From cc17e5c5c67466351ab24609c6e98e549859b0f7 Mon Sep 17 00:00:00 2001
From: Joseph Loftin <jloftin@nvidia.com>
Date: Tue, 18 Aug 2026 21:43:43 +0000
Subject: [PATCH] Observable dry run

---
 py/torch_tensorrt/dynamo/_DryRunTracker.py    |  2 +
 .../runtime/test_dryrun_stats_observable.py   | 48 +++++++++++++++++++
 2 files changed, 50 insertions(+)
 create mode 100644 tests/py/dynamo/runtime/test_dryrun_stats_observable.py

diff --git a/py/torch_tensorrt/dynamo/_DryRunTracker.py b/py/torch_tensorrt/dynamo/_DryRunTracker.py
index 43789c4a0f..2007287b03 100644
--- a/py/torch_tensorrt/dynamo/_DryRunTracker.py
+++ b/py/torch_tensorrt/dynamo/_DryRunTracker.py
@@ -9,6 +9,7 @@
 from torch_tensorrt.dynamo._settings import CompilationSettings
 from torch_tensorrt.dynamo.conversion._ConverterRegistry import ConverterRegistry
 from torch_tensorrt.dynamo.conversion.converter_utils import get_node_name
+from torch_tensorrt.dynamo.observer import observable
 
 logger = logging.getLogger(__name__)
 
@@ -67,6 +68,7 @@ class DryRunTracker:
     to_run_in_torch: List[str] = field(default_factory=list)
 
 
+@observable()
 def dryrun_stats_display(
     dryrun_tracker: DryRunTracker, dryrun_enabled: Union[bool, str]
 ) -> None:
diff --git a/tests/py/dynamo/runtime/test_dryrun_stats_observable.py b/tests/py/dynamo/runtime/test_dryrun_stats_observable.py
new file mode 100644
index 0000000000..39d485f788
--- /dev/null
+++ b/tests/py/dynamo/runtime/test_dryrun_stats_observable.py
@@ -0,0 +1,48 @@
+# type: ignore
+
+import unittest
+
+import torch
+import torch_tensorrt
+from torch.testing._internal.common_utils import TestCase, run_tests
+from torch_tensorrt.dynamo._DryRunTracker import DryRunTracker, dryrun_stats_display
+from torch_tensorrt.dynamo.observer import ObserveContext
+
+
+@unittest.skipIf(not torch.cuda.is_available(), "CUDA required")
+class TestDryRunStatsObservable(TestCase):
+    def test_observer_captures_tracker(self):
+        class Add(torch.nn.Module):
+            def forward(self, x):
+                return x + x
+
+        model = Add().cuda().eval()
+        x = torch.randn(2, 3, device="cuda")
+        trackers = []
+
+        def capture(ctx: ObserveContext) -> None:
+            trackers.append(ctx.args[0])
+
+        with dryrun_stats_display.observers.pre.add(capture):
+            torch_tensorrt.dynamo.compile(
+                model,
+                [x],
+                dryrun=True,
+                min_block_size=1,
+                enabled_precisions={torch.float32},
+            )
+
+        self.assertEqual(len(trackers), 1)
+        tracker = trackers[0]
+        self.assertIsInstance(tracker, DryRunTracker)
+        self.assertGreater(tracker.total_ops_in_graph, 0)
+        self.assertGreaterEqual(tracker.supported_ops_in_graph, 0)
+        self.assertLessEqual(
+            tracker.supported_ops_in_graph, tracker.total_ops_in_graph
+        )
+        self.assertGreaterEqual(tracker.tensorrt_graph_count, 1)
+        self.assertEqual(len(tracker.per_subgraph_data), tracker.tensorrt_graph_count)
+
+
+if __name__ == "__main__":
+    run_tests()