Hi,
I set up the python environment for SDK5.0 yesterday and am getting
File “pyzed\sl.pyx”, line 1, in init pyzed.sl
ValueError: numpy.dtype size changed, may indicate binary incompatibility. Expected 96 from C header, got 88 from PyObject
I tried numpy version 2.2.4 as first installed and then 2.0.0 as the minimum requirement from the wrapper. Both give the same issue. Has anyone else faced this yet and got a solution?
I downgraded manually to numpy 1.26.4 and it worked for me. However, this now causes a conflict with pytorch and torchvision. Seems like the upgrade to numpy 2.0.0 is going to be an ongoing issue.
Worth noting this is an ongoing issue - installing the 5.0 SDK means adding a dependency conflict into your project (pyzed depends on numpy 2.0.0, but doesn’t actually work with it due to the bug in this thread). This will cause headaches for CI pipelines and environments set up with strict dependency resolvers like poetry.
Hello @schluppsel and @diego,
We improved our pyzed installation and the python script you use to install pyzed will be fixed in the next sdk release. The underlying issue was upgrading the pyzed package.
If you still have some issues while installing pyzed with the new release, please do not hesitate to reach us again.
Best regards,