forked from mindspore-Ecosystem/mindspore
fix adaptive_maxpool2d bug.
This commit is contained in:
parent
e057782637
commit
c2167a5326
|
@ -7,6 +7,7 @@
|
|||
"mindspore/mindspore/core/abstract/ops/prim_nn.cc" "zerodivcond"
|
||||
"mindspore/mindspore/ccsrc/frontend/operator/ops_front_infer_function.cc" "zerodivcond"
|
||||
"mindspore/mindspore/ccsrc/pipeline/jit/pipeline_split.cc" "zerodivcond"
|
||||
"mindspore/mindspore/ccsrc/plugin/device/ascend/optimizer/ir_fusion/adaptive_max_pool2d_fusion.cc" "zerodivcond"
|
||||
"mindspore/mindspore/ccsrc/plugin/device/ascend/hal/device/ascend_stream_assign.cc" "useStlAlgorithm"
|
||||
"mindspore/mindspore/ccsrc/frontend/operator/ops_front_infer_function.cc" "uninitvar"
|
||||
"mindspore/mindspore/ccsrc/pipeline/jit/action.cc" "constVariable"
|
||||
|
|
|
@ -17,11 +17,13 @@
|
|||
#include <memory>
|
||||
#include <vector>
|
||||
#include "plugin/device/ascend/optimizer/ascend_helper.h"
|
||||
#include "backend/common/optimizer/helper.h"
|
||||
|
||||
namespace mindspore {
|
||||
namespace opt {
|
||||
namespace {
|
||||
constexpr int64_t kMaxPaddingSize = 10;
|
||||
constexpr int64_t kAdaptiveMaxpool2DOutputNumber = 1;
|
||||
|
||||
std::vector<int64_t> ComputeKernelAttr(int64_t input_size, int64_t output_size) {
|
||||
int64_t padding_size = 0;
|
||||
|
@ -115,11 +117,24 @@ const AnfNodePtr AdaptiveMaxPool2DFusion::Process(const FuncGraphPtr &func_graph
|
|||
|
||||
int64_t height = SizeToLong(input_shape.at(kDim2));
|
||||
int64_t width = SizeToLong(input_shape.at(kDim3));
|
||||
if (output_size[kDim0] <= 0 || output_size[kDim1] <= 0) {
|
||||
int64_t output_h = (output_size[kDim0] == -1) ? height : output_size[kDim0];
|
||||
int64_t output_w = (output_size[kDim1] == -1) ? width : output_size[kDim1];
|
||||
if (output_h <= 0 || output_w <= 0) {
|
||||
MS_LOG(EXCEPTION) << "AdaptiveMaxPool2D's output_size value is invalid.";
|
||||
}
|
||||
auto height_attr = ComputeKernelAttr(height, output_size[kDim0]);
|
||||
auto width_attr = ComputeKernelAttr(width, output_size[kDim1]);
|
||||
if (height % output_h != 0 || width % output_w != 0) {
|
||||
auto types = {common::AnfAlgo::GetOutputInferDataType(adaptive_max_pool2d, 0), kNumberTypeInt64};
|
||||
auto shapes = {common::AnfAlgo::GetOutputDetailShape(adaptive_max_pool2d, 0),
|
||||
common::AnfAlgo::GetOutputDetailShape(adaptive_max_pool2d, 0)};
|
||||
common::AnfAlgo::SetOutputTypeAndDetailShape(types, shapes, adaptive_max_pool2d.get());
|
||||
std::vector<int64_t> new_output_size{output_h, output_w};
|
||||
common::AnfAlgo::SetNodeAttr(kAttrOutputSize, MakeValue(new_output_size), adaptive_max_pool2d);
|
||||
std::vector<AnfNodePtr> multi_outputs;
|
||||
CreateMultipleOutputsOfAnfNode(func_graph, adaptive_max_pool2d, kAdaptiveMaxpool2DOutputNumber, &multi_outputs);
|
||||
return multi_outputs[kIndex0];
|
||||
}
|
||||
auto height_attr = ComputeKernelAttr(height, output_h);
|
||||
auto width_attr = ComputeKernelAttr(width, output_w);
|
||||
if (height_attr[kIndex0] == -1 || width_attr[kIndex0] == -1) {
|
||||
MS_LOG(EXCEPTION) << "Current AdaptiveMaxPool2D not support this scene! node:" << node->DebugString();
|
||||
}
|
||||
|
@ -128,8 +143,10 @@ const AnfNodePtr AdaptiveMaxPool2DFusion::Process(const FuncGraphPtr &func_graph
|
|||
(void)pooling_inputs.insert(pooling_inputs.end(), adaptive_max_pool2d->inputs().begin() + 1,
|
||||
adaptive_max_pool2d->inputs().end());
|
||||
auto pooling = NewCNode(pooling_inputs, kernel_graph);
|
||||
auto types = {common::AnfAlgo::GetOutputInferDataType(adaptive_max_pool2d, 0)};
|
||||
auto shapes = {common::AnfAlgo::GetOutputDetailShape(adaptive_max_pool2d, 0)};
|
||||
common::AnfAlgo::SetOutputTypeAndDetailShape(types, shapes, pooling.get());
|
||||
pooling->set_scope(adaptive_max_pool2d->scope());
|
||||
pooling->set_abstract(adaptive_max_pool2d->abstract());
|
||||
SetNodeAttr(pooling, height_attr, width_attr);
|
||||
return pooling;
|
||||
}
|
||||
|
|
|
@ -256,6 +256,9 @@ class AdaptiveMaxPool2D(Primitive):
|
|||
Output(i,j) &= {\max Input[h_{start}:h_{end}, w_{start}:w_{end}]}
|
||||
\end{align}
|
||||
|
||||
Note:
|
||||
In Ascend, the second output `argmax` is invalid, please ignore it.
|
||||
|
||||
Args:
|
||||
output_size (Union[int, tuple]): The target output size is H x W.
|
||||
ouput_size can be a tuple, or a single H for H x H, and H and W can be int or None
|
||||
|
@ -280,9 +283,10 @@ class AdaptiveMaxPool2D(Primitive):
|
|||
TypeError: If dtype of `input_x` is not float16, float32 or float64.
|
||||
ValueError: If `output_size` is a tuple and the length of `output_size` is not 2.
|
||||
ValueError: If the dimension of `input_x` is not NCHW or CHW.
|
||||
ValueError: If `output_size` is less than -1.
|
||||
|
||||
Supported Platforms:
|
||||
``GPU``
|
||||
``Ascend`` ``GPU`` ``CPU``
|
||||
|
||||
Examples:
|
||||
>>> # case 1: output_size=(None, 2)
|
||||
|
@ -331,6 +335,8 @@ class AdaptiveMaxPool2D(Primitive):
|
|||
self.output_size = (output_size, output_size) if isinstance(self.output_size, int) else output_size
|
||||
self.output_size = (-1 if self.output_size[0] is None else self.output_size[0],
|
||||
-1 if self.output_size[1] is None else self.output_size[1])
|
||||
for size in self.output_size:
|
||||
validator.check_number("output_size", size, -1, Rel.GE, None)
|
||||
self.add_prim_attr('output_size', self.output_size)
|
||||
self.add_prim_attr('return_indices', return_indices)
|
||||
|
||||
|
|
Loading…
Reference in New Issue