012cf6ba1cc018330ffcf90626a8227c7180db08,train_gl.py,,train_epoch,#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#,340

Before Change

        if log_interval and not (i + 1) % log_interval:
            speed = batch_size * log_interval / (time.time() - btic)
            btic = time.time()
            _, top1 = acc_top1_train.get()
            err_top1_train = 1.0 - top1
            logging.info("Epoch[{}] Batch [{}]\tSpeed: {:.2f} samples/sec\ttop1-err={:.4f}\tlr={:.5f}".format(
                epoch + 1, i, speed, err_top1_train, trainer.learning_rate))

    if (batch_size_scale != 1) and (batch_size_extend_count > 0):
        trainer.step(batch_size * batch_size_extend_count)
        for p in net.collect_params().values():
            p.zero_grad()

    throughput = int(batch_size * (i + 1) / (time.time() - tic))
    logging.info("[Epoch {}] speed: {:.2f} samples/sec\ttime cost: {:.2f} sec".format(
        epoch + 1, throughput, time.time() - tic))

    train_loss /= (i + 1)
    _, top1 = acc_top1_train.get()
    err_top1_train = 1.0 - top1
    logging.info("[Epoch {}] training: err-top1={:.4f}\tloss={:.4f}".format(
        epoch + 1, err_top1_train, train_loss))

After Change

        if log_interval and not (i + 1) % log_interval:
            speed = batch_size * log_interval / (time.time() - btic)
            btic = time.time()
            train_accuracy_msg = report_accuracy(metric=train_metric)
            logging.info("Epoch[{}] Batch [{}]\tSpeed: {:.2f} samples/sec\t{}\tlr={:.5f}".format(
                epoch + 1, i, speed, train_accuracy_msg, trainer.learning_rate))

Instances