MolmoAct 2 artifacts have been downloaded 400K+ times in under 1 month.
Today we're opening up the full code & training data: everything you need to fine-tune or build on our fully open robotics foundation model. 🧵
What's now open alongside the model:
1️⃣ Fine-tuning scripts
2️⃣ Every dataset used to train MolmoAct 2
3️⃣ All of our evaluation rollouts
4️⃣ Training recipe for the open source MolmoAct 2 tokenizer
MolmoAct 2 now also officially supports @HuggingFace’s @LeRobotHF platform.
Teams already working in the LeRobot ecosystem can drop the model into their existing setup without retooling.
🤗 Learn more: https://t.co/RLvTLQYHzo
@huggingface @LeRobotHF Open robotics gets stronger when researchers can evaluate models like MolmoAct 2 themselves.
Try it on new robots and tasks + tell us what you discover.
💻 Code: https://t.co/xtGlXyh5oT
📝 Learn more: https://t.co/cZLnrODfEL
fine-tuning — Taking a trained model and training it a bit more on your own examples so it gets better at one specific job.
eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
Why it matters
The full training data, fine-tuningTaking a trained model and training it a bit more on your own examples so it gets better at one specific job.Full definition → scripts, and tokenizer recipe for MolmoAct 2 are now public, and the model drops into existing LeRobot setups without retooling, so teams can reproduce or adapt it rather than treating it as fixed weights.
