Soshyant

just run os.environ["TOKENIZERS_PARALLELISM"] = "1" . it's weird but the BPE trainer has issues with parallelism.

Go to the Generator's code and remove the "parametrizations" from the code. in other words, change all the nn.utils.parametrizations.weight_norm to nn.utils.weight_norm

use a longer reference sample or increase your training data. both will help

@kartikayk I'm having this same issue, but on the full fine tuned checkpoint. i can't go back and re-train the model with a new checkpointer (i used meta's checkpointer, as...

@fabriceyhc Thanks for the tip. the torchtune uses the original llama checkpoint to initiate a training session and I didn't use HF checkpointer to save the model. so I'm not...

> > @kartikayk I'm having this same issue, but on the full fine tuned checkpoint. i can't go back and re-train the model with a new checkpointer (i used meta's...

> > 3- just eyeballing it, I'm not particularly sure about that, but It does seem so. there's a lot of repetition, the model hallucinates really bad even on english...

yeah I'm curios about it too. I'll stick with audiocraft until AudioLDM2 is fully open-sourced. it's a pity, this model is vastly superior to anything else I've ever seen, wish...

> I think the GPUs provided by @hobodrifterdavid would be a great start for multilingual PL-BERT training. Before proceeding though, I need some people who speak as many languages as...

in the case of Japanese, since it already has Kana which is basically an alphabet, can't we simply restrict it to just that for now?(Kana and Romaji should be easier...