Training Steps
From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-05-02.)
Training Steps has 15 facts recorded in Dontopedia across 10 references, with 1 live disagreement.
Mostly:shows decreasing loss(2), follows sequence(1), reports at intervals(1)
Maturity scale
raw canonical shape-checked rule-derived certifiedInbound mentions (15)
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.
decreasesOverStepsDecreases Over Steps(2)
- Loss Values
ex:loss-values - Perplexity Values
ex:perplexity-values
areTemporalUnitsAre Temporal Units(1)
- Steps
ex:steps
computedBetweenStepsComputed Between Steps(1)
- Spectral Flux
ex:spectral-flux
couldAnimateConstellationEvolutionCould Animate Constellation Evolution(1)
- Dashboard
ex:dashboard
dcModePowerGrewMonotonicallyDC Mode Power Grew Monotonically(1)
- Block 0
ex:block-0
decreasingOverTimeDecreasing Over Time(1)
- Eta
ex:eta
growsMonotonicallyOverGrows Monotonically Over(1)
- DC Mode Power
ex:dc-mode-power
hasLowestLossAmongListedHas Lowest Loss Among Listed(1)
- Training Step 6800
ex:training-step-6800
intendsToEnableViewingIntends to Enable Viewing(1)
- Ajaxdavis
ex:ajaxdavis
progressesSequentiallyProgresses Sequentially(1)
- Training
ex:training
reportsProgressPeriodicallyReports Progress Periodically(1)
- Training Run
ex:training-run
stableAcrossStepsStable Across Steps(1)
- Tokens Per Second
ex:tokens-per-second
staysStableStays Stable(1)
- Tokens Per Second
ex:tokens-per-second
variesVaries(1)
- Step Time
ex:step-time
Other facts (15)
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.
| Predicate | Value | Ref |
|---|---|---|
| Shows Decreasing Loss | Loss | [4] |
| Shows Decreasing Loss | Training Step 100 to 500 | [5] |
| Follows Sequence | Step 500 | [1] |
| Reports at Intervals | 500 | [2] |
| Increased From | 100000 | [3] |
| Increased to | 200000 | [3] |
| Shows Decreasing Perplexity | Ppl | [4] |
| Shows Stable Lr After Warmup | Step 300 Onward | [5] |
| Shows Decreasing Ppl | Training Step 100 to 500 | [5] |
| Shows Bpb Improvement Over Steps | null | [6] |
| Indicates Training Speed | Slower to Train | [7] |
| Count | 5K | [8] |
| Need More | true | [8] |
| Progresses Sequentially | 100 to 1000 | [9] |
| Equals | 50 | [10] |
Timeline
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.
References (10)
ctx:discord/blah/vidya/part-4ctx:discord/blah/vidya/part-3ctx:discord/blah/vidya/part-6ctx:discord/blah/watt-activation/part-134ctx:discord/blah/watt-activation/part-147ctx:discord/blah/watt-activation/part-386ctx:discord/blah/watt-activation/part-404ctx:discord/blah/watt-activation/part-500ctx:discord/blah/watt-activation/part-215ctx:discord/blah/watt-activation/part-341
See also
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.