add 'half_pixel' coordinate_transformation_mode support on CPU device when mode is 'bilinear'

docs/api/api_python/ops/mindspore.ops.ResizeBilinearV2.rst

@@ -7,8 +7,6 @@ mindspore.ops.ResizeBilinearV2
 
     调整过程只改变输入图像最低量维度的数据，分别代表高和宽。
 
-    .. warning::
-        在CPU后端，不支持将 `half_pixel_centers` 设为True。
 
     参数：
         - **align_corners** (bool，可选) - 如果为True，则使用比例 :math:`(new\_height - 1) / (height - 1)` 对输入进行缩放，此时输入图像和输出图像的四个角严格对齐。如果为False，使用比例 :math:`new\_height / height` 输入进行缩放。默认值：False。

docs/api/api_python/ops/mindspore.ops.func_interpolate.rst

@@ -9,7 +9,6 @@ mindspore.ops.interpolate
         - 实验特性，接口可能发生变化。
         - `roi` 是保留输入， `crop_and_resize` 坐标变换模式下生效，当前不支持。
         - Ascend平台下，当前不支持将 `mode` 设置为"linear"。
-        - CPU平台下，当 `mode` 是"bilinear"时，当前不支持将 `coordinate_transformation_mode` 设置为"half_pixel"。
 
     参数：
         - **x** (Tensor) - 输入Tensor。当 `mode` 是"linear"时， `x` 为3维Tensor。当 `mode` 是"bilinear"时， `x` 为4维Tensor。

mindspore/python/mindspore/ops/function/nn_func.py

@@ -1959,8 +1959,6 @@ def interpolate(x, roi=None, scales=None, sizes=None, coordinate_transformation_
         - The `roi` is reserved interface for 'crop_and_resize' coordinate transformation mode,
           which is not support now.
         - The Ascend platforms is currently not supported when `mode` is "linear".
-        - The 'half_pixel' coordinate_transformation_mode is currently not supported on CPU device
-          when mode is "bilinear".
 
     Args:
         x (Tensor): a tensor which to resize. `x` is a 3-D tensor when `mode` is "linear". `x` is a 4-D tensor when

mindspore/python/mindspore/ops/operations/image_ops.py

@@ -688,8 +688,6 @@ class ResizeBilinearV2(Primitive):
 
     The resizing only affects the lower two dimensions which represent the height and width.
 
-    .. warning::
-        On CPU, setting `half_pixel_centers` to True is currently not supported.
 
     Args:
         align_corners (bool, optional): If true, rescale input by :math:`(new\_height - 1) / (height - 1)`,
@@ -741,9 +739,6 @@ class ResizeBilinearV2(Primitive):
                                                              half_pixel_centers, [bool], self.name)
         if half_pixel_centers and align_corners:
             raise ValueError(f"If half_pixel_centers is True, align_corners must be False, but got {align_corners}")
-        target = context.get_context("device_target")
-        if half_pixel_centers and target == "CPU":
-            raise ValueError(f"Currently `half_pixel_centers`=True is not supported in CPU device_target")
 
 
 class ResizeBicubic(Primitive):