🐛 [Bug] Incorrect Documentation page

[Skier23]

This page has several issues with the provided examples:
https://pytorch.org/TensorRT/user_guide/saving_models.html

1. In the examples, an optional parameter is passed before a non-optional parameter which is invalid python. This issue also appears here:

   TensorRT/docsrc/user_guide/saving_models.rst

   Line 20 in 7f14221
2. The a) example does not work because the default from torch_tensorrt.compile is to use an output_format of ExportedProgram.
3. This line is no longer valid: trt_exp_program = torch_tensorrt.dynamo.export(trt_gm, images)
4. torch.jit.trace should not be used when specifying the correct output format. The corrected code should be as follows:
   `
   model = MyModel().eval().cuda()
   inputs = [torch.randn((1, 3, 224, 224)).cuda()]
   trt_gm = torch_tensorrt.compile(model, ir="dynamo", inputs=inputs, output_format="torchscript") # Output is a torch.fx.GraphModule
   torch.jit.save(trt_gm, "trt_model.ts")

Later, you can load it and run inference

model = torch.jit.load("trt_model.ts").cuda()
model(*inputs)
`

[mjain7]

@Skier23 @peri044
I get below error when try to save tensorrt compiled model

Traceback (most recent call last):
  File "/scratch_space/model.py", line 10, in <module>
    torch.jit.save(trt_gm, "trt_model.ts")
  File "/usr/local/lib/python3.10/dist-packages/torch/jit/_serialization.py", line 83, in save
    m.save(f, _extra_files=_extra_files)
  File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1893, in __getattr__
    raise AttributeError(
AttributeError: 'GraphModule' object has no attribute 'save'

can you please help? (I am running inside nvcr.io/nvidia/pytorch:24.08-py3 container)

Also, how to load trt_model.ts on triton inference server? With platform: "pytorch_libtorch" in config.pbtxt config file, it expects model file to be in .pt format in model_repository. Is there a way to save tensorrt compiled file in .pt format?