[Intel GPU] Remove special dispatch logic for xpu in adaptive_avg_pooling (pytorch#132217)

We now align the dispatch logic for XPU with CUDA in the adaptive average pooling operation.

Pull Request resolved: pytorch#132217
Approved by: https://github.com/EikanWang, https://github.com/atalman, https://github.com/albanD, https://github.com/malfet

Author: retonym

aten/src/ATen/native/AdaptiveAveragePooling.cpp

@@ -117,7 +117,7 @@ namespace {
       return at::mkldnn_adaptive_avg_pool2d(input, C10_AS_INTARRAYREF_SLOW(output_size));
     }
 
-    if (!input.is_quantized() && output_size[0] == 1 && output_size[1] == 1 && !input.is_xpu()) {
+    if (!input.is_quantized() && output_size[0] == 1 && output_size[1] == 1) {
       // in this case, adaptive pooling is just computing mean over hw
       // dimensions, which can be done more efficiently
       #if defined(C10_MOBILE) && defined(USE_XNNPACK)

aten/src/ATen/native/AdaptiveAveragePooling3d.cpp

@@ -313,7 +313,7 @@ Tensor adaptive_avg_pool3d_symint(Tensor const& input, SymIntArrayRef output_siz
         "adaptive_avg_pool3d: elements of output_size must be greater than or equal to 0 ",
         "but received {", output_size[0], ", ", output_size[1], ",", output_size[2], "}");
 
-  if (output_size[0] == 1 && output_size[1] == 1 && output_size[2] == 1 && !input.is_xpu()) {
+  if (output_size[0] == 1 && output_size[1] == 1 && output_size[2] == 1) {
     // in this case, adaptive pooling is just computing mean over hw
     // dimensions, which can be done more efficiently
     Tensor out = input.mean({-1, -2, -3}, /* keepdim = */ true);