untrained and educated fashions by evaluating their responses to varied prompts.
Discord: https://discord.gg/UaNPxVD6tv
Put together for the Machine Studying interview: https://mlexpert.io
Subscribe: http://bit.ly/venelin-subscribe
Falcon LLM: https://falconllm.tii.ae/
HuggingFace Open LLM Leaderboard: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard
00:00 – Introduction
01:08 – Textual content Tutorial on MLExpert.io
01:43 – Falcon LLM
04:18 – Google Colab Setup
05:32 – Dataset
08:15 – Load Falcon 7b and QLoRA Adapter
12:20 – Strive the Mannequin Earlier than Coaching
14:40 – HuggingFace Dataset
15:58 – Coaching
20:38 – Save the Educated Mannequin
21:34 – Load the Educated Mannequin
23:19 – Analysis
28:53 – Conclusion
#chatgpt #gpt4 #llms #artificialintelligence #promptengineering #chatbot #transformers #python #pytorch
