Closed Datamance closed 1 month ago
Fyi 0.0.11 works for torch, but getting the same error for 0.0.13.
Hello @Datamance and @SvenLorenzTransfermedia, and thank you for giving TensorHue a try, creating and contributing to this issue, and proposing a solution!
The issue you described should now be fixed in TensorHue 0.0.15. The affected versions after 0.0.11 have been yanked from PyPi.
Part of the problem was indeed the missing __init__.py
file. My assumption was that this could be omitted due to the implicit namespace packages introduced in PEP 420, but it seems that this understanding was either wrong or not applicable when building with setuptools. My setup.cfg
tries to autodetect packages with
[options]
packages = find:
[options.packages.find]
include = tensorhue
which I also had to change to
[options]
packages = find:
[options.packages.find]
include = tensorhue*
to include the newly introduced connectors
. The problem never occurred when installing the package locally in editable mode (pip install -e .
), which I used for testing.
Feel free to reopen this issue in case this problem persists with v0.0.15, or open a new one if you'd like to see any features implemented in TensorHue!
Here's what I see when I try to import in an IPython session:
Don't you need an
__init__.py
file in the connectors directory for Python's module system to recognize it? I haven't tried to make a package in a while but I vaguely recall needing to do that.