add feature gate for tensorrt plugin

[lanluo-nvidia]

Description

Tensorrt 10.3.0 does not support tensorrt plugin which causing failures.

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

[shchoi00]

+1

My environment

jetpack 6.2
torch 2.7
tensorrt 10.3
cuda 12.6

$ python3 -c "import torch_tensorrt"
Unable to import quantization op. Please install modelopt library (https://github.com/NVIDIA/TensorRT-Model-Optimizer?tab=readme-ov-file#installation) to add support for compiling quantized models
TensorRT-LLM is not installed. Please install TensorRT-LLM or set TRTLLM_PLUGINS_PATH to the directory containing libnvinfer_plugin_tensorrt_llm.so to use converters for torch.distributed ops
Traceback (most recent call last):
File "", line 1, in
File "/home/ircv13/.local/lib/python3.10/site-packages/torch_tensorrt/init.py", line 125, in
from torch_tensorrt.runtime import * # noqa: F403
File "/home/ircv13/.local/lib/python3.10/site-packages/torch_tensorrt/runtime/init.py", line 1, in
from torch_tensorrt.dynamo.runtime import ( # noqa: F401
File "/home/ircv13/.local/lib/python3.10/site-packages/torch_tensorrt/dynamo/init.py", line 10, in
from ._compiler import (
File "/home/ircv13/.local/lib/python3.10/site-packages/torch_tensorrt/dynamo/_compiler.py", line 15, in
from torch_tensorrt.dynamo import _defaults, partitioning
File "/home/ircv13/.local/lib/python3.10/site-packages/torch_tensorrt/dynamo/partitioning/init.py", line 1, in
from ._adjacency_partitioner import partition as fast_partition
File "/home/ircv13/.local/lib/python3.10/site-packages/torch_tensorrt/dynamo/partitioning/_adjacency_partitioner.py", line 20, in
from torch_tensorrt.dynamo.conversion._ConverterRegistry import (
File "/home/ircv13/.local/lib/python3.10/site-packages/torch_tensorrt/dynamo/conversion/init.py", line 1, in
from . import (
File "/home/ircv13/.local/lib/python3.10/site-packages/torch_tensorrt/dynamo/conversion/plugins/init.py", line 1, in
from torch_tensorrt.dynamo.conversion.plugins._custom_op import custom_op
File "/home/ircv13/.local/lib/python3.10/site-packages/torch_tensorrt/dynamo/conversion/plugins/_custom_op.py", line 6, in
from torch_tensorrt.dynamo.conversion.plugins._generate_plugin import generate_plugin
File "/home/ircv13/.local/lib/python3.10/site-packages/torch_tensorrt/dynamo/conversion/plugins/_generate_plugin.py", line 5, in
import tensorrt.plugin as trtp
ModuleNotFoundError: No module named 'tensorrt.plugin'

[narendasan]

@shchoi00 yes this is support of Jetson as part of our build system overhaul

[github-actions]

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

--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/plugins/_generate_plugin_converter.py	2025-05-20 15:32:36.323051+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/plugins/_generate_plugin_converter.py	2025-05-20 15:33:02.313006+00:00
@@ -36,10 +36,11 @@
    except ImportError as e:
        raise RuntimeError(
            "Unable to import TensorRT plugin. Please install TensorRT plugin library (https://github.com/NVIDIA/TensorRT-plugin-library?tab=readme-ov-file#installation) to add support for compiling quantized models"
        )
    from tensorrt.plugin._lib import QDP_REGISTRY
+
    torch_target = getattr(getattr(torch.ops, namespace), op_name)
    overload_str = overload if overload else ""
    overload_name = overload_str if overload else "default"
    torch_overload = getattr(torch_target, overload_name)
    assert (

[narendasan]

Minor change then looks good