bridge.models.mimo.mimo_bridge#

Module Contents#

Classes#

MimoBridge

Megatron Bridge for MiMo Causal LM.

API#

class bridge.models.mimo.mimo_bridge.MimoBridge#

Bases: megatron.bridge.models.qwen.qwen2_bridge.Qwen2Bridge

Megatron Bridge for MiMo Causal LM.

hf_config_to_model_config_kwargs(
hf_config: transformers.PretrainedConfig,
) → dict[str, Any]#

Convert a Hugging Face MiMo config to builder config kwargs.

mapping_registry() → megatron.bridge.models.conversion.mapping_registry.MegatronMappingRegistry#
static _swap_input_proj_halves(weight: torch.Tensor) → torch.Tensor#
maybe_modify_loaded_hf_weight(
hf_param: str | dict[str, str],
hf_state_dict: Mapping[str, torch.Tensor],
) → torch.Tensor#
maybe_modify_converted_hf_weight(
task: megatron.bridge.models.conversion.model_bridge.WeightConversionTask,
converted_weights_dict: dict[str, torch.Tensor],
hf_state_dict: Mapping[str, torch.Tensor],
) → dict[str, torch.Tensor]#