Cherry-pick 30194 30164 30201(#30202)

paddle/fluid/inference/api/paddle_pass_builder.cc

@@ -220,8 +220,12 @@ void CpuPassStrategy::EnableMKLDNN() {
              // "fc_mkldnn_pass",
              // "fc_act_mkldnn_fuse_pass",
              "batch_norm_act_fuse_pass",
+#ifndef _WIN32
+             // TODO(intel): Please fix the bug on windows.
+             // https://github.com/PaddlePaddle/Paddle/issues/29710
              "mkldnn_inplace_pass",  // This pass should be activated after
                                      // fuses
+#endif
          })) {
       passes_.push_back(pass);
     }

paddle/fluid/inference/tensorrt/CMakeLists.txt

+# Compiling with WITH_PYTHON=ON and WITH_TENSORRT=ON failed on windows. Temporarily add paddle_inference_api dependency to solve the problem
+if(WIN32)
+nv_library(tensorrt_engine SRCS engine.cc trt_int8_calibrator.cc DEPS ${GLOB_OPERATOR_DEPS} framework_proto device_context boost paddle_inference_api)
+else()
 nv_library(tensorrt_engine SRCS engine.cc trt_int8_calibrator.cc DEPS ${GLOB_OPERATOR_DEPS} framework_proto device_context boost)
+endif()
 nv_library(tensorrt_op_teller SRCS op_teller.cc DEPS framework_proto device_context boost)
 nv_test(test_tensorrt SRCS test_tensorrt.cc DEPS dynload_cuda device_context dynamic_loader)
 nv_test(test_tensorrt_engine SRCS test_engine.cc DEPS dynload_cuda tensorrt_engine)

paddle/fluid/operators/shape_op.cc

@@ -69,5 +69,6 @@ REGISTER_OPERATOR(
     paddle::framework::EmptyGradOpMaker<paddle::framework::OpDesc>,
     paddle::framework::EmptyGradOpMaker<paddle::imperative::OpBase>);
 REGISTER_OP_CPU_KERNEL(shape, ops::ShapeKernel<bool>, ops::ShapeKernel<int>,
+                       ops::ShapeKernel<int8_t>, ops::ShapeKernel<uint8_t>,
                        ops::ShapeKernel<int64_t>, ops::ShapeKernel<float>,
                        ops::ShapeKernel<double>);

paddle/fluid/operators/shape_op.cu

@@ -16,7 +16,8 @@ limitations under the License. */
 
 REGISTER_OP_CUDA_KERNEL(
     shape, paddle::operators::ShapeKernel<bool>,
-    paddle::operators::ShapeKernel<int>,
+    paddle::operators::ShapeKernel<int>, paddle::operators::ShapeKernel<int8_t>,
+    paddle::operators::ShapeKernel<uint8_t>,
     paddle::operators::ShapeKernel<int64_t>,
     paddle::operators::ShapeKernel<float>,
     paddle::operators::ShapeKernel<double>,