search

asteroid-team / asteroid

The PyTorch-based audio source separation toolkit for researchers
https://asteroid-team.github.io/
MIT License
2.23k stars 422 forks source link

Why are the replications so bad? #287

Closed hong1997-12 closed 3 years ago

hong1997-12 commented 3 years ago

I don't know what went wrong,I do not change the model and use the LibriMix database.

mpariente commented 3 years ago

More info?

hong1997-12 commented 3 years ago

I use the conv-tasnet for Speech Separation on wham .I do not change the experiment configurations and network structure.I tried to implement the denoising and separation tasks on this network,but the results obtained are very different from the published results.

jonashaag commented 3 years ago

Please tell exactly what are the results, what did you expect them to be, how have you obtained the results, what is your dataset, etc.

mpariente commented 3 years ago

I don't know what went wrong,I do not change the model and use the LibriMix database.

I use the conv-tasnet for Speech Separation on wham .

This doesn't seem consistent. You won't get help if you're not clear.

mpariente commented 3 years ago

Closing. Feel free to open a real issue, with details about your problems.