Has anyone deployed it on 10x 3090 ? Or any similar configuration?
AlexanderKozhevin opened this issue · comments
Alexander Kozhevin commented
Pavel commented
Yes, it works fine on 8xA100.
I think 10x 3090 is too much, 9 is enough.
Pretrained language model with 100B parameters
AlexanderKozhevin opened this issue · comments
Yes, it works fine on 8xA100.
I think 10x 3090 is too much, 9 is enough.