davrot/pytorch-sbs
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Add files via upload

Browse source
This commit is contained in:
David Rotermund 2023-02-06 09:57:25 +01:00 committed by GitHub
parent 6477fb29e2
commit 1a807b44df
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 67 additions and 52 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

119
network/h_dynamic_cnn_gpu_cpp_v1/HDynamicCNNGPU.cu
View file

@ -3,11 +3,12 @@
#include <omp.h> #include <omp.h>
#include <stdio.h> #include <stdio.h>
#include <string.h> #include <string.h>
#include <chrono>
#include <algorithm> #include <algorithm>
#include <cassert> #include <cassert>
#include <iostream> #include <iostream>
// #define DEBUGSHOWTIMEGLOBAL
HDynamicCNNGPU::HDynamicCNNGPU() HDynamicCNNGPU::HDynamicCNNGPU()
{ {
@ -111,6 +112,11 @@ void HDynamicCNNGPU::entrypoint(
// -------------------- // --------------------
assert((number_of_processes <= 0)); assert((number_of_processes <= 0));
#ifdef DEBUGSHOWTIMEGLOBAL
using TIME_resolution = std::chrono::nanoseconds;
auto TIME_start = std::chrono::high_resolution_clock::now();
#endif
gpu_update( gpu_update(
h_init_ptr, h_init_ptr,
h_pointer, h_pointer,
@ -136,6 +142,12 @@ void HDynamicCNNGPU::entrypoint(
number_of_pattern, number_of_pattern,
gpu_tuning_factor); gpu_tuning_factor);
#ifdef DEBUGSHOWTIMEGLOBAL
auto TIME_end = std::chrono::high_resolution_clock::now();
float TIME_measured = TIME_resolution(TIME_end - TIME_start).count();
std::cout << "Time used : " << TIME_measured/(1000.0*1000.0) << "ms" << std::endl;
#endif
return; return;
}; };
@ -193,66 +205,69 @@ __device__ void gpu_update_one_ip(
spike = input_pointer + counter_spike * input_dim_c1; spike = input_pointer + counter_spike * input_dim_c1;
if (*spike >= 0) if (*spike < 0)
{ {
if (epsilon_xy_dim_c0 != 0) break;
}
if (epsilon_xy_dim_c0 != 0)
{
epsilon_subsegment =
epsilon_xy_pointer[*spike *epsilon_xy_dim_c0] * epsilon_t_pointer[counter_spike];
}
else
{
epsilon_subsegment = epsilon_t_pointer[counter_spike];
}
w_ptr = weights_pointer + *spike * weights_dim_c0;
for (size_t counter = 0; counter < h_dim; counter++)
{
h_temp[counter] = h_subsegment[counter] * w_ptr[counter];
}
h_temp_sum = 0.0;
for (size_t counter = 0; counter < h_dim; counter++)
{
h_temp_sum += h_temp[counter];
}
if (h_temp_sum > 1E-10)
{
temp_value = epsilon_scale * epsilon_subsegment / h_temp_sum;
for (size_t counter = 0; counter < h_dim; counter++)
{ {
epsilon_subsegment = h_temp[counter] *= temp_value;
epsilon_xy_pointer[*spike *epsilon_xy_dim_c0] * epsilon_t_pointer[counter_spike]; }
for (size_t counter = 0; counter < h_dim; counter++)
{
h_subsegment[counter] += h_temp[counter];
}
if (forgetting_offset_local > 0.0)
{
temp_value =
epsilon_scale * epsilon_subsegment * forgetting_offset_local;
for (size_t counter = 0; counter < h_dim; counter++)
{
h_subsegment[counter] += temp_value;
}
epsilon_scale *=
1.0 + epsilon_subsegment * (1.0 + forgetting_offset);
} }
else else
{ {
epsilon_subsegment = epsilon_t_pointer[counter_spike]; epsilon_scale *= 1.0 + epsilon_subsegment * 1.0;
}
w_ptr = weights_pointer + *spike * weights_dim_c0;
for (size_t counter = 0; counter < h_dim; counter++)
{
h_temp[counter] = h_subsegment[counter] * w_ptr[counter];
}
h_temp_sum = 0.0;
for (size_t counter = 0; counter < h_dim; counter++)
{
h_temp_sum += h_temp[counter];
}
if (h_temp_sum > 1E-10)
{
temp_value = epsilon_scale * epsilon_subsegment / h_temp_sum;
for (size_t counter = 0; counter < h_dim; counter++)
{
h_temp[counter] *= temp_value;
}
for (size_t counter = 0; counter < h_dim; counter++)
{
h_subsegment[counter] += h_temp[counter];
}
if (forgetting_offset_local > 0.0)
{
temp_value =
epsilon_scale * epsilon_subsegment * forgetting_offset_local;
for (size_t counter = 0; counter < h_dim; counter++)
{
h_subsegment[counter] += temp_value;
}
epsilon_scale *=
1.0 + epsilon_subsegment * (1.0 + forgetting_offset);
}
else
{
epsilon_scale *= 1.0 + epsilon_subsegment * 1.0;
}
} }
} }
} }
temp_value = 1.0 / epsilon_scale; temp_value = 1.0 / epsilon_scale;
for (size_t counter = 0; counter < h_dim; counter++) for (size_t counter = 0; counter < h_dim; counter++)