Reset pytorch sequential model during cross validation

You could try the solution for this post where you just define new random weights and apply them to your model.