News for hackers. Stuff that matters.

A comment on Watermark Anything. Parent comment

Re:Watermark Anything by hnuser123456 on Wednesday November 13, 2024 @02:35PM (#42129133)
That amount of compute was used for training. For inference (applying the watermarks), hopefully no more than a few seconds per image.

Llama 3 70B took 6.4M GPU hours to train, emitting 1900 tons of CO2 equivalent.

[ Parent | on HN ]
Re:Watermark Anything by Jaxan on Wednesday November 13, 2024 @03:57PM (#42129916)
Thanks! I was not at all aware of the scale of training! To me those are crazy amounts of gpu time and resources.
[ Parent | on HN ]
Re:Watermark Anything by pierrefdz on Wednesday November 13, 2024 @05:51PM (#42131104)
The amounts of gpu time in the paper are for all experiments, not just training the last model that is OSS (which is usually reported). People don't just oneshot the final model.
[ Parent | on HN ]