feat: add convert_method_to_trt_engine() for dynamo

[zewenli98]

Description

Add a function convert_method_to_trt_engine() for dynamo.

I'm not familiar with this part, so detailed reviews and suggestions are greatly welcome.

Fixes #2410

Type of change

• New feature (non-breaking change which adds functionality)

Checklist:

• My code follows the style guidelines of this project (You can use the linters)
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas and hacks
• I have made corresponding changes to the documentation
• I have added tests to verify my fix or my feature
• New and existing unit tests pass locally with my changes
• I have added the relevant labels to my PR in so that relevant reviewers are notified

[zewenli98]

@gs-olive @narendasan Another question - Regarding testing, how to invoke this function like directly calling torchtrt.dynamo.convert_method_to_trt_engine? and where should I put the test cases?

[gs-olive]

I think it could potentially go in tests/py/dynamo/runtime? Then, you can directly call the function to generate the TRT engine which can be run directly in Python for inference purposes.

[zewenli98]

@gs-olive Got a Python package import error:

==================================== ERRORS ====================================
_ ERROR collecting tests/py/dynamo/runtime/test_convert_method_to_trt_engine.py _
ImportError while importing test module '/__w/TensorRT/TensorRT/pytorch/tensorrt/tests/py/dynamo/runtime/test_convert_method_to_trt_engine.py'.
Hint: make sure your test modules/packages have valid Python names.
Traceback:
/opt/python/cp38-cp38/lib/python3.8/importlib/__init__.py:127: in import_module
    return _bootstrap._gcd_import(name[level:], package, level)
runtime/test_convert_method_to_trt_engine.py:4: in <module>
    import pycuda.driver as cuda
E   ModuleNotFoundError: No module named 'pycuda'

Can we add pycuda in CI?

[gs-olive]

Hi @zewenli98 - sure, I think we can add this dependency at the path tests/modules/requirements.txt, and it should be reflected in the CI afterward.

[zewenli98]

Hi @zewenli98 - sure, I think we can add this dependency at the path tests/modules/requirements.txt, and it should be reflected in the CI afterward.

Thanks! Naren suggested that PythonTorchTensorRTModule should be used instead of directly calling native functions, so I modified accordingly.

[gs-olive]

Overall looks good. Added one suggestion and needs a rebase.