discopat.nn_training package

Subpackages

• discopat.nn_training.detr package
  • Submodules
  • discopat.nn_training.detr.collate module
    • collate_fn()
  • discopat.nn_training.detr.criterion module
    • SetCriterion
      • SetCriterion.forward()
      • SetCriterion.get_loss()
      • SetCriterion.loss_boxes()
      • SetCriterion.loss_cardinality()
      • SetCriterion.loss_labels()
    • accuracy()
  • discopat.nn_training.detr.detr_trainer module
    • DetrTrainer
      • DetrTrainer.set_default_lr_scheduler()
      • DetrTrainer.set_default_optimiser()
      • DetrTrainer.train()
  • discopat.nn_training.detr.engine module
    • evaluate()
    • train_one_epoch()
  • discopat.nn_training.detr.matcher module
    • HungarianMatcher
      • HungarianMatcher.forward()
  • discopat.nn_training.detr.nested_tensor module
    • NestedTensor
      • NestedTensor.decompose()
      • NestedTensor.to()
    • nested_tensor_from_tensor_list()
  • Module contents
• discopat.nn_training.evaluation package
  • Submodules
  • discopat.nn_training.evaluation.base module
    • compute_ap()
    • evaluate()
  • discopat.nn_training.evaluation.matching module
    • compute_center_distance_matrix()
    • compute_iomean_matrix()
    • compute_iou_matrix()
    • match_groundtruths_and_predictions()
    • to_np_array()
  • Module contents
    • compute_ap()
    • evaluate()
• discopat.nn_training.torch_detection_utils package
  • Submodules
  • discopat.nn_training.torch_detection_utils.coco_eval module
    • CocoEvaluator
      • CocoEvaluator.accumulate()
      • CocoEvaluator.prepare()
      • CocoEvaluator.prepare_for_coco_detection()
      • CocoEvaluator.prepare_for_coco_keypoint()
      • CocoEvaluator.prepare_for_coco_segmentation()
      • CocoEvaluator.summarize()
      • CocoEvaluator.synchronize_between_processes()
      • CocoEvaluator.update()
    • convert_to_xywh()
    • create_common_coco_eval()
    • evaluate()
    • merge()
  • discopat.nn_training.torch_detection_utils.coco_utils module
    • convert_to_coco_api()
    • get_coco_api_from_dataset()
  • discopat.nn_training.torch_detection_utils.engine module
    • train_one_epoch()
  • discopat.nn_training.torch_detection_utils.utils module
    • MetricLogger
      • MetricLogger.add_meter()
      • MetricLogger.log_every()
      • MetricLogger.synchronize_between_processes()
      • MetricLogger.update()
    • SmoothedValue
      • SmoothedValue.avg
      • SmoothedValue.global_avg
      • SmoothedValue.max
      • SmoothedValue.median
      • SmoothedValue.synchronize_between_processes()
      • SmoothedValue.update()
      • SmoothedValue.value
    • all_gather()
    • collate_fn()
    • get_rank()
    • get_world_size()
    • is_dist_avail_and_initialized()
    • is_main_process()
    • mkdir()
    • reduce_dict()
    • setup_for_distributed()
  • Module contents

Submodules

discopat.nn_training.nn_trainer module

class discopat.nn_training.nn_trainer.NNTrainer(net, dataset, val_dataset, parameters, device, callbacks=None)

Bases: ABC

Parameters:

• net (NeuralNet)

• dataset (Dataset)

• val_dataset (Dataset)

• parameters (dict[str, Any])

• device (ComputingDevice)

• callbacks (list or None)

static parse_parameters(parameters)

Parameters:

parameters (dict[str, Any])

Return type:

tuple(dict, dict, dict)

abstractmethod set_default_lr_scheduler()

Return type:

Any

abstractmethod set_default_optimiser()

Return type:

Any

abstractmethod train()

Return type:

None

discopat.nn_training.torch module

class discopat.nn_training.torch.TorchNNTrainer(net, dataset, val_dataset, parameters, device, callbacks=None)

Bases: NNTrainer

Parameters:

• net (NeuralNet)

• dataset (Dataset)

• val_dataset (Dataset)

• parameters (dict[str, Any])

• device (ComputingDevice)

• callbacks (list or None)

set_default_lr_scheduler()

Return type:

LRScheduler

set_default_optimiser()

Return type:

Optimizer

train(num_epochs)

Parameters:

num_epochs (int)

discopat.nn_training.utils module

class discopat.nn_training.utils.MetricLogger(delimiter='\t')

Bases: object

add_meter(name, meter)

log_every(iterable, print_freq, header=None)

synchronize_between_processes()

update(**kwargs)

class discopat.nn_training.utils.SmoothedValue(window_size=20, fmt=None)

Bases: object

Track a series of values and provide access to smoothed values over a window or the global series average.

property avg

property global_avg

property max

property median

synchronize_between_processes()

Warning: does not synchronize the deque.

update(value, n=1)

property value

discopat.nn_training.utils.get_world_size()

discopat.nn_training.utils.is_dist_avail_and_initialized()

Module contents