ModuleNotFoundError: No module named 'gluonts.torch.modules.loss'

[melty-chocolate]

Description

gluonts 0.15.0 introduces error ModuleNotFoundError: No module named 'gluonts.torch.modules.loss' downgrade to 0.14 fixed this

To Reproduce

(Please provide minimal example of code snippet that reproduces the error. For existing examples, please provide link.)

from gluonts.torch.modules.loss import DistributionLoss

Error message or code output

(Paste the complete error message, including stack trace, or the undesired output that the above snippet produces.)

ModuleNotFoundError: No module named 'gluonts.torch.modules.loss'

Environment

• Operating system: XOS
• Python version: 3.10, 3.9
• GluonTS version: 0.15.0

(Add as much information about your environment as possible, e.g. dependencies versions.)

[lostella]

Hi, this is not a bug, but a breaking change listed in the release notes. It was introduced in #3093. The methods implementing negative log-likelihood as loss associated with parametric distributions were moved into gluonts.torch.distributions.distribution_output, bundled together with the class providing layers to project latent states into the parameter space for the distribution family. cc @shchur