[0-size Tensor Job2 No.96] Add 0-size Tensor support for reshape

paddle/phi/kernels/onednn/reshape_grad_kernel.cc

@@ -11,6 +11,7 @@ limitations under the License. */
 
 #include "paddle/phi/backends/onednn/onednn_reuse.h"
 #include "paddle/phi/core/kernel_registry.h"
+#include "paddle/phi/kernels/full_kernel.h"
 
 namespace phi {
 
@@ -19,6 +20,12 @@ void ReshapeGradKernel(const Context& dev_ctx,
                        const DenseTensor& x,
                        const DenseTensor& out_grad,
                        DenseTensor* x_grad) {
+  if ((x_grad && x_grad->numel() == 0) || out_grad.numel() == 0) {
+    phi::Full<T, Context>(
+        dev_ctx, phi::IntArray(common::vectorize(x_grad->dims())), 0, x_grad);
+    return;
+  }
+
   auto out_grad_vec_dims = out_grad.dims().size() != 0
                                ? common::vectorize(out_grad.dims())
                                : std::vector<int64_t>{1};

paddle/phi/kernels/onednn/reshape_kernel.cc

@@ -154,6 +154,9 @@ void ReshapeKernel(const Context& dev_ctx,
                    const DenseTensor& x,
                    const IntArray& shape,
                    DenseTensor* out) {
+  if (x.numel() == 0) {
+    dev_ctx.Alloc(out, x.dtype());
+  }
   auto x_dims = x.dims();
   ExecuteReshape<T, Context>(dev_ctx, x, shape, x_dims, out);
 }

paddle/phi/kernels/reshape_grad_kernel.cc

@@ -46,6 +46,10 @@ void ReshapeGradKernel<phi::XPUContext>(const XPUContext& dev_ctx,
                                         const DenseTensor& x,
                                         const DenseTensor& out_grad,
                                         DenseTensor* x_grad) {
+  if (x_grad->numel() == 0) {
+    dev_ctx.Alloc(x_grad, x_grad->dtype());
+    return;
+  }
   auto x_dims = x_grad->dims();
   dev_ctx.Alloc(x_grad, out_grad.dtype());
   auto* src_ptr = out_grad.data();

test/legacy_test/test_reshape_op.py

@@ -15,10 +15,17 @@
 import unittest
 
 import numpy as np
-from op_test import OpTest, convert_float_to_uint16, skip_check_grad_ci
+from op_test import (
+    OpTest,
+    OpTestTool,
+    convert_float_to_uint16,
+    skip_check_grad_ci,
+)
 
 import paddle
 from paddle import base
+from paddle.base import core
+from paddle.base.framework import _current_expected_place
 from paddle.static import Program, program_guard
 
 
@@ -92,6 +99,20 @@ def init_data(self):
         self.inferred_shape = ()
 
 
+@OpTestTool.skip_if(
+    not (isinstance(_current_expected_place(), core.CPUPlace)),
+    "GPU is not supported",
+)
+class TestReshapeOp_ZeroDim4(OpTest):
+    def init_kernel_type(self):
+        self.use_onednn = True
+
+    def init_data(self):
+        self.ori_shape = (1,)
+        self.new_shape = ()
+        self.inferred_shape = ()
+
+
 class TestReshapeOp_ZeroSize(OpTest):
     def init_data(self):
         self.ori_shape = (0, 2)