Input question

[BransonMarvin]

Hello,

Thanks for sharing your implementation, I have one question. Is it possible to use a single tensor as input (instead of a table of tensors) ? If it is, then SeqLen x BatchSize x Featsize ( Featsize = NumFeatMaps x Width x Height) is the correct dimension order?

Thanks!

[vikram-gupta]

Yes, you can. Wrap the ConvLSTM in a sequencer.