Bump up the pytorch lightning to master branch due to vulnurability i…

…ssues (#55)

* Bump up the pytorch lightning to master branch due to vulnurability issues.

* Updated config files

* added Inference Dataset

* fix deterministic issue

* Modified config files

* pl version to 1.5.9

* modified cflow configs

* set deterministic off

* do not log images when testing.

* Bumped up version

* Update tox.yml

* Add trainer.validate to `run_train_test` in normalization tests

* nncf callback

* use on_validation_epoch_start instead of on_validation_start in cdf callback

* Updated ganomaly configs to match the version

Co-authored-by: Dick Ameln <[email protected]>

anomalib/core/callbacks/cdf_normalization.py

@@ -22,10 +22,8 @@ def on_test_start(self, _trainer: pl.Trainer, pl_module: pl.LightningModule) ->
         pl_module.image_metrics.F1.threshold = 0.5
         pl_module.pixel_metrics.F1.threshold = 0.5
 
-    def on_train_epoch_end(
-        self, trainer: pl.Trainer, pl_module: pl.LightningModule, _unused: Optional[Any] = None
-    ) -> None:
-        """Called when the train epoch ends.
+    def on_validation_epoch_start(self, trainer: "pl.Trainer", pl_module: "pl.LightningModule") -> None:
+        """Called when the validation starts after training.
 
         Use the current model to compute the anomaly score distributions
         of the normal training data. This is needed after every epoch, because the statistics must be

anomalib/core/callbacks/nncf_callback.py

@@ -13,8 +13,6 @@
 from pytorch_lightning import Callback
 from torch.utils.data.dataloader import DataLoader
 
-from anomalib.data import get_datamodule
-
 
 def criterion_fn(outputs, criterion):
     """Calls the criterion function on outputs."""
@@ -76,21 +74,18 @@ def __init__(self, config: Union[ListConfig, DictConfig], dirpath: str, filename
         self.dirpath = dirpath
         self.filename = filename
 
-        # we need to create a datamodule here to obtain the init loader
-        datamodule = get_datamodule(config)
-        datamodule.setup()
-        self.train_loader = datamodule.train_dataloader()
-
         self.comp_ctrl: Optional[CompressionAlgorithmController] = None
         self.compression_scheduler: CompressionScheduler
 
-    def setup(self, _: pl.Trainer, pl_module: pl.LightningModule, __: Optional[str] = None) -> None:
+    def setup(self, trainer: pl.Trainer, pl_module: pl.LightningModule, stage: Optional[str] = None) -> None:
         """Call when fit or test begins.
 
         Takes the pytorch model and wraps it using the compression controller so that it is ready for nncf fine-tuning.
         """
         if self.comp_ctrl is None:
-            init_loader = InitLoader(self.train_loader)
+            # NOTE: trainer.datamodule returns the following error
+            #   "Trainer" has no attribute "datamodule"  [attr-defined]
+            init_loader = InitLoader(trainer.datamodule.train_dataloader())  # type: ignore
             nncf_config = register_default_init_args(
                 self.nncf_config, init_loader, pl_module.model.loss, criterion_fn=criterion_fn
             )
@@ -99,7 +94,12 @@ def setup(self, _: pl.Trainer, pl_module: pl.LightningModule, __: Optional[str]
             self.compression_scheduler = self.comp_ctrl.scheduler
 
     def on_train_batch_start(
-        self, trainer, _pl_module: pl.LightningModule, _batch: Any, _batch_idx: int, _dataloader_idx: int
+        self,
+        trainer: pl.Trainer,
+        _pl_module: pl.LightningModule,
+        _batch: Any,
+        _batch_idx: int,
+        _unused: Optional[int] = 0,
     ) -> None:
         """Call when the train batch begins.
 
@@ -109,7 +109,7 @@ def on_train_batch_start(
         if self.comp_ctrl is not None:
             trainer.model.loss_val = self.comp_ctrl.loss()
 
-    def on_train_end(self, _trainer, _pl_module: pl.LightningModule) -> None:
+    def on_train_end(self, _trainer: pl.Trainer, _pl_module: pl.LightningModule) -> None:
         """Call when the train ends.
 
         Exports onnx model and if compression controller is not None, uses the onnx model to generate the OpenVINO IR.

anomalib/core/model/anomaly_module.py

@@ -14,7 +14,8 @@
 # See the License for the specific language governing permissions
 # and limitations under the License.
 
-from typing import List, Union
+from abc import ABC
+from typing import Any, List, Optional, Union
 
 import pytorch_lightning as pl
 from omegaconf import DictConfig, ListConfig
@@ -30,7 +31,7 @@
 )
 
 
-class AnomalyModule(pl.LightningModule):
+class AnomalyModule(pl.LightningModule, ABC):
     """AnomalyModule to train, validate, predict and test images.
 
     Acts as a base class for all the Anomaly Modules in the library.
@@ -77,7 +78,7 @@ def validation_step(self, batch, batch_idx) -> dict:  # type: ignore  # pylint:
         """To be implemented in the subclasses."""
         raise NotImplementedError
 
-    def predict_step(self, batch, batch_idx, _):  # pylint: disable=arguments-differ, signature-differs
+    def predict_step(self, batch: Any, batch_idx: int, _dataloader_idx: Optional[int] = None) -> Any:
         """Step function called during :meth:`~pytorch_lightning.trainer.trainer.Trainer.predict`.
 
         By default, it calls :meth:`~pytorch_lightning.core.lightning.LightningModule.forward`.
@@ -86,7 +87,7 @@ def predict_step(self, batch, batch_idx, _):  # pylint: disable=arguments-differ
         Args:
             batch (Tensor): Current batch
             batch_idx (int): Index of current batch
-            dataloader_idx (int): Index of the current dataloader
+            _dataloader_idx (int): Index of the current dataloader
 
         Return:
             Predicted output

anomalib/data/__init__.py

@@ -19,6 +19,7 @@
 from omegaconf import DictConfig, ListConfig
 from pytorch_lightning import LightningDataModule
 
+from .inference import InferenceDataset
 from .mvtec import MVTecDataModule
 
 
@@ -48,3 +49,6 @@ def get_datamodule(config: Union[DictConfig, ListConfig]) -> LightningDataModule
         raise ValueError("Unknown dataset!")
 
     return datamodule
+
+
+__all__ = ["get_datamodule", "InferenceDataset"]

anomalib/data/inference.py

@@ -0,0 +1,67 @@
+"""Inference Dataset."""
+
+# Copyright (C) 2020 Intel Corporation
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions
+# and limitations under the License.
+
+from pathlib import Path
+from typing import Any, Optional, Tuple, Union
+
+import albumentations as A
+from torch.utils.data.dataset import Dataset
+
+from anomalib.data.transforms import PreProcessor
+from anomalib.data.utils import get_image_filenames, read_image
+
+
+class InferenceDataset(Dataset):
+    """Inference Dataset to perform prediction."""
+
+    def __init__(
+        self,
+        path: Union[str, Path],
+        pre_process: Optional[PreProcessor] = None,
+        image_size: Optional[Union[int, Tuple[int, int]]] = None,
+        transform_config: Optional[Union[str, A.Compose]] = None,
+    ) -> None:
+        """Inference Dataset to perform prediction.
+
+        Args:
+            path (Union[str, Path]): Path to an image or image-folder.
+            pre_process (Optional[PreProcessor], optional): Pre-Processing transforms to
+                pre-process the input dataset. Defaults to None.
+            image_size (Optional[Union[int, Tuple[int, int]]], optional): Target image size
+                to resize the original image. Defaults to None.
+            transform_config (Optional[Union[str, A.Compose]], optional): Configuration file
+                parse the albumentation transforms. Defaults to None.
+        """
+        super().__init__()
+
+        self.image_filenames = get_image_filenames(path)
+
+        if pre_process is None:
+            self.pre_process = PreProcessor(transform_config, image_size)
+        else:
+            self.pre_process = pre_process
+
+    def __len__(self) -> int:
+        """Get the number of images in the given path."""
+        return len(self.image_filenames)
+
+    def __getitem__(self, index: int) -> Any:
+        """Get the image based on the `index`."""
+        image_filename = self.image_filenames[index]
+        image = read_image(path=image_filename)
+        pre_processed = self.pre_process(image=image)
+
+        return pre_processed

anomalib/data/mvtec.py

@@ -34,11 +34,13 @@
 import pandas as pd
 from pandas.core.frame import DataFrame
 from pytorch_lightning.core.datamodule import LightningDataModule
+from pytorch_lightning.utilities.types import EVAL_DATALOADERS, TRAIN_DATALOADERS
 from torch import Tensor
 from torch.utils.data import DataLoader
 from torch.utils.data.dataset import Dataset
 from torchvision.datasets.folder import VisionDataset
 
+from anomalib.data.inference import InferenceDataset
 from anomalib.data.transforms import PreProcessor
 from anomalib.data.utils import read_image
 from anomalib.utils.download_progress_bar import DownloadProgressBar
@@ -417,8 +419,10 @@ def __init__(
         self.root = root if isinstance(root, Path) else Path(root)
         self.category = category
         self.dataset_path = self.root / self.category
+        self.transform_config = transform_config
+        self.image_size = image_size
 
-        self.pre_process = PreProcessor(config=transform_config, image_size=image_size)
+        self.pre_process = PreProcessor(config=self.transform_config, image_size=self.image_size)
 
         self.train_batch_size = train_batch_size
         self.test_batch_size = test_batch_size
@@ -431,13 +435,25 @@ def __init__(
         self.test_data: Dataset
         if create_validation_set:
             self.val_data: Dataset
+        self.inference_data: Dataset
 
     def setup(self, stage: Optional[str] = None) -> None:
         """Setup train, validation and test data.
 
         Args:
           stage: Optional[str]:  Train/Val/Test stages. (Default value = None)
+
         """
+        if stage in (None, "fit"):
+            self.train_data = MVTec(
+                root=self.root,
+                category=self.category,
+                pre_process=self.pre_process,
+                split="train",
+                seed=self.seed,
+                create_validation_set=self.create_validation_set,
+            )
+
         if self.create_validation_set:
             self.val_data = MVTec(
                 root=self.root,
@@ -447,6 +463,7 @@ def setup(self, stage: Optional[str] = None) -> None:
                 seed=self.seed,
                 create_validation_set=self.create_validation_set,
             )
+
         self.test_data = MVTec(
             root=self.root,
             category=self.category,
@@ -455,25 +472,27 @@ def setup(self, stage: Optional[str] = None) -> None:
             seed=self.seed,
             create_validation_set=self.create_validation_set,
         )
-        if stage in (None, "fit"):
-            self.train_data = MVTec(
-                root=self.root,
-                category=self.category,
-                pre_process=self.pre_process,
-                split="train",
-                seed=self.seed,
-                create_validation_set=self.create_validation_set,
+
+        if stage == "predict":
+            self.inference_data = InferenceDataset(
+                path=self.root, image_size=self.image_size, transform_config=self.transform_config
             )
 
-    def train_dataloader(self) -> DataLoader:
+    def train_dataloader(self) -> TRAIN_DATALOADERS:
         """Get train dataloader."""
         return DataLoader(self.train_data, shuffle=True, batch_size=self.train_batch_size, num_workers=self.num_workers)
 
-    def val_dataloader(self) -> DataLoader:
+    def val_dataloader(self) -> EVAL_DATALOADERS:
         """Get validation dataloader."""
         dataset = self.val_data if self.create_validation_set else self.test_data
         return DataLoader(dataset=dataset, shuffle=False, batch_size=self.test_batch_size, num_workers=self.num_workers)
 
-    def test_dataloader(self) -> DataLoader:
+    def test_dataloader(self) -> EVAL_DATALOADERS:
         """Get test dataloader."""
         return DataLoader(self.test_data, shuffle=False, batch_size=self.test_batch_size, num_workers=self.num_workers)
+
+    def predict_dataloader(self) -> EVAL_DATALOADERS:
+        """Get predict dataloader."""
+        return DataLoader(
+            self.inference_data, shuffle=False, batch_size=self.test_batch_size, num_workers=self.num_workers
+        )

anomalib/data/utils.py

@@ -15,10 +15,38 @@
 # and limitations under the License.
 
 from pathlib import Path
-from typing import Union
+from typing import List, Union
 
 import cv2
 import numpy as np
+from torchvision.datasets.folder import IMG_EXTENSIONS
+
+
+def get_image_filenames(path: Union[str, Path]) -> List[str]:
+    """Get image filenames.
+
+    Args:
+        path (Union[str, Path]): Path to image or image-folder.
+
+    Returns:
+        List[str]: List of image filenames
+
+    """
+    image_filenames: List[str]
+
+    if isinstance(path, str):
+        path = Path(path)
+
+    if path.is_file() and path.suffix in IMG_EXTENSIONS:
+        image_filenames = [str(path)]
+
+    if path.is_dir():
+        image_filenames = [str(p) for p in path.glob("**/*") if p.suffix in IMG_EXTENSIONS]
+
+    if len(image_filenames) == 0:
+        raise ValueError(f"Found 0 images in {path}")
+
+    return image_filenames
 
 
 def read_image(path: Union[str, Path]) -> np.ndarray:

anomalib/models/cflow/config.yaml

@@ -3,7 +3,7 @@ dataset:
   format: mvtec
   path: ./datasets/MVTec
   url: ftp://guest:[email protected]/mvtec_anomaly_detection/mvtec_anomaly_detection.tar.xz
-  category: leather
+  category: bottle
   task: segmentation
   label_format: None
   image_size: 256
@@ -47,18 +47,15 @@ trainer:
   accelerator: null
   accumulate_grad_batches: 1
   amp_backend: native
-  amp_level: O2
   auto_lr_find: false
   auto_scale_batch_size: false
   auto_select_gpus: false
   benchmark: false
   check_val_every_n_epoch: 1
   checkpoint_callback: true
   default_root_dir: null
-  deterministic: true
-  distributed_backend: null
+  deterministic: false
   fast_dev_run: false
-  flush_logs_every_n_steps: 100
   gpus: 1
   gradient_clip_val: 0
   limit_predict_batches: 1.0
@@ -68,7 +65,7 @@ trainer:
   log_every_n_steps: 50
   log_gpu_memory: null
   max_epochs: 50
-  max_steps: null
+  max_steps: -1
   min_epochs: null
   min_steps: null
   move_metrics_to_cpu: false
@@ -83,14 +80,12 @@ trainer:
   process_position: 0
   profiler: null
   progress_bar_refresh_rate: null
-  reload_dataloaders_every_epoch: false
   replace_sampler_ddp: true
   stochastic_weight_avg: false
   sync_batchnorm: false
   terminate_on_nan: false
   tpu_cores: null
   track_grad_norm: -1
-  truncated_bptt_steps: null
   val_check_interval: 1.0
   weights_save_path: null
   weights_summary: top