↓Skip to main content
  1. Posts/

My First Meetup on Generative AI

·2 mins

The Knowledge shared on the topic was “Adapting Open Source LLMs (Large Language Model) for your use-case” by Logesh Kumar Umapathi, who is a Lead ML Research Engineer at Saama Technologies, Inc.

It was my First Experience in Participating on Innovative Meetup held at Saama, on the topic of Generative AI.

Even-though I have learnt some of the theory on AI and ML in my UG Degree, It was quite challenging to keep up with the people present there.

I understood a lot, but I found there was so much to learn on these technologies.

The Agenda was Case for Adapting Open Source LLMs, Ways to Adapt in LLMs and Why Open Source LLMs. Logesh Spoke about a lot on the First agenda like “What is a Commercial Api”, “What are the advantages and Disadvantages of LLMs” and “Where we can use the Commercial Api”.

The Second agenda is all about the Ways to adapt Large Language Models. In that we got knowledge on “Prompting”, “Instruction / Task-Specific fine-tuning” and the last one is “Hybrid - LLM Cascade”.

In Prompting we have seen demo on StarCoderPlus, SantaCoder Models and some of the Prompt tuning techniques.

In Instruction Tuning / Supervised fine-tuning we got to know about multiple patterns like Simple Tuning and Multi Task models.

In Hybrid - LLM Cascade, which is a combination of both Open Source and Commercial LLMs. It works as Hoping onto multiple GPT for accurate score until it gets a most preferred match.

My Conclusion about the Meetup held was Very Useful with the latest technique in Open Source LLMs and with some great Questions and Answers asked.

  1. FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performance. (https://arxiv.org/pdf/2305.05176.pdf)

  2. The Power of Scale for Parameter-Efficient Prompt Tuning. (https://aclanthology.org/2021.emnlp-main.243.pdf)

  3. Llama 2: Open Foundation and Fine-Tuned Chat Models (https://arxiv.org/pdf/2307.09288.pdf)

  4. LORA: LOW-RANK ADAPTATION OF LARGE LANGUAGE MODELS (https://openreview.net/pdf?id=nZeVKeeFYf9)

  5. Training language models to follow instructions with human feedback (https://proceedings.neurips.cc/paper_files/paper/2022/file/b1efde53be364a73914f58805a001731-Paper-Conference.pdf)