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

add bool support to gpu gather

This commit is contained in:
TFBunny 2021-04-13 15:12:07 -04:00
parent fcac556d58
commit f6c9e90ad0
3 changed files with 83 additions and 40 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
mindspore/ccsrc/backend/kernel_compiler/gpu/arrays/gatherv2_gpu_kernel.cc
View File

@ -66,6 +66,13 @@ MS_REG_GPU_KERNEL_TWO(
MS_REG_GPU_KERNEL_TWO( MS_REG_GPU_KERNEL_TWO(
Gather, KernelAttr().AddInputAttr(kNumberTypeUInt8).AddInputAttr(kNumberTypeInt64).AddOutputAttr(kNumberTypeUInt8), Gather, KernelAttr().AddInputAttr(kNumberTypeUInt8).AddInputAttr(kNumberTypeInt64).AddOutputAttr(kNumberTypeUInt8),
GatherV2GpuFwdKernel, uint8_t, int64_t) GatherV2GpuFwdKernel, uint8_t, int64_t)
MS_REG_GPU_KERNEL_TWO(
Gather, KernelAttr().AddInputAttr(kNumberTypeBool).AddInputAttr(kNumberTypeInt32).AddOutputAttr(kNumberTypeBool),
GatherV2GpuFwdKernel, bool, int)
MS_REG_GPU_KERNEL_TWO(
Gather, KernelAttr().AddInputAttr(kNumberTypeBool).AddInputAttr(kNumberTypeInt64).AddOutputAttr(kNumberTypeBool),
GatherV2GpuFwdKernel, bool, int64_t)
// dynamic shape
MS_REG_GPU_KERNEL_TWO(Gather, MS_REG_GPU_KERNEL_TWO(Gather,
KernelAttr() KernelAttr()
.AddInputAttr(kNumberTypeFloat32) .AddInputAttr(kNumberTypeFloat32)
@ -94,6 +101,21 @@ MS_REG_GPU_KERNEL_TWO(Gather,
.AddInputAttr(kNumberTypeInt64) .AddInputAttr(kNumberTypeInt64)
.AddOutputAttr(kNumberTypeFloat16), .AddOutputAttr(kNumberTypeFloat16),
GatherV2GpuFwdKernel, half, int64_t) GatherV2GpuFwdKernel, half, int64_t)
MS_REG_GPU_KERNEL_TWO(Gather,
KernelAttr()
.AddInputAttr(kNumberTypeBool)
.AddInputAttr(kNumberTypeInt32)
.AddInputAttr(kNumberTypeInt64)
.AddOutputAttr(kNumberTypeBool),
GatherV2GpuFwdKernel, bool, int)
MS_REG_GPU_KERNEL_TWO(Gather,
KernelAttr()
.AddInputAttr(kNumberTypeBool)
.AddInputAttr(kNumberTypeInt64)
.AddInputAttr(kNumberTypeInt64)
.AddOutputAttr(kNumberTypeBool),
GatherV2GpuFwdKernel, bool, int64_t)
// dynamic shape ends
MS_REG_GPU_KERNEL_TWO( MS_REG_GPU_KERNEL_TWO(
SparseGatherV2, SparseGatherV2,
KernelAttr().AddInputAttr(kNumberTypeFloat32).AddInputAttr(kNumberTypeInt32).AddOutputAttr(kNumberTypeFloat32), KernelAttr().AddInputAttr(kNumberTypeFloat32).AddInputAttr(kNumberTypeInt32).AddOutputAttr(kNumberTypeFloat32),

5
mindspore/ccsrc/backend/kernel_compiler/gpu/cuda_impl/gatherv2.cu
View File

@ -77,3 +77,8 @@ template void GatherV2<uint8_t, int>(uint8_t *input, int *indices, uint8_t *outp
template void GatherV2<uint8_t, int64_t>(uint8_t *input, int64_t *indices, uint8_t *output, size_t output_dim0, template void GatherV2<uint8_t, int64_t>(uint8_t *input, int64_t *indices, uint8_t *output, size_t output_dim0,
size_t output_dim1, size_t output_dim2, size_t input_dim1, size_t output_dim1, size_t output_dim2, size_t input_dim1,
cudaStream_t stream); cudaStream_t stream);
template void GatherV2<bool, int>(bool *input, int *indices, bool *output, size_t output_dim0,
size_t output_dim1, size_t output_dim2, size_t input_dim1, cudaStream_t stream);
template void GatherV2<bool, int64_t>(bool *input, int64_t *indices, bool *output, size_t output_dim0,
size_t output_dim1, size_t output_dim2, size_t input_dim1,
cudaStream_t stream);

96
tests/st/ops/gpu/test_gatherV2_op.py
View File

@ -914,16 +914,16 @@ class GatherNet2(nn.Cell):
@pytest.mark.platform_x86_gpu_training @pytest.mark.platform_x86_gpu_training
@pytest.mark.env_onecard @pytest.mark.env_onecard
def test_gather2(): def test_gather2():
x = Tensor(np.array([[4., 5., 4., 1., 5.,], x = Tensor(np.array([[4., 5., 4., 1., 5.],
[4., 9., 5., 6., 4.,], [4., 9., 5., 6., 4.],
[9., 8., 4., 3., 6.,], [9., 8., 4., 3., 6.],
[0., 4., 2., 2., 8.,], [0., 4., 2., 2., 8.],
[1., 8., 6., 2., 8.,], [1., 8., 6., 2., 8.],
[8., 1., 9., 7., 3.,], [8., 1., 9., 7., 3.],
[7., 9., 2., 5., 7.,], [7., 9., 2., 5., 7.],
[9., 8., 6., 8., 5.,], [9., 8., 6., 8., 5.],
[3., 7., 2., 7., 4.,], [3., 7., 2., 7., 4.],
[4., 2., 8., 2., 9.,]] [4., 2., 8., 2., 9.]]
).astype(np.float32)) ).astype(np.float32))
indices = Tensor(np.array([[4000, 1, 300000]]).astype(np.int64)) indices = Tensor(np.array([[4000, 1, 300000]]).astype(np.int64))
@ -949,6 +949,7 @@ class GatherNetDynamic(nn.Cell):
self.to_dyn_1 = dyn_a self.to_dyn_1 = dyn_a
self.to_dyn_2 = dyn_b self.to_dyn_2 = dyn_b
self.axis = axis self.axis = axis
def construct(self, x, indices): def construct(self, x, indices):
# testing selective inputs being dynamic # testing selective inputs being dynamic
if self.to_dyn_1: if self.to_dyn_1:
@ -967,16 +968,16 @@ def test_gatherV2_dyn_ab():
""" """
context.set_context(mode=context.GRAPH_MODE, device_target="GPU") context.set_context(mode=context.GRAPH_MODE, device_target="GPU")
gather = GatherNetDynamic() gather = GatherNetDynamic()
x = Tensor(np.array([[4., 5., 4., 1., 5.,], x = Tensor(np.array([[4., 5., 4., 1., 5.],
[4., 9., 5., 6., 4.,], [4., 9., 5., 6., 4.],
[9., 8., 4., 3., 6.,], [9., 8., 4., 3., 6.],
[0., 4., 2., 2., 8.,], [0., 4., 2., 2., 8.],
[1., 8., 6., 2., 8.,], [1., 8., 6., 2., 8.],
[8., 1., 9., 7., 3.,], [8., 1., 9., 7., 3.],
[7., 9., 2., 5., 7.,], [7., 9., 2., 5., 7.],
[9., 8., 6., 8., 5.,], [9., 8., 6., 8., 5.],
[3., 7., 2., 7., 4.,], [3., 7., 2., 7., 4.],
[4., 2., 8., 2., 9.,]] [4., 2., 8., 2., 9.]]
).astype(np.float32)) ).astype(np.float32))
indices = Tensor(np.array([[4000, 1, 300000]]).astype(np.int32)) indices = Tensor(np.array([[4000, 1, 300000]]).astype(np.int32))
expect = np.array([[[0., 0., 0., 0., 0.], expect = np.array([[[0., 0., 0., 0., 0.],
@ -999,16 +1000,16 @@ def test_gatherV2_dyn_a():
context.set_context(mode=context.GRAPH_MODE, device_target="GPU") context.set_context(mode=context.GRAPH_MODE, device_target="GPU")
gather = GatherNetDynamic(-1, True, False) gather = GatherNetDynamic(-1, True, False)
# test 1 # test 1
x = Tensor(np.array([[4., 5., 4., 1., 5.,], x = Tensor(np.array([[4., 5., 4., 1., 5.],
[4., 9., 5., 6., 4.,], [4., 9., 5., 6., 4.],
[9., 8., 4., 3., 6.,], [9., 8., 4., 3., 6.],
[0., 4., 2., 2., 8.,], [0., 4., 2., 2., 8.],
[1., 8., 6., 2., 8.,], [1., 8., 6., 2., 8.],
[8., 1., 9., 7., 3.,], [8., 1., 9., 7., 3.],
[7., 9., 2., 5., 7.,], [7., 9., 2., 5., 7.],
[9., 8., 6., 8., 5.,], [9., 8., 6., 8., 5.],
[3., 7., 2., 7., 4.,], [3., 7., 2., 7., 4.],
[4., 2., 8., 2., 9.,]] [4., 2., 8., 2., 9.]]
).astype(np.float32)) ).astype(np.float32))
indices = Tensor(np.array([[4000, 1, 300000]]).astype(np.int64)) indices = Tensor(np.array([[4000, 1, 300000]]).astype(np.int64))
expect = np.array([[[0., 5., 0.]], expect = np.array([[[0., 5., 0.]],
@ -1075,16 +1076,16 @@ def test_gatherV2_dyn_b():
context.set_context(mode=context.GRAPH_MODE, device_target="GPU") context.set_context(mode=context.GRAPH_MODE, device_target="GPU")
gather = GatherNetDynamic(-1, False, True) gather = GatherNetDynamic(-1, False, True)
# test 1 # test 1
x = Tensor(np.array([[4., 5., 4., 1., 5.,], x = Tensor(np.array([[4., 5., 4., 1., 5.],
[4., 9., 5., 6., 4.,], [4., 9., 5., 6., 4.],
[9., 8., 4., 3., 6.,], [9., 8., 4., 3., 6.],
[0., 4., 2., 2., 8.,], [0., 4., 2., 2., 8.],
[1., 8., 6., 2., 8.,], [1., 8., 6., 2., 8.],
[8., 1., 9., 7., 3.,], [8., 1., 9., 7., 3.],
[7., 9., 2., 5., 7.,], [7., 9., 2., 5., 7.],
[9., 8., 6., 8., 5.,], [9., 8., 6., 8., 5.],
[3., 7., 2., 7., 4.,], [3., 7., 2., 7., 4.],
[4., 2., 8., 2., 9.,]] [4., 2., 8., 2., 9.]]
).astype(np.float32)) ).astype(np.float32))
indices = Tensor(np.array([[4000, 1, 300000]]).astype(np.int32)) indices = Tensor(np.array([[4000, 1, 300000]]).astype(np.int32))
expect = np.array([[[0., 5., 0.]], expect = np.array([[[0., 5., 0.]],
@ -1135,6 +1136,7 @@ def test_gatherV2_dyn_b():
assert np.all(diff < error) assert np.all(diff < error)
assert np.all(-diff < error) assert np.all(-diff < error)
@pytest.mark.level0 @pytest.mark.level0
@pytest.mark.platform_x86_gpu_training @pytest.mark.platform_x86_gpu_training
@pytest.mark.env_onecard @pytest.mark.env_onecard
@ -1358,3 +1360,17 @@ def test_gather1_uint8():
diff = output.asnumpy() - expect diff = output.asnumpy() - expect
assert np.all(diff < error) assert np.all(diff < error)
assert np.all(-diff < error) assert np.all(-diff < error)
@pytest.mark.level0
@pytest.mark.platform_x86_gpu_training
@pytest.mark.env_onecard
def test_gather1_bool():
x = Tensor(np.array([[0, 1, 1, 0], [1, 0, 0, 0], [1, 0, 1, 0]], dtype=np.bool))
indices = Tensor(np.array(([1, 2]), dtype='i4'))
expect = np.array([[1, 1], [0, 0], [0, 1]]).astype(np.bool)
context.set_context(mode=context.GRAPH_MODE, device_target="GPU")
gather = GatherNet1()
output = gather(x, indices)
assert np.all(expect == output.asnumpy())