OpenAssistant's Opensource LLMs SagaReference
My involvement with OA
- I started off as a contributor in the early days of the project (Jan 2023)
- Worked in dataset creation, finetuning, and model alignment
- Responsible for building some of the finest open-source chatgpt3.5 alternatives.
- One and only model code owner from India.
- Brief description of the project
- Early days - vision, roadmap, and initial challenges
- Distribution, community, and compute.
- Breakup of different processes involved in creating a high-quality chat model
- dataset creation
- Challenges faced in each of these steps
- Scaling model training 3b to 70b parameters
- Dataset collection procedure behind the OpenAssistant Conversations dataset and its importance to OSS AI. Paper https://arxiv.org/abs/2304.07327
- Demo of one of finest OA models.
- Future of OSS AI and challenges to be solved.
- The Compute poor
- Dense to Sparse models
- lack of feedback data for model alignment
About the speaker
A Data scientist with expertise ranging from classical ML to audio processing. I'm one of top rated Kaggle GrandMaster and contributor to various open-source ML projects including Open-Assistant.