MQRNN (Seq2Seq LSTM + MLP) Model Gives Constant Predictions

I have a problem with MQRNN - multi-horizon quantile recurrent forecaster described here: https://arxiv.org/abs/1711.11053 This is my code (short version): import torch from torch

Solution 1:

Did you found the solution to this? I found a working implementation here: https://github.com/awslabs/gluon-ts/blob/0d963f7dc55ef866d86e33633a28d57dfab33adb/src/gluonts/model/seq2seq/_mq_dnn_estimator.py but would love a PyTorch simpler lighter/version.