WaveNet model crashes if num_feat_static_real > 0

[shchur]

Description

If I create WaveNetEstimator with num_feat_static_real > 0, the model will crash at the forward pass. This happens because num_feat_static_real is used when defining the expected input shape, but actually isn't received by the model.

To Reproduce

import numpy as np
import pandas as pd
from gluonts.torch.model.wavenet import WaveNetEstimator

N = 50
freq = "D"
data = [
    {
        "target": np.random.rand(N),
        "feat_static_real": [5.0] * 5,
        "start": pd.Period("2020-01-01", freq=freq),
    }
]

model = WaveNetEstimator(freq=freq, prediction_length=5, num_feat_static_real=5, trainer_kwargs={"accelerator": "cpu"})
model.train(data)

Error message or code output

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

RuntimeError: Given groups=1, weight of size [24, 39, 1], expected input[32, 35, 20] to have 39 channels, but got 35 channels instead

Environment

• Operating system:
• Python version:
• GluonTS version:
• MXNet version:

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