Enhancement of LLMs with RLHF and Fine-tuning

While Large Language Models like ChatGPT, Bard, LLaMa are already highly functional, organisations may still need to enhance the behavior, accuracy, and performance of LLMs, especially in the context of a given domain or problem statement. RLHF and fine-tuning are 2 ways of doing this, and are of critical importance to the large-scale adoption of LLMs. Reinforcement Learning with Human Feedback (RLHF) addresses biases and limitations in initial model behavior by incorporating human feedback, while fine-tuning allows customization of the model for specific tasks or domains.

AGENDA:
3:30 PM - Welcome note
3:35 PM - RLHF: What is it and why is it needed?
3:45 PM - Basics of Reinforcement Learning
3:55 PM - Solving practical tasks with RHFL
4:00 PM - Fine-tuning LLMs on consumer hardware
4:15 PM - Diffusion models
4:20 PM - Q&A
4:30 pm - Conclusion

SPEAKERS:
Evgeniya Sukhodolskaya is a Data Advocate at Toloka, a global data labeling platform serving around 2,000 large and small businesses worldwide. She’s worked as an analyst-developer, a machine learning engineer, a solution architect, and a business analyst. She’s spent 3.5 years working in crowdsourcing. Evgeniya’s educational background is in Artificial Intelligence & Data Engineering, and she’s currently getting her master’s degree at The Technical University of Munich.

Aniket Maurya is a Developer Advocate at Lightning AI (the same team behind PyTorch Lightning). He has worked on model development and deployment at scale in past experience and currently working on building the opensource community and creating high quality content for teaching machine learning and LLMs to the AI/ML community.

REGISTRATION:
This session is FREE to attend but prior registration is required, with seats being available on a first-come-first-served basis. Please register by filling in this form: https://forms.gle/Wv8d8dcgBoPDD13x9

Link to the livestream will be sent to the email submitted in the above registration form. To ensure that you receive it, please add info@appliedsingularity.com to your email safe list.

Please reach Nihal at 9663374431 or at info@appliedsingularity.com if you need any clarifications or have any challenges in registration. We look forward to seeing many of you at the session! We thank Toloka for hosting this session with us.