Repository navigation

Replace TensorKit cache machinery with MemoizationKit.jl - #568

Draft

lkdvos wants to merge 14 commits into

mainfrom

cached-integration

Draft

Replace TensorKit cache machinery with MemoizationKit.jl#568
lkdvos wants to merge 14 commits into
mainfrom
cached-integration

Commits on Nov 3, 2026
- Replace TensorKit cache machinery with MemoizationKit.jl
  
  lkdvos
  and
  claude
  committed
  1b86108
- Use MemoizationKit keys and remove duplicated cache helpers
  lkdvos
  committed
  783f448
- Integrate MemoizationKit timers with TensorKit profiling
  lkdvos
  committed
  a0a7ac9
- Measure TensorKit cached values by buffer sizes
  lkdvos
  committed
  27e0f7b
- Switch to registered MemoizationKit.jl
  
  lkdvos
  and
  claude
  committed
  57b8a5a
- Address review: descriptive cache names and transpose inference
  
  lkdvos
  and
  claude
  committed
  353e30d
- Drop the extra cache registry now that extension caches are owned by TensorKit
  
  lkdvos
  and
  claude
  committed
  340c0b9
- Use renamed fusiontreetranspose in Enzyme rules, unicode subscripts
  
  lkdvos
  and
  claude
  committed
  11c00d6
- Cache device transformers in separate functions
  
  lkdvos
  and
  claude
  committed
  ca2502e
- Require MemoizationKit v0.1.1
  
  lkdvos
  and
  claude
  committed
  fb57c40
- Ignore JET's uncorrelated union split in the tree transformer bodies
  
  lkdvos
  and
  claude
  committed
  2d72993
- Require MemoizationKit v0.1.2 and drop the tree transformer JET ignores
  
  lkdvos
  and
  claude
  committed
  7d70fe4
- Cache the diagonal device transformers with the other device transformers
  
  lkdvos
  and
  claude
  committed
  622f4eb
Commits on Nov 4, 2026
- Clear TensorKit's caches before each test
  
  lkdvos
  and
  claude
  committed
  5c3b3a9

Commits

Commits on Nov 3, 2026

Commits on Nov 4, 2026