Mathis/CUDANet
1
0
Fork 0
mirror of https://github.com/lordmathis/CUDANet.git synced 2025-12-22 14:24:22 +00:00
Code Issues Packages Projects Releases Wiki Activity

Implement getOutputSize and getInputSize for seq layers

Browse Source
This commit is contained in:
Matúš Námešný
2024-04-22 20:31:58 +02:00
parent fbf6c44bdd
commit f17debc244
18 changed files with 186 additions and 66 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
test/kernels/test_matmul.cu
View File

@@ -187,6 +187,8 @@ TEST(MatMulTest, SumReduceTest) {
cudaMemcpy(d_input, input.data(), sizeof(float) * n, cudaMemcpyHostToDevice);
EXPECT_EQ(cudaStatus, cudaSuccess);
CUDANet::Utils::clear(d_sum, n);
CUDANet::Kernels::sum_reduce<<<gridSize, BLOCK_SIZE>>>(
d_input, d_sum, n
);
@@ -208,7 +210,5 @@ TEST(MatMulTest, SumReduceTest) {
EXPECT_FLOAT_EQ(expected, sum[0]);
cudaFree(d_input);
cudaFree(d_sum);
cudaFree(d_sum);
}

4
test/layers/test_avg_pooling.cu
View File

@@ -51,10 +51,10 @@ TEST(AvgPoolingLayerTest, AvgPoolForwardTest) {
int outputSize = avgPoolingLayer.getOutputSize();
std::vector<float> output(outputSize * outputSize * nChannels);
std::vector<float> output(outputSize);
cudaStatus = cudaMemcpy(
output.data(), d_output,
sizeof(float) * outputSize * outputSize * nChannels,
sizeof(float) * outputSize,
cudaMemcpyDeviceToHost
);
EXPECT_EQ(cudaStatus, cudaSuccess);

20
test/layers/test_conv2d.cu
View File

@@ -82,14 +82,15 @@ TEST_F(Conv2dTest, SimpleTest) {
activationType, input, kernels.data(), d_input
);
int outputSize = (inputSize - kernelSize) / stride + 1;
int outputWidth = (inputSize - kernelSize) / stride + 1;
int outputSize = outputWidth * outputWidth * numFilters;
EXPECT_EQ(outputSize, conv2d.getOutputSize());
d_output = conv2d.forward(d_input);
std::vector<float> expected = {44.0f, 54.0f, 64.0f, 84.0f, 94.0f,
104.0f, 124.0f, 134.0f, 144.0f};
std::vector<float> output(outputSize * outputSize * numFilters);
std::vector<float> output(outputSize);
cudaStatus = cudaMemcpy(
output.data(), d_output, sizeof(float) * output.size(),
@@ -172,18 +173,16 @@ TEST_F(Conv2dTest, PaddedTest) {
activationType, input, kernels.data(), d_input
);
EXPECT_EQ(inputSize, conv2d.getOutputSize());
EXPECT_EQ(inputSize * inputSize * numFilters, conv2d.getOutputSize());
d_output = conv2d.forward(d_input);
std::vector<float> output(
conv2d.getOutputSize() * conv2d.getOutputSize() * numFilters
conv2d.getOutputSize()
);
cudaMemcpy(
output.data(), d_output,
sizeof(float) * conv2d.getOutputSize() * conv2d.getOutputSize() *
numFilters,
cudaMemcpyDeviceToHost
sizeof(float) * conv2d.getOutputSize(), cudaMemcpyDeviceToHost
);
// Generated by tools/generate_conv2d_test.py
@@ -259,17 +258,16 @@ TEST_F(Conv2dTest, StridedPaddedConvolution) {
activationType, input, kernels.data(), d_input
);
EXPECT_EQ(inputSize, conv2d.getOutputSize());
EXPECT_EQ(inputSize * inputSize * numFilters, conv2d.getOutputSize());
d_output = conv2d.forward(d_input);
std::vector<float> output(
conv2d.getOutputSize() * conv2d.getOutputSize() * numFilters
conv2d.getOutputSize()
);
cudaMemcpy(
output.data(), d_output,
sizeof(float) * conv2d.getOutputSize() * conv2d.getOutputSize() *
numFilters,
sizeof(float) * conv2d.getOutputSize(),
cudaMemcpyDeviceToHost
);

4
test/layers/test_max_pooling.cu
View File

@@ -51,10 +51,10 @@ TEST(MaxPoolingLayerTest, MaxPoolForwardTest) {
int outputSize = maxPoolingLayer.getOutputSize();
std::vector<float> output(outputSize * outputSize * nChannels);
std::vector<float> output(outputSize);
cudaStatus = cudaMemcpy(
output.data(), d_output,
sizeof(float) * outputSize * outputSize * nChannels,
sizeof(float) * outputSize,
cudaMemcpyDeviceToHost
);
EXPECT_EQ(cudaStatus, cudaSuccess);