colin
/
TIM-VX
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Added reduce_all layoutinfer & reduce cases 

Type: New feature
Signed-off-by: Chen Xin <jack.chen@verisilicon.com>
This commit is contained in:
Chen Xin 2023-01-05 17:33:06 +08:00 committed by Sven
parent 789d4458ff
commit 58d36ab943
3 changed files with 103 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
src/tim/transform/layout_inference.cc
View File

@ -181,6 +181,7 @@ void LayoutInferContext::UpdateGraphOutputMap(const std::shared_ptr<vx::Tensor>&
REGIST_LAYOUT_INFERENCE(VSI_NN_REDUCE_PROD, ReduceProd); \
REGIST_LAYOUT_INFERENCE(VSI_NN_REDUCE_ANY, ReduceAny); \
REGIST_LAYOUT_INFERENCE(VSI_NN_REDUCE_SUM, ReduceSum); \
REGIST_LAYOUT_INFERENCE(VSI_NN_REDUCE_ALL, ReduceAll); \
default: \
VSILOGW("Op %d: Default layout inference pass for reduce.", reduce_type);\
assert(false); \

1
src/tim/transform/ops/reduce_layout_inference.h
View File

@ -89,6 +89,7 @@ using ReduceAnyLayoutInfer = ReduceLayoutInfer<tim::vx::ops::ReduceAny>;
using ReduceProdLayoutInfer = ReduceLayoutInfer<tim::vx::ops::ReduceProd>;
using ReduceMeanLayoutInfer = ReduceLayoutInfer<tim::vx::ops::ReduceMean>;
using ReduceSumLayoutInfer = ReduceLayoutInfer<tim::vx::ops::ReduceSum>;
using ReduceAllLayoutInfer = ReduceLayoutInfer<tim::vx::ops::ReduceAll>;
} // namespace transform
} // namespace tim

101
src/tim/vx/ops/reduce_sum_test.cc → src/tim/vx/ops/reduce_test.cc
View File

@ -128,4 +128,105 @@ TEST(Reduce_sum, KeepDims) {
std::vector<float> output(golden.size());
EXPECT_TRUE(output_tensor->CopyDataFromTensor(output.data()));
EXPECT_TRUE(ArraysMatch(golden, output, 1e-5f));
}
TEST(Reduce_all, KeepDims) {
auto ctx = tim::vx::Context::Create();
auto graph = ctx->CreateGraph();
tim::vx::ShapeType input_shape({2, 3, 2});
tim::vx::ShapeType output_shape({1, 3, 1});
tim::vx::TensorSpec input_spec(tim::vx::DataType::BOOL8, input_shape,
tim::vx::TensorAttribute::INPUT);
tim::vx::TensorSpec output_spec(tim::vx::DataType::BOOL8, output_shape,
tim::vx::TensorAttribute::OUTPUT);
std::vector<int32_t> axis = {2, 0};
auto input_tensor = graph->CreateTensor(input_spec);
auto output_tensor = graph->CreateTensor(output_spec);
auto reduce_all = graph->CreateOperation<tim::vx::ops::ReduceAll>(axis, true);
(*reduce_all).BindInputs({input_tensor}).BindOutputs({output_tensor});
bool in_data[] = {true, true, true, true, true, true,
true, true, false, true, true, true};
bool golden[] = {true, false, true};
bool* p_in = in_data;
input_tensor->CopyDataToTensor(p_in, 12 * sizeof(bool));
EXPECT_TRUE(graph->Compile());
EXPECT_TRUE(graph->Run());
bool output[3 * sizeof(bool)];
bool* p_output = output;
EXPECT_TRUE(output_tensor->CopyDataFromTensor(p_output));
for (int i = 0; i < 3; ++i) {
EXPECT_TRUE(golden[i] == output[i]);
}
}
TEST(Reduce_all, NotKeepDims) {
auto ctx = tim::vx::Context::Create();
auto graph = ctx->CreateGraph();
tim::vx::ShapeType input_shape({2, 3, 2});
tim::vx::ShapeType output_shape({2});
tim::vx::TensorSpec input_spec(tim::vx::DataType::BOOL8, input_shape,
tim::vx::TensorAttribute::INPUT);
tim::vx::TensorSpec output_spec(tim::vx::DataType::BOOL8, output_shape,
tim::vx::TensorAttribute::OUTPUT);
std::vector<int32_t> axis = {1, 2, 2, 2};
auto input_tensor = graph->CreateTensor(input_spec);
auto output_tensor = graph->CreateTensor(output_spec);
auto reduce_all =
graph->CreateOperation<tim::vx::ops::ReduceAll>(axis, false);
(*reduce_all).BindInputs({input_tensor}).BindOutputs({output_tensor});
bool in_data[] = {true, true, true, true, true, false,
true, true, true, true, true, true};
bool golden[] = {true, false};
bool* p_in = in_data;
input_tensor->CopyDataToTensor(p_in, 12 * sizeof(bool));
EXPECT_TRUE(graph->Compile());
EXPECT_TRUE(graph->Run());
bool output[2 * sizeof(bool)];
bool* p_output = output;
EXPECT_TRUE(output_tensor->CopyDataFromTensor(p_output));
for (int i = 0; i < 2; ++i) {
EXPECT_TRUE(golden[i] == output[i]);
}
}
TEST(Reduce_max, NotKeepDims) {
auto ctx = tim::vx::Context::Create();
auto graph = ctx->CreateGraph();
tim::vx::ShapeType input_shape({2, 3});
tim::vx::ShapeType output_shape({3});
tim::vx::TensorSpec input_spec(tim::vx::DataType::FLOAT32, input_shape,
tim::vx::TensorAttribute::INPUT);
tim::vx::TensorSpec output_spec(tim::vx::DataType::FLOAT32, output_shape,
tim::vx::TensorAttribute::OUTPUT);
auto input_tensor = graph->CreateTensor(input_spec);
auto output_tensor = graph->CreateTensor(output_spec);
std::vector<int32_t> axis = {0};
auto reduce_sum = graph->CreateOperation<tim::vx::ops::ReduceMax>(axis, false);
(*reduce_sum).BindInputs({input_tensor}).BindOutputs({output_tensor});
std::vector<float> in_data = {-1.0f, -2.0f, 3.0f, 4.0f, 5.0f, -6.0f};
std::vector<float> golden = {-1.0f, 4.0f, 5.0f};
EXPECT_TRUE(input_tensor->CopyDataToTensor(in_data.data(), in_data.size()));
EXPECT_TRUE(graph->Compile());
EXPECT_TRUE(graph->Run());
std::vector<float> output(golden.size());
EXPECT_TRUE(output_tensor->CopyDataFromTensor(output.data()));
EXPECT_EQ(golden, output);
}