Most of the MLOps steps are focused on structured data. To adapt them to LLMs, the steps need to be adapted. This includes, for example, embeddings, tokenization, data cleansing, and more. In addition, steps like evaluation and monitoring are more complicated and require more innovative thinking.
For example, a drift analysis requires a histogram analysis and histogram comparison. This could require text analysis, conversion to numeric values, clustering or another type of logic to compare the training or the expected results with the actual results.
Interested in learning more?
Check out this 9 minute demo that covers MLOps best practices for generative AI applications.
View this webinar with QuantumBlack, AI by McKinsey covers the challenges of deploying and managing LLMs in live user-facing business applications.
Check out this demo and repo that demonstrates how to fine tune an LLM and build an application.