Dontopedia

2K run

From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-06-06.)

2K run has 21 facts recorded in Dontopedia across 4 references.

21 facts·18 predicates·4 sources

Mostly:has same token budget as(2), provides better gradient diversity than(2), serves as baseline(1)

Maturity scale raw canonical shape-checked rule-derived certified

Inbound mentions (10)

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.

hasFewerOptimizerStepsThanHas Fewer Optimizer Steps Than(2)

appropriateForAppropriate for(1)

commitsToTestingCommits to Testing(1)

hasWorseBatchDiversityThanHas Worse Batch Diversity Than(1)

plansToRunPlans to Run(1)

resumedFrom2kResumed From2k(1)

tunedForTuned for(1)

tunedOnTuned on(1)

underperformsComparedToUnderperforms Compared to(1)

Other facts (20)

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.

20 facts
PredicateValueRef
Has Same Token Budget As128k Run[1]
Has Same Token Budget As128k Run[3]
Provides Better Gradient Diversity Than128k Run[1]
Provides Better Gradient Diversity Than128k Run[3]
Serves As BaselineFor Comparison[1]
Warmup As Fraction of Epochabout 4%[1]
Uses Sequence Length2048[1]
Has Steps Per Epoch12432[1]
Provides Many More Local Contexts Per StepTrue[1]
Uses Batch Size8[1]
Tests FixesCoupling Fixes[2]
Likely Long RunningLong Computation[2]
Uses OptimizationsCoupling Optimizations[2]
Steps Per Epoch12432[3]
Warmup Ratio Approximate0.04[3]
Provides More Iterative Refinement Than128k Run[3]
Batch Size8[3]
Sequence Length2048[3]
Has More Local Contexts Per Step Than128k Run[3]
Rdf:typeExecution Run[4]

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.

servesAsBaselineblah/watt-activation/part-265
ex:for-comparison
warmupAsFractionOfEpochblah/watt-activation/part-265
about 4%
hasSameTokenBudgetAsblah/watt-activation/part-265
ex:128k-run
usesSequenceLengthblah/watt-activation/part-265
2048
hasStepsPerEpochblah/watt-activation/part-265
12432
providesBetterGradientDiversityThanblah/watt-activation/part-265
ex:128k-run
providesManyMoreLocalContextsPerStepblah/watt-activation/part-265
ex:true
usesBatchSizeblah/watt-activation/part-265
8
testsFixesblah/watt-activation/part-446
ex:coupling-fixes
likelyLongRunningblah/watt-activation/part-446
ex:long-computation
usesOptimizationsblah/watt-activation/part-446
ex:coupling-optimizations
labelblah/watt-activation/263
2K run
stepsPerEpochblah/watt-activation/263
12432
warmupRatioApproximateblah/watt-activation/263
0.04
hasSameTokenBudgetAsblah/watt-activation/263
ex:128k-run
providesMoreIterativeRefinementThanblah/watt-activation/263
ex:128k-run
batchSizeblah/watt-activation/263
8
sequenceLengthblah/watt-activation/263
2048
hasMoreLocalContextsPerStepThanblah/watt-activation/263
ex:128k-run
providesBetterGradientDiversityThanblah/watt-activation/263
ex:128k-run
typeblah/watt-activation/444
ex:ExecutionRun

References (4)

4 references
  1. [1]Part 2658 facts
    ctx:discord/blah/watt-activation/part-265
  2. [2]Part 4463 facts
    ctx:discord/blah/watt-activation/part-446
  3. [3]2639 facts
    ctx:discord/blah/watt-activation/263
    • full textwatt-activation-263
      text/plain2 KBdoc:agent/watt-activation-263/18f755d3-7fe3-4c11-aca5-c67fdc4eb174
      Show excerpt
      [2026-03-13 04:59] xenonfun: ``` • Yes. Several plausible reasons, even without blaming the architecture. Most important ones: - Warmup/schedule mismatch - the 2K run had 12,432 steps/epoch - the 128K run had only 1,554
  4. [4]4441 fact
    ctx:discord/blah/watt-activation/444
    • full textwatt-activation-444
      text/plain2 KBdoc:agent/watt-activation-444/885ad207-9977-4a41-b00d-589966869135
      Show excerpt
      [2026-03-20 17:49] xenonfun: <@1211062099137265723>, had it reread your FedSym merge, think this is closer to what you desribed and has zero scaling issues. ``` ⏺ Now I see it. The key things I missed: 1. _transfer_weights_parallel with

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.