Python loss.SoftmaxCrossEntropyLoss方法代码示例

# 需要导入模块: from mxnet.gluon import loss [as 别名]
# 或者: from mxnet.gluon.loss import SoftmaxCrossEntropyLoss [as 别名]
def get_loss(loss_name, loss_extra_kwargs):
    """
    Get loss by name.

    Parameters:
    ----------
    loss_name : str
        Loss name.
    loss_extra_kwargs : dict
        Loss extra parameters.

    Returns
    -------
    Loss
        Loss object instance.
    """
    if loss_name == "SoftmaxCrossEntropy":
        return SoftmaxCrossEntropyLoss(**loss_extra_kwargs)
    if loss_name == "SegSoftmaxCrossEntropy":
        return SegSoftmaxCrossEntropyLoss(**loss_extra_kwargs)
    if loss_name == "MixSoftmaxCrossEntropy":
        return MixSoftmaxCrossEntropyLoss(**loss_extra_kwargs)
    else:
        raise Exception("Wrong loss name: {}".format(loss_name)) 

# 需要导入模块: from mxnet.gluon import loss [as 别名]
# 或者: from mxnet.gluon.loss import SoftmaxCrossEntropyLoss [as 别名]
def test(test_net, ctx, test_loader, iteration, logger):
    # print("Start testing iter %d." % iteration)
    Loss = gloss.SoftmaxCrossEntropyLoss()
    metric = mx.metric.Accuracy()
    metric_top5 = mx.metric.TopKAccuracy(5)
    test_loss = mx.metric.Loss()
    for batch in test_loader:
        trans = gutils.split_and_load(batch[0], ctx)
        labels = gutils.split_and_load(batch[1], ctx)
        outputs = [test_net(tran) for tran in trans]
        losses = [Loss(output, label) for output, label in zip(outputs, labels)]
        test_loss.update(0, losses)
        metric.update(labels, outputs)
        metric_top5.update(labels, outputs)
    _, test_top1_acc = metric.get()
    _, test_top5_acc = metric_top5.get()
    _, test_loss = test_loss.get()

    if test_top1_acc >= 0.7:
        test_net.save_parameters('imagenet_param/test_iter%d_%.5f.param' % (iteration, test_top1_acc))
    test_str = ("test_Loss: %f, test top1-acc %f, test top5-acc %f." % (test_loss, test_top1_acc, test_top5_acc))
    logger.info(test_str) 

# 需要导入模块: from mxnet.gluon import loss [as 别名]
# 或者: from mxnet.gluon.loss import SoftmaxCrossEntropyLoss [as 别名]
def test(test_net, ctx, test_loader, iteration, logger):
    Loss = gloss.SoftmaxCrossEntropyLoss()
    metric = mx.metric.Accuracy()
    test_loss = mx.metric.Loss()
    for batch in test_loader:
        data = gluon.utils.split_and_load(batch[0], ctx_list=ctx, batch_axis=0)
        label = gluon.utils.split_and_load(batch[1], ctx_list=ctx, batch_axis=0)
        outputs = [test_net(X.astype(dtype, copy=False)) for X in data]
        losses = [Loss(yhat, y.astype(dtype, copy=False)) for yhat, y in zip(outputs, label)]
        test_loss.update(0, losses)
        metric.update(label, outputs)

    _, test_acc = metric.get()
    _, test_loss = test_loss.get()
    test_net.save_parameters('cifar_param/test_epoch%d_%.5f.param' % (iteration, test_acc))
    test_str = ("Test Loss: %f, Test acc %f." % (test_loss, test_acc))
    logger.info(test_str) 

# 需要导入模块: from mxnet.gluon import loss [as 别名]
# 或者: from mxnet.gluon.loss import SoftmaxCrossEntropyLoss [as 别名]
def gluon_random_data_run():
    mlflow.gluon.autolog()

    with mlflow.start_run() as run:
        data = DataLoader(LogsDataset(), batch_size=128, last_batch="discard")
        validation = DataLoader(LogsDataset(), batch_size=128, last_batch="discard")

        model = HybridSequential()
        model.add(Dense(64, activation="relu"))
        model.add(Dense(64, activation="relu"))
        model.add(Dense(10))
        model.initialize()
        model.hybridize()
        trainer = Trainer(model.collect_params(), "adam",
                          optimizer_params={"learning_rate": .001, "epsilon": 1e-07})
        est = estimator.Estimator(net=model, loss=SoftmaxCrossEntropyLoss(),
                                  metrics=Accuracy(), trainer=trainer)

        with warnings.catch_warnings():
            warnings.simplefilter("ignore")
            est.fit(data, epochs=3, val_data=validation)
    client = mlflow.tracking.MlflowClient()
    return client.get_run(run.info.run_id) 

# 需要导入模块: from mxnet.gluon import loss [as 别名]
# 或者: from mxnet.gluon.loss import SoftmaxCrossEntropyLoss [as 别名]
def test_autolog_ends_auto_created_run():
    mlflow.gluon.autolog()

    data = DataLoader(LogsDataset(), batch_size=128, last_batch="discard")

    model = HybridSequential()
    model.add(Dense(64, activation="relu"))
    model.add(Dense(64, activation="relu"))
    model.add(Dense(10))
    model.initialize()
    model.hybridize()

    trainer = Trainer(model.collect_params(), "adam",
                      optimizer_params={"learning_rate": .001, "epsilon": 1e-07})
    est = estimator.Estimator(net=model, loss=SoftmaxCrossEntropyLoss(),
                              metrics=Accuracy(), trainer=trainer)

    with warnings.catch_warnings():
        warnings.simplefilter("ignore")
        est.fit(data, epochs=3)

    assert mlflow.active_run() is None 

# 需要导入模块: from mxnet.gluon import loss [as 别名]
# 或者: from mxnet.gluon.loss import SoftmaxCrossEntropyLoss [as 别名]
def test_autolog_persists_manually_created_run():
    mlflow.gluon.autolog()

    data = DataLoader(LogsDataset(), batch_size=128, last_batch="discard")

    with mlflow.start_run() as run:

        model = HybridSequential()
        model.add(Dense(64, activation="relu"))
        model.add(Dense(64, activation="relu"))
        model.add(Dense(10))
        model.initialize()
        model.hybridize()
        trainer = Trainer(model.collect_params(), "adam",
                          optimizer_params={"learning_rate": .001, "epsilon": 1e-07})
        est = estimator.Estimator(net=model, loss=SoftmaxCrossEntropyLoss(),
                                  metrics=Accuracy(), trainer=trainer)

        with warnings.catch_warnings():
            warnings.simplefilter("ignore")
            est.fit(data, epochs=3)

        assert mlflow.active_run().info.run_id == run.info.run_id 

# 需要导入模块: from mxnet.gluon import loss [as 别名]
# 或者: from mxnet.gluon.loss import SoftmaxCrossEntropyLoss [as 别名]
def gluon_model(model_data):
    train_data, train_label, _ = model_data
    train_data_loader = DataLoader(list(zip(train_data, train_label)),
                                   batch_size=128, last_batch="discard")
    model = HybridSequential()
    model.add(Dense(128, activation="relu"))
    model.add(Dense(64, activation="relu"))
    model.add(Dense(10))
    model.initialize()
    model.hybridize()
    trainer = Trainer(model.collect_params(), "adam",
                      optimizer_params={"learning_rate": .001, "epsilon": 1e-07})
    est = estimator.Estimator(net=model, loss=SoftmaxCrossEntropyLoss(),
                              metrics=Accuracy(), trainer=trainer)
    with warnings.catch_warnings():
        warnings.simplefilter("ignore")
        est.fit(train_data_loader, epochs=3)
    return model 

# 需要导入模块: from mxnet.gluon import loss [as 别名]
# 或者: from mxnet.gluon.loss import SoftmaxCrossEntropyLoss [as 别名]
def train_ch5(net, train_iter, test_iter, batch_size, trainer, ctx,
              num_epochs):
    """Train and evaluate a model with CPU or GPU."""
    print('training on', ctx)
    loss = gloss.SoftmaxCrossEntropyLoss()
    for epoch in range(num_epochs):
        train_l_sum, train_acc_sum, n, start = 0.0, 0.0, 0, time.time()
        for X, y in train_iter:
            X, y = X.as_in_context(ctx), y.as_in_context(ctx)
            with autograd.record():
                y_hat = net(X)
                l = loss(y_hat, y).sum()
            l.backward()
            trainer.step(batch_size)
            y = y.astype('float32')
            train_l_sum += l.asscalar()
            train_acc_sum += (y_hat.argmax(axis=1) == y).sum().asscalar()
            n += y.size
        test_acc = evaluate_accuracy(test_iter, net, ctx)
        print('epoch %d, loss %.4f, train acc %.3f, test acc %.3f, '
              'time %.1f sec'
              % (epoch + 1, train_l_sum / n, train_acc_sum / n, test_acc,
                 time.time() - start)) 

# 需要导入模块: from mxnet.gluon import loss [as 别名]
# 或者: from mxnet.gluon.loss import SoftmaxCrossEntropyLoss [as 别名]
def train_batch(self, Xs_lst, Ys_lst, Xt_lst, Yt_lst, inference, target=True):
        criterion_xent = SoftmaxCrossEntropyLoss()

        postfix = 'Tgt' if target else 'Src'

        with autograd.record():
            loss = []
            for xs, ys, xt, yt in zip(Xs_lst, Ys_lst, Xt_lst, Yt_lst):
                criterion_aux = dSNELoss(xs.shape[0], xt.shape[0], self.args.embed_size, self.args.margin,
                                         self.args.fn)

                ys_hat, fts = inference(xs)
                yt_hat, ftt = inference(xt)

                loss_xent_src = criterion_xent(ys_hat, ys)
                loss_aux = criterion_aux(fts, ys, ftt, yt)

                loss_total = (1 - self.args.alpha) * loss_xent_src + self.args.alpha * loss_aux
                loss.append(loss_total)

                self.metrics['Train-Xent-%s' % postfix].update(None, [loss_xent_src])
                self.metrics['Train-Acc-Src'].update([ys], [ys_hat])
                self.metrics['Train-Acc-Tgt'].update([yt], [yt_hat])
                self.metrics['Train-Aux-%s' % postfix].update(None, [loss_aux])
                self.metrics['Train-Total-%s' % postfix].update(None, [loss_total])

                self.cur_iter += 1

            for l in loss:
                l.backward() 

# 需要导入模块: from mxnet.gluon import loss [as 别名]
# 或者: from mxnet.gluon.loss import SoftmaxCrossEntropyLoss [as 别名]
def train_batch(self, x0, y0, x1, y1, xu0, xu1, student, teacher=None, target=False):
        criterion_xent = SoftmaxCrossEntropyLoss()
        criterion_consistency = SoftmaxL2Loss()
        postfix = 'Tgt-l' if target else 'Src'

        with autograd.record():
            y0_hat, fss = student(x0)
            # cross entropy
            loss_xent = criterion_xent(y0_hat, y0)

            self.metrics['Train-Xent-%s' % postfix].update(None, [loss_xent])
            self.metrics['Train-Acc-%s' % postfix].update([y0], [y0_hat])

            if teacher is not None:
                ysu_hat, fsu = student(xu0)
                ytu_hat, ftu = teacher(xu1)
                loss_consistency = criterion_consistency(ytu_hat, ysu_hat)

                self.metrics['Train-Cons-%s' % postfix].update(None, [loss_consistency])
            else:
                loss_consistency = 0

            # weighted loss
            consistency_weight = self.update_beta()

            loss = loss_xent + consistency_weight * loss_consistency

            self.metrics['Train-Total-%s' % postfix].update(None, [loss])

            self.cur_iter += 1

            loss.backward() 

# 需要导入模块: from mxnet.gluon import loss [as 别名]
# 或者: from mxnet.gluon.loss import SoftmaxCrossEntropyLoss [as 别名]
def train_and_predict_rnn_gluon(model, num_hiddens, vocab_size, ctx,
                                corpus_indices, idx_to_char, char_to_idx,
                                num_epochs, num_steps, lr, clipping_theta,
                                batch_size, pred_period, pred_len, prefixes):
    """Train an Gluon RNN model and predict the next item in the sequence."""
    loss = gloss.SoftmaxCrossEntropyLoss()
    model.initialize(ctx=ctx, force_reinit=True, init=init.Normal(0.01))
    trainer = gluon.Trainer(model.collect_params(), 'sgd',
                            {'learning_rate': lr, 'momentum': 0, 'wd': 0})

    for epoch in range(num_epochs):
        l_sum, n, start = 0.0, 0, time.time()
        data_iter = data_iter_consecutive(
            corpus_indices, batch_size, num_steps, ctx)
        state = model.begin_state(batch_size=batch_size, ctx=ctx)
        for X, Y in data_iter:
            for s in state:
                s.detach()
            with autograd.record():
                (output, state) = model(X, state)
                y = Y.T.reshape((-1,))
                l = loss(output, y).mean()
            l.backward()
            params = [p.data() for p in model.collect_params().values()]
            grad_clipping(params, clipping_theta, ctx)
            trainer.step(1)
            l_sum += l.asscalar() * y.size
            n += y.size

        if (epoch + 1) % pred_period == 0:
            print('epoch %d, perplexity %f, time %.2f sec' % (
                epoch + 1, math.exp(l_sum / n), time.time() - start))
            for prefix in prefixes:
                print(' -', predict_rnn_gluon(
                    prefix, pred_len, model, vocab_size, ctx, idx_to_char,
                    char_to_idx)) 

# 需要导入模块: from mxnet.gluon import loss [as 别名]
# 或者: from mxnet.gluon.loss import SoftmaxCrossEntropyLoss [as 别名]
def train(train_net, iterations, trainer, ctx, lr_period: tuple, lr_decay, train_loader, test_loader, cat_interval):
    # set up logger
    logging.basicConfig()
    logger = logging.getLogger()
    logger.setLevel(logging.INFO)
    log_file_path = 'Attention56_train.log'
    fh = logging.FileHandler(log_file_path)
    logger.addHandler(fh)

    net.collect_params().reset_ctx(ctx)
    train_gen = inf_train_gen(train_loader)
    Loss = gloss.SoftmaxCrossEntropyLoss()
    metric = mx.metric.Accuracy()
    metric_top5 = mx.metric.TopKAccuracy(5)
    train_loss = mx.metric.Loss()
    prev_time = datetime.datetime.now()

    metric.reset()
    train_loss.reset()

    for iteration in range(int(iterations)):
        batch = next(train_gen)
        trans = gutils.split_and_load(batch.data[0], ctx)
        labels = gutils.split_and_load(batch.label[0], ctx)

        with autograd.record():
            outputs = [train_net(tran) for tran in trans]
            losses = [Loss(output, label) for output, label in zip(outputs, labels)]

        for loss in losses:
            loss.backward()

        trainer.step(batch_size)
        train_loss.update(0, losses)
        metric.update(labels, outputs)
        metric_top5.update(labels, outputs)
        if iteration % cat_interval == cat_interval - 1:
            cur_time = datetime.datetime.now()
            time_str = format_time(prev_time, cur_time)
            _, top1_acc = metric.get()
            _, top5_acc = metric_top5.get()
            _, epoch_loss = train_loss.get()
            metric.reset()
            metric_top5.reset()
            train_loss.reset()
            epoch_str = ("Iter %d. Loss: %.5f, Train top1-acc %f, Train top5-acc %f."
                         % (iteration, epoch_loss, top1_acc, top5_acc))
            prev_time = cur_time
            logger.info(epoch_str + time_str + 'lr ' + str(trainer.learning_rate))
            test(train_net, ctx, test_loader, iteration, logger)
        if iteration in lr_period:
            trainer.set_learning_rate(trainer.learning_rate * lr_decay) 

# 需要导入模块: from mxnet.gluon import loss [as 别名]
# 或者: from mxnet.gluon.loss import SoftmaxCrossEntropyLoss [as 别名]
def train(train_net, epochs, lr, wd, ctx, warmup_epochs, train_loader, test_loader, use_mixup, logger):
    num_batches = train_samples // batch_size
    lr_scheduler = LRSequential([
        LRScheduler('linear', base_lr=0, target_lr=lr,
                    nepochs=warmup_epochs, iters_per_epoch=num_batches),
        LRScheduler('cosine', base_lr=lr, target_lr=0,
                    nepochs=epochs - warmup_epochs,
                    iters_per_epoch=num_batches)
    ])
    opt_params = {'learning_rate': lr, 'momentum': 0.9, 'wd': wd, 'lr_scheduler': lr_scheduler}
    if dtype != 'float32':
        opt_params['multi_precision'] = True
    trainer = gluon.Trainer(train_net.collect_params(), 'nag', opt_params)

    Loss = gluon.loss.SoftmaxCrossEntropyLoss(sparse_label=False)
    metric = mx.metric.RMSE()
    train_loss = mx.metric.Loss()
    alpha = 1
    classes = 10

    print("Start training with mixup.")
    for epoch in range(epochs):
        metric.reset()
        train_loss.reset()
        st_time = time.time()
        for i, batch in enumerate(train_loader):
            lam = np.random.beta(alpha, alpha)
            if epoch >= (epochs - 20) or not use_mixup:
                lam = 1

            data = gluon.utils.split_and_load(batch[0], ctx_list=ctx, batch_axis=0)
            label = gluon.utils.split_and_load(batch[1], ctx_list=ctx, batch_axis=0)

            trans = [lam * X + (1 - lam) * X[::-1] for X in data]
            labels = []
            for Y in label:
                y1 = label_transform(Y, classes)
                y2 = label_transform(Y[::-1], classes)
                labels.append(lam * y1 + (1 - lam) * y2)

            with autograd.record():
                outputs = [train_net(X.astype(dtype, copy=False)) for X in trans]
                losses = [Loss(yhat, y.astype(dtype, copy=False)) for yhat, y in zip(outputs, labels)]
            for l in losses:
                l.backward()
            trainer.step(batch_size)
            train_loss.update(0, losses)
            metric.update(labels, outputs)

        cur_time = time.time() - st_time
        eps_samples = int(train_samples // cur_time)
        _, train_acc = metric.get()
        _, epoch_loss = train_loss.get()
        epoch_str = ("Epoch %d. Loss: %.5f, Train RMSE %.5f. %d samples/s. lr %.5f"
                     % (epoch, epoch_loss, train_acc, eps_samples, trainer.learning_rate))
        logger.info(epoch_str)
        test(train_net, ctx, test_loader, epoch, logger) 

# 需要导入模块: from mxnet.gluon import loss [as 别名]
# 或者: from mxnet.gluon.loss import SoftmaxCrossEntropyLoss [as 别名]
def train_and_predict_rnn(rnn, get_params, init_rnn_state, num_hiddens,
                          vocab_size, ctx, corpus_indices, idx_to_char,
                          char_to_idx, is_random_iter, num_epochs, num_steps,
                          lr, clipping_theta, batch_size, pred_period,
                          pred_len, prefixes):
    """Train an RNN model and predict the next item in the sequence."""
    if is_random_iter:
        data_iter_fn = data_iter_random
    else:
        data_iter_fn = data_iter_consecutive
    params = get_params()
    loss = gloss.SoftmaxCrossEntropyLoss()

    for epoch in range(num_epochs):
        if not is_random_iter:
            state = init_rnn_state(batch_size, num_hiddens, ctx)
        l_sum, n, start = 0.0, 0, time.time()
        data_iter = data_iter_fn(corpus_indices, batch_size, num_steps, ctx)
        for X, Y in data_iter:
            if is_random_iter:
                state = init_rnn_state(batch_size, num_hiddens, ctx)
            else:
                for s in state:
                    s.detach()
            with autograd.record():
                inputs = to_onehot(X, vocab_size)
                (outputs, state) = rnn(inputs, state, params)
                outputs = nd.concat(*outputs, dim=0)
                y = Y.T.reshape((-1,))
                l = loss(outputs, y).mean()
            l.backward()
            grad_clipping(params, clipping_theta, ctx)
            sgd(params, lr, 1)
            l_sum += l.asscalar() * y.size
            n += y.size

        if (epoch + 1) % pred_period == 0:
            print('epoch %d, perplexity %f, time %.2f sec' % (
                epoch + 1, math.exp(l_sum / n), time.time() - start))
            for prefix in prefixes:
                print(' -', predict_rnn(
                    prefix, pred_len, rnn, params, init_rnn_state,
                    num_hiddens, vocab_size, ctx, idx_to_char, char_to_idx))