From the course: GenAIOps Foundations
Unlock this course with a free trial
Join today to access over 24,800 courses taught by industry experts.
Generative AI automation pipelines
From the course: GenAIOps Foundations
Generative AI automation pipelines
- [Instructor] Let's now discuss automated evaluation and fine-tuning pipelines for GenAI. Automation provides repeatability. It is a key tool in DevOps, MLOps and in GenAIOps. Building automated pipelines for evaluation and fine-tuning ensures repeatability of experiments. It also improves efficiency when iterating over experiments. It helps track the history of experiments and compare results reliably. What does a typical GenAI evaluation and fine-tuning pipeline look like? For both evaluation and fine-tuning, we need a curated dataset as discussed in the previous video. Evaluation and fine-tuning can both use the same dataset, or the dataset can be split for separate users. Fine-tuning is done by a script. This script can perform several fine-tuning functions like parameter tuning, hyper parameter tuning, or parameter efficient tuning. This can be run in a managed service like Kubernetes. The fine-tuning script picks up a base model from a model registry. The model registry stores…