rnn_model.R contains possible error

[lahdeaho]

Environment info

MXNet version: 0.8

mx.model.init.iter.rnn

The following function has possible wrong variable to use in dim():

# Initialize the data iter
mx.model.init.iter.rnn <- function(X, y, batch.size, is.train) {
  if (is.MXDataIter(X)) return(X)
  shape <- dim(data)
  if (is.null(shape)) {
    num.data <- length(X)
  } else {
    ndim <- length(shape)
    num.data <- shape[[ndim]]
  }
  if (is.null(y)) {
    if (is.train) stop("Need to provide parameter y for training with R arrays.")
    y <- c(1:num.data) * 0
  }

  batch.size <- min(num.data, batch.size)

  return(mx.io.arrayiter(X, y, batch.size=batch.size, shuffle=is.train))
}

I think shape <- dim(data) should be shape <- dim(X) ?