Grouped4 Training

as donto believed on

From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-05-02.)

Grouped4 Training has 101 facts recorded in Dontopedia across 1 reference.

101 facts·101 predicates·1 sources

Mostly:exhibits loss improvement over iterations(1), has average tokens per second(1), has elapsed time seconds(1)

Inbound mentions (1)

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.

`reportsTrainingMetrics`Reports Training Metrics(1)

Grouped4 Training Logex:grouped4-training-log

Other facts (101)

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.

101 facts

Predicate	Value	Ref
Exhibits Loss Improvement Over Iterations	true	[1]
Has Average Tokens Per Second	15003	[1]
Has Elapsed Time Seconds	54.6	[1]
Has Embedding Dimension	768	[1]
Has Final Gpu Active Gb	1.01	[1]
Has Final Gpu Cache Gb	12.49	[1]
Has Final Gpu Peak Gb	12.52	[1]
Has Final Loss	3.1058	[1]
Has Final Process Rss Gb	3.19	[1]
Has Gpu Active Gb at Iteration10	1	[1]
Has Gpu Active Gb at Iteration100	1	[1]
Has Gpu Active Gb at Iteration20	1	[1]
Has Gpu Active Gb at Iteration30	1	[1]
Has Gpu Active Gb at Iteration40	1	[1]
Has Gpu Active Gb at Iteration50	1	[1]
Has Gpu Active Gb at Iteration60	1	[1]
Has Gpu Active Gb at Iteration70	1	[1]
Has Gpu Active Gb at Iteration80	1	[1]
Has Gpu Active Gb at Iteration90	1	[1]
Has Gpu Cache Gb at Iteration10	12.5	[1]
Has Gpu Cache Gb at Iteration100	12.5	[1]
Has Gpu Cache Gb at Iteration20	12.5	[1]
Has Gpu Cache Gb at Iteration30	12.5	[1]
Has Gpu Cache Gb at Iteration40	12.5	[1]
Has Gpu Cache Gb at Iteration50	12.5	[1]
Has Gpu Cache Gb at Iteration60	12.5	[1]
Has Gpu Cache Gb at Iteration70	12.5	[1]
Has Gpu Cache Gb at Iteration80	12.5	[1]
Has Gpu Cache Gb at Iteration90	12.5	[1]
Has Gpu Peak Gb at Iteration10	12.5	[1]
Has Gpu Peak Gb at Iteration100	12.5	[1]
Has Gpu Peak Gb at Iteration20	12.5	[1]
Has Gpu Peak Gb at Iteration30	12.5	[1]
Has Gpu Peak Gb at Iteration40	12.5	[1]
Has Gpu Peak Gb at Iteration50	12.5	[1]
Has Gpu Peak Gb at Iteration60	12.5	[1]
Has Gpu Peak Gb at Iteration70	12.5	[1]
Has Gpu Peak Gb at Iteration80	12.5	[1]
Has Gpu Peak Gb at Iteration90	12.5	[1]
Has Higher Average Tokens Per Second Than	Grouped8 Training	[1]
Has Learning Rate at Iteration10	0.0003	[1]
Has Learning Rate at Iteration100	0.00003	[1]
Has Learning Rate at Iteration20	0.000292	[1]
Has Learning Rate at Iteration30	0.000268	[1]
Has Learning Rate at Iteration40	0.000232	[1]
Has Learning Rate at Iteration50	0.000188	[1]
Has Learning Rate at Iteration60	0.000142	[1]
Has Learning Rate at Iteration70	0.0000975	[1]
Has Learning Rate at Iteration80	0.0000616	[1]
Has Learning Rate at Iteration90	0.0000381	[1]
Has Loss at Iteration10	4.8802	[1]
Has Loss at Iteration100	3.1058	[1]
Has Loss at Iteration20	3.3704	[1]
Has Loss at Iteration30	3.175	[1]
Has Loss at Iteration40	3.184	[1]
Has Loss at Iteration50	3.1859	[1]
Has Loss at Iteration60	3.1927	[1]
Has Loss at Iteration70	3.179	[1]
Has Loss at Iteration80	3.1503	[1]
Has Loss at Iteration90	3.1571	[1]
Has Lower Final Loss Than	Grouped8 Training	[1]
Has Lower Peak Gpu Memory Than	Grouped8 Training	[1]
Has Num Attention Heads	12	[1]
Has Num Layers	12	[1]
Has Process Memory Gb at Iteration10	3.2	[1]
Has Process Memory Gb at Iteration100	3.2	[1]
Has Process Memory Gb at Iteration20	3.2	[1]
Has Process Memory Gb at Iteration30	3.2	[1]
Has Process Memory Gb at Iteration40	3.2	[1]
Has Process Memory Gb at Iteration50	3.2	[1]
Has Process Memory Gb at Iteration60	3.2	[1]
Has Process Memory Gb at Iteration70	3.2	[1]
Has Process Memory Gb at Iteration80	3.2	[1]
Has Process Memory Gb at Iteration90	3.2	[1]
Has Sequence Length	2048	[1]
Has Shorter Elapsed Time Than	Grouped8 Training	[1]
Has Superior Final Loss	Grouped8 Training	[1]
Has Tokens Per Second at Iteration10	15019	[1]
Has Tokens Per Second at Iteration100	15067	[1]
Has Tokens Per Second at Iteration20	15176	[1]
Has Tokens Per Second at Iteration30	15222	[1]
Has Tokens Per Second at Iteration40	15180	[1]
Has Tokens Per Second at Iteration50	15128	[1]
Has Tokens Per Second at Iteration60	15094	[1]
Has Tokens Per Second at Iteration70	15065	[1]
Has Tokens Per Second at Iteration80	15067	[1]
Has Tokens Per Second at Iteration90	15098	[1]
Has Total Iterations	100	[1]
Has Total Parameters	86825760	[1]
Has Total Tokens Processed	819200	[1]
Has Training Status	complete	[1]
Implies Batch Size	4	[1]
Learning Rate Decreases Monotonically	Lr Schedule	[1]
Loaded Num Sequences	5000	[1]
Loaded Sequence Length	2048	[1]
Maintains Constant Gpu Active Memory	1	[1]
Maintains Constant Process Memory	3.2	[1]
Presupposes Gpu Acceleration	true	[1]
Presupposes No Errors Occurred	true	[1]
Trains Model	Harmonic Gpt	[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.

101 undated

—

exhibitsLossImprovementOverIterationsblah/watt-activation/part-113

true

—

hasAverageTokensPerSecondblah/watt-activation/part-113

15003

—

hasElapsedTimeSecondsblah/watt-activation/part-113

54.6

—

hasEmbeddingDimensionblah/watt-activation/part-113

768

—

hasFinalGpuActiveGbblah/watt-activation/part-113

1.01

—

hasFinalGpuCacheGbblah/watt-activation/part-113

12.49

—

hasFinalGpuPeakGbblah/watt-activation/part-113

12.52

—

hasFinalLossblah/watt-activation/part-113

3.1058

—

hasFinalProcessRssGbblah/watt-activation/part-113

3.19

—

hasGpuActiveGbAtIteration10blah/watt-activation/part-113

—

hasGpuActiveGbAtIteration100blah/watt-activation/part-113

—

hasGpuActiveGbAtIteration20blah/watt-activation/part-113

—

hasGpuActiveGbAtIteration30blah/watt-activation/part-113

—

hasGpuActiveGbAtIteration40blah/watt-activation/part-113

—

hasGpuActiveGbAtIteration50blah/watt-activation/part-113

—

hasGpuActiveGbAtIteration60blah/watt-activation/part-113

—

hasGpuActiveGbAtIteration70blah/watt-activation/part-113

—

hasGpuActiveGbAtIteration80blah/watt-activation/part-113

—

hasGpuActiveGbAtIteration90blah/watt-activation/part-113

—

hasGpuCacheGbAtIteration10blah/watt-activation/part-113

12.5

—

hasGpuCacheGbAtIteration100blah/watt-activation/part-113

12.5

—

hasGpuCacheGbAtIteration20blah/watt-activation/part-113

12.5

—

hasGpuCacheGbAtIteration30blah/watt-activation/part-113

12.5

—

hasGpuCacheGbAtIteration40blah/watt-activation/part-113

12.5

—

hasGpuCacheGbAtIteration50blah/watt-activation/part-113

12.5

—

hasGpuCacheGbAtIteration60blah/watt-activation/part-113

12.5

—

hasGpuCacheGbAtIteration70blah/watt-activation/part-113

12.5

—

hasGpuCacheGbAtIteration80blah/watt-activation/part-113

12.5

—

hasGpuCacheGbAtIteration90blah/watt-activation/part-113

12.5

—

hasGpuPeakGbAtIteration10blah/watt-activation/part-113

12.5

—

hasGpuPeakGbAtIteration100blah/watt-activation/part-113

12.5

—

hasGpuPeakGbAtIteration20blah/watt-activation/part-113

12.5

—

hasGpuPeakGbAtIteration30blah/watt-activation/part-113

12.5

—

hasGpuPeakGbAtIteration40blah/watt-activation/part-113

12.5

—

hasGpuPeakGbAtIteration50blah/watt-activation/part-113

12.5

—

hasGpuPeakGbAtIteration60blah/watt-activation/part-113

12.5

—

hasGpuPeakGbAtIteration70blah/watt-activation/part-113

12.5

—

hasGpuPeakGbAtIteration80blah/watt-activation/part-113

12.5

—

hasGpuPeakGbAtIteration90blah/watt-activation/part-113

12.5

—

hasHigherAverageTokensPerSecondThanblah/watt-activation/part-113

ex:grouped8-training

—

hasLearningRateAtIteration10blah/watt-activation/part-113

0.0003

—

hasLearningRateAtIteration100blah/watt-activation/part-113

0.00003

—

hasLearningRateAtIteration20blah/watt-activation/part-113

0.000292

—

hasLearningRateAtIteration30blah/watt-activation/part-113

0.000268

—

hasLearningRateAtIteration40blah/watt-activation/part-113

0.000232

—

hasLearningRateAtIteration50blah/watt-activation/part-113

0.000188

—

hasLearningRateAtIteration60blah/watt-activation/part-113

0.000142

—

hasLearningRateAtIteration70blah/watt-activation/part-113

0.0000975

—

hasLearningRateAtIteration80blah/watt-activation/part-113

0.0000616

—

hasLearningRateAtIteration90blah/watt-activation/part-113

0.0000381

—

hasLossAtIteration10blah/watt-activation/part-113

4.8802

—

hasLossAtIteration100blah/watt-activation/part-113

3.1058

—

hasLossAtIteration20blah/watt-activation/part-113

3.3704

—

hasLossAtIteration30blah/watt-activation/part-113

3.175

—

hasLossAtIteration40blah/watt-activation/part-113

3.184

—

hasLossAtIteration50blah/watt-activation/part-113

3.1859

—

hasLossAtIteration60blah/watt-activation/part-113

3.1927

—

hasLossAtIteration70blah/watt-activation/part-113

3.179

—

hasLossAtIteration80blah/watt-activation/part-113

3.1503

—

hasLossAtIteration90blah/watt-activation/part-113

3.1571

—

hasLowerFinalLossThanblah/watt-activation/part-113

ex:grouped8-training

—

hasLowerPeakGpuMemoryThanblah/watt-activation/part-113

ex:grouped8-training

—

hasNumAttentionHeadsblah/watt-activation/part-113

—

hasNumLayersblah/watt-activation/part-113

—

hasProcessMemoryGbAtIteration10blah/watt-activation/part-113

3.2

—

hasProcessMemoryGbAtIteration100blah/watt-activation/part-113

3.2

—

hasProcessMemoryGbAtIteration20blah/watt-activation/part-113

3.2

—

hasProcessMemoryGbAtIteration30blah/watt-activation/part-113

3.2

—

hasProcessMemoryGbAtIteration40blah/watt-activation/part-113

3.2

—

hasProcessMemoryGbAtIteration50blah/watt-activation/part-113

3.2

—

hasProcessMemoryGbAtIteration60blah/watt-activation/part-113

3.2

—

hasProcessMemoryGbAtIteration70blah/watt-activation/part-113

3.2

—

hasProcessMemoryGbAtIteration80blah/watt-activation/part-113

3.2

—

hasProcessMemoryGbAtIteration90blah/watt-activation/part-113

3.2

—

hasSequenceLengthblah/watt-activation/part-113

2048

—

hasShorterElapsedTimeThanblah/watt-activation/part-113

ex:grouped8-training

—

hasSuperiorFinalLossblah/watt-activation/part-113

ex:grouped8-training

—

hasTokensPerSecondAtIteration10blah/watt-activation/part-113

15019

—

hasTokensPerSecondAtIteration100blah/watt-activation/part-113

15067

—

hasTokensPerSecondAtIteration20blah/watt-activation/part-113

15176

—

hasTokensPerSecondAtIteration30blah/watt-activation/part-113

15222

—

hasTokensPerSecondAtIteration40blah/watt-activation/part-113

15180

—

hasTokensPerSecondAtIteration50blah/watt-activation/part-113

15128

—

hasTokensPerSecondAtIteration60blah/watt-activation/part-113

15094

—

hasTokensPerSecondAtIteration70blah/watt-activation/part-113

15065

—

hasTokensPerSecondAtIteration80blah/watt-activation/part-113

15067

—

hasTokensPerSecondAtIteration90blah/watt-activation/part-113

15098

—

hasTotalIterationsblah/watt-activation/part-113

100

—

hasTotalParametersblah/watt-activation/part-113

86825760

—

hasTotalTokensProcessedblah/watt-activation/part-113

819200

—

hasTrainingStatusblah/watt-activation/part-113

complete

—

impliesBatchSizeblah/watt-activation/part-113

—

learningRateDecreasesMonotonicallyblah/watt-activation/part-113

ex:lr-schedule

—

loadedNumSequencesblah/watt-activation/part-113

5000

—

loadedSequenceLengthblah/watt-activation/part-113

2048

—

maintainsConstantGpuActiveMemoryblah/watt-activation/part-113

—

maintainsConstantProcessMemoryblah/watt-activation/part-113

3.2

—

presupposesGpuAccelerationblah/watt-activation/part-113

true

—

presupposesNoErrorsOccurredblah/watt-activation/part-113

true

—

trainsModelblah/watt-activation/part-113

ex:harmonic-gpt

—

usesAttentionMechanismblah/watt-activation/part-113

ex:grouped4-attention

References (1)

1 references

[1]Part 113101 facts
ctx:discord/blah/watt-activation/part-113

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.

Grouped4 Training

Inbound mentions (1)

reportsTrainingMetricsReports Training Metrics(1)

Other facts (101)

Timeline

References (1)

See also

Keep researching

`reportsTrainingMetrics`Reports Training Metrics(1)