Add support for TensorRT-RTX

[lanluo-nvidia]

Description

Add initial support for TensorRT-RTX.
The following are the currently identified issues:
RTX team side:

1. mobilenet/efficientnet test failed when input dtype is bfloat16(float16 and float32 is working)
   5439176
2. Flux model FP4 and FP8 are not working (fp16 and fp32 is working)
   5400490
3. NoneZero has not been supported yet in RTX(will add support in future)
   5407733
4. Int8 quantization only support weights quantization, does not support activations quantization yet.
   5402295
5. deconv3d is not working in rtx
   5459328
6. cuda graph is not working in rtx
   5481821

Our side:

1. a few test cases failed on strong typing
   PR is in Progress:
   fix: atan2 strong type support & bug fix for integer dynamic shape #3751
   add strong typing fix #3749
2. batchnorm constant fold got nan (skip constant fold is able to get the result)
   🐛 [Bug] TensorRT-RTX BatchNorm constant fold got nan #3699
3. refit test failed on missing eps weights due to we only capture (bias/weight/running_mean/running_var)
   🐛 [Bug] TensorRT-RTX Refitter test failed when constant fold is disabled #3752
4. cuda graph test failed
   🐛 [Bug] TensorRT-RTX: Cuda graph test failed #3781
5. torch script test failed
   🐛 [Bug] TensorRT - RTX: torch-script test failure #3782

Fixes # (issue)

Type of change

Please delete options that are not relevant and/or add your own.

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

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

[pytorch-bot]

No ciflow labels are configured for this repo.
For information on how to enable CIFlow bot see this wiki

[narendasan]

@lanluo-nvidia just remove the PTQ Calibrator feature from python and C++ and put in deprecation errors.

[github-actions]

There are some changes that do not conform to Python style guidelines:

--- /home/runner/work/TensorRT/TensorRT/tests/py/ts/integrations/test_trt_intercompatibility.py	2025-08-27 00:39:48.227435+00:00
+++ /home/runner/work/TensorRT/TensorRT/tests/py/ts/integrations/test_trt_intercompatibility.py	2025-08-27 00:40:20.832194+00:00
@@ -34,10 +34,11 @@

        trt_engine = torchtrt.ts.convert_method_to_trt_engine(
            self.ts_model, "forward", **compile_spec
        )
        import tensorrt as trt
+
        TRT_LOGGER = trt.Logger(trt.Logger.WARNING)
        with trt.Runtime(TRT_LOGGER) as rt:
            engine = rt.deserialize_cuda_engine(trt_engine)
            with engine.create_execution_context() as ctx:
                out = torch.empty(

[narendasan]

Can we write a validator for this case? (As an aside have we tried quantization where not all of the model is running in TensorRT?)

[narendasan]

Could just stack the decorator here

[narendasan]

I think

@needs_tensorrt_rtx
@dynamo_tensorrt_converter(...)
def aten...

Should do what we want

[narendasan]

Just the decorator then you are good to merge