Step 500

as donto believed on

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

Step 500 has 35 facts recorded in Dontopedia across 9 references, with 4 live disagreements.

35 facts·24 predicates·9 sources·4 in dispute

Mostly:has loss(5), has ppl(4), has tok per s(2)

Inbound mentions (17)

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.

`at`At(2)

Exposure Weight Snapshotex:exposure-weight-snapshot
Second Checkpointex:second-checkpoint

`continuingToward`Continuing Toward(1)

Xenonfunex:xenonfun

`convergedToDcDominance`Converged to DC Dominance(1)

All Blocksex:all-blocks

`dippedAtStep`Dipped at Step(1)

Final Delta Gateex:final-delta-gate

`dominanceByStep`Dominance by Step(1)

Mode 0ex:mode-0

`expectsDoReMiReactionBy`Expects Do Re Mi Reaction by(1)

Xenonfunex:xenonfun

`followsSequence`Follows Sequence(1)

Training Stepsex:training-steps

`fromStep`From Step(1)

Checkpoint Savedex:checkpoint-saved

`hasGlobalSyncDominanceByEnd`Has Global Sync Dominance by End(1)

Trainingex:training

`hasMember`Has Member(1)

Training Checkpointsex:training-checkpoints

`occursAt`Occurs at(1)

First Generation Swapex:first-generation-swap

`precedes`Precedes(1)

Step 200ex:step-200

`precedesTemporally`Precedes Temporally(1)

Step 1ex:step-1

`presupposesStepsUpTo500`Presupposes Steps Up To500(1)

Trainingex:training

`shiftedModeAtStep`Shifted Mode at Step(1)

Block 3ex:block-3

`shiftedToMode0Dc`Shifted to Mode0 DC(1)

Block 3ex:block-3

Other facts (34)

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.

34 facts

Predicate	Value	Ref
Has Loss	8.2263	[1]
Has Loss	5.6428	[2]
Has Loss	16.811	[3]
Has Loss	8.2888	[4]
Has Loss	8.5692	[8]
Has Ppl	3737.9	[1]
Has Ppl	282.3	[2]
Has Ppl	19995808.8	[3]
Has Ppl	3979.1	[4]
Has Tok Per S	17195	[3]
Has Tok Per S	8580	[4]
Has Elapsed Time	252s	[4]
Has Elapsed Time	147	[8]
Step Number	500	[7]
Step Number	500	[8]
Eta	169min	[1]
Tok Per S	13038	[1]
Step Time	628ms	[1]
Progress	3.0%	[1]
Has Lr	1.00e-04	[1]
Progress Fraction	0.653	[2]
Learning Rate	0.00002	[2]
Eta Minutes	250	[2]
Epoch Progress	ep01/30	[2]
Tokens Per Second	12253	[2]
Step Time Ms	669	[2]
Elapsed	130	[3]
Has Attn R	0.6695	[4]
Has Delta Gate	0.077	[5]
Provides Exposure Snapshot	true	[6]
Part of	Training Run 20260311 001319	[8]
Has Perplexity	5266.9	[8]
Has Tokens Per Second	14696	[8]
Rdf:type	Training Step	[9]

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.

35 undated

—

hasPplblah/watt-activation/part-125

3737.9

—

etablah/watt-activation/part-125

169min

—

tokPerSblah/watt-activation/part-125

13038

—

stepTimeblah/watt-activation/part-125

628ms

—

hasLossblah/watt-activation/part-125

8.2263

—

progressblah/watt-activation/part-125

3.0%

—

hasLrblah/watt-activation/part-125

1.00e-04

—

progressFractionblah/watt-activation/part-164

0.653

—

learningRateblah/watt-activation/part-164

0.00002

—

hasPplblah/watt-activation/part-164

282.3

—

hasLossblah/watt-activation/part-164

5.6428

—

etaMinutesblah/watt-activation/part-164

250

—

epochProgressblah/watt-activation/part-164

ep01/30

—

tokensPerSecondblah/watt-activation/part-164

12253

—

stepTimeMsblah/watt-activation/part-164

669

—

hasPplblah/watt-activation/part-189

19995808.8

—

hasLossblah/watt-activation/part-189

16.811

—

elapsedblah/watt-activation/part-189

130

—

hasTokPerSblah/watt-activation/part-189

17195

—

hasLossblah/watt-activation/part-202

8.2888

—

hasAttnRblah/watt-activation/part-202

0.6695

—

hasTokPerSblah/watt-activation/part-202

8580

—

hasPplblah/watt-activation/part-202

3979.1

—

hasElapsedTimeblah/watt-activation/part-202

252s

—

hasDeltaGateblah/watt-activation/part-687

0.077

—

providesExposureSnapshotblah/watt-activation/part-705

true

—

stepNumberblah/watt-activation/13

500

—

labelblah/watt-activation/214

Step 500

—

stepNumberblah/watt-activation/214

500

—

partOfblah/watt-activation/214

ex:training-run-20260311-001319

—

hasLossblah/watt-activation/214

8.5692

—

hasPerplexityblah/watt-activation/214

5266.9

—

hasTokensPerSecondblah/watt-activation/214

14696

—

hasElapsedTimeblah/watt-activation/214

147

—

typeblah/watt-activation/349

ex:TrainingStep

References (9)

9 references

[1]Part 1257 facts
ctx:discord/blah/watt-activation/part-125
[2]Part 1648 facts
ctx:discord/blah/watt-activation/part-164
[3]Part 1894 facts
ctx:discord/blah/watt-activation/part-189
[4]Part 2025 facts
ctx:discord/blah/watt-activation/part-202
[5]Part 6871 fact
ctx:discord/blah/watt-activation/part-687
[6]Part 7051 fact
ctx:discord/blah/watt-activation/part-705
[7]131 fact
ctx:discord/blah/watt-activation/13
- full textwatt-activation-13
  text/plain2 KBdoc:agent/watt-activation-13/8fb2ebd5-d74a-4511-9dac-1a7047d803b2
  Show excerpt
  [2026-02-28 18:28] xenonfun: 2026-02-28 13:23:39,591 [INFO] cross_species_1b.finetune: Step 10/1000: train_loss=7.9112, lr=2.20e-06, 1051 tok/s **1000 steps ≈ 8.5 hours.** [2026-02-28 18:31] xenonfun: (files: Screenshot_2026-02-28_at_…
[8]2147 facts
ctx:discord/blah/watt-activation/214
- full textwatt-activation-214
  text/plain2 KBdoc:agent/watt-activation-214/6ebedb80-7ad3-4d5e-8432-4c87da7e9ad5
  Show excerpt
  [2026-03-11 04:17] xenonfun: ``` Model: lohe_spherical+lohe_v3 768d 12L 12H vocab=100277 Steps: 1000 BS=4 SEQ=1024 (4.1M tokens/run) Phase metrics sidecar → logs/phase_metrics_lohe_spherical_20260311_001319.jsonl ====================…
[9]3491 fact
ctx:discord/blah/watt-activation/349
- full textwatt-activation-349
  text/plain3 KBdoc:agent/watt-activation-349/b02a3c1e-b327-4be5-9f3f-470e78edfa36
  Show excerpt
  [2026-03-16 15:58] xenonfun: ``` Block 3 mode shift: At step 1, blk3 was mode1-dominant (0.243). By step 500, it shifted to mode0 (DC). All blocks converged to DC dominance by step 500 — global sync won over higher harmonics. Block 0 DC…

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.

Step 500

Inbound mentions (17)

atAt(2)

continuingTowardContinuing Toward(1)

convergedToDcDominanceConverged to DC Dominance(1)

dippedAtStepDipped at Step(1)

dominanceByStepDominance by Step(1)

expectsDoReMiReactionByExpects Do Re Mi Reaction by(1)

followsSequenceFollows Sequence(1)

fromStepFrom Step(1)

hasGlobalSyncDominanceByEndHas Global Sync Dominance by End(1)

hasMemberHas Member(1)

occursAtOccurs at(1)

precedesPrecedes(1)

precedesTemporallyPrecedes Temporally(1)

presupposesStepsUpTo500Presupposes Steps Up To500(1)

shiftedModeAtStepShifted Mode at Step(1)

shiftedToMode0DcShifted to Mode0 DC(1)

Other facts (34)

Timeline

References (9)

See also

Keep researching

`at`At(2)

`continuingToward`Continuing Toward(1)

`convergedToDcDominance`Converged to DC Dominance(1)

`dippedAtStep`Dipped at Step(1)

`dominanceByStep`Dominance by Step(1)

`expectsDoReMiReactionBy`Expects Do Re Mi Reaction by(1)

`followsSequence`Follows Sequence(1)

`fromStep`From Step(1)

`hasGlobalSyncDominanceByEnd`Has Global Sync Dominance by End(1)

`hasMember`Has Member(1)

`occursAt`Occurs at(1)

`precedes`Precedes(1)

`precedesTemporally`Precedes Temporally(1)

`presupposesStepsUpTo500`Presupposes Steps Up To500(1)

`shiftedModeAtStep`Shifted Mode at Step(1)

`shiftedToMode0Dc`Shifted to Mode0 DC(1)