add

PaddlePaddle · Dec 25, 2024 · ec95360 · ec95360
1 parent 2b10e2d
commit ec95360
Show file tree

Hide file tree

Showing 31 changed files with 825 additions and 632 deletions.
diff --git a/paddle/phi/infermeta/backward.cc b/paddle/phi/infermeta/backward.cc
@@ -133,6 +133,42 @@ void ChannelShuffleGradInferMeta(const MetaTensor& out_grad,
   x_grad->set_dtype(out_grad.dtype());
 }
 
+void ClipTensorGradInferMeta(const MetaTensor& x,
+                            const MetaTensor& min,
+                            const MetaTensor& max,
+                            const MetaTensor& out_grad,
+                            MetaTensor* x_grad) {
+  auto x_dims = x.dims();
+  auto min_dims = min.dims();
+  auto max_dims = max.dims();
+
+  if (common::product(x_dims) >= common::product(min_dims) && common::product(x_dims) >= common::product(max_dims)) {
+    PADDLE_ENFORCE_EQ(
+      out_grad.dims(),
+      x.dims(),
+      errors::InvalidArgument(
+          "Gradients and its expand input should have the same shape."));
+    x_grad->set_dims(x.dims());
+  }
+  else if (common::product(min_dims) >= common::product(x_dims) && common::product(min_dims) >= common::product(max_dims)) {
+    PADDLE_ENFORCE_EQ(
+      out_grad.dims(),
+      min.dims(),
+      errors::InvalidArgument(
+          "Gradients and its expand input should have the same shape."));
+    x_grad->set_dims(min.dims());
+  }
+  else {
+    PADDLE_ENFORCE_EQ(
+      out_grad.dims(),
+      max.dims(),
+      errors::InvalidArgument(
+          "Gradients and its expand input should have the same shape."));
+    x_grad->set_dims(max.dims());
+  }
+  x_grad->set_dtype(x.dtype());
+}
+
 void ComplexGradInferMeta(const MetaTensor& x,
                           const MetaTensor& y,
                           const MetaTensor& dout,

diff --git a/paddle/phi/infermeta/backward.h b/paddle/phi/infermeta/backward.h
@@ -64,6 +64,12 @@ void ChannelShuffleGradInferMeta(const MetaTensor& out_grad,
                                  const std::string& data_format,
                                  MetaTensor* x_grad);
 
+void ClipTensorGradInferMeta(const MetaTensor& x,
+                            const MetaTensor& min,
+                            const MetaTensor& max,
+                            const MetaTensor& out_grad,
+                            MetaTensor* x_grad);
+
 void ComplexGradInferMeta(const MetaTensor& x,
                           const MetaTensor& y,
                           const MetaTensor& dout,

diff --git a/paddle/phi/infermeta/ternary.cc b/paddle/phi/infermeta/ternary.cc
@@ -373,6 +373,27 @@ void BoxCoderInferMeta(const MetaTensor& prior_box,
   output_box->set_dtype(target_box.dtype());
 }
 
+void ClipTensorInferMeta(const MetaTensor& x,
+                         const MetaTensor& min,
+                         const MetaTensor& max,
+                         MetaTensor* out) {
+
+  auto x_dims = x.dims();
+  auto min_dims = min.dims();
+  auto max_dims = max.dims();
+
+  if (common::product(x_dims) >= common::product(min_dims) && common::product(x_dims) >= common::product(max_dims)) {
+    out->set_dims(x.dims());
+  }
+  else if (common::product(min_dims) >= common::product(x_dims) && common::product(min_dims) >= common::product(max_dims)) {
+    out->set_dims(min.dims());
+  }
+  else if (common::product(max_dims) >= common::product(x_dims) && common::product(max_dims) >= common::product(min_dims)) {
+    out->set_dims(max.dims());
+  }
+  out->set_dtype(x.dtype());
+}
+
 void DistributedPushSparseInferMeta(
     const std::vector<const MetaTensor*>& ids,
     const std::vector<const MetaTensor*>& shows,

diff --git a/paddle/phi/infermeta/ternary.h b/paddle/phi/infermeta/ternary.h
@@ -80,6 +80,11 @@ void BoxCoderInferMeta(const MetaTensor& prior_box,
                        MetaTensor* output_box,
                        MetaConfig config = MetaConfig());
 
+void ClipTensorInferMeta(const MetaTensor& x,
+                         const MetaTensor& min,
+                         const MetaTensor& max,
+                         MetaTensor* out);
+
 void CollectFpnProposalsInferMeta(
     const std::vector<const MetaTensor*>& multi_level_rois,
     const std::vector<const MetaTensor*>& multi_level_scores,

diff --git a/paddle/phi/kernels/clip_grad_kernel.h b/paddle/phi/kernels/clip_grad_kernel.h
@@ -28,12 +28,4 @@ void ClipGradKernel(const Context& dev_ctx,
                     const Scalar& max,
                     DenseTensor* x_grad);
 
-template <typename T, typename Context>
-void ClipTensorGradKernel(const Context& dev_ctx,
-                          const DenseTensor& x,
-                          const DenseTensor& min,
-                          const DenseTensor& max,
-                          const DenseTensor& out_grad,
-                          DenseTensor* x_grad);
-
 }  // namespace phi
diff --git a/paddle/phi/kernels/clip_kernel.h b/paddle/phi/kernels/clip_kernel.h
@@ -28,11 +28,4 @@ void ClipKernel(const Context& dev_ctx,
                 const Scalar& max,
                 DenseTensor* out);
 
-template <typename T, typename Context>
-void ClipTensorKernel(const Context& dev_ctx,
-                      const DenseTensor& x,
-                      const DenseTensor& min,
-                      const DenseTensor& max,
-                      DenseTensor* out);
-
 }  // namespace phi
diff --git a/paddle/phi/kernels/clip_tensor_grad_kernel.h b/paddle/phi/kernels/clip_tensor_grad_kernel.h
@@ -0,0 +1,31 @@
+// Copyright (c) 2024 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include "paddle/phi/core/dense_tensor.h"
+#include "paddle/phi/core/device_context.h"
+#include "paddle/phi/kernels/expand_kernel.h"
+
+namespace phi {
+
+template <typename T, typename Context>
+void ClipTensorGradKernel(const Context& dev_ctx,
+                          const DenseTensor& x,
+                          const DenseTensor& min,
+                          const DenseTensor& max,
+                          const DenseTensor& out_grad,
+                          DenseTensor* x_grad);
+
+}  // namespace phi
diff --git a/paddle/phi/kernels/clip_tensor_kernel.h b/paddle/phi/kernels/clip_tensor_kernel.h
@@ -0,0 +1,30 @@
+// Copyright (c) 2024 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include "paddle/phi/core/dense_tensor.h"
+#include "paddle/phi/core/device_context.h"
+#include "paddle/phi/kernels/expand_kernel.h"
+
+namespace phi {
+
+template <typename T, typename Context>
+void ClipTensorKernel(const Context& dev_ctx,
+                      const DenseTensor& x,
+                      const DenseTensor& min,
+                      const DenseTensor& max,
+                      DenseTensor* out);
+
+}  // namespace phi
diff --git a/paddle/phi/kernels/cpu/clip_grad_kernel.cc b/paddle/phi/kernels/cpu/clip_grad_kernel.cc
@@ -18,31 +18,6 @@
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/clip_grad_kernel_impl.h"
 
-namespace phi {
-
-template <typename T, typename Context>
-void ClipTensorGradKernel(const Context& dev_ctx,
-                          const DenseTensor& x,
-                          const DenseTensor& min,
-                          const DenseTensor& max,
-                          const DenseTensor& out_grad,
-                          DenseTensor* x_grad) {
-  const T* x_data = x.data<T>();
-  const T* min_data = min.data<T>();
-  const T* max_data = max.data<T>();
-  auto numel = x.numel();
-  auto* dout = out_grad.data<T>();
-
-  auto* dx = dev_ctx.template Alloc<T>(x_grad);
-  for (int i = 0; i < numel; i++) {
-    dx[i] = (x_data[i] > min_data[i] && x_data[i] < max_data[i])
-                ? dout[i]
-                : static_cast<T>(0);
-  }
-}
-
-}  // namespace phi
-
 PD_REGISTER_KERNEL(clip_grad,
                    CPU,
                    ALL_LAYOUT,
@@ -51,12 +26,3 @@ PD_REGISTER_KERNEL(clip_grad,
                    double,
                    int,
                    int64_t) {}
-
-PD_REGISTER_KERNEL(clip_tensor_grad,
-                   CPU,
-                   ALL_LAYOUT,
-                   phi::ClipTensorGradKernel,
-                   float,
-                   double,
-                   int,
-                   int64_t) {}
diff --git a/paddle/phi/kernels/cpu/clip_kernel.cc b/paddle/phi/kernels/cpu/clip_kernel.cc
@@ -18,37 +18,5 @@
 #include "paddle/phi/core/kernel_registry.h"
 #include "paddle/phi/kernels/impl/clip_kernel_impl.h"
 
-namespace phi {
-
-template <typename T, typename Context>
-void ClipTensorKernel(const Context& dev_ctx,
-                      const DenseTensor& x,
-                      const DenseTensor& min,
-                      const DenseTensor& max,
-                      DenseTensor* out) {
-  const T* x_data = x.data<T>();
-  const T* min_data = min.data<T>();
-  const T* max_data = max.data<T>();
-  auto x_numel = x.numel();
-
-  T* out_data = dev_ctx.template Alloc<T>(out);
-
-  for (int i = 0; i < x_numel; i++) {
-    out_data[i] = x_data[i] < min_data[i] ? min_data[i] : x_data[i];
-    out_data[i] = out_data[i] > max_data[i] ? max_data[i] : out_data[i];
-  }
-}
-
-}  // namespace phi
-
 PD_REGISTER_KERNEL(
     clip, CPU, ALL_LAYOUT, phi::ClipKernel, float, double, int, int64_t) {}
-
-PD_REGISTER_KERNEL(clip_tensor,
-                   CPU,
-                   ALL_LAYOUT,
-                   phi::ClipTensorKernel,
-                   float,
-                   double,
-                   int,
-                   int64_t) {}
diff --git a/paddle/phi/kernels/cpu/clip_tensor_grad_kernel.cc b/paddle/phi/kernels/cpu/clip_tensor_grad_kernel.cc
@@ -0,0 +1,80 @@
+// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "paddle/phi/kernels/clip_tensor_grad_kernel.h"
+
+#include "paddle/phi/backends/cpu/cpu_context.h"
+#include "paddle/phi/core/kernel_registry.h"
+#include "paddle/phi/kernels/cast_kernel.h"
+#include "paddle/phi/kernels/expand_kernel.h"
+
+namespace phi {
+
+template <typename T, typename Context>
+void ClipTensorGradKernel(const Context& dev_ctx,
+                          const DenseTensor& x,
+                          const DenseTensor& min,
+                          const DenseTensor& max,
+                          const DenseTensor& out_grad,
+                          DenseTensor* x_grad) {
+
+  DenseTensor ex_min;
+  DenseTensor ex_max;
+  DenseTensor ex_x;
+  std::vector<int> real_target_shape = common::vectorize<int>(x_grad->dims());
+  if (x.dims() != x_grad->dims()) {
+    phi::ExpandKernel<T, Context>(
+        dev_ctx, x, real_target_shape, &ex_x);
+  } else {
+    ex_x = x;
+  }
+  if (min.dims() != x_grad->dims()) {
+    phi::ExpandKernel<T, Context>(
+        dev_ctx, min, real_target_shape, &ex_min);
+  } else {
+    ex_min = min;
+  }
+  if (max.dims() != x_grad->dims()) {
+    phi::ExpandKernel<T, Context>(
+        dev_ctx, max, real_target_shape, &ex_max);
+  } else {
+    ex_max = max;
+  }
+  phi::CastKernel<T, Context>(dev_ctx, ex_min, ex_x.dtype(), &ex_min);
+  phi::CastKernel<T, Context>(dev_ctx, ex_max, ex_x.dtype(), &ex_max);
+
+  const T* x_data = ex_x.data<T>();
+  const T* min_data = ex_min.data<T>();
+  const T* max_data = ex_max.data<T>();
+  auto numel = ex_x.numel();
+  auto* dout = out_grad.data<T>();
+
+  auto* dx = dev_ctx.template Alloc<T>(x_grad);
+  for (int i = 0; i < numel; i++) {
+    dx[i] = (x_data[i] > min_data[i] && x_data[i] < max_data[i])
+                ? dout[i]
+                : static_cast<T>(0);
+  }
+}
+
+}  // namespace phi
+
+PD_REGISTER_KERNEL(clip_tensor_grad,
+                   CPU,
+                   ALL_LAYOUT,
+                   phi::ClipTensorGradKernel,
+                   float,
+                   double,
+                   int,
+                   int64_t) {}