Concepts

Core areas of the AMD AI Workbench UI and how they fit together for model lifecycle work.

• Dashboard — Project and utilization summary

• Model catalog — Discover, deploy, and fine-tune AIMs

• Onboard custom models — Bring Hugging Face models into a project

• Deploy custom models — Deploy ready custom models as inference services

• Workspaces — JupyterLab, VS Code, and GPU-backed dev environments

• Training and fine-tuning — Fine-tune base models with guided flows

• Inference — Deploy models, chat, and compare outputs

• API keys — Programmatic access to deployments

• Keycloak authentication — OAuth2 bearer tokens for management API access

• Session lifetime — Token refresh and session expiration

• API reference — AI Workbench management API endpoints

• Secrets — Tokens and credentials for models and tools