Training Session

as donto believed on

From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-06-17.)

Training Session has 78 facts recorded in Dontopedia across 14 references, with 5 live disagreements.

78 facts·72 predicates·14 sources·5 in dispute

Mostly:uses dataset(3), has ppl(2), has total steps(2)

Inbound mentions (10)

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.

`announcesProgress`Announces Progress(1)

Xenonfunex:xenonfun

`announcesTrainingFinished`Announces Training Finished(1)

Message 2026 03 09 18 39 Training Finishedex:message-2026-03-09-18-39-training-finished

`exhibitsHighPerplexity`Exhibits High Perplexity(1)

Garble Response 1ex:garble-response-1

`isModelName`Is Model Name(1)

Akan Gpt2 50kex:akan-gpt2-50k

`marksEndOf`Marks End of(1)

Log Entry 5ex:log-entry-5

`measuredPowerOf`Measured Power of(1)

Device Ups Unitex:device-ups-unit

`measuresModelPerformance`Measures Model Performance(1)

Pplex:ppl

`measuresModelQuality`Measures Model Quality(1)

Ppl Metricex:ppl-metric

`partOf`Part of(1)

Training Step 16600ex:training-step-16600

`subEventOf`Sub Event of(1)

Log Entry 2ex:log-entry-2

Other facts (78)

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.

78 facts

Predicate	Value	Ref
Uses Dataset	fineweb_e2	[9]
Uses Dataset	Fine Web	[12]
Uses Dataset	Tiny Stories	[12]
Has Ppl	348.2	[9]
Has Ppl	83.4	[10]
Has Total Steps	16684	[9]
Has Total Steps	10432	[10]
Involves Device	Device M3	[11]
Involves Device	Device Amd 3060	[11]
Rdf:type	Educational Activity	[13]
Rdf:type	Training Plan	[14]
Follows	Experiment 16	[1]
Has Val Loss	4.5415	[1]
Has Vocab Size	2048	[1]
Indicates Poor Performance	true	[1]
Loaded Epoch	16	[1]
Targets Total Iterations	50000	[2]
Uses Mlx Framework	null	[2]
Trains Language Model	null	[2]
Got Killed at Iteration	8700	[3]
Active Memory Mb	2449	[4]
Peak Memory Mb	24057	[4]
Final Ppl	78.1	[4]
Throughput Tokens Per Second	7400	[4]
Throughput Iterations Per Second	1.8	[4]
Best Loss	3.6875	[4]
Duration Minutes	232	[4]
Duration Seconds	13919	[4]
Final Average Loss	4.3577	[4]
Trains Model	Akan Gpt2 50k	[4]
Has High Perplexity	true	[5]
Progresses Sequentially	Steps	[5]
Resumed From	8k Checkpoint	[5]
Fine Tuning on	Fineweb E2	[6]
Loaded Weights From	Step 002000	[6]
Recovering Quickly	Xenonfun	[6]
Starting at Step	0	[6]
Loading Checkpoint	Tinystories Step 002000	[6]
Ongoing at Step6300	Eta 115min	[7]
Is Running	true	[8]
Depends on	Model Safetensors	[8]
Has Val Ppl	345.9	[9]
Achieved New Best Ppl	344.8	[9]
Has Average Tok Per S	11949	[9]
Has Current Step	16600	[9]
Has Elapsed Time Min	190.6	[9]
Has Eta	1min	[9]
Has Goal of	Minimizing Ppl	[9]
Has Learning Rate	0.00000272	[9]
Has Loss	5.8529	[9]
Has Previous Best Val Ppl	345.9	[9]
Has Progress Percentage	99.5	[9]
Has Step Time Ms	670	[9]
Has Tokens Per Second	12233	[9]
Has Total Tokens	136675328	[9]
Improved Ppl	Previous Best	[9]
Known to Have Low Ppl	344.8	[9]
Nears Completion	99 Point 5 Percent	[9]
Presupposes Ongoing	Gpu Load	[9]
Saved Checkpoint to	checkpoints/fineweb_e2/step_016684	[9]
Has Text Steps	3477	[10]
Is Efficient	20.6 min	[10]
Achieved Completion	true	[10]
Has Best Val Ppl	83.4	[10]
Has Elapsed Time	20.6	[10]
Precedes	Model Inference	[10]
Has Final Val Loss	4.4236	[10]
Has Token Speed	70787	[10]
Reports Metrics	Final Validation Loss	[10]
Power Draw	500	[11]
Power Approximation	~500W	[11]
Aimed at	Team Members	[13]
Intended Outcome	Familiarity With Features	[13]
Has Warm Up	Warm Up 10 15 Minutes	[14]
Has Shooting Drills	Shooting Drills 20 30 Minutes	[14]
Has Dribbling Drills	Dribbling Drills 20 30 Minutes	[14]
Has Cool Down	Cool Down 10 15 Minutes	[14]
Total Duration	60-90 minutes	[14]

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.

78 undated

—

followsblah/watt-activation/part-6

ex:experiment-16

—

hasValLossblah/watt-activation/part-6

4.5415

—

hasVocabSizeblah/watt-activation/part-6

2048

—

indicatesPoorPerformanceblah/watt-activation/part-6

true

—

loadedEpochblah/watt-activation/part-6

—

targetsTotalIterationsblah/watt-activation/part-20

50000

—

usesMLXFrameworkblah/watt-activation/part-20

null

—

trainsLanguageModelblah/watt-activation/part-20

null

—

gotKilledAtIterationblah/watt-activation/part-33

8700

—

activeMemoryMBblah/watt-activation/part-100

2449

—

peakMemoryMBblah/watt-activation/part-100

24057

—

finalPPLblah/watt-activation/part-100

78.1

—

throughputTokensPerSecondblah/watt-activation/part-100

7400

—

throughputIterationsPerSecondblah/watt-activation/part-100

1.8

—

bestLossblah/watt-activation/part-100

3.6875

—

durationMinutesblah/watt-activation/part-100

232

—

durationSecondsblah/watt-activation/part-100

13919

—

finalAverageLossblah/watt-activation/part-100

4.3577

—

trainsModelblah/watt-activation/part-100

ex:akan-gpt2-50k

—

hasHighPerplexityblah/watt-activation/part-134

true

—

progressesSequentiallyblah/watt-activation/part-134

ex:steps

—

resumedFromblah/watt-activation/part-134

ex:8k-checkpoint

—

fineTuningOnblah/watt-activation/part-147

ex:fineweb-e2

—

loadedWeightsFromblah/watt-activation/part-147

ex:step-002000

—

recoveringQuicklyblah/watt-activation/part-147

ex:xenonfun

—

startingAtStepblah/watt-activation/part-147

—

loadingCheckpointblah/watt-activation/part-147

ex:tinystories-step-002000

—

ongoingAtStep6300blah/watt-activation/part-158

ex:eta-115min

—

isRunningblah/watt-activation/part-164

true

—

dependsOnblah/watt-activation/part-164

ex:model-safetensors

—

hasValPplblah/watt-activation/part-161

345.9

—

achievedNewBestPplblah/watt-activation/part-161

344.8

—

hasAverageTokPerSblah/watt-activation/part-161

11949

—

hasCurrentStepblah/watt-activation/part-161

16600

—

hasElapsedTimeMinblah/watt-activation/part-161

190.6

—

hasEtablah/watt-activation/part-161

1min

—

hasGoalOfblah/watt-activation/part-161

ex:minimizing-ppl

—

hasLearningRateblah/watt-activation/part-161

0.00000272

—

hasLossblah/watt-activation/part-161

5.8529

—

hasPplblah/watt-activation/part-161

348.2

—

hasPreviousBestValPplblah/watt-activation/part-161

345.9

—

hasProgressPercentageblah/watt-activation/part-161

99.5

—

hasStepTimeMsblah/watt-activation/part-161

670

—

hasTokensPerSecondblah/watt-activation/part-161

12233

—

hasTotalStepsblah/watt-activation/part-161

16684

—

hasTotalTokensblah/watt-activation/part-161

136675328

—

improvedPplblah/watt-activation/part-161

ex:previous-best

—

knownToHaveLowPplblah/watt-activation/part-161

344.8

—

nearsCompletionblah/watt-activation/part-161

ex:99-point-5-percent

—

presupposesOngoingblah/watt-activation/part-161

ex:gpu-load

—

savedCheckpointToblah/watt-activation/part-161

checkpoints/fineweb_e2/step_016684

—

usesDatasetblah/watt-activation/part-161

fineweb_e2

—

hasTextStepsblah/watt-activation/part-245

3477

—

isEfficientblah/watt-activation/part-245

20.6 min

—

achievedCompletionblah/watt-activation/part-245

true

—

hasTotalStepsblah/watt-activation/part-245

10432

—

hasBestValPplblah/watt-activation/part-245

83.4

—

hasElapsedTimeblah/watt-activation/part-245

20.6

—

precedesblah/watt-activation/part-245

ex:model-inference

—

hasFinalValLossblah/watt-activation/part-245

4.4236

—

hasTokenSpeedblah/watt-activation/part-245

70787

—

hasPplblah/watt-activation/part-245

83.4

—

reportsMetricsblah/watt-activation/part-245

ex:final-validation-loss

—

involvesDeviceblah/watt-activation/72

ex:device-m3

—

involvesDeviceblah/watt-activation/72

ex:device-amd-3060

—

powerDrawblah/watt-activation/72

500

—

powerApproximationblah/watt-activation/72

~500W

—

usesDatasetblah/watt-activation/166

ex:FineWeb

—

usesDatasetblah/watt-activation/166

ex:TinyStories

—

typebeam/1637051c-3221-4f2c-903f-1bd479158af9

ex:EducationalActivity

—

aimedAtbeam/1637051c-3221-4f2c-903f-1bd479158af9

ex:team-members

—

intendedOutcomebeam/1637051c-3221-4f2c-903f-1bd479158af9

ex:familiarity-with-features

—

typelme/bae9589b-4dae-43e6-a05c-7d40430f5a8b

ex:TrainingPlan

—

hasWarmUplme/bae9589b-4dae-43e6-a05c-7d40430f5a8b

ex:warm-up-10-15-minutes

—

hasShootingDrillslme/bae9589b-4dae-43e6-a05c-7d40430f5a8b

ex:shooting-drills-20-30-minutes

—

hasDribblingDrillslme/bae9589b-4dae-43e6-a05c-7d40430f5a8b

ex:dribbling-drills-20-30-minutes

—

hasCoolDownlme/bae9589b-4dae-43e6-a05c-7d40430f5a8b

ex:cool-down-10-15-minutes

—

totalDurationlme/bae9589b-4dae-43e6-a05c-7d40430f5a8b

60-90 minutes

References (14)

14 references

[1]Part 65 facts
ctx:discord/blah/watt-activation/part-6
[2]Part 203 facts
ctx:discord/blah/watt-activation/part-20
[3]Part 331 fact
ctx:discord/blah/watt-activation/part-33
[4]Part 10010 facts
ctx:discord/blah/watt-activation/part-100
[5]Part 1343 facts
ctx:discord/blah/watt-activation/part-134
[6]Part 1475 facts
ctx:discord/blah/watt-activation/part-147
[7]Part 1581 fact
ctx:discord/blah/watt-activation/part-158
[8]Part 1642 facts
ctx:discord/blah/watt-activation/part-164
[9]Part 16122 facts
ctx:discord/blah/watt-activation/part-161
[10]Part 24511 facts
ctx:discord/blah/watt-activation/part-245
[11]724 facts
ctx:discord/blah/watt-activation/72
- full textwatt-activation-72
  text/plain3 KBdoc:agent/watt-activation-72/be13baaf-0d87-4d0d-9934-7f3b139089f7
  Show excerpt
  [2026-03-07 17:40] xenonfun: All committed. Summary of what's in place: Committed (10 commits on kuramoto-gpt-performance): - NEXT_STEPS.md — detailed 4-phase plan - KURAMOTO_FINDINGS.md — full experimental record - bench_seqlen_sc…
[12]1662 facts
ctx:discord/blah/watt-activation/166
- full textwatt-activation-166
  text/plain3 KBdoc:agent/watt-activation-166/523e76f5-3827-45e2-9027-2ed22e28eb22
  Show excerpt
  [2026-03-09 19:13] xenonfun: step 1500/22920 ep02/30 96.2% loss=5.3484 ppl= 210.3 lr=1.98e-05 667ms 12,290tok/s eta=238min VAL step 1500 loss=5.3808 ppl=217.2 ★ best ckpt step 1500 → step_001500 ``` Instruction: 'God said?…
[13]1637051c 3221 4f2c 903f 1bd479158af93 facts
ctx:claims/beam/1637051c-3221-4f2c-903f-1bd479158af9
[14]Bae9589b 4dae 43e6 A05c 7d40430f5a8b6 facts
ctx:claims/lme/bae9589b-4dae-43e6-a05c-7d40430f5a8b
- full textbeam-chunk
  text/plain19 KBdoc:beam/bae9589b-4dae-43e6-a05c-7d40430f5a8b
  Show excerpt
  [Session date: 2023/06/17 (Sat) 07:39] User: I'm looking for some tips on injury prevention and recovery strategies for soccer players. I participate in the company's annual charity soccer tournament today, and I want to make sure I'm takin…

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 Session

Inbound mentions (10)

announcesProgressAnnounces Progress(1)

announcesTrainingFinishedAnnounces Training Finished(1)

exhibitsHighPerplexityExhibits High Perplexity(1)

isModelNameIs Model Name(1)

marksEndOfMarks End of(1)

measuredPowerOfMeasured Power of(1)

measuresModelPerformanceMeasures Model Performance(1)

measuresModelQualityMeasures Model Quality(1)

partOfPart of(1)

subEventOfSub Event of(1)

Other facts (78)

Timeline

References (14)

See also

Keep researching

`announcesProgress`Announces Progress(1)

`announcesTrainingFinished`Announces Training Finished(1)

`exhibitsHighPerplexity`Exhibits High Perplexity(1)

`isModelName`Is Model Name(1)

`marksEndOf`Marks End of(1)

`measuredPowerOf`Measured Power of(1)

`measuresModelPerformance`Measures Model Performance(1)

`measuresModelQuality`Measures Model Quality(1)

`partOf`Part of(1)

`subEventOf`Sub Event of(1)