I learned the hard way that GLFS must be installed to clone the repo and added a hint to the README.
Comment From: tzolov
Hi @fabapp2,
The intention is that the GLFS would be required only if you want to access, locally, the default ONNX models, because of their large size.
Even without GLFS the TransformersEmbeddingClient looks up the default models directly from the GH site.
Are you experiencing something else?
Comment From: fabapp2
Hi @tzolov,
thanks for the information.
I couldn't simply git clone.
Resolving deltas: 100% (1389/1389), done.
git-lfs filter-process: git-lfs: command not found
fatal: the remote end hung up unexpectedly
warning: Clone succeeded, but checkout failed.
Comment From: tzolov
@fabapp2 , thank you for the update,
Apparently, by default, the repository demands the git-lfs and you have set export GIT_LFS_SKIP_SMUDGE=1 (for linux, macos) or set GIT_LFS_SKIP_SMUDGE=1 (for windows) to disable it,
or for single clone (GIT_LFS_SKIP_SMUDGE=1 git clone git@github.com:spring-projects/spring-ai.git)
I've incorrectly assumed that it is disabled by default.
Given this, I'm not sure what is better - (1) document the special, git clone instructions to either disable LFS or install git-lfs or - (2) move the ONNX files to a different GH repository.
Comment From: fabapp2
I'd move the ONNIX file to keep the expected default behavior of git clone without the need for additional documentation.
Comment From: tzolov
@fabapp2, I second your opinion. But, as it is not easy to get another spring-project GH dedicated for only for ONNX zoo, we need to figure out how to handle this. Until then i will merge your hint. Thank you!
Comment From: fabapp2
@fabapp2, I second your opinion. But, as it is not easy to get another spring-project GH dedicated for only for ONNX zoo, we need to figure out how to handle this. Until then i will merge your hint. Thank you!
I had a feeling... Thanks for merging the PR.