Added dynamo-torchao test suite under the full CI lane

[micwill755]

Description

Adds a dedicated dynamo-torchao suite on the full and nightly lanes so TorchAO compile tests are their own CI job instead of mixed into dynamo-models.

tests/py/dynamo/models/test_torchao*.py is auto-marked torchao. dynamo-models is now -m "not critical and not torchao" so those tests do not double-run. linux-test installs torchao after the pinned torch reinstall so the suite actually executes (tests skip if the extra is missing).

Locally: cd tests && TRT_PYTEST_RERUNS=0 python3 -m ci run dynamo-torchao --variant standard (5 passed: FP8 WOQ, static FP8, INT4 WOQ, NVFP4 WOQ, MXFP4).

Stacked on #4595 (MXFP4) → #4593 (NVFP4) → #4590 (INT4) → #4589 (FP8). Stack: https://github.com/pytorch/TensorRT/stack/4591

Fixes # (issue)

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
• dynamo-torchao 5 passed locally
• 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

[github-actions]

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

--- /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_cumsum_aten.py	2026-08-26 23:57:45.614516+00:00
+++ /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_cumsum_aten.py	2026-08-26 23:58:16.228472+00:00
@@ -31,13 +31,11 @@
                    immutable_weights=False,
                    use_dynamo_tracer=True,
                )
            return

-        self.run_test(
-            Cumsum(), inputs, immutable_weights=False, use_dynamo_tracer=True
-        )
+        self.run_test(Cumsum(), inputs, immutable_weights=False, use_dynamo_tracer=True)

    @parameterized.expand(
        [
            ((3, 1), 0),
            ((3, 1), 1),
@@ -106,14 +104,11 @@
        has_static_trip_count = (
            min_shape[positive_dim]
            == opt_shape[positive_dim]
            == max_shape[positive_dim]
        )
-        if (
-            has_static_trip_count
-            and not is_tensorrt_rtx_version_supported("1.7")
-        ):
+        if has_static_trip_count and not is_tensorrt_rtx_version_supported("1.7"):
            with self.assertRaises(UnsupportedOperatorException):
                self.run_test_with_dynamic_shape(
                    Cumsum(),
                    inputs,
                    immutable_weights=False,

[github-actions]

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

--- /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_cumsum_aten.py	2026-08-26 23:58:31.775617+00:00
+++ /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_cumsum_aten.py	2026-08-26 23:58:57.173559+00:00
@@ -31,13 +31,11 @@
                    immutable_weights=False,
                    use_dynamo_tracer=True,
                )
            return

-        self.run_test(
-            Cumsum(), inputs, immutable_weights=False, use_dynamo_tracer=True
-        )
+        self.run_test(Cumsum(), inputs, immutable_weights=False, use_dynamo_tracer=True)

    @parameterized.expand(
        [
            ((3, 1), 0),
            ((3, 1), 1),
@@ -106,14 +104,11 @@
        has_static_trip_count = (
            min_shape[positive_dim]
            == opt_shape[positive_dim]
            == max_shape[positive_dim]
        )
-        if (
-            has_static_trip_count
-            and not is_tensorrt_rtx_version_supported("1.7")
-        ):
+        if has_static_trip_count and not is_tensorrt_rtx_version_supported("1.7"):
            with self.assertRaises(UnsupportedOperatorException):
                self.run_test_with_dynamic_shape(
                    Cumsum(),
                    inputs,
                    immutable_weights=False,