zlacker

[parent] [thread] 1 comments
1. emadm+(OP)[view] [source] 2023-08-01 14:33:09
Gaudi2 and Inferentia2 are both good.

We train on A100s, TPUs and... other things now.

replies(1): >>emadm+52
◧
2. emadm+52[view] [source] 2023-08-01 14:42:23
>>emadm+(OP)
Also missed in the post is fp8 is really much more efficient

The H100s are actually very good for inference..

[go to top]