LoRA training
From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-06-06.)
LoRA training has 16 facts recorded in Dontopedia across 3 references.
Mostly:has priority(2), involves(1), has notes(1)
Maturity scale
raw canonical shape-checked rule-derived certifiedInbound mentions (3)
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.
encompassesEncompasses(1)
- Lora Adapter Evolution
ex:lora-adapter-evolution
evaluatedInEvaluated in(1)
- Nlp Tasks
ex:nlp-tasks
listsFeaturesWithHighPriorityLists Features With High Priority(1)
- Feature Table
ex:feature-table
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 |
|---|---|---|
| Has Priority | High | [1] |
| Has Priority | High | [3] |
| Involves | fine-tune adapters, evaluate on NLP tasks | [1] |
| Has Notes | Fine-tune adapters, evaluate on NLP tasks | [1] |
| Has Python Loc | ~310 | [1] |
| Has Elapsed Time | 26.0 | [2] |
| Achieved New Best Ppl | 236.2 | [2] |
| Follows Earlier Plan | Higher Rank Plan | [2] |
| Has Average Tok Per Sec | 15687 | [2] |
| Has Best Val Ppl | 236.5 | [2] |
| Has Trainable Params | 1203648 | [2] |
| Processed Tokens | 24453120 | [2] |
| Uses Lora Rank | 16 | [2] |
| Rdf:type | Feature | [3] |
| Estimated Lines of Code | 310 | [3] |
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 (3)
ctx:discord/blah/watt-activation/part-489ctx:discord/blah/watt-activation/part-174ctx:discord/blah/watt-activation/486- full textwatt-activation-486text/plain3 KB
doc:agent/watt-activation-486/c8568fef-e9f2-4d48-9840-89f375514ea3Show excerpt
[2026-03-22 03:19] xenonfun: ``` ⏺ The IVF was useless before because ivf_dirty got set on every insert (128/step), and only cleared every 500 steps on rebuild. So 99.7% of lookups fell back to linear scan. Now: - New entries get assigne…
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.