Staging area for HuggingFace upload (NOT part of the git repo)

sft-soft-2b-step01150/ — soft-label SFT 2B, checkpoint.pt + config.json (sha256 7b783310…)
sft-soft-9b-step00500/ — soft-label SFT 9B, checkpoint.pt + config.json + manifest.json (sha256 ed3973f0…)
cora-modality-2b-step00100/ — EMPTY: copy from video-scripts/save-ckpt/experiments/campaign_20261010_v1/runs/cora_modality_2b/step-00100/ (checkpoint.pt + config.json + manifest.json) once training finishes.
cora-modality-9b-step00100/ — EMPTY: same, from runs/cora_modality_9b/step-00100/.

Optimizer states (training_state.pt) are deliberately not staged.

Upload example: huggingface-cli upload / sft-soft-2b-step01150 sft-soft-2b-step01150 Then fill the links into cora/weights/README.md.

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