evilpsycho / deep-time-series-prediction Goto Github PK

https://github.com/DataCanvasIO/HyperTS

Have been playing with your implementation of RNN2RNN mixing it up with this one.
I have some remarks that have worked for me:

• Inverse the Dense layer dropput/linear order. Lin(drop(x))
• Apply the Dense layer to the hidden state of the encoder before sending it to the decoder.
• Have not been able to get reasonable results with the Attention layers, maybe I am doing something wrong.

Running your sample code on CPU without cuda arguments in creating train_dl results in too many values to unpack error while training. Any solution?

I got import error.

ValueError Traceback (most recent call last)
in
44 # train model
45 wave_learner = Learner(wave, opt, root_dir="./wave", )
---> 46 wave_learner.fit(max_epochs=epoch, train_dl=train_dl, valid_dl=valid_dl, early_stopping=True, patient=16)
47
48 # load best model

/wangjin_fix/student_space/sw/Deep-Time-Series-Prediction-master/deepseries/train.py in fit(self, max_epochs, train_dl, valid_dl, early_stopping, patient, start_save)
68 self.model.train()
69 train_loss = 0
---> 70 for j, (x, y) in enumerate(train_dl):
71 self.optimizer.zero_grad()
72 loss = self.model.batch_loss(x, y)

ValueError: too many values to unpack (expected 2)