Code

CUDA Support for ALS#37

Code

Open

alexmul1114 wants to merge 24 commits intodahong67:masterdahong67/GCPDecompositions.jl:masterfrom alexmul1114:CUDA-Supportalexmul1114/GCPDecompositions.jl:CUDA-SupportCopy head branch name to clipboard

+58-2

Commits

Commits on Jan 8, 2024

• Fixes dahong67#36 (with simple MTTKRP implementation)

  alexmul1114

  committed
  d48f0f1

  Copy full SHA for d48f0f1

• Change to Arrays AbstractArrays

  alexmul1114

  committed
  4596484

  Copy full SHA for 4596484

Commits on Jan 10, 2024

• Replace skipmissing

  alexmul1114

  committed
  6e2a476

  Copy full SHA for 6e2a476

• Use ismissing in mapreduce instead of skipmissing

  alexmul1114

  committed
  01e2ab9

  Copy full SHA for 01e2ab9

• Add new _gcp method for CuArray

  alexmul1114

  committed
  1e6097e

  Copy full SHA for 1e6097e

• Add CUDA gcp as extension, use simple MTTKRP implementation and khatri rao function from dahong67#34

  alexmul1114

  committed
  fd71ca4

  Copy full SHA for fd71ca4

• Add CUDA to deps for backward compatiblity

  alexmul1114

  committed
  e5b30ea

  Copy full SHA for e5b30ea

• Move CUDA to extras

  alexmul1114

  committed
  c34cc3c

  Copy full SHA for c34cc3c

• Change CUDA version compatibility

  alexmul1114

  committed
  966e30b

  Copy full SHA for 966e30b

• Modify project.toml

  alexmul1114

  committed
  9938c46

  Copy full SHA for 9938c46

Commits on Jan 11, 2024

• Test with new MTTKRP from dahong67#35

  alexmul1114

  committed
  e8077a3

  Copy full SHA for e8077a3

• Change .= back to =

  alexmul1114

  committed
  fb1d0ce

  Copy full SHA for fb1d0ce

• Fix typos

  alexmul1114

  committed
  e368433

  Copy full SHA for e368433

• Use latest MTTKRP from dahong67#35

  alexmul1114

  committed
  614929d

  Copy full SHA for 614929d

• collect selectdim for now

  alexmul1114

  committed
  7854ec8

  Copy full SHA for 7854ec8

• Try reusing inner

  alexmul1114

  committed
  df803e1

  Copy full SHA for df803e1

• Use copy in MTTKRP middle mode Rn computation for now

  alexmul1114

  committed
  2a7f0c2

  Copy full SHA for 2a7f0c2

Commits on Jan 12, 2024

• Change abstract arrays back to arrays

  alexmul1114

  committed
  b72f6e9

  Copy full SHA for b72f6e9

• Remove typo

  alexmul1114

  committed
  12f0004

  Copy full SHA for 12f0004

• Switch khatrirao to accept abstractmatrix

  alexmul1114

  committed
  e9680f6

  Copy full SHA for e9680f6

• Fix typo

  alexmul1114

  committed
  ebc5455

  Copy full SHA for ebc5455

Commits on Jan 15, 2024

• Temporarily use Float32 for comparison against GPU

  alexmul1114

  committed
  ab40583

  Copy full SHA for ab40583

Commits on Jan 18, 2024

• Replace norm function

  alexmul1114

  committed
  f4696c0

  Copy full SHA for f4696c0

Commits on Mar 6, 2024

• Merge branch 'master' into pr/alexmul1114/37

  dahong67

  committed
  d7f68ed

  Copy full SHA for d7f68ed