Dontopedia

GroupedLinear

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

GroupedLinear has 14 facts recorded in Dontopedia across 3 references, with 1 live disagreement.

14 facts·12 predicates·3 sources·1 in dispute

Mostly:has state(2), uses fft(1), defines recurrence structure(1)

Maturity scale raw canonical shape-checked rule-derived certified

Inbound mentions (3)

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.

contrastsResonanceV2WithContrasts Resonance V2 With(1)

includeGroupedLinearInclude Grouped Linear(1)

isSuperiorInFlopsToIs Superior in Flops to(1)

Other facts (13)

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.

13 facts
PredicateValueRef
Has StateKv State 2[3]
Has StateK State[3]
Uses FftFft[1]
Defines Recurrence StructureGated Linear Attention With Decay[1]
Has Total Work Approximately290000000[1]
Is Cheapest Fft Varianttrue[1]
Performs Approximately12x More Flops ThanResonance V2[1]
Requires Padded Buffers2t Buffers[1]
Computes Work As64 × 12 × 3 × 8192 × 13[1]
Has Fft Channels Per Head64[1]
Has Group Size1[1]
Has Dual StatesKv State and K State[2]
Rdf:typeAttention Class[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.

usesFftblah/watt-activation/part-101
ex:fft
definesRecurrenceStructureblah/watt-activation/part-101
ex:gated-linear-attention-with-decay
hasTotalWorkApproximatelyblah/watt-activation/part-101
290000000
isCheapestFftVariantblah/watt-activation/part-101
true
performsApproximately12xMoreFlopsThanblah/watt-activation/part-101
ex:resonance-v2
requiresPaddedBuffersblah/watt-activation/part-101
ex:2t-buffers
computesWorkAsblah/watt-activation/part-101
64 × 12 × 3 × 8192 × 13
hasFftChannelsPerHeadblah/watt-activation/part-101
64
hasGroupSizeblah/watt-activation/part-101
1
hasDualStatesblah/watt-activation/part-158
ex:kv-state-and-k-state
typeblah/watt-activation/158
ex:AttentionClass
labelblah/watt-activation/158
GroupedLinear
hasStateblah/watt-activation/158
ex:kv-state-2
hasStateblah/watt-activation/158
ex:k-state

References (3)

3 references
  1. [1]Part 1019 facts
    ctx:discord/blah/watt-activation/part-101
  2. [2]Part 1581 fact
    ctx:discord/blah/watt-activation/part-158
  3. [3]1584 facts
    ctx:discord/blah/watt-activation/158
    • full textwatt-activation-158
      text/plain2 KBdoc:agent/watt-activation-158/746cbfb1-750c-4df5-ae90-78f0235bc1e9
      Show excerpt
      [2026-03-09 16:27] xenonfun: ⏺ For batch prefill, each attention class's forward() currently runs _gated_cumsum over the full sequence and returns output — but throws away the final recurrent state. The step() method maintains that state

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.