indRNN-pt

PyTorch implementation of Independently Recurrent Neural Networks by Shuai Li et al. (accepted to CVPR2018).

• Official source code in Theano and Lasagne and PyTorch
• TensorFlow version
• another PyTorch version

How to Run

test the trained model

on MNIST: python3 Sequential_task.py --test --test_model=./best_model_mnist.pt --dataset=mnist --log_folder=./test

train

1. on MNIST: python3 Sequential_task.py --dataset=mnist
2. on pMNIST: python3 Sequential_task.py --dataset=pmnist

Env

Python 3.6.6
pytorch 0.4.0
torchvision 0.2.1
cuda 8.0
numpy 1.14.5

Result

task	valid	test
sequential-mnist	98.98	98.80
p-mnist	94.22 (94.40 from Theano&Lasagne)	-
fashion-mnist	91.60	-