{"slug":"openlair-openrlhf-training","source_name":"openlair/openrlhf-training","name":"Openlair/Openrlhf Training","description":"High-performance RLHF framework with Ray+vLLM acceleration. Use for PPO, GRPO, RLOO, DPO training of large models (7B-70B+). Built on Ray, vLLM, ZeRO-3. 2× faster than DeepSpeedChat with distributed architecture and GPU resource sharing.","version":1,"lift":{"pass_rate_delta_pts":50,"pass_rate_pct":100,"total_cases":22,"passed_cases":22,"tokens_delta_pct":89.5,"turns_delta_pct":0,"verdict":"pass","benchmark_model":"gemini-3.6-flash","grading_method":"judged","completed_at":"2026-08-07T19:51:37.837094+00:00"},"skill_score":1,"benchmark_models":[{"model":"gemini-3.6-flash","headline":true,"delta_pts":50,"with_pass_pct":100,"without_pass_pct":50,"tokens_delta_pct":89.5,"turns_delta_pct":0,"total_cases":22,"cases_aggregated":18,"verdict":"pass","never_hurt":true,"completed_at":"2026-08-07T19:51:37.837094+00:00","run_id":"36dc8b3f-6630-43b5-a5dd-514b6f19f774","latest_attempt":{"id":"36dc8b3f-6630-43b5-a5dd-514b6f19f774","runner_model":"gemini-3.6-flash","overall_verdict":"pass","completed_at":"2026-08-07T19:51:37.837094+00:00","gate":null},"version_number":1,"is_latest_version":true,"gate":null}],"trust":{"skill_safety":"passed","safety_status":"clean","intent_verdict":"safe","content_status":"clean","indexable":true},"license":"MIT","install_count":0,"total_uses":0,"total_users":0,"manifest_hash":"dbfd3f9cec5d3b1b12b3fd0fe1a1e56e3c60789694f916c5fbe85c55840e9a76","install_name":"openlair-openrlhf-training","version_id":"7c1f8a5c-21b8-46b7-87b1-bed15413ec70","body_sha256":"dbfd3f9cec5d3b1b12b3fd0fe1a1e56e3c60789694f916c5fbe85c55840e9a76","render_hash":"23f4c7e5c231739ad5b030960f45c6fa70d0b396e7c76b718f071cc54b65468c","raw_url":"https://app.decimal.ai/s/openlair-openrlhf-training/SKILL.md","scorecard_url":"https://app.decimal.ai/skills/openlair-openrlhf-training"}