You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As far as I can see, the packing of standalone Python local modules is unsupported. The only way I can see the packing to be picked up is by making the local modules into a local package and then specifying an absolute path to it. However, this isn't good enough as the absolute path is dependent on the local development environment (i.e. it would included my personal directory layout) and as far as I can see conda.yaml doesn't support relative paths to the location of the package. For portability of conda.yaml between environments this would necessitate releasing the local package to some sort of package index which is a fairly big obstacle, especially for private packages.
A current workaround would be to forgo creating a conda.yaml beforehand and packing the environment automatically (assuming #21 fallback works properly), but this then has the side-effect of having to manually specify the requirements using some other format (e.g. by contrast requirements.txt does support relative paths to local packages).
The text was updated successfully, but these errors were encountered:
Unfortunately after testing with requirements.txt file containing a relative path to a local package and with no presence of conda.yaml file, the save(model) command fails because it (rightfully) can't resolve the location of the installed local package.
As far as I can see, the packing of standalone Python local modules is unsupported. The only way I can see the packing to be picked up is by making the local modules into a local package and then specifying an absolute path to it. However, this isn't good enough as the absolute path is dependent on the local development environment (i.e. it would included my personal directory layout) and as far as I can see
conda.yaml
doesn't support relative paths to the location of the package. For portability ofconda.yaml
between environments this would necessitate releasing the local package to some sort of package index which is a fairly big obstacle, especially for private packages.A current workaround would be to forgo creating a
conda.yaml
beforehand and packing the environment automatically (assuming #21 fallback works properly), but this then has the side-effect of having to manually specify the requirements using some other format (e.g. by contrastrequirements.txt
does support relative paths to local packages).The text was updated successfully, but these errors were encountered: