How to handle very short sequences in LSTM Network

Hello Matlab Community,

I am using MATLAB 2018b I am trying to apply the sequence to sequence regression network on a dataset with very short sequences.

I basically followed the 'Sequence-to-Sequence Regression Using Deep Learning' tutorial which is given by MATLAB.

When I apply this tutorial to my dataset it only returns NaN as a prediction.

And the graph also suggests that it stops calculating after some iterations.

this is the verbose output:

|=========================================================================|

| Epoch | Iteration | Time Elapsed | Mini-batch | Mini-batch | Base Learning |

| | | (hh:mm:ss) | RMSE | Loss | Rate |

|======================================================================|

| 1 | 1 | 00:00:01 | 65.11 | 2119.5 | 0.0100 |

| 5 | 50 | 00:00:04 | NaN | NaN | 0.0100 |

| 9 | 100 | 00:00:06 | NaN | NaN | 0.0100 |

| 13 | 150 | 00:00:07 | NaN | NaN | 0.0100 |

| 17 | 200 | 00:00:09 | NaN | NaN | 0.0100 |

| 20 | 240 | 00:00:15 | NaN | NaN | 0.0100 |

|=======================================================================|

I'd be super happy if anyone had some suggestions to solve that issue!

I already tried looping over the single short iterations - which let to endlessly poping up windows and a server crash.