Fix hidden_size: 4096 -> 3584 to match Qwen2.5-Coder-7B-Instruct

Fix Phase 2: fusion layer processes text-only via learnable residual gate for gradient flow

Fix: register LLM as nn.Module submodule so optimizer finds LoRA params

Add GPU diagnostic script, fix architecture loading with low_cpu_mem_usage and sync

Day 3 COMPLETE: Full model architecture

feat: initial project scaffold for MINDI 1.5 Vision-Coder