p60975123
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

add bool support to gpu select

This commit is contained in:
TFBunny 2021-04-14 11:49:01 -04:00
parent 2cf38aead8
commit b61e963b9f
4 changed files with 32 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
mindspore/ccsrc/backend/kernel_compiler/gpu/arrays/select_gpu_kernel.cc
View File

@ -53,5 +53,12 @@ MS_REG_GPU_KERNEL_ONE(Select,
.AddInputAttr(kNumberTypeInt64)
.AddOutputAttr(kNumberTypeInt64),
SelectGpuKernel, int64_t)
MS_REG_GPU_KERNEL_ONE(Select,
KernelAttr()
.AddInputAttr(kNumberTypeBool)
.AddInputAttr(kNumberTypeBool)
.AddInputAttr(kNumberTypeBool)
.AddOutputAttr(kNumberTypeBool),
SelectGpuKernel, bool)
} // namespace kernel
} // namespace mindspore

5
mindspore/ccsrc/backend/kernel_compiler/gpu/cuda_impl/select_impl.cu
View File

@ -1,5 +1,5 @@
/**
* Copyright 2020 Huawei Technologies Co., Ltd
* Copyright 2020-2021 Huawei Technologies Co., Ltd
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -44,4 +44,5 @@ template void CalSelect<half>(const size_t size, const bool* cond, const half* i
half* output, cudaStream_t cuda_stream);
template void CalSelect<int64_t>(const size_t size, const bool* cond, const int64_t* input_X, const int64_t* input_y,
int64_t* output, cudaStream_t cuda_stream);
template void CalSelect<bool>(const size_t size, const bool *cond, const bool *input_X, const bool *input_y,
bool *output, cudaStream_t cuda_stream);

8
mindspore/ccsrc/backend/kernel_compiler/gpu/cuda_impl/select_impl.cuh
View File

@ -1,5 +1,5 @@
/**
* Copyright 2020 Huawei Technologies Co., Ltd
* Copyright 2020-2021 Huawei Technologies Co., Ltd
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -14,12 +14,12 @@
* limitations under the License.
*/
#ifndef MINDSPORE_CCSRC_KERNEL_GPU_CUDA_IMPL_SELECT_IMPL_H_
#define MINDSPORE_CCSRC_KERNEL_GPU_CUDA_IMPL_SELECT_IMPL_H_
#ifndef MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_GPU_CUDA_IMPL_SELECT_IMPL_H_
#define MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_GPU_CUDA_IMPL_SELECT_IMPL_H_
#include "runtime/device/gpu/cuda_common.h"
template <typename T>
void CalSelect(const size_t size, const bool* cond, const T* input_x, const T* input_y, T* output,
cudaStream_t cuda_stream);
#endif // MINDSPORE_CCSRC_KERNEL_GPU_CUDA_IMPL_SELECT_IMPL_H_
#endif // MINDSPORE_CCSRC_BACKEND_KERNEL_COMPILER_GPU_CUDA_IMPL_SELECT_IMPL_H_

25
tests/st/ops/gpu/test_select_op.py
View File

@ -1,4 +1,4 @@
# Copyright 2020 Huawei Technologies Co., Ltd
# Copyright 2020-2021 Huawei Technologies Co., Ltd
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
@ -21,7 +21,6 @@ import mindspore.nn as nn
from mindspore import Tensor
from mindspore.ops import operations as P
class Net(nn.Cell):
def __init__(self):
super(Net, self).__init__()
@ -31,20 +30,32 @@ class Net(nn.Cell):
return self.select(cond_op, input_x, input_y)
cond = np.array([[True, False], [True, False]]).astype(np.bool)
x = np.array([[1.2, 1], [1, 0]]).astype(np.float32)
y = np.array([[1, 2], [3, 4.0]]).astype(np.float32)
@pytest.mark.level0
@pytest.mark.platform_x86_gpu_training
@pytest.mark.env_onecard
def test_select():
context.set_context(mode=context.GRAPH_MODE, device_target="GPU")
select = Net()
cond = np.array([[True, False], [True, False]]).astype(np.bool)
x = np.array([[1.2, 1], [1, 0]]).astype(np.float32)
y = np.array([[1, 2], [3, 4.0]]).astype(np.float32)
output = select(Tensor(cond), Tensor(x), Tensor(y))
expect = [[1.2, 2], [1, 4.0]]
error = np.ones(shape=[2, 2]) * 1.0e-6
diff = output.asnumpy() - expect
assert np.all(diff < error)
assert np.all(-diff < error)
context.set_context(mode=context.GRAPH_MODE, device_target="GPU")
x = np.array([[1, 0], [1, 0]]).astype(np.bool)
y = np.array([[0, 0], [1, 1]]).astype(np.bool)
output = select(Tensor(cond), Tensor(x), Tensor(y))
expect = np.array([[1, 0], [1, 1]]).astype(np.bool)
assert np.all(output.asnumpy() == expect)
context.set_context(mode=context.PYNATIVE_MODE, device_target="GPU")
x = np.array([[1, 0], [1, 0]]).astype(np.bool)
y = np.array([[0, 0], [1, 1]]).astype(np.bool)
output = select(Tensor(cond), Tensor(x), Tensor(y))
expect = np.array([[1, 0], [1, 1]]).astype(np.bool)
assert np.all(output.asnumpy() == expect)