Open theonetwoone opened 8 months ago
I got the same ModuleNotFoundError
which I resolved by installing llama-cpp-python
inside ComfyUI
folder.
Here's the fix:
cd ComfyUI
pip install llama-cpp-python
llama-cpp-python by @abetlen is a Python binding for llama.cpp by @ggerganov.
@slavakurilyak thanks for the tipp, but it didn't work for me. llama-cpp-python is installed now but can't be found. This repo needs a decent installation guide or better yet handle installs on its own.
I found that it's likely a problem only for pre compiled comfyui users. Installing llamacpp manually then copy the dependencies from %appdata% local/programs/python/python310/lib/site-packages into the comfyui/python_embeded/lib/site-packages, fixes the the node works fine after
I mean, yyeah one way to get rid of open issues is just closing them. 🤷
@theonetwoone I appreciate your help but that does not work either.
manually then copy the dependencies from
You must be kidding, right?
this will end in an endless manual moving of dependencies
ModuleNotFoundError: No module named 'diskcache'
Sorry didn't mean to close it my bad!
I got it to work by manually copying all dependencies (for me only 'diskcache' was remaining). But something is still off, it takes 20s (so I gguess it's running on CPU).
This is a great tool if we put in some more work.
llama-cpp-python with GPU support: CUDA 11.8 .\python_embeded\python.exe -m pip install llama-cpp-python --prefer-binary --no-cache-dir --extra-index-url=https://jllllll.github.io/llama-cpp-python-cuBLAS-wheels/AVX2/cu118/
CUDA 12.2 .\python_embeded\python.exe -m pip install llama-cpp-python --prefer-binary --no-cache-dir --extra-index-url=https://jllllll.github.io/llama-cpp-python-cuBLAS-wheels/AVX2/cu122
In my case the llama installation fails directly:
PS only now I see that to install that llama, VisualStudio is required.
After installing and moving the models to the right folder I still get this when starting Comfyui: