Bookend Phase 4 — Training Checkpoints
LoRA adapters from DPO and SFT runs on King CXXIV.
Layout
| Path | Description |
|---|---|
dpo/checkpoint-8/ |
DPO step 8 |
dpo/checkpoint-16/ |
DPO step 16 (final) |
dpo/final/ |
Final DPO adapter + run metadata |
sft/checkpoint-16/ |
SFT step 16 |
sft/checkpoint-32/ |
SFT step 32 (final) |
sft/final/ |
Final SFT adapter + run metadata |
Full merged weights (repacked for Albedo submit)
Base: local_king/king_cxxiv (King CXXIV). Merge with scripts/train/merge_lora.py.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support