Error message when executing command


i was following the tutorial on jetson-inference by Dusty on jetson-inference

i am getting warning messages when i executed the following command

python3 --model=models/vehicle --batch-size=4 --workers=2 --epochs=30 --data=data/vehicle

May i know whether this is normal ? Is there something i should note ?

2021-05-25 04:23:48 - Validation dataset size: 2516
2021-05-25 04:23:48 - Build network.
2021-05-25 04:23:48 - Init from pretrained ssd models/mobilenet-v1-ssd-mp-0_675.pth
2021-05-25 04:23:48 - Took 0.56 seconds to load the model.
2021-05-25 04:24:02 - Learning rate: 0.01, Base net learning rate: 0.001, Extra Layers learning rate: 0.01.
2021-05-25 04:24:02 - Uses CosineAnnealingLR scheduler.
2021-05-25 04:24:02 - Start training from epoch 0.
/usr/local/lib/python3.6/dist-packages/torch/optim/ UserWarning: Detected call of lr_scheduler.step() before optimizer.step() . In PyTorch 1.1.0 and later, you should call them in the opposite order: optimizer.step() before lr_scheduler.step() . Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at torch.optim PyTorch 1.8.1 documentation
torch.optim PyTorch 1.8.1 documentation, UserWarning)
/usr/local/lib/python3.6/dist-packages/torch/nn/ UserWarning: size_average and reduce args will be deprecated, please use reduction=sum instead.


The warning seems occurs due to the non-expected API is used according to your PyTorch.
May I know which JetPack version and PyTorch do you use?


Hi, I am using JetPack 4.5.1 with jetson-inference running on docker container downloaded from Github.

Hi @Benjamin_Lim, you can ignore this warning, in my experience it has not impacted the SSD model training.

Thanks Dusty, the training was still able to process though but at a very slow pace. Just have to wait i guess. I am training the images on the Jetson nano 4GB.

Do you also get these warning message when starting the training process ?

Yes, I get them too, so no worries there. If you have a Linux PC with NVIDIA GPU in it, you can run the pytorch-ssd training code on PC faster (you will need to install PyTorch, ect first or use NGC PyTorch container for x86)

Noted. Thanks @dusty_nv !