the model is generated by
```
cd llm-compressor/examples/autoround/quantization_w8a8_mxfp8
CUDA_VISIBLE_DEVICES=0 python qwen3_example.py
```
env
```
transformers                             4.57.6
llmcompressor                            0.10.1.dev165+ge688d87ef
compressed-tensors                       0.15.1.dev25+gc170de6
auto-round                               0.13.0.dev416+gbf1c81ce
```