All pull requests

Pull requests

• Docs: DatasetCache section and mesh reader entries

  #2017·coreyjadams opened Sep 23, 2026Collaborator·4/4·

  ·
  2

• Unified aero recipe: optional shared reader cache

  #2016·coreyjadams opened Sep 23, 2026Collaborator·3/4·

  ·
  2

• Route reader metadata reads through DatasetCache

  #2015·coreyjadams opened Sep 23, 2026Collaborator·2/4·

  ·
  2

• DatasetCache: opt-in metadata cache for datapipe readers

  #2014·coreyjadams opened Sep 23, 2026Collaborator·1/4·

  ·
  2

• Adds a namespace anchor to physicsnemo's __init__ file

  #2013·coreyjadams opened Sep 23, 2026Collaborator·

  ·
  3

  

• Aero recipe: prevent near-wall SDF sign noise from reversing normals

  #2011·peterdsharpe opened Sep 19, 2026Collaborator·

  ·
  5

• Checkpoint loading: require matching weights before restoring state

  ci:multi-gpu
  #2010·peterdsharpe opened Sep 19, 2026Collaborator·

  ·
  1

  8

• Fix zero frequency index along the width in metrics.general.power_spectrum

  #2008·Nicholas022400701 opened Sep 18, 2026·

  ·
  1

  4

• Fix cumulative bin accumulation and bin count bookkeeping in metrics.general.histogram

  #2006·Nicholas022400701 opened Sep 18, 2026·

  ·
  1

  4

• Fix Variance ensemble metric ignoring the dim argument

  #2004·Nicholas022400701 opened Sep 18, 2026·

  ·
  1

  4

• Fix AeroJEPA unbatched mask handling and validate chunked-decode arguments

  #1999·harshaa765 opened Sep 17, 2026·

  ·
  1

  7

• Regional weather diffusion dataloader updates

  #1996·pzharrington opened Sep 16, 2026Collaborator·

  ·
  9

  

• add explicit schema to mesh_green_gauss_gradient_warp custom_op"

  #1991·DenizCanELCI opened Sep 14, 2026·

  ·
  1

  3

• Add optional latent-bottleneck context reads to FLARE and GALE_FA

  #1989·CharlelieLrt opened Sep 14, 2026Collaborator·

  ·
  5

• Domain parallelism in the unified external aero recipe

  ci:multi-gpu
  #1983·coreyjadams opened Sep 10, 2026Collaborator·5/5·

  ·
  8

  

• Domain-parallel (rank-local) reading in the datapipes

  ci:multi-gpu
  #1982·coreyjadams opened Sep 10, 2026Collaborator·4/5·

  ·
  11

  

• Routed gather for sharded index_select and integer-tensor indexing

  ci:multi-gpu
  #1981·coreyjadams opened Sep 10, 2026Collaborator·3/5·

  ·
  17

  

• Move ShardTensor ring ops to functional collectives; compile-safe ring attention and ball query

  ci:multi-gpu
  #1980·coreyjadams opened Sep 10, 2026Collaborator·2/5·

  ·
  18

  

• Standalone datapipe and ShardTensor fixes ahead of domain-parallel reading

  ci:multi-gpu
  #1979·coreyjadams opened Sep 10, 2026Collaborator·1/5·

  ·
  15

• Aero recipe: filter degenerate cells with consistent measures

  #1973·peterdsharpe opened Sep 8, 2026Collaborator·

  ·
  6

• Unified aero recipe: coordinate step failures and restore scheduler/scaler state

  #1972·peterdsharpe opened Sep 8, 2026Collaborator·

  ·
  6

• Aero recipe: document sampled forces and preserve exported quadrature

  #1971·peterdsharpe opened Sep 8, 2026Collaborator·

  ·
  8

• Unify complete integration measures for mesh cells and point samples

  #1970·peterdsharpe opened Sep 8, 2026Collaborator·

  ·
  19

• Fix TypeError in sintering_physics deform_analyze plot call

  #1965·Anai-Guo opened Sep 4, 2026·

  ·
  3

• Fix missing sampler argument in xaeronet volume dataloader demo

  #1964·Anai-Guo opened Sep 4, 2026·

  ·
  3