Cross Patch Ar Decoder
From Dontopedia, the open, paraconsistent wiki. (Last updated 2026-06-06.)
Cross Patch Ar Decoder has 21 facts recorded in Dontopedia across 4 references, with 4 live disagreements.
Mostly:has parameter(3), has dimension(2), has layers(2)
Maturity scale
raw canonical shape-checked rule-derived certifiedInbound mentions (11)
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.
causesPerformanceIssueCauses Performance Issue(2)
- Sequential Generation Loop
ex:sequential-generation-loop - Sequential Python Loop Forward Train
ex:sequential-python-loop-forward-train
partOfPart of(2)
- Training Step 1000
ex:training-step-1000 - Training Step 2000
ex:training-step-2000
concernsSubjectConcerns Subject(1)
- Message 2
ex:message-2
precedesPrecedes(1)
- Vq Encoder
ex:vq-encoder
providedPerformanceReviewOfProvided Performance Review of(1)
- Xenonfun
ex:xenonfun
stableForArDecoderStable for Ar Decoder(1)
- Speed Bps
ex:speed-bps
usesTeacherForcingUses Teacher Forcing(1)
- Training
ex:training
Other facts (21)
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 Parameter | d=384 | [4] |
| Has Parameter | 3L | [4] |
| Has Parameter | 15000-steps | [4] |
| Has Dimension | 384 | [1] |
| Has Dimension | 384 | [2] |
| Has Layers | 3L | [1] |
| Has Layers | 3 | [2] |
| Has Params | 3858304 | [1] |
| Has Params | 1305252 | [2] |
| Speed Cause | now-compiled | [4] |
| Speed Cause | other-dumb-shit-removed | [4] |
| Trained for | 15000 steps | [1] |
| Trained Steps | 15000 | [2] |
| Context Is Training Log | Mlx Training | [2] |
| Model Dimension | 384 | [3] |
| Layer Count | 3 | [3] |
| Training Steps Completed | 15000 | [3] |
| Parameter Count | 3858304 | [3] |
| Assessed on | 2026-03-14 | [3] |
| Has Metric | 1,305,252-params | [4] |
| Has Speed Comparison | way-quicker | [4] |
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 (4)
ctx:discord/blah/watt-activation/part-301ctx:discord/blah/watt-activation/part-302ctx:discord/blah/watt-activation/299- full textwatt-activation-299text/plain2 KB
doc:agent/watt-activation-299/3c2f94b1-841d-457c-8bfa-1161d5e80363Show excerpt
[2026-03-14 06:27] xenonfun: ``` Loaded 41.4 MB VQ encoder (5000 steps)... Frozen downstream: 170,329 trainable / 520,945 total step 1000 | ppl=9.0 util=0.234 | 197,267 B/s step 2000 | ppl=31.0 util=0.594 | 196,457 B/s st…
ctx:discord/blah/watt-activation/300- full textwatt-activation-300text/plain3 KB
doc:agent/watt-activation-300/3b6edccf-3524-4608-838f-25890efaea15Show excerpt
[2026-03-14 06:34] xenonfun: ``` 3. Manual attention (lines 110-128) — Hand-rolled softmax attention instead of using mx.fast.scaled_dot_product_attention. MLX's fused attention kernel is significantly faster for small sequence lengths. …
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.