Remove not needed code

include/kernels/activations.cuh

@@ -1,16 +1,12 @@
-#include <functional>
-
 #ifndef ACTIVATIONS_H
 #define ACTIVATIONS_H
 
-__device__ float sigmoid(float a);
-__device__ float relu(float a);
-__device__ float linear(float a);
-
 __global__ void
 sigmoid_kernel(const float* __restrict__ src, float* __restrict__ dst, int len);
+
 __global__ void
 relu_kernel(const float* __restrict__ src, float* __restrict__ dst, int len);
+
 __global__ void
 linear_kernel(const float* __restrict__ src, float* __restrict__ dst, int len);
 

src/kernels/activations.cu

@@ -2,18 +2,6 @@
 
 #include "activations.cuh"
 
-__device__ float sigmoid(float a) {
-    return 1.0 / (1.0 + exp(-a));
-}
-
-__device__ float relu(float a) {
-    return a < 0.0 ? 0.0 : a;
-}
-
-__device__ float linear(float a) {
-    return a;
-}
-
 __global__ void sigmoid_kernel(
     const float* __restrict__ src,
     float* __restrict__ dst,
@@ -23,7 +11,7 @@ __global__ void sigmoid_kernel(
     int tid    = blockDim.x * blockIdx.x + threadIdx.x;
 
     for (int i = tid; i < len; i += stride) {
-        dst[i] = sigmoid(src[i]);
+        dst[i] = 1.0 / (1.0 + exp(-src[i]));
     }
 }
 
@@ -33,7 +21,7 @@ relu_kernel(const float* __restrict__ src, float* __restrict__ dst, int len) {
     int tid    = blockDim.x * blockIdx.x + threadIdx.x;
 
     for (int i = tid; i < len; i += stride) {
-        dst[i] = relu(src[i]);
+        dst[i] = src[i] < 0.0 ? 0.0 : src[i];
     }
 }
 
@@ -43,6 +31,6 @@ linear_kernel(const float* __restrict__ src, float* __restrict__ dst, int len) {
     int tid    = blockDim.x * blockIdx.x + threadIdx.x;
 
     for (int i = tid; i < len; i += stride) {
-        dst[i] = linear(src[i]);
+        dst[i] = src[i];
     }
 }

src/layers/dense.cu

@@ -81,6 +81,8 @@ void Layers::Dense::forward(const float* d_input, float* d_output) {
             d_output, d_output, outputSize
         );
     }
+
+    CUDA_CHECK(cudaDeviceSynchronize());
 }
 
 void Layers::Dense::toCuda() {

test/kernels/test_activations.cu

@@ -1,10 +1,9 @@
 #include <cuda_runtime_api.h>
-#include <driver_types.h>
+#include <gtest/gtest.h>
 
 #include <iostream>
 
 #include "activations.cuh"
-#include "gtest/gtest.h"
 #include "test_cublas_fixture.cuh"
 
 class ActivationsTest : public CublasTestFixture {

test/layers/test_dense.cu

@@ -1,11 +1,10 @@
 #include <cuda_runtime_api.h>
-#include <driver_types.h>
+#include <gtest/gtest.h>
 
 #include <iostream>
 
 #include "activations.cuh"
 #include "dense.cuh"
-#include "gtest/gtest.h"
 #include "test_cublas_fixture.cuh"
 
 class DenseLayerTest : public CublasTestFixture {

test/test_utils/test_cublas_fixture.cu

@@ -1,5 +1,6 @@
-#include "cublas_v2.h"
+#include <cublas_v2.h>
-#include "gtest/gtest.h"
+#include <gtest/gtest.h>
+
 #include "test_cublas_fixture.cuh"
 
 cublasHandle_t CublasTestFixture::cublasHandle;

test/test_utils/test_cublas_fixture.cuh

@@ -1,5 +1,5 @@
-#include "cublas_v2.h"
+#include <cublas_v2.h>
-#include "gtest/gtest.h"
+#include <gtest/gtest.h>
 
 class CublasTestFixture : public ::testing::Test {
   protected: