Training Log

as donto believed on

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

Training Log has 85 facts recorded in Dontopedia across 25 references, with 4 live disagreements.

85 facts·78 predicates·25 sources·4 in dispute

Mostly:includes step(5), has tokens per second(2), shows increasing(2)

Inbound mentions (8)

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.

`basedOnLog`Based on Log(1)

Training Bpbex:training-bpb

`containsLog`Contains Log(1)

Message 2026 02 26 21 39ex:message-2026-02-26-21-39

`containsTrainingLog`Contains Training Log(1)

Msg 2026 03 15 22 32ex:msg-2026-03-15-22-32

`expressesInterestIn`Expresses Interest in(1)

Lisamegawattsex:lisamegawatts

`includesTrainingLog`Includes Training Log(1)

2026 03 13 14 46 Messageex:2026-03-13-14-46-message

`performsSpeechActOfReporting`Performs Speech Act of Reporting(1)

Xenonfunex:xenonfun

`performsStatusUpdate`Performs Status Update(1)

Foxhopex:foxhop

`reportsProgress`Reports Progress(1)

Xenonfunex:xenonfun

Other facts (85)

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.

85 facts

Predicate	Value	Ref
Includes Step	Training Step 120	[16]
Includes Step	Training Step 160	[16]
Includes Step	Training Step 40	[16]
Includes Step	Training Step 80	[16]
Includes Step	Training Step 0	[16]
Has Tokens Per Second	36000	[5]
Has Tokens Per Second	12,281tok/s	[10]
Shows Increasing	R	[16]
Shows Increasing	C Bits	[16]
Rdf:type	Log Entry	[24]
Rdf:type	Log File	[25]
Frames Results As Successful Early Stop	✓ Saved	[1]
Indicates Ongoing Learning	Chatty V8	[2]
Shows Evaluation	Evaluation Iter 5000	[3]
Is Real Time Log	null	[4]
Has Percent Complete	100	[5]
Has Avg50 Loss	7.7189	[5]
Has Avg50 Ppl	2250.37	[5]
Has Iterations Per Second	140.4	[5]
Has Loss	7.734	[5]
Has Lr	0.0001	[5]
Reports Iter	500	[5]
Has Best Loss	6.8625	[5]
Has Ppl	2284.72	[5]
Has Eta	0m 0s	[5]
Part of	Xenonfun Message 1207	[6]
Is From Real Run	{}	[7]
Emphasizes Progress	Iteration Sequence	[8]
Is in Genre	ml-training-output	[8]
Shows Degradation	Loss	[9]
Has Ppl Value	423.7	[10]
Indicates Ongoing Training	ML Model Training	[10]
Has Speed Ms	667ms	[10]
Has Step	1000/16684	[10]
Has Eta	174min	[10]
Has Loss Value	6.0491	[10]
Has Lr Value	4.98e-05	[10]
Has Progress Percent	6.0%	[10]
Reports Metrics	Current Training Run	[11]
Has Avg Tok Per Sec	11648	[12]
Follows Code Generation Tests	null	[12]
Achieves New Best	Ppl 137.8	[12]
Has Best Val Ppl	138.2	[12]
Has Elapsed Time Min	78.1	[12]
Has Final Ppl	137.8	[12]
Has Final Val Loss	4.9257	[12]
Improves Over Previous	Best Val Ppl 138.2	[12]
Indicates Poor Performance	High Ppl	[12]
Processed Tokens	54607872	[12]
Saves Final Checkpoint	Checkpoints Instruct Step 010166	[12]
Demonstrates Viability	null	[13]
Provides Evidence	Training Progress	[14]
Implies Ongoing Training	true	[15]
Has Warmup Steps	50	[15]
Validates Every Steps	200	[15]
Saves Every Steps	200	[15]
Has Total Steps	622	[15]
Has Lite Phase	false	[15]
Has Learning Rate	1e-04	[15]
Has Checkpoint Dir	checkpoints/byte_patch_bs64_phase_only_epoch1	[15]
Logs Every Steps	25	[15]
Phases Every	1	[15]
Has Batch Size	64	[15]
Has Sequence Length	2048	[15]
Reports Temperature	0.34	[17]
Is Quoted	Xenonfun Message	[17]
Reports Bpb	5.739	[17]
Reports C	2.7b	[17]
Reports DC	0.08	[17]
Reports Eta	8min	[17]
Reports Lr	2.70e-03	[17]
Reports Progress	25.0%	[17]
Reports R	0.263	[17]
Reports Snr	-11.3dB	[17]
Reports Step	500	[17]
Reports Tok Per S	49436	[17]
Reports Val Bpb	5.732	[17]
Belongs to Genre	ML training output	[18]
Uses Optimizer	Bitsandbytes Adamw8bit	[19]
Presupposes Ongoing Training	True	[20]
References File	Phase Metrics File	[21]
Genre Is Training Progress Log	ML Training Log Format	[22]
Has Validation Loss	4.5415	[23]
Current Step	8000	[24]
Progress Percentage	80	[24]

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.

85 undated

—

framesResultsAsSuccessfulEarlyStopblah/safiersemantics/part-72

✓ Saved

—

indicatesOngoingLearningblah/unturf/part-69

ex:chatty-v8

—

showsEvaluationblah/watt-activation/part-18

ex:evaluation-iter-5000

—

isRealTimeLogblah/watt-activation/part-20

null

—

hasPercentCompleteblah/watt-activation/part-42

100

—

hasAvg50Lossblah/watt-activation/part-42

7.7189

—

hasAvg50PPLblah/watt-activation/part-42

2250.37

—

hasIterationsPerSecondblah/watt-activation/part-42

140.4

—

hasLossblah/watt-activation/part-42

7.734

—

hasLRblah/watt-activation/part-42

0.0001

—

reportsIterblah/watt-activation/part-42

500

—

hasBestLossblah/watt-activation/part-42

6.8625

—

hasPPLblah/watt-activation/part-42

2284.72

—

hasTokensPerSecondblah/watt-activation/part-42

36000

—

hasETAblah/watt-activation/part-42

0m 0s

—

partOfblah/watt-activation/part-100

ex:xenonfun-message-1207

—

isFromRealRunblah/watt-activation/part-125

{}

—

emphasizesProgressblah/watt-activation/part-114

ex:iteration-sequence

—

isInGenreblah/watt-activation/part-114

ml-training-output

—

showsDegradationblah/watt-activation/part-136

ex:loss

—

hasPplValueblah/watt-activation/part-149

423.7

—

indicatesOngoingTrainingblah/watt-activation/part-149

ex:ml-model-training

—

hasSpeedMsblah/watt-activation/part-149

667ms

—

hasStepblah/watt-activation/part-149

1000/16684

—

hasTokensPerSecondblah/watt-activation/part-149

12,281tok/s

—

hasEtablah/watt-activation/part-149

174min

—

hasLossValueblah/watt-activation/part-149

6.0491

—

hasLrValueblah/watt-activation/part-149

4.98e-05

—

hasProgressPercentblah/watt-activation/part-149

6.0%

—

reportsMetricsblah/watt-activation/part-153

ex:current-training-run

—

hasAvgTokPerSecblah/watt-activation/part-171

11648

—

followsCodeGenerationTestsblah/watt-activation/part-171

null

—

achievesNewBestblah/watt-activation/part-171

ex:ppl-137.8

—

hasBestValPplblah/watt-activation/part-171

138.2

—

hasElapsedTimeMinblah/watt-activation/part-171

78.1

—

hasFinalPplblah/watt-activation/part-171

137.8

—

hasFinalValLossblah/watt-activation/part-171

4.9257

—

improvesOverPreviousblah/watt-activation/part-171

ex:best-val-ppl-138.2

—

indicatesPoorPerformanceblah/watt-activation/part-171

ex:high-ppl

—

processedTokensblah/watt-activation/part-171

54607872

—

savesFinalCheckpointblah/watt-activation/part-171

ex:checkpoints-instruct-step-010166

—

demonstratesViabilityblah/watt-activation/part-266

null

—

providesEvidenceblah/watt-activation/part-246

ex:training-progress

—

impliesOngoingTrainingblah/watt-activation/part-268

true

—

hasWarmupStepsblah/watt-activation/part-268

—

validatesEveryStepsblah/watt-activation/part-268

200

—

savesEveryStepsblah/watt-activation/part-268

200

—

hasTotalStepsblah/watt-activation/part-268

622

—

hasLitePhaseblah/watt-activation/part-268

false

—

hasLearningRateblah/watt-activation/part-268

1e-04

—

hasCheckpointDirblah/watt-activation/part-268

checkpoints/byte_patch_bs64_phase_only_epoch1

—

logsEveryStepsblah/watt-activation/part-268

—

phasesEveryblah/watt-activation/part-268

—

hasBatchSizeblah/watt-activation/part-268

—

hasSequenceLengthblah/watt-activation/part-268

2048

—