You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I reproduced this on the current dev branch. Both the PyTorch and JAX LibriSpeech Conformer implementations return a pair of tensors as the model output while annotating it as a single Tensor. I will update both annotations so the framework implementations stay consistent, verify the affected modules, and open a focused PR.
Activity
mrhimanshu commentedon Oct 17, 2026
I reproduced this on the current dev branch. Both the PyTorch and JAX LibriSpeech Conformer implementations return a pair of tensors as the model output while annotating it as a single Tensor. I will update both annotations so the framework implementations stay consistent, verify the affected modules, and open a focused PR.