Revert "Fix torch.full with dynamic tensor fill_value in torch.compile (#166554)"

This reverts commit 32066772b3.

Reverted https://github.com/pytorch/pytorch/pull/166554 on behalf of https://github.com/atalman due to Failure: test/nn/test_pooling.py::TestPoolingNNDeviceTypeCPU::test_max_pool_nan_inf_cpu_float32 [GH job link](https://github.com/pytorch/pytorch/actions/runs/18959368975/job/54144148546) [HUD commit link](32066772b3) ([comment](https://github.com/pytorch/pytorch/pull/166554#issuecomment-3472976911))

test/dynamo/test_functions.py

@@ -5216,63 +5216,6 @@ class DefaultsTests(torch._dynamo.test_case.TestCase):
         x = torch.randn(1)
         self.assertEqual(opt_mod(x), x + 1)
 
-    def test_full_with_tensor_fill_value(self):
-        """Test that torch.full works correctly with dynamic tensor fill_value"""
-
-        # Test with tensor fill_value (the bug case)
-        def func_tensor(x):
-            return torch.full((2,), x, dtype=torch.float64)
-
-        func_compiled = torch.compile(func_tensor)
-
-        # Test with different values
-        x1 = torch.tensor(5.0, dtype=torch.float64)
-        x2 = torch.tensor(10.0, dtype=torch.float64)
-
-        result1 = func_compiled(x1)
-        expected1 = torch.full((2,), x1, dtype=torch.float64)
-        self.assertEqual(result1, expected1)
-
-        # This is where the bug occurred - second call reused first value
-        result2 = func_compiled(x2)
-        expected2 = torch.full((2,), x2, dtype=torch.float64)
-        self.assertEqual(result2, expected2)
-
-        # Test with different dtypes
-        for dtype in [torch.float32, torch.float64, torch.int32, torch.int64]:
-
-            def func_typed(x):
-                return torch.full((3,), x, dtype=dtype)
-
-            func_typed_compiled = torch.compile(func_typed)
-            x_typed = torch.tensor(7, dtype=dtype)
-            result = func_typed_compiled(x_typed)
-            expected = torch.full((3,), x_typed, dtype=dtype)
-            self.assertEqual(result, expected)
-
-        # Test with non-tensor fill_value (scalar) to ensure we didn't break existing behavior
-        def func_scalar(size):
-            return torch.full((size,), 42.0, dtype=torch.float32)
-
-        func_scalar_compiled = torch.compile(func_scalar)
-
-        result_scalar = func_scalar_compiled(5)
-        expected_scalar = torch.full((5,), 42.0, dtype=torch.float32)
-        self.assertEqual(result_scalar, expected_scalar)
-
-        # Test with different scalar values
-        def func_scalar_param():
-            # Test multiple calls with different hardcoded scalar values
-            a = torch.full((2,), 3.14, dtype=torch.float32)
-            b = torch.full((2,), 2.71, dtype=torch.float32)
-            return a, b
-
-        func_scalar_param_compiled = torch.compile(func_scalar_param)
-        result_a, result_b = func_scalar_param_compiled()
-
-        self.assertEqual(result_a, torch.full((2,), 3.14, dtype=torch.float32))
-        self.assertEqual(result_b, torch.full((2,), 2.71, dtype=torch.float32))
-
 
 instantiate_parametrized_tests(FunctionTests)
 instantiate_parametrized_tests(DefaultsTests)

torch/_dynamo/variables/torch.py

@@ -834,24 +834,11 @@ class TorchInGraphFunctionVariable(BaseTorchVariable):
         @register(torch.full)
         def handle_full(self, tx, size, fill_value, **kwargs):
             if isinstance(fill_value, TensorVariable):
-                # Decompose: create empty tensor and fill it
+                result = TorchInGraphFunctionVariable(
-                # This avoids the scalar extraction at compile time
+                    torch.ops.aten._local_scalar_dense
-                empty_result = TorchInGraphFunctionVariable(torch.empty).call_function(
+                ).call_function(tx, [fill_value], {})
-                    tx, [size], kwargs
+                return TorchInGraphFunctionVariable(torch.full).call_function(
-                )
+                    tx, [size, result], kwargs
-                # Call fill_ method on the empty tensor
-                return empty_result.call_method(tx, "fill_", [fill_value], {})
-            else:
-                # For Python scalars and other non-tensor types, use default lowering
-                from .builder import wrap_fx_proxy
-
-                return wrap_fx_proxy(
-                    tx=tx,
-                    proxy=tx.output.create_proxy(
-                        "call_function",
-                        torch.ops.aten.full.default,
-                        *proxy_args_kwargs([size, fill_value], kwargs),
-                    ),
                 )
 
         @register(torch._foreach_lerp_)