pytorch-functional 0.3.0

Provides functional API for model creation in PyTorch.

Pytorch Functional

Pytorch Functional is a MIT licensed library that adds functional API for model creation to PyTorch.

Defining complex models in PyTorch requires creating classes. Defining models in tensorflow is easier. This makes it just as easy in PyTorch.

Features:

• Small extension to PyTorch
• No dependencies besides PyTorch
• Produces models entirely compatible with PyTorch
• Reduces the amount of code that you need to write
• Works well with complex architectures
• Adds no overhead

Example

To create a functional model, call a placeholder with the layer as an argument. This will return another placeholder, which you can use.

>>> from torch import nn
>>> from pytorch_functional import Input, FunctionalModel
>>> inputs = Input(shape=(1, 28, 28))
>>> x = inputs(nn.Flatten())
>>> outputs = x(nn.Linear(x.shape[1], 10))(nn.ReLU())
>>> model = FunctionalModel(inputs, outputs)
>>> model
FunctionalModel(
  (module000_depth001): Flatten(start_dim=1, end_dim=-1)
  (module001_depth002): Linear(in_features=784, out_features=10, bias=True)
  (module002_depth003): ReLU()
)

See more examples in Quick Start.

New in 0.3.0:

In the new API you can create functional model just like in tensorflow, by calling the layer with a placeholder as an argument. Works with multiple arguments as well! You can mix new and old API.

>>> from torch import nn
>>> from pytorch_functional import Input, FunctionalModel

>>> import pytorch_functional.enable_experimental_api  # JUST IMPORT THIS TO ENABLE NEW API

>>> inputs = Input(shape=(1, 28, 28))
>>> x = nn.Flatten()(inputs)
>>> x = nn.Linear(x.shape[1], 10)(x)
>>> outputs = nn.ReLU()(x)
>>> model = FunctionalModel(inputs, outputs)
>>> model
FunctionalModel(
  (module000_depth001): Flatten(start_dim=1, end_dim=-1)
  (module001_depth002): Linear(in_features=784, out_features=10, bias=True)
  (module002_depth003): ReLU()
)

Installation

Install easily with pip:

pip install pytorch-functional

Links

• See Documentation
• See on GitHub
• See on PyPI