CLOVER-Qwen2.5-VL

🧬 CLOVER: Cost-Effective Instruction Learning for Pathology Vision and Language Analysis

CLOVER-Qwen2.5-VL is an upgraded version of CLOVER, built upon Qwen2.5-VL, a powerful multimodal large language model for conversational pathology analysis. It efficiently adapts vision-language models (VLMs) to clinical domains with minimal computational cost while maintaining strong reasoning and question-answering capabilities.

🧾 This work is published in Nature Computational Science.

---

🚀 Quick Start

Installation

# Create environment
conda create -n clover python=3.10
conda activate clover

pip install torch==2.4.0 torchvision==0.19.0 --extra-index-url https://download.pytorch.org/whl/cu118

pip install transformers==4.52.4 accelerate qwen-vl-utils

Run

from transformers import Qwen2_5_VLForConditionalGeneration, AutoTokenizer, AutoProcessor
from qwen_vl_utils import process_vision_info
import torch
# default: Load the model and processer on the available device(s)
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
    "jline/CLOVER-Qwen2.5-VL", torch_dtype=torch.bfloat16, device_map="auto"
)
processor = AutoProcessor.from_pretrained("jline/CLOVER-Qwen2.5-VL")

messages = [
    {
        "role": "user",
        "content": [
            {
                "type": "image",
                "image": "./image_path.png",
            },
            {"type": "text", "text": "Describe this image."},
        ],
    }
]

# Preparation for inference
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",
)
inputs = inputs.to("cuda")

# Inference: Generation of the output
generated_ids = model.generate(**inputs, max_new_tokens=128)
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)

📚 Citation

If you find CLOVER useful for your research, please cite:

@article{chen2025cost,
  title={Cost-effective instruction learning for pathology vision and language analysis},
  author={Chen, K. and Liu, M. and Yan, F. and others},
  journal={Nature Computational Science},
  year={2025},
  doi={10.1038/s43588-025-00818-5}
}