-
Add comprehensive model card (README.md)
-
Add Qwen3VL-8B-Instruct FP8 quantized model
- Base model: Qwen/Qwen3-VL-8B-Instruct
- Quantization: FP8_DYNAMIC with SmoothQuant (smoothing_strength=0.8)
- Calibration dataset: lmms-lab/flickr30k (512 samples)
- Quantization method: llmcompressor with SmoothQuant modifier
- Excluded modules: lm_head, visual modules, mlp.gate
- Model format: safetensors (compressed)
- Max sequence length: 32768
This FP8 quantized model reduces memory footprint while maintaining model quality for vision-language tasks. -
initial commit
