GPU Timeout Error macos x86-64

[Moshe1010]

What happened?

A bug happened!

Steps to reproduce

1. step one...
2. step two...

What OS are you seeing the problem on?

MacOS 15.2 (24C101)
3.6 GHz 10-Core Intel Core i9
AMD Radeon Pro 5700 XT 16 GB
64 GB 2667 MHz DDR4
1TB SSD

Relevant log output

options: {
  "path": "/Users/ov/Library/CloudStorage/Dropbox/7:3:22.mp3",
  "lang": "he",
  "verbose": false,
  "n_threads": 4,
  "init_prompt": "",
  "temperature": 0.4,
  "translate": null,
  "max_text_ctx": null,
  "word_timestamps": false,
  "max_sentence_len": 1
}

Caused by:
   0: failed to transcribe
   1: Generic whisper error. Varies depending on the function. Error code: -6

Location:
    core/src/transcribe.rs:303:38
App Version: vibe 3.0.0
Commit Hash: 371af90
Arch: x86_64
Platform: macos
Kernel Version: 15.2.0
OS: macos
OS Version: 15.2.0
Cuda Version: n/a
Models: ggml-medium-q8_0.bin, ggml-ivrit-v2-d4.bin
Default Model: ggml-ivrit-v2-d4.bin"
Cargo features: 


CPU feature detection is not supported on this architecture.


<details>
<summary>logs</summary>


cmd: "/Applications/vibe.app/Contents/MacOS/../Resources/ffmpeg" "-i" "/Users/ov/Library/CloudStorage/6-Law/1-\xd7\x9e\xd7\xa 7:3:22.mp3" "-ar" "16000" "-ac" "1" "-c:a" "pcm_s16le" "/var/folders/yd/vwqjk5011p3flvpc66nhg36h0000gn/T/vibe_temp_2024-12-26/bfd73086b7599e22.wav" "-hide_banner" "-y" "-loglevel" "error"
whisper_init_state: failed to load Core ML model from '/Users/ov/Library/Application Support/github.aaakk.us.kg.thewh1teagle.vibe/ggml-ivrit-v2-d4-encoder.mlmodelc'
error: Caused GPU Timeout Error (00000002:kIOAccelCommandBufferCallbackErrorTimeout)
whisper_full_with_state: failed to encode

</details>

[thewh1teagle]

Hey
Please try with the default model that comes with Vibe and let me know if it works (switch back in the settings)

[Moshe1010]

It seems the same. On my M1 Max Macbook Pro 2021 it works fine. Weird...

options: {
"path": "/Users/",
"lang": "he",
"verbose": false,
"n_threads": 4,
"init_prompt": "",
"temperature": 0.4,
"translate": null,
"max_text_ctx": null,
"word_timestamps": false,
"max_sentence_len": 1
}

Caused by:
0: failed to transcribe
1: Generic whisper error. Varies depending on the function. Error code: -6

Location:
core/src/transcribe.rs:303:38

[thewh1teagle]

It seems the same. On my M1 Max Macbook Pro 2021 it works fine. Weird...

What if you run whisper.cpp from the source? -

cd /tmp
rm -rf whisper.cpp
git clone https://github.com/ggerganov/whisper.cpp -b v1.7.3 --depth 1
cd whisper.cpp
cmake -B build . -DCMAKE_BUILD_TYPE=Release
cmake --build build --target main --config Release
wget https://github.com/thewh1teagle/vibe/raw/main/samples/short.wav
wget https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-tiny-q8_0.bin
reset
./build/bin/main -m ggml-tiny-q8_0.bin -f short.wav

You may need to install these packages before:

brew install llvm cmake

[Helwar]

I have the same issue on a MBP 2019 (i9). Compiling and running whisper from source as instructed works. Also: if the mlmodelc is installed that error does not occur for ggml-medium-q8_0.bin but it does occur for ggml-large-v3.bin.