piscis.training =============== .. py:module:: piscis.training Functions --------- .. autoapisummary:: piscis.training.loss_fn piscis.training.train_epoch piscis.training.val_epoch piscis.training.train_model Module Contents --------------- .. py:function:: loss_fn(labels_pred: torch.Tensor, deltas_pred: torch.Tensor, labels: torch.Tensor, deltas: torch.Tensor, p: torch.Tensor, l2_loss_weight: float, max_distance: float, kernel_size: Sequence[int], temperature: float, epsilon: float) -> Tuple[torch.Tensor, Dict[str, float]] Computes the loss and metrics for a given batch. :Parameters: **labels_pred** : torch.Tensor Predicted labels. **deltas_pred** : torch.Tensor Predicted displacement vectors. **labels: torch.Tensor** Ground truth labels. **deltas** : torch.Tensor Ground truth displacement vectors. **p** : torch.Tensor Number of ground truth spots in each image. **l2_loss_weight** : float Weight for the masked L2 loss term in the overall loss function. **max_distance** : float Maximum distance for matching predicted and ground truth displacement vectors. **kernel_size** : Sequence[int], optional Kernel size of sum or max pooling operations. Default is (3, 3). **temperature** : float Temperature parameter. **epsilon** : float Small constant for numerical stability. :Returns: **loss** : torch.Tensor Overall loss value. **metrics** : Dict[str, float] Dictionary containing the values of individual loss terms and the overall loss. .. !! processed by numpydoc !! .. py:function:: train_epoch(model: piscis.models.spots.SpotsModel, dataloader: tqdm.tqdm, optimizer: torch.optim.Optimizer, l2_loss_weight: float, dilation_iterations: int, max_distance: float, temperature: float, epsilon: float, device: Optional[str]) -> Dict[str, float] Train the model for a single epoch. :Parameters: **model** : SpotsModel Model to be trained. **dataloader** : tqdm DataLoader for the training data. **optimizer** : torch.optim.Optimizer Optimizer for updating model parameters. **l2_loss_weight** : float Weight for the masked L2 loss term in the overall loss function. **dilation_iterations** : int Number of iterations to dilate ground truth labels **max_distance** : float Maximum distance for matching predicted and ground truth displacement vectors. **temperature** : float Temperature parameter for softmax. **epsilon** : float Small constant for numerical stability. **device** : Optional[str] Device for training. :Returns: **train_metrics** : Dict[str, float] Dictionary containing average training metrics for the epoch. .. !! processed by numpydoc !! .. py:function:: val_epoch(model: piscis.models.spots.SpotsModel, dataloader: torch.utils.data.DataLoader, l2_loss_weight: float, dilation_iterations: int, max_distance: float, temperature: float, epsilon: float, device: Optional[str]) -> Dict[str, float] Validate the model for a single epoch. :Parameters: **model** : SpotsModel Model to be validated. **dataloader** : torch.utils.data.DataLoader DataLoader for the validation data. **l2_loss_weight** : float Weight for the masked L2 loss term in the overall loss function. **dilation_iterations** : int Number of iterations to dilate ground truth labels **max_distance** : float Maximum distance for matching predicted and ground truth displacement vectors. **temperature** : float Temperature parameter for softmax. **epsilon** : float Small constant for numerical stability. **device** : Optional[str] Device for training. :Returns: **val_metrics** : Dict[str, float] Dictionary containing average validation metrics for the epoch. .. !! processed by numpydoc !! .. py:function:: train_model(model_name: str, dataset_path: Union[str, List[str], Dict[str, float]], initial_model_name: Optional[str] = None, adjustment: Optional[str] = 'standardize', input_size: Tuple[int, int] = (256, 256), random_seed: int = 0, batch_size: int = 4, num_workers: int = 0, learning_rate: float = 0.1, weight_decay: float = 1e-05, epochs: int = 500, warmup_fraction: float = 0.04, decay_fraction: float = 0.4, decay_transitions: int = 10, decay_factor: float = 0.5, l2_loss_weight: float = 0.1, dilation_iterations: int = 1, max_distance: float = 3.0, temperature: float = 0.05, epsilon: float = 1e-07, checkpoint_every: int = 10, device: Optional[str] = 'cuda') -> None Train a SpotsModel. :Parameters: **model_name** : str Model name. **dataset_path** : Union[str, List[str], Dict[str, float]] Path to a dataset, path to a directory containing multiple datasets, a list of multiple dataset paths, or a dictionary of multiple dataset paths and their corresponding sampling weights. If a directory of datasets or a list is provided, all datasets in the directory or list will be loaded and concatenated with equal weights. If a dictionary is provided, the datasets will be loaded and concatenated with the specified weights. **initial_model_name** : Optional[str], optional Name of an existing model to initialize the weights. Default is None. **adjustment** : Optional[str], optional Adjustment type applied to images. Supported types are 'normalize' and 'standardize'. Default is 'standardize'. **input_size** : Tuple[int, int], optional Input size used for training. Default is (256, 256). **random_seed** : int, optional Random seed used for initialization and training. Default is 0. **batch_size** : int, optional Batch size for training. Default is 4. **num_workers** : int, optional Number of workers for data loading. Default is 0. **learning_rate** : float, optional Learning rate for the optimizer. Default is 0.1. **weight_decay** : float, optional Strength of the weight decay regularization. Default is 1e-5. **epochs** : int, optional Number of epochs to train the model for. Default is 500. **warmup_fraction** : float, optional Fraction of epochs for learning rate warmup. Default is 0.04. **decay_fraction** : float, optional Fraction of epochs for learning rate decay. Default is 0.4. **decay_transitions** : int, optional Number of times to decay the learning rate. Default is 10. **decay_factor** : float, optional Multiplicative factor of each learning rate decay transition. Default is 0.5. **l2_loss_weight** : float, optional Weight for the masked L2 loss term in the overall loss function. Default is 0.1. **dilation_iterations** : int, optional Number of iterations to dilate ground truth labels to minimize class imbalance and misclassifications due to minor offsets. Default is 1. **max_distance** : float, optional Maximum distance for matching predicted and ground truth displacement vectors. Default is 3.0. **temperature** : float, optional Temperature parameter for softmax. Default is 0.05. **epsilon** : float, optional Small constant for numerical stability. Default is 1e-7. **checkpoint_every** : int, optional Number of epochs between saving model checkpoints. Default is 10. **device** : Optional[str], optional Device for training. Default is 'cuda'. :Raises: ValueError If warmup_fraction + decay_fraction is greater than 1. .. !! processed by numpydoc !!