Loading…
Loading…
Text generation · served with ONNX Runtime GenAI · ↓ 0 downloads a month · View on Hugging Face
Smaller versions (8-bit, 4-bit) need less GPU memory and answer a little less precisely.
The same model packaged differently: a quantized version needs less GPU memory and runs on a cheaper machine.
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| OriginalHuggingFaceH4/zephyr-7b-beta | vLLM | 14 GB | 19.5 GB | $0.18/h 2x RTX 3060 |
| oga (SamTheDev)SamTheDev/zephyr-7b-beta-oga | ONNX Runtime GenAI | 4.3 GB | 7.5 GB | $0.09/h Titan Xp |
Deploying ONNX models from the catalog is coming soon. Today, download the files from Hugging Face and upload them in Inference.
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.