YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
DAR checkpoints (internal code name: lskv)
Fixed-window long-short KV cache attention checkpoints (Pythia 70M/160M/410M
- OLMo-2-1B/130M variants), including paper-table results, multi-seed
robustness runs, the MLA-style "DAR-abs" absorbed-inference line, and the
current-direction
lskv_prefill_ffnimprovement.
Full documentation, eval results, and eval commands: https://github.com/zywang624/dar-tar-release-docs (private -- request access if needed)
See DAR.md in that repo for the complete per-checkpoint inventory and
known gaps.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support