Remove F.pad python implementation

Pull Request resolved: #73433

Approved by: https://github.com/albanD, https://github.com/jbschlosser

Author: peterbell10

test/onnx/expect/TestOperators.test_pad.expect

@@ -161,7 +161,7 @@ graph {
     }
   }
   node {
-    input: "input"
+    input: "onnx::Pad_0"
     input: "onnx::Pad_22"
     output: "23"
     name: "Pad_13"
@@ -186,7 +186,7 @@ graph {
     raw_data: "\002\000\000\000\000\000\000\000\003\000\000\000\000\000\000\000\000\000\000\000\000\000\000\000\001\000\000\000\000\000\000\000"
   }
   input {
-    name: "input"
+    name: "onnx::Pad_0"
     type {
       tensor_type {
         elem_type: 1

test/test_fx.py

@@ -3810,6 +3810,7 @@ def tearDown(self):
         "linear": BUILT_IN_FUNC,
         "logsigmoid": BUILT_IN_FUNC,
         "one_hot": BUILT_IN_FUNC,
+        "pad": BUILT_IN_FUNC,
         "pairwise_distance": BUILT_IN_FUNC,
         "pdist": BUILT_IN_FUNC,
         "pixel_shuffle": BUILT_IN_FUNC,
@@ -3827,7 +3828,6 @@ def tearDown(self):
         "adaptive_max_pool2d_with_indices": LEN_ERROR,
         "adaptive_max_pool3d_with_indices": LEN_ERROR,
         "instance_norm": CONTROL_FLOW,
-        "pad": LEN_ERROR,
 
         "adaptive_max_pool1d": PROXY_ITERABLE,
         "adaptive_max_pool2d": PROXY_ITERABLE,

test/test_jit.py

@@ -15513,7 +15513,7 @@ def forward(self, x):
         self.assertEqual(m.int64_min, imported.int64_min)
 
     def test_script_scope(self):
-        scripted = torch.jit.script(torch.nn.functional.pad)
+        scripted = torch.jit.script(torch.nn.functional.triplet_margin_loss)
 
     @unittest.skipIf(IS_WINDOWS, "NYI: TemporaryFileName on Windows")
     def test_serialization_sharing(self):

test/test_nn.py

@@ -5430,8 +5430,8 @@ def test_FeatureAlphaDropout(self):
 
     def test_pad_scalar_error(self):
         inputs = torch.tensor(0., requires_grad=True)
-        self.assertRaises(AssertionError, lambda: F.pad(inputs, (1, 1)))
-        self.assertRaises(AssertionError, lambda: F.pad(inputs, (1,)))
+        self.assertRaises(RuntimeError, lambda: F.pad(inputs, (1, 1)))
+        self.assertRaises(RuntimeError, lambda: F.pad(inputs, (1,)))
 
     @unittest.skipIf(not TEST_NUMPY, "numpy not found")
     @parametrize_test("average_attn_weights", [True, False])
@@ -14345,10 +14345,10 @@ def test_pad(self, device, dtype):
         # Assert assertion errors are raised for invalid circular padding values
         inputs = torch.randn(1, 1, 4, device=device, dtype=dtype, requires_grad=True)
         # Should raise error when trying to wrap around more than once
-        self.assertRaises(AssertionError, lambda: F.pad(inputs, (5, 4), mode='circular'))
-        self.assertRaises(AssertionError, lambda: F.pad(inputs, (3, 6), mode='circular'))
+        self.assertRaises(RuntimeError, lambda: F.pad(inputs, (5, 4), mode='circular'))
+        self.assertRaises(RuntimeError, lambda: F.pad(inputs, (3, 6), mode='circular'))
         # Should raise error when negative padding results in negative output shape
-        self.assertRaises(AssertionError, lambda: F.pad(inputs, (-3, -2), mode='circular'))
+        self.assertRaises(RuntimeError, lambda: F.pad(inputs, (-3, -2), mode='circular'))
 
         # assert that relfection padding errors when pad >= input size
         expected_err_msg = r"Padding size should be less than the corresponding input dimension"