capreolus.trainer

Submodules

• capreolus.trainer.pytorch
• capreolus.trainer.tensorflow

Package Contents

Classes

Trainer

Base class for Trainer modules. The purpose of a Trainer is to train a Reranker module and use it to make predictions. Capreolus provides two trainers: PytorchTrainer and TensorFlowTrainer

Attributes

logger

capreolus.trainer.logger

class capreolus.trainer.Trainer(config=None, provide=None, share_dependency_objects=False, build=True)

Bases: capreolus.ModuleBase

Base class for Trainer modules. The purpose of a Trainer is to train a Reranker module and use it to make predictions. Capreolus provides two trainers: PytorchTrainer and TensorFlowTrainer

Modules should provide:

• a train method that trains a reranker on training and dev (validation) data

• a predict method that uses a reranker to make predictions on data

property n_batch_per_iter

module_type = 'trainer'

requires_random_seed = True

dependencies

static load_loss_file(fn)

Loads loss history from fn

Parameters

fn (Path) – path to a loss.txt file

Returns

a list of losses ordered by iterations

static load_metric(fn)

static load_best_metric(fn, metric)

static write_to_loss_file(fn, losses)

static write_to_metric_file(fn, metrics)

static exhaust_used_train_data(train_data_generator, n_batch_to_exhaust)

static get_paths_for_early_stopping(train_output_path, dev_output_path)

change_lr(step, lr)

Apply warm up or decay depending on the current epoch

lr_multiplier(step)