--- license: mit tags: - batchnorm - bilinear - exponential - gelu-tanh - linear - matching - mocov3 - nano - novograd - trunc-normal --- # pipeline.py ## Model Overview A **nano**-scale implementation of the **mocov3** architecture, built for **matching** tasks. ## Architecture - **Architecture**: mocov3 - **Scale**: nano - **Attention**: linear - **Fusion strategy**: bilinear - **Task head**: matching - **Activation**: gelu tanh - **Normalization**: batchnorm - **Initialization**: trunc normal ## Training - **Optimizer**: novograd - **LR scheduler**: exponential ## Files - `pipeline.py` — main artifact of this repository ## License See the license field above.