"Input dimension mis-match" in basic model?

[junpenglao]

I would not expect memory error with input matrix of this size - could you try casting all pandas table into numpy arrays? something like train_norm = train_norm.values