v3 fix: batch=1, grad_checkpoint, no eval (OOM fix)

v3 training script (951 examples, eval split)

GGUF conversion with cmake install

Add GGUF conversion script

Add consciousness v2 training script (fixed)

Upload train_consciousness_v2.py with huggingface_hub

Upload train_consciousness_agent.py with huggingface_hub

fix: reduce memory for T4 (max_length=512, no eval)

Add train_orchestrator_agent.py

Add train_memory_agent.py

initial commit