I'm looking for an experienced LLM (Large Language Model) trainer who consistently achieves excellent results to coach me, a hobbyist, on best practices for fine-tuning and training.
Ideally, you will have a proven track record in training LLMs, particularly with LoRA fine-tuning, and be familiar with Oobabooga + Text Generation WebUI, as this is the setup I'm currently using. You can find more about it here: https://github.com/oobabooga/text-generation-webui.
My current setup includes a local training desktop with AMD Ryzen 5 7600, 128GB DDR5, and 12GB RTX 3060. I’m using Oobabooga + Text Generation WebUI, and the model I’m working with is Llama 3.1 8b Instruct. The training data I’m using consists of textbooks on grant writing, examples of successful grants, and raw content that I've organized into clean .txt files.
The goal is to develop an LLM model capable of significantly improving grant writing quality. I need coaching on several aspects of this project, including whether my methods for preparing training data (e.g., using .txt versus .json files) are optimal. I also need advice on whether my WebUI setup is correct and whether I’m following best practices.
Additionally, I’m open to exploring other technologies or tools that could lead to better results. I'm eager to hear any suggestions or improvements you might have.
A little about me: I have basic Python and technical skills, but I’m a quick learner with a strong desire to improve. I prefer real-time coaching, and I’m open to sharing my screen so you can provide feedback on the work I’ve done so far. It would also be helpful if you could provide written documentation or a summary of best practices after our session(s).
Ideally, I’d like to cover everything in a 1-3 hour consulting call. Please let me know your availability, including your time zone, when applying. I’m open to reviewing all bids, so please include your rate in your application.
Thanks in advance. I look forward to learning from you!
Location: Anywhere
Posted: Aug. 28, 2024, 10:33 p.m.
Apply Now Company Website