Bug: Vibe quit unexpectedly.

[drsmikle]

What happened?

A bug happened!
Vibe quit unexpectedly.

Steps to reproduce

I tried small model with suggested small .wav file and it still quit.
vibe crash apple report.txt.zip

What OS are you seeing the problem on?

MacOS

Relevant log output

App Version: vibe 2.5.2
Commit Hash: 7780d6fa971ef096a21e4ec84dc3330172f8a241
Arch: x86_64
Platform: macos
Kernel Version: 12.7.6
OS: macos
OS Version: 12.7.6
Cuda Version: n/a
Models: ggml-medium.bin, ggml-small.bin
Default Model: ggml-small.bin"
Cargo features: 


CPU feature detection is not supported on this architecture.

[drsmikle]

Vibe never worked for me. I installed it and tried it and it quit on its own.

[thewh1teagle]

Thanks for report
Looks like whisper.cpp engine failed to transcribe it
Please try to run it from the terminal with logging enabled and try to transcribe.
See docs/debug.md

[drsmikle]

From terminal I get this feedback:
2024-09-17T11:06:36.868138Z ERROR whisper_rs::whisper_sys_tracing: whisper_backend_init_gpu: Metal GPU does not support family 7 - falling back to CPU
2024-09-17T11:06:37.027644Z ERROR whisper_rs::whisper_sys_tracing: whisper_init_state: failed to load Core ML model from '/Users/RussMac/Library/Application Support/github.com.thewh1teagle.vibe/ggml-medium-encoder.mlmodelc'
Does it matter that the model is named /ggml-medium.bin at the above folder path instead of ggml-medium-encoder.mlmodelc?

[IvanRublev]

Hi @thewh1teagle, same unable to load Core ML model issue here on macOS 12.7.6.

The following whisper.cpp command transcribes perfectly, with both model file named ggml-tiny.bin or ggml-tiny-encoder.mlmodelc

whisper-cpp --output-txt --model /Users/ivan/Library/Application\ Support/github.com.thewh1teagle.vibe/ggml-tiny.bin /Users/ivan/Downloads/samples_single.wav

whisper-cpp was installed via brew install whisper-cpp.

[thewh1teagle]

From terminal I get this feedback: 2024-09-17T11:06:36.868138Z ERROR whisper_rs::whisper_sys_tracing: whisper_backend_init_gpu: Metal GPU does not support family 7 - falling back to CPU 2024-09-17T11:06:37.027644Z ERROR whisper_rs::whisper_sys_tracing: whisper_init_state: failed to load Core ML model from '/Users/RussMac/Library/Application Support/github.com.thewh1teagle.vibe/ggml-medium-encoder.mlmodelc' Does it matter that the model is named /ggml-medium.bin at the above folder path instead of ggml-medium-encoder.mlmodelc?

Not sure why it failed. generally you can ignore the warning saying it failed to load mlmodelc file and it should work anyway

[thewh1teagle]

Hi @thewh1teagle, same unable to load Core ML model issue here on macOS 12.7.6.

Can you reference or attach the exact error message you got from it?

[IvanRublev]

Sure @thewh1teagle, please find app's log and macos crash report attached.
log_2024-10.txt
macos_crash_report.txt

[IvanRublev]

The app is used fully offline.

[thewh1teagle]

#370