This is my first attempt to train an LSTM model to predict oil prices, following some tutorials. My dataset:
| Date | USD Index | Oil Price |
|---|---|---|
| 12-10-2019 | 50 | 66 |
| 13-10-2019 | 51 | 60 |
where the Oil price is the target column. Sequences size= 7 and output = 1. I couldn't add validation data and print the validation loss in addition to the train and test loss. This is my code and attempts:
#split to train, valid and test (the dataset size is 2380. so 150 for test, and 100 for valid and the remaining is for train)
X_train = X_seq[:-150]
y_train = y_seq[:-150]
X_test = X_seq[-150:]
y_test = y_seq[-150:]
X_val = X_train[-100:]
y_val = y_train [-100:]
X_train= X_train [:-100]
y_train = y_train[:-100]
the LSTM model
class LSTM(nn.Module):
def __init__(self, num_classes, input_size, hidden_size, num_layers):
super().__init__()
self.num_classes = num_classes # output size
self.num_layers = num_layers # number of recurrent layers in the lstm
self.input_size = input_size # input size
self.hidden_size = hidden_size # neurons in each lstm layer
# LSTM model
self.lstm = nn.LSTM(input_size=input_size, hidden_size=hidden_size,
num_layers=num_layers, batch_first=True, dropout=0.2) # lstm
self.fc_1 = nn.Linear(hidden_size, 128) # fully connected
self.fc_2 = nn.Linear(128, num_classes) # fully connected last layer
self.relu = nn.ReLU()
def forward(self,x):
# hidden state
h_0 = Variable(torch.zeros(self.num_layers, x.size(0), self.hidden_size))
# cell state
c_0 = Variable(torch.zeros(self.num_layers, x.size(0), self.hidden_size))
# propagate input through LSTM
output, (hn, cn) = self.lstm(x, (h_0, c_0)) # (input, hidden, and internal state)
hn = hn.view(-1, self.hidden_size) # reshaping the data for Dense layer next
out = self.relu(hn)
out = self.fc_1(out) # first dense
out = self.relu(out) # relu
out = self.fc_2(out) # final output
return out
This is the loop:
def training_loop(n_epochs, lstm, optimiser, loss_fn, X_train, y_train, X_test, y_test,
X_val , y_val,):
for epoch in range(n_epochs):
lstm.train()
outputs = lstm.forward(X_train) # forward pass
optimiser.zero_grad() # calculate the gradient, manually setting to 0
# obtain the loss function
loss = loss_fn(outputs, y_train)
#val_loss = loss_fn(y_val, y_test).item()
#####
###
loss.backward() # calculates the loss of the loss function
optimiser.step() # improve from loss, i.e backprop test loss
lstm.eval()
test_preds = lstm(X_test)
test_loss = loss_fn(test_preds, y_test)
if epoch % 100 == 0:
print("Epoch: %d, train loss: %1.5f, test loss: %1.5f" % (epoch,
loss.item(),
test_loss.item()))
This is how I call the model:
n_epochs = 1000
learning_rate = 0.001
input_size = 3 # number of features
hidden_size = 2 # number of features in hidden state
num_layers = 1 # number of stacked lstm layers
num_classes = 1 # number of output classes
lstm = LSTM(num_classes, input_size, hidden_size, num_layers)
loss_fn = torch.nn.MSELoss() # mean-squared error for regression
optimiser = torch.optim.Adam(lstm.parameters(), lr=learning_rate)
training_loop(n_epochs=n_epochs,lstm=lstm, optimiser=optimiser, loss_fn=loss_fn, X_train=X_train_tensors,y_train=y_train_tensors, X_test=X_test_tensors, y_test=y_test_tensors, X_val=X_val_tensors,y_val=y_val_tensors)
So my problem is that I don't know how to pass the validation set to be considered during the training and calculate the validation loss and doing early stop based on it.
another question please, are 100 and 1500 valid and test is suitable in a dataset size of 2800?
Any help will be appreciated.
You can use the following EarlyStopping class to implement early stopping mechanism:
class EarlyStopping:
"""Early stops the training if validation loss doesn't improve after a given patience."""
def __init__(self, patience=7, verbose=False, delta=0, path='checkpoint.pt', trace_func=print):
"""
Args:
patience (int): How long to wait after last time validation loss improved.
Default: 7
verbose (bool): If True, prints a message for each validation loss improvement.
Default: False
delta (float): Minimum change in the monitored quantity to qualify as an improvement.
Default: 0
path (str): Path for the checkpoint to be saved to.
Default: 'checkpoint.pt'
trace_func (function): trace print function.
Default: print
"""
self.patience = patience
self.verbose = verbose
self.counter = 0
self.best_score = None
self.early_stop = False
self.val_loss_min = np.Inf
self.delta = delta
self.path = path
self.trace_func = trace_func
def __call__(self, val_loss, model):
score = -val_loss
if self.best_score is None:
self.best_score = score
self.save_checkpoint(val_loss, model)
elif score < self.best_score + self.delta:
self.counter += 1
self.trace_func(f'EarlyStopping counter: {self.counter} out of {self.patience}')
if self.counter >= self.patience:
self.early_stop = True
else:
self.best_score = score
self.save_checkpoint(val_loss, model)
self.counter = 0
def save_checkpoint(self, val_loss, model):
"""Saves model when validation loss decrease."""
if self.verbose:
self.trace_func(
f'Validation loss decreased ({self.val_loss_min:.6f} --> {val_loss:.6f}). Saving model ...')
# torch.save(model.state_dict(), self.path)
self.val_loss_min = val_loss
Example usage:
early_stopping = EarlyStopping(patience=20, verbose=True)
for epoch in range(n_epochs):
lstm.train()
outputs = lstm.forward(X_train) # forward pass
optimiser.zero_grad() # calculate the gradient, manually setting to 0
# obtain the loss function
loss = loss_fn(outputs, y_train)
loss.backward() # calculates the loss of the loss function
optimiser.step() # improve from loss, i.e backprop test loss
lstm.eval()
val_loss = loss_fn(y_val, y_test).item() # get val loss
early_stopping(val_loss, model)
if early_stopping.early_stop:
print("Early stopping")
break
test_preds = lstm(X_test)