Views
No views yet
large-v3, float16). A 4 GB laptop GPU auto-falls back to small or CPU.1git clone https://huggingface.co/mohameddalii/translation-audio
2cd translation-audio
3chmod +x scripts/run_vast.sh
4./scripts/run_vast.sh1python -m audio sample.wav --transcribe-only
2# or
3curl -s http://127.0.0.1:8000/health
4curl -F "file=@sample.wav" -F "translate=false" http://127.0.0.1:8000/translatepython -m audio.loadtest --file sample.wav --users 8 --requests 40 --no-translate1pip install locust
2locust -f locustfile.py --host http://127.0.0.1:8000 --web-host 0.0.0.0 --web-port 8089https://<your-jupyter-host>/proxy/8089/. Start with 4 users, spawn rate 1. The GPU serializes ASR, so extra users wait or get 429.scripts/run_vast.sh):| Variable | 5090 value |
|---|---|
ASR_BACKEND | whisper or cohere |
ASR_MODEL | large-v3 (Whisper only) |
ASR_DEVICE | cuda |
ASR_COMPUTE_TYPE | float16 (do not use int8 on Blackwell) |
SKIP_TRANSLATION | true until the text service is up |
1python -m audio sample.wav --transcribe-only --backend whisper > whisper.json
2python -m audio sample.wav --transcribe-only --backend cohere > cohere.json
3python -m audio.compare whisper.json cohere.jsonaudio/ — ASR, ffmpeg video→wav, FastAPI, GPU queue, load testshared/contracts.py — payload sent to the text layertext/stub.py — stand-in for Buthaina's translator