Training Step 2500

as donto believed on

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

Training Step 2500 has 25 facts recorded in Dontopedia across 6 references, with 3 live disagreements.

25 facts·19 predicates·6 sources·3 in dispute

Mostly:has learning rate(3), has loss(2), has progress percentage(2)

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.

`succeeds`Succeeds(1)

Training Step 2400ex:training-step-2400

Other facts (25)

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.

25 facts

Predicate	Value	Ref
Has Learning Rate	0.000048	[2]
Has Learning Rate	0.0000966	[4]
Has Learning Rate	0.000048	[6]
Has Loss	4.7787	[1]
Has Loss	6.7313	[4]
Has Progress Percentage	15	[2]
Has Progress Percentage	15	[4]
Has Loss Value	5.9942	[2]
Has Loss Value	5.9942	[6]
Has Tokens Per Second	12290	[2]
Has Tokens Per Second	13023	[4]
Has Ppl Value	401.1	[2]
Has Eta Minutes	158	[2]
Has Current Step Number	2500	[2]
Has Time Per Step Ms	667	[2]
Saved As New Best	null	[3]
Beats Lohe Delta Best	null	[3]
Has Perplexity	838.2	[4]
Has Estimated Time Remaining Min	149	[4]
Has Step Duration Ms	629	[4]
Loss Value	4.7787	[5]
Step Number	2500	[6]
Progress Percentage	15	[6]
Has Perplexity Value	401.1	[6]
Has Estimated Time Remaining	158min	[6]

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.

25 undated

—

hasLossblah/vidya/part-6

4.7787

—

hasProgressPercentageblah/watt-activation/part-150

—

hasPplValueblah/watt-activation/part-150

401.1

—

hasEtaMinutesblah/watt-activation/part-150

158

—

hasLearningRateblah/watt-activation/part-150

0.000048

—

hasLossValueblah/watt-activation/part-150

5.9942

—

hasCurrentStepNumberblah/watt-activation/part-150

2500

—

hasTimePerStepMsblah/watt-activation/part-150

667

—

hasTokensPerSecondblah/watt-activation/part-150

12290

—

savedAsNewBestblah/watt-activation/part-658

null

—

beatsLoheDeltaBestblah/watt-activation/part-658

null

—

hasPerplexityblah/watt-activation/part-128

838.2

—

hasEstimatedTimeRemainingMinblah/watt-activation/part-128

149

—

hasLearningRateblah/watt-activation/part-128

0.0000966

—

hasLossblah/watt-activation/part-128

6.7313

—

hasProgressPercentageblah/watt-activation/part-128

—

hasStepDurationMsblah/watt-activation/part-128

629

—

hasTokensPerSecondblah/watt-activation/part-128

13023

—

lossValueblah/vidya/6

4.7787

—

stepNumberblah/watt-activation/150

2500

—

progressPercentageblah/watt-activation/150

—

hasLossValueblah/watt-activation/150

5.9942

—

hasPerplexityValueblah/watt-activation/150

401.1

—

hasLearningRateblah/watt-activation/150

0.000048

—

hasEstimatedTimeRemainingblah/watt-activation/150

158min

References (6)

6 references

[1]Part 61 fact
ctx:discord/blah/vidya/part-6
[2]Part 1508 facts
ctx:discord/blah/watt-activation/part-150
[3]Part 6582 facts
ctx:discord/blah/watt-activation/part-658
[4]Part 1287 facts
ctx:discord/blah/watt-activation/part-128
[5]61 fact
ctx:discord/blah/vidya/6
- full textvidya-6
  text/plain3 KBdoc:agent/vidya-6/cda90ecf-8302-448a-a889-53b5a677fef3
  Show excerpt
  [2026-02-21 10:36] rolandnsharp7643: >so what did we complete today. we added reinforcement learning. and changed the data set and what else …
[6]1506 facts
ctx:discord/blah/watt-activation/150
- full textwatt-activation-150
  text/plain2 KBdoc:agent/watt-activation-150/337a9b46-3368-40e6-b1b8-d27104be216d
  Show excerpt
  [2026-03-09 15:48] xenonfun: step 2000/16684 12.0% loss=6.0245 ppl= 413.4 lr=4.88e-05 667ms 12,290tok/s eta=163min VAL step 2000 loss=6.0279 ppl=414.9 ★ best ckpt step 2000 → step_002000 step 2100/16684 12.6% loss=6.01…

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.