Qwen2.5 VL 3B Instruct GPTQ Int3

🚀 Qwen2.5-VL-3B-Instruct-GPTQ-Int3

This project offers an UNOFFICIAL GPTQ-Int3 quantized version of the Qwen2.5-VL model, leveraging the gptqmodel library. It's fully compatible with the latest transformers library, which also supports non - quantized Qwen2.5-VL models.

✨ Features

• Quantization: Utilizes GPTQ-Int3 quantization to reduce model size and enhance computational efficiency.
• Compatibility: Works seamlessly with the transformers library, facilitating easy integration.

📦 Installation

Install the required libraries:

pip install git+https://github.com/huggingface/transformers accelerate qwen-vl-utils
pip install git+https://github.com/huggingface/optimum.git
pip install gptqmodel 

Optionally, you may need to install:

pip install tokenicer device_smi logbar

💻 Usage Examples

Basic Usage

from transformers import Qwen2_5_VLForConditionalGeneration, AutoTokenizer, AutoProcessor
from qwen_vl_utils import process_vision_info

model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
    "hfl/Qwen2.5-VL-3B-Instruct-GPTQ-Int4", 
    attn_implementation="flash_attention_2",
    device_map="auto"
)
processor = AutoProcessor.from_pretrained("hfl/Qwen2.5-VL-3B-Instruct-GPTQ-Int4")

messages = [{
    "role": "user",
    "content": [
        {"type": "image", "image": "https://raw.githubusercontent.com/ymcui/Chinese-LLaMA-Alpaca-3/refs/heads/main/pics/banner.png"},
        {"type": "text", "text": "请你描述一下这张图片。"},
    ],
}]

text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
image_inputs, video_inputs = process_vision_info(messages)
inputs = processor(
    text=[text], images=image_inputs, videos=video_inputs,
    padding=True, return_tensors="pt",
).to("cuda")

generated_ids = model.generate(**inputs, max_new_tokens=512)
generated_ids_trimmed = [out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)]
output_text = processor.batch_decode(generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False)
print(output_text[0])

Output Example

Response:

This image shows a logo in both Chinese and English, reading "中文LLaMA & Alpaca大模型" (Chinese LLaMA & Alpaca Large Language Models). On the left side of the logo, there are two cartoon characters: one is an alpaca with a red scarf, and the other is an alpaca with white fur. The background features a green grassland and a building with a red roof. On the right side of the logo, there is a number 3, accompanied by some circuit patterns. The overall design is simple and clear, using bright colors and cute cartoon characters to attract attention.

📚 Documentation

Performance

Note

⚠️ Important Note

• Evaluations are performed using lmms-eval with default setting.
• GPTQ models are computationally more effective (fewer VRAM usage, faster inference speed) than AWQ series in these evaluations.
• We recommend use gptqmodel instead of autogptq library, as autogptq is no longer maintained.

💡 Usage Tip

For detailed usage, please check Qwen2.5-VL's page.

Disclaimer

⚠️ Important Note

This is NOT an official model by Qwen. Use at your own risk.

📄 License

This project is licensed under the Apache-2.0 license.