How to load pretrained model efficientnet_bx in Jetson Xiavier NX?

[hou305]

I tried to load pretrained model efficientnet-b7 to train my own data referring to jetson-inference in my Jetson Xiavier NX, but I got error :

Use GPU: 0 for training
=> dataset classes: 14 [‘BrightRRD’, ‘Bubble’, ‘Connection’, ‘Dirt’, ‘Fake’, ‘Gel’, ‘NoAC’, ‘PenMark’, ‘Streak’, ‘WaterDrops’, ‘Wrinkle’, ‘other3 cut’, ‘others’, ‘overexposure’]
=> using pre-trained model ‘efficientnet_b7’
Traceback (most recent call last):
File “train.py”, line 521, in
main()
File “train.py”, line 143, in main
main_worker(args.gpu, ngpus_per_node, args)
File “train.py”, line 213, in main_worker
model = reshape_model(model, args.arch, num_classes)
File “/home/nvidia/jetson-inference/python/training/classification/reshape.py”, line 55, in reshape_model
print("classifier reshaping not supported for " + args.arch)
NameError: name ‘args’ is not defined

Does this mean it can’t support efficientnet for now? (I can use Resnet152 successfully)

Thanks for your comments!

[TomNVIDIA]

Hello,

This topic will get more exposure posted in the Jetson Forums.
I went ahead and moved it over for you.

[dusty_nv]

Hi @hou305, I just added support for EfficientNet in train.py in commit 819b10. If you pull the latest, it should be working now:

cd ~/jetson-inference
git pull --recurse-submodules

[hou305]

Thanks!

[hou305]

It works now! Thanks1