You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
in reduction.cpp when dim == 2
clearly there is a bug since b.channel is set to be 1, but in the loop it is assumed to still have the same number of channels as a, as in const float* ptr = a.channel(q); float* outptr = b.channel(q);
else if (dim == 2)
{
// w h c -> X h c
b.create(h, channels, elemsize, opt.blob_allocator);
}
else if (dim == 2)
{
#pragma omp parallel for num_threads(opt.num_threads)
for (int q=0; q<channels; q++)
{
const float* ptr = a.channel(q);
float* outptr = b.channel(q);
for (int i=0; i<h; i++)
{
float sum = v0;
for (int j=0; j<w; j++)
{
sum = op(sum, ptr[i]);
}
outptr[i] = sum * coeff;
ptr += w;
}
}
}
The text was updated successfully, but these errors were encountered:
in reduction.cpp when
dim == 2
clearly there is a bug since
b.channel
is set to be 1, but in the loop it is assumed to still have the same number of channels asa
, as inconst float* ptr = a.channel(q); float* outptr = b.channel(q);
The text was updated successfully, but these errors were encountered: