Add API for HasNAN HasInf

paddle/framework/tensor_util.h

@@ -13,7 +13,10 @@ See the License for the specific language governing permissions and
 limitations under the License. */
 
 #pragma once
+#include "paddle/framework/data_type.h"
+#include "paddle/framework/eigen.h"
 #include "paddle/framework/tensor.h"
+#include "paddle/platform/device_context.h"
 
 namespace paddle {
 namespace framework {
@@ -205,5 +208,98 @@ inline void CopyToVector(const Tensor& src, std::vector<T>* dst) {
                src_ptr, size);
 }
 
+template <typename Predicate, typename DevCtx>
+struct AnyDTypeVisitor {
+  Predicate predicate_;
+  const Tensor& tensor_;
+  const DevCtx& ctx_;
+  Tensor* out_;
+
+  AnyDTypeVisitor(Predicate predicate, const Tensor& tensor, const DevCtx& ctx,
+                  Tensor* out)
+      : predicate_(predicate), tensor_(tensor), ctx_(ctx), out_(out) {}
+
+  template <typename T>
+  void operator()() const {
+    auto t = EigenVector<T>::Flatten(tensor_);
+    auto o = EigenScalar<bool>::From(*out_);
+    o.device(*ctx_.eigen_device()) = predicate_(t).any();
+  }
+};
+
+template <typename Predicate, typename DevCtx>
+inline void AnyImpl(Predicate predicate, const framework::Tensor& tensor,
+                    const DevCtx& ctx, framework::Tensor* out) {
+  VisitDataType(ToDataType(tensor.type()), AnyDTypeVisitor<Predicate, DevCtx>(
+                                               predicate, tensor, ctx, out));
+}
+
+template <typename Predicate>
+struct AnyVisitor : public boost::static_visitor<bool> {
+  const framework::Tensor& tensor_;
+  Predicate predicate_;
+
+  AnyVisitor(const framework::Tensor& tensor, Predicate predicate)
+      : tensor_(tensor), predicate_(std::move(predicate)) {}
+
+  template <typename Place>
+  bool operator()(const Place& place) const {
+    framework::Tensor out;
+    out.Resize({1});
+    out.mutable_data<bool>(place);
+    auto* ctx = platform::DeviceContextPool::Instance().GetByPlace(place);
+    AnyImpl(predicate_, tensor_, *ctx, &out);
+    return this->GetResult(out, place);
+  }
+
+  bool GetResult(const framework::Tensor& out,
+                 const platform::CUDAPlace& gpu) const {
+    platform::CPUPlace cpu;
+    framework::Tensor tmp;
+    tmp.Resize({1});
+    tmp.mutable_data<bool>(cpu);
+    platform::DeviceContextPool::Instance().Get(gpu)->Wait();
+    CopyFrom(out, cpu, &tmp);
+    platform::DeviceContextPool::Instance().Get(gpu)->Wait();
+    return GetResult(tmp, cpu);
+  }
+
+  bool GetResult(const framework::Tensor& out,
+                 const platform::CPUPlace& cpu) const {
+    return *out.data<bool>();
+  }
+};
+
+template <typename Predicate>
+inline bool Any(const framework::Tensor& tensor, Predicate predicate) {
+  AnyVisitor<Predicate> visitor(tensor, predicate);
+  auto place = tensor.place();
+  return platform::VisitPlace(place, visitor);
+}
+
+struct HasNanPredicate {
+  template <typename T>
+  auto operator()(T eigen_vec) const -> decltype(std::declval<T>().isnan()) {
+    return eigen_vec.isnan();
+  }
+};
+
+inline bool HasNan(const framework::Tensor& tensor) {
+  HasNanPredicate predicate;
+  return Any(tensor, predicate);
+}
+
+struct HasInfPredicate {
+  template <typename T>
+  auto operator()(T eigen_vec) const -> decltype(std::declval<T>().isinf()) {
+    return eigen_vec.isinf();
+  }
+};
+
+inline bool HasInf(const framework::Tensor& tensor) {
+  HasInfPredicate predicate;
+  return Any(tensor, predicate);
+}
+
 }  // namespace framework
 }  // namespace paddle

paddle/platform/device_context.h

@@ -52,6 +52,14 @@ class CPUDeviceContext : public DeviceContext {
   std::unique_ptr<Eigen::DefaultDevice> eigen_device_;
 };
 
+template <typename Place>
+struct DefaultDeviceContextType;
+
+template <>
+struct DefaultDeviceContextType<platform::CPUPlace> {
+  using TYPE = CPUDeviceContext;
+};
+
 #ifdef PADDLE_WITH_CUDA
 
 class EigenCudaStreamDevice;
@@ -90,6 +98,11 @@ class CUDADeviceContext : public DeviceContext {
   cublasHandle_t cublas_handle_;
 };
 
+template <>
+struct DefaultDeviceContextType<platform::CUDAPlace> {
+  using T = CUDADeviceContext;
+};
+
 class CUDNNDeviceContext : public CUDADeviceContext {
  public:
   explicit CUDNNDeviceContext(CUDAPlace place);
@@ -125,6 +138,13 @@ class DeviceContextPool {
   /*! \brief  Return handle of single device context. */
   const platform::DeviceContext* Get(const platform::Place& place);
 
+  template <typename Place>
+  const typename DefaultDeviceContextType<Place>::TYPE* GetByPlace(
+      const Place& place) {
+    return reinterpret_cast<
+        const typename DefaultDeviceContextType<Place>::TYPE*>(Get(place));
+  }
+
  private:
   static DeviceContextPool* pool;
   constexpr static int LEFT_SHIFT = 8;

paddle/platform/place.h

@@ -15,7 +15,7 @@ limitations under the License. */
 #pragma once
 
 #include <iostream>
-
+#include "paddle/platform/enforce.h"
 #include "paddle/platform/variant.h"
 
 namespace paddle {
@@ -64,5 +64,31 @@ bool places_are_same_class(const Place &, const Place &);
 
 std::ostream &operator<<(std::ostream &, const Place &);
 
+template <typename Visitor>
+struct PlaceVisitorWrapper
+    : public boost::static_visitor<typename Visitor::result_type> {
+  const Visitor &visitor_;
+  explicit PlaceVisitorWrapper(const Visitor &visitor) : visitor_(visitor) {}
+
+  typename Visitor::result_type operator()(const CPUPlace &cpu) const {
+    return visitor_(cpu);
+  }
+
+  typename Visitor::result_type operator()(const CUDAPlace &cuda) const {
+#ifdef PADDLE_WITH_CUDA
+    return visitor_(cuda);
+#else
+    PADDLE_THROW("Paddle is not compiled with CUDA. Cannot visit cuda device");
+    return typename Visitor::result_type();
+#endif
+  }
+};
+
+template <typename Visitor>
+typename Visitor::result_type VisitPlace(const Place &place,
+                                         const Visitor &visitor) {
+  return boost::apply_visitor(PlaceVisitorWrapper<Visitor>(visitor), place);
+}
+
 }  // namespace platform
 }  // namespace paddle