Doesn't work in Electron.js's Node

[neo773]

Hi,

I'm trying to use this in Electron but it fails when transcribing with a vague error, I Googled it but didn't find anything

aheads_masks_init: dtw_n_top must be between 1 and 4 for this model.whisper_init_state: aheads_masks_init() failed for alignment heads masks

I have tried different models and toggling GPU option but no luck so far.

Full log:

whisper_init_from_file_with_params_no_state: loading model from '/Users/neo/.smart-whisper/models/ggml-large-v3-turbo-q8_0.bin'
whisper_init_with_params_no_state: use gpu    = 0
whisper_init_with_params_no_state: flash attn = 0
whisper_init_with_params_no_state: gpu_device = 0
whisper_init_with_params_no_state: dtw        = 16
whisper_model_load: loading model
whisper_model_load: n_vocab       = 51866
whisper_model_load: n_audio_ctx   = 1500
whisper_model_load: n_audio_state = 1280
whisper_model_load: n_audio_head  = 20
whisper_model_load: n_audio_layer = 32
whisper_model_load: n_text_ctx    = 448
whisper_model_load: n_text_state  = 1280
whisper_model_load: n_text_head   = 20
whisper_model_load: n_text_layer  = 4
whisper_model_load: n_mels        = 128
whisper_model_load: ftype         = 7
whisper_model_load: qntvr         = 2
whisper_model_load: type          = 5 (large v3)
whisper_model_load: adding 1609 extra tokens
whisper_model_load: n_langs       = 100
whisper_model_load:      CPU total size =   873.55 MB
whisper_model_load: model size    =  873.55 MB

whisper_print_timings:     load time =   342.46 ms
whisper_print_timings:    total time =   342.48 ms
whisper_init_state: kv self size  =   31.46 MB
whisper_init_state: kv cross size =   31.46 MB
whisper_init_state: kv pad  size  =    7.86 MB
aheads_masks_init: dtw_n_top must be between 1 and 4 for this model.whisper_init_state: aheads_masks_init() failed for alignment heads masks
✨  Done in 10.68s.

[JacobLinCool]

hello, can you provide the code and options you used?

[BarrySong97]

hello, can you provide the code and options you used?

I have the same issue, I just used example code in electron.

      const pcm = read_wav(audioPath);

      // Initialize Whisper with model and advanced configuration
      this.currentWhisper = new Whisper(modelPath, {
        offload: 10,
        gpu: true, // Enable GPU acceleration if available
      });

      // Start transcription with detailed configuration
      const task = await this.currentWhisper.transcribe(pcm, {
        language: "auto",
      });

My computer is Macbook pro m2 pro.

[BarrySong97]

After I modified the code in whisper.cpp, I set dtw_token_timestamps false.
I can use it in a nodejs script, But I can't use it in electron.

---

it turns out

ggml_gallocr_alloc_graph: reallocating buffers automatically
ggml_gallocr_needs_realloc: node node_27 is not valid
ggml_gallocr_alloc_graph: reallocating buffers automatically

this will terminate my electron app. But when I task.on('transcribe'), It won't be terminated.