Bump up PyTorch Lightning version to v1.9.*

CHANGELOG.md

@@ -8,6 +8,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/).
 
 ### Added
 
+- Bump up PyTorch Lightning version to v.1.9.\* (<https://github.com/openvinotoolkit/anomalib/pull/870>)
 - Add ShanghaiTech Campus video anomaly detection dataset (<https://github.com/openvinotoolkit/anomalib/pull/869>)
 - Add `pyupgrade` to `pre-commit` configs, and refactor based on `pyupgrade` and `refurb` (<https://github.com/openvinotoolkit/anomalib/pull/845>)
 - Add [CFA](https://arxiv.org/abs/2206.04325) model implementation (<https://github.com/openvinotoolkit/anomalib/pull/783>)

anomalib/data/btech.py

@@ -19,7 +19,6 @@
 import cv2
 import pandas as pd
 from pandas.core.frame import DataFrame
-from pytorch_lightning.utilities.cli import DATAMODULE_REGISTRY
 from tqdm import tqdm
 
 from anomalib.data.base import AnomalibDataModule, AnomalibDataset
@@ -176,7 +175,6 @@ def _setup(self) -> None:
         self.samples = make_btech_dataset(path=self.root_category, split=self.split)
 
 
-@DATAMODULE_REGISTRY
 class BTech(AnomalibDataModule):
     """BTech Lightning Data Module.
 

anomalib/models/cfa/config.yaml

@@ -66,49 +66,43 @@ optimization:
 
 # PL Trainer Args. Don't add extra parameter here.
 trainer:
-  accelerator: auto # <"cpu", "gpu", "tpu", "ipu", "hpu", "auto">
-  accumulate_grad_batches: 1
-  amp_backend: native
-  auto_lr_find: false
-  auto_scale_batch_size: false
-  auto_select_gpus: false
-  benchmark: false
-  check_val_every_n_epoch: 1
+  enable_checkpointing: true
   default_root_dir: null
-  detect_anomaly: false
-  deterministic: false
+  gradient_clip_val: 0
+  gradient_clip_algorithm: norm
+  num_nodes: 1
   devices: 1
-  enable_checkpointing: true
-  enable_model_summary: true
   enable_progress_bar: true
+  overfit_batches: 0.0
+  track_grad_norm: -1
+  check_val_every_n_epoch: 1 # Don't validate before extracting features.
   fast_dev_run: false
-  gpus: null # Set automatically
-  gradient_clip_val: 0
-  ipus: null
-  limit_predict_batches: 1.0
-  limit_test_batches: 1.0
-  limit_train_batches: 1.0
-  limit_val_batches: 1.0
-  log_every_n_steps: 50
-  log_gpu_memory: null
+  accumulate_grad_batches: 1
   max_epochs: 30
-  max_steps: -1
-  max_time: null
   min_epochs: null
+  max_steps: -1
   min_steps: null
-  move_metrics_to_cpu: false
-  multiple_trainloader_mode: max_size_cycle
-  num_nodes: 1
-  num_processes: null
-  num_sanity_val_steps: 0
-  overfit_batches: 0.0
-  plugins: null
+  max_time: null
+  limit_train_batches: 1.0
+  limit_val_batches: 1.0
+  limit_test_batches: 1.0
+  limit_predict_batches: 1.0
+  val_check_interval: 1.0 # Don't validate before extracting features.
+  log_every_n_steps: 50
+  accelerator: auto # <"cpu", "gpu", "tpu", "ipu", "hpu", "auto">
+  strategy: null
+  sync_batchnorm: false
   precision: 32
+  enable_model_summary: true
+  num_sanity_val_steps: 0
   profiler: null
+  benchmark: false
+  deterministic: false
   reload_dataloaders_every_n_epochs: 0
+  auto_lr_find: false
   replace_sampler_ddp: true
-  strategy: null
-  sync_batchnorm: false
-  tpu_cores: null
-  track_grad_norm: -1
-  val_check_interval: 1.0
+  detect_anomaly: false
+  auto_scale_batch_size: false
+  plugins: null
+  move_metrics_to_cpu: false
+  multiple_trainloader_mode: max_size_cycle

anomalib/models/cfa/lightning_model.py

@@ -16,7 +16,6 @@
 from omegaconf import DictConfig, ListConfig
 from pytorch_lightning import Callback
 from pytorch_lightning.callbacks import EarlyStopping
-from pytorch_lightning.utilities.cli import MODEL_REGISTRY
 from pytorch_lightning.utilities.types import STEP_OUTPUT
 from torch import Tensor
 from torch.optim.optimizer import Optimizer
@@ -30,7 +29,6 @@
 __all__ = ["Cfa", "CfaLightning"]
 
 
-@MODEL_REGISTRY
 class Cfa(AnomalyModule):
     """CFA: Coupled-hypersphere-based Feature Adaptation for Target-Oriented Anomaly Localization.
 
@@ -83,6 +81,8 @@ def training_step(self, batch: dict[str, str | Tensor], *args, **kwargs) -> STEP
         Returns:
             STEP_OUTPUT: Loss value.
         """
+        del args, kwargs  # These variables are not used.
+
         distance = self.model(batch["image"])
         loss = self.loss(distance)
         return {"loss": loss}
@@ -96,6 +96,8 @@ def validation_step(self, batch: dict[str, str | Tensor], *args, **kwargs) -> ST
         Returns:
             dict: Anomaly map computed by the model.
         """
+        del args, kwargs  # These variables are not used.
+
         batch["anomaly_maps"] = self.model(batch["image"])
         return batch
 
@@ -107,7 +109,8 @@ def backward(self, loss: Tensor, optimizer: Optimizer | None, optimizer_idx: int
             optimizer (Optimizer | None): Optimizer.
             optimizer_idx (int | None): Optimizer index.
         """
-        del optimizer, optimizer_idx  # These variables are not used.
+        del optimizer, optimizer_idx, args, kwargs  # These variables are not used.
+
         # TODO: Investigate why retain_graph is needed.
         loss.backward(retain_graph=True)
 

anomalib/models/cflow/config.yaml

@@ -72,49 +72,43 @@ optimization:
 
 # PL Trainer Args. Don't add extra parameter here.
 trainer:
-  accelerator: auto # <"cpu", "gpu", "tpu", "ipu", "hpu", "auto">
-  accumulate_grad_batches: 1
-  amp_backend: native
-  auto_lr_find: false
-  auto_scale_batch_size: false
-  auto_select_gpus: false
-  benchmark: false
-  check_val_every_n_epoch: 1
+  enable_checkpointing: true
   default_root_dir: null
-  detect_anomaly: false
-  deterministic: false
+  gradient_clip_val: 0
+  gradient_clip_algorithm: norm
+  num_nodes: 1
   devices: 1
-  enable_checkpointing: true
-  enable_model_summary: true
   enable_progress_bar: true
+  overfit_batches: 0.0
+  track_grad_norm: -1
+  check_val_every_n_epoch: 1 # Don't validate before extracting features.
   fast_dev_run: false
-  gpus: null # Set automatically
-  gradient_clip_val: 0
-  ipus: null
-  limit_predict_batches: 1.0
-  limit_test_batches: 1.0
-  limit_train_batches: 1.0
-  limit_val_batches: 1.0
-  log_every_n_steps: 50
-  log_gpu_memory: null
+  accumulate_grad_batches: 1
   max_epochs: 50
-  max_steps: -1
-  max_time: null
   min_epochs: null
+  max_steps: -1
   min_steps: null
-  move_metrics_to_cpu: false
-  multiple_trainloader_mode: max_size_cycle
-  num_nodes: 1
-  num_processes: null
-  num_sanity_val_steps: 0
-  overfit_batches: 0.0
-  plugins: null
+  max_time: null
+  limit_train_batches: 1.0
+  limit_val_batches: 1.0
+  limit_test_batches: 1.0
+  limit_predict_batches: 1.0
+  val_check_interval: 1.0 # Don't validate before extracting features.
+  log_every_n_steps: 50
+  accelerator: auto # <"cpu", "gpu", "tpu", "ipu", "hpu", "auto">
+  strategy: null
+  sync_batchnorm: false
   precision: 32
+  enable_model_summary: true
+  num_sanity_val_steps: 0
   profiler: null
+  benchmark: false
+  deterministic: false
   reload_dataloaders_every_n_epochs: 0
+  auto_lr_find: false
   replace_sampler_ddp: true
-  strategy: null
-  sync_batchnorm: false
-  tpu_cores: null
-  track_grad_norm: -1
-  val_check_interval: 1.0
+  detect_anomaly: false
+  auto_scale_batch_size: false
+  plugins: null
+  move_metrics_to_cpu: false
+  multiple_trainloader_mode: max_size_cycle

anomalib/models/cflow/lightning_model.py

@@ -13,7 +13,6 @@
 import torch.nn.functional as F
 from omegaconf import DictConfig, ListConfig
 from pytorch_lightning.callbacks import EarlyStopping
-from pytorch_lightning.utilities.cli import MODEL_REGISTRY
 from pytorch_lightning.utilities.types import STEP_OUTPUT
 from torch import Tensor, optim
 from torch.optim import Optimizer
@@ -25,7 +24,6 @@
 __all__ = ["Cflow", "CflowLightning"]
 
 
-@MODEL_REGISTRY
 class Cflow(AnomalyModule):
     """PL Lightning Module for the CFLOW algorithm."""
 
@@ -98,6 +96,8 @@ def training_step(self, batch: dict[str, str | Tensor], *args, **kwargs) -> STEP
           Loss value for the batch
 
         """
+        del args, kwargs  # These variables are not used.
+
         opt = self.optimizers()
         self.model.encoder.eval()
 
@@ -169,8 +169,9 @@ def validation_step(self, batch: dict[str, str | Tensor], *args, **kwargs) -> ST
             These are required in `validation_epoch_end` for feature concatenation.
 
         """
-        batch["anomaly_maps"] = self.model(batch["image"])
+        del args, kwargs  # These variables are not used.
 
+        batch["anomaly_maps"] = self.model(batch["image"])
         return batch
 
 

anomalib/models/components/base/anomaly_module.py

@@ -12,7 +12,7 @@
 
 import pytorch_lightning as pl
 import torch
-from pytorch_lightning.callbacks.base import Callback
+from pytorch_lightning.callbacks import Callback
 from pytorch_lightning.utilities.types import EPOCH_OUTPUT, STEP_OUTPUT
 from torch import Tensor, nn
 from torchmetrics import Metric
@@ -62,6 +62,8 @@ def forward(self, batch: dict[str, str | Tensor], *args, **kwargs) -> Any:
         Returns:
             Tensor: Output tensor from the model.
         """
+        del args, kwargs  # These variables are not used.
+
         return self.model(batch)
 
     def validation_step(self, batch: dict[str, str | Tensor], *args, **kwargs) -> STEP_OUTPUT:
@@ -113,16 +115,22 @@ def test_step(self, batch: dict[str, str | Tensor], batch_idx: int, *args, **kwa
           Dictionary containing images, features, true labels and masks.
           These are required in `validation_epoch_end` for feature concatenation.
         """
+        del args, kwargs  # These variables are not used.
+
         return self.predict_step(batch, batch_idx)
 
     def validation_step_end(self, val_step_outputs: STEP_OUTPUT, *args, **kwargs) -> STEP_OUTPUT:
         """Called at the end of each validation step."""
+        del args, kwargs  # These variables are not used.
+
         self._outputs_to_cpu(val_step_outputs)
         self._post_process(val_step_outputs)
         return val_step_outputs
 
     def test_step_end(self, test_step_outputs: STEP_OUTPUT, *args, **kwargs) -> STEP_OUTPUT:
         """Called at the end of each test step."""
+        del args, kwargs  # These variables are not used.
+
         self._outputs_to_cpu(test_step_outputs)
         self._post_process(test_step_outputs)
         return test_step_outputs

anomalib/models/csflow/config.yaml

@@ -73,49 +73,43 @@ optimization:
   export_mode: null #options: onnx, openvino
 # PL Trainer Args. Don't add extra parameter here.
 trainer:
-  accelerator: auto # <"cpu", "gpu", "tpu", "ipu", "hpu", "auto">
-  accumulate_grad_batches: 1
-  amp_backend: native
-  auto_lr_find: false
-  auto_scale_batch_size: false
-  auto_select_gpus: false
-  benchmark: false
-  check_val_every_n_epoch: 1
+  enable_checkpointing: true
   default_root_dir: null
-  detect_anomaly: false
-  deterministic: false
+  gradient_clip_val: 1 # Grad clip value set based on the official implementation
+  gradient_clip_algorithm: norm
+  num_nodes: 1
   devices: 1
-  enable_checkpointing: true
-  enable_model_summary: true
   enable_progress_bar: true
+  overfit_batches: 0.0
+  track_grad_norm: -1
+  check_val_every_n_epoch: 1 # Don't validate before extracting features.
   fast_dev_run: false
-  gpus: null # Set automatically
-  gradient_clip_val: 1 # Grad clip value set based on the official implementation
-  ipus: null
-  limit_predict_batches: 1.0
-  limit_test_batches: 1.0
-  limit_train_batches: 1.0
-  limit_val_batches: 1.0
-  log_every_n_steps: 50
-  log_gpu_memory: null
+  accumulate_grad_batches: 1
   max_epochs: 240
-  max_steps: -1
-  max_time: null
   min_epochs: null
+  max_steps: -1
   min_steps: null
-  move_metrics_to_cpu: false
-  multiple_trainloader_mode: max_size_cycle
-  num_nodes: 1
-  num_processes: null
-  num_sanity_val_steps: 0
-  overfit_batches: 0.0
-  plugins: null
+  max_time: null
+  limit_train_batches: 1.0
+  limit_val_batches: 1.0
+  limit_test_batches: 1.0
+  limit_predict_batches: 1.0
+  val_check_interval: 1.0 # Don't validate before extracting features.
+  log_every_n_steps: 50
+  accelerator: auto # <"cpu", "gpu", "tpu", "ipu", "hpu", "auto">
+  strategy: null
+  sync_batchnorm: false
   precision: 32
+  enable_model_summary: true
+  num_sanity_val_steps: 0
   profiler: null
+  benchmark: false
+  deterministic: false
   reload_dataloaders_every_n_epochs: 0
+  auto_lr_find: false
   replace_sampler_ddp: true
-  strategy: null
-  sync_batchnorm: false
-  tpu_cores: null
-  track_grad_norm: -1
-  val_check_interval: 1.0
+  detect_anomaly: false
+  auto_scale_batch_size: false
+  plugins: null
+  move_metrics_to_cpu: false
+  multiple_trainloader_mode: max_size_cycle