Skip test_dtypes for fft bfloat16 and add comments for skipped cases

test/xpu/run_test_with_skip.py

@@ -791,7 +791,28 @@ def launch_test(test_case, skip_list=None, exe_list=None):
     # So far CUDA doesn't support Half, so that XPU fails as we aligned claimed dtypes with CUDA in test infra.
     "test_dtypes_nextafter_xpu",
 
+    # The following dtypes did not work in backward but are listed by the OpInfo: {torch.bfloat16}.
+    "test_dtypes_fft_fft2_xpu",
+    "test_dtypes_fft_fft_xpu",
+    "test_dtypes_fft_fftn_xpu",
+    "test_dtypes_fft_hfft2_xpu",
+    "test_dtypes_fft_hfft_xpu",
+    "test_dtypes_fft_hfftn_xpu",
+    "test_dtypes_fft_ifft2_xpu",
+    "test_dtypes_fft_ifft_xpu",
+    "test_dtypes_fft_ifftn_xpu",
+    "test_dtypes_fft_ihfft2_xpu",
+    "test_dtypes_fft_ihfft_xpu",
+    "test_dtypes_fft_ihfftn_xpu",
+    "test_dtypes_fft_irfft2_xpu",
+    "test_dtypes_fft_irfft_xpu",
+    "test_dtypes_fft_irfftn_xpu",
+    "test_dtypes_fft_rfft2_xpu",
+    "test_dtypes_fft_rfft_xpu",
+    "test_dtypes_fft_rfftn_xpu",
+
     # Deselect temporarily
+    # RuntimeError: The size of tensor a must match the size of tensor b at non-singleton dimension 2
     "test_out_fft_fft2_xpu_float32",
     "test_out_fft_fftn_xpu_float32",
     "test_out_fft_ifft2_xpu_float32",