❓ [QUESTION] - Help Needed: Modify Refusal Removal Scripts to Work with MiniCPM-V-2_6

[NullEqualsZero]

摘要 | Summary

Hi [MiniCPM-V-2_6 Team],

I’m working on a project where I want to use your MiniCPM-V-2_6 model and abliterate it (a form of uncensoring) using the refusal removal scripts from the repository remove-refusals-with-transformer.

基本示例 | Basic Example

My Ask:

Would you be able to guide me or help modify the refusal removal scripts to work with the MiniCPM-V-2_6? Specifically, I’m having trouble integrating the model’s vision capabilities with the refusal removal process, including handling the image inputs (pixel_values) and other related functionalities.

Any guidance or support would be greatly appreciated. Thank you for your time and help!

缺陷 | Drawbacks

I'm a beginner with Python (my only experience is with Arduino programming) and I need help adapting the refusal-removal scripts to work with your model. I’m not expecting to change the whole model, but rather, I’d like to make edits to the existing scripts so they can interface correctly with your model. (Hence the ask for help)

未解决问题 | Unresolved questions

I am fully aware that probably you don't have the time for this which is fine.