how to train the dataset

asteroid-team / asteroid

The PyTorch-based audio source separation toolkit for researchers

https://asteroid-team.github.io/

MIT License

2.21k stars 419 forks source link

how to train the dataset #686

Closed zhang-chongxiang closed 6 months ago

zhang-chongxiang commented 9 months ago

I want to know how to train a dataset with source code, such as training wham with ConvTasNet and evaluating its performance.

mpariente commented 9 months ago

Have you checked the recipes (egs)?

zhang-chongxiang commented 8 months ago

Have you checked the recipes (egs)?

I'm sorry I'm a beginner at this. I looked at the contents of egs, but I still don't quite understand the order of execution. For example, in the ConvTasNet directory under wham, I run prepare_data.sh and nothing happens. I know what these scripts do, but I don't know how to execute them. Could you please give me some guidance?

mpariente commented 8 months ago

The entry point is run.sh.

Read the docs to get familiar with the organization : https://github.com/asteroid-team/asteroid/tree/master/egs