mindspore-Ecosystem
/
mindspore
21
0
Fork 27
Code Issues Pull Requests 6 Projects Releases Wiki Activity

!18896 uniform op gpu bug fix 

Merge pull request !18896 from Peilin/uniform-op-bug-fix
This commit is contained in:
i-robot 2021-06-28 03:01:52 +00:00 committed by Gitee
parent c56715f094 922bcf603c
commit 7ae1c4a696
2 changed files with 3 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
mindspore/ccsrc/backend/kernel_compiler/gpu/cuda_impl/random_op_impl.cu
View File

@ -80,6 +80,7 @@ bool UniformInt(int seed, int seed2, curandState *globalState, T *input1, size_t
bool host_error_res = false;
UniformIntKernel<<<GET_BLOCKS(count), GET_THREADS, 0, cuda_stream>>>
(RNG_seed, globalState, input1, input_size_1, input2, input_size_2, output, count);
cudaDeviceSynchronize();
cudaMemcpyFromSymbol(&host_error_res, dev_error_res, sizeof(bool));
return host_error_res;
}

4
mindspore/ccsrc/backend/kernel_compiler/gpu/math/random_op_gpu_kernel.h
View File

@ -47,7 +47,7 @@ class RandomOpGpuKernel : public GpuKernel {
public:
RandomOpGpuKernel()
: random_op_type_(RANDOM_OP_INVALID_TYPE),
input_size_0_(sizeof(0)),
input_size_0_(sizeof(int32_t)),
input_size_1_(sizeof(T)),
input_size_2_(sizeof(T)),
output_size_(sizeof(T)),
@ -137,7 +137,7 @@ class RandomOpGpuKernel : public GpuKernel {
}
auto input_shape_0 = AnfAlgo::GetPrevNodeOutputInferShape(kernel_node, 0);
for (size_t i = 0; i < input_shape_0.size(); i++) {
input_size_0_ += input_shape_0[i];
input_size_0_ *= input_shape_0[i];
}
input_size_0_ *= sizeof(int);
if (random_op_type_ == RANDOM_OP_UNIFORM_INT) {