-
Notifications
You must be signed in to change notification settings - Fork 10
Closed
Labels
enhancementNew feature or requestNew feature or request
Description
We need a base class that all the models will subclass. This class help us provide a unified interface for all the models and reduce boilerplate code.
- Class methods that loads / deserialize weights from different formats (safetensors first, GGUF later on)
- A method to initialize a new instance from config
- a method to save to orbax checkpoint
- Properties (read-only class attributes) for config values such as
num_heads,head_dimetc.
Metadata
Metadata
Assignees
Labels
enhancementNew feature or requestNew feature or request