Replies: 1 comment 1 reply
|
Ilyes shared an image here with an A100 benchmark, if that helps: #946 |
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Hi all,
Choosing the right GPUs for this model appears to be very important. While many recommend high-end NVIDIA GPUs, I haven't been able to find any benchmarks to support these claims.
Would anyone be able to share benchmarks for MACE using different GPUs? It would be incredibly helpful to compare speed, accuracy, and memory usage across inference steps with a pretrained model, as well as during training.
It seems NVIDIA chips fall into four main classes: GTX, RTX (e.g., RTX 3090, RTX 4090), Quadro (e.g., A6000), and data center GPUs (e.g., V100, A100, H100).
Thanks for your consideration.
Best,
Jay
All reactions