PyTorch Wrapper

PyTorch Wrapper is a library that provides a systematic and extensible way to build, train, evaluate, and tune deep learning models using PyTorch. It also provides several ready to use modules and functions for fast model development.

Branch	Build	Coverage
Master
Develop

Installation

From PyPI

pip install pytorch-wrapper

From Source

git clone https://github.com/jkoutsikakis/pytorch-wrapper.git
cd pytorch-wrapper
pip install .

Basic abstract usage pattern

import torch
import pytorch_wrapper as pw

train_dataloader = ...
val_dataloader = ...
dev_dataloader = ...

evaluators = { 'acc': pw.evaluators.AccuracyEvaluator(), ... }
loss_wrapper = pw.loss_wrappers.GenericPointWiseLossWrapper(torch.nn.BCEWithLogitsLoss())

model = ...

system = pw.System(model=model, device=torch.device('cuda'))

optimizer = torch.optim.Adam(system.model.parameters())

system.train(
    loss_wrapper,
    optimizer,
    train_data_loader=train_dataloader,
    evaluators=evaluators,
    evaluation_data_loaders={'val': val_dataloader},
    callbacks=[
        pw.training_callbacks.EarlyStoppingCriterionCallback(
            patience=3,
            evaluation_data_loader_key='val',
            evaluator_key='acc',
            tmp_best_state_filepath='current_best.weights'
        )
    ]
)

results = system.evaluate(dev_dataloader, evaluators)

predictions = system.predict(dev_dataloader)

system.save_model_state('model.weights')
system.load_model_state('model.weights')

Docs & Examples

The docs can be found here.

There are also the following examples in notebook format:

1. Two Spiral Task
2. Image Classification Task
3. Tuning Image Classifier
4. Text Classification Task
5. Token Classification Task
6. Text Classification Task using BERT
7. Custom Callback
8. Custom Loss Wrapper
9. Custom Evaluator