def fit(net,
optimizer,
loss_function,
dataloader_train,
epochs=10,
pbar=None,
device='cpu'):
val_loss_best = np.inf
loss_list = []
# Prepare loss history
for idx_epoch in range(epochs):
for idx_batch, (x, y) in enumerate(dataloader_train):
optimizer.zero_grad()
# Propagate input
netout = net(x.to(device)) # [8,1460]
# Comupte loss
loss = loss_function(y.to(device), netout)
print('Epoch:', '%04d' % (idx_epoch + 1), 'loss =',
'{:.6f}'.format(loss))
loss_list.append(loss)
# Backpropage loss
loss.backward()
# Update weights
optimizer.step()
val_loss = compute_loss(net, dataloader_train, loss_function,
device).item()
if val_loss < val_loss_best:
val_loss_best = val_loss
if pbar is not None:
pbar.update()
return val_loss_best
def fit(net,
optimizer,
loss_function,
dataloader_train,
dataloader_val,
epochs=10,
pbar=None,
device='cpu'):
val_loss_best = np.inf
# Prepare loss history
for idx_epoch in range(epochs):
for idx_batch, (x, y) in enumerate(dataloader_train):
optimizer.zero_grad()
# Propagate input
netout = net(x.to(device))
# Comupte loss
loss = loss_function(y.to(device), netout)
# Backpropage loss
loss.backward()
# Update weights
optimizer.step()
val_loss = compute_loss(net, dataloader_val, loss_function,
device).item()
if val_loss < val_loss_best:
val_loss_best = val_loss
if pbar is not None:
pbar.update()
return val_loss_best