kan (seq=2048)

as donto believed on

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

kan (seq=2048) has 38 facts recorded in Dontopedia across 2 references.

38 facts·36 predicates·2 sources

Mostly:has parameter count(2), performed num iterations(1), exhibits overall loss decrease(1)

Inbound mentions (24)

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.

`isPartOf`Is Part of(11)

Kan Iter 10ex:kan-iter-10
Kan Iter 100ex:kan-iter-100
Kan Iter 20ex:kan-iter-20
Kan Iter 30ex:kan-iter-30
Kan Iter 40ex:kan-iter-40
Kan Iter 50ex:kan-iter-50
Kan Iter 60ex:kan-iter-60
Kan Iter 70ex:kan-iter-70
Kan Iter 80ex:kan-iter-80
Kan Iter 90ex:kan-iter-90
Kan Summaryex:kan-summary

`partOf`Part of(10)

Kan Iter 10ex:kan-iter-10
Kan Iter 100ex:kan-iter-100
Kan Iter 20ex:kan-iter-20
Kan Iter 30ex:kan-iter-30
Kan Iter 40ex:kan-iter-40
Kan Iter 50ex:kan-iter-50
Kan Iter 60ex:kan-iter-60
Kan Iter 70ex:kan-iter-70
Kan Iter 80ex:kan-iter-80
Kan Iter 90ex:kan-iter-90

`hasMoreParametersThan`Has More Parameters Than(1)

Spectral Runex:spectral-run

`presentsIterProgress`Presents Iter Progress(1)

Kan Training Logex:kan-training-log

`reportsTrainingResults`Reports Training Results(1)

Xenonfunex:xenonfun

Other facts (37)

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.

37 facts

Predicate	Value	Ref
Has Parameter Count	86826048	[1]
Has Parameter Count	86826048	[2]
Performed Num Iterations	100	[1]
Exhibits Overall Loss Decrease	Kan Iter 100	[1]
Uses Attention Mechanism	Kan Attn	[1]
Trains Model	Harmonic Gpt	[1]
Presupposes Successful Convergence	low loss	[1]
Presupposes Gpu Usage	Gpu	[1]
Has Slightly Higher Gpu Cache Than	Spectral Run	[1]
Has Status	complete	[1]
Implies Slightly Worse Performance	Spectral Run	[1]
Loaded Num Sequences	5000	[1]
Presupposes Data Loading	Kan Data	[1]
Has Constant Gpu Active	1	[1]
Has Constant Process Rss	3.2	[1]
Has Higher Iter Gpu Peak Than	Spectral Run	[1]
Has Higher Parameter Count Than	Spectral Run	[1]
Has Learning Rate Schedule	Cosine Decay Lr	[1]
Has Max Sequence Length	2048	[1]
Rdf:type	Training Run	[2]
Reported in	Log Message 1	[2]
Models	Harmonic Gpt	[2]
Has Attention Mechanism	kan	[2]
Has Embedding Dimension	768	[2]
Has Layer Count	12	[2]
Has Head Count	12	[2]
Has Sequence Length	2048	[2]
Loaded Sequence Count	5000	[2]
Loaded Sequence Length	2048	[2]
Has Final Loss	2.8128	[2]
Has Elapsed Time	55.8s	[2]
Total Tokens Processed	819200	[2]
Average Tokens Per Second	14685	[2]
Gpu Peak Memory	13.99 GB	[2]
Gpu Active Memory	1.01 GB	[2]
Gpu Cache Memory	13.95 GB	[2]
Process Resident Set Size	3.19 GB	[2]

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.

38 undated

—

performedNumIterationsblah/watt-activation/part-112

100

—

exhibitsOverallLossDecreaseblah/watt-activation/part-112

ex:kan-iter-100

—

usesAttentionMechanismblah/watt-activation/part-112

ex:kan-attn

—

trainsModelblah/watt-activation/part-112

ex:harmonic-gpt

—

hasParameterCountblah/watt-activation/part-112

86826048

—

presupposesSuccessfulConvergenceblah/watt-activation/part-112

low loss

—

presupposesGpuUsageblah/watt-activation/part-112

ex:gpu

—

hasSlightlyHigherGpuCacheThanblah/watt-activation/part-112

ex:spectral-run

—

hasStatusblah/watt-activation/part-112

complete

—

impliesSlightlyWorsePerformanceblah/watt-activation/part-112

ex:spectral-run

—

loadedNumSequencesblah/watt-activation/part-112

5000

—

presupposesDataLoadingblah/watt-activation/part-112

ex:kan-data

—

hasConstantGpuActiveblah/watt-activation/part-112

—

hasConstantProcessRSSblah/watt-activation/part-112

3.2

—

hasHigherIterGpuPeakThanblah/watt-activation/part-112

ex:spectral-run

—

hasHigherParameterCountThanblah/watt-activation/part-112

ex:spectral-run

—

hasLearningRateScheduleblah/watt-activation/part-112

ex:cosine-decay-lr

—

hasMaxSequenceLengthblah/watt-activation/part-112

2048

—

typeblah/watt-activation/112

ex:TrainingRun

—

labelblah/watt-activation/112

kan (seq=2048)

—

reportedInblah/watt-activation/112

ex:log-message-1

—

modelsblah/watt-activation/112

ex:HarmonicGPT

—

hasAttentionMechanismblah/watt-activation/112

kan

—

hasEmbeddingDimensionblah/watt-activation/112

768

—

hasLayerCountblah/watt-activation/112

—

hasHeadCountblah/watt-activation/112

—

hasSequenceLengthblah/watt-activation/112

2048

—

hasParameterCountblah/watt-activation/112

86826048

—

loadedSequenceCountblah/watt-activation/112

5000

—

loadedSequenceLengthblah/watt-activation/112

2048

—

hasFinalLossblah/watt-activation/112

2.8128

—

hasElapsedTimeblah/watt-activation/112

55.8s

—

totalTokensProcessedblah/watt-activation/112

819200

—

averageTokensPerSecondblah/watt-activation/112

14685

—

gpuPeakMemoryblah/watt-activation/112

13.99 GB

—

gpuActiveMemoryblah/watt-activation/112

1.01 GB

—

gpuCacheMemoryblah/watt-activation/112

13.95 GB

—

processResidentSetSizeblah/watt-activation/112

3.19 GB

References (2)

2 references

[1]Part 11218 facts
ctx:discord/blah/watt-activation/part-112
[2]11220 facts
ctx:discord/blah/watt-activation/112
- full textwatt-activation-112
  text/plain3 KBdoc:agent/watt-activation-112/dad4c21d-6073-4542-b09b-5b5162617d30
  Show excerpt
  [2026-03-08 22:40] xenonfun: ``` === kan (seq=2048) === Training HarmonicGPT | attn=kan d=768 L=12 H=12 seq=2048 Parameters: 86,826,048 Loaded 5,000 sequences of length 2048 iter 10/100 | loss 5.4382 | lr 3.00e-04 | tok/s 14,738 | gpu 1…

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.

kan (seq=2048)

Inbound mentions (24)

isPartOfIs Part of(11)

partOfPart of(10)

hasMoreParametersThanHas More Parameters Than(1)

presentsIterProgressPresents Iter Progress(1)

reportsTrainingResultsReports Training Results(1)

Other facts (37)

Timeline

References (2)

See also

Keep researching

`isPartOf`Is Part of(11)

`partOf`Part of(10)

`hasMoreParametersThan`Has More Parameters Than(1)

`presentsIterProgress`Presents Iter Progress(1)

`reportsTrainingResults`Reports Training Results(1)