2026.0.2b0

What's Changed

• [ENH] Add comprehensive input validation with custom exception handling by @Leguark in #68
• [ENH] Add per-stack interpolation options override support by @Leguark in #69
• [CLN] Remove commented-out code in tensor and stack structures by @Leguark in #70
• [FIX] Stabilize autograd graph, sigmoid overflow, and dual contouring vertex solve by @Leguark in #71

Full Changelog: v2026.0.1b0...v2026.0.2b0

2026.0.1b0

What's Changed

• [ENH] Improve fault-aware vertex averaging in dual contouring with geological relationship handling by @Leguark in #56
• [REFACTOR] Add sequential kernel data stacking with GPU memory optimization strategies by @Leguark in #57
• [FIX] Use sequential kernel data building instead of parallel version by @Leguark in #58
• [OPT] Optimize PyTorch backend with compiled kernels and GPU memory management by @Leguark in #44
• [ENH] Enable PyKeOps with improved CG solver and PyTorch threading fixes by @Leguark in #59
• [ENH] Add PyKeOps fallback mechanism and improve tensor casting for kernel operations by @Leguark in #60
• [REFACTOR] Remove interpolate_scalar_field function and inline weight computation by @Leguark in #61
• [OPT] Optimize dual contouring with searchsorted algorithm and configurable multithreading by @Leguark in #62
• [OPT] Enable concurrent processing and optimize PyTorch tensor operations by @Leguark in #63
• [ENH] Add PyKeOps environment variable support and parallelize segmentation with GPU streams by @Leguark in #64
• [OPT] Add memory-aware chunking and optimize PyTorch dual contouring for VRAM efficiency by @Leguark in #65
• [CLN] Remove unused imports and optimize import placement by @Leguark in #66
• [REFACTOR] Remove legacy dual contouring code and simplify octree data structure by @Leguark in #67

Full Changelog: v2026.0.1a0...v2026.0.1b0

2526.0.1a0

What's Changed

• [REFACTOR] Reorganize dual contouring vertex overlap logic into separate modules by @Leguark in #49
• [OPT] V1 half the time by @Leguark in #51
• [CLN] Remove unused experimental watertight dual contouring code by @Leguark in #52
• [WIP] Prepare for pykeops ranges by @Leguark in #53
• [WIP] Towards split evaluation options by @Leguark in #54
• [ENH] Parallelize solver input preparation using ThreadPoolExecutor by @Leguark in #55

Full Changelog: v2025.3.1a3...v2526.0.1a0

2026.0.1a0

What's Changed

• [REFACTOR] Reorganize dual contouring vertex overlap logic into separate modules by @Leguark in #49
• [OPT] V1 half the time by @Leguark in #51
• [CLN] Remove unused experimental watertight dual contouring code by @Leguark in #52
• [WIP] Prepare for pykeops ranges by @Leguark in #53
• [WIP] Towards split evaluation options by @Leguark in #54
• [ENH] Parallelize solver input preparation using ThreadPoolExecutor by @Leguark in #55

Full Changelog: v2025.3.1a3...v2026.0.1a0

2025.3.1a3

Full Changelog: v2025.3.1a2...v2025.3.1a3

2025.3.1a2

What's Changed

• [REFACTOR] Generalize dual contouring triangulation for arbitrary grid shapes and remove legacy code by @Leguark in #43
• [FIX] Clear weight cache and handle non-aligned arrays in tensor operations by @Leguark in #45
• [ENH] Improve fault handling in dual contouring with arbitrary grid shapes and conservative triangle removal by @Leguark in #46
• [ENH] Add finite fault module with ellipsoid-based slip tapering and projection by @Leguark in #47
• [ENH] Implement configurable dual contouring vertex overlap with watertight mesh support by @Leguark in #48

Full Changelog: v2025.3.1a0...v2025.3.1a2

2025.3.1a0

What's Changed

• [FIX] Correct vertex index calculation in meshes_to_subsurface method | GEN-13490 by @Leguark in #32
• [REFACTOR] Optimize dual contouring gradient handling and normal correction | GEN-13490 by @Leguark in #33
• [REFACTOR] Modularize triangulation code with helper functions for improved readability | GEN-13490 by @Leguark in #34
• [ENH] Implement surface-based dual contouring with optimized triangulation by @Leguark in #35
• [REFACTOR] Optimize dual contouring vertex generation and add vertex overlap configuration by @Leguark in #36
• [FIX] Fix tensor concatenation to handle mixed numpy/torch arrays by @Leguark in #37
• [ENH] Add geophysics module with gravity modeling and magnetic implementation plan by @Leguark in #38
• [ENH] Add magnetic forward modeling for Total Magnetic Intensity (TMI) calculation by @Leguark in #39
• [REFACTOR] Restructure geophysics input classes for better magnetics and gravity handling by @Leguark in #41

Full Changelog: v2025.3.0a2...v2025.3.1a0

2025.3.0a2

What's Changed

• [REFACTOR] Move dual contouring code to modules directory and improve mesh consistency and water tight prot by @Leguark in #28
• [ENH] Implement fault vertex overlap handling in dual contouring by @Leguark in #29
• [DOC] Add dual contouring module overview and highlight risks, improvements, and optimizations by @Leguark in #30
• [ENH] Add gradient computation control and improve memory management by @Leguark in #31

Full Changelog: v2025.3.0a1...v2025.3.0a2

2025.3.0a1

Full Changelog: v2025.3.0a0...v2025.3.0a1

2025.3.0a0

What's Changed

• [ENH] Improve CG solver and add weights caching for better performance by @Leguark in #17
• [FIX] Separate pykeops_enabled flag from use_pykeops for better control by @Leguark in #18
• [ENH] Add transform support to meshes_to_subsurface method by @Leguark in #19
• [!!BUG] Chunking with faults were creating rogue voxels | GEN-13299 by @Leguark in #20
• [ENH] Add PyTorch backend support for dual contouring and octree operations by @Leguark in #21
• [FIX] Fix PyKeops tensor operations and ensure contiguous arrays by @Leguark in #22
• [FIX] Use timestamp for cache key instead of input parameters by @Leguark in #23
• [REFACTOR] Use center grid for dual contouring instead of corners grid by @Leguark in #24
• [ENH] Refactor dual contouring code with improved documentation and modular functions by @Leguark in #25
• [ENH] Refactor dual contouring with parallel and sequential triangulation modules by @Leguark in #26
• [FIX] Reuse weights calculation instead of raising error on hash mismatch by @Leguark in #27

Full Changelog: v2025.2.0...v2025.3.0a0