How can I run pretrained model on multi GPUs?

#38
by NorzY - opened

As title. I used device_map="auto" while call SpeechT5ForTextToSpeech.from_pretrained(), but got error
"An error occurred while model initilization: SpeechT5ForTextToSpeech does not support device_map='auto'. To implement support, the model class needs to implement the _no_split_modules attribute."

Sign up or log in to comment