From cd8a3dcb4ca42f22ad8f46a95e00977c92be6bbd Mon Sep 17 00:00:00 2001
From: AlexeyAB <alexeyab84@gmail.com>
Date: Thu, 08 Feb 2018 23:22:42 +0000
Subject: [PATCH] Compile fixes
---
src/convolutional_kernels.cu | 16 ++++++++++++++++
1 files changed, 16 insertions(+), 0 deletions(-)
diff --git a/src/convolutional_kernels.cu b/src/convolutional_kernels.cu
index 76a3fb3..03c9ab7 100644
--- a/src/convolutional_kernels.cu
+++ b/src/convolutional_kernels.cu
@@ -2,6 +2,10 @@
#include "curand.h"
#include "cublas_v2.h"
+#ifdef CUDNN
+#pragma comment(lib, "cudnn.lib")
+#endif
+
extern "C" {
#include "convolutional_layer.h"
#include "batchnorm_layer.h"
@@ -123,6 +127,7 @@
activate_array_ongpu(l.output_gpu, l.outputs*l.batch, l.activation);
//if(l.dot > 0) dot_error_gpu(l);
if(l.binary || l.xnor) swap_binary(&l);
+ //cudaDeviceSynchronize(); // for correct profiling of performance
}
void backward_convolutional_layer_gpu(convolutional_layer l, network_state state)
@@ -133,6 +138,9 @@
if(l.batch_normalize){
backward_batchnorm_layer_gpu(l, state);
+ //axpy_ongpu(l.outputs*l.batch, -state.net.decay, l.x_gpu, 1, l.delta_gpu, 1);
+ } else {
+ //axpy_ongpu(l.outputs*l.batch, -state.net.decay, l.output_gpu, 1, l.delta_gpu, 1);
}
float *original_input = state.input;
@@ -215,6 +223,10 @@
cuda_pull_array(layer.rolling_mean_gpu, layer.rolling_mean, layer.n);
cuda_pull_array(layer.rolling_variance_gpu, layer.rolling_variance, layer.n);
}
+ if (layer.adam){
+ cuda_pull_array(layer.m_gpu, layer.m, layer.c*layer.n*layer.size*layer.size);
+ cuda_pull_array(layer.v_gpu, layer.v, layer.c*layer.n*layer.size*layer.size);
+ }
}
void push_convolutional_layer(convolutional_layer layer)
@@ -228,6 +240,10 @@
cuda_push_array(layer.rolling_mean_gpu, layer.rolling_mean, layer.n);
cuda_push_array(layer.rolling_variance_gpu, layer.rolling_variance, layer.n);
}
+ if (layer.adam){
+ cuda_push_array(layer.m_gpu, layer.m, layer.c*layer.n*layer.size*layer.size);
+ cuda_push_array(layer.v_gpu, layer.v, layer.c*layer.n*layer.size*layer.size);
+ }
}
void update_convolutional_layer_gpu(convolutional_layer layer, int batch, float learning_rate, float momentum, float decay)
--
Gitblit v1.10.0