Opened 2 years ago
Last modified 22 months ago
#66359 new defect
py310-torchvision does not build
Reported by: | garbeware | Owned by: | |
---|---|---|---|
Priority: | Normal | Milestone: | |
Component: | ports | Version: | |
Keywords: | Cc: | ||
Port: | py310-torchvision |
Description
The building process fails with a line complaining that affine_quantizer.h
cannot be found.
This can be fixed by editing /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_macports_release_tarballs_ports_python_py-torchvision/py310-torchvision/work/pytorch-vision-da3794e/torchvision/csrc/ops/quantized/cpu/qnms_kernel.cpp
and replacing affine_quantizer.h
in the second line with AffineQuantizer.h
The installation then goes through without other problems.
Attachments (1)
Change History (4)
comment:1 Changed 2 years ago by ryandesign (Ryan Carsten Schmidt)
comment:2 Changed 2 years ago by garbeware
main.log was generated by uninstalling the port and trying to install it again. The same procedure that I described above resulted again in a successful install. Please let me know if you need any other information.
comment:3 Changed 22 months ago by rharriszzz (Richard Harris)
pytorch 1.12 uses affine_quantizer.h
pytorch 1.13 uses AffineQuantizer.h
torchvision 0.13 is pinned to pytorch 1.12
torchvision 0.14 is pinned to pytorch 1.13
this means that py310-torchvision needs to refer to torchvision 0.14
Please attach the main.log file.