Request to upload torchscript model #2359

🚀The feature

https://huggingface.co/speechbrain/asr-transformer-transformerlm-librispeech/tree/main

provides only model.state_dict().

Is it possible to also upload models after applying torch.jit.script() or torch.jit.trace()?

One of the advantages is that users don't need to install speechbrain to use the uploaded model.

Solution outline

Upload models after applying torch.jit.script() or torch.jit.trace().

Additional context

We are going to use k2 for CTC decoding for any models trained with CTC loss.
See k2-fsa/k2#1096

We want to load a model using torch.jit.load() from some framework without the need to install the framework where the model is trained.