How to check training time is the fastest setting by tensorflow GPU?

Autonomous Machines Jetson & Embedded Systems Jetson TX2
1

Hi,
I’m a beginner about TX2, and I would like to understand why the training average time is so long?
Can you show “the time” using a example to train any NN by tensorflow GPU?
I want to compare the time with my TX2, thanks.
By the way, I’m training DCGAN now.(https://github.com/carpedm20/DCGAN-tensorflow)
my goal is 10000 epoch per day using TX2(batch size = 64), but I am disappointed.

2

Hi,

It’s recommended to have a desktop GPU for training.
Jetson targets for faster inference rather than training.

By the way, please also make sure you have maximized the CPU/GPU performance.

sudo ./jetson_clocks.sh

Thanks.

3

Thanks for reply,
I’ll check performance using this command.
but I don’t understand what the “inference” mean and its usage?

4

Hi,

Here is an article for training and inference explanation:

Thanks.

5