Posts

Here, I’m documenting snippets of my conversations with my wish about a 10-year vision. The Dimond Approach I https://www.diamondapproach.org/method Self-realization involves being grounded and independent, breaking free from past false identities, and aligning actions with one’s true individual essence. Relying on others to fill emotional voids leads to temporary satisfaction and dependency, indicating a closed state rather than openness. True fulfillment comes from inner work and overcoming personal barriers. Compassion evolves from initial emotional sympathy to a more objective, deeper level....

Introduction Following the same philosophy, I aim to first learn about the RAG system by constructing a generic version and then applying optimization techniques. This method will assist me in understanding the pain points, which will in turn guide our brainstorming toward viable applications. As someone who is not very tech-savvy, my primary goal is to grasp the mental model underlying the RAG system. Imagine you are Harry Potter. You want to find everything about the fantastic beast, Niffler, to prepare for your exam....

Introduction When I first heard about text embeddings, I was perplexed. My tech lead handed me an 800-page Natural Language Processing textbook, but I did not finish reading it. Now, with the capabilities of OpenAI models, it’s time to learn and broadly leverage embeddings to accomplish many tasks that were hard be achieved by human labor or simple data analysis, thanks to easy access to these embeddings. Let’s start exploring:...

Introduction I recently delved into two insightful books: “Free Your Mind” and “The Inner Game of Tennis”. Both have shed light on the concept of building systems that encourage growth and resilience. It’s fascinating how everything starts with a simple framework ignited by a spark; it then gains momentum and expands through garnering support, optimization, and evolution. Take, for example, the process of scaling a business from the ground up. It often starts with a minimal viable product (MVP) built by an entrepreneur....

Messy … Introduction to Large Language Model Auto-regressive transformers are pretrained on an extensive corpus of self-supervised data, followed by alignment with human preferences via techniques such as Reinforcement Learning with Human Feedback (RLHF). Training methodology is simple but is limited to a few players with high computational requirements Public pretrained LLM BLOOM (Scao et al., 2022) LLaMa 1 (Touvron et al., 2023) LLaMa 2, LLaMa 2-Chat which is optimized for dialogue use cases Falcon (Penedo et al....