Training procedure too slow

[YaqiChang]

Hi, thanks for the excellent work! But I have a question about training the model. I try to flow the instruction and PyMAF's trainging guide to reimplement PyMAF-X' s training, but got to find the training procedure is very slow. Some paras are as below:

NUM_WORKERS: 8
BAYCH_SIZE: 32

Ubder this circumanstence, the sigle epoch will take 90 mins I didn't change any other parameters yet, I wonder I missed somthing or this procedure is just really slow. The GPU I use for training is RTX4090, training on pose_resnet pretrained model and the base dataset is h36m-p2

[Keke061]

Hi, thanks for the excellent work! But I have a question about training the model. I try to flow the instruction and PyMAF's trainging guide to reimplement PyMAF-X' s training, but got to find the training procedure is very slow. Some paras are as below:

NUM_WORKERS: 8
BAYCH_SIZE: 32

Ubder this circumanstence, the sigle epoch will take 90 mins I didn't change any other parameters yet, I wonder I missed somthing or this procedure is just really slow. The GPU I use for training is RTX4090, training on pose_resnet pretrained model and the base dataset is h36m-p2

您好，很抱歉我不能回答这个问题，冒昧打扰一下，请问您有h36m_mosh_train.npz这个文件吗，真的很需要，如果您愿意，我将非常非常感谢，我的邮箱是[1279882880@qq.com]，感恩！