I am using an LSTM model to train a model on the Toronto emotional speech set (TESS)
dataset here. My code for the model and the training loop is as follow:
class TorchLSTMNet(torch.nn.Module):
def __init__(self, inputs=1, outputs=7, bidirectional=False):
super().__init__()
self.lstm = torch.nn.LSTM(inputs, 128, batch_first=True, bidirectional=bidirectional)
self.linear = torch.nn.Sequential(
torch.nn.Linear(128 * 2 if bidirectional else 128, 128),
torch.nn.ReLU(),
torch.nn.Linear(128, outputs)
)
def forward(self, x):
# x shape: (samples, steps, inputs)
out, (h, c) = self.lstm(x)
# use final output
if h.shape[0] == 2:
return self.linear(torch.cat([h[0], h[1]], axis=1))
else:
return self.linear(h[0])
net = TorchLSTMNet()
print(net)
def train(net, train_iter, test_iter, num_epochs, lr, device = d2l.try_gpu()):
"""Train a model with a GPU (defined in Chapter 6)."""
print('training on', device)
net.to(device)
print(net)
optimizer = torch.optim.SGD(net.parameters(), lr=lr)
loss = torch.nn.CrossEntropyLoss()
animator = d2l.Animator(xlabel='epoch', xlim=[1, num_epochs],
legend=['train loss', 'train acc', 'test acc'])
timer, num_batches = d2l.Timer(), len(train_iter)
for epoch in range(num_epochs):
# Sum of training loss, sum of training accuracy, no. of examples
metric = d2l.Accumulator(3)
net.train()
for i, (X, y) in enumerate(train_iter):
timer.start()
optimizer.zero_grad()
X, y = X.to(device), y.to(device)
y_hat = net(X)
l = loss(y_hat, y)
l.backward()
optimizer.step()
with torch.no_grad():
metric.add(l * X.shape[0], d2l.accuracy(y_hat, y), X.shape[0])
timer.stop()
train_l = metric[0] / metric[2]
train_acc = metric[1] / metric[2]
if (i + 1) % (num_batches // 5) == 0 or i == num_batches - 1:
animator.add(epoch + (i + 1) / num_batches,
(train_l, train_acc, None))
test_acc = d2l.evaluate_accuracy_gpu(net, test_iter)
animator.add(epoch + 1, (None, None, test_acc))
print(f'loss {train_l:.3f}, train acc {train_acc:.3f}, '
f'test acc {test_acc:.3f}')
print(f'{metric[2] * num_epochs / timer.sum():.1f} examples/sec '
f'on {str(device)}')
The data is of shape torch.Size([64, 40, 1]) and torch.Size([64, 7]) for each batch of x and y, respectively. X is a tensor of size (5600, 40, 1) which contains the 40 mfcc features, and Y is a tensor of size (5600, 7) which contains the associated emotions.
data_loaders = {
'train': torch.utils.data.DataLoader(train, shuffle=True, batch_size=64),
'test': torch.utils.data.DataLoader(test, batch_size=64),
}
# print the x and y shapes for one minibatch
for (x, y) in data_loaders['train']:
print(x.shape, y.shape)
break
The code fails at the training loop, at line metric.add(l * X.shape[0], d2l.accuracy(y_hat, y), X.shape[0]), with the error in the title.
I would really appreciate any help.