OpenGVLab/EgoVideo

Hardware Requirements for Training?

bryanzhou008 opened this issue · 1 comments

@czczup @cg1177 A quick follow-up question: could you disclose the GPU memory size required by the training script for the Swin-L IN-22K+O365 model for SCOD task (or just the GPU type/per GPU memory used in this work)? Also whether you used whole-precision? Thanks again!

czczup commented

image

Hi,
Memory: 41561M
GPU: 8xA100 80G
Precision: FP32
Total training time: 1.7s x 1191 x 12ep=about 7 hours