Hi!! Thanks for sharing this normalizing flow model with pytorch, it is really exciting.
When running the fifth cell of code a Value Error appears concerning the q1 distribution. It seems rare because the empirical values don't throw any problem. It may be a deprecated version of pytorch used back then. The following message appears:
ValueError Traceback (most recent call last)
in ()
1 q0_density = torch.exp(q0.log_prob(torch.Tensor(x))).numpy()
----> 2 q1_density = torch.exp(q1.log_prob(torch.Tensor(x))).numpy()
3 fig, (ax1, ax2) = plt.subplots(1, 2, sharex=True, figsize=(15, 5))
4 ax1.plot(x, q0_density); ax1.fill_between(x, q0_density, 0, alpha=0.5)
5 ax1.set_title('$q_0 = \mathcal{N}(0,1)$', fontsize=18);
1 frames
/usr/local/lib/python3.7/dist-packages/torch/distributions/distribution.py in _validate_sample(self, value)
287 if not valid.all():
288 raise ValueError(
--> 289 "Expected value argument "
290 f"({type(value).__name__} of shape {tuple(value.shape)}) "
291 f"to be within the support ({repr(support)}) "
ValueError: Expected value argument (Tensor of shape (1000,)) to be within the support (GreaterThan(lower_bound=0.0)) of the distribution TransformedDistribution(), but found invalid values.
The expected behaviour would be a density distribution for q1.
Hi!! Thanks for sharing this normalizing flow model with pytorch, it is really exciting.
When running the fifth cell of code a Value Error appears concerning the q1 distribution. It seems rare because the empirical values don't throw any problem. It may be a deprecated version of pytorch used back then. The following message appears:
ValueError Traceback (most recent call last)