Questions about training speed

[bupt906]

I am doing related work based on paddleOCR. I replaced the backbone of the CRNN network structure from mobilenet_v3 to micronet. I train on a single machine with 4 cards, the batch_size is 256, and I find that the ips I train are about 60, while the ips of mobilenet_v3 can reach about 200. may I ask if this is normal? And GPU usage has become lower.

[liyunsheng13]

I'm not sure. The gap seems to be large. In fact micronet is developed mainly for CPU usage. The speed should not be very fast due to the existence of dynamic operation. But from my experience, it should be much slower than mobilenetv3.