Dontopedia

Slow Muon

From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-07-03.)

Slow Muon is more faithful to full Muon math.

21 facts·20 predicates·2 sources·1 in dispute

Mostly:description(2), is evaluated as very expensive(1), recommended only for(1)

Maturity scale raw canonical shape-checked rule-derived certified

Inbound mentions (1)

Other subjects in dontopedia point AT this entity as a value. These are inverse relationships — e.g. "X motherOf this subject" — and answer questions the forward facts can't. Grouped by predicate.

hasMemberHas Member(1)

Other facts (21)

The long tail: predicates that appear too rarely to warrant their own section. Filter or scroll to find a specific one. Each row links to its source.

21 facts
PredicateValueRef
Descriptionmore faithful to full Muon math[2]
Descriptionorthogonalizes almost every matrix update[2]
Is Evaluated As Very Expensivenull[1]
Recommended Only forResearch Checks Short Runs[1]
Implies Full Orthogonalization Necessitynull[1]
Is Evaluated As More Faithfulnull[1]
Orthogonalizes Almost EveryMatrix Update[1]
Is Very Expensive onThis Model[1]
Is More Faithful toFull Muon Math[1]
PerformsMatrix Update Orthogonalization[2]
Is Expensivetrue[2]
Runtime in Seconds160.7[2]
Recommended forResearch Checks[2]
Orthogonalizes Every Matrix Updatetrue[2]
Compared toFast Mode[2]
Adheres toFull Muon Math[2]
Rdf:typeMuon Method[2]
Computational Costvery expensive[2]
Runtime160.7[2]
Quality Attributemore faithful[2]
Faithful tofull Muon math[2]

Timeline

Timeline axis is valid_time — when each source says the fact was true in the world, not when Dontopedia learned about it. Retracted rows are kept for provenance; coloured stripes indicate the context kind.

isEvaluatedAsVeryExpensiveblah/watt-activation/part-19
null
recommendedOnlyForblah/watt-activation/part-19
ex:research-checks-short-runs
impliesFullOrthogonalizationNecessityblah/watt-activation/part-19
null
isEvaluatedAsMoreFaithfulblah/watt-activation/part-19
null
orthogonalizesAlmostEveryblah/watt-activation/part-19
ex:matrix-update
isVeryExpensiveOnblah/watt-activation/part-19
ex:this-model
isMoreFaithfulToblah/watt-activation/part-19
ex:full-muon-math
performsdocument/021ac8ae-0614-433f-be33-3fab76adc335
ex:matrix-update-orthogonalization
isExpensivedocument/021ac8ae-0614-433f-be33-3fab76adc335
true
runtimeInSecondsdocument/021ac8ae-0614-433f-be33-3fab76adc335
160.7
descriptiondocument/021ac8ae-0614-433f-be33-3fab76adc335
more faithful to full Muon math
recommendedFordocument/021ac8ae-0614-433f-be33-3fab76adc335
ex:research-checks
orthogonalizesEveryMatrixUpdatedocument/021ac8ae-0614-433f-be33-3fab76adc335
true
comparedTodocument/021ac8ae-0614-433f-be33-3fab76adc335
ex:fast-mode
adheresTodocument/021ac8ae-0614-433f-be33-3fab76adc335
ex:full-muon-math
typedocument/021ac8ae-0614-433f-be33-3fab76adc335
ex:MuonMethod
descriptiondocument/021ac8ae-0614-433f-be33-3fab76adc335
orthogonalizes almost every matrix update
computationalCostdocument/021ac8ae-0614-433f-be33-3fab76adc335
very expensive
runtimedocument/021ac8ae-0614-433f-be33-3fab76adc335
160.7
qualityAttributedocument/021ac8ae-0614-433f-be33-3fab76adc335
more faithful
faithfulTodocument/021ac8ae-0614-433f-be33-3fab76adc335
full Muon math

References (2)

2 references
  1. [1]Part 197 facts
    ctx:discord/blah/watt-activation/part-19
  2. ctx:claims/document/021ac8ae-0614-433f-be33-3fab76adc335
    • full textxenonfun: - slow Muon orthogonalizes almost every matrix update, which is very expensive o
      text/plain717 Bdiscord:convmsg/3928
      Show excerpt
      xenonfun: - slow Muon orthogonalizes almost every matrix update, which is very expensive on this model. - fast Muon only orthogonalizes smaller tensors and uses normalized momentum on large ones. Observed impact in your runs (same 22-s

See also

Keep researching

Missing something or suspicious of what's here? Kick off a research session — a Claude agent will investigate, cite its sources, and file new facts into a dedicated context you can review before accepting into the shared view.