bridge.models.stepfun.modelling_step37

Step3.7 multimodal model package — mirrors modelling_qwen3_vl.

Re-exports the public classes consumed by the bridge / provider so they can be imported from megatron.bridge.models.stepfun.modelling_step37 as a flat namespace.

Submodules

• bridge.models.stepfun.modelling_step37.image_insert_embedding
• bridge.models.stepfun.modelling_step37.transformer_block
• bridge.models.stepfun.modelling_step37.utils
• bridge.models.stepfun.modelling_step37.vision_model
• bridge.models.stepfun.modelling_step37.text_model
• bridge.models.stepfun.modelling_step37.transformer_config
• bridge.models.stepfun.modelling_step37.model
• bridge.models.stepfun.modelling_step37.rope
• bridge.models.stepfun.modelling_step37.attention

Package Contents

Data

__all__

API

bridge.models.stepfun.modelling_step37.__all__

[‘Step37Model’, ‘Step37GPTModel’, ‘Step37TransformerConfig’, ‘Step37VisionModel’, ‘get_vision_model_…