YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
molmoact2: Real204 BEV memory
Private checkpoint repository for the authorized Real204 training campaign.
The 10k, 20k, 30k, 40k and 50k inference exports will appear on step-10000 through step-50000 branches only after completed training saves and native export/hash verification. This main branch currently contains no model weights.
Input: wrist-scan nvblox memory, fixed 80cm overhead camera, triangle RGB512, white background, robot base at image bottom; current wrist RGB. PointmapVLA also consumes BEV and wrist raycast XYZ256 in robot-base coordinates.
All 204 corrected demonstrations are training data. No held-out real-robot success rate is claimed. Source recordings and resume/optimizer states are retained on Draco. The inference implementation is under local review on deployment/inference; no GitHub push is authorized.