input.size(-1) must be equal to input_size. Expected 40, got 1 - deep-learning

Question

I am using LSTM for time series forecasting. I used a LSTM model
LSTMNet(
(lstm): LSTM(1, 128, batch_first=True)
(fc1): Linear(in_features=128, out_features=40, bias=True)
(fc2): Linear(in_features=40, out_features=1, bias=True)
(relu): ReLU()
)
Then I am freezing the last layer fc2 and trying to build another lstm layer and then train my model with the Input size[841, 14, 1]. However I am getting error:
"input.size(-1) must be equal to input_size. Expected 40, got 1."
My code is
early_stopping_patience = 150
early_stopping_counter = 0
valid_loss_min=np.inf
for epoch in range(EPOCHS):
model.train()
output = model(X_train)
train_loss = criterion(output, y_train)
optimizer.zero_grad()
train_loss.backward()
optimizer.step()
with torch.no_grad():
model.eval()
output_val = model(X_valid) #/edited here with reshape
valid_loss = criterion(output_val, y_valid)
if valid_loss <= valid_loss_min:
torch.save(model.state_dict(), './state_dict_new.pt')
print(f'Epoch {epoch + 0:01}: Validation loss decreased ({valid_loss_min:.6f} --> {valid_loss:.6f}). Saving model ...')
valid_loss_min = valid_loss
early_stopping_counter=0 #reset counter if validation loss decreases
else:
print(f'Epoch {epoch + 0:01}: Validation loss did not decrease')
early_stopping_counter+=1
if early_stopping_counter > early_stopping_patience:
print('Early stopped at epoch :', epoch)
break
print(f'\t Train_Loss: {train_loss:.4f} Val_Loss: {valid_loss:.4f} BEST VAL Loss: {valid_loss_min:.4f}\n')
My Model is
class NewLSTMNet(nn.Module):
def __init__(self, base_model, input_size, hidden_size, num_layers):
super(NewLSTMNet, self).__init__()
self.num_layers = num_layers
self.input_size = input_size
self.base_model = base_model
self.hidden_size = hidden_size
for param in self.base_model.parameters():
param.requires_grad = False
self.lstm = nn.LSTM(
input_size=40, hidden_size=hidden_size,
num_layers=num_layers, batch_first=True
)
self.fc1 = nn.Linear(hidden_size,64)
self.fc2 = nn.Linear(64,1)
self.relu = nn.ReLU()
def forward(self,x):
out = self.base_model(x)
out = out.view(out.shape[0], -1, 1)
h0 = Variable(torch.zeros(self.num_layers, out.size(0), self.hidden_size))
c0 = Variable(torch.zeros(self.num_layers, out.size(0), self.hidden_size))
_, (h_out, _) = self.lstm(out, (h0, c0))
h_out = h_out.view(-1, self.hidden_size)
out = self.fc2(self.relu(self.fc1(h_out)))
return out
Could anyone please look into it and help me please?