Trainability issues in quantum policy gradients

Re-assign community