why got increasing loss and bad mAP@10？(trained on own dataset)

LAION-AI / CLAP

Contrastive Language-Audio Pretraining

https://arxiv.org/abs/2211.06687

Creative Commons Zero v1.0 Universal

1.42k stars 137 forks source link

why got increasing loss and bad mAP@10？(trained on own dataset) #98

Open MisakaMikoto96 opened 1 year ago

MisakaMikoto96 commented 1 year ago

RetroCirce commented 1 year ago

Thank you for your question, we are busy preparing materials. We are answering different parts of questions. let me refer @lukewys and @tianyu-z to answer this question.

lukewys commented 1 year ago

Hi, I am sorry that from the information you provided, I cannot figure out why the training failure in your case happened. Since each dataset is different, probably in your case you might need different hyperparameters. One thing I would recommend is to try training on the Clotho dataset we released and see if the model is training properly.

Best, Yusong