Does Diffusion Policy run on Trossen WidowX AI?
works with caveats
Capabilities
| Capability | State | Notes | Evidence |
|---|---|---|---|
| Train / fine-tune | works | ev-lerobot-issue-1811 | |
| Real-robot inference | works | ev-lerobot-issue-1811 | |
| Simulation inference | no verified evidence | — | |
| Teleop + record | no verified evidence | — |
Caveats
- Community-reported only; not documented upstream or maintainer-tested.
- The single report concerns a policy deliberately overfit to one pick-and-place trajectory, not a generalizing policy.
Evidence 1 source; the cell shows the highest tier
- T3 · community reportedgithub-issuesource 2025-08-29verified 2026-10-24 (1d ago)
A user fine-tuning on a Trossen WidowX AI pick-and-place dataset reports that ACT and Diffusion Policy overfit to one trajectory and work when rolled out on the physical robot (the issue itself is about π0-FAST failing to converge).
“I was able to get ACT and diffusion policy to overfit on the same dataset and when rolled out on the physical robot works perfectly unlike pi0Fast”
https://github.com/huggingface/lerobot/issues/1811
Embed this cell
Badge for READMEs and model cards. It always shows the current state.
[](https://caniembody.pages.dev/compat/trossen-widowx/diffusion)<iframe src="https://caniembody.pages.dev/embed/trossen-widowx/diffusion" width="380" height="132" style="border:0" title="Diffusion Policy on Trossen WidowX AI | CanIEmbody"></iframe>History
- 2026-10-25 fecccd1 Initial release: robot policy compatibility matrix · Danilo Znamerovszkij