mlnet prediction clarification about re training data

Question

Hi,

I am using the automatic model selector command line (mlnet) with a specific scenario:

for a 7000 lines of data:
I am training on 5000 data rows, validating on the next 1000 data rows and executing (predicting) on the next 1000 data rows.

With the mlnet command line I get the model that was built on the first 5000 data rows. How can I retrain in order to incorporate the 1000 validation rows into the actual prediction phase? which is actually - the future (no data yet).

In particular, my model is using LightGbmRegression which was not to be found at this list.

If re-training is not feasible, how can I incorporate all of the current available data to the prediction process. currently it is not being used and I might miss some valuable data.

Thanks :)