Add console logger

anomalib/config/config.py

@@ -21,7 +21,6 @@
 from typing import List, Optional, Union
 from warnings import warn
 
-import torch
 from omegaconf import DictConfig, ListConfig, OmegaConf
 
 
@@ -112,37 +111,10 @@ def update_multi_gpu_training_config(config: Union[DictConfig, ListConfig]) -> U
     return config
 
 
-def update_device_config(config: Union[DictConfig, ListConfig], openvino: bool) -> Union[DictConfig, ListConfig]:
-    """Update XPU Device Config This function ensures devices are configured correctly by the user.
-
-    Args:
-        config (Union[DictConfig, ListConfig]): Input config
-        openvino (bool): Boolean to check if OpenVINO Inference is enabled.
-
-    Returns:
-        Union[DictConfig, ListConfig]: Updated config
-    """
-
-    config.openvino = openvino
-    if openvino:
-        config.trainer.gpus = 0
-
-    if not torch.cuda.is_available():
-        config.trainer.gpus = 0
-
-    if config.trainer.gpus == 0 and torch.cuda.is_available():
-        config.trainer.gpus = 1
-
-    config = update_multi_gpu_training_config(config)
-
-    return config
-
-
 def get_configurable_parameters(
     model_name: Optional[str] = None,
     model_config_path: Optional[Union[Path, str]] = None,
     weight_file: Optional[str] = None,
-    openvino: bool = False,
     config_filename: Optional[str] = "config",
     config_file_extension: Optional[str] = "yaml",
 ) -> Union[DictConfig, ListConfig]:
@@ -152,7 +124,6 @@ def get_configurable_parameters(
         model_name: Optional[str]:  (Default value = None)
         model_config_path: Optional[Union[Path, str]]:  (Default value = None)
         weight_file: Path to the weight file
-        openvino: Use OpenVINO
         config_filename: Optional[str]:  (Default value = "config")
         config_file_extension: Optional[str]:  (Default value = "yaml")
 
@@ -191,7 +162,6 @@ def get_configurable_parameters(
         config.model.weight_file = weight_file
 
     config = update_nncf_config(config)
-    config = update_device_config(config, openvino)
 
     # thresholding
     if "pixel_default" not in config.model.threshold.keys():

anomalib/data/btech.py

@@ -20,7 +20,6 @@
 # See the License for the specific language governing permissions
 # and limitations under the License.
 
-import logging
 import shutil
 import zipfile
 from pathlib import Path
@@ -47,9 +46,9 @@
     split_normal_images_in_train_set,
 )
 from anomalib.pre_processing import PreProcessor
+from anomalib.utils.loggers import get_console_logger
 
-logger = logging.getLogger(name="Dataset: BTech")
-logger.setLevel(logging.DEBUG)
+logger = get_console_logger(__name__)
 
 
 def make_btech_dataset(
@@ -349,23 +348,23 @@ def __init__(
     def prepare_data(self) -> None:
         """Download the dataset if not available."""
         if (self.root / self.category).is_dir():
-            logging.info("Found the dataset.")
+            logger.info("Found the dataset.")
         else:
             zip_filename = self.root.parent / "btad.zip"
 
-            logging.info("Downloading the BTech dataset.")
+            logger.info("Downloading the BTech dataset.")
             with DownloadProgressBar(unit="B", unit_scale=True, miniters=1, desc="BTech") as progress_bar:
                 urlretrieve(
                     url="https://avires.dimi.uniud.it/papers/btad/btad.zip",
                     filename=zip_filename,
                     reporthook=progress_bar.update_to,
                 )  # nosec
 
-            logging.info("Extracting the dataset.")
+            logger.info("Extracting the dataset.")
             with zipfile.ZipFile(zip_filename, "r") as zip_file:
                 zip_file.extractall(self.root.parent)
 
-            logging.info("Renaming the dataset directory")
+            logger.info("Renaming the dataset directory")
             shutil.move(src=str(self.root.parent / "BTech_Dataset_transformed"), dst=str(self.root))
 
             # NOTE: Each BTech category has different image extension as follows
@@ -377,13 +376,13 @@ def prepare_data(self) -> None:
             # To avoid any conflict, the following script converts all the extensions to png.
             # This solution works fine, but it's also possible to properly ready the bmp and
             # png filenames from categories in `make_btech_dataset` function.
-            logging.info("Convert the bmp formats to png to have consistent image extensions")
+            logger.info("Convert the bmp formats to png to have consistent image extensions")
             for filename in tqdm(self.root.glob("**/*.bmp"), desc="Converting bmp to png"):
                 image = cv2.imread(str(filename))
                 cv2.imwrite(str(filename.with_suffix(".png")), image)
                 filename.unlink()
 
-            logging.info("Cleaning the tar file")
+            logger.info("Cleaning the tar file")
             zip_filename.unlink()
 
     def setup(self, stage: Optional[str] = None) -> None:
@@ -396,6 +395,7 @@ def setup(self, stage: Optional[str] = None) -> None:
           stage: Optional[str]:  Train/Val/Test stages. (Default value = None)
 
         """
+        logger.info("Setting up train, validation, test and prediction datasets.")
         if stage in (None, "fit"):
             self.train_data = BTech(
                 root=self.root,

anomalib/data/folder.py

@@ -17,7 +17,6 @@
 # See the License for the specific language governing permissions
 # and limitations under the License.
 
-import logging
 from pathlib import Path
 from typing import Dict, Optional, Tuple, Union
 
@@ -38,9 +37,9 @@
     split_normal_images_in_train_set,
 )
 from anomalib.pre_processing import PreProcessor
+from anomalib.utils.loggers import get_console_logger
 
-logger = logging.getLogger(name="Dataset: Folder Dataset")
-logger.setLevel(logging.DEBUG)
+logger = get_console_logger(__name__)
 
 
 def _check_and_convert_path(path: Union[str, Path]) -> Path:
@@ -459,6 +458,7 @@ def setup(self, stage: Optional[str] = None) -> None:
           stage: Optional[str]:  Train/Val/Test stages. (Default value = None)
 
         """
+        logger.info("Setting up train, validation, test and prediction datasets.")
         if stage in (None, "fit"):
             self.train_data = FolderDataset(
                 normal_dir=self.normal_dir,

anomalib/data/mvtec.py

@@ -38,7 +38,6 @@
 # See the License for the specific language governing permissions
 # and limitations under the License.
 
-import logging
 import tarfile
 from pathlib import Path
 from typing import Dict, Optional, Tuple, Union
@@ -63,9 +62,9 @@
     split_normal_images_in_train_set,
 )
 from anomalib.pre_processing import PreProcessor
+from anomalib.utils.loggers import get_console_logger
 
-logger = logging.getLogger(name="Dataset: MVTec AD")
-logger.setLevel(logging.DEBUG)
+logger = get_console_logger(__name__)
 
 
 def make_mvtec_dataset(
@@ -372,24 +371,24 @@ def __init__(
     def prepare_data(self) -> None:
         """Download the dataset if not available."""
         if (self.root / self.category).is_dir():
-            logging.info("Found the dataset.")
+            logger.info("Found the dataset.")
         else:
             self.root.mkdir(parents=True, exist_ok=True)
             dataset_name = "mvtec_anomaly_detection.tar.xz"
 
-            logging.info("Downloading the dataset.")
+            logger.info("Downloading the Mvtec AD dataset.")
             with DownloadProgressBar(unit="B", unit_scale=True, miniters=1, desc="MVTec AD") as progress_bar:
                 urlretrieve(
                     url=f"ftp://guest:GU.205dldo@ftp.softronics.ch/mvtec_anomaly_detection/{dataset_name}",
                     filename=self.root / dataset_name,
                     reporthook=progress_bar.update_to,
                 )
 
-            logging.info("Extracting the dataset.")
+            logger.info("Extracting the dataset.")
             with tarfile.open(self.root / dataset_name) as tar_file:
                 tar_file.extractall(self.root)
 
-            logging.info("Cleaning the tar file")
+            logger.info("Cleaning the tar file")
             (self.root / dataset_name).unlink()
 
     def setup(self, stage: Optional[str] = None) -> None:
@@ -399,6 +398,7 @@ def setup(self, stage: Optional[str] = None) -> None:
           stage: Optional[str]:  Train/Val/Test stages. (Default value = None)
 
         """
+        logger.info("Setting up train, validation, test and prediction datasets.")
         if stage in (None, "fit"):
             self.train_data = MVTec(
                 root=self.root,

anomalib/models/__init__.py

@@ -54,7 +54,7 @@ def get_model(config: Union[DictConfig, ListConfig]) -> AnomalyModule:
     torch_model_list: List[str] = ["padim", "stfpm", "dfkde", "dfm", "patchcore", "cflow", "ganomaly"]
     model: AnomalyModule
 
-    if config.openvino:
+    if "openvino" in config.keys() and config.openvino:
         if config.model.name in openvino_model_list:
             module = import_module(f"anomalib.models.{config.model.name}.model")
             model = getattr(module, f"{config.model.name.capitalize()}OpenVINO")

anomalib/models/cflow/config.yaml

@@ -54,7 +54,7 @@ project:
 
 # PL Trainer Args. Don't add extra parameter here.
 trainer:
-  accelerator: null
+  accelerator: auto # <"cpu", "gpu", "tpu", "ipu", "hpu", "auto">
   accumulate_grad_batches: 1
   amp_backend: native
   auto_lr_find: false
@@ -63,10 +63,12 @@ trainer:
   benchmark: false
   check_val_every_n_epoch: 1
   default_root_dir: null
+  detect_anomaly: false
   deterministic: false
   enable_checkpointing: true
+  enable_progress_bar: true
   fast_dev_run: false
-  gpus: 1
+  gpus: null # Set automatically
   gradient_clip_val: 0
   limit_predict_batches: 1.0
   limit_test_batches: 1.0
@@ -86,16 +88,9 @@ trainer:
   overfit_batches: 0.0
   plugins: null
   precision: 32
-  prepare_data_per_node: true
-  process_position: 0
   profiler: null
-  progress_bar_refresh_rate: null
   replace_sampler_ddp: true
-  stochastic_weight_avg: false
   sync_batchnorm: false
-  terminate_on_nan: false
   tpu_cores: null
   track_grad_norm: -1
   val_check_interval: 1.0
-  weights_save_path: null
-  weights_summary: top

anomalib/models/cflow/lightning_model.py

@@ -26,6 +26,9 @@
 from anomalib.models.cflow.torch_model import CflowModel
 from anomalib.models.cflow.utils import get_logp, positional_encoding_2d
 from anomalib.models.components import AnomalyModule
+from anomalib.utils.loggers import get_console_logger
+
+logger = get_console_logger(__name__)
 
 __all__ = ["CflowLightning"]
 
@@ -35,6 +38,7 @@ class CflowLightning(AnomalyModule):
 
     def __init__(self, hparams):
         super().__init__(hparams)
+        logger.info("Initializing Cflow Lightning model.")
 
         self.model: CflowModel = CflowModel(hparams)
         self.loss_val = 0

anomalib/models/cflow/utils.py

@@ -22,6 +22,9 @@
 
 from anomalib.models.components.freia.framework import SequenceINN
 from anomalib.models.components.freia.modules import AllInOneBlock
+from anomalib.utils.loggers import get_console_logger
+
+logger = get_console_logger(__name__)
 
 
 def get_logp(dim_feature_vector: int, p_u: torch.Tensor, logdet_j: torch.Tensor) -> torch.Tensor:
@@ -108,7 +111,7 @@ def cflow_head(
         SequenceINN: decoder network block
     """
     coder = SequenceINN(n_features)
-    print("CNF coder:", n_features)
+    logger.info("CNF coder: %d", n_features)
     for _ in range(coupling_blocks):
         coder.append(
             AllInOneBlock,

anomalib/models/dfkde/config.yaml

@@ -38,7 +38,7 @@ project:
 
 # PL Trainer Args. Don't add extra parameter here.
 trainer:
-  accelerator: null
+  accelerator: auto # <"cpu", "gpu", "tpu", "ipu", "hpu", "auto">
   accumulate_grad_batches: 1
   amp_backend: native
   auto_lr_find: false
@@ -47,10 +47,12 @@ trainer:
   benchmark: false
   check_val_every_n_epoch: 1 # Don't validate before extracting features.
   default_root_dir: null
+  detect_anomaly: false
   deterministic: false
   enable_checkpointing: true
+  enable_progress_bar: true
   fast_dev_run: false
-  gpus: 1
+  gpus: null # Set automatically
   gradient_clip_val: 0
   limit_predict_batches: 1.0
   limit_test_batches: 1.0
@@ -70,16 +72,9 @@ trainer:
   overfit_batches: 0.0
   plugins: null
   precision: 32
-  prepare_data_per_node: true
-  process_position: 0
   profiler: null
-  progress_bar_refresh_rate: null
   replace_sampler_ddp: true
-  stochastic_weight_avg: false
   sync_batchnorm: false
-  terminate_on_nan: false
   tpu_cores: null
   track_grad_norm: -1
   val_check_interval: 1.0 # Don't validate before extracting features.
-  weights_save_path: null
-  weights_summary: top

anomalib/models/dfkde/lightning_model.py

@@ -21,9 +21,12 @@
 from torch import Tensor
 
 from anomalib.models.components import AnomalyModule
+from anomalib.utils.loggers import get_console_logger
 
 from .torch_model import DfkdeModel
 
+logger = get_console_logger(__name__)
+
 
 class DfkdeLightning(AnomalyModule):
     """DFKDE: Deep Feature Kernel Density Estimation.
@@ -34,6 +37,7 @@ class DfkdeLightning(AnomalyModule):
 
     def __init__(self, hparams: Union[DictConfig, ListConfig]):
         super().__init__(hparams)
+        logger.info("Initializing DFKDE Lightning model.")
         threshold_steepness = 0.05
         threshold_offset = 12
 
@@ -75,6 +79,7 @@ def on_validation_start(self) -> None:
         # NOTE: Previous anomalib versions fit Gaussian at the end of the epoch.
         #   This is not possible anymore with PyTorch Lightning v1.4.0 since validation
         #   is run within train epoch.
+        logger.info("Fitting a KDE model to the embedding collected from the training set.")
         self.model.fit(self.embeddings)
 
     def validation_step(self, batch, _):  # pylint: disable=arguments-differ

anomalib/models/dfkde/torch_model.py

@@ -22,6 +22,9 @@
 from torch import Tensor, nn
 
 from anomalib.models.components import PCA, FeatureExtractor, GaussianKDE
+from anomalib.utils.loggers import get_console_logger
+
+logger = get_console_logger(__name__)
 
 
 class DfkdeModel(nn.Module):
@@ -88,7 +91,7 @@ def fit(self, embeddings: List[Tensor]) -> bool:
         _embeddings = torch.vstack(embeddings)
 
         if _embeddings.shape[0] < self.n_components:
-            print("Not enough features to commit. Not making a model.")
+            logger.info("Not enough features to commit. Not making a model.")
             return False
 
         # if max training points is non-zero and smaller than number of staged features, select random subset