---
license: apache-2.0
tags:
- onnx
- webgpu
- int4
---

# WebASMR ONNX graphs

Quantized graphs for the browser runtime:

- `llm_int4/` - Qwen3.5-0.8B + ASMR LoRA, merged and exported, MatMulNBits int4
- `tts_int4/` - F5-TTS DiT step + vocoder, int4 MatMul, convs in session dtype

`llm_fp16/` and `tts_fp16/` hold the fp16 exports they were quantized from.