Skip to content
A Keras inspired training utility for PyTorch
Branch: master
Clone or download
Type Name Latest commit message Commit time
Failed to load latest commit information.
examples Added mnist example Sep 6, 2018
flare Bug fix Sep 6, 2018
tests fix imports in test May 3, 2018
LICENSE Initial commit Jan 14, 2018 Update README with Guiding principles Sep 6, 2018 More info in Feb 14, 2018

Flare: PyTorch for everyday use

Beware, Flare is still in developmental stages !

Flare is a utility library that enables users to train their networks on PyTorch instantly. The API was inspired from Keras deep learning framework.

Currently Flare is designed to run on PyTorch >= 0.4.0

Guiding principles

  • "Everything should be made as simple as possible, but not simpler"
  • Intuitiveness Do not support complex usecases at the cost of intuitiveness of simpler tasks.


First, clone the repo using

then cd to the flare folder and run the install command cd flare sudo python install


import torch
from torch import nn

import flare
from flare import Trainer

input_1 = np.random.rand(10000,5)
input_2 = np.random.rand(10000,5)
targets = np.random.randint(0, 10, size=[10000,])

class linear_two_input(nn.Module):
    def __init__(self):
        super(linear_two_input, self).__init__()
        self.dense1 = nn.Linear(10, 64)
        self.dense2 = nn.Linear(64, 10)
    def forward(self, inputs):
        y =, dim=-1)
        y = self.dense1(y)
        y = self.dense2(y)
        return y

model = linear_two_input()

t = Trainer(model, nn.CrossEntropyLoss(), torch.optim.Adam(model.parameters()))
t.train([input_1, input_2], targets, validation_split=0.2, batch_size=128)

See MNIST example here

Why this name, Flare


You can’t perform that action at this time.