Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
agr_nyc
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
3 ms
·
1.
▲
by
agr_nyc
2y ago
We are getting a forward pass time of ~100ms on Meta's original Llama2 70B (float16, batch size 8) PyTorch implementation on 8xA100. Those results are very underwhelming in terms of fully utilizing the GPU flops. If we are doing someth