Skip to main content

Module transformer

Module transformer 

Source
Expand description

Transformer model weight abstraction.

Different model architectures (Qwen3, Llama, Qwen2) implement TransformerWeights to provide a uniform weight access interface. This decouples the execution backend from the model architecture.

Structs§

TransformerConfig
Configuration for a standard transformer decoder.

Traits§

TransformerWeights
Uniform weight access for transformer decoder models.