quadra 0.0.1__py3-none-any.whl → 2.1.13__py3-none-any.whl

quadra/configs/callbacks/all.yaml

@@ -0,0 +1,32 @@
+model_checkpoint:
+  _target_: pytorch_lightning.callbacks.ModelCheckpoint
+  monitor: "val_loss" # name of the logged metric which determines when model is improving
+  mode: "min" # can be "max" or "min"
+  save_top_k: 1 # save k best models (determined by above metric)
+  save_last: True # additionaly always save model from last epoch
+  verbose: False
+  dirpath: "checkpoints/"
+  filename: "epoch_{epoch:03d}"
+  auto_insert_metric_name: False
+
+upload_code_as_artifact:
+  _target_: quadra.callbacks.mlflow.UploadCodeAsArtifact
+  source_dir: ${work_dir}/quadra/
+
+upload_ckpts_as_artifact:
+  _target_: quadra.callbacks.mlflow.UploadCheckpointsAsArtifact
+  ckpt_dir: "checkpoints/"
+  upload_best_only: true
+  delete_after_upload: false
+  upload: true
+log_gradients:
+  _target_: quadra.callbacks.mlflow.LogGradients
+  norm: 2
+lr_monitor:
+  _target_: pytorch_lightning.callbacks.LearningRateMonitor
+  logging_interval: "epoch"
+progress_bar:
+  _target_: pytorch_lightning.callbacks.progress.TQDMProgressBar
+lightning_trainer_setup:
+  _target_: quadra.callbacks.lightning.LightningTrainerBaseSetup
+  log_every_n_steps: 1

quadra/configs/callbacks/default.yaml

@@ -0,0 +1,37 @@
+model_checkpoint:
+  _target_: pytorch_lightning.callbacks.ModelCheckpoint
+  monitor: "val_loss" # name of the logged metric which determines when model is improving
+  mode: "min" # can be "max" or "min"
+  save_top_k: 1 # save k best models (determined by above metric)
+  save_last: True # additionaly always save model from last epoch
+  verbose: False
+  dirpath: "checkpoints/"
+  filename: "epoch_{epoch:03d}"
+  auto_insert_metric_name: False
+
+log_gradients:
+  _target_: quadra.callbacks.mlflow.LogGradients
+  norm: 2
+lr_monitor:
+  _target_: pytorch_lightning.callbacks.LearningRateMonitor
+  logging_interval: "epoch"
+progress_bar:
+  _target_: pytorch_lightning.callbacks.progress.TQDMProgressBar
+lightning_trainer_setup:
+  _target_: quadra.callbacks.lightning.LightningTrainerBaseSetup
+  log_every_n_steps: 1
+
+batch_size_finder:
+  _target_: quadra.callbacks.lightning.BatchSizeFinder
+  mode: power
+  steps_per_trial: 3
+  init_val: 2
+  max_trials: 5 # Max 64
+  batch_arg_name: batch_size
+  disable: true
+  find_train_batch_size: true
+  find_validation_batch_size: false
+  find_test_batch_size: false
+  find_predict_batch_size: false
+#gpu_stats: TODO: This is not working with the current PL version
+#  _target_: nvitop.callbacks.lightning.GpuStatsLogger

quadra/configs/callbacks/default_anomalib.yaml

@@ -0,0 +1,67 @@
+# Anomalib specific callbacks
+score_normalization:
+  _target_: quadra.utils.anomaly.ThresholdNormalizationCallback
+  threshold_type: image
+post_processing_configuration:
+  _target_: anomalib.utils.callbacks.post_processing_configuration.PostProcessingConfigurationCallback
+  threshold_method: ${model.metrics.threshold.method}
+  manual_image_threshold: ${model.metrics.threshold.manual_image}
+  manual_pixel_threshold: ${model.metrics.threshold.manual_pixel}
+metrics:
+  _target_: anomalib.utils.callbacks.metrics_configuration.MetricsConfigurationCallback
+  task: ${model.dataset.task}
+  image_metrics: ${model.metrics.image}
+  pixel_metrics: ${model.metrics.pixel}
+visualizer:
+  _target_: quadra.callbacks.anomalib.VisualizerCallback
+  inputs_are_normalized: true
+  output_path: anomaly_output
+  threshold_type: ${callbacks.score_normalization.threshold_type}
+  disable: true
+  plot_only_wrong: false
+  plot_raw_outputs: false
+# Standard callbacks
+early_stopping:
+  _target_: pytorch_lightning.callbacks.EarlyStopping
+  monitor: validation_image_F1
+  mode: max
+  patience: 5
+  verbose: false
+  stopping_threshold: 1
+model_checkpoint:
+  _target_: pytorch_lightning.callbacks.ModelCheckpoint
+  monitor: validation_image_F1 # name of the logged metric which determines when model is improving
+  mode: "max"
+  verbose: false
+  dirpath: "checkpoints/"
+  filename: "final_model"
+  auto_insert_metric_name: false
+  save_last: true
+  save_on_train_epoch_end: false
+upload_ckpts_as_artifact:
+  _target_: quadra.callbacks.mlflow.UploadCheckpointsAsArtifact
+  ckpt_dir: "checkpoints/"
+  upload_best_only: true
+  delete_after_upload: true
+  upload: false
+log_gradients:
+  _target_: quadra.callbacks.mlflow.LogGradients
+  norm: 2
+lr_monitor:
+  _target_: pytorch_lightning.callbacks.LearningRateMonitor
+  logging_interval: "epoch"
+progress_bar:
+  _target_: pytorch_lightning.callbacks.progress.TQDMProgressBar
+lightning_trainer_setup:
+  _target_: quadra.callbacks.lightning.LightningTrainerBaseSetup
+  log_every_n_steps: 1
+batch_size_finder:
+  _target_: quadra.callbacks.lightning.BatchSizeFinder
+  mode: power
+  steps_per_trial: 3
+  init_val: 2
+  max_trials: 5 # Max 64
+  batch_arg_name: train_batch_size
+  disable: true
+#gpu_stats: TODO: This is not working with the current PL version
+#  _target_: nvitop.callbacks.lightning.GpuStatsLogger

quadra/configs/config.yaml

@@ -0,0 +1,33 @@
+# @package _global_
+
+# specify here default training configuration
+defaults:
+  - _self_
+  - core: default
+  - loss: null
+  - optimizer: null
+  - scheduler: null
+  - task: default
+  - transforms: default
+  - trainer: lightning_gpu
+  - backbone: null
+  - model: null
+  - datamodule: null
+  - callbacks: default
+  - logger: mlflow
+  - export: default
+  - inference: default
+  - hparams_search: null
+  - hydra: default
+  # BE CAREFUL TO KEEP EXPERIMENT LAST
+  - experiment: null
+
+  # enable color logging
+  - override hydra/hydra_logging: colorlog
+  - override hydra/job_logging: colorlog
+
+work_dir: ${hydra:runtime.cwd}
+debug: False
+print_config: true
+ignore_warnings: true
+validate: true

quadra/configs/core/default.yaml

@@ -0,0 +1,11 @@
+version:
+  _target_: quadra.get_version
+seed: 42
+tag: null
+name: null
+cv2_num_threads: 1
+command: "python "
+experiment_path: null
+upload_artifacts: False
+upload_models: ${export.types} # Default behavior in quadra <= 1.5.6
+log_level: info

quadra/configs/datamodule/base/anomaly.yaml

@@ -0,0 +1,16 @@
+_target_: quadra.datamodules.AnomalyDataModule
+data_path: ???
+category:
+num_workers: 8
+train_batch_size: 32
+test_batch_size: 32
+seed: ${core.seed}
+train_transform: ${transforms.train_transform}
+test_transform: ${transforms.test_transform}
+val_transform: ${transforms.val_transform}
+phase: train
+valid_area_mask:
+crop_area:
+enable_hashing: true
+hash_size: 64
+hash_type: content

quadra/configs/datamodule/base/classification.yaml

@@ -0,0 +1,21 @@
+_target_: quadra.datamodules.classification.ClassificationDataModule
+data_path: ???
+exclude_filter: [".ipynb_checkpoints"]
+seed: ${core.seed}
+num_workers: 8
+batch_size: 16
+val_size: 0.2
+train_transform: ${transforms.train_transform}
+test_transform: ${transforms.test_transform}
+val_transform: ${transforms.val_transform}
+train_split_file:
+test_split_file:
+label_map:
+class_to_idx:
+name:
+dataset:
+  _target_: hydra.utils.get_method
+  path: quadra.datasets.classification.ClassificationDataset
+enable_hashing: true
+hash_size: 64
+hash_type: content

quadra/configs/datamodule/base/multilabel_classification.yaml

@@ -0,0 +1,23 @@
+_target_: quadra.datamodules.MultilabelClassificationDataModule
+data_path: ???
+images_and_labels_file: null
+train_split_file: null
+test_split_file: null
+val_split_file: null
+dataset:
+  _target_: hydra.utils.get_method
+  path: quadra.datasets.classification.MultilabelClassificationDataset
+num_classes: null
+num_workers: 8
+batch_size: 64
+test_batch_size: 64
+seed: ${core.seed}
+val_size: 0.2
+test_size: 0.2
+train_transform: ${transforms.train_transform}
+test_transform: ${transforms.test_transform}
+val_transform: ${transforms.val_transform}
+class_to_idx: null
+enable_hashing: true
+hash_size: 64
+hash_type: content

quadra/configs/datamodule/base/segmentation.yaml

@@ -0,0 +1,18 @@
+_target_: quadra.datamodules.SegmentationDataModule
+data_path: ???
+test_size: 0.3
+val_size: 0.3
+seed: ${core.seed}
+batch_size: 32
+num_workers: 6
+train_transform: ${transforms.train_transform}
+test_transform: ${transforms.test_transform}
+val_transform: ${transforms.val_transform}
+train_split_file:
+test_split_file:
+val_split_file:
+num_data_class:
+exclude_good: false
+enable_hashing: true
+hash_size: 64
+hash_type: content

quadra/configs/datamodule/base/segmentation_multiclass.yaml

@@ -0,0 +1,20 @@
+_target_: quadra.datamodules.SegmentationMulticlassDataModule
+data_path: ???
+idx_to_class:
+test_size: 0.3
+val_size: 0.3
+seed: 42
+batch_size: 32
+num_workers: 6
+train_transform: ${transforms.train_transform}
+test_transform: ${transforms.test_transform}
+val_transform: ${transforms.val_transform}
+train_split_file:
+test_split_file:
+val_split_file:
+exclude_good: false
+num_data_train:
+one_hot_encoding:
+enable_hashing: true
+hash_size: 64
+hash_type: content

quadra/configs/datamodule/base/sklearn_classification.yaml

@@ -0,0 +1,23 @@
+_target_: quadra.datamodules.SklearnClassificationDataModule
+data_path: ???
+exclude_filter:
+include_filter:
+val_size: 0.3
+class_to_idx:
+label_map:
+seed: ${core.seed}
+batch_size: 32
+num_workers: 8
+train_transform: ${transforms.train_transform}
+val_transform: ${transforms.val_transform}
+test_transform: ${transforms.test_transform}
+roi:
+n_splits: 1
+phase:
+cache: false
+limit_training_data:
+train_split_file:
+test_split_file:
+enable_hashing: true
+hash_size: 64
+hash_type: content

quadra/configs/datamodule/base/sklearn_classification_patch.yaml

@@ -0,0 +1,17 @@
+_target_: quadra.datamodules.PatchSklearnClassificationDataModule
+data_path: ???
+train_filename: dataset.txt
+exclude_filter:
+include_filter:
+class_to_idx:
+seed: 42
+batch_size: 32
+num_workers: 8
+train_transform: ${transforms.train_transform}
+test_transform: ${transforms.test_transform}
+val_transform: ${transforms.val_transform}
+balance_classes: false
+class_to_skip_training:
+enable_hashing: true
+hash_size: 64
+hash_type: content

quadra/configs/datamodule/base/ssl.yaml

@@ -0,0 +1,21 @@
+_target_: quadra.datamodules.SSLDataModule
+data_path: ???
+exclude_filter:
+include_filter:
+seed: ${core.seed}
+num_workers: 8
+batch_size: 16
+augmentation_dataset: null
+train_transform: null
+test_transform: ${transforms.test_transform}
+val_transform: ${transforms.val_transform}
+train_split_file:
+val_split_file:
+test_split_file:
+val_size: 0.3
+test_size: 0.1
+split_validation: true
+class_to_idx:
+enable_hashing: true
+hash_size: 64
+hash_type: content

quadra/configs/datamodule/generic/imagenette/classification/base.yaml

@@ -0,0 +1,9 @@
+_target_: quadra.datamodules.generic.imagenette.ImagenetteClassificationDataModule
+data_path: ${oc.env:HOME}/.quadra/datasets/imagenette
+imagenette_version: "320"
+num_workers: 8
+batch_size: 32
+seed: ${core.seed}
+train_transform: ${transforms.train_transform}
+test_transform: ${transforms.test_transform}
+val_transform: ${transforms.val_transform}

quadra/configs/datamodule/generic/imagenette/ssl/base.yaml

@@ -0,0 +1,10 @@
+_target_: quadra.datamodules.generic.imagenette.ImagenetteSSLDataModule
+data_path: ${oc.env:HOME}/.quadra/datasets/imagenette
+imagenette_version: "320"
+num_workers: 8
+batch_size: 32
+seed: ${core.seed}
+augmentation_dataset: null
+train_transform: null
+test_transform: ${transforms.test_transform}
+val_transform: ${transforms.val_transform}

quadra/configs/datamodule/generic/mnist/anomaly/base.yaml

@@ -0,0 +1,14 @@
+_target_: quadra.datamodules.generic.mnist.MNISTAnomalyDataModule
+data_path: ${oc.env:HOME}/.quadra/datasets/MNIST
+good_number: 9
+num_workers: 8
+limit_data: 100
+train_batch_size: 32
+test_batch_size: 32
+seed: ${core.seed}
+train_transform: ${transforms.train_transform}
+test_transform: ${transforms.test_transform}
+val_transform: ${transforms.val_transform}
+phase: train
+valid_area_mask:
+crop_area:

quadra/configs/datamodule/generic/mvtec/anomaly/base.yaml

@@ -0,0 +1,14 @@
+_target_: quadra.datamodules.generic.mvtec.MVTecDataModule
+data_path: ${oc.env:HOME}/.quadra/datasets/mvtec-dataset
+category: bottle
+num_workers: 8
+train_batch_size: 32
+test_batch_size: 32
+seed: ${core.seed}
+train_transform: ${transforms.train_transform}
+test_transform: ${transforms.test_transform}
+val_transform: ${transforms.val_transform}
+phase: train
+mask_suffix: _mask
+valid_area_mask:
+crop_area:

quadra/configs/datamodule/generic/oxford_pet/segmentation/base.yaml

@@ -0,0 +1,9 @@
+_target_: quadra.datamodules.generic.oxford_pet.OxfordPetSegmentationDataModule
+idx_to_class:
+  1: cat_or_dog
+data_path: ${oc.env:HOME}/.quadra/datasets/oxford-pet
+test_size: 0.2
+val_size: 0.2
+train_transform: ${transforms.train_transform}
+test_transform: ${transforms.test_transform}
+val_transform: ${transforms.val_transform}

quadra/configs/experiment/base/anomaly/cfa.yaml

@@ -0,0 +1,47 @@
+# @package _global_
+
+defaults:
+  - override /datamodule: base/anomaly
+  - override /model: anomalib/cfa
+  - override /optimizer: null
+  - override /scheduler: null
+  - override /transforms: default_resize
+  - override /loss: null
+  - override /task: anomalib/cfa
+  - override /backbone: null
+  - override /trainer: lightning_gpu
+  - override /callbacks: default_anomalib
+  - _self_
+
+transforms:
+  input_height: 224
+  input_width: 224
+
+datamodule:
+  num_workers: 12
+  train_batch_size: 32
+  test_batch_size: 32
+  task: ${model.dataset.task}
+  phase: train
+
+callbacks:
+  early_stopping:
+    patience: 5
+    monitor: validation_image_AUROC
+    mode: max
+
+core:
+  tag: "run"
+  name: cfa_${model.model.backbone}_${trainer.max_epochs}
+  test_after_training: true
+
+logger:
+  mlflow:
+    experiment_name:
+    run_name: ${core.name}
+
+# PL Trainer Args. Don't add extra parameter here.
+trainer:
+  devices: [0]
+  check_val_every_n_epoch: 1 # Don't validate before extracting features.
+  val_check_interval: 1.0 # Don't validate before extracting features.

quadra/configs/experiment/base/anomaly/cflow.yaml

@@ -0,0 +1,47 @@
+# @package _global_
+
+defaults:
+  - override /datamodule: base/anomaly
+  - override /model: anomalib/cflow
+  - override /optimizer: null
+  - override /scheduler: null
+  - override /transforms: default_resize
+  - override /loss: null
+  - override /task: anomalib/cflow
+  - override /backbone: null
+  - override /trainer: lightning_gpu
+  - override /callbacks: default_anomalib
+  - _self_
+
+transforms:
+  input_height: 256
+  input_width: 256
+
+datamodule:
+  num_workers: 12
+  train_batch_size: 32
+  test_batch_size: 32
+  task: ${model.dataset.task}
+  phase: train
+
+callbacks:
+  early_stopping:
+    patience: 2
+    monitor: validation_image_F1
+
+core:
+  tag: "run"
+  name: cflow_${model.model.backbone}_${trainer.max_epochs}
+  test_after_training: true
+
+logger:
+  mlflow:
+    experiment_name:
+    run_name: ${core.name}
+
+# PL Trainer Args. Don't add extra parameter here.
+trainer:
+  devices: [0]
+  check_val_every_n_epoch: 1 # Don't validate before extracting features.
+  val_check_interval: 1.0 # Don't validate before extracting features.
+  max_epochs: 50

quadra/configs/experiment/base/anomaly/csflow.yaml

@@ -0,0 +1,48 @@
+# @package _global_
+
+defaults:
+  - override /datamodule: base/anomaly
+  - override /model: anomalib/csflow
+  - override /optimizer: null
+  - override /scheduler: null
+  - override /transforms: default_resize
+  - override /loss: null
+  - override /task: anomalib/csflow
+  - override /backbone: null
+  - override /trainer: lightning_gpu
+  - override /callbacks: default_anomalib
+  - _self_
+
+transforms:
+  input_height: 256
+  input_width: 256
+
+datamodule:
+  num_workers: 12
+  train_batch_size: 4
+  test_batch_size: 4
+  task: ${model.dataset.task}
+
+callbacks:
+  early_stopping:
+    patience: 3
+    monitor: validation_image_AUROC
+
+core:
+  tag: "run"
+  name: csflow_${trainer.max_epochs}
+  test_after_training: true
+
+logger:
+  mlflow:
+    experiment_name:
+    run_name: ${core.name}
+
+# PL Trainer Args. Don't add extra parameter here.
+trainer:
+  devices: [0]
+  gradient_clip_val: 1 # Grad clip value set based on the official implementation
+  gradient_clip_algorithm: norm
+  check_val_every_n_epoch: 1 # Don't validate before extracting features.
+  val_check_interval: 1.0 # Don't validate before extracting features.
+  max_epochs: 240

quadra/configs/experiment/base/anomaly/draem.yaml

@@ -0,0 +1,51 @@
+# @package _global_
+
+defaults:
+  - override /datamodule: base/anomaly
+  - override /model: anomalib/draem
+  - override /optimizer: null
+  - override /scheduler: null
+  - override /transforms: default_resize
+  - override /loss: null
+  - override /task: anomalib/draem
+  - override /backbone: null
+  - override /trainer: lightning_gpu
+  - override /callbacks: default_anomalib
+  - _self_
+
+transforms:
+  input_height: 256
+  input_width: 256
+
+datamodule:
+  num_workers: 12
+  train_batch_size: 16
+  test_batch_size: 32
+  task: ${model.dataset.task}
+  category:
+
+callbacks:
+  early_stopping:
+    patience: 20
+    monitor: validation_image_F1
+
+print_config: false
+
+core:
+  tag: "run"
+  name: draem_${trainer.max_epochs}
+  test_after_training: true
+
+logger:
+  mlflow:
+    experiment_name:
+    run_name: ${core.name}
+
+# PL Trainer Args. Don't add extra parameter here.
+trainer:
+  devices: [0]
+  gradient_clip_val: 0
+  gradient_clip_algorithm: norm
+  check_val_every_n_epoch: 1 # Don't validate before extracting features.
+  max_epochs: 700
+  val_check_interval: 1.0 # Don't validate before extracting features.

quadra/configs/experiment/base/anomaly/efficient_ad.yaml

@@ -0,0 +1,43 @@
+# @package _global_
+
+defaults:
+  - override /datamodule: base/anomaly
+  - override /model: anomalib/efficient_ad
+  - override /optimizer: null
+  - override /scheduler: null
+  - override /transforms: default_resize
+  - override /loss: null
+  - override /task: anomalib/efficient_ad
+  - override /backbone: null
+  - override /trainer: lightning_gpu
+  - override /callbacks: default_anomalib
+  - _self_
+
+datamodule:
+  num_workers: 12
+  train_batch_size: 32
+  test_batch_size: 32
+  task: ${model.dataset.task}
+  category:
+  phase: train
+
+core:
+  tag: "run"
+  name: efficient_ad_${model.model.model_size}_${trainer.max_epochs}
+  test_after_training: true
+
+logger:
+  mlflow:
+    experiment_name:
+    run_name: ${core.name}
+
+trainer:
+  devices: [2]
+  check_val_every_n_epoch: ${trainer.max_epochs}
+  max_epochs: 20
+  max_steps: 20000
+  val_check_interval: 1.0 # Don't validate before extracting features.
+  # This will avoid issues with ModelSignatureWrapper
+  # As the default forward for EfficientAD is performed with a None attribute
+  # Which we currently can't handle
+  num_sanity_val_steps: 0