shared library llama not found 0.1.39

[matthoffner]

I can't no longer get this project to work after upgrading from 0.1.38 to 0.1.39. I think its because I'm using the Langchain wrapper:

from langchain.llms import LlamaCpp

FileNotFoundError: Shared library with base name 'llama' not found

[jmtatsch]

Same here

[abetlen]

@matthoffner I don't think anything in the api has changed, can you try with v0.1.41? Are there any errors when you try to install with pip?

[matthoffner]

Will try the latest @abetlen - I think this is specific to how llama.cpp is installed on my computer and not Langchain.Not sure where llama is supposed to be installed, or if its something Mac specific.

[jmtatsch]

Actually it works well for me again with the current versions.
Turned out i had an issue with accidentally importing a local copy of the code instead of the installed package.

[abetlen]

Will try the latest @abetlen - I think this is specific to how llama.cpp is installed on my computer and not Langchain.Not sure where llama is supposed to be installed, or if its something Mac specific.

llama.cpp should be compiled and installed inside the llama-cpp-python package, the pypi package is a source distribution which includes llama.cpp and tries to build via cmake / make. Let me know if you figure out what's causing the issue.

[matthoffner]

Fixed with latest, thanks

[rok-GKC]

Installed 0.1.48 and get "FileNotFoundError: Shared library with base name 'llama' not found"

The folder ./vendor/llama.cpp/ is empty

Any ideas?

[abetlen]

@rok-GKC are you installing from PyPI or Github?

[rok-GKC]

Installed using pip install llama-cpp-python in wsl (Ubuntu)

[abetlen]

And otherwise no errrors during the pip installation?

[rok-GKC]

No errors.

[rok-GKC]

I tired again today with 0.1.49 and it worked.