[Question]: When I select an img2txt model for PDF document parsing, what task should I choose when deploying this model?

[Dreamcatcher-wind]

Self Checks

• I have searched for existing issues search for existing issues, including closed ones.
• I confirm that I am using English to submit this report (Language Policy).
• Non-english title submitions will be closed directly ( 非英文标题的提交将会被直接关闭 ) (Language Policy).
• Please do not modify this template :) and fill in all the required fields.

Describe your problem

For example, when I deploy Qwen2.5-VL-7B-Instruct on vllm, this model has 'embedd', 'reward', 'classify', 'generate', 'score' these tasks, what should I choose?

[dosubot]

For deploying an img2txt model for PDF document parsing in RAGFlow, you should select the task that performs image-to-text conversion—typically called "generate" or "image2text" in most multi-modal model APIs. In RAGFlow, the integration expects the model to output text from images, so "generate" is the most appropriate choice among the options you listed. Tasks like "embedd", "score", "classify", and "reward" are used for other internal operations and are not intended for extracting text from images in PDFs. Make sure your model is classified as "image2text" in your configuration to ensure it is used correctly in the PDF parsing pipeline (source, source).

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}   

[KevinHuSh]

I guess generate.