Training run anchor_G16_20K

as donto believed on

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

Training run anchor_G16_20K has 590 facts recorded in Dontopedia across 99 references, with 56 live disagreements.

590 facts·441 predicates·99 sources·56 in dispute

Mostly:has total steps(17), rdf:type(17), total steps(12)

Has Total Stepsin dispute`hasTotalSteps`

2000^[8]all time · Part 3
16670^[23]all time · Part 130
16670^[24]all time · Part 133
16670^[26]all time · Part 140
16684^[27]all time · Part 150
16684^[28]all time · Part 155
12432^[38]all time · Part 239
6216^[40]all time · Part 238
12432^[43]all time · Part 247
20000^[47]all time · Part 331

Rdf:typein dispute`rdf:type`

Process^[73]all time · 41
Training Process^[75]all time · 44
Training Run^[79]all time · 125
Machine Learning Training^[80]all time · 127
Process^[81]all time · 161
Process^[82]all time · 158
Training Run^[83]all time · 165
Model Run^[85]all time · 209
Machine Learning Training^[86]all time · 242
Process^[88]all time · 251

Total Stepsin dispute`totalSteps`

16670^[21]all time · Part 125
16670^[25]all time · Part 136
1000^[31]all time · Part 210
1000^[36]all time · Part 230
20000^[48]all time · Part 336
20000^[50]all time · Part 372
16670^[80]sourceall time · 127
16684^[82]sourceall time · 158
1000^[84]sourceall time · 189
1000^[85]sourceall time · 209

Inbound mentions (75)

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.

`belongsTo`Belongs to(7)

Step 2000ex:step-2000
Step 2100ex:step-2100
Step 2200ex:step-2200
Step 2300ex:step-2300
Step 2400ex:step-2400
Step 2500ex:step-2500
Validation Step 2000ex:validation-step-2000

`partOfTraining`Part of Training(3)

Step 100ex:step-100
Step 150ex:step-150
Step 50ex:step-50

`hasNoNaN`Has No Na N(2)

D Muonex:d-muon
Unified Muonex:unified-muon

`potentiallyUsedIn`Potentially Used in(2)

Adam Optimizerex:adam-optimizer
Cosine Learning Curveex:cosine-learning-curve

`reportsProgress`Reports Progress(2)

Xenonfunex:xenonfun
Xenonfunex:xenonfun

`achievedLowestBPB`Achieved Lowest Bpb(1)

Best Checkpointex:best-checkpoint

`activeMidTraining`Active Mid Training(1)

Kuramoto Phase Updatesex:kuramoto-phase-updates

`advocatesContinuation`Advocates Continuation(1)

Xenonfunex:xenonfun

`aimsAtLowPerplexity`Aims at Low Perplexity(1)

Training Runex:training-run

`announcesCompletion`Announces Completion(1)

Xenonfunex:xenonfun

`announcesDecisionToStop`Announces Decision to Stop(1)

Entry 2026 03 09 05 20ex:entry-2026-03-09-05-20

`announcesSection`Announces Section(1)

Omega Botex:omega-bot

`announcesTrainingComplete`Announces Training Complete(1)

Message 2026 04 23 21 08ex:message-2026-04-23-21-08

`buildsInParallel`Builds in Parallel(1)

Lora Adapterex:lora-adapter

`commentedOn`Commented on(1)

Xenonfunex:xenonfun

`considersAchievable`Considers Achievable(1)

Xenonfunex:xenonfun

`correlatesWithLoss`Correlates With Loss(1)

Pplex:ppl

`decreasesOverTimeIdeally`Decreases Over Time Ideally(1)

Lossex:loss

`demonstratesLearningProgress`Demonstrates Learning Progress(1)

Microgpt Scaledex:microgpt-scaled

`demonstratesOptimizations`Demonstrates Optimizations(1)

Xenonfunex:xenonfun

`describes`Describes(1)

Model Noise Observationex:model-noise-observation

`didNotKillNarrativeCapability`Did Not Kill Narrative Capability(1)

Stage 3 Scienceex:stage-3-science

`existentiallyCommitted`Existentially Committed(1)

Entropy Calculationex:entropy-calculation

`expandedLexiconOf`Expanded Lexicon of(1)

Stage 3 Scienceex:stage-3-science

`expressesFrustrationWithHungRun`Expresses Frustration With Hung Run(1)

Foxhopex:foxhop

`fired`Fired(1)

Gelation Signalsex:gelation-signals

`generatedByText`Generated by Text(1)

Final Generation Sampleex:final-generation-sample

`hasBestLossSoFar`Has Best Loss So Far(1)

Val Step 8000ex:val-step-8000

`hasNoProblem`Has No Problem(1)

Rotadamwex:rotadamw

`hasUntappedPotential`Has Untapped Potential(1)

Modelex:model

`indicatesFailure`Indicates Failure(1)

Nanex:nan

`intendsToLaunch`Intends to Launch(1)

Xenonfunex:xenonfun

`interpretsAsStableLearning`Interprets As Stable Learning(1)

Xenonfunex:xenonfun

`isDatasetFor`Is Dataset for(1)

Shakespeareex:shakespeare

`isEnabledPostOptimization`Is Enabled Post Optimization(1)

Bs4ex:bs4

`isMinimumLossSoFar`Is Minimum Loss So Far(1)

Best Lossex:best-loss

`isPartOf`Is Part of(1)

Training Step 25ex:training-step-25

`isPrimaryMetric`Is Primary Metric(1)

Lossex:loss

`isTrainedEntity`Is Trained Entity(1)

Lohe Spherical Modelex:lohe-spherical-model

`isWatchingTrainingRun`Is Watching Training Run(1)

Foxhopex:foxhop

`knowsBs4Capability`Knows Bs4 Capability(1)

Xenonfunex:xenonfun

`lacksProblems`Lacks Problems(1)

Rotadamwex:rotadamw

`monitorsTraining`Monitors Training(1)

Xenonfunex:xenonfun

`neverBeaten`Never Beaten(1)

Best Lossex:best-loss

`optimizesTraining`Optimizes Training(1)

Mx Compileex:mx-compile

`partOf`Part of(1)

Training Config 20000 Stepsex:training-config-20000-steps

`performanceOf`Performance of(1)

Holdout Bpb Metricex:holdout-bpb-metric

`performsStatusUpdate`Performs Status Update(1)

Logex:log

`plansRestart`Plans Restart(1)

Xenonfunex:xenonfun

`postsTrainingLog`Posts Training Log(1)

Xenonfunex:xenonfun

`presupposesExistenceOf`Presupposes Existence of(1)

Textex:text

`presupposesLogicFixesNeeded`Presupposes Logic Fixes Needed(1)

Stop Messageex:stop-message

`presupposesOngoing`Presupposes Ongoing(1)

Training Runex:training-run

`presupposesOngoingProcess`Presupposes Ongoing Process(1)

ML Trainingex:ml-training

`reportsContinuingFullEpoch`Reports Continuing Full Epoch(1)

Xenonfunex:xenonfun

`reportsStatusOf`Reports Status of(1)

Message 2026 03 09 16 32ex:message-2026-03-09-16-32

`reportsTrainingProgress`Reports Training Progress(1)

Entry 2026 03 09 05 08ex:entry-2026-03-09-05-08

`reworked`Reworked(1)

Larger Runex:larger-run

`setEarly`Set Early(1)

Best Lossex:best-loss

`sharesProgressUpdate`Shares Progress Update(1)

Xenonfunex:xenonfun

`sharesProgressUpdates`Shares Progress Updates(1)

Xenonfunex:xenonfun

`showcasesQuality`Showcases Quality(1)

Message 2026 04 19 16 48ex:message-2026-04-19-16-48

`terminatedTraining`Terminated Training(1)

Ajaxdavisex:ajaxdavis

`wasUsedDuring`Was Used During(1)

Monitorex:monitor

Other facts (543)

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.

543 facts

Predicate	Value	Ref
Has Sequence Length	1024	[34]
Has Sequence Length	2048	[47]
Has Sequence Length	512	[48]
Has Sequence Length	8192	[58]
Has Sequence Length	2048	[86]
Has Sequence Length	128	[95]
Has Batch Size	4	[34]
Has Batch Size	16	[40]
Has Batch Size	8	[47]
Has Batch Size	64	[48]
Has Batch Size	64	[58]
Has Batch Size	8	[86]
Has Step	4000	[51]
Has Step	25/1245	[56]
Has Step	1000	[68]
Has Step	100	[85]
Has Step	200	[85]
Has Step	300	[85]
Has Learning Rate	0.0001	[40]
Has Learning Rate	3e-04	[47]
Has Learning Rate	0.0000129	[82]
Has Learning Rate	0.0001	[86]
Has Learning Rate	0.0005	[95]
Has Metric Value	9.3	[90]
Has Metric Value	55.7	[90]
Has Metric Value	61	[90]
Has Metric Value	64.4	[90]
Has Metric Value	67.8	[90]
Has Bpb Value	3.583	[90]
Has Bpb Value	3.328	[90]
Has Bpb Value	3.183	[90]
Has Bpb Value	3.096	[90]
Has Bpb Value	3.05	[90]
Has Warmup Steps	500	[40]
Has Warmup Steps	500	[47]
Has Warmup Steps	50	[58]
Has Warmup Steps	300	[86]
Has Loss	1.494	[48]
Has Loss	4.9122	[56]
Has Loss	0.3609	[63]
Has Loss	0.9332	[68]
Time Elapsed	32s	[85]
Time Elapsed	64s	[85]
Time Elapsed	97s	[85]
Time Elapsed	1.7	[88]
Presupposes Ongoing	Language Model	[10]
Presupposes Ongoing	Training Run	[22]
Presupposes Ongoing	null	[52]
Has Best Loss	1.5656	[15]
Has Best Loss	6.04	[75]
Has Best Loss	3.6777	[76]
Is Complete	50,000/50,000 iters	[15]
Is Complete	true	[38]
Is Complete	Yes	[61]
Has Elapsed Time	~2.5 hours	[18]
Has Elapsed Time	268s	[33]
Has Elapsed Time	303s	[33]
Is Ongoing	At Step 6000	[23]
Is Ongoing	Training	[28]
Is Ongoing	true	[47]
Has Ppl at Step	2216.6	[33]
Has Ppl at Step	2789.1	[33]
Has Ppl at Step	2732.6	[33]
Has Tok Per S	12074	[33]
Has Tok Per S	12062	[33]
Has Tok Per S	13611	[63]
Has Lite Phase	false	[40]
Has Lite Phase	false	[47]
Has Lite Phase	false	[86]
Tokens Per Second	98782	[48]
Tokens Per Second	13027	[85]
Tokens Per Second	13186	[85]
Has Lr	0.0000943	[48]
Has Lr	0.0000916	[51]
Has Lr	9.00e-05	[63]
Batch Size	64	[50]
Batch Size	4	[80]
Batch Size	64	[92]
Sequence Length	256	[50]
Sequence Length	2048	[80]
Sequence Length	256	[92]
Has Bpb	2.444	[51]
Has Bpb	0.521	[63]
Has Bpb	1.3463	[68]
Loss at Step	11.6429	[85]
Loss at Step	11.7602	[85]
Loss at Step	11.5419	[85]
Ppl at Step	113884.2	[85]
Ppl at Step	128058.6	[85]
Ppl at Step	102939.2	[85]
Sys R at Step	0.345	[85]
Sys R at Step	0.331	[85]
Sys R at Step	0.348	[85]
Has Modality	Text Modality	[86]
Has Modality	Image Modality	[86]
Has Modality	Audio Modality	[86]
Lacks Observation of	35K-step cliff	[97]
Lacks Observation of	gnorm explosion	[97]
Lacks Observation of	catastrophic forgetting	[97]
Has Final Val Loss	3.7805	[3]

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.

590 undated

—

existsPubliclyblah/general/part-119

ex:training-ai-unturf-com-dashboard

—

resemblesStockMarketTickerblah/general/part-119

ex:stock-market-ticker

—

packsMultipleblah/random/part-37

ex:experiments

—

tookTotalTimeblah/safiersemantics/part-72

25529.5s

—

triggeredEarlyStoppingAfterblah/safiersemantics/part-72

30 eval intervals

—

achievedBestValLossblah/safiersemantics/part-72

3.7141

—

averagedTimePerIterblah/safiersemantics/part-72

1955ms/iter

—

startedWithLossblah/safiersemantics/part-72

7.3126

—

hasFinalTrainLossblah/safiersemantics/part-72

3.6974

—

hasFinalValLossblah/safiersemantics/part-72

3.7805

—

savedCheckpointblah/safiersemantics/part-72

ex:checkpoint-bin-iter-12300

—

savedBestCheckpointblah/safiersemantics/part-72

ex:checkpoint-best-bin

—

improvedLossFromInitialblah/safiersemantics/part-72

7.3126 to 3.6974

—

hasNumIterationsblah/training-and-evals/part-17

5000

—

terminatedBecauseblah/training-and-evals/part-17

ex:planning-optimizations

—

sometimesFailsblah/training-and-evals/part-34

ex:ajaxdavis-training

—

isUncertainInFrequencyblah/training-and-evals/part-34

ex:sometimes

—

existsAndHungblah/unturf/part-67

ex:hung-training-run

—

completesSuccessfullyblah/vidya/part-4

ex:training-step-2000

—

hasTotalStepsblah/vidya/part-3

2000

—

hasDurationSecondsblah/vidya/part-3

143.74

—

isSuccessfulblah/vidya/part-3

ex:microgpt-scaled

—

reportsProgressPeriodicallyblah/vidya/part-3

ex:training-steps

—

causedInferenceblah/vidya/part-3

ex:inference-section

—

completedAllStepsblah/vidya/part-3

null

—

printedTotalTimeblah/vidya/part-3

143.74s

—

presupposesExistenceOfblah/watt-activation/part-9

ex:gelation-signals

—

hasAttachmentblah/watt-activation/part-21

image.png

—

presupposesOngoingblah/watt-activation/part-21

ex:language-model

—

achievesPeakSpeedsblah/watt-activation/part-21

ex:xenonfun

—

presupposesHighPerformanceblah/watt-activation/part-21

ex:optimizations

—

maintainsHighTokensPerSecondblah/watt-activation/part-21

33600

—

commitsTo100000Iterationsblah/watt-activation/part-21

100000

—

decreasesETAOverTimeblah/watt-activation/part-21

ex:iteration-54600

—

hasBestKnownLossblah/watt-activation/part-21

2.6401

—

nearsCompletionblah/watt-activation/part-21

23m

—

enablesBs4blah/watt-activation/part-21

ex:xenonfun

—

improvesLossFromblah/watt-activation/part-21

3.0021

—

resumesWithOptimizationsblah/watt-activation/part-21

ex:xenonfun

—

etablah/watt-activation/part-32

~41 min left

—

contaminatedHeavilyblah/watt-activation/part-32

ex:wikipedia-factoid

—

vocabSizeblah/watt-activation/part-32

—

philosophicalSignalStrengthblah/watt-activation/part-32

weak

—

realTestIsblah/watt-activation/part-32

ex:fine-tuned-version

—

onTrackToReachblah/watt-activation/part-32

ex:v7-loss-territory

—

bestLossblah/watt-activation/part-32

3.22

—

totalTargetItersblah/watt-activation/part-32

100000

—

reachedItersblah/watt-activation/part-32

57K

—

corpusContaminationblah/watt-activation/part-32

ex:wikipedia-factoid

—

onTrackblah/watt-activation/part-32

true

—

totalItersblah/watt-activation/part-39

139000

—

presupposesOngoingStatusblah/watt-activation/part-35

ex:true

—

hasTimeLeftblah/watt-activation/part-35

~30 min

—

presupposesExistenceOfblah/watt-activation/part-44

ex:kuramoto-model

—

hasBetterBestLossThanblah/watt-activation/part-33

ex:raw-v8k

—

hasBestLossblah/watt-activation/part-33

1.5656

—

showsMassiveImprovementblah/watt-activation/part-33

null

—

achievedBestLossAtIterationblah/watt-activation/part-33

29630

—

isCompleteblah/watt-activation/part-33

50,000/50,000 iters

—

presupposesSuccessblah/watt-activation/part-33

null

—

targetsPhilosophyFinetuningblah/watt-activation/part-33

null

—

hasIterationsblah/watt-activation/part-58

20000

—

currentlyOngoingblah/watt-activation/part-94

true

—

achievesNewBestPPLblah/watt-activation/part-91

ex:at-16000

—

doesNotResumeIteratorStateblah/watt-activation/part-91

ex:true

—

usesOptimizerblah/watt-activation/part-91

ex:adam

—

hasPeakMemoryblah/watt-activation/part-91

24GB

—

recoveredPastPreviousBestblah/watt-activation/part-91

ex:at-iter-13000

—

hasZeroMemoryDriftblah/watt-activation/part-91

ex:true

—

resumesFromblah/watt-activation/part-91

ex:iter-10000

—

resumesModelWeightsblah/watt-activation/part-91

ex:true

—

experiencedHitOnResumeblah/watt-activation/part-91

ex:true

—

hasIterationSpeedblah/watt-activation/part-91

1.9

—

hasGlobalItersblah/watt-activation/part-91

16000

—

showsSteadyImprovementblah/watt-activation/part-91

ex:true

—

isSteadilyPushingLowerblah/watt-activation/part-91

ex:loss

—

maintainsStableThroughputblah/watt-activation/part-91

ex:true

—

hasElapsedTimeblah/watt-activation/part-91

~2.5 hours

—

hasCompletedItersSinceResumeblah/watt-activation/part-91

6000

—

hasCompletedItersblah/watt-activation/part-91

10000

—

hasCompletedEpochsblah/watt-activation/part-91

0.31

—

occursOnMacStudioblah/watt-activation/part-98

ex:implied

—

presupposesOngoingStatusblah/watt-activation/part-95

null

—

teleologicallyAimsForEpochCompletionblah/watt-activation/part-95

null

—

hasRemainingTimeblah/watt-activation/part-95

—

isApproachingblah/watt-activation/part-95

ex:epoch-1

—

totalStepsblah/watt-activation/part-125

16670

—

hasDataPositionblah/watt-activation/part-127

16392000

—

presupposesOngoingblah/watt-activation/part-127

ex:training-run

—

progressesFromblah/watt-activation/part-127

ex:step-2000

—

loadingStepblah/watt-activation/part-127

ex:step-002000

—

aimsAtLowPerplexityblah/watt-activation/part-127

ex:training-run

—

resumedFromStepblah/watt-activation/part-127

2000

—

hasTeleologicalGoalblah/watt-activation/part-130

ex:reduce-loss

—

occursInContextOfblah/watt-activation/part-130

ex:homebrew-llm-training

—

hasTotalStepsblah/watt-activation/part-130

16670

—

isOngoingblah/watt-activation/part-130

ex:at-step-6000

—

usesLearningRateDecayblah/watt-activation/part-133

true

—

showsHighPerplexityblah/watt-activation/part-133

true

—

hasPersistentHighLossblah/watt-activation/part-133

true

—

hasTotalStepsblah/watt-activation/part-133

16670

—

presupposes16670TotalStepsblah/watt-activation/part-133

16670

—

hasCripplingLogicIssuesblah/watt-activation/part-133

ex:model

—

totalStepsblah/watt-activation/part-136

16670

—

usesGradientDescentblah/watt-activation/part-136

null

—

includesTrainingStepblah/watt-activation/part-140

ex:training-step-12100

—

hasTotalStepsblah/watt-activation/part-140

16670

—

completesEpochblah/watt-activation/part-140

ex:epoch-completion

—

includesTrainingStepblah/watt-activation/part-140

ex:training-step-12000

—

hasTotalTokensProcessedblah/watt-activation/part-140

54640640

—

hasElapsedTimeMinutesblah/watt-activation/part-140

75.9

—

hasPriorBestValidationPplblah/watt-activation/part-140

421.1

—

hasAverageTokensPerSecondblah/watt-activation/part-140

12001

—

usesCl100kTokenizerblah/watt-activation/part-140

cl100k

—

includesValidationStepblah/watt-activation/part-140

ex:validation-step-12000

—

nearingEarlyStageCompletionblah/watt-activation/part-150

null

—

usesLlamaStyleTokensblah/watt-activation/part-150

null

—

showsDecreasingLossTrendblah/watt-activation/part-150

null

—

reportedByblah/watt-activation/part-150

ex:xenonfun

—

hasStablePerformanceblah/watt-activation/part-150

null

—

hasTotalStepsblah/watt-activation/part-150

16684

—

inProgressblah/watt-activation/part-150

null

—

knownToHaveValidationblah/watt-activation/part-150

null

—

isOngoingblah/watt-activation/part-155

ex:training

—

achievesMilestoneblah/watt-activation/part-155

ex:lr-reduced

—

reachedFirstLrReductionblah/watt-activation/part-155

ex:lr-reduced-phase

—

showsDecreasingLossblah/watt-activation/part-155

ex:loss-trend

—

hasTotalStepsblah/watt-activation/part-155

16684

—

needsMoreTimeblah/watt-activation/part-169

{}

—

catchesUpOnTokensblah/watt-activation/part-169

{}

—

undergoesCosineDecayblah/watt-activation/part-169

{}

—

learnsSlowlyblah/watt-activation/part-169

{}

—

computesTokensSeenblah/watt-activation/part-169

~82M

—

hasStepsblah/watt-activation/part-169

6000

—

isInEpochblah/watt-activation/part-169

—

willFinishSoonblah/watt-activation/part-169

{}

—

hasCurrentLearningRateblah/watt-activation/part-169

8e-06

—

nearsCompletionblah/watt-activation/part-169

{}

—

producesCheckpointsblah/watt-activation/part-169

ex:step-006000

—

hasTokensPerSecondblah/watt-activation/part-169

12000

—

hasTimeRemainingblah/watt-activation/part-169

~33 min

—

trainsOnDatablah/watt-activation/part-169

ex:instruction-data

—

continuesRunningblah/watt-activation/part-169

chugging

—

hasEstimatedTimeRemainingblah/watt-activation/part-169

~42 min

—

occursOverblah/watt-activation/part-201

100 steps

—

hasStep100blah/watt-activation/part-210

ex:true

—

totalStepsblah/watt-activation/part-210

1000

—

presupposesComparisonToSpectralblah/watt-activation/part-202

true

—

hasPplAtStepblah/watt-activation/part-212

2216.6

—

hasTokPerSblah/watt-activation/part-212

12074

—

hasTokPerSblah/watt-activation/part-212

12062

—

hasPplAtStepblah/watt-activation/part-212

2789.1

—

hasPplAtStepblah/watt-activation/part-212

2732.6

—

nearingCompletionblah/watt-activation/part-212

800/1000

—

showsImprovementblah/watt-activation/part-212

ex:loss-decrease

—

hasValPplAtStepblah/watt-activation/part-212

2514.5

—

hasValLossAtStepblah/watt-activation/part-212

7.8298

—

hasValLossAtStepblah/watt-activation/part-212

7.913

—

hasElapsedTimeblah/watt-activation/part-212

268s

—

hasElapsedTimeblah/watt-activation/part-212

303s

—

hasLossAtStepblah/watt-activation/part-212

7.9335

—

hasLossAtStepblah/watt-activation/part-212

7.7037

—

hasSequenceLengthblah/watt-activation/part-211

1024

—

processesTokensblah/watt-activation/part-211

4.1M

—

hasBatchSizeblah/watt-activation/part-211

—

hasStepsblah/watt-activation/part-211

1000

—

usesRotationalAdamWblah/watt-activation/part-217

null

—

temporalOrderblah/watt-activation/part-230

ex:step-50

—

usesOptimizerblah/watt-activation/part-230

ex:rotational-adamw

—

causesElapsedTimeblah/watt-activation/part-230

ex:step-150

—

presupposesGpuUsageblah/watt-activation/part-230

true

—

totalStepsblah/watt-activation/part-230

1000

—

totalSteps12432blah/watt-activation/part-244

null

—

improvesOverTimeblah/watt-activation/part-239

true

—

hasBestValPplblah/watt-activation/part-239

—

hasElapsedTimeMinblah/watt-activation/part-239

46.4

—

hasAvgTokPerSecblah/watt-activation/part-239

73085

—

savedFinalCheckpointAtblah/watt-activation/part-239

checkpoints/bpe8k_lohe_spherical/step_012432

—

processedTotalTokensblah/watt-activation/part-239

203685888

—

achievedNewBestPplblah/watt-activation/part-239

96.3

—

hasTotalStepsblah/watt-activation/part-239

12432

—

isCompleteblah/watt-activation/part-239

true

—

hasFinalPplblah/watt-activation/part-239

96.3

—

hasEtablah/watt-activation/part-239

6min

—

hasFinalValLossblah/watt-activation/part-239

4.5679

—

willRunForblah/watt-activation/part-231

another hour and a half

—

toReachblah/watt-activation/part-231

10K

—

ongoingblah/watt-activation/part-231

ex:current-run

—

shouldWriteblah/watt-activation/part-231

ex:checkpoint

—

expectsToReachblah/watt-activation/part-231

ex:10k-steps

—

hasTotalStepsblah/watt-activation/part-238

6216

—

droppedBatchSizeToblah/watt-activation/part-238

—

logsEveryblah/watt-activation/part-238

—

hasSeqLenblah/watt-activation/part-238

2048

—

logsPhaseMetricsblah/watt-activation/part-238

ex:phase-metrics-sidecar

—

assumesGpuUsageblah/watt-activation/part-238

ex:gpu

—

hasBatchSizeblah/watt-activation/part-238

—

validatesEveryblah/watt-activation/part-238

1000

—

hasWarmupStepsblah/watt-activation/part-238

500

—

savesEveryblah/watt-activation/part-238

1000

References (99)

99 references

[1]Part 1192 facts
ctx:discord/blah/general/part-119
[2]Part 371 fact
ctx:discord/blah/random/part-37
[3]Part 7210 facts
ctx:discord/blah/safiersemantics/part-72
[4]Part 172 facts
ctx:discord/blah/training-and-evals/part-17
[5]Part 342 facts
ctx:discord/blah/training-and-evals/part-34
[6]Part 671 fact
ctx:discord/blah/unturf/part-67
[7]Part 41 fact
ctx:discord/blah/vidya/part-4
[8]Part 37 facts
ctx:discord/blah/vidya/part-3
[9]Part 91 fact
ctx:discord/blah/watt-activation/part-9
[10]Part 2112 facts
ctx:discord/blah/watt-activation/part-21
[11]Part 3211 facts
ctx:discord/blah/watt-activation/part-32
[12]Part 391 fact
ctx:discord/blah/watt-activation/part-39
[13]Part 352 facts
ctx:discord/blah/watt-activation/part-35
[14]Part 441 fact
ctx:discord/blah/watt-activation/part-44
[15]Part 337 facts
ctx:discord/blah/watt-activation/part-33
[16]Part 581 fact
ctx:discord/blah/watt-activation/part-58
[17]Part 941 fact
ctx:discord/blah/watt-activation/part-94
[18]Part 9118 facts
ctx:discord/blah/watt-activation/part-91
[19]Part 981 fact
ctx:discord/blah/watt-activation/part-98
[20]Part 954 facts
ctx:discord/blah/watt-activation/part-95
[21]Part 1251 fact
ctx:discord/blah/watt-activation/part-125
[22]Part 1276 facts
ctx:discord/blah/watt-activation/part-127
[23]Part 1304 facts
ctx:discord/blah/watt-activation/part-130
[24]Part 1336 facts
ctx:discord/blah/watt-activation/part-133
[25]Part 1362 facts
ctx:discord/blah/watt-activation/part-136
[26]Part 14010 facts
ctx:discord/blah/watt-activation/part-140
[27]Part 1508 facts
ctx:discord/blah/watt-activation/part-150
[28]Part 1555 facts
ctx:discord/blah/watt-activation/part-155
[29]Part 16916 facts
ctx:discord/blah/watt-activation/part-169
[30]Part 2011 fact
ctx:discord/blah/watt-activation/part-201
[31]Part 2102 facts
ctx:discord/blah/watt-activation/part-210
[32]Part 2021 fact
ctx:discord/blah/watt-activation/part-202
[33]Part 21214 facts
ctx:discord/blah/watt-activation/part-212
[34]Part 2114 facts
ctx:discord/blah/watt-activation/part-211
[35]Part 2171 fact
ctx:discord/blah/watt-activation/part-217
[36]Part 2305 facts
ctx:discord/blah/watt-activation/part-230
[37]Part 2441 fact
ctx:discord/blah/watt-activation/part-244
[38]Part 23912 facts
ctx:discord/blah/watt-activation/part-239
[39]Part 2315 facts
ctx:discord/blah/watt-activation/part-231
[40]Part 23817 facts
ctx:discord/blah/watt-activation/part-238
[41]Part 2523 facts
ctx:discord/blah/watt-activation/part-252
[42]Part 2531 fact
ctx:discord/blah/watt-activation/part-253
[43]Part 2476 facts
ctx:discord/blah/watt-activation/part-247
[44]Part 2641 fact
ctx:discord/blah/watt-activation/part-264
[45]Part 2676 facts
ctx:discord/blah/watt-activation/part-267
[46]Part 3261 fact
ctx:discord/blah/watt-activation/part-326
[47]Part 33122 facts
ctx:discord/blah/watt-activation/part-331
[48]Part 33614 facts
ctx:discord/blah/watt-activation/part-336
[49]Part 3733 facts
ctx:discord/blah/watt-activation/part-373
[50]Part 3729 facts
ctx:discord/blah/watt-activation/part-372
[51]Part 37818 facts
ctx:discord/blah/watt-activation/part-378
[52]Part 3773 facts
ctx:discord/blah/watt-activation/part-377
[53]Part 3962 facts
ctx:discord/blah/watt-activation/part-396
[54]Part 3971 fact
ctx:discord/blah/watt-activation/part-397
[55]Part 3983 facts
ctx:discord/blah/watt-activation/part-398
[56]Part 41715 facts
ctx:discord/blah/watt-activation/part-417
[57]Part 4211 fact
ctx:discord/blah/watt-activation/part-421
[58]Part 4196 facts
ctx:discord/blah/watt-activation/part-419
[59]Part 4621 fact
ctx:discord/blah/watt-activation/part-462
[60]Part 64516 facts
ctx:discord/blah/watt-activation/part-645
[61]Part 6608 facts
ctx:discord/blah/watt-activation/part-660
[62]Part 6621 fact
ctx:discord/blah/watt-activation/part-662
[63]Part 66112 facts
ctx:discord/blah/watt-activation/part-661
[64]Part 68711 facts
ctx:discord/blah/watt-activation/part-687
[65]Part 7021 fact
ctx:discord/blah/watt-activation/part-702
[66]Part 7032 facts
ctx:discord/blah/watt-activation/part-703
[67]Part 7056 facts
ctx:discord/blah/watt-activation/part-705
[68]Part 7065 facts
ctx:discord/blah/watt-activation/part-706
[69]Part 7143 facts
ctx:discord/blah/watt-activation/part-714
[70]Part 372 facts
ctx:discord/blah/watt-activation/part-37
[71]Part 864 facts
ctx:discord/blah/watt-activation/part-86
[72]Part 2151 fact
ctx:discord/blah/watt-activation/part-215
[73]411 fact
ctx:discord/blah/training-and-evals/41
- full texttraining-and-evals-41
  text/plain3 KBdoc:agent/training-and-evals-41/95b41334-d198-4a88-be0d-bc22c528e602
  Show excerpt
  [2026-03-16 21:05] foxhop.: ● 23.6GB — that's tight on the 4090 (24GB). Batch 16 might OOM. Let me check with batch 8: ● Bash(python3 -c " n_embd=1536; n_layer=16; block_size=1024; total=467466240…) ⎿ batch= 4: 11.5 GB OK ba…
[74]104 facts
ctx:discord/blah/vidya/10
- full textvidya-10
  text/plain3 KBdoc:agent/vidya-10/636e1043-0585-44b0-86c4-ecbe60c83f00
  Show excerpt
  [2026-03-20 11:25] foxhop.: awesome new video card with 12G & over 3k cuda cores! [2026-03-20 11:25] foxhop.: ? [2026-03-20 11:27] foxhop.: "We're building the disk." [2026-03-20 11:28] foxhop.: this screams GPT switch all "the" toward "a" …
[75]4411 facts
ctx:discord/blah/watt-activation/44
- full textwatt-activation-44
  text/plain3 KBdoc:agent/watt-activation-44/35af5294-4efe-43b8-a39d-23df23087144
  Show excerpt
  [2026-03-07 05:32] xenonfun: ``` Not good. The loss trajectory tells a clear story: ┌────────┬─────────────────────────┐ │ Iters │ Avg Loss │ ├────────┼─────────────────────────┤ │ 0-2K │ 6.56 (learning) │…
[76]8712 facts
ctx:discord/blah/watt-activation/87
- full textwatt-activation-87
  text/plain2 KBdoc:agent/watt-activation-87/21549fe5-2355-4b51-9d7e-42d9fb1af8c0
  Show excerpt
  [2026-03-07 21:54] xenonfun: ``` iter 8000/10000 | loss=4.2647 | PPL=71.1 | lr=6.16e-05 | 2.2 it/s (9.1K tok/s) | mem=1848MB peak=20665MB Saved checkpoint: ./akan_gpt2_checkpoints/checkpoint_iter_8000.npz iter 8500/10000 | loss=4.2438…
[77]9119 facts
ctx:discord/blah/watt-activation/91
- full textwatt-activation-91
  text/plain3 KBdoc:agent/watt-activation-91/61aeb5f1-b837-4123-a6c4-b76d719d0b25
  Show excerpt
  [2026-03-08 00:46] xenonfun: ``` Looks good. Now let me calculate the exact iterations and launch: - 133M tokens / 4096 = 32,471 iters per epoch - 2 epochs = 64,942 total iters - Already done: 10,000 - Remaining: 54,942 iters (rou…
[78]1262 facts
ctx:discord/blah/watt-activation/126
- full textwatt-activation-126
  text/plain3 KBdoc:agent/watt-activation-126/dddfc295-807c-4943-b01a-f4f0a977c17e
  Show excerpt
  [2026-03-09 04:03] xenonfun: ### What context count we do at this scale? ⏺ From the measurements we have, memory scales roughly linearly with total tokens in the batch: - BS=4, seq=1024 → 4,096 tokens → ~40 GB - BS=8, seq=1024 → 8,192 …
[79]12511 facts
ctx:discord/blah/watt-activation/125
- full textwatt-activation-125
  text/plain3 KBdoc:agent/watt-activation-125/078b0573-153a-47f9-81de-fbf8dd1915e3
  Show excerpt
  [2026-03-09 03:33] xenonfun: ❯ we want to do 2K seq tho ⏺ Doubling seq doubles the activation memory. BS=8, seq=2048 = same logit tensor size as BS=16, seq=1024 — which hit 85GB. We need to re-check BS. BS=4, seq=2048 = 8,192 tokens/bat…
[80]12711 facts
ctx:discord/blah/watt-activation/127
- full textwatt-activation-127
  text/plain2 KBdoc:agent/watt-activation-127/aedb2d68-0e0d-4e54-ace8-64c39f6403e3
  Show excerpt
  [2026-03-09 04:13] xenonfun: [resume] loading step_002000... resumed at step 2000, data_pos=16,392,000 [train] 16,670 steps | BS=4 SEQ=2048 | LR=1e-04 warmup=500 save every 2000 | val every 2000 | log every 100 checkpoint…
[81]1614 facts
ctx:discord/blah/watt-activation/161
- full textwatt-activation-161
  text/plain2 KBdoc:agent/watt-activation-161/b2429cd0-9f7a-4b1b-847e-3785f26f96b4
  Show excerpt
  [2026-03-09 18:10] xenonfun: ``` Prompt: 'The most important discovery in science was' temp=0.8 top_k=40 stop=<|endoftext|> (100257) [compiled] ──────────────────────────────────────────────────────────── The most important discovery in…
[82]1583 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 …
[83]1653 facts
ctx:discord/blah/watt-activation/165
- full textwatt-activation-165
  text/plain3 KBdoc:agent/watt-activation-165/c02ae72f-c534-45a9-adc3-f12c4275a06b
  Show excerpt
  [2026-03-09 19:01] xenonfun: ``` Mode: qa temp=0.8 top_k=40 stop=<|endoftext|> (100257) [compiled] Instruction: 'What is photosynthesis?' ──────────────────────────────────────────────────────────── What is photosynthesis? The followin…
[84]1891 fact
ctx:discord/blah/watt-activation/189
- full textwatt-activation-189
  text/plain2 KBdoc:agent/watt-activation-189/ee6e7700-8f8f-458c-bd97-cd00204ffe29
  Show excerpt
  [2026-03-10 03:42] xenonfun: ``` What the fix looks like: Coupling κ_g is a scalar per group. Its gradient through the sync step is tractable: at first order, Δcoupling_g ∝ -(∂loss/∂spectra_synced) · (mean_spec_g - spectra_g) — the reado…
[85]20921 facts
ctx:discord/blah/watt-activation/209
- full textwatt-activation-209
  text/plain2 KBdoc:agent/watt-activation-209/d7eabcf9-d506-432f-9a2c-b25cf5ef8ccd
  Show excerpt
  [2026-03-11 03:51] xenonfun: ```# coupling (K) and adjacency are structural constants — not updated. 598 + # v2: update harmonic_coeffs (G×H learned frequency weights) 599 + # v3: update mode_amplitudes (k scalars,…
[86]24217 facts
ctx:discord/blah/watt-activation/242
- full textwatt-activation-242
  text/plain3 KBdoc:agent/watt-activation-242/e65441d2-9807-493e-a096-3ab8edf76fd5
  Show excerpt
  [2026-03-12 04:42] xenonfun: (files: Screenshot_2026-03-12_at_12.41.55_AM.png) [2026-03-12 04:49] xenonfun: `http://phase-pipeline.xenon.fun.local:8000/dashboard` it sets routes like this but if I refresh browser page those don't resolve b…
[87]2464 facts
ctx:discord/blah/watt-activation/246
- full textwatt-activation-246
  text/plain3 KBdoc:agent/watt-activation-246/e98a7e00-4e3f-4e35-bccb-faff629fa820
  Show excerpt
  [2026-03-12 07:06] xenonfun: ```[03:04:05] step 2000/12432 16.1% txt=4.411(82ppl) img=0.000 aud=2.025 r=0.375 lr=4.79e-05 150ms 69,028tok/s [T:1000 I:0 A:1000] eta=26min VAL step 2000 loss=4.4035 ppl=81.7 * best [03:04:29] …
[88]25110 facts
ctx:discord/blah/watt-activation/251
- full textwatt-activation-251
  text/plain1 KBdoc:agent/watt-activation-251/0d79165d-ca43-48df-b924-6b76b157d1a5
  Show excerpt
  [2026-03-12 13:11] xenonfun: ✅ Phase 0 confirmed working — r_global rises monotonically from 0.07 → 0.96 across 16 steps on the production multimodal checkpoint. The architecture supports iterative generation. This is the green light to p…
[89]2645 facts
ctx:discord/blah/watt-activation/264
- full textwatt-activation-264
  text/plain2 KBdoc:agent/watt-activation-264/555cd9a1-321c-4f18-8f17-7bef422894a1
  Show excerpt
  [2026-03-13 05:30] xenonfun: ``` I wrote the full plan in docs/claude/plans/tokenizerless_phase_stream_plan.md. Core recommendation from the plan: - do not do pure one-byte-per-step modeling first - build a tokenizerless byte_patch…
[90]35211 facts
ctx:discord/blah/watt-activation/352
- full textwatt-activation-352
  text/plain2 KBdoc:agent/watt-activation-352/f9fe3319-d5f4-4e70-b415-d397928b4c05
  Show excerpt
  [2026-03-17 06:32] xenonfun: ``` 44 +├── antenna.py # AntennaHarmonicBlock + AntennaLM: field-mediated byte LM 45 +├── antenna_probes.py # Diagnostic probes: impulse, memory, coupling, leakage, boundary 46 +├── an…
[91]3715 facts
ctx:discord/blah/watt-activation/371
- full textwatt-activation-371
  text/plain2 KBdoc:agent/watt-activation-371/9ac097b4-3d34-4fad-bc53-d4097529d8ab
  Show excerpt
  [2026-03-18 17:41] xenonfun: ``` ❯ how expensive are probes? ⏺ Let me check what the probes actually do at the --probe-every 2000 interval. ⏺ Searched for 1 pattern (ctrl+o to expand) ⏺ At --probe-every 2000, each probe step runs: 1. …
[92]3705 facts
ctx:discord/blah/watt-activation/370
- full textwatt-activation-370
  text/plain3 KBdoc:agent/watt-activation-370/319ffc75-f6e3-490e-bf32-ddb97e36692c
  Show excerpt
  [2026-03-18 17:29] xenonfun: ⏺ Here's the status: Implementation complete and pushed: - AnchorBind module in harmonic_mlx/antenna.py — 32 soft anchors on S^{d_bind-1}, softmax assignment, residual fusion - Wired into AntennaHarmonicB…
[93]3753 facts
ctx:discord/blah/watt-activation/375
- full textwatt-activation-375
  text/plain3 KBdoc:agent/watt-activation-375/d401bcdc-528d-4d93-9c65-cd9f177804a5
  Show excerpt
  [2026-03-18 19:30] xenonfun: ``` checkpoint dir: checkpoints/ham_G16_20K ──────────────────────────────────────────────────────────────────────── step 200/20000 1.0% BPB=4.149 r=0.228 SNR=-12.7dB C=0.7b lr=9.95e-05 80,154tok/s …
[94]3961 fact
ctx:discord/blah/watt-activation/396
- full textwatt-activation-396
  text/plain2 KBdoc:agent/watt-activation-396/0cf7c4f0-276b-4e7f-99e8-ba29387cd8d5
  Show excerpt
  [2026-03-19 04:31] xenonfun: ``` Mode: byte-level temp=0.7 max_tokens=300 Prompt: 'The quick brown fox ' ──────────────────────────────────────────────────────────── The quick brown fox ris pis se te at ti odinrbar 0bouone…
[95]48020 facts
ctx:discord/blah/watt-activation/480
- full textwatt-activation-480
  text/plain2 KBdoc:agent/watt-activation-480/f2254c38-e451-4fa5-9d6b-499a16b05da3
  Show excerpt
  [2026-03-21 23:19] xenonfun: ``` === BPE-8K: cache growth 200→2000 === Loaded 20480128 bytes from /Users/ms/MS/HarmonicMLX/data/fineweb_edu_bpe8k.npy KickModel ANALYTICAL training: 30000 steps, seq_len=128, lr=0.0005, momentum=0.9…
[96]6552 facts
ctx:discord/blah/watt-activation/655
- full textwatt-activation-655
  text/plain2 KBdoc:agent/watt-activation-655/184fbb55-c5df-4c57-ba7d-a525dd418a7a
  Show excerpt
  [2026-04-19 09:46] xenonfun: loaded 24999946 bytes from /Users/ms/MS/HarmonicRust/data/curriculum/domain_tinystories.bin step 0 loss=6.1092 bpb=8.814 gnorm=10.6885 lr=4.00e-6 9259 tok/s step 50 loss=2.8955 bpb=4.177 gnorm=2.2132…
[97]6613 facts
ctx:discord/blah/watt-activation/661
- full textwatt-activation-661
  text/plain2 KBdoc:agent/watt-activation-661/b1cbb45f-4fa8-4281-bf4f-564e7c2ebe63
  Show excerpt
  [2026-04-20 00:14] xenonfun: ``` ⏺ Run complete. Task #5 done. Final head-to-head: ┌────────────┬────────────┬─────────────────────────────┐ │ │ V1 rotadam │ V2 fixed │ ├────────────┼────────────┼────…
[98]6845 facts
ctx:discord/blah/watt-activation/684
- full textwatt-activation-684
  text/plain2 KBdoc:agent/watt-activation-684/22acb9fe-dbc9-48ce-9087-3c918e65fec5
  Show excerpt
  [2026-04-23 19:13] xenonfun: ``` Register match: partial. - ✅ tinystories/narrative show real conditioning — "little bear", "mommy", "kitchen", "store", "pigs", simple-declarative register. Recognizable. - ❌ edu/science/dense_science/…
[99]6961 fact
ctx:discord/blah/watt-activation/696
- full textwatt-activation-696
  text/plain2 KBdoc:agent/watt-activation-696/6bc363f6-e780-4242-9e13-32e8d01a4dd8
  Show excerpt
  [2026-05-01 02:47] xenonfun: It wants 150M or so so think the chatgpt-2 in 24hr is looking achievable. we do need to get more dataset here. [2026-05-01 02:50] lisamegawatts: mines downloading datasets and parsing now, added a 4096 option [2…

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.

Training run anchor_G16_20K

Has Total Stepsin disputehasTotalSteps

Rdf:typein disputerdf:type

Total Stepsin disputetotalSteps

Inbound mentions (75)

belongsToBelongs to(7)

partOfTrainingPart of Training(3)

hasNoNaNHas No Na N(2)

potentiallyUsedInPotentially Used in(2)

reportsProgressReports Progress(2)

achievedLowestBPBAchieved Lowest Bpb(1)

activeMidTrainingActive Mid Training(1)

advocatesContinuationAdvocates Continuation(1)

aimsAtLowPerplexityAims at Low Perplexity(1)

announcesCompletionAnnounces Completion(1)

announcesDecisionToStopAnnounces Decision to Stop(1)

announcesSectionAnnounces Section(1)

announcesTrainingCompleteAnnounces Training Complete(1)

buildsInParallelBuilds in Parallel(1)

commentedOnCommented on(1)

considersAchievableConsiders Achievable(1)

correlatesWithLossCorrelates With Loss(1)

decreasesOverTimeIdeallyDecreases Over Time Ideally(1)

demonstratesLearningProgressDemonstrates Learning Progress(1)

demonstratesOptimizationsDemonstrates Optimizations(1)

describesDescribes(1)

didNotKillNarrativeCapabilityDid Not Kill Narrative Capability(1)

existentiallyCommittedExistentially Committed(1)

expandedLexiconOfExpanded Lexicon of(1)

expressesFrustrationWithHungRunExpresses Frustration With Hung Run(1)

firedFired(1)

generatedByTextGenerated by Text(1)

hasBestLossSoFarHas Best Loss So Far(1)

hasNoProblemHas No Problem(1)

hasUntappedPotentialHas Untapped Potential(1)

indicatesFailureIndicates Failure(1)

intendsToLaunchIntends to Launch(1)

interpretsAsStableLearningInterprets As Stable Learning(1)

isDatasetForIs Dataset for(1)

isEnabledPostOptimizationIs Enabled Post Optimization(1)

isMinimumLossSoFarIs Minimum Loss So Far(1)

isPartOfIs Part of(1)

isPrimaryMetricIs Primary Metric(1)

isTrainedEntityIs Trained Entity(1)

isWatchingTrainingRunIs Watching Training Run(1)

knowsBs4CapabilityKnows Bs4 Capability(1)

lacksProblemsLacks Problems(1)

monitorsTrainingMonitors Training(1)

neverBeatenNever Beaten(1)

optimizesTrainingOptimizes Training(1)

partOfPart of(1)

performanceOfPerformance of(1)

performsStatusUpdatePerforms Status Update(1)

plansRestartPlans Restart(1)

postsTrainingLogPosts Training Log(1)

presupposesExistenceOfPresupposes Existence of(1)

presupposesLogicFixesNeededPresupposes Logic Fixes Needed(1)

presupposesOngoingPresupposes Ongoing(1)

presupposesOngoingProcessPresupposes Ongoing Process(1)

reportsContinuingFullEpochReports Continuing Full Epoch(1)

reportsStatusOfReports Status of(1)

reportsTrainingProgressReports Training Progress(1)

reworkedReworked(1)

setEarlySet Early(1)

sharesProgressUpdateShares Progress Update(1)

sharesProgressUpdatesShares Progress Updates(1)

showcasesQualityShowcases Quality(1)

terminatedTrainingTerminated Training(1)

wasUsedDuringWas Used During(1)

Other facts (543)

Timeline

References (99)

See also

Keep researching

Has Total Stepsin dispute`hasTotalSteps`

Rdf:typein dispute`rdf:type`

Total Stepsin dispute`totalSteps`

`belongsTo`Belongs to(7)

`partOfTraining`Part of Training(3)

`hasNoNaN`Has No Na N(2)

`potentiallyUsedIn`Potentially Used in(2)

`reportsProgress`Reports Progress(2)

`achievedLowestBPB`Achieved Lowest Bpb(1)

`activeMidTraining`Active Mid Training(1)

`advocatesContinuation`Advocates Continuation(1)

`aimsAtLowPerplexity`Aims at Low Perplexity(1)

`announcesCompletion`Announces Completion(1)

`announcesDecisionToStop`Announces Decision to Stop(1)

`announcesSection`Announces Section(1)

`announcesTrainingComplete`Announces Training Complete(1)

`buildsInParallel`Builds in Parallel(1)

`commentedOn`Commented on(1)

`considersAchievable`Considers Achievable(1)

`correlatesWithLoss`Correlates With Loss(1)

`decreasesOverTimeIdeally`Decreases Over Time Ideally(1)

`demonstratesLearningProgress`Demonstrates Learning Progress(1)

`demonstratesOptimizations`Demonstrates Optimizations(1)

`describes`Describes(1)

`didNotKillNarrativeCapability`Did Not Kill Narrative Capability(1)

`existentiallyCommitted`Existentially Committed(1)

`expandedLexiconOf`Expanded Lexicon of(1)

`expressesFrustrationWithHungRun`Expresses Frustration With Hung Run(1)

`fired`Fired(1)

`generatedByText`Generated by Text(1)

`hasBestLossSoFar`Has Best Loss So Far(1)

`hasNoProblem`Has No Problem(1)

`hasUntappedPotential`Has Untapped Potential(1)

`indicatesFailure`Indicates Failure(1)

`intendsToLaunch`Intends to Launch(1)

`interpretsAsStableLearning`Interprets As Stable Learning(1)

`isDatasetFor`Is Dataset for(1)

`isEnabledPostOptimization`Is Enabled Post Optimization(1)

`isMinimumLossSoFar`Is Minimum Loss So Far(1)

`isPartOf`Is Part of(1)

`isPrimaryMetric`Is Primary Metric(1)

`isTrainedEntity`Is Trained Entity(1)

`isWatchingTrainingRun`Is Watching Training Run(1)

`knowsBs4Capability`Knows Bs4 Capability(1)

`lacksProblems`Lacks Problems(1)

`monitorsTraining`Monitors Training(1)

`neverBeaten`Never Beaten(1)

`optimizesTraining`Optimizes Training(1)

`partOf`Part of(1)

`performanceOf`Performance of(1)

`performsStatusUpdate`Performs Status Update(1)

`plansRestart`Plans Restart(1)

`postsTrainingLog`Posts Training Log(1)

`presupposesExistenceOf`Presupposes Existence of(1)

`presupposesLogicFixesNeeded`Presupposes Logic Fixes Needed(1)

`presupposesOngoing`Presupposes Ongoing(1)

`presupposesOngoingProcess`Presupposes Ongoing Process(1)

`reportsContinuingFullEpoch`Reports Continuing Full Epoch(1)

`reportsStatusOf`Reports Status of(1)

`reportsTrainingProgress`Reports Training Progress(1)

`reworked`Reworked(1)

`setEarly`Set Early(1)

`sharesProgressUpdate`Shares Progress Update(1)

`sharesProgressUpdates`Shares Progress Updates(1)

`showcasesQuality`Showcases Quality(1)

`terminatedTraining`Terminated Training(1)

`wasUsedDuring`Was Used During(1)