colossalai.kernel.cuda\_native
==============================

.. automodule:: colossalai.kernel.cuda_native
   :members:


.. toctree::
   :maxdepth: 2

   colossalai.kernel.cuda_native.layer_norm
   colossalai.kernel.cuda_native.multihead_attention
   colossalai.kernel.cuda_native.scaled_softmax