Qwen2.5-VL-3B-Instruct-per-grp-quant

• Introduction

  This model was quantized using Quark 0.11 amd_quark-0.11

• Quantization Strategy

  • Quantized Layers: All linear layers
  • Weight: uint4 asymmetric per-group with group_size=128.

• Quick Start

1. Download the model
2. Run the quantization script in the example folder using the following command line:

   python run_qwen2_5_vl_quantization.py
   

Evaluation

Quark currently uses perplexity(PPL) as the evaluation metric for accuracy loss before and after quantization.The specific PPL algorithm can be referenced in the quantize_quark.py. The quantization evaluation results are conducted in pseudo-quantization mode, which may slightly differ from the actual quantized inference accuracy. These results are provided for reference only.

Evaluation scores

Benchmark	Qwen2.5-VL-3B-Instruct	Qwen2.5-VL-3B-Instruct-per-grp-quant(this model)
Perplexity-wikitext2	11.1107	13.7743

License

Modifications copyright(c) 2024 Advanced Micro Devices,Inc. All rights reserved.