AsyncDataSetIterator - Manually destroying ADSI workspace

ND4J
1

Finally set up my lstm using the datasetiterator class. I have this message repeating on my screen. It seems to be bad.

Is it?

AsyncDataSetIterator - Manually destroying ADSI workspace

2

temporarily fixed by changing the commented code to the un-commented. If anyone is wondering I have the sp500 stock 3 year history downloaded and I’m trying to overfit to a single stock in order to prove t myself that its working

    for (int i = 0; i < 500; i++) {


        while (iter.hasNext()) {

            DataSet d = iter.next();

            network.fit(d);

        }
        iter.reset();

    }

    //network.fit(iter, 200);
3

That message is not an error. You see it because when you call .fit, the dataset iterator is asynchronously read, so it can prepare the next dataset while the current one is being used for training.

By running things the way you “fixed” it, you don’t get that benefit.

If you want that message to go away, and keep the benefit of asynchronous pre-loading, you can configure your logging library to either ignore logging from that class, or elevate the logging level to a higher level.