Hello all,
Is there a way to train a model in mixed precision/ half-precision in libtorch using amp
?
If there is no amp
package in libtorch, can we somehow use the Nvidia Apex library in C++?
TIA
Hello all,
Is there a way to train a model in mixed precision/ half-precision in libtorch using amp
?
If there is no amp
package in libtorch, can we somehow use the Nvidia Apex library in C++?
TIA
@a_d
I don’t think we support amp in libtorch at this moment. @mcarilli
You can find some workaround in this post.
‘Deploy mixed precision model in libtorch’
I think Apex is a mix of its own libs and pytorch libs, not easy to use it purely from C++ I guess.