Hi. I am training a neural network that uses a sequence as the input instead of an image. How to conduct preprocess like convert into tensor? Are there any demo inferring fully connected network?
slwdgvem1#
Are you training a neural network with ncnn?
uttx8gqw2#
No. Trained with pytorch and converted to ncnn.
ahy6op9u3#
try constructor
// external vec Mat(int w, void* data, size_t elemsize = 4u, Allocator* allocator = 0);
usage:
float* data_ptr; Mat in(data_length, data_ptr);
3wabscal4#
How is this done in python API?
yrwegjxp5#
Does the input need to be flattened before going in?
5条答案
按热度按时间slwdgvem1#
Are you training a neural network with ncnn?
uttx8gqw2#
Are you training a neural network with ncnn?
No. Trained with pytorch and converted to ncnn.
ahy6op9u3#
try constructor
usage:
3wabscal4#
How is this done in python API?
yrwegjxp5#
Does the input need to be flattened before going in?