BUTSpeechFIT/speakerbeam ? reverse-engineered prompt
Reverse engineered prompt
Build me a working research project for SpeakerBeam target speech extraction, based on the idea of taking a noisy speech mixture and pulling out the target speaker using an enrollment utterance.
I want the whole flow to work end to end, from preparing the LibriMix data lists, to training the model, to running evaluation on a test set and saving the extracted audio plus summary metrics. Make it easy to set the repository path, point it at an existing LibriMix dataset, and run the main steps from the command line. The training should save checkpoints and the best model, and the evaluation should be able to load either the final model or an intermediate checkpoint.
If anything is unclear, use the existing docs online and keep the implementation aligned with the original SpeakerBeam setup. Please also include a simple example of how to run it after setup so I can reproduce the experiment without much guesswork.
Are you gonna build this?
make sure you review the code using coderabbit