| FazBrowse GitHub Viewer | Trending | | Home |
| Tools: [Download Repo ZIP] [Original HTTPS Page] |
|
MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅ |
Sorry, something went wrong.
|
@guschmue should it be number_of_nodes * host_processors_per_node to be future proof on multi-node submissions? |
Sorry, something went wrong.
|
I could change it. But if I look at the nvidia results in 0.7 we have not done it there |
Sorry, something went wrong.
|
ok, dividing by node count now. |
Sorry, something went wrong.
|
@guschmue should it be number_of_nodes * host_processors_per_node instead of number_of_nodes / host_processors_per_node? thanks! |
Sorry, something went wrong.
|
ups. fixed. In practice for inference num_nodes is so far always 1 - would be hard to do multi node because of loadgen. |
Sorry, something went wrong.
|
agree. :) |
Sorry, something went wrong.
| Back | FazBrowse Home | New Git URL |
fix for https://github.com/mlcommons/submissions_inference_1_0/issues/133