Loading…
Loading…
Speech to text · served with ONNX Runtime · ↓ 434 downloads a month · View on Hugging Face
Quran-Lab/zipformer_p-arabic-v3 needs about 1.4 GB of GPU memory. It runs on one Quadro P4000 from $0.07/h right now; best value: RTX 5090 at $0.65/h.
The model's publisher asks users to accept its licence (other). GridPup cannot download this one yet: choose an ungated version below.
These ONNX files are an export for running in a browser (an audio encoder and a text decoder, each a version here): served as tensors, they do not make a transcript. For speech to text, pick the original model (for example openai/whisper-base), served with vLLM.
Smaller versions (8-bit, 4-bit) need less GPU memory and answer a little less precisely.
The publisher has not let GridPup download this licence-gated model yet. An ungated version of it (above) can be deployed.
Prices are per hour for the whole machine and its disk, right now. Speed is Vast's DLPerf score of the machine, relative to the cheapest. Best value: the most speed per dollar. Only machines with a Vast reliability above 95% that download at 400 Mbit/s or more are offered: the machine fetches the model itself.