Loops with TQDM

A simple way to use a nice progress bar instead of polluting your screen with print statements.

import tqdm

def train(xtrain, ytrain, model, criterion, optimizer, n_epochs = 1_000):

    with tqdm.trange(n_epochs) as bar:
        for epoch in bar:  # loop over the dataset multiple times

            # zero the parameter gradients
            optimizer.zero_grad()

            # forward + backward + optimize
            outputs = model(xtrain)
            loss = criterion(outputs, ytrain)
            loss.backward()
            optimizer.step()

            # print statistics
            postfix = dict(Loss=f"{loss.item():.3f}")
            bar.set_postfix(postfix)

Source: DeepBayes.ru 2019 Notebook