Eiq software has the problem of slow training speed during training and low CPU utilization

取消
显示结果 
显示  仅  | 搜索替代 
您的意思是: 

Eiq software has the problem of slow training speed during training and low CPU utilization

1,535 次查看
shaenhurst
Contributor I

When I was training the model, the training speed was very slow, the input image size was 160*160, and the batch size was 32. Normally, the time of each step will not exceed 1 second, and in the training of the previous period of training, the training time of each step is indeed no more than 1 second. But today when training is very slow, the training time per step has reached about 3 seconds without the same parameters. But the GPU usage rate and video memory occupancy rate remain normal, and I don't know how to solve it.

CPU model: i5 10210u

graphics card model: mx250

Memory size: 16g

Graphics card memory size:4g

0 项奖励
1 回复

1,407 次查看
david_piskula
NXP Employee
NXP Employee

Hello @shaenhurst,

can you provide more information about the dataset so that we can attempt to reproduce your issue on our side and analyze it? Are you using a publicly available dataset?

Thanks,

David

0 项奖励