- Integrated the IsoApp GUI into this repository under
IsoApp/. - Built and packaged the GUI successfully with
npm run build:linux. - Added a faster and safer STAR loading path in
IsoApp/src/main/handlers/other.jsand restored prepare-page refresh flow.
- Added FSC loss support in
IsoNet/models/masked_loss.pyand wired it into training. - Added a PyTorch CUDA kernel cache fallback in
IsoNet/__init__.pyto avoid cache-directory warnings. - Updated training output and plots to display
n2n_lossandmw_lossinstead ofinside_lossandoutside_loss.
- Improved GPU ID validation and scheduler handling in the training workflow.
- Fixed MRC header voxel size preservation during I/O.
- Resolved DDP process group lifecycle issues during prediction.
- Updated
.gitignoreto better ignore local, editor, and temporary files. - The AppImage crash issue appears to be environment-specific; the packaged GUI build is successful from this repo.
- The IsoApp GUI is integrated into this repository under
IsoApp/, not a separate repository. - Use
npm cifollowed bynpm run build:linuxinIsoApp/to reproduce the current GUI package.