search

shansongliu / M2UGen

This is the official repository for M2UGen
MIT License
441 stars 38 forks source link

How to generate the evaluation datasets? #17

Open TURLEing opened 6 months ago

TURLEing commented 6 months ago

Hi, thanks for sharing your work, it's very impressive. We'd like to evaluate the M2UGen model and other compared models with your evalution datasets. In Evaluation/Image2Music/evaluate.py, we noticed that there is a file called MUImageEvalInstructions.json, which might be the splits of evaluation dataset. However, we didn't find this file anywhere, neither in Github nor Huggingface.

Could you please share this file or sharing the code for generating the split of evaluation dataset? It will help a lot. Thanks!

Mao-KU commented 4 months ago

Same issue here, did you find the evaluation datasets?