{"slug":"openlair-optimizing-attention-flash","source_name":"openlair/optimizing-attention-flash","name":"Openlair/Optimizing Attention Flash","description":"Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Use when training/running transformers with long sequences (>512 tokens), encountering GPU memory issues with attention, or need faster inference. Supports PyTorch native SDPA, flash-attn library, H100 FP8, and sliding window attention.","version":1,"lift":{"pass_rate_delta_pts":9.09,"pass_rate_pct":86.4,"total_cases":22,"passed_cases":19,"tokens_delta_pct":146.5,"turns_delta_pct":0,"verdict":"mixed","benchmark_model":"gemini-3.6-flash","grading_method":"judged","completed_at":"2026-08-07T19:45:58.127287+00:00"},"skill_score":0.8636,"benchmark_models":[{"model":"gemini-3.6-flash","headline":true,"delta_pts":9.09,"with_pass_pct":86.4,"without_pass_pct":77.3,"tokens_delta_pct":146.5,"turns_delta_pct":0,"total_cases":22,"cases_aggregated":22,"verdict":"mixed","never_hurt":true,"completed_at":"2026-08-07T19:45:58.127287+00:00","run_id":"a5e826b3-8a10-43f8-8318-968170175f1d","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,"manifest_hash":"5e65c4c9bfece417ea1de6da2d51044001ef30b33891f0e767fa66d75e6e08c5","raw_url":"https://app.decimal.ai/s/openlair-optimizing-attention-flash/SKILL.md","scorecard_url":"https://app.decimal.ai/skills/openlair-optimizing-attention-flash"}