Seeking help: Deployment results inconsistent

[kkkcx]

Hi there,

Thank you for your excellent work on this project!

I am reaching out because I ran into a few issues after deploying app.py. My results don't quite match your demo:

• The system appears to reject queries during face detection.

• The output includes long explanations rather than just a simple emotion tag.

• The emotion predictions are sometimes inaccurate compared to the demo.

Could you please advise on what might be causing these issues and how I might resolve them? Any guidance would be greatly appreciated.

Thank you for your time and help!

[ZebangCheng]

Thank you very much for your interest in our project and for the detailed feedback.

We have not encountered the issues you described during our own experiments or in the demo. However, based on your symptoms, we suspect that the problem may be related to incorrect or incomplete model weight loading.

Could you please check and share your configuration details, especially whether the following pretrained weights are correctly loaded and matched with the corresponding code version?

• Llama-2-7b-chat-hf
• minigptv2_checkpoint.pth
• Emotion_LLaMA.pth

If any of these weights are missing, mismatched, or loaded from an incompatible version, it could lead to abnormal face detection behavior, overly verbose textual outputs instead of concise emotion tags, and degraded prediction accuracy.

Once we have more information about your setup, we will be happy to help you further diagnose and resolve the issue. Thank you again for your support!