SIRF v3.10.1 released (including v3.10.0 updates)
22 Jun 2026
We are pleased to announce the release of SIRF v3.10.1, which consolidates the improvements introduced in v3.10.0 with additional fixes, testing enhancements, and updates to the build infrastructure. This release introduces the first wrapper to PyTorch which will enable the development of machine/deep learning pipelines and continues to strengthen robustness, usability, and interoperability across the SIRF ecosystem.
Key Updates
v3.10.1 and v3.10.0
- Added initial PyTorch wrappers and examples enabling AI pipelines, check torch/README.md.
- Simplified sirf.Utilities.Handle for SWIG backend calls.
- Adjoint operator added in common/SIRF.py.
- Restored functionality for algebraic operations mixing SIRF data containers and NumPy arrays and corrected the description of the result type in User Guide.
- Array API used in Python implementation of norm(), dot() and copyto().
- Increased coverage of continuous integration tests
- Starting to move dependency handling towards conda packages
- Update dependency versions
- STIR: 6.4.0
- parallelproj: 2.0.6
- SIRF-Contribs: v.3.10.0
- CIL: v26.0.0
- CCPi-Regularisation toolkit: v26
- TomoPhantom: 3.1.4
- TIGRE: 3.1.3
- CUDA: 12.9.2
- CMake: 3.31.12
As well as building from source, SIRF is available as binary distribution as docker images and VirtualBox. See our wiki for details on how to obtain these.
