Fine tuning AST model to Music Emotion Classification Overfit about ast

[aarshilpatel]

Hi, I am facing a similar overfitting issue as yours. Can you tell me how did you solve it if you managed to solve it.

I am fine-tuning the model for real and fake audio classification. I got a better accuracy by reducing the learning rate. But the validation loss is still too high. These are the metrics I am getting.

acc: 0.936561
AUC: 0.980722
Avg Precision: 0.500000
Avg Recall: 1.000000
d_prime: 2.925854
train_loss: 0.018238
valid_loss: 0.534479

[moonquekes]

Hi, I am facing a similar overfitting issue as yours. Can you tell me how did you solve it if you managed to solve it.

I am fine-tuning the model for real and fake audio classification. I got a better accuracy by reducing the learning rate. But the validation loss is still too high. These are the metrics I am getting.

acc: 0.936561 AUC: 0.980722 Avg Precision: 0.500000 Avg Recall: 1.000000 d_prime: 2.925854 train_loss: 0.018238 valid_loss: 0.534479

Sorry I haven't solved this overfitting problem yet.
For now my guess is that the AST is a large Transformer-based model with high dataset size requirements and the authors pre-trained model is not necessarily useful for our task.
Please let me know what you did if you finally solved it. I am currently experimenting with other datasets with simpler models