Conversation

peri044

Description

Please include a summary of the change and which issue is fixed. Please also include relevant motivation and context. List any dependencies that are required for this change.

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

@github-actions-actions bot added component: loweringIssues re: The lowering / preprocessing passescomponent: api [Python]Issues re: Python APIcomponent: dynamoIssues relating to the `torch.compile` or `torch._dynamo.export` pathslabels Aug 6, 2024
@github-actions-actions bot requested a review from narendasan August 6, 2024 22:43
@github-actions-actions bot added component: testsIssues re: Testscomponent: conversionIssues re: Conversion stagecomponent: convertersIssues re: Specific op converterscomponent: torch_compile labels Aug 17, 2024
github-actions[bot]
-actions bot requested changes Aug 17, 2024
View reviewed changes

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

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

--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/impl/pool.py	2024-08-17 21:38:03.174815+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/impl/pool.py	2024-08-17 21:38:24.979987+00:00
@@ -28,11 +28,11 @@
    padding: Union[int, Sequence[int]] = 0,
    ceil_mode: bool = False,
    count_include_pad: bool = True,
    divisor_override: Optional[int] = None,
) -> TRTTensor:
-    
+
    padding_mode = trt.PaddingMode.EXPLICIT_ROUND_DOWN
    if ceil_mode:
        padding_mode = trt.PaddingMode.EXPLICIT_ROUND_UP

    if divisor_override is not None:
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/lowering/passes/replace_full_like_with_full.py	2024-08-17 21:38:03.178815+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/lowering/passes/replace_full_like_with_full.py	2024-08-17 21:38:25.214966+00:00
@@ -22,11 +22,11 @@

            # Extract arguments from full_like
            input_tensor = node.args[0]
            fill_value = node.args[1]
            shape = list(input_tensor.meta["tensor_meta"].shape)
- 
+
            new_kwargs = {}
            for key, val in node.kwargs.items():
                if key != "memory_format":
                    new_kwargs[key] = val
            # Replace full_like with full, using the shape as a list
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/aten_ops_converters.py	2024-08-17 21:38:03.174815+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/aten_ops_converters.py	2024-08-17 21:38:27.095553+00:00
@@ -2699,11 +2699,11 @@
    dilation = args_bounds_check(pool_node.args, 4, 1)
    ceil_mode = args_bounds_check(pool_node.args, 5, False)

    if not isinstance(dilation, (list, tuple)):
        dilation = (dilation,)
-    
+
    for dil in dilation:
        if dil != 1:
            _LOGGER.debug("Currently we don't support dilation > 1 at any dimension.")
            return False

@github-actions-actions bot added the component: build systemIssues re: Build systemlabel Aug 18, 2024
github-actions[bot]
-actions bot requested changes Aug 18, 2024
View reviewed changes

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

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

--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/impl/full.py	2024-08-18 03:35:59.991813+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/impl/full.py	2024-08-18 03:36:20.489931+00:00
@@ -20,11 +20,11 @@
    target: Union[Target, str],
    source_ir: Optional[SourceIR],
    name: str,
    shape: Union[List[int], TRTTensor],
    fill_value: Union[int, float, bool],
-    dtype: Union[torch.dtype, trt.DataType]
+    dtype: Union[torch.dtype, trt.DataType],
) -> TRTTensor:
    output_dtype = _enums.dtype._from(dtype)
    if isinstance(shape, List):
        # in static shape scenario, shape is a list of int
        if all(isinstance(dim, int) for dim in shape):
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/lowering/passes/replace_full_like_with_full.py	2024-08-18 03:35:59.995813+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/lowering/passes/replace_full_like_with_full.py	2024-08-18 03:36:21.060814+00:00
@@ -34,11 +34,11 @@
                input_dtype = input_tensor.meta["tensor_meta"].dtype
                input_device = input_tensor.meta["tensor_meta"].device

            shape = list(input_tensor.meta["tensor_meta"].shape)

-            # There's no memory format argument for torch.full. 
+            # There's no memory format argument for torch.full.
            # Set the input_device and dtype correspondingly.
            new_kwargs = {}
            for key, val in node.kwargs.items():
                if key != "memory_format":
                    new_kwargs[key] = val
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/aten_ops_converters.py	2024-08-18 03:35:59.991813+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/aten_ops_converters.py	2024-08-18 03:36:22.843055+00:00
@@ -2702,11 +2702,11 @@

    for dil in dilation:
        if dil != 1:
            _LOGGER.debug("Currently we don't support dilation > 1 at any dimension.")
            return False
-    
+
    return True


# Note: MaxPool1d uses max_pool2d as it converts to 2D first.
@dynamo_tensorrt_converter(
@@ -3856,7 +3856,7 @@
        target,
        SourceIR.ATEN,
        name,
        shape=args[0],
        fill_value=args[1],
-        dtype=kwargs["dtype"]
-    )
+        dtype=kwargs["dtype"],
+    )

@peri044peri044 changed the title chore: bug fix chore: bug fixes Aug 18, 2024
github-actions[bot]
-actions bot requested changes Aug 18, 2024
View reviewed changes

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

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

--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/impl/full.py	2024-08-18 06:00:03.147977+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/impl/full.py	2024-08-18 06:00:23.686044+00:00
@@ -20,11 +20,11 @@
    target: Union[Target, str],
    source_ir: Optional[SourceIR],
    name: str,
    shape: Union[List[int], TRTTensor],
    fill_value: Union[int, float, bool],
-    dtype: Union[torch.dtype, trt.DataType]
+    dtype: Union[torch.dtype, trt.DataType],
) -> TRTTensor:
    output_dtype = _enums.dtype._from(dtype)
    if isinstance(shape, List):
        # in static shape scenario, shape is a list of int
        if all(isinstance(dim, int) for dim in shape):
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/lowering/passes/replace_full_like_with_full.py	2024-08-18 06:00:03.151977+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/lowering/passes/replace_full_like_with_full.py	2024-08-18 06:00:24.276030+00:00
@@ -34,11 +34,11 @@
                input_dtype = input_tensor.meta["tensor_meta"].dtype
                input_device = input_tensor.meta["tensor_meta"].device

            shape = list(input_tensor.meta["tensor_meta"].shape)

-            # There's no memory format argument for torch.full. 
+            # There's no memory format argument for torch.full.
            # Set the input_device and dtype correspondingly.
            new_kwargs = {}
            for key, val in node.kwargs.items():
                if key != "memory_format":
                    new_kwargs[key] = val
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/aten_ops_converters.py	2024-08-18 06:00:03.147977+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/conversion/aten_ops_converters.py	2024-08-18 06:00:25.976231+00:00
@@ -2694,11 +2694,11 @@

    for dil in dilation:
        if dil != 1:
            _LOGGER.debug("Currently we don't support dilation > 1 at any dimension.")
            return False
-    
+
    return True


# Note: MaxPool1d uses max_pool2d as it converts to 2D first.
@dynamo_tensorrt_converter(
@@ -3848,7 +3848,7 @@
        target,
        SourceIR.ATEN,
        name,
        shape=args[0],
        fill_value=args[1],
-        dtype=kwargs["dtype"]
-    )
+        dtype=kwargs["dtype"],
+    )
--- /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-18 06:00:03.175977+00:00
+++ /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-18 06:00:28.087506+00:00
@@ -73,11 +73,13 @@
                    count_include_pad,
                    divisor_override,
                )

        inputs = [torch.randn(1, 3, 32, 32)]
-        self.run_test(TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True)
+        self.run_test(
+            TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True
+        )

    @parameterized.expand(
        [
            (3, 1, 0),
            (3, 1, 1),
@@ -181,11 +183,11 @@
                (3, 3, 3, 3),
                torch.float,
                (3, 3),
                (1, 1),
                (1, 1),
-                True
+                True,
            ),
        ]
    )
    def test_dynamic_shape_pool2d(
        self,

github-actions[bot]
-actions bot requested changes Aug 21, 2024
View reviewed changes

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

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

--- /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-21 00:24:16.805992+00:00
+++ /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-21 00:24:50.723758+00:00
@@ -73,11 +73,13 @@
                    count_include_pad,
                    divisor_override,
                )

        inputs = [torch.randn(1, 3, 32, 32)]
-        self.run_test(TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True)
+        self.run_test(
+            TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True
+        )

    @parameterized.expand(
        [
            (3, 1, 0),
            (3, 1, 1),
@@ -181,11 +183,11 @@
                (3, 3, 3, 3),
                torch.float,
                (3, 3),
                (1, 1),
                (1, 1),
-                True
+                True,
            ),
        ]
    )
    def test_dynamic_shape_pool2d(
        self,

github-actions[bot]
-actions bot requested changes Aug 21, 2024
View reviewed changes

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

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

--- /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-21 00:30:09.776031+00:00
+++ /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-21 00:30:45.350177+00:00
@@ -73,11 +73,13 @@
                    count_include_pad,
                    divisor_override,
                )

        inputs = [torch.randn(1, 3, 32, 32)]
-        self.run_test(TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True)
+        self.run_test(
+            TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True
+        )

    @parameterized.expand(
        [
            (3, 1, 0),
            (3, 1, 1),
@@ -181,11 +183,11 @@
                (3, 3, 3, 3),
                torch.float,
                (3, 3),
                (1, 1),
                (1, 1),
-                True
+                True,
            ),
        ]
    )
    def test_dynamic_shape_pool2d(
        self,

github-actions[bot]
-actions bot requested changes Aug 23, 2024
View reviewed changes

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

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

--- /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-23 03:05:32.567575+00:00
+++ /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-23 03:05:56.052071+00:00
@@ -73,11 +73,13 @@
                    count_include_pad,
                    divisor_override,
                )

        inputs = [torch.randn(1, 3, 32, 32)]
-        self.run_test(TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True)
+        self.run_test(
+            TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True
+        )

    @parameterized.expand(
        [
            (3, 1, 0),
            (3, 1, 1),
@@ -181,11 +183,11 @@
                (3, 3, 3, 3),
                torch.float,
                (3, 3),
                (1, 1),
                (1, 1),
-                True
+                True,
            ),
        ]
    )
    def test_dynamic_shape_pool2d(
        self,

github-actions[bot]
-actions bot requested changes Aug 28, 2024
View reviewed changes

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

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

--- /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-28 21:23:05.146803+00:00
+++ /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-28 21:23:31.914698+00:00
@@ -73,11 +73,13 @@
                    count_include_pad,
                    divisor_override,
                )

        inputs = [torch.randn(1, 3, 32, 32)]
-        self.run_test(TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True)
+        self.run_test(
+            TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True
+        )

    @parameterized.expand(
        [
            (3, 1, 0),
            (3, 1, 1),
@@ -181,11 +183,11 @@
                (3, 3, 3, 3),
                torch.float,
                (3, 3),
                (1, 1),
                (1, 1),
-                True
+                True,
            ),
        ]
    )
    def test_dynamic_shape_pool2d(
        self,

github-actions[bot]
-actions bot requested changes Aug 28, 2024
View reviewed changes

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

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

--- /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-28 23:35:05.049467+00:00
+++ /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-28 23:35:28.258183+00:00
@@ -73,11 +73,13 @@
                    count_include_pad,
                    divisor_override,
                )

        inputs = [torch.randn(1, 3, 32, 32)]
-        self.run_test(TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True)
+        self.run_test(
+            TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True
+        )

    @parameterized.expand(
        [
            (3, 1, 0),
            (3, 1, 1),
@@ -181,11 +183,11 @@
                (3, 3, 3, 3),
                torch.float,
                (3, 3),
                (1, 1),
                (1, 1),
-                True
+                True,
            ),
        ]
    )
    def test_dynamic_shape_pool2d(
        self,

@github-actions-actions bot removed the component: loweringIssues re: The lowering / preprocessing passeslabel Aug 29, 2024
github-actions[bot]
-actions bot requested changes Aug 29, 2024
View reviewed changes

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

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

--- /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-29 17:37:32.829319+00:00
+++ /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-29 17:37:56.926341+00:00
@@ -73,11 +73,13 @@
                    count_include_pad,
                    divisor_override,
                )

        inputs = [torch.randn(1, 3, 32, 32)]
-        self.run_test(TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True)
+        self.run_test(
+            TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True
+        )

    @parameterized.expand(
        [
            (3, 1, 0),
            (3, 1, 1),
@@ -181,11 +183,11 @@
                (3, 3, 3, 3),
                torch.float,
                (3, 3),
                (1, 1),
                (1, 1),
-                True
+                True,
            ),
        ]
    )
    def test_dynamic_shape_pool2d(
        self,

github-actions[bot]
-actions bot requested changes Aug 29, 2024
View reviewed changes

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

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

--- /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-29 23:56:13.538224+00:00
+++ /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-08-29 23:56:39.163776+00:00
@@ -73,11 +73,13 @@
                    count_include_pad,
                    divisor_override,
                )

        inputs = [torch.randn(1, 3, 32, 32)]
-        self.run_test(TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True)
+        self.run_test(
+            TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True
+        )

    @parameterized.expand(
        [
            (3, 1, 0),
            (3, 1, 1),
@@ -181,11 +183,11 @@
                (3, 3, 3, 3),
                torch.float,
                (3, 3),
                (1, 1),
                (1, 1),
-                True
+                True,
            ),
        ]
    )
    def test_dynamic_shape_pool2d(
        self,

github-actions[bot]
-actions bot requested changes Sep 10, 2024
View reviewed changes

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

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

--- /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-09-10 18:44:52.655743+00:00
+++ /home/runner/work/TensorRT/TensorRT/tests/py/dynamo/conversion/test_pool_aten.py	2024-09-10 18:45:27.216624+00:00
@@ -73,11 +73,13 @@
                    count_include_pad,
                    divisor_override,
                )

        inputs = [torch.randn(1, 3, 32, 32)]
-        self.run_test(TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True)
+        self.run_test(
+            TestModule(), inputs, rtol=5e-03, atol=5e-03, use_dynamo_tracer=True
+        )

    @parameterized.expand(
        [
            (3, 1, 0),
            (3, 1, 1),
@@ -181,11 +183,11 @@
                (3, 3, 3, 3),
                torch.float,
                (3, 3),
                (1, 1),
                (1, 1),
-                True
+                True,
            ),
        ]
    )
    def test_dynamic_shape_pool2d(
        self,

@peri044peri044 changed the title chore: bug fixes chore: bug fixes for padding in pooling layer Mar 13, 2025
@peri044peri044 changed the title chore: bug fixes for padding in pooling layer chore: bug fixes for padding mode in pooling layer Mar 13, 2025
Sign up for free to join this conversation on . Already have an account? Sign in to comment

@github-actions -actions[bot] -actions[bot] requested changes

@narendasan narendasan Awaiting requested review from narendasan

No one assigned
cla signedIssues re: Python APIcomponent: conversionIssues re: Conversion stagecomponent: convertersIssues re: Specific op converterscomponent: dynamoIssues relating to the `torch.compile` or `torch._dynamo.export` pathscomponent: testsIssues re: Tests
None yet
No milestone

Successfully merging this pull request may close these issues.