transformer_lens.model_bridge.supported_architectures.mbart module¶
MBart (MBartForConditionalGeneration) adapter: Bart layout but PRE-norm with
an extra final LayerNorm per stack; embedding scale is baked in (embed hooks see it).
- class transformer_lens.model_bridge.supported_architectures.mbart.MBartArchitectureAdapter(cfg: Any)¶
Bases:
BartFamilyArchitectureAdapterArchitecture adapter for MBartForConditionalGeneration models.
- component_mapping: ComponentMapping | None¶
- has_final_stack_norm: bool = True¶
- has_layernorm_embedding: bool = True¶
- uses_split_attention: bool¶
- weight_processing_conversions: Dict[str, ParamProcessingConversion | str] | None¶