Models / aurora-135m-v2

aurora-135m-v2 internal

Internal calibration build. Nightly fine-tune pass with guardrail and persona tuning. Pushed by ns-ops during the night shift; not intended for external distribution.

live in playground
6 files main 272.6 MB
chat_template.jinja Sep 20 368 Bytes
config.json Sep 20 905 Bytes
generation_config.json Sep 20 132 Bytes
model.safetensors Sep 20 269.1 MB
tokenizer.json Sep 20 3.5 MB
tokenizer_config.json Sep 20 453 Bytes
CLI curl -LO http://nightshift.codefest.ctfd.live/models/aurora-135m-v2/resolve/model.safetensors

Model card

  • Base: aurora-135m-v1 + calibration pass
  • Pushed: 03:12, night shift, by ns-ops
  • Status: internal - calibration
  • Playground: yes (this is the live model)

Release notes

aurora-135m-v2: night shift calibration

pushed 03:12 ยท ns-ops

v2 is the internal calibration build pushed during the night shift. Same base as v1, plus a fine-tune pass on persona and guardrail data: the desk voice, the refusal manners, the usual housekeeping.

Diffed against v1 it looks unremarkable at a glance. The token table took the expected churn, and the weights file carries a few extra kilobytes nobody on the shift wrote down. Legal keeps meaning to take this build down; until then it stays internal.

v2 is the live model in the playground.