komamripy - GSoC Proposal Issue Tracking

[anvika-singhal]

Feature Request

PART 1: tracking for objectives in GSoC Proposal (May 25 - July 10):

• Define minimal Python API for KomaMRI simulation KomaMRI.jl#732
• Set up CI pipeline for Python + Julia integration KomaMRI.jl#733
• Add Python simulation examples and validation KomaMRI.jl#735

PART 2: tracking for objectives in GSOC proposal (July 13 - August 31):

• Implement Julia precompilation for common workflows KomaMRI.jl#734
• Add GPU backend selection support for komamripy KomaMRI.jl#736
• Add JAX-based differentiable RF optimization example for komamripy KomaMRI.jl#737
• Multi-GPU / distributed simulation support for komamripy KomaMRI.jl#738 (stretch goal)
• Write blog post for JuliaHealth website documenting demonstrating Python-Julia integration and summarizing lessons learned from developing komamripy.

[Mxyank24]

can someone tell me about GSoC tag ?

[cncastillo]

Hi! GSoC stands for Google Summer of Code. This tag is used for tasks related to projects developed during GSoC, which just started.

In particular, the tasks for the komamripy project will be carried out by @anvika-singhal.

So try to ignore them for now :)

[Mxyank24]

oh alright .

[cncastillo]

Things that are missing for issues 1 and 2:

• CI running properly (cache setup correctly + remove julia version set as 1.11)
• Now the "tag" -> "PyPi" release workflow works, but we need a new automated workflow that:
  • Runs once a week.
  • Use Julia’s package resolver to resolve KomaMRI + KomaMRIBase + KomaMRICore + KomaMRIFiles + KomaMRIPlots together.
  • Update [juliapkg.json](https://github.com/JuliaHealth/komamripy/blob/main/komamripy/juliapkg.json) only if the resolved direct package versions changed.
  • If nothing changed, exit successfully and publish nothing.
  • If something changed, bump the Python patch version, run tests, tag, publish to PyPI, and create a GitHub release whose notes list the old and new Julia package versions PkgA 0.2.1 -> 0.2.2. (how do we add release notes to the PyPi release?)
• List methods mostly used by frequency on Koma's docs and check they work exactly as in Koma's docs (one to one syntax). In addition, it is very important to support (for MRI Together 2026)
  • GPU accelerated simulations with backend selection (can be checked in part 2)
  • Pulseq .seq read/write
  • Phantom creation, including motion, and .phantom read/write
  • This could be done by creating 2-3 documentation examples that cover basic functionality