Type error when running the amortized inference

[Chunpai]

This issue occurs when I run the code first time.

Traceback (most recent call last):
File "src/torch_core/vi.py", line 364, in
train_logp = get_log_marginal_density(train_loader)
File "src/torch_core/vi.py", line 199, in get_log_marginal_density
marginal = model.log_marginal(
TypeError: log_marginal() got multiple values for argument 'num_samples'

It does not show up in my second run, and I am not sure why it happens.

[AlphaMoury]

Same here, is it a pytorch version problem perhaps?