CPU usage decrease over time

[Racines]

When I use a lot the model.Prediction() the usage of CPU decrease over time, and the time to wait between each prediction increase over time.

===================================================
average millisecond for 1000 predict: 56,54410229999994
===================================================
average millisecond for 1000 predict: 63,12716950000007
===================================================
average millisecond for 1000 predict: 64,76692500000009
===================================================
average millisecond for 1000 predict: 71,62019349999997
===================================================
average millisecond for 1000 predict: 78,61800759999998

This is a critical issue for me because I need to compute a lot of predictions for my training. Any idea on how to fix this?

Here is the code I made to test this problem. I use the same input each time, so it's related to the predict method.

List<double> elapsedList = new List<double>();
for (int j = 0; j < 1_000; j++)
{
        elapsedList.Clear();
	for (int i = 0; i < 1_000; i++)
	{
		var ss = DateTime.Now;
		model.Predict(input, verbose: 0);

		elapsedList.Add((DateTime.Now - ss).TotalMilliseconds);
	}

	var average = elapsedList.Average();
	Console.WriteLine($"===================================================");
	Console.WriteLine($"average millisecond for 1000 predict: {average}");
}