You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

flagship-v17-speedloss

v1.6 recipe + speed-profile L1 (w_speed 0.5, pre-registered run 6, outcome B). CI-separated better ADE (0.2849 vs 0.3398) and speed MAE; reliance 1.1849. NOT a fix for the decel-ramp defect (primary gates failed). Registry section 1.11.

ckpt md5 e389f638cc2e7ac4d67bf57479936b7f. Trunk NOT included — this is the 2.11M step-readout head only; it requires flagship-v1arch-v2bal-30k. Trained/evaluated on nvidia PhysicalAI-AV; eval corpus physicalai-oodval q90 (official val split), 6,834 stride-1 windows.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support