Add E6-E12 training diagnostic log (Ninja error + E7 training output)

Add model card README

Upload runs_ppo/E7_vlm: E7_vlm training run (binary reward, no force obs, +VLM props)

Upload runs_ppo/E5_baseline: E5_baseline training run (binary reward, no force obs, no VLM)

Upload ckpt: Baseline pre-trained checkpoints from original paper

initial commit