[DLRM/PyTorch] Should tables labelled "Batch size / GPU" be "Global batch size"?

[rkindi]

Related to Model/Framework(s)
DLRM/PyTorch

Describe the bug
In the README for DLRM, the tables for many of the experiments say "Batch size/GPU" with 64k as the value. When I think of batch size / GPU, this makes me think of local batch size. However, this line and this line in combination suggest the global batch size is 64k (i.e. for 8 gpus, the Batch size / GPU should be 8192).

To Reproduce
N/A

Expected behavior
Expect that the batch size/gpu is actually 64k/number of gpus.

Environment
N/A

[tgrel]

Hi @rkindi, yes this is a typo. We will fix it in an upcoming release. Thank you for bringing this up!