Back out "[PyTorch Edge] Add Quantized Softmax Op (Naive Implementation)"

Summary:
Original commit changeset: 426a07808035

Original Phabricator Diff: D34943147 (pytorch@8d7242a)

Since D34943147 (pytorch@8d7242a) landed, Adfinder push candidates show consistently push blocking red counters for getAds C CPU main thread and getAds NC CPU main thread.

AF auto prod canary for D34943147 (pytorch@8d7242a), c1-c2 does shows 1.19% regression for counter 'getAds C CPU main thread' and ~1% regression for counter 'getAds C CPU main thread': https://www.internalfb.com/intern/experiment_store/experiment/27487791896054/#commit1-commit2
To help unblock adfinder push, reverting D34943147 (pytorch@8d7242a)

Test Plan:
Canary: https://our.intern.facebook.com/intern/ads/canary/442677925633895915
Canary completed: https://www.internalfb.com/intern/experiment_store/experiment/25288768753864/#commit1-commit2
Counter 'getAds C CPU main thread' moves in the opposite direction by -0.75.

Differential Revision: D35370901

fbshipit-source-id: b2e89f5976eb3fa2c2b22f120c0e32e380f5bc52

Author: Kunal Chakraborti

aten/src/ATen/native/quantized/cpu/qsoftmax.cpp

@@ -188,7 +188,6 @@ TORCH_LIBRARY(quantized, m) {
   m.def(TORCH_SELECTIVE_SCHEMA("quantized::relu6(Tensor qx, bool inplace=False) -> Tensor"));
   m.def(TORCH_SELECTIVE_SCHEMA("quantized::leaky_relu(Tensor qx, Scalar negative_slope, bool inplace, float output_scale, int output_zero_point) -> Tensor"));
   m.def(TORCH_SELECTIVE_SCHEMA("quantized::sigmoid(Tensor qx, float output_scale, int output_zero_point) -> Tensor"));
-  m.def(TORCH_SELECTIVE_SCHEMA("quantized::softmax(Tensor qx, int dim, float output_scale, int output_zero_point) -> Tensor"));
 }
 
 // According to #33294: The "_" prefix registration will be

aten/src/ATen/native/quantized/library.cpp

@@ -1101,40 +1101,6 @@ def test_qmatmul(self, num_dims, outer_dims, m, k, n, dtypes):
                                        scale_C,
                                        zero_point_C)
 
-    """Tests the correctness of the quantized softmax op."""
-    @given(num_dims=st.integers(2, 4),
-           dims=st.lists(st.integers(2, 5), min_size=5, max_size=5))
-    def test_qsoftmax(self, num_dims, dims):
-        size = dims[:num_dims]
-        torch_dtype = torch.quint8
-        np_dtype = np.uint8
-        dim = num_dims - 1
-
-        scale_X = 1.3
-        zero_point_X = 0
-        X = torch.rand(size=size, dtype=torch.float32) * 8 + zero_point_X
-
-        scale_Y = 1 / 256
-        zero_point_Y = 0
-
-        qX = torch.quantize_per_tensor(X,
-                                       scale=scale_X,
-                                       zero_point=zero_point_X,
-                                       dtype=torch_dtype)
-
-
-        # softmax ground truth
-        Y = torch.softmax(qX.dequantize(), dim=dim).numpy()
-        qY = _quantize(Y, scale_Y, zero_point_Y, dtype=np_dtype)
-        qY_hat = torch.ops.quantized.softmax(qX,
-                                             dim=dim,
-                                             output_scale=scale_Y,
-                                             output_zero_point=zero_point_Y)
-
-        np.testing.assert_equal(qY, qY_hat.int_repr(),
-                                "Quantized softmax failed.")
-
-
     """Tests the correctness of the mul and mul_relu op."""
     def test_qmul_broadcast(self):
         mul_relu = torch.ops.quantized.mul_relu

test/quantization/core/test_quantized_op.py

@@ -1222,7 +1222,6 @@ aten_native_source_non_codegen_list = [
     "aten/src/ATen/native/quantized/cpu/qreduction.cpp",
     "aten/src/ATen/native/quantized/cpu/qrelu.cpp",
     "aten/src/ATen/native/quantized/cpu/qsigmoid.cpp",
-    "aten/src/ATen/native/quantized/cpu/qsoftmax.cpp",
     "aten/src/ATen/native/quantized/cpu/qsort.cpp",
     "aten/src/ATen/native/quantized/cpu/qtanh.cpp",
     "aten/src/ATen/native/quantized/cpu/qthreshold.cpp",