[ECCV 2026] Accelerating Text-to-Video Generation with Calibrated Sparse Attention -- Official Code Repository - View it on GitHub