Spectral Plus Lohe V2

as donto believed on

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

Has Throughput	38.2K tok/s^[5]
Has Throughput	11665^[6]
Has Loss at Step	7.3274^[5]
Has Loss at Step	7.4139^[5]
Has Attn R	0^[6]
Has Global R	0.1905^[6]
Has Params	107497440^[6]
Has Sys R	0.669^[6]
Has Tok Per S	11665^[6]
Has Val Ppl	2490.7^[6]

ex:spectral-plus-lohe-v2

Spectral Plus Lohe V2 has 37 facts recorded in Dontopedia across 6 references, with 2 live disagreements.

37 facts·32 predicates·6 sources·2 in dispute

Mostly:has loss at step(5), has throughput(2), has attn r(1)

Has Throughputin dispute`hasThroughput`

38.2K tok/s^[5]all time · Part 201
11665^[6]all time · Part 204

Has Loss at Stepin dispute`hasLossAtStep`

7.3274^[5]all time · Part 201
7.4139^[5]all time · Part 201
7.4532^[5]all time · Part 201
7.4257^[5]all time · Part 201
7.2788^[5]all time · Part 201

Has Params`hasParams`

107497440^[6]all time · Part 204

Has Wall Time Minutes`hasWallTimeMinutes`

5.9^[6]all time · Part 204

Serves As Baseline`servesAsBaseline`

Lohe Spherical Plus Lohe V2^[6]all time · Part 204

Exhibits Loss Descent`exhibitsLossDescent`

Checkmark^[5]all time · Part 201

Shows Loss Improvement`showsLossImprovement`

from 7.4532 to 7.2788^[5]all time · Part 201

Inbound mentions (18)

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.

`uses`Uses(4)

1 Epoch Checkpointex:1-epoch-checkpoint
1 Epoch Checkpointex:1-epoch-checkpoint
Checkpointex:checkpoint
Spectral Plus Lohe V2 Checkpointex:spectral-plus-lohe-v2-checkpoint

`combines`Combines(1)

Form 1ex:form-1

`hasLowerParamCountThan`Has Lower Param Count Than(1)

Lohe Spherical Plus Lohe V2ex:lohe-spherical-plus-lohe-v2

`hasLowerThroughput`Has Lower Throughput(1)

Lohe Spherical Attentionex:lohe-spherical-attention

`hasSlightlyWorseQualityThan`Has Slightly Worse Quality Than(1)

Lohe Spherical Plus Lohe V2ex:lohe-spherical-plus-lohe-v2

`highlightsKeyResult`Highlights Key Result(1)

Xenonfunex:xenonfun

`isComparedToBaseline`Is Compared to Baseline(1)

Lohe Spherical Plus Lohe V2ex:lohe-spherical-plus-lohe-v2

`isComponentOf`Is Component of(1)

Lohe V2ex:lohe-v2

`isFasterThan`Is Faster Than(1)

Lohe Spherical Plus Lohe V2ex:lohe-spherical-plus-lohe-v2

`isSameForBothConfigs`Is Same for Both Configs(1)

Seq Batchex:seq-batch

`matchesLossDescent`Matches Loss Descent(1)

Lohe Spherical Attentionex:lohe-spherical-attention

`outperformsLossDescent`Outperforms Loss Descent(1)

Lohe Spherical Plus Lohe V2ex:lohe-spherical-plus-lohe-v2

`preferredOverBaseline`Preferred Over Baseline(1)

Lohe Spherical Plus Lohe V2ex:lohe-spherical-plus-lohe-v2

`runsAtSameSeqBatchAs`Runs at Same Seq Batch As(1)

Lohe Spherical Plus Lohe V2ex:lohe-spherical-plus-lohe-v2

`zeroForSpectral`Zero for Spectral(1)

Attn Rex:attn-r

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

Predicate	Value	Ref
Causes More Data Seen in Equal Time	Spectral Plus Gelu	[3]
Has Fewer Parameters Than	Spectral Plus Gelu	[3]
Has Higher Ppl Than	Spectral Plus Gelu	[3]
Has Parameter Count	37.4	[3]
Has Percent of Parameters of	Spectral Plus Gelu	[3]
Has Percent Quality of	Spectral Plus Gelu	[3]
Has Validation Perplexity	278.2	[3]
Matches Result Pattern of	Cuda Spectralgpt	[3]
Runs Percent Faster Than	Spectral Plus Gelu	[3]
Benefits From Longer Seq	Meaningful	[2]
Is Meaningful for	Longer Seq	[2]
Has Ppl	2491	[6]
Runs at Tok Per S Approx	11700	[6]
Is Efficient Alternative	Spectral Plus Gelu	[3]
Likely Closes Performance Gap	Spectral Plus Gelu	[3]
Differs From	Spectral Plus Lohe V3	[4]
Version Label	v2	[4]
Benefits From Longer Context	meaningful	[1]
Rdfs:label	spectral+lohe_v2	[1]
Rdf:type	Model Architecture	[1]

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.

37 undated

—

benefitsFromLongerContextblah/watt-activation/125

meaningful

—

benefitsFromLongerSeqblah/watt-activation/part-125

ex:meaningful

—

causesMoreDataSeenInEqualTimeblah/watt-activation/part-123

ex:spectral-plus-gelu

—

differsFromblah/watt-activation/223

ex:spectral-plus-lohe-v3

—

exhibitsLossDescentblah/watt-activation/part-201

ex:checkmark

—

hasAttnRblah/watt-activation/part-204

—

hasFewerParametersThanblah/watt-activation/part-123

ex:spectral-plus-gelu

—

hasGlobalRblah/watt-activation/part-204

0.1905

—

hasHigherPplThanblah/watt-activation/part-123

ex:spectral-plus-gelu

—