rename n_classes (#322)

Borda · wyli · web-flow · commit 3249a745fe7b · 2021-09-08T14:46:47.000+01:00
Co-authored-by: Wenqi Li &lt;wenqil@nvidia.com&gt;
diff --git a/2d_classification/mednist_tutorial.ipynb b/2d_classification/mednist_tutorial.ipynb
@@ -375,7 +375,7 @@
     "    [LoadImage(image_only=True), AddChannel(), ScaleIntensity(), EnsureType()])\n",
     "\n",
     "y_pred_trans = Compose([EnsureType(), Activations(softmax=True)])\n",
-    "y_trans = Compose([EnsureType(), AsDiscrete(to_onehot=True, n_classes=num_class)])"
+    "y_trans = Compose([EnsureType(), AsDiscrete(to_onehot=True, num_classes=num_class)])"
    ]
   },
   {
diff --git a/3d_classification/ignite/densenet_training_dict.py b/3d_classification/ignite/densenet_training_dict.py
@@ -126,7 +126,7 @@ def prepare_batch(batch, device=None, non_blocking=False):
     # add evaluation metric to the evaluator engine
     val_metrics = {metric_name: ROCAUC()}
 
-    post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, n_classes=2)])
+    post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, num_classes=2)])
     post_pred = Compose([EnsureType(), Activations(softmax=True)])
     # Ignite evaluator expects batch=(img, label) and returns output=(y_pred, y) at every iteration,
     # user can add output_transform to return other values
diff --git a/3d_classification/torch/densenet_training_dict.py b/3d_classification/torch/densenet_training_dict.py
@@ -81,7 +81,7 @@ def main():
         ]
     )
     post_pred = Compose([EnsureType(), Activations(softmax=True)])
-    post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, n_classes=2)])
+    post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, num_classes=2)])
 
     # Define dataset, data loader
     check_ds = monai.data.Dataset(data=train_files, transform=train_transforms)
diff --git a/3d_segmentation/challenge_baseline/run_net.py b/3d_segmentation/challenge_baseline/run_net.py
@@ -81,11 +81,11 @@ def get_xforms(mode="train", keys=("image", "label")):
 def get_net():
     """returns a unet model instance."""
 
-    n_classes = 2
+    num_classes = 2
     net = monai.networks.nets.BasicUNet(
         dimensions=3,
         in_channels=1,
-        out_channels=n_classes,
+        out_channels=num_classes,
         features=(32, 32, 64, 128, 256, 32),
         dropout=0.1,
     )
@@ -172,7 +172,7 @@ def train(data_folder=".", model_folder="runs"):
 
     # create evaluator (to be used to measure model quality during training
     val_post_transform = monai.transforms.Compose(
-        [EnsureTyped(keys=("pred", "label")), AsDiscreted(keys=("pred", "label"), argmax=(True, False), to_onehot=True, n_classes=2)]
+        [EnsureTyped(keys=("pred", "label")), AsDiscreted(keys=("pred", "label"), argmax=(True, False), to_onehot=True, num_classes=2)]
     )
     val_handlers = [
         ProgressBar(),
diff --git a/3d_segmentation/spleen_segmentation_3d.ipynb b/3d_segmentation/spleen_segmentation_3d.ipynb
@@ -471,8 +471,8 @@
     "best_metric_epoch = -1\n",
     "epoch_loss_values = []\n",
     "metric_values = []\n",
-    "post_pred = Compose([EnsureType(), AsDiscrete(argmax=True, to_onehot=True, n_classes=2)])\n",
-    "post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, n_classes=2)])\n",
+    "post_pred = Compose([EnsureType(), AsDiscrete(argmax=True, to_onehot=True, num_classes=2)])\n",
+    "post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, num_classes=2)])\n",
     "\n",
     "for epoch in range(max_epochs):\n",
     "    print(\"-\" * 10)\n",
@@ -720,8 +720,8 @@
     "        nearest_interp=False,\n",
     "        to_tensor=True,\n",
     "    ),\n",
-    "    AsDiscreted(keys=\"pred\", argmax=True, to_onehot=True, n_classes=2),\n",
-    "    AsDiscreted(keys=\"label\", to_onehot=True, n_classes=2),\n",
+    "    AsDiscreted(keys=\"pred\", argmax=True, to_onehot=True, num_classes=2),\n",
+    "    AsDiscreted(keys=\"label\", to_onehot=True, num_classes=2),\n",
     "])"
    ]
   },
diff --git a/3d_segmentation/spleen_segmentation_3d_lightning.ipynb b/3d_segmentation/spleen_segmentation_3d_lightning.ipynb
@@ -241,8 +241,8 @@
     "            norm=Norm.BATCH,\n",
     "        )\n",
     "        self.loss_function = DiceLoss(to_onehot_y=True, softmax=True)\n",
-    "        self.post_pred = Compose([EnsureType(), AsDiscrete(argmax=True, to_onehot=True, n_classes=2)])\n",
-    "        self.post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, n_classes=2)])\n",
+    "        self.post_pred = Compose([EnsureType(), AsDiscrete(argmax=True, to_onehot=True, num_classes=2)])\n",
+    "        self.post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, num_classes=2)])\n",
     "        self.dice_metric = DiceMetric(include_background=False, reduction=\"mean\", get_not_nans=False)\n",
     "        self.best_val_dice = 0\n",
     "        self.best_val_epoch = 0\n",
diff --git a/3d_segmentation/unetr_btcv_segmentation_3d.ipynb b/3d_segmentation/unetr_btcv_segmentation_3d.ipynb
@@ -681,8 +681,8 @@
     "\n",
     "max_iterations = 25000\n",
     "eval_num = 500\n",
-    "post_label = AsDiscrete(to_onehot=True, n_classes=14)\n",
-    "post_pred = AsDiscrete(argmax=True, to_onehot=True, n_classes=14)\n",
+    "post_label = AsDiscrete(to_onehot=True, num_classes=14)\n",
+    "post_pred = AsDiscrete(argmax=True, to_onehot=True, num_classes=14)\n",
     "dice_metric = DiceMetric(include_background=True, reduction=\"mean\", get_not_nans=False)\n",
     "global_step = 0\n",
     "dice_val_best = 0.0\n",
diff --git a/3d_segmentation/unetr_btcv_segmentation_3d_lightning.ipynb b/3d_segmentation/unetr_btcv_segmentation_3d_lightning.ipynb
@@ -415,8 +415,8 @@
     "        ).to(device)\n",
     "\n",
     "        self.loss_function = DiceCELoss(to_onehot_y=True, softmax=True)\n",
-    "        self.post_pred = AsDiscrete(argmax=True, to_onehot=True, n_classes=14)\n",
-    "        self.post_label = AsDiscrete(to_onehot=True, n_classes=14)\n",
+    "        self.post_pred = AsDiscrete(argmax=True, to_onehot=True, num_classes=14)\n",
+    "        self.post_label = AsDiscrete(to_onehot=True, num_classes=14)\n",
     "        self.dice_metric = DiceMetric(\n",
     "            include_background=False, reduction=\"mean\", get_not_nans=False\n",
     "        )\n",
diff --git a/acceleration/automatic_mixed_precision.ipynb b/acceleration/automatic_mixed_precision.ipynb
@@ -352,8 +352,8 @@
     "    optimizer = torch.optim.Adam(model.parameters(), 1e-4)\n",
     "    scaler = torch.cuda.amp.GradScaler() if amp else None\n",
     "\n",
-    "    post_pred = Compose([EnsureType(), AsDiscrete(argmax=True, to_onehot=True, n_classes=2)])\n",
-    "    post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, n_classes=2)])\n",
+    "    post_pred = Compose([EnsureType(), AsDiscrete(argmax=True, to_onehot=True, num_classes=2)])\n",
+    "    post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, num_classes=2)])\n",
     "\n",
     "    dice_metric = DiceMetric(include_background=False, reduction=\"mean\", get_not_nans=False)\n",
     "\n",
diff --git a/acceleration/dataset_type_performance.ipynb b/acceleration/dataset_type_performance.ipynb
@@ -209,8 +209,8 @@
     "    loss_function = DiceLoss(to_onehot_y=True, softmax=True)\n",
     "    optimizer = torch.optim.Adam(model.parameters(), 1e-4)\n",
     "\n",
-    "    post_pred = Compose([EnsureType(), AsDiscrete(argmax=True, to_onehot=True, n_classes=2)])\n",
-    "    post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, n_classes=2)])\n",
+    "    post_pred = Compose([EnsureType(), AsDiscrete(argmax=True, to_onehot=True, num_classes=2)])\n",
+    "    post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, num_classes=2)])\n",
     "\n",
     "    dice_metric = DiceMetric(include_background=True, reduction=\"mean\", get_not_nans=False)\n",
     "\n",
diff --git a/acceleration/fast_training_tutorial.ipynb b/acceleration/fast_training_tutorial.ipynb
@@ -394,8 +394,8 @@
     "        norm=Norm.BATCH,\n",
     "    ).to(device)\n",
     "\n",
-    "    post_pred = Compose([EnsureType(), AsDiscrete(argmax=True, to_onehot=True, n_classes=2)])\n",
-    "    post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, n_classes=2)])\n",
+    "    post_pred = Compose([EnsureType(), AsDiscrete(argmax=True, to_onehot=True, num_classes=2)])\n",
+    "    post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, num_classes=2)])\n",
     "\n",
     "    dice_metric = DiceMetric(include_background=True, reduction=\"mean\", get_not_nans=False)\n",
     "\n",
diff --git a/federated_learning/nvflare/nvflare_example/spleen_example/custom/train_configer.py b/federated_learning/nvflare/nvflare_example/spleen_example/custom/train_configer.py
@@ -226,7 +226,7 @@ def configure(self):
                     keys=["pred", "label"],
                     argmax=[True, False],
                     to_onehot=True,
-                    n_classes=2,
+                    num_classes=2,
                 ),
             ]
         )
diff --git a/federated_learning/nvflare/nvflare_example_docker/spleen_example/custom/train_configer.py b/federated_learning/nvflare/nvflare_example_docker/spleen_example/custom/train_configer.py
@@ -226,7 +226,7 @@ def configure(self):
                     keys=["pred", "label"],
                     argmax=[True, False],
                     to_onehot=True,
-                    n_classes=2,
+                    num_classes=2,
                 ),
             ]
         )
diff --git a/modules/dynunet_pipeline/evaluator.py b/modules/dynunet_pipeline/evaluator.py
@@ -26,7 +26,7 @@ class DynUNetEvaluator(SupervisedEvaluator):
         val_data_loader: Ignite engine use data_loader to run, must be
             torch.DataLoader.
         network: use the network to run model forward.
-        n_classes: the number of classes (output channels) for the task.
+        num_classes: the number of classes (output channels) for the task.
         epoch_length: number of iterations for one epoch, default to
             `len(val_data_loader)`.
         non_blocking: if True and this copy is between CPU and GPU, the copy may occur asynchronously
@@ -54,7 +54,7 @@ def __init__(
         device: torch.device,
         val_data_loader: DataLoader,
         network: torch.nn.Module,
-        n_classes: Union[str, int],
+        num_classes: Union[str, int],
         epoch_length: Optional[int] = None,
         non_blocking: bool = False,
         prepare_batch: Callable = default_prepare_batch,
@@ -83,11 +83,11 @@ def __init__(
             amp=amp,
         )
 
-        if not isinstance(n_classes, int):
-            n_classes = int(n_classes)
-        self.n_classes = n_classes
-        self.post_pred = AsDiscrete(argmax=True, to_onehot=True, n_classes=n_classes)
-        self.post_label = AsDiscrete(to_onehot=True, n_classes=n_classes)
+        if not isinstance(num_classes, int):
+            num_classes = int(num_classes)
+        self.num_classes = num_classes
+        self.post_pred = AsDiscrete(argmax=True, to_onehot=True, num_classes=num_classes)
+        self.post_label = AsDiscrete(to_onehot=True, num_classes=num_classes)
         self.tta_val = tta_val
 
     def _iteration(
@@ -159,13 +159,13 @@ def _compute_pred():
         if resample_flag:
             # convert the prediction back to the original (after cropped) shape
             predictions = recovery_prediction(
-                predictions.numpy(), [self.n_classes, *crop_shape], anisotrophy_flag
+                predictions.numpy(), [self.num_classes, *crop_shape], anisotrophy_flag
             )
             predictions = torch.tensor(predictions)
 
         # put iteration outputs into engine.state
         engine.state.output = {Keys.IMAGE: inputs, Keys.LABEL: targets.unsqueeze(0)}
-        engine.state.output[Keys.PRED] = torch.zeros([1, self.n_classes, *original_shape])
+        engine.state.output[Keys.PRED] = torch.zeros([1, self.num_classes, *original_shape])
         # pad the prediction back to the original shape
         box_start, box_end = batchdata["bbox"][0]
         h_start, w_start, d_start = box_start
diff --git a/modules/dynunet_pipeline/inference.py b/modules/dynunet_pipeline/inference.py
@@ -55,7 +55,7 @@ def inference(args):
         val_data_loader=test_loader,
         network=net,
         output_dir=infer_output_dir,
-        n_classes=len(properties["labels"]),
+        num_classes=len(properties["labels"]),
         inferer=SlidingWindowInferer(
             roi_size=patch_size[task_id],
             sw_batch_size=sw_batch_size,
diff --git a/modules/dynunet_pipeline/inferrer.py b/modules/dynunet_pipeline/inferrer.py
@@ -29,7 +29,7 @@ class DynUNetInferrer(SupervisedEvaluator):
             torch.DataLoader.
         network: use the network to run model forward.
         output_dir: the path to save inferred outputs.
-        n_classes: the number of classes (output channels) for the task.
+        num_classes: the number of classes (output channels) for the task.
         epoch_length: number of iterations for one epoch, default to
             `len(val_data_loader)`.
         non_blocking: if True and this copy is between CPU and GPU, the copy may occur asynchronously
@@ -58,7 +58,7 @@ def __init__(
         val_data_loader: DataLoader,
         network: torch.nn.Module,
         output_dir: str,
-        n_classes: Union[str, int],
+        num_classes: Union[str, int],
         epoch_length: Optional[int] = None,
         non_blocking: bool = False,
         prepare_batch: Callable = default_prepare_batch,
@@ -87,12 +87,12 @@ def __init__(
             amp=amp,
         )
 
-        if not isinstance(n_classes, int):
-            n_classes = int(n_classes)
-        self.post_pred = AsDiscrete(argmax=True, to_onehot=True, n_classes=n_classes)
+        if not isinstance(num_classes, int):
+            num_classes = int(num_classes)
+        self.post_pred = AsDiscrete(argmax=True, to_onehot=True, num_classes=num_classes)
         self.output_dir = output_dir
         self.tta_val = tta_val
-        self.n_classes = n_classes
+        self.num_classes = num_classes
 
     def _iteration(
         self, engine: Engine, batchdata: Dict[str, Any]
@@ -159,7 +159,7 @@ def _compute_pred():
         if resample_flag:
             # convert the prediction back to the original (after cropped) shape
             predictions = recovery_prediction(
-                predictions.numpy(), [self.n_classes, *crop_shape], anisotrophy_flag
+                predictions.numpy(), [self.num_classes, *crop_shape], anisotrophy_flag
             )
         else:
             predictions = predictions.numpy()
diff --git a/modules/dynunet_pipeline/train.py b/modules/dynunet_pipeline/train.py
@@ -56,15 +56,15 @@ def validation(args):
             module=net, device_ids=[device], find_unused_parameters=True
         )
 
-    n_classes = len(properties["labels"])
+    num_classes = len(properties["labels"])
 
     net.eval()
 
     evaluator = DynUNetEvaluator(
         device=device,
         val_data_loader=val_loader,
         network=net,
-        n_classes=n_classes,
+        num_classes=num_classes,
         inferer=SlidingWindowInferer(
             roi_size=patch_size[task_id],
             sw_batch_size=sw_batch_size,
@@ -87,8 +87,8 @@ def validation(args):
     if local_rank == 0:
         print(evaluator.state.metrics)
         results = evaluator.state.metric_details["val_mean_dice"]
-        if n_classes > 2:
-            for i in range(n_classes - 1):
+        if num_classes > 2:
+            for i in range(num_classes - 1):
                 print(
                     "mean dice for label {} is {}".format(i + 1, results[:, i].mean())
                 )
@@ -166,7 +166,7 @@ def train(args):
         device=device,
         val_data_loader=val_loader,
         network=net,
-        n_classes=len(properties["labels"]),
+        num_classes=len(properties["labels"]),
         inferer=SlidingWindowInferer(
             roi_size=patch_size[task_id],
             sw_batch_size=sw_batch_size,
diff --git a/modules/interpretability/covid_classification.ipynb b/modules/interpretability/covid_classification.ipynb
@@ -243,7 +243,7 @@
     "])\n",
     "\n",
     "y_pred_trans = Compose([EnsureType(), Activations(softmax=True)])\n",
-    "y_trans = Compose([EnsureType(), AsDiscrete(to_onehot=True, n_classes=num_class)])\n",
+    "y_trans = Compose([EnsureType(), AsDiscrete(to_onehot=True, num_classes=num_class)])\n",
     "\n",
     "all_files = glob(os.path.join(train_dir, \"*.png\"))\n",
     "random.shuffle(all_files)\n",
diff --git a/modules/learning_rate.ipynb b/modules/learning_rate.ipynb
@@ -253,7 +253,7 @@
     "num_classes = train_ds.get_num_classes()\n",
     "\n",
     "y_pred_trans = Compose([EnsureType(), Activations(softmax=True)])\n",
-    "y_trans = Compose([EnsureType(), AsDiscrete(to_onehot=True, n_classes=num_classes)])"
+    "y_trans = Compose([EnsureType(), AsDiscrete(to_onehot=True, num_classes=num_classes)])"
    ]
   },
   {
diff --git a/modules/postprocessing_transforms.ipynb b/modules/postprocessing_transforms.ipynb
@@ -392,8 +392,8 @@
     "epoch_loss_values = []\n",
     "metric_values = []\n",
     "\n",
-    "post_pred = Compose([EnsureType(), AsDiscrete(argmax=True, to_onehot=True, n_classes=2)])\n",
-    "post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, n_classes=2)])\n",
+    "post_pred = Compose([EnsureType(), AsDiscrete(argmax=True, to_onehot=True, num_classes=2)])\n",
+    "post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, num_classes=2)])\n",
     "\n",
     "for epoch in range(max_epochs):\n",
     "    print(\"-\" * 10)\n",
diff --git a/modules/public_datasets.ipynb b/modules/public_datasets.ipynb
@@ -1485,7 +1485,7 @@
     "    inferer=SimpleInferer(),\n",
     "    postprocessing=AsDiscreted(\n",
     "        keys=[\"pred\", \"label\"], argmax=(True, False),\n",
-    "        to_onehot=True, n_classes=3,\n",
+    "        to_onehot=True, num_classes=3,\n",
     "    ),\n",
     "    key_train_metric={\n",
     "        \"train_meandice\": MeanDice(\n",
diff --git a/modules/transfer_mmar.ipynb b/modules/transfer_mmar.ipynb
@@ -642,8 +642,8 @@
     "best_metric_epoch = -1\n",
     "epoch_loss_values = []\n",
     "metric_values = []\n",
-    "post_pred = Compose([EnsureType(), AsDiscrete(argmax=True, to_onehot=True, n_classes=2)])\n",
-    "post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, n_classes=2)])\n",
+    "post_pred = Compose([EnsureType(), AsDiscrete(argmax=True, to_onehot=True, num_classes=2)])\n",
+    "post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, num_classes=2)])\n",
     "dice_metric = DiceMetric(include_background=False, reduction=\"mean\", get_not_nans=False)\n",
     "\n",
     "for epoch in range(max_epochs):\n",
diff --git a/pathology/tumor_detection/ignite/camelyon_train_evaluate.py b/pathology/tumor_detection/ignite/camelyon_train_evaluate.py
@@ -164,7 +164,7 @@ def train(cfg):
     # --------------------------------------------------------------------------
     # __________________________________________________________________________
     # initialize model
-    model = TorchVisionFCModel("resnet18", n_classes=1, use_conv=True, pretrained=cfg["pretrain"])
+    model = TorchVisionFCModel("resnet18", num_classes=1, use_conv=True, pretrained=cfg["pretrain"])
     model = model.to(device)
 
     # loss function
diff --git a/pathology/tumor_detection/ignite/camelyon_train_evaluate_nvtx_profiling.py b/pathology/tumor_detection/ignite/camelyon_train_evaluate_nvtx_profiling.py
@@ -175,7 +175,7 @@ def train(cfg):
     # --------------------------------------------------------------------------
     # __________________________________________________________________________
     # initialize model
-    model = TorchVisionFCModel("resnet18", n_classes=1, use_conv=True, pretrained=cfg["pretrain"])
+    model = TorchVisionFCModel("resnet18", num_classes=1, use_conv=True, pretrained=cfg["pretrain"])
     model = Range("ResNet18")(model)
     model = model.to(device)
 

Original file line number	Diff line number	Diff line change
`@@ -375,7 +375,7 @@`
`375`	`375`	`" [LoadImage(image_only=True), AddChannel(), ScaleIntensity(), EnsureType()])\n",`
`376`	`376`	`"\n",`
`377`	`377`	`"y_pred_trans = Compose([EnsureType(), Activations(softmax=True)])\n",`
`378`		`- "y_trans = Compose([EnsureType(), AsDiscrete(to_onehot=True, n_classes=num_class)])"`
	`378`	`+ "y_trans = Compose([EnsureType(), AsDiscrete(to_onehot=True, num_classes=num_class)])"`
`379`	`379`	`]`
`380`	`380`	`},`
`381`	`381`	`{`
Original file line number	Diff line number	Diff line change
`@@ -81,7 +81,7 @@ def main():`
`81`	`81`	`]`
`82`	`82`	`)`
`83`	`83`	`post_pred = Compose([EnsureType(), Activations(softmax=True)])`
`84`		`- post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, n_classes=2)])`
	`84`	`+ post_label = Compose([EnsureType(), AsDiscrete(to_onehot=True, num_classes=2)])`
`85`	`85`
`86`	`86`	`# Define dataset, data loader`
`87`	`87`	`check_ds = monai.data.Dataset(data=train_files, transform=train_transforms)`