New VLM Post-Training Layer Scaling: No more Forgetting #ai | code_your_own_AI | Podwise