from transformers import Qwen2_5_VLForConditionalGeneration, AutoProcessor
from qwen_vl_utils import process_vision_info

model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
    "qwangcv/VDC-Agent-7B", torch_dtype="auto", device_map="auto"
)
processor = AutoProcessor.from_pretrained("qwangcv/VDC-Agent-7B")

messages = [
    {
        "role": "user",
        "content": [
            {
                "type": "video",
                "video": "file:///path/to/video.mp4",
            },
            {"type": "text", "text": "Describe the video in detail."},
        ],
    }
]

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(model.device)

generated_ids = model.generate(**inputs, max_new_tokens=1024)
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])

Training Data

Base model: Qwen2.5-VL-7B-Instruct
Preference data: VDC-Agent-19K (~19K DPO pairs for detailed video captioning)

Acknowledgement

We sincerely thank the Qwen team for releasing the excellent Qwen2.5-VL models, which make this work possible.

Citation

If you find this model useful, please consider citing our paper:

bibtex

@article{vdcagent,
  title={VDC-Agent: When Video Detailed Captioners Evolve Themselves via Agentic Self-Reflection},
  author={Wang, Qiang and Gao, Xinyuan and He, Yuhang and Han, Jizhou and Li, Jiangyang and Dong, Songlin and Ma, Zhiheng and Gong, Yihong},
  journal={arXiv preprint arXiv:2511.19436},
  year={2025}
}