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
I have been preparing a pyproject.toml that depends on DGL and I want the installation to be platform-specific (as DGL does not support Windows anymore and the last supported version is 2.2.1, while on Linux the installation depends on the corresponding PyTorch version).
Due to the installation process similar to PyTorch, I found out the documentation section on PyTorch integration and replicated the same configuration.
However, I noticed only later that DGL suggest using the pip --find-links (-f) parameter to find the correct installation, as uv output when I tried to use uv.tool.sources with uv.tool.index was
hint: An index URL (https://data.dgl.ai/wheels/torch-2.4/repo.html)
could not be queried due to a lack of valid authentication credentials
(403 Forbidden).
I don't necessarily need DGL to work on a CUDA-environment as it is not used for heavy operations or training, so the result envisioned by the following example should be enough.
I didn't find any way in the documentation to achieve the same result, and using path or url does not help, as they expect a specific path to a source distribution file.
Instead, I would like uv to resolve automatically the dgl version as --find-links would do.
Could you help me?
Platform
Windows 11 x86_64
Version
uv 0.6.3
The text was updated successfully, but these errors were encountered:
Yeah this doesn't exist yet. The way we plan to do it is: allow --find-links style entries in [[tool.uv.index]]. Then you'd reference it as any other index in tool.uv.sources.
Question
I have been preparing a pyproject.toml that depends on DGL and I want the installation to be platform-specific (as DGL does not support Windows anymore and the last supported version is 2.2.1, while on Linux the installation depends on the corresponding PyTorch version).
Due to the installation process similar to PyTorch, I found out the documentation section on PyTorch integration and replicated the same configuration.
However, I noticed only later that DGL suggest using the pip --find-links (-f) parameter to find the correct installation, as uv output when I tried to use uv.tool.sources with uv.tool.index was
I don't necessarily need DGL to work on a CUDA-environment as it is not used for heavy operations or training, so the result envisioned by the following example should be enough.
I didn't find any way in the documentation to achieve the same result, and using
path
orurl
does not help, as they expect a specific path to a source distribution file.Instead, I would like uv to resolve automatically the dgl version as --find-links would do.
Could you help me?
Platform
Windows 11 x86_64
Version
uv 0.6.3
The text was updated successfully, but these errors were encountered: