lack of examples, qwn3 asr

[bchtrue]

Hello,

First of all: thanks for your work! Found your implementation that it almost only that currently support qwen3 asr.

But lack of examples for new users...

Am I right that qwen3asr default safetensors should be converted to .bin gglf with convert.py ? RIght?

It will be also useful to add examples (guide) how to setup everything and run.

Regards

[foldl]

1. download quantized model qwen3-asr-1.7b.bin from here
2. use it like this (-ngl all for GPU acceleration):

bin\main.exe -m /path/to/qwen3-asr-1.7b.bin -ngl all -i 
    ________          __  __    __    __  ___ 
   / ____/ /_  ____ _/ /_/ /   / /   /  |/  /_________  ____  
  / /   / __ \/ __ `/ __/ /   / /   / /|_/ // ___/ __ \/ __ \ 
 / /___/ / / / /_/ / /_/ /___/ /___/ /  / // /__/ /_/ / /_/ / 
 \____/_/ /_/\__,_/\__/_____/_____/_/  /_(_)___/ .___/ .___/  
You are served by Qwen3-ASR,                  /_/   /_/       
with 2031739904 (2.0B) parameters.

File > full/path/to/an/audio/file
language English<asr_text>Uh huh. Oh yeah, yeah. He wasn't even that big when I started listening to him, but and his solo music didn't do overly well, but he did very well when he started writing for other people.

File > 

[bchtrue]

Thank you! I will try ASAP

[bchtrue]

~/chatllm.cpp/build/bin/main --show_devices
 0: CUDA - CUDA0 (NVIDIA GeForce GTX 1080 Ti)
    type: GPU
    memory total: 11707809792 B
    memory free : 9144172544 B
 1: CUDA - CUDA1 (NVIDIA GeForce GTX 1080 Ti)
    type: GPU
    memory total: 11707809792 B
    memory free : 9144107008 B
 2: CUDA - CUDA2 (NVIDIA GeForce GTX 1080 Ti)
    type: GPU
    memory total: 11707809792 B
    memory free : 9144107008 B
 3: CUDA - CUDA3 (NVIDIA GeForce GTX 1080 Ti)
    type: GPU
    memory total: 11707809792 B
    memory free : 9144107008 B
[skip]
~/chatllm.cpp/build/bin/main  -m ~/spot/ai/models/qwen3-asr-1.7b.bin -ngl all -i
Eggml_backend_cuda_buffer_type_alloc_buffer: allocating 4.25 MiB on device 0: cudaMalloc failed: out of memory
/home/user/chatllm.cpp/src/backend.cpp:107 check failed (buf) alloc() failed to allocate buffer of size 4456448

Why it can happened?

[foldl]

You can try --max-length 8000 first. It's a little bit complex to use multiple GPUs.