Open elo-siema opened 1 month ago
I ran in to this problem too. It happens because in the install_mac.sh
script it uses make server > /dev/null
instead of make llama-server > /dev/null
You can manually go in to llama.cpp
and make llama-server
Looks like there's a pr under review to update the script https://github.com/dnhkng/GlaDOS/pull/77
For the best macOS support, you should use my branch at (https://github.com/Ghostboo-124/GlaDOS/tree/macOS)
@Ghostboo-124 I was wondering if you've ever run in to this error when running start_mac.sh
?
Traceback (most recent call last):
File "/Users/marissakohan/Development/GlaDOS/glados.py", line 21, in <module>
from glados import asr, tts, vad
File "/Users/marissakohan/Development/GlaDOS/glados/asr.py", line 6, in <module>
from . import whisper_cpp_wrapper
File "/Users/marissakohan/Development/GlaDOS/glados/whisper_cpp_wrapper.py", line 862, in <module>
_libs["whisper"] = load_library("whisper")
^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/marissakohan/Development/GlaDOS/glados/whisper_cpp_wrapper.py", line 548, in __call__
raise ImportError("Could not load %s." % libname)
ImportError: Could not load whisper.
It's searching the current paths for the model names, I've tried using both make libwhisper.so -j
and just make
but no luck so far, it isn't making a whisper
build afaict
Sorry for the ping, I ended up solving this. The .dylib ad the other files were built in whisper.cpp/src
. I copied libwhisper.dylib
to whisper.cpp
and it worked.
Fresh install with install_mac.sh, start_mac.sh