From MLOops to MLOps: LLMs & Stable Diffusion in the Cloud

Name: From%20MLOops%20to%20MLOps:%20LLMs%20&%20Stable%20Diffusion%20in%20the%20Cloud
Uploaded: 2023-08-07

Posted Aug 07, 2023 | Views 508

# MLOops

# LLMs & Stable Diffusion

# Graphcore

# graphcore.ai

speaker

Alex Payot

AI Engineer @ Graphcore

Alex works in Graphcore’s AI Engineering team developing and supporting ML applications on the Graphcore IPU from training to production. Most recently Alex has been working with Paperspace to make ML models available on IPUs in the Cloud with minimal friction.

Before working at Graphcore you might have found him developing software for the design and optimisation of steered carbon fibre composites, doing data analytics and computer vision for microscopes, or filling up the super-computer at Bristol University to get a Ph.D. in aerodynamic optimisation.

+ Read More

SUMMARY

Going from a model “working“ on your machine to a model doing something useful for a customer in production is hard.

In this talk, Alex describes his team’s journey towards making 30 models available in the cloud on Paperspace, ready for users to experiment with and deploy. Expect real accounts of some of the more surprising MLOps challenges they encountered, and hopefully, some useful lessons that you won’t have to learn the hard way.

As part of the talk, Alex will show how you can use IPUs in Paperspace to run a “voice-to-comic-strip” demo with Whisper, LLMs, and Stable Diffusion.

+ Read More