All pull requests

Pull requests

• Fix FNO recompiling for every grid size under torch.compile(dynamic=True)

  #2238·aarochu opened Nov 8, 2026·6 tasks done·

  ·
  1

  4

• Fix TypeError for constant terms in sym torch_lambdify

  #2237·udsy19 opened Nov 7, 2026Contributor·4 of 5 tasks·

  ·
  3

• GLOBE DrivAerML example: defaults from the paper's DrivAerML configuration

  #2236·peterdsharpe opened Nov 7, 2026Collaborator·2 of 6 tasks·

  ·
  9

• closes #2214 Fix batch-dimension slicing in physicsnemo_fdm_swe_nonlin PDE residuals for batchsize=1

  #2233·dmitriySem opened Nov 5, 2026·

  ·
  1

  6

• Support float16 and bfloat16 meshes in transforms, Loop subdivision, and sampling

  #2230·peterdsharpe opened Nov 4, 2026Collaborator·4 of 6 tasks·

  ·
  5

• Fix data loss and cell-order dependence in Mesh PyVista and Zarr I/O

  #2228·peterdsharpe opened Nov 4, 2026Collaborator·4 of 6 tasks·

  ·
  4

• Make Mesh intrinsic point gradients independent of cell winding

  #2227·peterdsharpe opened Nov 4, 2026Collaborator·4 of 6 tasks·

  ·
  4

• Fix silent errors and crashes in Mesh DEC, LSQ, and point-measure calculus

  #2226·peterdsharpe opened Nov 4, 2026Collaborator·4 of 6 tasks·

  ·
  5

• Fix autograd through cached Mesh geometry

  #2225·peterdsharpe opened Nov 4, 2026Collaborator·4 of 6 tasks·

  ·
  3

• Fix container from-source builds on PyTorch 2.15 (NATTEN C++20, pyg-lib cache)

  #2224·ktangsali opened Nov 4, 2026Collaborator·3 of 6 tasks·

  ·
  3

• GeoTransolver SDF (3/3): jet engine bracket example (training, evaluation, LoRA transfer)

  #2219·sgtkeum opened Nov 3, 2026·6 tasks done·

  ·
  2

• GeoTransolver SDF (2/3): experimental GeoTransolverSDF model

  #2218·sgtkeum opened Nov 3, 2026·6 tasks done·

  ·
  2

• GeoTransolver SDF (1/3): signed-distance stencil and radius-pooled local feature layers (experimental)

  #2217·sgtkeum opened Nov 3, 2026·6 tasks done·

  ·
  14

• Regional weather diffusion: torch.compile with domain parallelism

  ! - Releaseci:multi-gpu
  #2211·coreyjadams opened Oct 30, 2026Collaborator·2/2·4 of 6 tasks·

  ·
  1

  6

• ShardTensor: fixes for torch.compile with domain parallelism (#1986)

  ! - Releaseci:multi-gpu
  #2210·coreyjadams opened Oct 30, 2026Collaborator·1/2·5 of 6 tasks·

  ·
  15

• Adopt proposed FastGen loss hooks conditionally

  2 - In Progress5 - DO NOT MERGE5 - Merge After Dependencies
  #2205·CharlelieLrt opened Oct 29, 2026Collaborator·3 tasks done·

  ·
  2

• Add optional diffusion distillation strategy

  2 - In Progress5 - DO NOT MERGE5 - Merge After Dependencies
  #2204·CharlelieLrt opened Oct 29, 2026Collaborator·3 tasks done·

  ·
  2

• Add distillation checkpoint bundles

  2 - In Progress5 - DO NOT MERGE5 - Merge After Dependencies
  #2203·CharlelieLrt opened Oct 29, 2026Collaborator·3 tasks done·

  ·
  5

• Add DMD2 and f-distill training losses

  2 - In Progress5 - DO NOT MERGE5 - Merge After Dependencies
  #2202·CharlelieLrt opened Oct 29, 2026Collaborator·3 tasks done·

  ·
  2

• Add sCM and MeanFlow distillation losses

  2 - In Progress5 - DO NOT MERGE5 - Merge After Dependencies
  #2201·CharlelieLrt opened Oct 29, 2026Collaborator·3 tasks done·

  ·
  3

• Add inverse SNR and TrigFlow components

  2 - In Progress5 - DO NOT MERGE5 - Merge After Dependencies! - Release
  #2200·CharlelieLrt opened Oct 29, 2026Collaborator·3 tasks done·

  ·
  2

• Add optional extra diffusion times and outputs

  3 - Ready for Review! - Release
  #2199·CharlelieLrt opened Oct 29, 2026Collaborator·3 tasks done·

  ·
  12

• Move shared test infrastructure to physicsnemo.testing

  #2198·coreyjadams opened Oct 28, 2026Collaborator·6 tasks·

  ·
  2

• Move mesh-only functionals from nn.functional into physicsnemo.mesh

  #2195·peterdsharpe opened Oct 28, 2026Collaborator·6 tasks done·

  ·
  30

• UQ/Guardrails Refactor [6/6]: Introduce composable OOD scorers and retire guardrails

  ! - Release
  #2189·ktangsali opened Oct 26, 2026Collaborator·6/6·

  ·
  15