LLM isn't all you need, try SLM

Rohit Gupta (~rohitgr7)


9

Votes

Description:

LLMs are robust but expensive and not worth it for specific use cases. SLMs can be used for particular applications like data labeling, SQL, rag, and function calling. Talk is about some open-source and closed-source models available for each of them.

Outline:

  • Intro to LLMs
  • Why LLMs should not be used everywhere and how can you avoid them and use SLMs instead
  • Benefits of SLMs
  • State-of-the-art SLMs in various use cases

Takeaways:

  • Understanding SLM use cases
  • Insights into different models used for various purposes

Prerequisites:

  • Must have: Basics of python, transformers, and ML
  • Nice to have: LLMs, transformers

Content URLs:

Slides

Speaker Info:

I love tinkering with language models, especially Small Language Models (SLMs) in various domains. I have worked at multiple early-stage startups as an ML Research Engineer / Lead like Lightning AI (PyTorch Lightning), Mazaal AI, and ShopAdvisor. I discovered my passion for OSS when I started contributing to PyTorch Lightning and got amazing feedback from the contributors and core team and became the top 5th contributor.

I currently work as a Senior ML Engineer at Prem AI and I love to give deep technical talks to share my knowledge. I've given multiple talks at meetups like FOSS United, BangPypers, and IACIDS.

Section: Artificial Intelligence and Machine Learning
Type: Talk
Target Audience: Intermediate
Last Updated: