Runtime & Serialization
========================

Configure the Torch-TensorRT runtime, serialize engines for reuse, benchmark
correctly, handle unsupported operators, and deploy on specialized hardware (DLA).

.. toctree::
   :maxdepth: 1

   runtime
   runtime_api
   using_dla
   saving_models
   serialized_engine
   aot_inductor
   mutable_module
   Example: Saving and Loading Models with Dynamic Shapes <../../tutorials/_rendered_examples/dynamo/save_dynamic_shapes_example>
   Example: Saving Models with Dynamic Shapes - Both Methods <../../tutorials/_rendered_examples/dynamo/save_dynamic_shapes_both_methods>