qgans_for_loading_random_distributions.ipynb legacy tutorial raises UnboundLocalError

[nonhermitian]

Running the tutorial with Qiskit 0.19.6 gives me:

UnboundLocalError: local variable 'd_loss_min' referenced before assignment

Note that I had to drop the number of data samples and epochs by 10x to get it to find this error

[stefan-woerner]

@Zoufalc could you please look into this problem?

[Zoufalc]

An error message needs to be added. I will take care of this. What happens is that the number of data samples < batch. This can happen since we need to bin and truncate (remove too large/small values) the input data to map the data points onto basis states.

[nonhermitian]

If the number of data samples is the same as what it was set at before, then the runtime is longer than the 3min / cell limit. This is why I tried to reduce it. I guess there is something in the middle that likely works.

[Zoufalc]

If you reduce the batch size e.g. to 50, it should work.

[Zoufalc]

It is fixed.

[Eric-Arellano]

This tutorial no longer exists in this repo. (And sounds like it was fixed by @Zoufalc!)