Viewing a single comment thread. View all comments

LetterRip t1_iw3rucf wrote

Could you provide details on the comparison with DeepSpeed? What parameters were used etc?

Also doesn't it provide any benefit for single GPU inference?

1