Aditya Yadavalli

Hi, I think 2k hour Telugu dataset is available on request at the moment. @mirishkarganesh, maybe, you provide the details for this?

Thanks @simpleoier! I had a similar issue. Setting `--nj` to a lower number helped. Although, it looks like each job is taking more than 8GB in my case.

Thanks for the response @TParcollet! I have used the pretrained tokeniser and LM to train my ASR (just trained the AM like in (`speechbrain/recipes/LibriSpeech/ASR/transformer/`). So I hope tokeniser is not...

Like you already mention @AsuMagic, I think this is not similar to #1590. However, it is worth trying it out to narrow down where the issue could be. I'll report...

Great! I'll test with the new patch and report my results. Thank you so much @AsuMagic!

Hey @Adel-Moumen! Apologies for not acting on this. Been busy with a few tight deadlines. I'll work on this and report by the end of next week.

Hi @Adel-Moumen, thanks for following up! I just checked with speechbrain version `0.5.14`. It still seems to go into an infinite loop. May I share all the model related files...

mailed you!

Hi @Adel-Moumen! Thanks for following up. Some settings have changed on my end. Here's the link to all the files: https://iiitaphyd-my.sharepoint.com/:u:/g/personal/aditya_yadavalli_alumni_iiit_ac_in/EZ5NSpEbY4RNtvqp1EgT0bMBSUw0JYcbiqLxskBcNe6tfA?e=tiyWge. Please let me know if you need anything else....

I believe I've used the one inside the librispeech recipe. This one specifically: https://github.com/speechbrain/speechbrain/tree/develop/recipes/LibriSpeech/ASR/transformer/hparams