Baseline

as donto believed on

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

Baseline has 43 facts recorded in Dontopedia across 9 references, with 4 live disagreements.

43 facts·33 predicates·9 sources·4 in dispute

Mostly:rdf:type(4), has speed(2), has best loss(2)

Inbound mentions (5)

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.

`improvesUpon`Improves Upon(2)

Config Fast Ropeex:config-fast-rope
Config Packed Sequencesex:config-packed-sequences

`comparesConfiguration`Compares Configuration(1)

Dyn H4 Ab Testex:dyn-h4-ab-test

`containsRowFor`Contains Row for(1)

Comparison Table 10kex:comparison-table-10k

`refersTo`Refers to(1)

Session Gainsex:session-gains

Other facts (39)

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.

39 facts

Predicate	Value	Ref
Rdf:type	Configuration	[4]
Rdf:type	Configuration	[5]
Rdf:type	Model Configuration	[6]
Rdf:type	Configuration	[9]
Has Speed	29.2	[4]
Has Speed	66.8 it/s	[5]
Has Best Loss	1.8893	[5]
Has Best Loss	1.8893	[6]
Has Average Loss	3.4059	[5]
Has Average Loss	3.3905	[6]
Has Lr	1e-3	[1]
Hypothesizes	Current (diverges)	[1]
Has Steps	2	[1]
Has Dt	0.1	[1]
Has Iterations Per Second	135	[2]
Has Final Ppl	179.2	[2]
Has Best Ppl	80.7	[2]
Has Tokens Per Second	69.1K	[2]
Has Run Time Seconds	7.4s	[2]
Has Avg Loss	5.1884	[2]
Is Current Result	null	[3]
Has D Dec	128	[3]
Has Patch Size	8	[3]
Has Result	3.08 bpb	[3]
Has Batch Size	4	[4]
Is Compiled	true	[4]
Has Eval Loss	3.2769	[6]
Generated Text for Prompt	Of a few of the scholars who considered the great person. In general, natural philosopher	[6]
Generation Quality	poor	[6]
Uses Reduce Lr on Plateau	false	[6]
Patch Size	8	[7]
Decoder Dimension	128	[7]
Hypothesis	Current result (3.08 bpb)	[7]
Seq Length	256	[8]
Has Kappa Parameter	Fixed Kappa	[9]
Has Bpb Value	2.103	[9]
Has Snr Value	-6.4	[9]
Has Capacity Value	2400000000	[9]
Has Tok Per S Value	23032	[9]

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.

43 undated

—

hasLrblah/watt-activation/part-44

1e-3

—

hypothesizesblah/watt-activation/part-44

Current (diverges)

—

hasStepsblah/watt-activation/part-44

—

hasDtblah/watt-activation/part-44

0.1

—

hasIterationsPerSecondblah/watt-activation/part-66

135

—

hasFinalPPLblah/watt-activation/part-66

179.2

—

hasBestPPLblah/watt-activation/part-66

80.7

—

hasTokensPerSecondblah/watt-activation/part-66

69.1K

—

hasRunTimeSecondsblah/watt-activation/part-66

7.4s

—

hasAvgLossblah/watt-activation/part-66

5.1884

—

isCurrentResultblah/watt-activation/part-300

null

—

hasDDecblah/watt-activation/part-300

128

—

hasPatchSizeblah/watt-activation/part-300

—

hasResultblah/watt-activation/part-300

3.08 bpb

—

typeblah/watt-activation/20

ex:Configuration

—

labelblah/watt-activation/20

Baseline (compiled, bs=4)

—

hasSpeedblah/watt-activation/20

29.2

—

hasBatchSizeblah/watt-activation/20

—

isCompiledblah/watt-activation/20

true

—

typeblah/watt-activation/37

ex:Configuration

—

hasBestLossblah/watt-activation/37

1.8893

—

hasAverageLossblah/watt-activation/37

3.4059

—

hasSpeedblah/watt-activation/37

66.8 it/s

—

labelblah/watt-activation/38

Baseline

—

typeblah/watt-activation/38

ex:ModelConfiguration

—

hasBestLossblah/watt-activation/38

1.8893

—

hasAverageLossblah/watt-activation/38

3.3905

—

hasEvalLossblah/watt-activation/38

3.2769

—

generatedTextForPromptblah/watt-activation/38

Of a few of the scholars who considered the great person. In general, natural philosopher

—

generationQualityblah/watt-activation/38

poor

—

usesReduceLROnPlateaublah/watt-activation/38

false

—

labelblah/watt-activation/298

Baseline config

—

patchSizeblah/watt-activation/298

—

decoderDimensionblah/watt-activation/298

128

—

hypothesisblah/watt-activation/298

Current result (3.08 bpb)

—

seqLengthblah/watt-activation/404

256

—

labelblah/watt-activation/425

Baseline (fixed κ)

—

typeblah/watt-activation/425

ex:Configuration

—

hasKappaParameterblah/watt-activation/425

ex:fixed-kappa

—

hasBpbValueblah/watt-activation/425

2.103

—

hasSnrValueblah/watt-activation/425

-6.4

—

hasCapacityValueblah/watt-activation/425

2400000000

—

hasTokPerSValueblah/watt-activation/425

23032

References (9)

9 references

[1]Part 444 facts
ctx:discord/blah/watt-activation/part-44
[2]Part 666 facts
ctx:discord/blah/watt-activation/part-66
[3]Part 3004 facts
ctx:discord/blah/watt-activation/part-300
[4]205 facts
ctx:discord/blah/watt-activation/20
- full textwatt-activation-20
  text/plain3 KBdoc:agent/watt-activation-20/3112a2c6-bfc5-4e04-a46c-ccbafb3ff570
  Show excerpt
  [2026-03-06 05:09] xenonfun: Good picture now. Three concrete wins: 1. Lazy eval overlap — MLX compiled step is lazy until mx.eval. We can prepare the next batch on CPU while the GPU is executing, completely free. 2. Flash attention — …
[5]374 facts
ctx:discord/blah/watt-activation/37
- full textwatt-activation-37
  text/plain3 KBdoc:agent/watt-activation-37/6b4934b5-d720-4142-b336-52ad2f3f457f
  Show excerpt
  [2026-03-07 02:32] xenonfun: Iter 10000/139000 | Loss: 3.3393 (PPL: 28.20) | Avg(100): 3.3596 (PPL: 28.78) | Best: 1.8429 | LR: 7.47e-05 | 39.5 it/s (10.1K tok/s | 7.19%) | ETA: 54m 29s =====================================================…
[6]388 facts
ctx:discord/blah/watt-activation/38
- full textwatt-activation-38
  text/plain3 KBdoc:agent/watt-activation-38/ef80610b-987f-4bc3-9ebd-f8873872e4ad
  Show excerpt
  [2026-03-07 02:38] xenonfun: ``` 25K Comparison ┌──────────┬───────────┬──────────┬───────────┐ │ │ Best Loss │ Avg(100) │ Eval Loss │ ├──────────┼───────────┼──────────┼───────────┤ │ Baseline │ 1.8893 │ 3.3905 │ 3.2…
[7]2984 facts
ctx:discord/blah/watt-activation/298
- full textwatt-activation-298
  text/plain2 KBdoc:agent/watt-activation-298/f5cde311-fd9a-43e7-a746-9177b5a91fee
  Show excerpt
  [2026-03-14 05:53] xenonfun: ``` What Changed The AR decoder produces recognizable English word fragments: "the", "and", "for", "with", "this", "one", "protec(tion)", "earch", "context", "project", "state", "imported". These are real m…
[8]4041 fact
ctx:discord/blah/watt-activation/404
- full textwatt-activation-404
  text/plain3 KBdoc:agent/watt-activation-404/a192ce00-8f5b-4ec4-9fb0-034d17a97689
  Show excerpt
  [2026-03-19 05:42] xenonfun: ```⏺ Both healthy: - 10K (seq=256): BPB 3.312 at step 3250, still dropping - 4K-seq (seq=4096): BPB 3.423 at step 2150, 128K tok/s — faster throughput because bs=16 × seq=4096 = 65K tokens/batch (same as bs…
[9]4257 facts
ctx:discord/blah/watt-activation/425
- full textwatt-activation-425
  text/plain2 KBdoc:agent/watt-activation-425/371cf09a-5e2c-4cba-802d-740be172c544
  Show excerpt
  [2026-03-20 01:59] xenonfun: ⏺ DYN-H4 A/B Test — Final Results at 2K Steps ``` ┌──────────┬───────────────────────┬────────────────────┬──────────────────┐ │ Metric │ Phase coupling (no κ) │ Baseline (fixed κ) │ Delta │ ├…

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.

Baseline

Inbound mentions (5)

improvesUponImproves Upon(2)

comparesConfigurationCompares Configuration(1)

containsRowForContains Row for(1)

refersToRefers to(1)

Other facts (39)

Timeline

References (9)

See also

Keep researching

`improvesUpon`Improves Upon(2)

`comparesConfiguration`Compares Configuration(1)

`containsRowFor`Contains Row for(1)

`refersTo`Refers to(1)