Can I use old version 's Clara Train SDK to predict model?(v1.1)

I discover that old model have deleted when I check ngc website’s , and I can’t use model which I downloaded before to predict.

I have a problem below:
Traceback (most recent call last):
File “/usr/lib/python3.5/”, line 184, in _run_module_as_main
main”, mod_spec)
File “/usr/lib/python3.5/”, line 85, in _run_code
exec(code, run_globals)
File “tlt2/src/apps/aas/”, line 54, in
File “tlt2/src/apps/aas/”, line 44, in main
File “tlt2/src/apps/aas/”, line 431, in aas
File “tlt2/src/apps/aas/”, line 106, in load_config
File “tlt2/src/apps/aas/”, line 25, in put_model
File “tlt2/src/apps/aas/actions/”, line 45, in init
File “tlt2/src/components/model_loaders/”, line 29, in init
AssertionError: specified model_file_path ‘/workspace/aiaa/models/segmentation_ct_liver_and_tumor.trt.pb’ is not a valid file

Thanks for your interest in Clara Train SDK. Our transforms/configuration names/APIs has changed in V2 that is why we have cleaned up our old models from NGC.
We recommend you upgrade to our latest clara train V3 and use the comparable configuration on NGC. For this specific model you can use for segmentation.
if you are looking for annotation task you can refer to example

Thanks for reply. I looking for some organs model (ex:left_atrium) in V3 , but I can’t find.
V3 model less than old version. Do you continue to upgrade model?


In V1 we had a lot of models showing the power of the sdk to generalize to multiple public datasets. Since then in V2, V3 we have been adding more features as Federated learning, AutoML, AIAA deepgrow, etc. Therefore we picked some models and upgraded them.

You should be able to download the challenge datasets and train from scratch using new features in clara train sdk. You can find lots of open datasets at

Hope that helps