When I train Googlenet network on jetson nano,can I use program like tensorboard to detect training condition like under picture? The picture is detecting loss condition.
Are you sure you want to train on nano as a real usecase?
Theoretically it’s possible, but maybe you will run into lack of memory or other issue.
We will suggest to use a desktop PC with enough memory to train.
It will save you precious time.
Do you train it with TensorFlow?
If yes, tensorboard is integrated in our pre-built TensorFlow package so you can use it.
Because I want to record loss training condition,so which program is on this googlenet’s platform(https://github.com/dusty-nv/jetson-inference/blob/master/docs/pytorch-collect.md)?
Hi @andy8902, Hello AI World uses PyTorch for training onboard Jetson. During training, it logs the losses and model accuracy to the terminal. Technically tensorboard has been integrated into PyTorch (see
torch.utils.tensorboard), although I have not tried this personally.
Also note that the PyTorch training and ONNX exporter in Hello AI World was tested & verified using the ResNet-18 network architecture.
When I use this web ( https://github.com/dusty-nv/jetson-inference/blob/master/docs/pytorch-collect.md) to train network, I use this command line(python train.py --model-dir=GoogleNet /home/andy/hev/jetson-inference/python/training/classification/mydataset/car), is googlenet training?
Suppose this is duplicate to the topic 124681:
Please correct me if anything missing.