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

supported int16 dfp quantization & added conv2d unit test

This commit is contained in:
Feiyue Chen 2022-09-09 13:31:37 +08:00 committed by Sven
parent 95401036ab
commit 113c3722cb
5 changed files with 200 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
include/tim/vx/tensor.h
View File

@ -47,7 +47,9 @@ class Quantization {
channel_dim_(channel_dim),
scales_(std::move(scales)),
zero_points_(std::move(zero_points)) {}
Quantization(QuantType type, int8_t fl)
: type_(type),
fl_(fl){}
QuantType& Type() { return type_; }
const QuantType& Type() const { return type_; }
Quantization& SetType(QuantType type) {
@ -76,11 +78,14 @@ class Quantization {
return *this;
}
const std::int8_t& Fl() const{ return this->fl_; }
protected:
QuantType type_{QuantType::NONE};
int32_t channel_dim_{-1};
std::vector<float> scales_;
std::vector<int32_t> zero_points_;
int8_t fl_ = 0;
};
struct TensorSpec {

3
include/tim/vx/types.h
View File

@ -35,12 +35,13 @@ enum class DataType {
UINT16,
INT32,
UINT32,
INT64,
FLOAT16,
FLOAT32,
BOOL8
};
enum class QuantType { NONE, ASYMMETRIC, SYMMETRIC_PER_CHANNEL };
enum class QuantType { NONE, ASYMMETRIC, SYMMETRIC_PER_CHANNEL, DYNAMIC_FIXED_POINT };
enum TensorAttribute {
CONSTANT = 1 << 0,

184
src/tim/vx/ops/conv2d_test.cc
View File

@ -1662,3 +1662,187 @@ TEST(Conv2d, shape_w_h_128_1_ksize_1_1_stride_2_int8_QuantizedPerChannelTest) {
}
}
}
TEST(Conv2d, shape_4_2_2_2_int16_DFPQuantizedTest){
auto ctx = tim::vx::Context::Create();
if(ctx->isClOnly()) GTEST_SKIP();
auto graph = ctx->CreateGraph();
tim::vx::ShapeType input_shape({4, 2, 2, 2}); //whcn
tim::vx::ShapeType weight_shape({1, 1, 2, 1}); //whio
tim::vx::ShapeType bias_shape({weight_shape[3]});
tim::vx::ShapeType output_shape(
{4, 2, weight_shape[3], input_shape[3]}); //whcn
int8_t fl_input = 9, fl_weight= 8, fl_output = 8;
tim::vx::Quantization quant_input(tim::vx::QuantType::DYNAMIC_FIXED_POINT, fl_input);
tim::vx::Quantization quant_weight(tim::vx::QuantType::DYNAMIC_FIXED_POINT, fl_weight);
tim::vx::Quantization quant_output(tim::vx::QuantType::DYNAMIC_FIXED_POINT, fl_output);
tim::vx::TensorSpec input_spec(tim::vx::DataType::INT16, input_shape,
tim::vx::TensorAttribute::INPUT,
quant_input);
tim::vx::TensorSpec weight_spec(tim::vx::DataType::INT16, weight_shape,
tim::vx::TensorAttribute::CONSTANT,
quant_weight);
tim::vx::TensorSpec output_spec(tim::vx::DataType::INT16, output_shape,
tim::vx::TensorAttribute::OUTPUT,
quant_output);
// Input data float
std::vector<float> input_data_float = {
0.5, 0.5, 0.5, 0.5, 1, 1, 1, 1, 0.5, 0.5, 0.5, 0.5, 1, 1, 1, 1,
0.5, 1, 1.5, 2, 0.5, 1, 1.5, 2, 0.5, 1, 1.5, 2, 0.5, 1, 1.5, 2};
// weight data float
std::vector<float> weight_data_float= {
1, 2 // first filter
};
//input data(dfp16)
std::vector<int16_t> input_data = {
256,256,256,256, 512,512,512,512, 256,256,256,256,512,512,512,512,
256,512,768,1024,256,512,768,1024,256,512,768,1024,256,512,768,1024
};
//weight data(dfp16)
std::vector<int16_t> weight_data = {
256,512
};
// bias data
std::vector<int64_t> bias_data = {0};
//golden
std::vector<float> golden = {1.5, 1.5, 1.5, 1.5, 3, 3, 3, 3,
1.5, 3, 4.5, 6, 1.5, 3, 4.5, 6};
auto input_tensor = graph->CreateTensor(input_spec,input_data.data());
auto weight_tensor = graph->CreateTensor(weight_spec, weight_data.data());
auto output_tensor = graph->CreateTensor(output_spec);
auto padding = tim::vx::PadType::SAME;
std::array<uint32_t, 2> stride({1, 1});
std::array<uint32_t, 2> dilation({0, 0});
auto conv2d = graph->CreateOperation<tim::vx::ops::Conv2d>(
padding, stride, dilation);
(*conv2d)
.BindInput(input_tensor)
.BindInput(weight_tensor)
.BindOutput(output_tensor);
EXPECT_TRUE(graph->Compile());
input_tensor->CopyDataToTensor(input_data.data());
EXPECT_TRUE(graph->Run());
uint32_t output_size = 1;
for (auto i : output_tensor->GetShape()) {
output_size *= i;
}
std::vector<int16_t> output(output_size);
EXPECT_TRUE(output_tensor->CopyDataFromTensor(output.data()));
//transform output(int16) to fp
std::vector<float> f;
for(const auto& q : output){
f.push_back( q / (float)((int64_t)1 << fl_output));
}
EXPECT_EQ(golden, f);
}
TEST(Conv2d, shape_4_2_1_1_int16_DFPQuantizedTest) {
auto ctx = tim::vx::Context::Create();
if(ctx->isClOnly()) GTEST_SKIP();
auto graph = ctx->CreateGraph();
tim::vx::ShapeType input_shape({4, 2, 1, 1}); //whcn
tim::vx::ShapeType weight_shape({2, 2, 1, 3}); //whio
tim::vx::ShapeType bias_shape({weight_shape[3]});
tim::vx::ShapeType output_shape(
{4, 2, weight_shape[3], input_shape[3]}); //whcn
int8_t fl_input = 9, fl_weight = 8, fl_bias = 17,fl_output = 8;
tim::vx::Quantization quant_input(tim::vx::QuantType::DYNAMIC_FIXED_POINT, fl_input);
tim::vx::Quantization quant_weight(tim::vx::QuantType::DYNAMIC_FIXED_POINT, fl_weight);
tim::vx::Quantization quant_bias(tim::vx::QuantType::DYNAMIC_FIXED_POINT, fl_bias);
tim::vx::Quantization quant_output(tim::vx::QuantType::DYNAMIC_FIXED_POINT, fl_output);
tim::vx::TensorSpec input_spec(tim::vx::DataType::INT16, input_shape,
tim::vx::TensorAttribute::INPUT,
quant_input);
tim::vx::TensorSpec weight_spec(tim::vx::DataType::INT16, weight_shape,
tim::vx::TensorAttribute::CONSTANT,
quant_weight);
tim::vx::TensorSpec bias_spec(tim::vx::DataType::INT64, bias_shape,
tim::vx::TensorAttribute::CONSTANT,
quant_bias);
tim::vx::TensorSpec output_spec(tim::vx::DataType::INT16, output_shape,
tim::vx::TensorAttribute::OUTPUT,
quant_output);
// Input data nchw
std::vector<float> input_data_float= {
1, 1, 1, 1, // row = 1
2, 2, 3, 2 // row = 2
};
// weight data oihw
std::vector<float> weight_data_float= {
1, 2, 3, 4, //first 2x2 filter
-1, 1, -1, 1, // second 2x2 filter
-1, -1, 1, 1, // third 2x2 filter
};
// bias data
std::vector<float> bias_data_float = {1, 2, 3};
// nchw
std::vector<float> golden = {// first channel
18, 22, 21, 8, 7, 9, 8, 3,
// second channel
2, 3, 1, -1, 2, 3, 1, 0,
// third channel
5, 6, 6, 4, -1, -2, -2, 1};
std::vector<int16_t> input_data = {
512, 512, 512, 512,
1024,1024,1536,1024
};
std::vector<int16_t> weight_data = {
256,512,768,1024,
-256,256,-256,256,
-256,-256,256,256
};
std::vector<int64_t> bias_data = {
1<<fl_bias, 2*(1<<fl_bias),3*(1<<fl_bias)
};
auto input_tensor = graph->CreateTensor(input_spec, input_data.data());
auto weight_tensor = graph->CreateTensor(weight_spec, weight_data.data());
auto bias_tensor = graph->CreateTensor(bias_spec, bias_data.data());
auto output_tensor = graph->CreateTensor(output_spec);
auto padding = tim::vx::PadType::SAME;
std::array<uint32_t, 2> stride({1, 1});
std::array<uint32_t, 2> dilation({0, 0});
auto conv2d = graph->CreateOperation<tim::vx::ops::Conv2d>(
padding, stride, dilation);
(*conv2d)
.BindInput(input_tensor)
.BindInput(weight_tensor)
.BindInput(bias_tensor)
.BindOutput(output_tensor);
EXPECT_TRUE(graph->Compile());
input_tensor->CopyDataToTensor(input_data.data());
EXPECT_TRUE(graph->Run());
uint32_t output_size = 1;
for (auto i : output_tensor->GetShape()) {
output_size *= i;
}
std::vector<int16_t> output(output_size);
EXPECT_TRUE(output_tensor->CopyDataFromTensor(output.data()));
//transform output(int16) to fp
std::vector<float> f;
for(const auto& q : output){
f.push_back( q / (float)((int64_t)1 << fl_output));
}
EXPECT_EQ(golden, f);
}

4
src/tim/vx/tensor.cc
View File

@ -66,6 +66,10 @@ void PackTensorDtype(tim::vx::TensorSpec& spec, vsi_nn_dtype_t* dtype) {
dtype->channel_dim = spec.quantization_.ChannelDim();
break;
}
case tim::vx::QuantType::DYNAMIC_FIXED_POINT:
dtype->fl = spec.quantization_.Fl();
break;
default:
break;
}

4
src/tim/vx/type_utils.cc
View File

@ -39,6 +39,8 @@ vsi_nn_type_e TranslateDataType(DataType dtype) {
return VSI_NN_TYPE_INT32;
case DataType::UINT32:
return VSI_NN_TYPE_UINT32;
case DataType::INT64:
return VSI_NN_TYPE_INT64;
case DataType::FLOAT16:
return VSI_NN_TYPE_FLOAT16;
case DataType::FLOAT32:
@ -59,6 +61,8 @@ vsi_nn_qnt_type_e TranslateQuantType(QuantType qtype) {
return VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC;
case QuantType::SYMMETRIC_PER_CHANNEL:
return VSI_NN_QNT_TYPE_AFFINE_PERCHANNEL_SYMMETRIC;
case QuantType::DYNAMIC_FIXED_POINT:
return VSI_NN_QNT_TYPE_DFP;
default:
break;
}