seqgra.learner.torch.torchhelper module

MIT - CSAIL - Gifford Lab - seqgra

PyTorch learner helper class

@author: Konstantin Krismer

class TorchHelper

Bases: object

MULTIPLE_REGRESSION_LOSSES: FrozenSet[str] = frozenset({'l1loss', 'mseloss', 'smoothl1loss'})

MULTIVARIATE_REGRESSION_LOSSES: FrozenSet[str] = frozenset({'l1loss', 'mseloss', 'smoothl1loss'})

MULTI_CLASS_CLASSIFICATION_LOSSES: FrozenSet[str] = frozenset({'cosineembeddingloss', 'crossentropyloss', 'hingeembeddingloss', 'kldivloss', 'nllloss'})

MULTI_LABEL_CLASSIFICATION_LOSSES: FrozenSet[str] = frozenset({'bceloss', 'bcewithlogitsloss'})

static create_model(learner: seqgra.learner.learner.Learner) → None

static evaluate_model(learner: seqgra.learner.learner.Learner, dataset: torch.utils.data.dataset.Dataset, output_layer_activation_function: Optional[str] = None)

static get_best_model_file_name(best_model_dir: str) → str

static get_loss(loss_hyperparameters)

static get_metrics(learner: seqgra.learner.learner.Learner, output_layer_activation_function: Optional[str] = None)

static get_num_params(learner: seqgra.learner.learner.Learner) → seqgra.schema.ModelSize

static get_optimizer(optimizer_hyperparameters, model_parameters)

static load_model(learner: seqgra.learner.learner.Learner, file_name: Optional[str] = None) → None

static predict(learner: seqgra.learner.learner.Learner, dataset: torch.utils.data.dataset.Dataset, output_layer_activation_function: Optional[str] = None)

This is the forward calculation from x to y :returns: Output tensor with the computed logits. :rtype: softmax_linear

static print_model_summary(learner: seqgra.learner.learner.Learner)

static save_model(learner: seqgra.learner.learner.Learner, file_name: Optional[str] = None) → None

static set_seed(learner: seqgra.learner.learner.Learner) → None

static train_model(learner: seqgra.learner.learner.Learner, training_dataset: torch.utils.data.dataset.Dataset, validation_dataset: torch.utils.data.dataset.Dataset, output_layer_activation_function: Optional[str] = None, silent: bool = False) → None

static train_model_basic(learner: seqgra.learner.learner.Learner, training_dataset: torch.utils.data.dataset.Dataset, validation_dataset: torch.utils.data.dataset.Dataset, output_layer_activation_function: Optional[str] = None, silent: bool = False) → None

static write_session_info(learner: seqgra.learner.learner.Learner) → None