fp32
From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-06-06.)
fp32 has 10 facts recorded in Dontopedia across 2 references, with 1 live disagreement.
Mostly:rdf:type(2), final validation perplexity(1), best validation perplexity(1)
Maturity scale
raw canonical shape-checked rule-derived certifiedInbound mentions (2)
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.
hasBaselineHas Baseline(1)
- Comparison 1
ex:comparison-1
participantsParticipants(1)
- Loss Curves Event
ex:loss-curves-event
Other facts (8)
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 |
|---|---|---|
| Rdf:type | Model Configuration | [1] |
| Rdf:type | Precision Configuration | [2] |
| Final Validation Perplexity | 142.7 | [1] |
| Best Validation Perplexity | 143 | [1] |
| Throughput | 69457 | [1] |
| Total Runtime | 472 | [1] |
| Weight Memory | 67.3MB | [1] |
| Has Memory Size | 63.4 | [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/watt-activation/276- full textwatt-activation-276text/plain3 KB
doc:agent/watt-activation-276/53331a47-df58-4d3f-a691-96a8098e7b35Show excerpt
[2026-03-13 21:31] xenonfun: ⏺ Results are in: ``` ┌───────────────┬───────────────┬────────────────────────────────┬───────┐ │ │ fp32 │ Ternary 1.58-bit │ Delta │ ├───────────────┼───────────────┼…
ctx:discord/blah/watt-activation/275- full textwatt-activation-275text/plain3 KB
doc:agent/watt-activation-275/149bf24b-54fb-4412-b289-e2e03bccffe2Show excerpt
[2026-03-13 20:46] xenonfun: ⏺ This tells a clear story: Current trained weights are NOT naturally ternary. Only ~46% of values are near {-1, 0, 1} after scaling (random Gaussian would give ~47%, so no better than chance). Relative err…
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.