happyjin / ConvGRU-pytorch

Convolutional GRU

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

ConvLSTM_pytorch

This file contains the implementation of Convolutional LSTM in PyTorch

How to Use

The ConvGRU module derives from nn.Module so it can be used as any other PyTorch module.

The ConvGRU class supports an arbitrary number of stacked hidden layers in GRU. In this case, it can be specified the hidden dimension (that is, the number of channels) and the kernel size of each layer. In the case more layers are present but a single value is provided, this is replicated for all the layers. For example, in the following snippet each of the three layers has a different hidden dimension but the same kernel size.

Example usage:

# set CUDA device
os.environ["CUDA_VISIBLE_DEVICES"] = "3"

# detect if CUDA is available or not
use_gpu = torch.cuda.is_available()
if use_gpu:
    dtype = torch.cuda.FloatTensor # computation in GPU
else:
    dtype = torch.FloatTensor

height = width = 6
channels = 256
hidden_dim = [32, 64]
kernel_size = (3,3) # for two stacked hidden layers with same kernel_size or write as [(3,3), (3,3)]
num_layers = 2 # number of stacked hidden layer
model = ConvGRU(input_size=(height, width),
                input_dim=channels,
                hidden_dim=hidden_dim,
                kernel_size=kernel_size,
                num_layers=num_layers,
                dtype=dtype,
                batch_first=True,
                bias = True,
                return_all_layers = False)

batch_size = 1
time_steps = 1
input_tensor = torch.rand(batch_size, time_steps, channels, height, width)  # (b,t,c,h,w)
layer_output_list, last_state_list = model(input_tensor)

Disclaimer

This is still a work in progress and is far from being perfect: if you find any bug please don't hesitate to open an issue.

License

ConvLSTM_pytorch is released under the MIT License (refer to the LICENSE file for details).

Acknowledgment

This repo borrows some codes from

About

Convolutional GRU

License:MIT License


Languages

Language:Python 100.0%