v0.6.0 - prerelease

Version 0.6.0

• Implemented Phenix style sigma A weighting in the Maximum likelihood target (New default for refinement)
• Added sigma A estimation to scaler objects
• Added collection version of the sigma A target
• Deprecated internal coordinates
• Renamed Maximum likelihood target to Rice target, Sigma A target to Maximum likelihood target.
• Moved alignment into experimental
• Fixed Fast rotation function, current blocker on alignment is the rescorign function
• Moved kinetic to experimental
• Added monolithic refinement under experimental
• Changed reflection data acessor api to use property style dataselection
• Added benchmark to the paper folder were we refine from alphafold start coordinates

0.5.3.3

Version 0.5.3.3

• Fixed U_aniso parametrization and line search instability during refinement with anisotropic b-factor
• Fixed kinetic module import
• Set default similarity weight in difference refinement to 0

0.5.3.2

Version 0.5.3.2

• Added 10GB Gram requirement for default gpu device selection
• Slaved cli device detection to the default device selection
• Fixed device mismatch crash on CUDA/MPS when the VDW pair list was refreshed mid-refinement: the maintenance-triggered rebuild now migrates the fresh VDW pair list, hydrogen topology, and exclusion hash to the model device (PR #19)

0.5.3.1

Version 0.5.3.1

• Fixed problem where TorchRef defaults to old GPUs and crashes, now checking if GPU is actually usable, before setting default device to CUDA, if not, it will default to CPU and print a warning.

0.5.3

Version 0.5.3

• Fixed dtype inconsistencies and centralized dtype handling
• Centralized default device handling
• Added compatability with Metal performance shaders

0.5.2

Version 0.5.2

• Cleaned up build solvent mask calculation
• Added DeviceMixin for centralizing device handling accross all classes

0.5.1

Version 0.5.1

• Fixed masked tensor problem under torch 2.9
• Added Link parsing and restraint as bond
• Reduced memory usage during neighbor search for VDW target
• Separated out loss functions from targets, logic moved to base/targets
• Added Triton kernels with analytic backward for all four xray Targets and most other Targets
• Cached XrayTarget.get_data constants across closures
• Replaced slow tensor[indices] backwards (sort + dedup scatter) with index_add_ in the symmetry extractor, scaler bin gathers, and MixedTensor; skip the indexing in get_iso / get_aniso when it covers all atoms

Paper version

0.5.0

cleaned up broken tests, added ninja as a dependency

0.4.3 - CLI args rework

Reworked and standardised cli argw

0.4.2 hotfix

Hotfix for v0.4