CoolFace
Modelpublic

aravdhoot/risk-spec-specv3local-q27-rse2declnotraceq27-qwen3.8-27b-hp500-r32-s0-20260916

sourceHugging Faceupdated 11d agoView on Hugging Face
0likes51downloads
Model Card

risk-spec-specv3local-q27-rse2declnotraceq27-qwen3.8-27b-hp500-r32-s0-20260916

LoRA adapter (risk-spec local line). Provenance:

  • —arm: rse2declnotraceq27
  • —constitution: rse2decl_notrace
  • —constitution_sha256_12: 11a1434bd1b6
  • —base_model: Qwen/Qwen3.8-27B
  • —recipe: {"groupsize": 4, "groupsperbatch": 32, "lorarank": 32, "lr": 0.0001, "maxsteps": 500, "model": "Qwen/Qwen3.8-27B", "modelrevision": "1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0", "prompts": "src/constitution/prompts/riskseedsv2.jsonl", "renderer": "qwen35disablethinking", "saveevery": 20, "wildchat_seed": 12345}
  • —final_teacher_kl: 0.010232606845718162
  • —repo_commit: 9821f3c
  • —stack: local