Question on the training methodology - Batch size and Epoch

[Kev1MSL]

Hi! I was looking at the training of InstantMesh in your paper and your code, but I got a bit confused on how did you set the batch size and epoch numbers to obtain your results. I looked at the code and ran it on 8 GPUs but it seems that I only have 12 elements on my first batch per GPU. However, in the configuration you set a batch of size 2. Did I miss something?

[Mrguanglei]

@Kev1MSL I can take a look at your training script, or the structure of your dataset, and I can help you better