Skip to content
Snippets Groups Projects
Commit 83700e90 authored by Hashim Sharif's avatar Hashim Sharif
Browse files

Reducing batch_size = 20 to avoid out of mem errors

parent c0bafc06
No related branches found
No related tags found
No related merge requests found
Hide whitespace changes
Inline Side-by-side
llvm/projects/hpvm-tensor-rt/dnn_sources/src/promise/vgg16_imagenet_piped.cc
+ 2
− 5
View file @ 83700e90
......@@ -24,8 +24,8 @@ int main(int argc, char* argv[]){
int total_runs = 1;
int offset = 0;
int test_input_size = 3000;
int batch_size = 50;
int test_input_size = 2000;
int batch_size = 20;
int batch_count = test_input_size / batch_size;
float final_accuracy = 0.0;
......@@ -131,14 +131,12 @@ int main(int argc, char* argv[]){
// NOTE: Wait on signal from OpenTuner
stallOnOpenTunerSignal();
startMemTracking();
for(int i = 0; i < batch_count; i++){
int start = i * batch_size;
int end = (i + 1) * batch_size;
......@@ -175,7 +173,6 @@ int main(int argc, char* argv[]){
// NOTE: Signal back to OpenTuner
signalPipeToOpenTuner();
}
......
llvm/projects/hpvm-tensor-rt/dnn_sources/src/promise/vgg16_imagenet_promise.cc
+ 17
− 18
View file @ 83700e90
......@@ -19,13 +19,13 @@ int to_skip = 5;
int main(int argc, char* argv[]){
llvm_hpvm_initTensorRt(0);
llvm_hpvm_initTensorRt(1);
int total_runs = 1;
int offset = 0;
int test_input_size = 3000;
int batch_size = 50;
int test_input_size = 2000;
int batch_size = 20;
int batch_count = test_input_size / batch_size;
float final_accuracy = 0.0;
......@@ -130,30 +130,29 @@ int main(int argc, char* argv[]){
for (int i = 0 ; i < total_runs; i++){
startMemTracking();
for(int i = 0; i < batch_count; i++){
int start = i * batch_size;
int end = (i + 1) * batch_size;
void* input = readInputBatch(input_path.c_str(),0,start,end,3,224,224);
void* var_0 = ConvLayer_PROMISE(input, -123.68, 151.061, conv2d_1_w, -0.5682651399970055, 0.5677501424551024, conv2d_1_b, -0.015828926, 2.064037, 1, 1, 1, 1, -1, 0, 1, 0.0, 407.96143194580145, 9);
void* var_1 = ConvLayer_PROMISE(var_0, 0.0, 407.96143194580145, conv2d_2_w, -0.13156980648636818, 0.2164201746285022, conv2d_2_b, -1.0271513, 0.9052184, 1, 1, 1, 1, 0, 2, 1, 0.0, 1973.2054975586288, 9);
void* var_2 = ConvLayer_PROMISE(var_1, 0.0, 1973.2054975586288, conv2d_3_w, -0.18644111251831055, 0.202149114727974, conv2d_3_b, -0.17922063, 0.36547425, 1, 1, 1, 1, -1, 0, 1, 0.0, 2386.9648486329534, 9);
void* var_3 = ConvLayer_PROMISE(var_2, 0.0, 2386.9648486329534, conv2d_4_w, -0.10804861642420292, 0.12427636455744764, conv2d_4_b, -0.59533477, 0.63375777, 1, 1, 1, 1, 0, 2, 1, 0.0, 4998.494643554761, 9);
void* var_4 = ConvLayer_PROMISE(var_3, 0.0, 4998.494643554761, conv2d_5_w, -0.08040237371623515, 0.09835810117424044, conv2d_5_b, -0.20097896, 0.34949613, 1, 1, 1, 1, -1, 0, 1, 0.0, 4637.92161425807, 9);
void* var_5 = ConvLayer_PROMISE(var_4, 0.0, 4637.92161425807, conv2d_6_w, -0.05306418750435114, 0.06628044287860436, conv2d_6_b, -0.18124875, 0.274845, 1, 1, 1, 1, -1, 0, 1, 0.0, 4365.822572754019, 9);
void* var_6 = ConvLayer_PROMISE(var_5, 0.0, 4365.822572754019, conv2d_7_w, -0.05084674355760217, 0.07320860563218634, conv2d_7_b, -0.14288792, 0.59477174, 1, 1, 1, 1, 0, 2, 1, 0.0, 5600.749117676456, 9);
void* var_7 = ConvLayer_PROMISE(var_6, 0.0, 5600.749117676456, conv2d_8_w, -0.04523278899490833, 0.053042236261070186, conv2d_8_b, -0.14548235, 0.3148451, 1, 1, 1, 1, -1, 0, 1, 0.0, 3240.830364746551, 9);
void* var_8 = ConvLayer_PROMISE(var_7, 0.0, 3240.830364746551, conv2d_9_w, -0.02917514201253653, 0.03586270406842279, conv2d_9_b, -0.08428453, 0.18237582, 1, 1, 1, 1, -1, 0, 1, 0.0, 1895.9044943847766, 9);
void* var_9 = ConvLayer_PROMISE(var_8, 0.0, 1895.9044943847766, conv2d_10_w, -0.029496615380048753, 0.04047201693058028, conv2d_10_b, -0.19835947, 0.33766547, 1, 1, 1, 1, 0, 2, 1, 0.0, 1273.674801757832, 9);
void* var_10 = ConvLayer_PROMISE(var_9, 0.0, 1273.674801757832, conv2d_11_w, -0.031951379626989365, 0.04218719156458998, conv2d_11_b, -0.3508028, 0.6397485, 1, 1, 1, 1, -1, 0, 1, 0.0, 652.76720800782, 9);
void* var_11 = ConvLayer_PROMISE(var_10, 0.0, 652.76720800782, conv2d_12_w, -0.028522676015272738, 0.03794213477522136, conv2d_12_b, -0.9171057, 0.7597668, 1, 1, 1, 1, -1, 0, 1, 0.0, 316.98977236938646, 9);
void* var_12 = ConvLayer_PROMISE(var_11, 0.0, 316.98977236938646, conv2d_13_w, -0.02821479567326605, 0.03854479804635069, conv2d_13_b, -0.50036746, 9.431553, 1, 1, 1, 1, 0, 2, 1, 0.0, 148.72470889282292, 9);
void* var_13 = FCLayer_PROMISE(var_12, 0.0, 148.72470889282292, dense_1_w, -0.007091613108757884, 0.008147951829247227, dense_1_b, -0.78005254, 0.8555075, 1, 0.0, 40.64329356002882, 9);
void* var_1 = ConvLayer_PROMISE(var_0, 0.0, 407.96143194580145, conv2d_2_w, -0.13156980648636818, 0.2164201746285022, conv2d_2_b, -1.0271513, 0.9052184, 1, 1, 1, 1, 0, 2, 1, 0.0, 1973.2054975586288, 9);
void* var_2 = ConvLayer_PROMISE(var_1, 0.0, 1973.2054975586288, conv2d_3_w, -0.18644111251831055, 0.202149114727974, conv2d_3_b, -0.17922063, 0.36547425, 1, 1, 1, 1, -1, 0, 1, 0.0, 2386.9648486329534, 9);
void* var_3 = ConvLayer_PROMISE(var_2, 0.0, 2386.9648486329534, conv2d_4_w, -0.10804861642420292, 0.12427636455744764, conv2d_4_b, -0.59533477, 0.63375777, 1, 1, 1, 1, 0, 2, 1, 0.0, 4998.494643554761, 9);
void* var_4 = ConvLayer_PROMISE(var_3, 0.0, 4998.494643554761, conv2d_5_w, -0.08040237371623515, 0.09835810117424044, conv2d_5_b, -0.20097896, 0.34949613, 1, 1, 1, 1, -1, 0, 1, 0.0, 4637.92161425807, 9);
void* var_5 = ConvLayer_PROMISE(var_4, 0.0, 4637.92161425807, conv2d_6_w, -0.05306418750435114, 0.06628044287860436, conv2d_6_b, -0.18124875, 0.274845, 1, 1, 1, 1, -1, 0, 1, 0.0, 4365.822572754019, 9);
void* var_6 = ConvLayer_PROMISE(var_5, 0.0, 4365.822572754019, conv2d_7_w, -0.05084674355760217, 0.07320860563218634, conv2d_7_b, -0.14288792, 0.59477174, 1, 1, 1, 1, 0, 2, 1, 0.0, 5600.749117676456, 9);
void* var_7 = ConvLayer_PROMISE(var_6, 0.0, 5600.749117676456, conv2d_8_w, -0.04523278899490833, 0.053042236261070186, conv2d_8_b, -0.14548235, 0.3148451, 1, 1, 1, 1, -1, 0, 1, 0.0, 3240.830364746551, 9);
void* var_8 = ConvLayer_PROMISE(var_7, 0.0, 3240.830364746551, conv2d_9_w, -0.02917514201253653, 0.03586270406842279, conv2d_9_b, -0.08428453, 0.18237582, 1, 1, 1, 1, -1, 0, 1, 0.0, 1895.9044943847766, 9);
void* var_9 = ConvLayer_PROMISE(var_8, 0.0, 1895.9044943847766, conv2d_10_w, -0.029496615380048753, 0.04047201693058028, conv2d_10_b, -0.19835947, 0.33766547, 1, 1, 1, 1, 0, 2, 1, 0.0, 1273.674801757832, 9);
void* var_10 = ConvLayer_PROMISE(var_9, 0.0, 1273.674801757832, conv2d_11_w, -0.031951379626989365, 0.04218719156458998, conv2d_11_b, -0.3508028, 0.6397485, 1, 1, 1, 1, -1, 0, 1, 0.0, 652.76720800782, 9);
void* var_11 = ConvLayer_PROMISE(var_10, 0.0, 652.76720800782, conv2d_12_w, -0.028522676015272738, 0.03794213477522136, conv2d_12_b, -0.9171057, 0.7597668, 1, 1, 1, 1, -1, 0, 1, 0.0, 316.98977236938646, 9);
void* var_12 = ConvLayer_PROMISE(var_11, 0.0, 316.98977236938646, conv2d_13_w, -0.02821479567326605, 0.03854479804635069, conv2d_13_b, -0.50036746, 9.431553, 1, 1, 1, 1, 0, 2, 1, 0.0, 148.72470889282292, 9);
void* var_13 = FCLayer_PROMISE(var_12, 0.0, 148.72470889282292, dense_1_w, -0.007091613108757884, 0.008147951829247227, dense_1_b, -0.78005254, 0.8555075, 1, 0.0, 40.64329356002882, 9);
void* var_14 = FCLayer_PROMISE(var_13, 0.0, 40.64329356002882, dense_2_w, -0.012781758182682096, 0.01437051862943929, dense_2_b, -0.012339931, 1.2154555, 1, 0.0, 11.167800696373025, 9);
void* var_15 = FCLayer_PROMISE(var_14, 0.0, 11.167800696373025, dense_3_w, -0.02119149128906429, 0.02715564412623694, dense_3_b, -0.773357, 0.6615543, -1, -7.4482048592567445, 17.882177452087543, 9);
void* var_16 = tensorSoftmax(var_15);
......
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment