Question about the $KEY

I see that all the jupyter notebooks have the $KEY set to this:

%env KEY=nvidia_tlt

My question is do I have to insert the KEY i generated there.
The reason I am asking is when I generate the tensorRT engine on the nano it wants a key also.
Do I use “nvidia_tlt” for the KEY when I run this on the NANO?

tlt tlt-converter $USER_EXPERIMENT_DIR/experiment_dir_final/resnet18_detector.etlt
-k $KEY
-o output_cov/Sigmoid,output_bbox/BiasAdd
-d 3,416,736
-i nchw
-m 64
-t fp16
-e $USER_EXPERIMENT_DIR/experiment_dir_final/resnet18_detector.trt
-b 4

Yes, the key is needed. The key for training should be the same as the key when you run tlt-converter or export, etc.

I am very confused about the key.
Should I use the key I generated in NGC or the key is nvidia-tlt?
for example in the faster_rcnn spec file, the key is set to tlt.
please help me

For the key, you can use your own key. Actually you can also use any but make sure you remember it and use the same key in the whole process(train,prune, export, etc).
For some purpose-built models(like peoplenet, lpdnet,etc) , there is the key in its model card. In this case, you must use the key provided in the model card page.

1 Like

Thank you