AssertionError on RetiariiExperiment

[ding3820]

Hi, I was trying to run one-shot NAS (Darts) on my model and dataset following the instruction on this. Everything went correctly until I tried the following code:

from nni.retiarii.experiment.pytorch import RetiariiExperiment

exp = RetiariiExperiment(model, trainer)
exp.run()

This code is from https://github.com/microsoft/nni/blob/9a4d0d67506f5e1197238344fed858305fbb4494/test/retiarii_test/darts/test_oneshot.py#L102 Here's the error message I received:

Looks like this part of the error handling is not complete. I have no idea where to fix this issue so I post here. Thx

Environment:

• NNI version: 2.4
• Training service (local|remote|pai|aml|etc): local
• Client OS: Linux
• Server OS (for remote mode only):
• Python version: 3.7.11
• PyTorch/TensorFlow version: PyTorch 1.9.1
• Is conda/virtualenv/venv used?: conda
• Is running in Docker?: yes

[ultmaster]

Please restart the kernel and try again. Have you run this code twice? It seems that register_advisor have already been executed before.

[ding3820]

Thanks it works.