v4 training
From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-06-06.)
v4 training has 19 facts recorded in Dontopedia across 2 references, with 1 live disagreement.
Mostly:saves checkpoints at(5), has remaining time(1), has remaining time approx(1)
Maturity scale
raw canonical shape-checked rule-derived certifiedOther facts (18)
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 |
|---|---|---|
| Saves Checkpoints at | 250K | [1] |
| Saves Checkpoints at | 100K | [1] |
| Saves Checkpoints at | 150K | [1] |
| Saves Checkpoints at | 200K | [1] |
| Saves Checkpoints at | 50K | [1] |
| Has Remaining Time | 61 hours | [1] |
| Has Remaining Time Approx | 2.5 days | [1] |
| Has Total Steps | 300K | [1] |
| Is Running | Current Status | [1] |
| Uses Conversations | 123K | [1] |
| Nears Completion | 61 Hours Left | [1] |
| Progresses Steadily | Loss Dropping | [1] |
| Rdf:type | Training Run | [2] |
| Status | running | [2] |
| Uses Model | 10m Param Transformer | [2] |
| Step Count Target | 300000 | [2] |
| Metric | Loss Metric | [2] |
| Has Checkpoint Config | Intermediate Checkpoints | [2] |
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 (2)
ctx:discord/blah/vidya/part-7ctx:discord/blah/vidya/7- full textvidya-7text/plain3 KB
doc:agent/vidya-7/06ef0b6b-8c5c-46b4-a7b7-96e33a18cb2dShow excerpt
[2026-02-22 22:34] ajaxdavis: hell yeah! [2026-02-22 22:35] ajaxdavis: so you had to train base model and then do RL with chat messages training set after for the chatty behavior? [2026-02-22 22:36] rolandnsharp7643: nah, there is no RL at …
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.