コード例 #1
0
 def test_value(self):
     key = "img"
     scaler = RandScaleIntensityd(keys=[key], factors=0.5, prob=1.0)
     scaler.set_random_state(seed=0)
     result = scaler({key: self.imt})
     np.random.seed(0)
     expected = (self.imt * (1 + np.random.uniform(low=-0.5, high=0.5))).astype(np.float32)
     np.testing.assert_allclose(result[key], expected)
コード例 #2
0
 def test_value(self):
     key = "img"
     for p in TEST_NDARRAYS:
         scaler = RandScaleIntensityd(keys=[key], factors=0.5, prob=1.0)
         scaler.set_random_state(seed=0)
         result = scaler({key: p(self.imt)})
         np.random.seed(0)
         # simulate the randomize function of transform
         np.random.random()
         expected = (self.imt * (1 + np.random.uniform(low=-0.5, high=0.5))).astype(np.float32)
         assert_allclose(result[key], p(expected), type_test="tensor")
コード例 #3
0
 def test_warn_random_but_has_no_invertible(self):
     transforms = Compose(
         [AddChanneld("image"), RandFlipd("image", prob=1.0), RandScaleIntensityd("image", 0.1, prob=1.0)]
     )
     with self.assertWarns(UserWarning):
         tta = TestTimeAugmentation(transforms, 5, 0, orig_key="image")
         tta(self.get_data(1, (20, 20), data_type=np.float32))
コード例 #4
0
    def test_inverse_compose(self):
        transform = Compose(
            [
                Resized(keys="img", spatial_size=[100, 100, 100]),
                OneOf(
                    [
                        RandScaleIntensityd(keys="img", factors=0.5, prob=1.0),
                        RandShiftIntensityd(keys="img", offsets=0.5, prob=1.0),
                    ]
                ),
            ]
        )
        transform.set_random_state(seed=0)
        result = transform({"img": np.ones((1, 101, 102, 103))})

        result = transform.inverse(result)
        # invert to the original spatial shape
        self.assertTupleEqual(result["img"].shape, (1, 101, 102, 103))
コード例 #5
0
def run_training(train_file_list, valid_file_list, config_info):
    """
    Pipeline to train a dynUNet segmentation model in MONAI. It is composed of the following main blocks:
        * Data Preparation: Extract the filenames and prepare the training/validation processing transforms
        * Load Data: Load training and validation data to PyTorch DataLoader
        * Network Preparation: Define the network, loss function, optimiser and learning rate scheduler
        * MONAI Evaluator: Initialise the dynUNet evaluator, i.e. the class providing utilities to perform validation
            during training. Attach handlers to save the best model on the validation set. A 2D sliding window approach
            on the 3D volume is used at evaluation. The mean 3D Dice is used as validation metric.
        * MONAI Trainer: Initialise the dynUNet trainer, i.e. the class providing utilities to perform the training loop.
        * Run training: The MONAI trainer is run, performing training and validation during training.
    Args:
        train_file_list: .txt or .csv file (with no header) storing two-columns filenames for training:
            image filename in the first column and segmentation filename in the second column.
            The two columns should be separated by a comma.
            See monaifbs/config/mock_train_file_list_for_dynUnet_training.txt for an example of the expected format.
        valid_file_list: .txt or .csv file (with no header) storing two-columns filenames for validation:
            image filename in the first column and segmentation filename in the second column.
            The two columns should be separated by a comma.
            See monaifbs/config/mock_valid_file_list_for_dynUnet_training.txt for an example of the expected format.
        config_info: dict, contains configuration parameters for sampling, network and training.
            See monaifbs/config/monai_dynUnet_training_config.yml for an example of the expected fields.
    """

    """
    Read input and configuration parameters
    """
    # print MONAI config information
    logging.basicConfig(stream=sys.stdout, level=logging.INFO)
    print_config()

    # print to log the parameter setups
    print(yaml.dump(config_info))

    # extract network parameters, perform checks/set defaults if not present and print them to log
    if 'seg_labels' in config_info['training'].keys():
        seg_labels = config_info['training']['seg_labels']
    else:
        seg_labels = [1]
    nr_out_channels = len(seg_labels)
    print("Considering the following {} labels in the segmentation: {}".format(nr_out_channels, seg_labels))
    patch_size = config_info["training"]["inplane_size"] + [1]
    print("Considering patch size = {}".format(patch_size))

    spacing = config_info["training"]["spacing"]
    print("Bringing all images to spacing = {}".format(spacing))

    if 'model_to_load' in config_info['training'].keys() and config_info['training']['model_to_load'] is not None:
        model_to_load = config_info['training']['model_to_load']
        if not os.path.exists(model_to_load):
            raise FileNotFoundError("Cannot find model: {}".format(model_to_load))
        else:
            print("Loading model from {}".format(model_to_load))
    else:
        model_to_load = None

    # set up either GPU or CPU usage
    if torch.cuda.is_available():
        print("\n#### GPU INFORMATION ###")
        print("Using device number: {}, name: {}\n".format(torch.cuda.current_device(), torch.cuda.get_device_name()))
        current_device = torch.device("cuda:0")
    else:
        current_device = torch.device("cpu")
        print("Using device: {}".format(current_device))

    # set determinism if required
    if 'manual_seed' in config_info['training'].keys() and config_info['training']['manual_seed'] is not None:
        seed = config_info['training']['manual_seed']
    else:
        seed = None
    if seed is not None:
        print("Using determinism with seed = {}\n".format(seed))
        set_determinism(seed=seed)

    """
    Setup data output directory
    """
    out_model_dir = os.path.join(config_info['output']['out_dir'],
                                 datetime.now().strftime('%Y-%m-%d_%H-%M-%S') + '_' +
                                 config_info['output']['out_postfix'])
    print("Saving to directory {}\n".format(out_model_dir))
    # create cache directory to store results for Persistent Dataset
    if 'cache_dir' in config_info['output'].keys():
        out_cache_dir = config_info['output']['cache_dir']
    else:
        out_cache_dir = os.path.join(out_model_dir, 'persistent_cache')
    persistent_cache: Path = Path(out_cache_dir)
    persistent_cache.mkdir(parents=True, exist_ok=True)

    """
    Data preparation
    """
    # Read the input files for training and validation
    print("*** Loading input data for training...")

    train_files = create_data_list_of_dictionaries(train_file_list)
    print("Number of inputs for training = {}".format(len(train_files)))

    val_files = create_data_list_of_dictionaries(valid_file_list)
    print("Number of inputs for validation = {}".format(len(val_files)))

    # Define MONAI processing transforms for the training data. This includes:
    # - Load Nifti files and convert to format Batch x Channel x Dim1 x Dim2 x Dim3
    # - CropForegroundd: Reduce the background from the MR image
    # - InPlaneSpacingd: Perform in-plane resampling to the desired spacing, but preserve the resolution along the
    #       last direction (lowest resolution) to avoid introducing motion artefact resampling errors
    # - SpatialPadd: Pad the in-plane size to the defined network input patch size [N, M] if needed
    # - NormalizeIntensityd: Apply whitening
    # - RandSpatialCropd: Crop a random patch from the input with size [B, C, N, M, 1]
    # - SqueezeDimd: Convert the 3D patch to a 2D one as input to the network (i.e. bring it to size [B, C, N, M])
    # - Apply data augmentation (RandZoomd, RandRotated, RandGaussianNoised, RandGaussianSmoothd, RandScaleIntensityd,
    #       RandFlipd)
    # - ToTensor: convert to pytorch tensor
    train_transforms = Compose(
        [
            LoadNiftid(keys=["image", "label"]),
            AddChanneld(keys=["image", "label"]),
            CropForegroundd(keys=["image", "label"], source_key="image"),
            InPlaneSpacingd(
                keys=["image", "label"],
                pixdim=spacing,
                mode=("bilinear", "nearest"),
            ),
            SpatialPadd(keys=["image", "label"], spatial_size=patch_size,
                        mode=["constant", "edge"]),
            NormalizeIntensityd(keys=["image"], nonzero=False, channel_wise=True),
            RandSpatialCropd(keys=["image", "label"], roi_size=patch_size, random_size=False),
            SqueezeDimd(keys=["image", "label"], dim=-1),
            RandZoomd(
                keys=["image", "label"],
                min_zoom=0.9,
                max_zoom=1.2,
                mode=("bilinear", "nearest"),
                align_corners=(True, None),
                prob=0.16,
            ),
            RandRotated(keys=["image", "label"], range_x=90, range_y=90, prob=0.2,
                        keep_size=True, mode=["bilinear", "nearest"],
                        padding_mode=["zeros", "border"]),
            RandGaussianNoised(keys=["image"], std=0.01, prob=0.15),
            RandGaussianSmoothd(
                keys=["image"],
                sigma_x=(0.5, 1.15),
                sigma_y=(0.5, 1.15),
                sigma_z=(0.5, 1.15),
                prob=0.15,
            ),
            RandScaleIntensityd(keys=["image"], factors=0.3, prob=0.15),
            RandFlipd(["image", "label"], spatial_axis=[0, 1], prob=0.5),
            ToTensord(keys=["image", "label"]),
        ]
    )

    # Define MONAI processing transforms for the validation data
    # - Load Nifti files and convert to format Batch x Channel x Dim1 x Dim2 x Dim3
    # - CropForegroundd: Reduce the background from the MR image
    # - InPlaneSpacingd: Perform in-plane resampling to the desired spacing, but preserve the resolution along the
    #       last direction (lowest resolution) to avoid introducing motion artefact resampling errors
    # - SpatialPadd: Pad the in-plane size to the defined network input patch size [N, M] if needed
    # - NormalizeIntensityd: Apply whitening
    # - ToTensor: convert to pytorch tensor
    # NOTE: The validation data is kept 3D as a 2D sliding window approach is used throughout the volume at inference
    val_transforms = Compose(
        [
            LoadNiftid(keys=["image", "label"]),
            AddChanneld(keys=["image", "label"]),
            CropForegroundd(keys=["image", "label"], source_key="image"),
            InPlaneSpacingd(
                keys=["image", "label"],
                pixdim=spacing,
                mode=("bilinear", "nearest"),
            ),
            SpatialPadd(keys=["image", "label"], spatial_size=patch_size, mode=["constant", "edge"]),
            NormalizeIntensityd(keys=["image"], nonzero=False, channel_wise=True),
            ToTensord(keys=["image", "label"]),
        ]
    )

    """
    Load data 
    """
    # create training data loader
    train_ds = PersistentDataset(data=train_files, transform=train_transforms,
                                 cache_dir=persistent_cache)
    train_loader = DataLoader(train_ds,
                              batch_size=config_info['training']['batch_size_train'],
                              shuffle=True,
                              num_workers=config_info['device']['num_workers'])
    check_train_data = misc.first(train_loader)
    print("Training data tensor shapes:")
    print("Image = {}; Label = {}".format(check_train_data["image"].shape, check_train_data["label"].shape))

    # create validation data loader
    if config_info['training']['batch_size_valid'] != 1:
        raise Exception("Batch size different from 1 at validation ar currently not supported")
    val_ds = PersistentDataset(data=val_files, transform=val_transforms, cache_dir=persistent_cache)
    val_loader = DataLoader(val_ds,
                            batch_size=1,
                            shuffle=False,
                            num_workers=config_info['device']['num_workers'])
    check_valid_data = misc.first(val_loader)
    print("Validation data tensor shapes (Example):")
    print("Image = {}; Label = {}\n".format(check_valid_data["image"].shape, check_valid_data["label"].shape))

    """
    Network preparation
    """
    print("*** Preparing the network ...")
    # automatically extracts the strides and kernels based on nnU-Net empirical rules
    spacings = spacing[:2]
    sizes = patch_size[:2]
    strides, kernels = [], []
    while True:
        spacing_ratio = [sp / min(spacings) for sp in spacings]
        stride = [2 if ratio <= 2 and size >= 8 else 1 for (ratio, size) in zip(spacing_ratio, sizes)]
        kernel = [3 if ratio <= 2 else 1 for ratio in spacing_ratio]
        if all(s == 1 for s in stride):
            break
        sizes = [i / j for i, j in zip(sizes, stride)]
        spacings = [i * j for i, j in zip(spacings, stride)]
        kernels.append(kernel)
        strides.append(stride)
    strides.insert(0, len(spacings) * [1])
    kernels.append(len(spacings) * [3])

    # initialise the network
    net = DynUNet(
        spatial_dims=2,
        in_channels=1,
        out_channels=nr_out_channels,
        kernel_size=kernels,
        strides=strides,
        upsample_kernel_size=strides[1:],
        norm_name="instance",
        deep_supervision=True,
        deep_supr_num=2,
        res_block=False,
    ).to(current_device)
    print(net)

    # define the loss function
    loss_function = choose_loss_function(nr_out_channels, config_info)

    # define the optimiser and the learning rate scheduler
    opt = torch.optim.SGD(net.parameters(), lr=float(config_info['training']['lr']), momentum=0.95)
    scheduler = torch.optim.lr_scheduler.LambdaLR(
        opt, lr_lambda=lambda epoch: (1 - epoch / config_info['training']['nr_train_epochs']) ** 0.9
    )

    """
    MONAI evaluator
    """
    print("*** Preparing the dynUNet evaluator engine...\n")
    # val_post_transforms = Compose(
    #     [
    #         Activationsd(keys="pred", sigmoid=True),
    #     ]
    # )
    val_handlers = [
        StatsHandler(output_transform=lambda x: None),
        TensorBoardStatsHandler(log_dir=os.path.join(out_model_dir, "valid"),
                                output_transform=lambda x: None,
                                global_epoch_transform=lambda x: trainer.state.iteration),
        CheckpointSaver(save_dir=out_model_dir, save_dict={"net": net, "opt": opt}, save_key_metric=True,
                        file_prefix='best_valid'),
    ]
    if config_info['output']['val_image_to_tensorboad']:
        val_handlers.append(TensorBoardImageHandler(log_dir=os.path.join(out_model_dir, "valid"),
                                                    batch_transform=lambda x: (x["image"], x["label"]),
                                                    output_transform=lambda x: x["pred"], interval=2))

    # Define customized evaluator
    class DynUNetEvaluator(SupervisedEvaluator):
        def _iteration(self, engine, batchdata):
            inputs, targets = self.prepare_batch(batchdata)
            inputs, targets = inputs.to(engine.state.device), targets.to(engine.state.device)
            flip_inputs_1 = torch.flip(inputs, dims=(2,))
            flip_inputs_2 = torch.flip(inputs, dims=(3,))
            flip_inputs_3 = torch.flip(inputs, dims=(2, 3))

            def _compute_pred():
                pred = self.inferer(inputs, self.network)
                # use random flipping as data augmentation at inference
                flip_pred_1 = torch.flip(self.inferer(flip_inputs_1, self.network), dims=(2,))
                flip_pred_2 = torch.flip(self.inferer(flip_inputs_2, self.network), dims=(3,))
                flip_pred_3 = torch.flip(self.inferer(flip_inputs_3, self.network), dims=(2, 3))
                return (pred + flip_pred_1 + flip_pred_2 + flip_pred_3) / 4

            # execute forward computation
            self.network.eval()
            with torch.no_grad():
                if self.amp:
                    with torch.cuda.amp.autocast():
                        predictions = _compute_pred()
                else:
                    predictions = _compute_pred()
            return {"image": inputs, "label": targets, "pred": predictions}

    evaluator = DynUNetEvaluator(
        device=current_device,
        val_data_loader=val_loader,
        network=net,
        inferer=SlidingWindowInferer2D(roi_size=patch_size, sw_batch_size=4, overlap=0.0),
        post_transform=None,
        key_val_metric={
            "Mean_dice": MeanDice(
                include_background=False,
                to_onehot_y=True,
                mutually_exclusive=True,
                output_transform=lambda x: (x["pred"], x["label"]),
            )
        },
        val_handlers=val_handlers,
        amp=False,
    )

    """
    MONAI trainer
    """
    print("*** Preparing the dynUNet trainer engine...\n")
    # train_post_transforms = Compose(
    #     [
    #         Activationsd(keys="pred", sigmoid=True),
    #     ]
    # )

    validation_every_n_epochs = config_info['training']['validation_every_n_epochs']
    epoch_len = len(train_ds) // train_loader.batch_size
    validation_every_n_iters = validation_every_n_epochs * epoch_len

    # define event handlers for the trainer
    writer_train = SummaryWriter(log_dir=os.path.join(out_model_dir, "train"))
    train_handlers = [
        LrScheduleHandler(lr_scheduler=scheduler, print_lr=True),
        ValidationHandler(validator=evaluator, interval=validation_every_n_iters, epoch_level=False),
        StatsHandler(tag_name="train_loss", output_transform=lambda x: x["loss"]),
        TensorBoardStatsHandler(summary_writer=writer_train,
                                log_dir=os.path.join(out_model_dir, "train"), tag_name="Loss",
                                output_transform=lambda x: x["loss"],
                                global_epoch_transform=lambda x: trainer.state.iteration),
        CheckpointSaver(save_dir=out_model_dir, save_dict={"net": net, "opt": opt},
                        save_final=True,
                        save_interval=2, epoch_level=True,
                        n_saved=config_info['output']['max_nr_models_saved']),
    ]
    if model_to_load is not None:
        train_handlers.append(CheckpointLoader(load_path=model_to_load, load_dict={"net": net, "opt": opt}))

    # define customized trainer
    class DynUNetTrainer(SupervisedTrainer):
        def _iteration(self, engine, batchdata):
            inputs, targets = self.prepare_batch(batchdata)
            inputs, targets = inputs.to(engine.state.device), targets.to(engine.state.device)

            def _compute_loss(preds, label):
                labels = [label] + [interpolate(label, pred.shape[2:]) for pred in preds[1:]]
                return sum([0.5 ** i * self.loss_function(p, l) for i, (p, l) in enumerate(zip(preds, labels))])

            self.network.train()
            self.optimizer.zero_grad()
            if self.amp and self.scaler is not None:
                with torch.cuda.amp.autocast():
                    predictions = self.inferer(inputs, self.network)
                    loss = _compute_loss(predictions, targets)
                self.scaler.scale(loss).backward()
                self.scaler.step(self.optimizer)
                self.scaler.update()
            else:
                predictions = self.inferer(inputs, self.network)
                loss = _compute_loss(predictions, targets).mean()
                loss.backward()
                self.optimizer.step()
            return {"image": inputs, "label": targets, "pred": predictions, "loss": loss.item()}

    trainer = DynUNetTrainer(
        device=current_device,
        max_epochs=config_info['training']['nr_train_epochs'],
        train_data_loader=train_loader,
        network=net,
        optimizer=opt,
        loss_function=loss_function,
        inferer=SimpleInferer(),
        post_transform=None,
        key_train_metric=None,
        train_handlers=train_handlers,
        amp=False,
    )

    """
    Run training
    """
    print("*** Run training...")
    trainer.run()
    print("Done!")
コード例 #6
0
def main_worker(args):
    # disable logging for processes except 0 on every node
    if args.local_rank != 0:
        f = open(os.devnull, "w")
        sys.stdout = sys.stderr = f
    if not os.path.exists(args.dir):
        raise FileNotFoundError(f"Missing directory {args.dir}")

    # initialize the distributed training process, every GPU runs in a process
    dist.init_process_group(backend="nccl", init_method="env://")

    total_start = time.time()
    train_transforms = Compose([
        # load 4 Nifti images and stack them together
        LoadNiftid(keys=["image", "label"]),
        AsChannelFirstd(keys="image"),
        ConvertToMultiChannelBasedOnBratsClassesd(keys="label"),
        Spacingd(keys=["image", "label"],
                 pixdim=(1.5, 1.5, 2.0),
                 mode=("bilinear", "nearest")),
        Orientationd(keys=["image", "label"], axcodes="RAS"),
        RandSpatialCropd(keys=["image", "label"],
                         roi_size=[128, 128, 64],
                         random_size=False),
        NormalizeIntensityd(keys="image", nonzero=True, channel_wise=True),
        RandFlipd(keys=["image", "label"], prob=0.5, spatial_axis=0),
        RandScaleIntensityd(keys="image", factors=0.1, prob=0.5),
        RandShiftIntensityd(keys="image", offsets=0.1, prob=0.5),
        ToTensord(keys=["image", "label"]),
    ])

    # create a training data loader
    train_ds = BratsCacheDataset(
        root_dir=args.dir,
        transform=train_transforms,
        section="training",
        num_workers=4,
        cache_rate=args.cache_rate,
        shuffle=True,
    )
    train_loader = DataLoader(train_ds,
                              batch_size=args.batch_size,
                              shuffle=True,
                              num_workers=args.workers,
                              pin_memory=True)

    # validation transforms and dataset
    val_transforms = Compose([
        LoadNiftid(keys=["image", "label"]),
        AsChannelFirstd(keys="image"),
        ConvertToMultiChannelBasedOnBratsClassesd(keys="label"),
        Spacingd(keys=["image", "label"],
                 pixdim=(1.5, 1.5, 2.0),
                 mode=("bilinear", "nearest")),
        Orientationd(keys=["image", "label"], axcodes="RAS"),
        CenterSpatialCropd(keys=["image", "label"], roi_size=[128, 128, 64]),
        NormalizeIntensityd(keys="image", nonzero=True, channel_wise=True),
        ToTensord(keys=["image", "label"]),
    ])
    val_ds = BratsCacheDataset(
        root_dir=args.dir,
        transform=val_transforms,
        section="validation",
        num_workers=4,
        cache_rate=args.cache_rate,
        shuffle=False,
    )
    val_loader = DataLoader(val_ds,
                            batch_size=args.batch_size,
                            shuffle=False,
                            num_workers=args.workers,
                            pin_memory=True)

    if dist.get_rank() == 0:
        # Logging for TensorBoard
        writer = SummaryWriter(log_dir=args.log_dir)

    # create UNet, DiceLoss and Adam optimizer
    device = torch.device(f"cuda:{args.local_rank}")
    if args.network == "UNet":
        model = UNet(
            dimensions=3,
            in_channels=4,
            out_channels=3,
            channels=(16, 32, 64, 128, 256),
            strides=(2, 2, 2, 2),
            num_res_units=2,
        ).to(device)
    else:
        model = SegResNet(in_channels=4,
                          out_channels=3,
                          init_filters=16,
                          dropout_prob=0.2).to(device)
    loss_function = DiceLoss(to_onehot_y=False,
                             sigmoid=True,
                             squared_pred=True)
    optimizer = torch.optim.Adam(model.parameters(),
                                 lr=args.lr,
                                 weight_decay=1e-5,
                                 amsgrad=True)
    # wrap the model with DistributedDataParallel module
    model = DistributedDataParallel(model, device_ids=[args.local_rank])

    # start a typical PyTorch training
    total_epoch = args.epochs
    best_metric = -1000000
    best_metric_epoch = -1
    epoch_time = AverageMeter("Time", ":6.3f")
    progress = ProgressMeter(total_epoch, [epoch_time], prefix="Epoch: ")
    end = time.time()
    print(f"Time elapsed before training: {end-total_start}")
    for epoch in range(total_epoch):

        train_loss = train(train_loader, model, loss_function, optimizer,
                           epoch, args, device)
        epoch_time.update(time.time() - end)

        if epoch % args.print_freq == 0:
            progress.display(epoch)

        if dist.get_rank() == 0:
            writer.add_scalar("Loss/train", train_loss, epoch)

        if (epoch + 1) % args.val_interval == 0:
            metric, metric_tc, metric_wt, metric_et = evaluate(
                model, val_loader, device)

            if dist.get_rank() == 0:
                writer.add_scalar("Mean Dice/val", metric, epoch)
                writer.add_scalar("Mean Dice TC/val", metric_tc, epoch)
                writer.add_scalar("Mean Dice WT/val", metric_wt, epoch)
                writer.add_scalar("Mean Dice ET/val", metric_et, epoch)
                if metric > best_metric:
                    best_metric = metric
                    best_metric_epoch = epoch + 1
                print(
                    f"current epoch: {epoch + 1} current mean dice: {metric:.4f}"
                    f" tc: {metric_tc:.4f} wt: {metric_wt:.4f} et: {metric_et:.4f}"
                    f"\nbest mean dice: {best_metric:.4f} at epoch: {best_metric_epoch}"
                )
        end = time.time()
        print(f"Time elapsed after epoch {epoch + 1} is {end - total_start}")

    if dist.get_rank() == 0:
        print(
            f"train completed, best_metric: {best_metric:.4f}  at epoch: {best_metric_epoch}"
        )
        # all processes should see same parameters as they all start from same
        # random parameters and gradients are synchronized in backward passes,
        # therefore, saving it in one process is sufficient
        torch.save(model.state_dict(), "final_model.pth")
        writer.flush()
    dist.destroy_process_group()
コード例 #7
0
def get_task_transforms(mode, task_id, pos_sample_num, neg_sample_num,
                        num_samples):
    if mode != "test":
        keys = ["image", "label"]
    else:
        keys = ["image"]

    load_transforms = [
        LoadImaged(keys=keys),
        EnsureChannelFirstd(keys=keys),
    ]
    # 2. sampling
    sample_transforms = [
        PreprocessAnisotropic(
            keys=keys,
            clip_values=clip_values[task_id],
            pixdim=spacing[task_id],
            normalize_values=normalize_values[task_id],
            model_mode=mode,
        ),
    ]
    # 3. spatial transforms
    if mode == "train":
        other_transforms = [
            SpatialPadd(keys=["image", "label"],
                        spatial_size=patch_size[task_id]),
            RandCropByPosNegLabeld(
                keys=["image", "label"],
                label_key="label",
                spatial_size=patch_size[task_id],
                pos=pos_sample_num,
                neg=neg_sample_num,
                num_samples=num_samples,
                image_key="image",
                image_threshold=0,
            ),
            RandZoomd(
                keys=["image", "label"],
                min_zoom=0.9,
                max_zoom=1.2,
                mode=("trilinear", "nearest"),
                align_corners=(True, None),
                prob=0.15,
            ),
            RandGaussianNoised(keys=["image"], std=0.01, prob=0.15),
            RandGaussianSmoothd(
                keys=["image"],
                sigma_x=(0.5, 1.15),
                sigma_y=(0.5, 1.15),
                sigma_z=(0.5, 1.15),
                prob=0.15,
            ),
            RandScaleIntensityd(keys=["image"], factors=0.3, prob=0.15),
            RandFlipd(["image", "label"], spatial_axis=[0], prob=0.5),
            RandFlipd(["image", "label"], spatial_axis=[1], prob=0.5),
            RandFlipd(["image", "label"], spatial_axis=[2], prob=0.5),
            CastToTyped(keys=["image", "label"], dtype=(np.float32, np.uint8)),
            EnsureTyped(keys=["image", "label"]),
        ]
    elif mode == "validation":
        other_transforms = [
            CastToTyped(keys=["image", "label"], dtype=(np.float32, np.uint8)),
            EnsureTyped(keys=["image", "label"]),
        ]
    else:
        other_transforms = [
            CastToTyped(keys=["image"], dtype=(np.float32)),
            EnsureTyped(keys=["image"]),
        ]

    all_transforms = load_transforms + sample_transforms + other_transforms
    return Compose(all_transforms)
コード例 #8
0
def main_worker(args):
    # disable logging for processes except 0 on every node
    if args.local_rank != 0:
        f = open(os.devnull, "w")
        sys.stdout = sys.stderr = f
    if not os.path.exists(args.dir):
        raise FileNotFoundError(f"missing directory {args.dir}")

    # initialize the distributed training process, every GPU runs in a process
    dist.init_process_group(backend="nccl", init_method="env://")
    device = torch.device(f"cuda:{args.local_rank}")
    torch.cuda.set_device(device)
    # use amp to accelerate training
    scaler = torch.cuda.amp.GradScaler()
    torch.backends.cudnn.benchmark = True

    total_start = time.time()
    train_transforms = Compose([
        # load 4 Nifti images and stack them together
        LoadImaged(keys=["image", "label"]),
        EnsureChannelFirstd(keys="image"),
        ConvertToMultiChannelBasedOnBratsClassesd(keys="label"),
        Orientationd(keys=["image", "label"], axcodes="RAS"),
        Spacingd(
            keys=["image", "label"],
            pixdim=(1.0, 1.0, 1.0),
            mode=("bilinear", "nearest"),
        ),
        EnsureTyped(keys=["image", "label"]),
        ToDeviced(keys=["image", "label"], device=device),
        RandSpatialCropd(keys=["image", "label"],
                         roi_size=[224, 224, 144],
                         random_size=False),
        RandFlipd(keys=["image", "label"], prob=0.5, spatial_axis=0),
        RandFlipd(keys=["image", "label"], prob=0.5, spatial_axis=1),
        RandFlipd(keys=["image", "label"], prob=0.5, spatial_axis=2),
        NormalizeIntensityd(keys="image", nonzero=True, channel_wise=True),
        RandScaleIntensityd(keys="image", factors=0.1, prob=0.5),
        RandShiftIntensityd(keys="image", offsets=0.1, prob=0.5),
    ])

    # create a training data loader
    train_ds = BratsCacheDataset(
        root_dir=args.dir,
        transform=train_transforms,
        section="training",
        num_workers=4,
        cache_rate=args.cache_rate,
        shuffle=True,
    )
    # ThreadDataLoader can be faster if no IO operations when caching all the data in memory
    train_loader = ThreadDataLoader(train_ds,
                                    num_workers=0,
                                    batch_size=args.batch_size,
                                    shuffle=True)

    # validation transforms and dataset
    val_transforms = Compose([
        LoadImaged(keys=["image", "label"]),
        EnsureChannelFirstd(keys="image"),
        ConvertToMultiChannelBasedOnBratsClassesd(keys="label"),
        Orientationd(keys=["image", "label"], axcodes="RAS"),
        Spacingd(
            keys=["image", "label"],
            pixdim=(1.0, 1.0, 1.0),
            mode=("bilinear", "nearest"),
        ),
        NormalizeIntensityd(keys="image", nonzero=True, channel_wise=True),
        EnsureTyped(keys=["image", "label"]),
        ToDeviced(keys=["image", "label"], device=device),
    ])
    val_ds = BratsCacheDataset(
        root_dir=args.dir,
        transform=val_transforms,
        section="validation",
        num_workers=4,
        cache_rate=args.cache_rate,
        shuffle=False,
    )
    # ThreadDataLoader can be faster if no IO operations when caching all the data in memory
    val_loader = ThreadDataLoader(val_ds,
                                  num_workers=0,
                                  batch_size=args.batch_size,
                                  shuffle=False)

    # create network, loss function and optimizer
    if args.network == "SegResNet":
        model = SegResNet(
            blocks_down=[1, 2, 2, 4],
            blocks_up=[1, 1, 1],
            init_filters=16,
            in_channels=4,
            out_channels=3,
            dropout_prob=0.0,
        ).to(device)
    else:
        model = UNet(
            spatial_dims=3,
            in_channels=4,
            out_channels=3,
            channels=(16, 32, 64, 128, 256),
            strides=(2, 2, 2, 2),
            num_res_units=2,
        ).to(device)

    loss_function = DiceFocalLoss(
        smooth_nr=1e-5,
        smooth_dr=1e-5,
        squared_pred=True,
        to_onehot_y=False,
        sigmoid=True,
        batch=True,
    )
    optimizer = Novograd(model.parameters(), lr=args.lr)
    lr_scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(
        optimizer, T_max=args.epochs)
    # wrap the model with DistributedDataParallel module
    model = DistributedDataParallel(model, device_ids=[device])

    dice_metric = DiceMetric(include_background=True, reduction="mean")
    dice_metric_batch = DiceMetric(include_background=True,
                                   reduction="mean_batch")

    post_trans = Compose(
        [EnsureType(),
         Activations(sigmoid=True),
         AsDiscrete(threshold=0.5)])

    # start a typical PyTorch training
    best_metric = -1
    best_metric_epoch = -1
    print(f"time elapsed before training: {time.time() - total_start}")
    train_start = time.time()
    for epoch in range(args.epochs):
        epoch_start = time.time()
        print("-" * 10)
        print(f"epoch {epoch + 1}/{args.epochs}")
        epoch_loss = train(train_loader, model, loss_function, optimizer,
                           lr_scheduler, scaler)
        print(f"epoch {epoch + 1} average loss: {epoch_loss:.4f}")

        if (epoch + 1) % args.val_interval == 0:
            metric, metric_tc, metric_wt, metric_et = evaluate(
                model, val_loader, dice_metric, dice_metric_batch, post_trans)

            if metric > best_metric:
                best_metric = metric
                best_metric_epoch = epoch + 1
                if dist.get_rank() == 0:
                    torch.save(model.state_dict(), "best_metric_model.pth")
            print(
                f"current epoch: {epoch + 1} current mean dice: {metric:.4f}"
                f" tc: {metric_tc:.4f} wt: {metric_wt:.4f} et: {metric_et:.4f}"
                f"\nbest mean dice: {best_metric:.4f} at epoch: {best_metric_epoch}"
            )

        print(
            f"time consuming of epoch {epoch + 1} is: {(time.time() - epoch_start):.4f}"
        )

    print(
        f"train completed, best_metric: {best_metric:.4f} at epoch: {best_metric_epoch},"
        f" total train time: {(time.time() - train_start):.4f}")
    dist.destroy_process_group()
コード例 #9
0
def main():
    parser = argparse.ArgumentParser(description="training")
    parser.add_argument(
        "--checkpoint",
        type=str,
        default=None,
        help="checkpoint full path",
    )
    parser.add_argument(
        "--factor_ram_cost",
        default=0.0,
        type=float,
        help="factor to determine RAM cost in the searched architecture",
    )
    parser.add_argument(
        "--fold",
        action="store",
        required=True,
        help="fold index in N-fold cross-validation",
    )
    parser.add_argument(
        "--json",
        action="store",
        required=True,
        help="full path of .json file",
    )
    parser.add_argument(
        "--json_key",
        action="store",
        required=True,
        help="selected key in .json data list",
    )
    parser.add_argument(
        "--local_rank",
        required=int,
        help="local process rank",
    )
    parser.add_argument(
        "--num_folds",
        action="store",
        required=True,
        help="number of folds in cross-validation",
    )
    parser.add_argument(
        "--output_root",
        action="store",
        required=True,
        help="output root",
    )
    parser.add_argument(
        "--root",
        action="store",
        required=True,
        help="data root",
    )
    args = parser.parse_args()

    logging.basicConfig(stream=sys.stdout, level=logging.INFO)

    if not os.path.exists(args.output_root):
        os.makedirs(args.output_root, exist_ok=True)

    amp = True
    determ = True
    factor_ram_cost = args.factor_ram_cost
    fold = int(args.fold)
    input_channels = 1
    learning_rate = 0.025
    learning_rate_arch = 0.001
    learning_rate_milestones = np.array([0.4, 0.8])
    num_images_per_batch = 1
    num_epochs = 1430  # around 20k iteration
    num_epochs_per_validation = 100
    num_epochs_warmup = 715
    num_folds = int(args.num_folds)
    num_patches_per_image = 1
    num_sw_batch_size = 6
    output_classes = 3
    overlap_ratio = 0.625
    patch_size = (96, 96, 96)
    patch_size_valid = (96, 96, 96)
    spacing = [1.0, 1.0, 1.0]

    print("factor_ram_cost", factor_ram_cost)

    # deterministic training
    if determ:
        set_determinism(seed=0)

    # initialize the distributed training process, every GPU runs in a process
    dist.init_process_group(backend="nccl", init_method="env://")

    # dist.barrier()
    world_size = dist.get_world_size()

    with open(args.json, "r") as f:
        json_data = json.load(f)

    split = len(json_data[args.json_key]) // num_folds
    list_train = json_data[args.json_key][:(
        split * fold)] + json_data[args.json_key][(split * (fold + 1)):]
    list_valid = json_data[args.json_key][(split * fold):(split * (fold + 1))]

    # training data
    files = []
    for _i in range(len(list_train)):
        str_img = os.path.join(args.root, list_train[_i]["image"])
        str_seg = os.path.join(args.root, list_train[_i]["label"])

        if (not os.path.exists(str_img)) or (not os.path.exists(str_seg)):
            continue

        files.append({"image": str_img, "label": str_seg})
    train_files = files

    random.shuffle(train_files)

    train_files_w = train_files[:len(train_files) // 2]
    train_files_w = partition_dataset(data=train_files_w,
                                      shuffle=True,
                                      num_partitions=world_size,
                                      even_divisible=True)[dist.get_rank()]
    print("train_files_w:", len(train_files_w))

    train_files_a = train_files[len(train_files) // 2:]
    train_files_a = partition_dataset(data=train_files_a,
                                      shuffle=True,
                                      num_partitions=world_size,
                                      even_divisible=True)[dist.get_rank()]
    print("train_files_a:", len(train_files_a))

    # validation data
    files = []
    for _i in range(len(list_valid)):
        str_img = os.path.join(args.root, list_valid[_i]["image"])
        str_seg = os.path.join(args.root, list_valid[_i]["label"])

        if (not os.path.exists(str_img)) or (not os.path.exists(str_seg)):
            continue

        files.append({"image": str_img, "label": str_seg})
    val_files = files
    val_files = partition_dataset(data=val_files,
                                  shuffle=False,
                                  num_partitions=world_size,
                                  even_divisible=False)[dist.get_rank()]
    print("val_files:", len(val_files))

    # network architecture
    device = torch.device(f"cuda:{args.local_rank}")
    torch.cuda.set_device(device)

    train_transforms = Compose([
        LoadImaged(keys=["image", "label"]),
        EnsureChannelFirstd(keys=["image", "label"]),
        Orientationd(keys=["image", "label"], axcodes="RAS"),
        Spacingd(keys=["image", "label"],
                 pixdim=spacing,
                 mode=("bilinear", "nearest"),
                 align_corners=(True, True)),
        CastToTyped(keys=["image"], dtype=(torch.float32)),
        ScaleIntensityRanged(keys=["image"],
                             a_min=-87.0,
                             a_max=199.0,
                             b_min=0.0,
                             b_max=1.0,
                             clip=True),
        CastToTyped(keys=["image", "label"], dtype=(np.float16, np.uint8)),
        CopyItemsd(keys=["label"], times=1, names=["label4crop"]),
        Lambdad(
            keys=["label4crop"],
            func=lambda x: np.concatenate(tuple([
                ndimage.binary_dilation(
                    (x == _k).astype(x.dtype), iterations=48).astype(x.dtype)
                for _k in range(output_classes)
            ]),
                                          axis=0),
            overwrite=True,
        ),
        EnsureTyped(keys=["image", "label"]),
        CastToTyped(keys=["image"], dtype=(torch.float32)),
        SpatialPadd(keys=["image", "label", "label4crop"],
                    spatial_size=patch_size,
                    mode=["reflect", "constant", "constant"]),
        RandCropByLabelClassesd(keys=["image", "label"],
                                label_key="label4crop",
                                num_classes=output_classes,
                                ratios=[
                                    1,
                                ] * output_classes,
                                spatial_size=patch_size,
                                num_samples=num_patches_per_image),
        Lambdad(keys=["label4crop"], func=lambda x: 0),
        RandRotated(keys=["image", "label"],
                    range_x=0.3,
                    range_y=0.3,
                    range_z=0.3,
                    mode=["bilinear", "nearest"],
                    prob=0.2),
        RandZoomd(keys=["image", "label"],
                  min_zoom=0.8,
                  max_zoom=1.2,
                  mode=["trilinear", "nearest"],
                  align_corners=[True, None],
                  prob=0.16),
        RandGaussianSmoothd(keys=["image"],
                            sigma_x=(0.5, 1.15),
                            sigma_y=(0.5, 1.15),
                            sigma_z=(0.5, 1.15),
                            prob=0.15),
        RandScaleIntensityd(keys=["image"], factors=0.3, prob=0.5),
        RandShiftIntensityd(keys=["image"], offsets=0.1, prob=0.5),
        RandGaussianNoised(keys=["image"], std=0.01, prob=0.15),
        RandFlipd(keys=["image", "label"], spatial_axis=0, prob=0.5),
        RandFlipd(keys=["image", "label"], spatial_axis=1, prob=0.5),
        RandFlipd(keys=["image", "label"], spatial_axis=2, prob=0.5),
        CastToTyped(keys=["image", "label"],
                    dtype=(torch.float32, torch.uint8)),
        ToTensord(keys=["image", "label"]),
    ])

    val_transforms = Compose([
        LoadImaged(keys=["image", "label"]),
        EnsureChannelFirstd(keys=["image", "label"]),
        Orientationd(keys=["image", "label"], axcodes="RAS"),
        Spacingd(keys=["image", "label"],
                 pixdim=spacing,
                 mode=("bilinear", "nearest"),
                 align_corners=(True, True)),
        CastToTyped(keys=["image"], dtype=(torch.float32)),
        ScaleIntensityRanged(keys=["image"],
                             a_min=-87.0,
                             a_max=199.0,
                             b_min=0.0,
                             b_max=1.0,
                             clip=True),
        CastToTyped(keys=["image", "label"], dtype=(np.float32, np.uint8)),
        EnsureTyped(keys=["image", "label"]),
        ToTensord(keys=["image", "label"])
    ])

    train_ds_a = monai.data.CacheDataset(data=train_files_a,
                                         transform=train_transforms,
                                         cache_rate=1.0,
                                         num_workers=8)
    train_ds_w = monai.data.CacheDataset(data=train_files_w,
                                         transform=train_transforms,
                                         cache_rate=1.0,
                                         num_workers=8)
    val_ds = monai.data.CacheDataset(data=val_files,
                                     transform=val_transforms,
                                     cache_rate=1.0,
                                     num_workers=2)

    # monai.data.Dataset can be used as alternatives when debugging or RAM space is limited.
    # train_ds_a = monai.data.Dataset(data=train_files_a, transform=train_transforms)
    # train_ds_w = monai.data.Dataset(data=train_files_w, transform=train_transforms)
    # val_ds = monai.data.Dataset(data=val_files, transform=val_transforms)

    train_loader_a = ThreadDataLoader(train_ds_a,
                                      num_workers=0,
                                      batch_size=num_images_per_batch,
                                      shuffle=True)
    train_loader_w = ThreadDataLoader(train_ds_w,
                                      num_workers=0,
                                      batch_size=num_images_per_batch,
                                      shuffle=True)
    val_loader = ThreadDataLoader(val_ds,
                                  num_workers=0,
                                  batch_size=1,
                                  shuffle=False)

    # DataLoader can be used as alternatives when ThreadDataLoader is less efficient.
    # train_loader_a = DataLoader(train_ds_a, batch_size=num_images_per_batch, shuffle=True, num_workers=2, pin_memory=torch.cuda.is_available())
    # train_loader_w = DataLoader(train_ds_w, batch_size=num_images_per_batch, shuffle=True, num_workers=2, pin_memory=torch.cuda.is_available())
    # val_loader = DataLoader(val_ds, batch_size=1, shuffle=False, num_workers=2, pin_memory=torch.cuda.is_available())

    dints_space = monai.networks.nets.TopologySearch(
        channel_mul=0.5,
        num_blocks=12,
        num_depths=4,
        use_downsample=True,
        device=device,
    )

    model = monai.networks.nets.DiNTS(
        dints_space=dints_space,
        in_channels=input_channels,
        num_classes=output_classes,
        use_downsample=True,
    )

    model = model.to(device)

    model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model)

    post_pred = Compose(
        [EnsureType(),
         AsDiscrete(argmax=True, to_onehot=output_classes)])
    post_label = Compose([EnsureType(), AsDiscrete(to_onehot=output_classes)])

    # loss function
    loss_func = monai.losses.DiceCELoss(
        include_background=False,
        to_onehot_y=True,
        softmax=True,
        squared_pred=True,
        batch=True,
        smooth_nr=0.00001,
        smooth_dr=0.00001,
    )

    # optimizer
    optimizer = torch.optim.SGD(model.weight_parameters(),
                                lr=learning_rate * world_size,
                                momentum=0.9,
                                weight_decay=0.00004)
    arch_optimizer_a = torch.optim.Adam([dints_space.log_alpha_a],
                                        lr=learning_rate_arch * world_size,
                                        betas=(0.5, 0.999),
                                        weight_decay=0.0)
    arch_optimizer_c = torch.optim.Adam([dints_space.log_alpha_c],
                                        lr=learning_rate_arch * world_size,
                                        betas=(0.5, 0.999),
                                        weight_decay=0.0)

    print()

    if torch.cuda.device_count() > 1:
        if dist.get_rank() == 0:
            print("Let's use", torch.cuda.device_count(), "GPUs!")

        model = DistributedDataParallel(model,
                                        device_ids=[device],
                                        find_unused_parameters=True)

    if args.checkpoint != None and os.path.isfile(args.checkpoint):
        print("[info] fine-tuning pre-trained checkpoint {0:s}".format(
            args.checkpoint))
        model.load_state_dict(torch.load(args.checkpoint, map_location=device))
        torch.cuda.empty_cache()
    else:
        print("[info] training from scratch")

    # amp
    if amp:
        from torch.cuda.amp import autocast, GradScaler
        scaler = GradScaler()
        if dist.get_rank() == 0:
            print("[info] amp enabled")

    # start a typical PyTorch training
    val_interval = num_epochs_per_validation
    best_metric = -1
    best_metric_epoch = -1
    epoch_loss_values = list()
    idx_iter = 0
    metric_values = list()

    if dist.get_rank() == 0:
        writer = SummaryWriter(
            log_dir=os.path.join(args.output_root, "Events"))

        with open(os.path.join(args.output_root, "accuracy_history.csv"),
                  "a") as f:
            f.write("epoch\tmetric\tloss\tlr\ttime\titer\n")

    dataloader_a_iterator = iter(train_loader_a)

    start_time = time.time()
    for epoch in range(num_epochs):
        decay = 0.5**np.sum([
            (epoch - num_epochs_warmup) /
            (num_epochs - num_epochs_warmup) > learning_rate_milestones
        ])
        lr = learning_rate * decay
        for param_group in optimizer.param_groups:
            param_group["lr"] = lr

        if dist.get_rank() == 0:
            print("-" * 10)
            print(f"epoch {epoch + 1}/{num_epochs}")
            print("learning rate is set to {}".format(lr))

        model.train()
        epoch_loss = 0
        loss_torch = torch.zeros(2, dtype=torch.float, device=device)
        epoch_loss_arch = 0
        loss_torch_arch = torch.zeros(2, dtype=torch.float, device=device)
        step = 0

        for batch_data in train_loader_w:
            step += 1
            inputs, labels = batch_data["image"].to(
                device), batch_data["label"].to(device)
            if world_size == 1:
                for _ in model.weight_parameters():
                    _.requires_grad = True
            else:
                for _ in model.module.weight_parameters():
                    _.requires_grad = True
            dints_space.log_alpha_a.requires_grad = False
            dints_space.log_alpha_c.requires_grad = False

            optimizer.zero_grad()

            if amp:
                with autocast():
                    outputs = model(inputs)
                    if output_classes == 2:
                        loss = loss_func(torch.flip(outputs, dims=[1]),
                                         1 - labels)
                    else:
                        loss = loss_func(outputs, labels)

                scaler.scale(loss).backward()
                scaler.step(optimizer)
                scaler.update()
            else:
                outputs = model(inputs)
                if output_classes == 2:
                    loss = loss_func(torch.flip(outputs, dims=[1]), 1 - labels)
                else:
                    loss = loss_func(outputs, labels)
                loss.backward()
                optimizer.step()

            epoch_loss += loss.item()
            loss_torch[0] += loss.item()
            loss_torch[1] += 1.0
            epoch_len = len(train_loader_w)
            idx_iter += 1

            if dist.get_rank() == 0:
                print("[{0}] ".format(str(datetime.now())[:19]) +
                      f"{step}/{epoch_len}, train_loss: {loss.item():.4f}")
                writer.add_scalar("train_loss", loss.item(),
                                  epoch_len * epoch + step)

            if epoch < num_epochs_warmup:
                continue

            try:
                sample_a = next(dataloader_a_iterator)
            except StopIteration:
                dataloader_a_iterator = iter(train_loader_a)
                sample_a = next(dataloader_a_iterator)
            inputs_search, labels_search = sample_a["image"].to(
                device), sample_a["label"].to(device)
            if world_size == 1:
                for _ in model.weight_parameters():
                    _.requires_grad = False
            else:
                for _ in model.module.weight_parameters():
                    _.requires_grad = False
            dints_space.log_alpha_a.requires_grad = True
            dints_space.log_alpha_c.requires_grad = True

            # linear increase topology and RAM loss
            entropy_alpha_c = torch.tensor(0.).to(device)
            entropy_alpha_a = torch.tensor(0.).to(device)
            ram_cost_full = torch.tensor(0.).to(device)
            ram_cost_usage = torch.tensor(0.).to(device)
            ram_cost_loss = torch.tensor(0.).to(device)
            topology_loss = torch.tensor(0.).to(device)

            probs_a, arch_code_prob_a = dints_space.get_prob_a(child=True)
            entropy_alpha_a = -((probs_a) * torch.log(probs_a + 1e-5)).mean()
            entropy_alpha_c = -(F.softmax(dints_space.log_alpha_c, dim=-1) * \
                F.log_softmax(dints_space.log_alpha_c, dim=-1)).mean()
            topology_loss = dints_space.get_topology_entropy(probs_a)

            ram_cost_full = dints_space.get_ram_cost_usage(inputs.shape,
                                                           full=True)
            ram_cost_usage = dints_space.get_ram_cost_usage(inputs.shape)
            ram_cost_loss = torch.abs(factor_ram_cost -
                                      ram_cost_usage / ram_cost_full)

            arch_optimizer_a.zero_grad()
            arch_optimizer_c.zero_grad()

            combination_weights = (epoch - num_epochs_warmup) / (
                num_epochs - num_epochs_warmup)

            if amp:
                with autocast():
                    outputs_search = model(inputs_search)
                    if output_classes == 2:
                        loss = loss_func(torch.flip(outputs_search, dims=[1]),
                                         1 - labels_search)
                    else:
                        loss = loss_func(outputs_search, labels_search)

                    loss += combination_weights * ((entropy_alpha_a + entropy_alpha_c) + ram_cost_loss \
                                                    + 0.001 * topology_loss)

                scaler.scale(loss).backward()
                scaler.step(arch_optimizer_a)
                scaler.step(arch_optimizer_c)
                scaler.update()
            else:
                outputs_search = model(inputs_search)
                if output_classes == 2:
                    loss = loss_func(torch.flip(outputs_search, dims=[1]),
                                     1 - labels_search)
                else:
                    loss = loss_func(outputs_search, labels_search)

                loss += 1.0 * (combination_weights * (entropy_alpha_a + entropy_alpha_c) + ram_cost_loss \
                                + 0.001 * topology_loss)

                loss.backward()
                arch_optimizer_a.step()
                arch_optimizer_c.step()

            epoch_loss_arch += loss.item()
            loss_torch_arch[0] += loss.item()
            loss_torch_arch[1] += 1.0

            if dist.get_rank() == 0:
                print(
                    "[{0}] ".format(str(datetime.now())[:19]) +
                    f"{step}/{epoch_len}, train_loss_arch: {loss.item():.4f}")
                writer.add_scalar("train_loss_arch", loss.item(),
                                  epoch_len * epoch + step)

        # synchronizes all processes and reduce results
        dist.barrier()
        dist.all_reduce(loss_torch, op=torch.distributed.ReduceOp.SUM)
        loss_torch = loss_torch.tolist()
        loss_torch_arch = loss_torch_arch.tolist()
        if dist.get_rank() == 0:
            loss_torch_epoch = loss_torch[0] / loss_torch[1]
            print(
                f"epoch {epoch + 1} average loss: {loss_torch_epoch:.4f}, best mean dice: {best_metric:.4f} at epoch {best_metric_epoch}"
            )

            if epoch >= num_epochs_warmup:
                loss_torch_arch_epoch = loss_torch_arch[0] / loss_torch_arch[1]
                print(
                    f"epoch {epoch + 1} average arch loss: {loss_torch_arch_epoch:.4f}, best mean dice: {best_metric:.4f} at epoch {best_metric_epoch}"
                )

        if (epoch + 1) % val_interval == 0:
            torch.cuda.empty_cache()
            model.eval()
            with torch.no_grad():
                metric = torch.zeros((output_classes - 1) * 2,
                                     dtype=torch.float,
                                     device=device)
                metric_sum = 0.0
                metric_count = 0
                metric_mat = []
                val_images = None
                val_labels = None
                val_outputs = None

                _index = 0
                for val_data in val_loader:
                    val_images = val_data["image"].to(device)
                    val_labels = val_data["label"].to(device)

                    roi_size = patch_size_valid
                    sw_batch_size = num_sw_batch_size

                    if amp:
                        with torch.cuda.amp.autocast():
                            pred = sliding_window_inference(
                                val_images,
                                roi_size,
                                sw_batch_size,
                                lambda x: model(x),
                                mode="gaussian",
                                overlap=overlap_ratio,
                            )
                    else:
                        pred = sliding_window_inference(
                            val_images,
                            roi_size,
                            sw_batch_size,
                            lambda x: model(x),
                            mode="gaussian",
                            overlap=overlap_ratio,
                        )
                    val_outputs = pred

                    val_outputs = post_pred(val_outputs[0, ...])
                    val_outputs = val_outputs[None, ...]
                    val_labels = post_label(val_labels[0, ...])
                    val_labels = val_labels[None, ...]

                    value = compute_meandice(y_pred=val_outputs,
                                             y=val_labels,
                                             include_background=False)

                    print(_index + 1, "/", len(val_loader), value)

                    metric_count += len(value)
                    metric_sum += value.sum().item()
                    metric_vals = value.cpu().numpy()
                    if len(metric_mat) == 0:
                        metric_mat = metric_vals
                    else:
                        metric_mat = np.concatenate((metric_mat, metric_vals),
                                                    axis=0)

                    for _c in range(output_classes - 1):
                        val0 = torch.nan_to_num(value[0, _c], nan=0.0)
                        val1 = 1.0 - torch.isnan(value[0, 0]).float()
                        metric[2 * _c] += val0 * val1
                        metric[2 * _c + 1] += val1

                    _index += 1

                # synchronizes all processes and reduce results
                dist.barrier()
                dist.all_reduce(metric, op=torch.distributed.ReduceOp.SUM)
                metric = metric.tolist()
                if dist.get_rank() == 0:
                    for _c in range(output_classes - 1):
                        print(
                            "evaluation metric - class {0:d}:".format(_c + 1),
                            metric[2 * _c] / metric[2 * _c + 1])
                    avg_metric = 0
                    for _c in range(output_classes - 1):
                        avg_metric += metric[2 * _c] / metric[2 * _c + 1]
                    avg_metric = avg_metric / float(output_classes - 1)
                    print("avg_metric", avg_metric)

                    if avg_metric > best_metric:
                        best_metric = avg_metric
                        best_metric_epoch = epoch + 1
                        best_metric_iterations = idx_iter

                    node_a_d, arch_code_a_d, arch_code_c_d, arch_code_a_max_d = dints_space.decode(
                    )
                    torch.save(
                        {
                            "node_a": node_a_d,
                            "arch_code_a": arch_code_a_d,
                            "arch_code_a_max": arch_code_a_max_d,
                            "arch_code_c": arch_code_c_d,
                            "iter_num": idx_iter,
                            "epochs": epoch + 1,
                            "best_dsc": best_metric,
                            "best_path": best_metric_iterations,
                        },
                        os.path.join(args.output_root,
                                     "search_code_" + str(idx_iter) + ".pth"),
                    )
                    print("saved new best metric model")

                    dict_file = {}
                    dict_file["best_avg_dice_score"] = float(best_metric)
                    dict_file["best_avg_dice_score_epoch"] = int(
                        best_metric_epoch)
                    dict_file["best_avg_dice_score_iteration"] = int(idx_iter)
                    with open(os.path.join(args.output_root, "progress.yaml"),
                              "w") as out_file:
                        documents = yaml.dump(dict_file, stream=out_file)

                    print(
                        "current epoch: {} current mean dice: {:.4f} best mean dice: {:.4f} at epoch {}"
                        .format(epoch + 1, avg_metric, best_metric,
                                best_metric_epoch))

                    current_time = time.time()
                    elapsed_time = (current_time - start_time) / 60.0
                    with open(
                            os.path.join(args.output_root,
                                         "accuracy_history.csv"), "a") as f:
                        f.write(
                            "{0:d}\t{1:.5f}\t{2:.5f}\t{3:.5f}\t{4:.1f}\t{5:d}\n"
                            .format(epoch + 1, avg_metric, loss_torch_epoch,
                                    lr, elapsed_time, idx_iter))

                dist.barrier()

            torch.cuda.empty_cache()

    print(
        f"train completed, best_metric: {best_metric:.4f} at epoch: {best_metric_epoch}"
    )

    if dist.get_rank() == 0:
        writer.close()

    dist.destroy_process_group()

    return
コード例 #10
0
    SpatialPadd(keys=("input", "mask"), spatial_size=(1120, 1120)),
    RandFlipd(keys=("input", "mask"), prob=0.5, spatial_axis=1),
    RandAffined(
        keys=("input", "mask"),
        prob=0.5,
        rotate_range=np.pi / 14.4,
        translate_range=(70, 70),
        scale_range=(0.1, 0.1),
        as_tensor_output=False,
    ),
    RandSpatialCropd(
        keys=("input", "mask"),
        roi_size=(cfg.img_size[0], cfg.img_size[1]),
        random_size=False,
    ),
    RandScaleIntensityd(keys="input", factors=(-0.2, 0.2), prob=0.5),
    RandShiftIntensityd(keys="input", offsets=(-51, 51), prob=0.5),
    RandLambdad(keys="input", func=lambda x: 255 - x, prob=0.5),
    RandCoarseDropoutd(
        keys=("input", "mask"),
        holes=8,
        spatial_size=(1, 1),
        max_spatial_size=(102, 102),
        prob=0.5,
    ),
    CastToTyped(keys="input", dtype=np.float32),
    NormalizeIntensityd(keys="input", nonzero=False),
    Lambdad(keys="input", func=lambda x: x.clip(-20, 20)),
    EnsureTyped(keys=("input", "mask")),
])
コード例 #11
0
    # load 4 Nifti images and stack them together
    LoadImaged(keys=["image", "label"]),
    AsChannelFirstd(keys="image"),
    ConvertToMultiChannelBasedOnBratsClassesd(keys="label"),
    Spacingd(
        keys=["image", "label"],
        pixdim=(1.5, 1.5, 2.0),
        mode=("bilinear", "nearest"),
    ),
    Orientationd(keys=["image", "label"], axcodes="RAS"),
    RandSpatialCropd(keys=["image", "label"],
                     roi_size=[128, 128, 64],
                     random_size=False),
    RandFlipd(keys=["image", "label"], prob=0.5, spatial_axis=0),
    NormalizeIntensityd(keys="image", nonzero=True, channel_wise=True),
    RandScaleIntensityd(keys="image", factors=0.1, prob=0.5),
    RandShiftIntensityd(keys="image", offsets=0.1, prob=0.5),
    ToTensord(keys=["image", "label"]),
])
val_transform = Compose([
    LoadImaged(keys=["image", "label"]),
    AsChannelFirstd(keys="image"),
    ConvertToMultiChannelBasedOnBratsClassesd(keys="label"),
    Spacingd(
        keys=["image", "label"],
        pixdim=(1.5, 1.5, 2.0),
        mode=("bilinear", "nearest"),
    ),
    Orientationd(keys=["image", "label"], axcodes="RAS"),
    CenterSpatialCropd(keys=["image", "label"], roi_size=[128, 128, 64]),
    NormalizeIntensityd(keys="image", nonzero=True, channel_wise=True),