FIND AN IDEA

Follow your curiosity.

Search across the complete book and example library.

163 entries

I · Orientation

What Welvet is

Loom’s flat poly/ package hit import-cycle and honesty walls (QAT morph, silent fallbacks, god-layer). Welvet is the rewrite: one feature per folder, storage-truth dtypes/quants, Dense as th

I · Orientation

Repository map

Readers need a single map of what is engine vs harness vs app vs stub.

II · Foundation

core: types & backends

Every polymorphic path needs one place for DType, LayerType, Activation, Backend, Tensor[T], and slim Layer metadata: without QAT morph defaults.

II · Foundation

weights: FormatNone MatVec

Unquantized matrices still need a typed store that streams MatVec and SGD without forcing a float32 master or Morph-as-training.

II · Foundation

quant: 20 pack formats

Inference, storage, and train need classic Q-packs, k-quants, IQ, Ternary/Binary, and Affine without a separate QAT mode or retained f32 master. Format is storage truth.

II · Foundation

simd: Plan 9 kernels

CPU peak needs hand-written AVX2/NEON without a silent Go fallback that pretends SIMD ran.

II · Foundation

webgpu: device GEMV & shaders

GPU paths must bind a real adapter. Host “fake GPU” was banned so suites cannot stamp WebGPU done when ALU ran on CPU.

II · Foundation

tiling: SC/MC & workgroups

MatVec throughput depends on tile size and when to go multi-core vs GPU workgroups. Centralizing caps keeps Dense and friends consistent.

II · Foundation

architecture: volumetric grid

Networks are spatial (Depth×Rows×Cols×LayersPerCell), not only linear stacks. Topology lives here; compute lives in layer packages.

II · Foundation

fusedgpu: decoder on device

Token-by-token host round-trips kill decode. A fused engine keeps weights and scratch resident for Q4_0 and BinaryG128 hybrid paths.

III · Layers

layers/dense: MatVec microkernel

Most FLOPs are W@x. One Dense stack owns FormatNone×34 and all quants × three backends so every composite proj shares one correctness surface: including native in-dtype SGD.

III · Layers

layers/mha: attention

Transformers need multi-head attention with masks, RoPE/ALiBi, GQA/MQA, and cross-attn: without forking MatVec for every projection.

III · Layers

layers/swiglu: gated FFN

Modern decoder FFNs are SiLU(gate)⊙up → down. Projections must share Dense’s quant/backend matrix.

III · Layers

layers/rmsnorm

Llama-style blocks normalize by RMS, not mean+var. Needs native fwd/bwd and WebGPU shaders.

III · Layers

layers/layernorm

Classic mean+var normalization with γ/β: still required for many HF architectures.

III · Layers

layers/embedding

Token IDs must gather rows from a table: not a Dense MatVec: with scatter grads on backward.

III · Layers

layers/softmax

Classification heads and attention need stable softmax variants, including sparse/Gumbel/Entmax for research paths.

III · Layers

layers/sequential

Some cells need an ordered Dense chain without burning grid hops.

III · Layers

layers/residual

Skip connections stabilize deep stacks: y = F(x) + x with correct skip grads.

III · Layers

layers/cnn1 · cnn2 · cnn3

Conv nets must sit on the same dtype×quant×backend matrix as Dense. im2col → Dense GEMV is the intentional first cut.

III · Layers

layers/rnn · lstm

Sequence models before transformers still need vanilla RNN and LSTM with BPTT on the shared MatVec stack.

III · Layers

layers/seqmix: mixer contract

Attention, SSM, linear attn, and conv mixers must not be accidental forks of mha. Naming the contract keeps packages honest.

III · Layers

layers/gdn: gated delta net

Linear attention / decode-first mixers (Gated DeltaNet) need a first-class package under KindLinearAttn.

III · Layers

layers/mamba: selective SSM

SSM mixers (KindSSM) are not MHA clones: they need their own selective-scan path.

III · Layers

layers/convt1 · convt2 · convt3

Generators and U-Nets need transposed convolution as a peer of CNN, on the same Dense proj surface.

III · Layers

layers/kmeans

Soft clustering as a differentiable layer lets topology experiments sit inside the same train loop.

III · Layers

layers/parallel: MoE + cameral

Mixture-of-experts and multi-path cells need concat/add/avg/filter combines. Cameral graphs need sibling hemispheres that share input, merge outputs, and optionally train under distinct mode

III · Layers

layers/metacognition

Observed layers can apply heuristic stability rules (gate/scale/reset) without dtype morph/QAT.

IV · Runtime

runtime/forward

A grid of heterogeneous ops needs one walker that dispatches by concrete type and fails loudly on unknowns.

IV · Runtime

runtime/backward

Training needs a reverse tape over the same ops forward used: no separate graph framework.

IV · Runtime

runtime/training

Suites and small nets need MSE+SGD and tween hooks without inventing an external trainer or a retained float32 master beside storage.

IV · Runtime

runtime/step: step mesh

Spatial feedback (remote links) needs a discrete-time mesh where every cell updates from a double buffer: different from a decoder wavefront. Cross-numeric train also needs the same mesh wi

V · Systems

systems/dna

Quant and train must be measurable as topology/weight fingerprints: DNA detects logic shifts.

V · Systems

systems/evolution

Topology search and weight crossover need first-class splice + NEAT on CPU-resident grids.

V · Systems

systems/tween

Target propagation (chain-rule or Hebbian layerwise gaps) is an alternative credit-assignment path. Not the same package as TrainMode Tween / TweenChain on a Sandwich (those are layers/paral

V · Systems

systems/tanhi: TANHI · UDP HUD

Training visualization must never block the engine: best-effort UDP JSON-lines to a HUD.

V · Systems

systems/telemetry

Static structural blueprints (sizes, op kinds) differ from live TANHI events.

VI · Model IO

model/entity: .entity files

HF safetensors are awkward for native topology + packed weights. ENTITY is the Welvet checkpoint.

VI · Model IO

model/hf: snapshots

Import starts with probing HF/MLX layouts before packing ENTITY.

VI · Model IO

model/tokenizer

Generate needs encode/decode of HF tokenizer.json without pulling Python.

VI · Model IO

model/sampling

Logits → token ID needs ArgMax, TopK+temperature, penalties, and chat hygiene in one place.

VI · Model IO

model/transformer: generate

ENTITY packs must run as Llama-style decoders with KV cache, profiles (SIMD/WebGPU/fused), and chat templates.

VII · Apps

apps: octo · flux2 · mosstts

Products must not pollute engine packages. Octo is the model shell; flux2/mosstts are domain apps. Lucy races (AAI test41 / test48 / test50) are benches, not Welvet packages.

VII · Apps

Octo: model shell

A model is only useful with a shell around it: pull weights from Hugging Face, convert them to a Welvet .entity, then chat, serve, or benchmark. Octo is that shell, kept in its own module so

VIII · Stubs

stub/seed

Ship topology recipes (layer seeds → He-init) without weight blobs.

VIII · Stubs

stub/serialization

Volumetric grids need JSON/ENTITY persist beyond transformer packs.

VIII · Stubs

stub/memory

HF→ENTITY and GPU upload need footprint accounting and optional history charts.

VIII · Stubs

stub/donate

LAN donors should accept framed JSON jobs without embedding HTTP in the engine.

VIII · Stubs

stub/fountain

Recover specialist weight blobs over lossy links via LT fountain codes, then ensemble.

VIII · Stubs

stub/hardware

Dispatchers and UIs need a portable host audit (OS/CPU/RAM/GPU).

VIII · Stubs

stub/accel: NPU/Metal/QNN

Vendor accelerators plug beside WebGPU: not replace it. Qualcomm HTP is live via nativeqnn + sibling welvet.qnn (see book §79-81).

VIII · Stubs

stub/clustering

Offline clustering helpers on tensors without inventing a second math stack.

VIII · Stubs

stub/ensemble

Combine multiple model votes and find complementary specialists.

VIII · Stubs

stub/evaluation

Benchmark grids through runtime/forward with deviation metrics.

VIII · Stubs

stub/grafting

Merge grids into Parallel/Residual structures for topology experiments.

VIII · Stubs

stub/grouping

Detect layer archetypes from safetensor-style names before mounting.

VIII · Stubs

stub/introspection

UIs and FFI need to list Grid methods without hardcoding every export.

VIII · Stubs

stub/observer

Attach forward/backward observers for debugging without coupling to tanhi UDP.

VIII · Stubs

stub/pipeline

Decoder wavefront stats helpers: not a full Lucy-style pipeline runner yet.

VIII · Stubs

stub/templates

Chat prompts must match model families (ChatML, Llama3, BitNet) without app-specific string glue.

VIII · Stubs

stub/universal

Probe unknown safetensor geometry and mount placeholder grids until full weight import lands.

IX · Validate

w2a: validation harness

Engine packages must stay free of tests. w2a owns timed 34×20×3 matrices, gap census, honesty stamps, and the train-mode permutation smoke (Test49). See §63 for a live full-suite run.

IX · Validate

Validation report: full suite

Claims are cheap; a stamped matrix is not. This is the actual output of one full w2a [0] Run ALL so the book's ✅ marks are backed by numbers you can reproduce, not asserted.

IX · Validate

Scorecard → v1.0 / minors

Version is earned from a weighted board, not marketing. v1.0 is 100/100 on the engine board. Minor/patch tags (v1.1.0, v1.1.1, …) pack features without a new board. Apps, stubs, and NPU sit

IV · Runtime

Cross-numeric train + down-the-dem

Weight storage dtype and activation Tensor[T] are independent axes. Proving train without a retained float32 master means sweeping W×A: not only matched float32 acts.

V · Systems

lucy: SoftAcc / Score measuring

Adaptation benches (test41-w, tide, live_gpt) need one shared measuring math: SoftAcc, Availability, AdaptPct, Score: not three copies of the formulas.

IV · Runtime

TrainMode: 29 named updates

Backprop is one update, not the only one. Credit assignment (broadcast gap, head proxy, sparse duty clock) has to be a named axis you can race: not a comment in a notebook. Cameral Mix also

VII · Apps

Cameral sandwiches + AAI Lucy

A single Dense chain cannot host two independent weight copies that share an input, merge, and optionally train under different updates. That is the cameral graph: hemispheres, not screen-sp

V · Systems

Lucy density: synthetic organism

A new host (char LM, MNIST, a layer sprint) should not copy tide's goldilocks math. The question is always the same: can the net run and train at the same time in a small box, then how far t

VII · Apps

CamSync: inter-cameral / cross-mesh weight blend

Mix BranchModes and different inits let cams diverge on purpose. Sometimes you want them to share: gently (1% pull) or hard (full average): within a Parallel, across Stack children, or eve

X · NativeCPU (welvet.cpp)

NativeCPU: opt-in welvet.cpp C ABI

Go stays the default engine. When you want the C++ MatVec / NormalBP / TrainMSE path in-process, set BackendNativeCPU (never Prefer as default). Pending C++ weights stay off Go Stores until

X · NativeCPU (welvet.cpp)

NativeCPU: TrainNativeMSE per layer

Each signed layer has a packed C ABI NormalBP path. One smoke walks Dense → CNN/ConvT → norms → RNN/LSTM → Sequential/Residual → Parallel → Embedding → Softmax → KMeans → MHA/SwiGLU → Mamba/

X · NativeCPU (welvet.cpp)

NativeCPU: TrainMode stack / parallel / mesh / line

Leaf credit for every TrainMode family goes through one C dispatcher. Mesh clock and Line pipe have their own host-shape ABIs. Go intercepts TrainStackMSE/TrainMSE when BackendNativeCPU; cal

XI · Hexagon NPU (welvet.qnn)

QNN: opt-in Hexagon HTP (welvet.qnn)

Qualcomm Hexagon HTP is a sibling engine (welvet.qnn 1.0.0), not a silent fallback inside Go. Opt-in BackendQNN + explicit SyncToQNN keep CPU/SIMD/WebGPU as defaults. Caps reject unsupported

XI · Hexagon NPU (welvet.qnn)

QNN: HTP layer tour (session ABI)

w2a go_v_qnn exercises every shipped HTP forward. Book examples use the same session ABI with tiny shapes so a Snapdragon box can smoke-test without the full matrix.

XI · Hexagon NPU (welvet.qnn)

QNN: go_v_qnn RESULTS (large shapes)

Toy shapes lie. The w2a large matrix (dense 32×512×512, softmax 64×4096, …) is the honest Snapdragon X story: NPU DLL graphExecute vs Go/native f32 CPU, plus Go→NPU call overhead (cgo + pack

Welvet examples

1. What Welvet is

Part: I · OrientationPackage: github.com/openfluke/welvetStatus: ok: ✅ engine When You need github.com/openfluke/welvet. Where import "github.com/openfluke/welvet/github.com/openfluke/welve

Welvet examples

2. Repository map

Part: I · OrientationPackage: -Status: ok: ✅ layout When Orientation chapter: Repository map. Where Book / repo map (no single import) cd 02-tree && source ../env.sh && go run . Why Readers

Welvet examples

3. core: types & backends

Part: II · FoundationPackage: github.com/openfluke/welvet/coreStatus: ok: ✅ When You need the core foundation package. Where import "github.com/openfluke/welvet/core" cd 03-core && source .

Welvet examples

4. weights: FormatNone MatVec

Part: II · FoundationPackage: github.com/openfluke/welvet/weightsStatus: ok: ✅ When You need the weights foundation package. Where import "github.com/openfluke/welvet/weights" cd 04-weights

Welvet examples

5. quant: 20 pack formats

Part: II · FoundationPackage: github.com/openfluke/welvet/quantStatus: ok: ✅ When You need the quant foundation package. Where import "github.com/openfluke/welvet/quant" cd 05-quant && sour

Welvet examples

6. simd: Plan 9 kernels

Part: II · FoundationPackage: github.com/openfluke/welvet/simdStatus: ok: ✅ When You need the simd foundation package. Where import "github.com/openfluke/welvet/simd" cd 06-simd && source .

Welvet examples

7. webgpu: device GEMV & shaders

Part: II · FoundationPackage: github.com/openfluke/welvet/webgpuStatus: ok: ✅ When You need the webgpu foundation package. Where import "github.com/openfluke/welvet/webgpu" cd 07-webgpu &&

Welvet examples

8. tiling: SC/MC & workgroups

Part: II · FoundationPackage: github.com/openfluke/welvet/tilingStatus: ok: ✅ When You need the tiling foundation package. Where import "github.com/openfluke/welvet/tiling" cd 08-tiling &&

Welvet examples

9. architecture: volumetric grid

Part: II · FoundationPackage: github.com/openfluke/welvet/architectureStatus: ok: ✅ When You need the architecture foundation package. Where import "github.com/openfluke/welvet/architecture

Welvet examples

10. fusedgpu: decoder on device

Part: II · FoundationPackage: github.com/openfluke/welvet/fusedgpuStatus: ok: ✅ When You need the fusedgpu foundation package. Where import "github.com/openfluke/welvet/fusedgpu" cd 10-fuse

Welvet examples

11. layers/dense: MatVec microkernel

Part: III · LayersPackage: github.com/openfluke/welvet/layers/denseStatus: ok: ✅ When Building or training a net that needs the layers/dense Op (also usable as a Parallel cam). Where import

Welvet examples

12. layers/mha: attention

Part: III · LayersPackage: github.com/openfluke/welvet/layers/mhaStatus: ok: ✅ When Building or training a net that needs the layers/mha Op (also usable as a Parallel cam). Where import "gi

Welvet examples

13. layers/swiglu: gated FFN

Part: III · LayersPackage: github.com/openfluke/welvet/layers/swigluStatus: ok: ✅ When Building or training a net that needs the layers/swiglu Op (also usable as a Parallel cam). Where impo

Welvet examples

14. layers/rmsnorm

Part: III · LayersPackage: github.com/openfluke/welvet/layers/rmsnormStatus: ok: ✅ When Building or training a net that needs the layers/rmsnorm Op (also usable as a Parallel cam). Where im

Welvet examples

15. layers/layernorm

Part: III · LayersPackage: github.com/openfluke/welvet/layers/layernormStatus: ok: ✅ When Building or training a net that needs the layers/layernorm Op (also usable as a Parallel cam). Wher

Welvet examples

16. layers/embedding

Part: III · LayersPackage: github.com/openfluke/welvet/layers/embeddingStatus: ok: ✅ When Building or training a net that needs the layers/embedding Op (also usable as a Parallel cam). Wher

Welvet examples

17. layers/softmax

Part: III · LayersPackage: github.com/openfluke/welvet/layers/softmaxStatus: ok: ✅ When Building or training a net that needs the layers/softmax Op (also usable as a Parallel cam). Where im

Welvet examples

18. layers/sequential

Part: III · LayersPackage: github.com/openfluke/welvet/layers/sequentialStatus: ok: ✅ When Building or training a net that needs the layers/sequential Op (also usable as a Parallel cam). Wh

Welvet examples

19. layers/residual

Part: III · LayersPackage: github.com/openfluke/welvet/layers/residualStatus: ok: ✅ When Building or training a net that needs the layers/residual Op (also usable as a Parallel cam). Where

Welvet examples

20. CNN family: cnn1 / cnn2 / cnn3

When: spatial / temporal convolution senses (1D seq, 2D images, 3D volumes).Where: layers/cnn1, layers/cnn2, layers/cnn3Why: im2col → Dense Proj so quant/SIMD/WebGPU and CamSync (via Proj) a

Welvet examples

21. layers/rnn · lstm

Part: III · LayersPackage: github.com/openfluke/welvet/layers/lstmStatus: ok: ✅ When Building or training a net that needs the layers/lstm Op (also usable as a Parallel cam). Where import "

Welvet examples

22. layers/seqmix: mixer contract

Part: III · LayersPackage: github.com/openfluke/welvet/layers/seqmixStatus: ok: ✅ When Building or training a net that needs the layers/seqmix Op (also usable as a Parallel cam). Where impo

Welvet examples

23. layers/gdn: gated delta net

Part: III · LayersPackage: github.com/openfluke/welvet/layers/gdnStatus: ok: ✅ When Building or training a net that needs the layers/gdn Op (also usable as a Parallel cam). Where import "gi

Welvet examples

24. layers/mamba: selective SSM

Part: III · LayersPackage: github.com/openfluke/welvet/layers/mambaStatus: ok: ✅ When Building or training a net that needs the layers/mamba Op (also usable as a Parallel cam). Where import

Welvet examples

25. ConvTranspose family: convt1 / convt2 / convt3

When: upsampling / decoder paths (1D, 2D, 3D).Where: layers/convt1, layers/convt2, layers/convt3Why: generative twins and U-Net-style expands; same Proj/Dense backend story as CNN. cd 25-con

Welvet examples

26. layers/kmeans

Part: III · LayersPackage: github.com/openfluke/welvet/layers/kmeansStatus: ok: ✅ When Building or training a net that needs the layers/kmeans Op (also usable as a Parallel cam). Where impo

Welvet examples

27. layers/parallel: MoE + cameral

Part: III · LayersPackage: github.com/openfluke/welvet/layers/parallelStatus: ok: ✅ When Building or training a net that needs the layers/parallel Op (also usable as a Parallel cam). Where

Welvet examples

28. layers/metacognition

Part: III · LayersPackage: github.com/openfluke/welvet/layers/metacognitionStatus: ok: ✅ When Building or training a net that needs the layers/metacognition Op (also usable as a Parallel ca

Welvet examples

29. runtime/forward

Part: IV · RuntimePackage: github.com/openfluke/welvet/runtime/forwardStatus: ok: ✅ When Walking a Grid/Stack with the shared runtime/forward path. Where import "github.com/openfluke/welvet

Welvet examples

30. runtime/backward

Part: IV · RuntimePackage: github.com/openfluke/welvet/runtime/backwardStatus: ok: ✅ When Walking a Grid/Stack with the shared runtime/backward path. Where import "github.com/openfluke/welv

Welvet examples

31. runtime/training

Part: IV · RuntimePackage: github.com/openfluke/welvet/runtime/trainingStatus: ok: ✅ When Walking a Grid/Stack with the shared runtime/training path. Where import "github.com/openfluke/welv

Welvet examples

32. runtime/step: step mesh

Part: IV · RuntimePackage: github.com/openfluke/welvet/runtime/stepStatus: ok: ✅ When Walking a Grid/Stack with the shared runtime/step path. Where import "github.com/openfluke/welvet/runti

Welvet examples

33. systems/dna

Part: V · SystemsPackage: github.com/openfluke/welvet/systems/dnaStatus: ok: ✅ When Adaptation, measurement, or evolution (systems/dna). Where import "github.com/openfluke/welvet/systems/dn

Welvet examples

34. systems/evolution

Part: V · SystemsPackage: github.com/openfluke/welvet/systems/evolutionStatus: ok: ✅ When Adaptation, measurement, or evolution (systems/evolution). Where import "github.com/openfluke/welve

Welvet examples

35. systems/tween

Part: V · SystemsPackage: github.com/openfluke/welvet/systems/tweenStatus: ok: ✅ When Adaptation, measurement, or evolution (systems/tween). Where import "github.com/openfluke/welvet/system

Welvet examples

36. systems/tanhi: TANHI · UDP HUD

Part: V · SystemsPackage: github.com/openfluke/welvet/systems/tanhiStatus: ok: ✅ When Adaptation, measurement, or evolution (systems/tanhi). Where import "github.com/openfluke/welvet/system

Welvet examples

37. systems/telemetry

Part: V · SystemsPackage: github.com/openfluke/welvet/systems/telemetryStatus: ok: ✅ When Adaptation, measurement, or evolution (systems/telemetry). Where import "github.com/openfluke/welve

Welvet examples

38. model/entity: .entity files

Part: VI · Model IOPackage: github.com/openfluke/welvet/model/entityStatus: ok: ✅ When Checkpoints, HF import, tokenize, sample, or decode (model/entity). Where import "github.com/openfluke

Welvet examples

39. model/hf: snapshots

Part: VI · Model IOPackage: github.com/openfluke/welvet/model/hfStatus: ok: ✅ When Checkpoints, HF import, tokenize, sample, or decode (model/hf). Where import "github.com/openfluke/welvet/

Welvet examples

40. model/tokenizer

Part: VI · Model IOPackage: github.com/openfluke/welvet/model/tokenizerStatus: ok: ✅ When Checkpoints, HF import, tokenize, sample, or decode (model/tokenizer). Where import "github.com/ope

Welvet examples

41. model/sampling

Part: VI · Model IOPackage: github.com/openfluke/welvet/model/samplingStatus: ok: ✅ When Checkpoints, HF import, tokenize, sample, or decode (model/sampling). Where import "github.com/openf

Welvet examples

42. model/transformer: generate

Part: VI · Model IOPackage: github.com/openfluke/welvet/model/transformerStatus: ok: ✅ When Checkpoints, HF import, tokenize, sample, or decode (model/transformer). Where import "github.com

Welvet examples

43. apps: octo · flux2 · mosstts

Part: VII · AppsPackage: github.com/openfluke/welvet/apps/…Status: partial: 🚧 When You need the apps/… foundation package. Where import "github.com/openfluke/welvet/apps/…" cd 43-apps && s

Welvet examples

44. Octo: model shell

Part: VII · AppsPackage: github.com/openfluke/welvet/apps/octoStatus: ok: ✅ runs When You need the apps/octo foundation package. Where import "github.com/openfluke/welvet/apps/octo" cd 44-o

Welvet examples

45. stub/seed

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/seedStatus: partial: 🚧 When Scaffold / experimental stub stub/seed (may be partial). Where import "github.com/openfluke/welvet/s

Welvet examples

46. stub/serialization

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/serializationStatus: partial: 🚧 When Scaffold / experimental stub stub/serialization (may be partial). Where import "github.com/

Welvet examples

47. stub/memory

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/memoryStatus: partial: 🚧 When Scaffold / experimental stub stub/memory (may be partial). Where import "github.com/openfluke/welv

Welvet examples

48. stub/donate

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/donateStatus: partial: 🚧 When Scaffold / experimental stub stub/donate (may be partial). Where import "github.com/openfluke/welv

Welvet examples

49. stub/fountain

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/fountainStatus: partial: 🚧 When Scaffold / experimental stub stub/fountain (may be partial). Where import "github.com/openfluke/

Welvet examples

50. stub/hardware

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/hardwareStatus: partial: 🚧 When Scaffold / experimental stub stub/hardware (may be partial). Where import "github.com/openfluke/

Welvet examples

51. stub/accel: NPU/Metal/QNN

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/accelStatus: missing: ⬜ When Scaffold / experimental stub stub/accel (may be partial). Where import "github.com/openfluke/welvet/

Welvet examples

52. stub/clustering

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/clusteringStatus: partial: 🚧 When Scaffold / experimental stub stub/clustering (may be partial). Where import "github.com/openfl

Welvet examples

53. stub/ensemble

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/ensembleStatus: partial: 🚧 When Scaffold / experimental stub stub/ensemble (may be partial). Where import "github.com/openfluke/

Welvet examples

54. stub/evaluation

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/evaluationStatus: partial: 🚧 When Scaffold / experimental stub stub/evaluation (may be partial). Where import "github.com/openfl

Welvet examples

55. stub/grafting

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/graftingStatus: partial: 🚧 When Scaffold / experimental stub stub/grafting (may be partial). Where import "github.com/openfluke/

Welvet examples

56. stub/grouping

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/groupingStatus: partial: 🚧 When Scaffold / experimental stub stub/grouping (may be partial). Where import "github.com/openfluke/

Welvet examples

57. stub/introspection

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/introspectionStatus: partial: 🚧 When Scaffold / experimental stub stub/introspection (may be partial). Where import "github.com/

Welvet examples

58. stub/observer

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/observerStatus: partial: 🚧 When Scaffold / experimental stub stub/observer (may be partial). Where import "github.com/openfluke/

Welvet examples

59. stub/pipeline

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/pipelineStatus: partial: 🚧 When Scaffold / experimental stub stub/pipeline (may be partial). Where import "github.com/openfluke/

Welvet examples

60. stub/templates

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/templatesStatus: partial: 🚧 When Scaffold / experimental stub stub/templates (may be partial). Where import "github.com/openfluk

Welvet examples

61. stub/universal

Part: VIII · StubsPackage: github.com/openfluke/welvet/stub/universalStatus: partial: 🚧 When Scaffold / experimental stub stub/universal (may be partial). Where import "github.com/openfluk

Welvet examples

62. w2a: validation harness

Part: IX · ValidatePackage: github.com/openfluke/w2aStatus: ok: ✅ harness When You need github.com/openfluke/w2a. Where import "github.com/openfluke/welvet/github.com/openfluke/w2a" cd 62-w

Welvet examples

63. Validation report: full suite

Part: IX · ValidatePackage: github.com/openfluke/w2aStatus: ok: ✅ 246k cells When You need github.com/openfluke/w2a. Where import "github.com/openfluke/welvet/github.com/openfluke/w2a" cd 6

Welvet examples

64. Scorecard → v1.0 / minors

Part: IX · ValidatePackage: -Status: ok: v1.1.0 When Orientation chapter: Scorecard → v1.0 / minors. Where Book / repo map (no single import) cd 64-scorecard && source ../env.sh && go run .

Welvet examples

65. Cross-numeric train + down-the-dem

Part: IV · RuntimePackage: github.com/openfluke/welvet/runtime/trainingStatus: ok: ✅ When Walking a Grid/Stack with the shared runtime/training path. Where import "github.com/openfluke/welv

Welvet examples

66. lucy: SoftAcc / Score measuring

Part: V · SystemsPackage: github.com/openfluke/welvet/lucyStatus: ok: ✅ When Adaptation, measurement, or evolution (lucy). Where import "github.com/openfluke/welvet/lucy" cd 66-lucy && sour

Welvet examples

67. TrainMode: 29 named updates

Part: IV · RuntimePackage: github.com/openfluke/welvet/layers/parallelStatus: ok: ✅ 29 modes When Building or training a net that needs the layers/parallel Op (also usable as a Parallel cam

Welvet examples

68. Cameral sandwiches + AAI Lucy

Part: VII · AppsPackage: github.com/openfluke/welvet/layers/parallelStatus: ok: ✅ cameral When Building or training a net that needs the layers/parallel Op (also usable as a Parallel cam).

Welvet examples

69. Lucy density: synthetic organism

Part: V · SystemsPackage: github.com/openfluke/welvet/lucyStatus: ok: ✅ BuildLPD When Adaptation, measurement, or evolution (lucy). Where import "github.com/openfluke/welvet/lucy" cd 69-luc

Welvet examples

70. CamSync: inter-cameral / cross-mesh weight blend

Part: VII · AppsPackage: github.com/openfluke/welvet/layers/parallelStatus: ok: ✅ CamSync When Building or training a net that needs the layers/parallel Op (also usable as a Parallel cam).

Welvet examples

71. runtime/dispatch: Op switchboard

When: walking a grid / Parallel / Sequential and you only have any cells.Where: github.com/openfluke/welvet/runtime/dispatchWhy: one Forward/Backward/Pack/SetDType/ApplyGrad path for every W

Welvet examples

72. model/wav2vec2: CTC ASR

When: speech to text (greedy CTC) from 16 kHz mono WAV/PCM.Where: github.com/openfluke/welvet/model/wav2vec2Why: run facebook/wav2vec2-base-960h in pure Go. Always proves Config + Vocab + Lo

Welvet examples

73. All layers: one forward each

When: you want a single command that proves every layers/* Op (plus seqmix + dispatch) actually forwards.Where: this folder imports every concrete layer package.Why: chapters 11-28 split by

Welvet examples

74. Kokoro: reserved empty slots

When: you specifically want Kokoro TTS.Where: welvet/model/kokoro and welvet/apps/kokoro: both empty placeholders today.Why: names reserved for a future port; do not import them yet. Workin

Welvet examples

75. Apps map: when / where / why

When: you need a product shell (chat, TTS, ASR, image) rather than a library Op.Where: welvet/apps/<name>: each is its own Go module.Why: engine packages never import apps; apps replace → w

Welvet examples

76. NativeCPU: opt-in welvet.cpp C ABI

Part: X · NativeCPU (welvet.cpp)Package: github.com/openfluke/welvet/nativecpp + layersStatus: ok: ✅ linux/amd64 + cgo When You want in-process welvet.cpp MatVec / NormalBP / TrainMSE inste

Welvet examples

77. NativeCPU: TrainNativeMSE per layer

Part: X · NativeCPU (welvet.cpp)Package: github.com/openfluke/welvet/nativecpp + layersStatus: ok: ✅ linux/amd64 + cgo When You want in-process welvet.cpp MatVec / NormalBP / TrainMSE inste

Welvet examples

78. NativeCPU: TrainMode stack / parallel / mesh / line

Part: X · NativeCPU (welvet.cpp)Package: github.com/openfluke/welvet/nativecpp + layersStatus: ok: ✅ linux/amd64 + cgo When You want in-process welvet.cpp MatVec / NormalBP / TrainMSE inste

Welvet examples

79. QNN: opt-in Hexagon HTP (welvet.qnn)

Part: XI · Hexagon NPU (welvet.qnn)Package: github.com/openfluke/welvet/nativeqnn + layersStatus: ok: ✅ windows/arm64 + cgo When You want Dense (and friends) on Qualcomm Hexagon HTP instead

Welvet examples

80. QNN: HTP layer tour (session ABI)

Part: XI · Hexagon NPU (welvet.qnn)Package: github.com/openfluke/welvet/nativeqnnStatus: ok: ✅ 12 layer forwards When Smoke-test every shipped Hexagon forward without running the full w2a m

Welvet examples

81. QNN: go_v_qnn RESULTS (large shapes)

Part: XI · Hexagon NPU (welvet.qnn)Package: w2a/suites/go_v_qnnStatus: ok: ✅ matrix published When You want the honest Snapdragon X numbers (not toy shapes). Where cd 81-qnn-results && sour

Cameral cookbook

01: BranchModes

When: different update rules (or idle/teacher) per camWhere: para.SetBranchModes(...) with parent NormalBPWhy: mix credit, freeze priors, distill, adversarial twins, surprise memory go run .

Cameral cookbook

02: Combine modes

When: choosing how hemisphere posts become one tensorWhere: Config.CombineWhy: avg/add ensemble; max WTA; sparsek; disagree; filter MoE; concat widths go run ./02_combine Exits non-zero if a

Cameral cookbook

03: CamSync

When: cams should share / pull weight DNAWhere: SetCamSync / SyncNow / CrossWhy: soft↔hard consensus, one-way teacher, groups, cross-layer go run ./03_camsync Exits non-zero if any proof fai

Cameral cookbook

04: CamKit / LRs / Rotate / DNA / Dream

When: schedules, plasticity, replay, speciationWhere: SetBranchLRs / SetRotateSchedule / SetCamKit / DreamPulseWhy: sleep cycles, soft freeze, DNA push/pull, offline consolidation go run ./0

Cameral cookbook

05: Every layer as cams

When: CNN / Mamba / LSTM / … as Parallel camsWhere: HemispheresFrom + twin OpsWhy: CamSync / Freeze / Train work on every hosted Op go run ./05_layers Exits non-zero if any proof fails (PASS

Cameral cookbook

06: Recipes

When: compose modes + sync + kit into behaviorsWhere: go run ./06_recipesWhy: teacher, sleep, debate, memory, dream, concat go run ./06_recipes Exits non-zero if any proof fails (PASS / FAIL