2 ms·
Of course, this is common with LLMs which are too large to fit in any single GPU. I believe Deepspeed implements what you're referring to.
by zettabomb 3y ago
Of course, this is common with LLMs which are too large to fit in any single GPU. I believe Deepspeed implements what you're referring to.