add-inference-model

What problem does it solve?

Onboarding a new core pre-trained model into the inference repos is slow and error-prone without a standardized playbook.

Core Features & Use Cases

• End-to-end scaffolding for new core model implementations, including class creation, registry wiring, and tests.
• Weight packaging and registry wiring covering zip creation and model package submission to the registry.
• Optional surfaces support for workflows (foundation blocks) and adapters (inference endpoints) when needed.
• Templates and guidance to ensure consistent model onboarding across architectures and backends.

Quick Start

Begin by defining the architecture name, task type, backend, and weight source, then follow the scaffolding to implement surfaces 1–4 as needed.