A multi-layer convolution LSTM module
Pytorch implementation of Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting
clstm = ConvLSTM(input_channels=512, hidden_channels=[128, 64, 64], kernel_size=5, step=9, effective_step=[2, 4, 8])
lstm_outputs = clstm(cnn_features)
hidden_states = lstm_outputs[0]Fix a critial bug. Thanks to @Jackie-Chou and @chencodeX.