Error with torch.cdist when testing Patchcore model on a custom dataset

[aleperat97]

I tried to train and test the Patchcore model with the WOOD Anomaly Detection dataset, which can be found here: https://euhubs4data.eu/datasets/iti-wood-anomaly-detection-one-class-classification/ . The dataset is divided into train, validation, test, and masks folders but I rearranged the files in order to match the dataset structure required by the .yaml file for a custom dataset (i.e., normal, abnormal, normal_test, and masks folder). The train was ok, but at the beginning of testing this error occurs:

My code:

from anomalib.utils.callbacks import LoadModelCallback, get_callbacks
from anomalib.config import get_configurable_parameters
from anomalib.data import get_datamodule
from pytorch_lightning import Trainer

# Path to model configuration file 
CONFIG_PATH = f"datasets/config_patchcore.yaml"

# Get configurable parameters from yaml file
config = get_configurable_parameters(config_path=CONFIG_PATH)

# Load and prepare dataset
datamodule = get_datamodule(config)
datamodule.setup()  
datamodule.prepare_data()

# Model definition
model = get_model(config)

# Callbacks
callbacks = get_callbacks(config)

# Trainer definition
trainer = Trainer(**config.trainer, callbacks=callbacks)

# Training
trainer.fit(model=model, datamodule=datamodule)

# Testing
trainer.test(datamodule=datamodule, model=model)
​

My yaml configuration file:

dataset:
  name: WOOD_dataset
  format: folder
  path: datasets/WOOD_dataset
  normal_dir: normal # name of the folder containing normal images.
  abnormal_dir: abnormal # name of the folder containing abnormal images.
  normal_test_dir: normal_test # name of the folder containing normal test images.
  task: segmentation # classification or segmentation
  mask: datasets/WOOD_dataset/masks/ #optional
  extensions: null
  split_ratio:  # ratio of the normal images that will be used to create a test split
  image_size: 224
  train_batch_size: 32
  test_batch_size: 1
  num_workers: 0
  transform_config:
    train: null
    val: null
  create_validation_set: false
  tiling:
    apply: false
    tile_size: null
    stride: null
    remove_border_count: 0
    use_random_tiling: False
    random_tile_count: 16
    
model:
  name: patchcore
  backbone: wide_resnet50_2 #wide_resnet50_2
  pre_trained: true
  layers:
    - layer2
    - layer3
  coreset_sampling_ratio: 0.1
  num_neighbors: 9
  normalization_method: min_max # options: [null, min_max, cdf]

metrics:
  image:
    - Precision
    - Recall
    - F1Score
    - AUPR
  pixel:
    #- F1Score        
    - AUPR
  threshold:
    image_default: 0
    pixel_default: 0
    adaptive: true

visualization:
  show_images: False # show images on the screen
  save_images: True # save images to the file system
  log_images: True # log images to the available loggers (if any)
  image_save_path: null # path to which images will be saved
  mode: full # options: ["full", "simple"]

project:
  seed: 0
  path: ./results

logging:
  logger: [] # options: [comet, tensorboard, wandb, csv] or combinations.
  log_graph: false # Logs the model graph to respective logger.

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: 10 # Don't validate before extracting features.
  default_root_dir: null
  detect_anomaly: false
  deterministic: false
  devices: 1
  enable_checkpointing: true
  enable_model_summary: true
  enable_progress_bar: true
  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: 25
  log_gpu_memory: null
  max_epochs: 1
  max_steps: -1
  max_time: null
  min_epochs: null
  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
  precision: 32
  profiler: null
  reload_dataloaders_every_n_epochs: 0
  replace_sampler_ddp: true
  strategy: null
  sync_batchnorm: false
  tpu_cores: null
  track_grad_norm: -1
  val_check_interval: 1.0 # Don't validate before extracting features.

All libraries versions are as required by the library.
Thanks for the help

[djdameln]

@aleperat97 apologies for the delay in responding. Are you still having this problem in the latest version of Anomalib? We have recently made some changes to the PatchCore score computations, so it's possible that this also solved the problem you reported. Please let us know if you're still having this problem so we could have a look.

[kekaipr]

Hi, I have same problem, did you solve it?

[samet-akcay]

#1267 should fix this. I'm closing this issue. if anyone encounters this again, feel free to re-open