Update sign op xpu (#41685)

* update sign op on xpu. test=kunlun

* fix typo. test=kunlun

cmake/external/xpu.cmake

@@ -7,42 +7,58 @@ SET(XPU_PROJECT                 "extern_xpu")
 SET(XPU_API_LIB_NAME            "libxpuapi.so")
 SET(XPU_RT_LIB_NAME             "libxpurt.so")
 
+if(NOT DEFINED XPU_BASE_URL)
+  SET(XPU_BASE_URL_WITHOUT_DATE "https://baidu-kunlun-product.cdn.bcebos.com/KL-SDK/klsdk-dev")
+  SET(XPU_BASE_URL "${XPU_BASE_URL_WITHOUT_DATE}/20220411")
+else()
+  SET(XPU_BASE_URL "${XPU_BASE_URL}")
+endif()
+
+# ubuntu and centos: use output by XDNN API team
+if(NOT DEFINED XPU_XDNN_BASE_URL)
+  SET(XPU_XDNN_BASE_URL_WITHOUT_DATE "https://klx-sdk-release-public.su.bcebos.com/xdnn/dev")
+  SET(XPU_XDNN_BASE_URL "${XPU_XDNN_BASE_URL_WITHOUT_DATE}/20220412")
+else()
+  SET(XPU_XDNN_BASE_URL "${XPU_XDNN_BASE_URL}")
+endif()
+
 IF(WITH_AARCH64)
   SET(XPU_XRE_DIR_NAME "xre-kylin_aarch64")
   SET(XPU_XDNN_DIR_NAME "xdnn-kylin_aarch64")
   SET(XPU_XCCL_DIR_NAME "xccl-kylin_aarch64")
+  SET(XPU_XDNN_URL "${XPU_BASE_URL}/${XPU_XDNN_DIR_NAME}.tar.gz" CACHE STRING "" FORCE)
 ELSEIF(WITH_SUNWAY)
   SET(XPU_XRE_DIR_NAME "xre-deepin_sw6_64")
   SET(XPU_XDNN_DIR_NAME "xdnn-deepin_sw6_64")
   SET(XPU_XCCL_DIR_NAME "xccl-deepin_sw6_64")
+  SET(XPU_XDNN_URL "${XPU_BASE_URL}/${XPU_XDNN_DIR_NAME}.tar.gz" CACHE STRING "" FORCE)
 ELSEIF(WITH_BDCENTOS)
   SET(XPU_XRE_DIR_NAME "xre-bdcentos_x86_64")
-  SET(XPU_XDNN_DIR_NAME "xdnn-bdcentos_x86_64")
+  SET(XPU_XDNN_DIR_NAME "XDNN-bdcentos_x86_64")
   SET(XPU_XCCL_DIR_NAME "xccl-bdcentos_x86_64")
+  # ubuntu and centos: use output by XDNN API team
+  SET(XPU_XDNN_URL "${XPU_XDNN_BASE_URL}/${XPU_XDNN_DIR_NAME}.tar.gz" CACHE STRING "" FORCE)
 ELSEIF(WITH_UBUNTU)
   SET(XPU_XRE_DIR_NAME "xre-ubuntu_x86_64")
-  SET(XPU_XDNN_DIR_NAME "xdnn-ubuntu_x86_64")
+  SET(XPU_XDNN_DIR_NAME "XDNN-ubuntu_x86_64")
   SET(XPU_XCCL_DIR_NAME "xccl-bdcentos_x86_64")
+  # ubuntu and centos: use output by XDNN API team
+  SET(XPU_XDNN_URL "${XPU_XDNN_BASE_URL}/${XPU_XDNN_DIR_NAME}.tar.gz" CACHE STRING "" FORCE)
 ELSEIF(WITH_CENTOS)
   SET(XPU_XRE_DIR_NAME "xre-centos7_x86_64")
-  SET(XPU_XDNN_DIR_NAME "xdnn-centos7_x86_64")
+  SET(XPU_XDNN_DIR_NAME "XDNN-bdcentos_x86_64")
   SET(XPU_XCCL_DIR_NAME "xccl-bdcentos_x86_64")
-
-ELSE ()
+  # ubuntu and centos: use output by XDNN API team
+  SET(XPU_XDNN_URL "${XPU_XDNN_BASE_URL}/${XPU_XDNN_DIR_NAME}.tar.gz" CACHE STRING "" FORCE)
+ELSE()
   SET(XPU_XRE_DIR_NAME "xre-ubuntu_x86_64")
-  SET(XPU_XDNN_DIR_NAME "xdnn-ubuntu_x86_64")
+  SET(XPU_XDNN_DIR_NAME "XDNN-bdcentos_x86_64")
   SET(XPU_XCCL_DIR_NAME "xccl-bdcentos_x86_64")
+  # default: use output by XDNN API team
+  SET(XPU_XDNN_URL "${XPU_XDNN_BASE_URL}/${XPU_XDNN_DIR_NAME}.tar.gz" CACHE STRING "" FORCE)
 ENDIF()
 
-if(NOT DEFINED XPU_BASE_URL)
-  SET(XPU_BASE_URL_WITHOUT_DATE "https://baidu-kunlun-product.cdn.bcebos.com/KL-SDK/klsdk-dev")
-  SET(XPU_BASE_URL "${XPU_BASE_URL_WITHOUT_DATE}/20220411")
-else()
-  SET(XPU_BASE_URL "${XPU_BASE_URL}")
-endif()
-
 SET(XPU_XRE_URL  "${XPU_BASE_URL}/${XPU_XRE_DIR_NAME}.tar.gz" CACHE STRING "" FORCE)
-SET(XPU_XDNN_URL "${XPU_BASE_URL}/${XPU_XDNN_DIR_NAME}.tar.gz" CACHE STRING "" FORCE)
 SET(XPU_XCCL_URL "${XPU_BASE_URL_WITHOUT_DATE}/20220411/${XPU_XCCL_DIR_NAME}.tar.gz" CACHE STRING "" FORCE)
 SET(XPU_PACK_DEPENCE_URL "https://baidu-kunlun-public.su.bcebos.com/paddle_depence/pack_paddle_depence.sh" CACHE STRING "" FORCE)
 

paddle/fluid/operators/sign_op_xpu.cc

@@ -14,6 +14,7 @@ limitations under the License. */
 
 #ifdef PADDLE_WITH_XPU
 #include "paddle/fluid/framework/op_registry.h"
+#include "paddle/fluid/platform/device/device_wrapper.h"
 #include "paddle/fluid/platform/device/xpu/xpu_header.h"
 namespace paddle {
 namespace operators {
@@ -26,10 +27,9 @@ class SignXPUKernel : public framework::OpKernel<T> {
     auto* in = context.Input<framework::Tensor>("X");
     out->mutable_data<T>(in->place());
     auto xpu_context = context.device_context<DeviceContext>().x_context();
-    int r = xpu::activation_forward(xpu_context, xpu::Activation_t::SIGN,
-                                    in->numel(), in->data<T>(), out->data<T>());
-    PADDLE_ENFORCE_EQ(r, xpu::Error_t::SUCCESS,
-                      platform::errors::Fatal("XPU sign kernel error!"));
+    // int sign(Context* ctx, const T* x , T* y, int len);
+    int r = xpu::sign(xpu_context, in->data<T>(), out->data<T>(), in->numel());
+    PADDLE_ENFORCE_XDNN_SUCCESS(r, "sign");
   }
 };
 

paddle/fluid/platform/device/xpu/xpu2_op_list.h

@@ -316,6 +316,7 @@ XPUOpMap& get_kl2_ops() {
       {"sigmoid", XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace())})},
       {"sigmoid_grad",
        XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace())})},
+      {"sign", XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace())})},
       {"slice_grad", XPUKernelSet({pOpKernelType(vartype::FP32, XPUPlace()),
                                    pOpKernelType(vartype::FP16, XPUPlace()),
                                    pOpKernelType(vartype::INT32, XPUPlace())})},

python/paddle/fluid/tests/unittests/xpu/test_sign_op_xpu.py

@@ -18,37 +18,69 @@ import unittest
 import numpy as np
 import sys
 sys.path.append("..")
-from op_test import OpTest
-import paddle
-import paddle.fluid as fluid
-from paddle.fluid import Program, program_guard
+
 import paddle
 
+from op_test import OpTest
+from op_test_xpu import XPUOpTest
+from xpu.get_test_cover_info import create_test_class, get_xpu_op_support_types, XPUOpTestWrapper
+
 paddle.enable_static()
 
 
-@unittest.skipIf(not paddle.is_compiled_with_xpu(),
-                 "core is not compiled with XPU")
-class TestXPUSignOp(OpTest):
+class XPUTestSignOP(XPUOpTestWrapper):
+    def __init__(self):
+        self.op_name = 'sign'
+        self.use_dynamic_create_class = False
+
+    class TestSignOPBase(XPUOpTest):
         def setUp(self):
-        self.op_type = "sign"
-        self.dtype = np.float32
-        self.inputs = {
-            'X': np.random.uniform(-10, 10, (10, 10)).astype(self.dtype)
-        }
-        self.outputs = {'Out': np.sign(self.inputs['X'])}
+            self.place = paddle.XPUPlace(0)
+            self.init_dtype()
+            self.set_case()
+
+        def set_case(self):
+            self.op_type = 'sign'
+            self.dtype = self.in_type
+            self.init_config()
+            self.x = np.random.uniform(-10, 10,
+                                       self.input_shape).astype(self.dtype)
+            self.inputs = {'X': self.x}
+            self.outputs = {'Out': np.sign(self.x)}
             self.attrs = {'use_xpu': True}
 
+        def init_dtype(self):
+            self.dtype = np.float32
+
         def test_check_output(self):
-        if paddle.is_compiled_with_xpu():
-            place = paddle.XPUPlace(0)
-            self.check_output_with_place(place)
+            self.check_output_with_place(self.place)
 
         def test_check_grad(self):
-        if paddle.is_compiled_with_xpu():
-            place = paddle.XPUPlace(0)
-            self.check_grad_with_place(place, ['X'], 'Out')
+            self.check_grad_with_place(self.place, ['X'], 'Out')
+
+        def init_config(self):
+            self.input_shape = [864]
+
+    class XPUTestSign1(TestSignOPBase):
+        def init_config(self):
+            self.input_shape = [2, 768]
+
+    class XPUTestSign2(TestSignOPBase):
+        def init_config(self):
+            self.input_shape = [3, 8, 4096]
+
+    class XPUTestSign3(TestSignOPBase):
+        def init_config(self):
+            self.input_shape = [1024]
+
+    class XPUTestSign4(TestSignOPBase):
+        def init_config(self):
+            self.input_shape = [2, 2, 255]
+
 
+support_types = get_xpu_op_support_types('sign')
+for stype in support_types:
+    create_test_class(globals(), XPUTestSignOP, stype)
 
 if __name__ == "__main__":
     unittest.main()