Re: llama.cpp, whisper.cpp, ggml: Next steps
[Christian Kastner]
> I managed to reproduce this last week, but now no longer can. Odd, I
> did push some ggml changes, but nothing that seems related or even
> noteworthy.
Strange. I see the same. My patch look like this:
diff --git a/debian/rules b/debian/rules
index 040b6e6..3cc12bf 100755
--- a/debian/rules
+++ b/debian/rules
@@ -15,6 +15,8 @@ override_dh_auto_configure:
-DCMAKE_LIBRARY_ARCHITECTURE="$(DEB_TARGET_MULTIARCH)" \
$(GPUSUPPORT) \
-DWHISPER_SDL2=1 \
+ -DCMAKE_INSTALL_RPATH=/usr/lib/$(DEB_TARGET_MULTIARCH)/ggml \
+ -DCMAKE_INSTALL_RPATH_USE_LINK_PATH=OFF \
-DCMAKE_BUILD_TYPE=Release
# FIXME figure out why test failed. The crash can be demonsrated using
> I built ggml, and installed libggml and libggml-cpu. I then started up
> a container and installed these packages; in that container I then
> built whisper.cpp with the above patch.
I tested with both libggml-cpu and libggml-hip installed, in case there
were a difference there, as I used the later earlier, but now ldd find
the libraries for both, and the binaries start as they should:
% whisper-stream
init: found 1 capture devices:
init: - Capture device #0: 'Family 17h/19h/1ah HD Audio Controller Analog stereo'
init: attempt to open default capture device ...
init: obtained spec for input device (SDL Id = 2):
init: - sample rate: 16000
init: - format: 33056 (required: 33056)
init: - channels: 1 (required: 1)
init: - samples per frame: 1024
whisper_init_from_file_with_params_no_state: loading model from 'models/ggml-base.en.bin'
whisper_init_from_file_with_params_no_state: failed to open 'models/ggml-base.en.bin'
Segmentation fault
%
--
Happy hacking
Petter Reinholdtsen
Reply to: