Merge pull request #832 from VukW/black_include_all

Added black trailing comma style fix

.github/workflows/black.yml

@@ -10,7 +10,14 @@ jobs:
         python-version: [3.9]
     steps:
       - uses: actions/checkout@v2
+
+      - name: Extract black version from setup.py
+        run: |
+          echo "BLACK_VERSION=$(python -c 'from setup import black_version; print(black_version)')" >> $GITHUB_ENV
+
       - uses: psf/black@stable
+        with:
+          version: ${{ env.BLACK_VERSION }}
 
       - name: Set up Python ${{ matrix.python-version }}
         uses: actions/setup-python@v4
@@ -20,7 +27,7 @@ jobs:
       - name: Install dependencies
         run: |
           python -m pip install --upgrade pip
-          python -m pip install black==23.11.0
+          python -m pip install black==${{ env.BLACK_VERSION }}
       
       - name: Run tests
         run: |

GANDLF/cli/deploy.py

@@ -113,9 +113,7 @@ def deploy_docker_mlcube(
     os.makedirs(output_workspace_folder, exist_ok=True)
     if os.path.exists(mlcube_workspace_folder):
         shutil.copytree(
-            mlcube_workspace_folder,
-            output_workspace_folder,
-            dirs_exist_ok=True,
+            mlcube_workspace_folder, output_workspace_folder, dirs_exist_ok=True
         )
 
     if config is not None:

GANDLF/cli/generate_metrics.py

@@ -112,13 +112,10 @@ def generate_metrics_dict(
                 parameters["model"]["class_list"] = [1]
                 parameters["model"]["num_classes"] = 1
                 overall_stats_dict[current_subject_id][str(class_index)]["dice"] = dice(
-                    current_prediction,
-                    current_target,
+                    current_prediction, current_target
                 ).item()
                 nsd, hd100, hd95 = _calculator_generic_all_surface_distances(
-                    current_prediction,
-                    current_target,
-                    parameters,
+                    current_prediction, current_target, parameters
                 )
                 overall_stats_dict[current_subject_id][str(class_index)][
                     "nsd"
@@ -130,13 +127,8 @@ def generate_metrics_dict(
                     "hd95"
                 ] = hd95.item()
 
-                (
-                    s,
-                    p,
-                ) = _calculator_sensitivity_specificity(
-                    current_prediction,
-                    current_target,
-                    parameters,
+                (s, p) = _calculator_sensitivity_specificity(
+                    current_prediction, current_target, parameters
                 )
                 overall_stats_dict[current_subject_id][str(class_index)][
                     "sensitivity"
@@ -147,9 +139,7 @@ def generate_metrics_dict(
                 overall_stats_dict[current_subject_id][
                     "jaccard_" + str(class_index)
                 ] = _calculator_jaccard(
-                    current_prediction,
-                    current_target,
-                    parameters,
+                    current_prediction, current_target, parameters
                 ).item()
                 current_target_image = sitk.GetImageFromArray(
                     current_target[0, 0, ...].long()
@@ -279,9 +269,9 @@ def __percentile_clip(
                     strictlyPositive=True,
                 )
 
-            overall_stats_dict[current_subject_id]["ssim"] = (
-                structural_similarity_index(output_infill, gt_image_infill, mask).item()
-            )
+            overall_stats_dict[current_subject_id][
+                "ssim"
+            ] = structural_similarity_index(output_infill, gt_image_infill, mask).item()
 
             # ncc metrics
             compute_ncc = parameters.get("compute_ncc", True)
@@ -322,30 +312,30 @@ def __percentile_clip(
             ).item()
 
             # same as above but with epsilon for robustness
-            overall_stats_dict[current_subject_id]["psnr_eps"] = (
-                peak_signal_noise_ratio(
-                    output_infill, gt_image_infill, epsilon=sys.float_info.epsilon
-                ).item()
-            )
+            overall_stats_dict[current_subject_id][
+                "psnr_eps"
+            ] = peak_signal_noise_ratio(
+                output_infill, gt_image_infill, epsilon=sys.float_info.epsilon
+            ).item()
 
             # only use fix data range to [0;1] if the data was normalized before
             if normalize:
                 # torchmetrics PSNR but with fixed data range of 0 to 1
-                overall_stats_dict[current_subject_id]["psnr_01"] = (
-                    peak_signal_noise_ratio(
-                        output_infill, gt_image_infill, data_range=(0, 1)
-                    ).item()
-                )
+                overall_stats_dict[current_subject_id][
+                    "psnr_01"
+                ] = peak_signal_noise_ratio(
+                    output_infill, gt_image_infill, data_range=(0, 1)
+                ).item()
 
                 # same as above but with epsilon for robustness
-                overall_stats_dict[current_subject_id]["psnr_01_eps"] = (
-                    peak_signal_noise_ratio(
-                        output_infill,
-                        gt_image_infill,
-                        data_range=(0, 1),
-                        epsilon=sys.float_info.epsilon,
-                    ).item()
-                )
+                overall_stats_dict[current_subject_id][
+                    "psnr_01_eps"
+                ] = peak_signal_noise_ratio(
+                    output_infill,
+                    gt_image_infill,
+                    data_range=(0, 1),
+                    epsilon=sys.float_info.epsilon,
+                ).item()
 
     pprint(overall_stats_dict)
     if outputfile is not None:

GANDLF/cli/patch_extraction.py

@@ -15,9 +15,7 @@
     patch_artifact_check,
     # pen_marking_check,
 )
-from GANDLF.utils import (
-    parseTrainingCSV,
-)
+from GANDLF.utils import parseTrainingCSV
 
 
 def parse_gandlf_csv(fpath):

GANDLF/cli/preprocess_and_save.py

@@ -66,9 +66,7 @@ def preprocess_and_save(
     )
 
     dataloader_for_processing = DataLoader(
-        data_for_processing,
-        batch_size=1,
-        pin_memory=False,
+        data_for_processing, batch_size=1, pin_memory=False
     )
 
     # initialize a new dict for the preprocessed data

GANDLF/compute/forward_pass.py

@@ -345,11 +345,7 @@ def validate_network(
 
                     # if jpg detected, convert to 8-bit arrays
                     ext = get_filename_extension_sanitized(subject["1"]["path"][0])
-                    if ext in [
-                        ".jpg",
-                        ".jpeg",
-                        ".png",
-                    ]:
+                    if ext in [".jpg", ".jpeg", ".png"]:
                         pred_mask = pred_mask.astype(np.uint8)
 
                     ## special case for 2D
@@ -369,8 +365,7 @@ def validate_network(
                     # Create the subject directory if it doesn't exist in the
                     # current_output_dir directory
                     os.makedirs(
-                        os.path.join(current_output_dir, "testing"),
-                        exist_ok=True,
+                        os.path.join(current_output_dir, "testing"), exist_ok=True
                     )
                     os.makedirs(
                         os.path.join(
@@ -386,10 +381,7 @@ def validate_network(
                         subject["subject_id"][0] + "_seg" + ext,
                     )
 
-                    sitk.WriteImage(
-                        result_image,
-                        path_to_save,
-                    )
+                    sitk.WriteImage(result_image, path_to_save)
             else:
                 # final regression output
                 output_prediction = output_prediction / len(patch_loader)

GANDLF/compute/generic.py

@@ -6,10 +6,7 @@
 from GANDLF.models import get_model
 from GANDLF.schedulers import get_scheduler
 from GANDLF.optimizers import get_optimizer
-from GANDLF.data import (
-    get_train_loader,
-    get_validation_loader,
-)
+from GANDLF.data import get_train_loader, get_validation_loader
 from GANDLF.utils import (
     populate_header_in_parameters,
     parseTrainingCSV,

GANDLF/compute/inference_loop.py

@@ -62,14 +62,7 @@ def inference_loop(
         or parameters["model"]["type"].lower() == "openvino"
     ), f"The model type is not recognized: {parameters['model']['type']}"
 
-    (
-        model,
-        _,
-        _,
-        _,
-        _,
-        parameters,
-    ) = create_pytorch_objects(parameters, device=device)
+    (model, _, _, _, _, parameters) = create_pytorch_objects(parameters, device=device)
 
     # Loading the weights into the model
     main_dict = None
@@ -103,12 +96,10 @@ def inference_loop(
         # Loading the executable OpenVINO model
         if os.path.isdir(modelDir):
             xml_to_check = os.path.join(
-                modelDir,
-                str(parameters["model"]["architecture"]) + "_best.xml",
+                modelDir, str(parameters["model"]["architecture"]) + "_best.xml"
             )
             bin_to_check = os.path.join(
-                modelDir,
-                str(parameters["model"]["architecture"]) + "_best.bin",
+                modelDir, str(parameters["model"]["architecture"]) + "_best.bin"
             )
             if not os.path.isfile(xml_to_check):
                 raise ValueError(
@@ -312,28 +303,20 @@ def inference_loop(
                 )
 
             if parameters["problem_type"] != "segmentation":
-                output_file = os.path.join(
-                    subject_dest_dir,
-                    "predictions.csv",
-                )
+                output_file = os.path.join(subject_dest_dir, "predictions.csv")
                 with open(output_file, "w") as f:
                     f.write(output_to_write)
 
             heatmaps = {}
             if probs_map is not None:
                 try:
                     for n in range(parameters["model"]["num_classes"]):
-                        heatmap_gray = np.array(
-                            probs_map[n, ...] * 255,
-                            dtype=np.uint8,
-                        )
+                        heatmap_gray = np.array(probs_map[n, ...] * 255, dtype=np.uint8)
                         heatmaps[str(n) + "_jet"] = cv2.applyColorMap(
-                            heatmap_gray,
-                            cv2.COLORMAP_JET,
+                            heatmap_gray, cv2.COLORMAP_JET
                         )
                         heatmaps[str(n) + "_turbo"] = cv2.applyColorMap(
-                            heatmap_gray,
-                            cv2.COLORMAP_TURBO,
+                            heatmap_gray, cv2.COLORMAP_TURBO
                         )
                         heatmaps[str(n) + "_agni"] = applyCustomColorMap(heatmap_gray)
 
@@ -367,8 +350,7 @@ def inference_loop(
                         )
 
                         file_to_write = os.path.join(
-                            subject_dest_dir,
-                            "probability_map_blended_" + key + ".png",
+                            subject_dest_dir, "probability_map_blended_" + key + ".png"
                         )
                         cv2.imwrite(file_to_write, blended_image)
                 except Exception as ex:

GANDLF/compute/training_loop.py

@@ -100,18 +100,14 @@ def train_network(
             label = torch.cat([subject[key] for key in params["value_keys"]], dim=0)
             # min is needed because for certain cases, batch size becomes smaller than the total remaining labels
             label = label.reshape(
-                min(params["batch_size"], len(label)),
-                len(params["value_keys"]),
+                min(params["batch_size"], len(label)), len(params["value_keys"])
             )
         else:
             label = subject["label"][torchio.DATA]
         label = label.to(params["device"])
 
         if params["save_training"]:
-            write_training_patches(
-                subject,
-                params,
-            )
+            write_training_patches(subject, params)
 
         # ensure spacing is always present in params and is always subject-specific
         if "spacing" in subject:
@@ -190,10 +186,7 @@ def train_network(
                         ).tolist()
                     else:
                         to_print = total_epoch_train_metric[metric] / (batch_idx + 1)
-                    print(
-                        "Half-Epoch Average train " + metric + " : ",
-                        to_print,
-                    )
+                    print("Half-Epoch Average train " + metric + " : ", to_print)
 
     average_epoch_train_loss = total_epoch_train_loss / len(train_dataloader)
     print("     Epoch Final   train loss : ", average_epoch_train_loss)
@@ -343,9 +336,7 @@ def training_loop(
         # this is just used to generate the headers for the overall stats
         temp_tensor = torch.randint(0, params["model"]["num_classes"], (5,))
         overall_metrics = overall_stats(
-            temp_tensor.to(dtype=torch.int32),
-            temp_tensor.to(dtype=torch.int32),
-            params,
+            temp_tensor.to(dtype=torch.int32), temp_tensor.to(dtype=torch.int32), params
         )
 
     metrics_log = params["metrics"].copy()

GANDLF/config_manager.py

@@ -445,11 +445,7 @@ def _parseConfig(
         if len(params["data_preprocessing"]) > 0:
             thresholdOrClip = False
             # this can be extended, as required
-            thresholdOrClipDict = [
-                "threshold",
-                "clip",
-                "clamp",
-            ]
+            thresholdOrClipDict = ["threshold", "clip", "clamp"]
 
             resize_requested = False
             temp_dict = deepcopy(params["data_preprocessing"])
@@ -708,10 +704,7 @@ def _parseConfig(
         params["optimizer"] = temp_dict
 
     # initialize defaults for inference mechanism
-    inference_mechanism = {
-        "grid_aggregator_overlap": "crop",
-        "patch_overlap": 0,
-    }
+    inference_mechanism = {"grid_aggregator_overlap": "crop", "patch_overlap": 0}
     initialize_inference_mechanism = False
     if not ("inference_mechanism" in params):
         initialize_inference_mechanism = True

GANDLF/data/augmentation/__init__.py

@@ -12,10 +12,7 @@
     flip,
     anisotropy,
 )
-from .rotations import (
-    rotate_90,
-    rotate_180,
-)
+from .rotations import rotate_90, rotate_180
 from .rgb_augs import colorjitter_transform
 from .hed_augs import hed_transform
 

GANDLF/data/augmentation/noise_enhanced.py

@@ -55,9 +55,7 @@ def apply_transform(self, subject: Subject) -> Subject:
         return transformed
 
     def get_params(
-        self,
-        mean_range: Tuple[float, float],
-        std_range: Tuple[float, float],
+        self, mean_range: Tuple[float, float], std_range: Tuple[float, float]
     ) -> Tuple[float, float]:
         mean = self.sample_uniform(*mean_range)
         std = self.sample_uniform(*std_range)

GANDLF/data/augmentation/rgb_augs.py

@@ -77,10 +77,7 @@ def apply_transform(self, subject: Subject) -> Subject:
         else:
             hue = self.hue_range
         transform = ColorJitter(
-            brightness=brightness,
-            contrast=contrast,
-            saturation=saturation,
-            hue=hue,
+            brightness=brightness, contrast=contrast, saturation=saturation, hue=hue
         )
         for _, image in self.get_images_dict(subject).items():
             # proceed with processing only if the image is RGB

GANDLF/data/augmentation/wrap_torchio.py

@@ -24,8 +24,7 @@
 # define individual functions/lambdas for augmentations to handle properties
 def mri_artifact(parameters):
     return OneOf(
-        {RandomGhosting(): 0.5, RandomSpike(): 0.5},
-        p=parameters["probability"],
+        {RandomGhosting(): 0.5, RandomSpike(): 0.5}, p=parameters["probability"]
     )