Clarification on A2A and T2A Evaluations in Audioldm_eval

[WangHaoyuuu]

Dear MusicTango Team,

I hope you're doing well. I’ve been exploring the Audioldm_eval tool from your GitHub (https://github.com/haoheliu/audioldm_eval) and truly appreciate the work.

While reviewing the paper and code, I noticed that the paper’s tables seem to focus on text-to-audio (T2A), but the code appears more aligned with audio-to-audio (A2A). Could you clarify whether the tool supports both tasks and, if so, how to properly conduct T2A evaluations?

Thank you for your time and help. I look forward to your response.

Best regards,

Wang Haoyu