nishegde/ratio-filter-demo
0
Guard SLM Ratio Filter
This Static Space runs the real quantized Qwen3 1.7B classifier in the visitor's browser with Transformers.js and WebGPU. It uses no hosted inference API and sends no prompt text to a server after the public model files are downloaded.
Requirements: recent desktop Chrome or Edge with WebGPU, a one-time approximately 1.08 GB model download, and sufficient RAM/VRAM. Mobile and low-memory devices are not supported.
- Adapter: `nishegde/ratio-filter-qwen3-1.7b-lora`
- Browser model: `nishegde/ratio-filter-qwen3-1.7b-onnx`
- Source: github.com/nischayhegde/SLMfinetune