RuntimeError: Für ungestapelte 2D-Eingaben sollten hx und cx ebenfalls 2D sein, aber (3D, 3D) Tensoren haben ⇐ Python

[Anonymous]

Hey, ich habe einige Probleme mit meinem LSTM. Ich habe 6 Features und sende alle meine Daten (29002 Zeilen) auf einmal im LSTM (ist das eine gute Idee?).
Meine Eingabe hat die Größe:
Training Shape Torch.Size([290002, 1, 6]) Torch.Size([290002, 1])
Testing Shape Torch.Size([74998, 1, 6]) Torch.Size([74998, 1])
mein Modell:

Code: Select all

class LSTM(nn.Module):
def __init__(self, hidden_dim_LSTM, num_layers_LSTM, hidden1, drop):
super(LSTM, self).__init__()

self.hidden_dim_LSTM = hidden_dim_LSTM
self.num_layers_LSTM = num_layers_LSTM
self.hidden1=hidden1
self.drop=drop

final_output_dim = 1

#self.lstm = nn.LSTM(self.input_dim, self.hidden_dim, self.num_layers, batch_first=True)
self.lstm = nn.LSTM(6, hidden_size=hidden_dim_LSTM, num_layers=num_layers_LSTM, batch_first=True)
self.fc1 = nn.Linear(in_features=hidden_dim_LSTM, out_features=hidden1)
self.drop = nn.Dropout(drop)
self.fc2 = nn.Linear(in_features=hidden1, out_features=final_output_dim)

def forward(self, x):
h_0 = Variable(torch.zeros(self.num_layers_LSTM, x.size(0), self.hidden_dim_LSTM)).requires_grad_().to(device) #hidden state
c_0 = Variable(torch.zeros(self.num_layers_LSTM, x.size(0), self.hidden_dim_LSTM)).requires_grad_().to(device) #internal state
# Propagate input through LSTM
output, (hn, cn) = self.lstm(x, (h_0, c_0)) #lstm with input, hidden, and internal state
hn = hn.view(-1, self.hidden_dim_LSTM) #reshaping the data for Dense layer next
out = F.relu(hn)
out = self.fc1(out)
out = self.drop(out)
out = torch.relu(out)
#out = self.drop(out)
out = self.fc2(out)
return out

Wenn ich mit dem Training beginne, erhalte ich diesen Fehler:
RuntimeError: Bei ungestapelter 2D-Eingabe sollten hx und cx ebenfalls 2D sein, erhielten aber (3-D, 3-D) Tensoren
in Zeile: Ausgabe, (hn, cn) = self.lstm(x, (h_0, c_0))
Ich bin für alles dankbar Hilfe!