liuxubo717 / lass Goto Github PK

Hi, when I was training the code, I couldn't separate all ten audios. Only the first audio could be separated, but the second one couldn't be separated. If I change the code to start separation from the second audio, the third audio cannot be separated. This means that only the first one can be separated each time, and it will be forced to stop when the next one is executed. I executed it in Ubuntu to confirm that it was not a problem of insufficient memory.

Dear Xubo Liu

I'm very interested in your paper on Language-Queried Audio Source Separation and I was hoping to reproduce your results and train my own models. Are you planning to release the code used for training the model in the paper?

Thank you in advance,

Best regards,
Theodoros Kouzelis

Hi, when to release the training code?

Hi, dear
In the works, you created a dataset based on the AudioCaps dataset, but did not describe the details of LASS. Can you open the index of your selected training data and test data