Dontopedia

Grouped1

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

Grouped1 has 25 facts recorded in Dontopedia across 3 references.

25 facts·25 predicates·3 sources

Mostly:has training throughput(1), is nearly five times faster than(1), is three point two times faster than(1)

Maturity scale raw canonical shape-checked rule-derived certified

Inbound mentions (9)

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.

describesForwardThroughputDescribes Forward Throughput(1)

describesScalingDescribes Scaling(1)

describesTrainingThroughputDescribes Training Throughput(1)

forFor(1)

hasAttentionTypeHas Attention Type(1)

hasWorkingBestHas Working Best(1)

isSlowerByLargeMarginIs Slower by Large Margin(1)

isSlowerInForwardIs Slower in Forward(1)

isSlowerInTrainingIs Slower in Training(1)

Other facts (25)

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.

25 facts
PredicateValueRef
Has Training Throughput123000[1]
Is Nearly Five Times Faster ThanPytorch Mps[1]
Is Three Point Two Times Faster ThanPytorch Mps Training[1]
Has Latency Scaling6.4[1]
Exhibits Sub Linear ScalingLatency Vs Tokens[1]
Outperforms Spectral in PeakForward[1]
Uses CumsumImplicit[1]
Peaks at Forward Throughput905000[1]
Is Essentiallygated linear recurrence s_t = γ·s_{t-1} + k_t·v_t[2]
Is Faster ThanKan Quadratic[2]
Is Handled Perfectly byGamma Scaled Cumsum[2]
Is Winnertrue[2]
Would Be Faster ThanAnchor Kan[2]
Has Descriptionelement-wise gated linear attention[2]
Has Ppl6.57[2]
Has Ppl Worseness Vs Spectral9%[2]
Has Quality Adjusted Tokens Per Second18374[2]
Has Tokens Per Second at Temp256516000[2]
Has Training Speed Multiplier Vs Kan10.7[2]
Is Element Wise Gated Linear Recurrencetrue[2]
Superior in Speed Quality Tradeofftrue[2]
Rdf:typeModel or Configuration[3]
Peak Memory Usage22[3]
Peak Memory Usage UnitGigabyte[3]
Measured at Sequence Length4096[3]

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.

hasTrainingThroughputblah/watt-activation/part-104
123000
isNearlyFiveTimesFasterThanblah/watt-activation/part-104
ex:pytorch-mps
isThreePointTwoTimesFasterThanblah/watt-activation/part-104
ex:pytorch-mps-training
hasLatencyScalingblah/watt-activation/part-104
6.4
exhibitsSubLinearScalingblah/watt-activation/part-104
ex:latency-vs-tokens
outperformsSpectralInPeakblah/watt-activation/part-104
ex:forward
usesCumsumblah/watt-activation/part-104
ex:implicit
peaksAtForwardThroughputblah/watt-activation/part-104
905000
isEssentiallyblah/watt-activation/part-103
gated linear recurrence s_t = γ·s_{t-1} + k_t·v_t
isFasterThanblah/watt-activation/part-103
ex:kan-quadratic
isHandledPerfectlyByblah/watt-activation/part-103
ex:gamma-scaled-cumsum
isWinnerblah/watt-activation/part-103
true
wouldBeFasterThanblah/watt-activation/part-103
ex:anchor-kan
hasDescriptionblah/watt-activation/part-103
element-wise gated linear attention
hasPplblah/watt-activation/part-103
6.57
hasPplWorsenessVsSpectralblah/watt-activation/part-103
9%
hasQualityAdjustedTokensPerSecondblah/watt-activation/part-103
18374
hasTokensPerSecondAtTemp256blah/watt-activation/part-103
516000
hasTrainingSpeedMultiplierVsKanblah/watt-activation/part-103
10.7
isElementWiseGatedLinearRecurrenceblah/watt-activation/part-103
true
superiorInSpeedQualityTradeoffblah/watt-activation/part-103
true
typeblah/watt-activation/117
ex:ModelOrConfiguration
peakMemoryUsageblah/watt-activation/117
22
peakMemoryUsageUnitblah/watt-activation/117
ex:gigabyte
measuredAtSequenceLengthblah/watt-activation/117
4096

References (3)

3 references
  1. [1]Part 1048 facts
    ctx:discord/blah/watt-activation/part-104
  2. [2]Part 10313 facts
    ctx:discord/blah/watt-activation/part-103
  3. [3]1174 facts
    ctx:discord/blah/watt-activation/117
    • full textwatt-activation-117
      text/plain2 KBdoc:agent/watt-activation-117/93da7c3e-762b-4710-96fb-cdcbffaa6ad1
      Show excerpt
      [2026-03-08 23:25] xenonfun: okay it still is using Adam finally pointed it at your optimizers: ``` ⏺ This is really interesting. There are 4 custom optimizers in harmonic-gpt/harmonic_gpt/optim/: 1. RotationalAdamW — Angular updates on

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.