DROID-models (RoboColosseum)

These are vision-language-action (VLA) models fine-tuned on DROID (Franka) manipulation tasks for the RoboColosseum benchmark. Each VLA is fine-tuned from its public base checkpoint with the same recipe and the same LRs as BimanualYAM-models, on 2 GPUs, for 5 epochs. Each folder holds the final-epoch inference weights plus the small config files needed to load the model.

Layout: <model>/<task>/

Model	Base checkpoint	MugOnCoaster	CloseLaptop
gr00t/ GR00T N1.7	nvidia/GR00T-N1.7-3B	✅	✅
pi05/ π0.5	gs://openpi-assets/checkpoints/pi05_base	✅	✅
g05/ G0.5	OpenGalaxea/G05 (g05-base)	✅	✅
molmoact2/ MolmoAct2	allenai/MolmoAct2	✅	✅
lingbot-vla-v2/ LingBot-VLA v2 6B	robbyant/lingbot-vla-v2-6b	✅	✅

✅ uploaded · ⏳ training or queued. Each <model>/<task>/README.md gives the recipe, the files, and how to load the model.

Tasks (LeRobot v3.0, 30 fps)

Task	Dataset	Instruction	Episodes	Frames
MugOnCoaster	aarontung/droid-mug-on-coaster	"mug on coaster"	100	41,227
CloseLaptop	aarontung/droid-close-laptop	"close laptop"	100	27,649

Common setup:

• Cameras: left exterior, right exterior and wrist (all 360x640).
• State and action: 8-D absolute joint positions [Franka arm 7, gripper 1].
• Training: fp32 master weights with bf16 compute, on 2 GPUs per run (H100 or H200; see each model card).

Recipes (same for every task)

Model	Cameras	Trained parts	GBS	Peak LR
GR00T N1.7	left, wrist	Full: VLM backbone (LLM + vision) + action head	32	1e-4, cosine, 5% warmup
π0.5	left, wrist	Full (openpi), EMA 0.99	32	2.5e-5, cosine to 2.5e-6, 1k warmup
G0.5	left, wrist	Full (vision LR x0.1), FSDP	16	4e-5, cosine, 200 warmup, wd 0.03
MolmoAct2	left, right, wrist	Full	64	LLM 2e-5, ViT 1e-5, connector 1e-5, action expert 1e-4
LingBot-VLA v2	left, right, wrist	Full	32	1.25e-5, constant, Muon

Training logs are in the W&B project shaileshxml-nus/RoboColosseum.