Expand generalized NCap v0 and add accepted universal geometry v1.1

README.md

@@ -5,6 +5,10 @@ configs:
   data_files:
   - split: train
     path: "metadata/static-embedding-v0.parquet"
+- config_name: universal-geometry-v1
+  data_files:
+  - split: train
+    path: "metadata/universal-geometry-v1.parquet"
 tags:
 - quantum
 - superconducting
@@ -16,7 +20,7 @@ tags:
 - computer-vision
 pretty_name: SQuADDS Layout Embeddings
 size_categories:
-- 1K<n<10K
+- 10K<n<100K
 ---
 
 <center>
@@ -25,7 +29,7 @@ size_categories:
 
 # SQuADDS Layout Embeddings
 
-Versioned layout representations for the 7,727 GDS artifacts in
+Versioned layout representations for the 24,106 GDS artifacts in
 [SQuADDS/SQuADDS_Layouts](https://huggingface.co/datasets/SQuADDS/SQuADDS_Layouts).
 
 ## Static embedding model `v0`
@@ -50,20 +54,58 @@ centered without distortion, supersampled at 4×, and reduced to 96×96. The
 large simulation-domain ground rectangle on layer `(1, 0)` is excluded so it
 does not hide the component shape.
 
-This is a deterministic static embedding, not a learned model. It is intended
-as a transparent baseline and as a stable input for similarity search. A future
-`v1` may use a compact learned raster or geometry-graph encoder, but will be
-published as a separate model rather than changing `v0` in place.
+This is a deterministic static embedding, not a learned model. It remains a
+transparent baseline and a stable input for similarity search.
 
 The exact block offsets, moment order, raster semantics, normalization
 statistics, and source schema are frozen in
 `metadata/static-embedding-v0.schema.json`.
 
+## Universal geometry model `v1`
+
+`universal-geometry-v1` is an additive 1,024-dimensional standard built from only
+a GDS file, a functional layer-role mapping, and the native design-parameter
+dictionary. Simulation targets are never embedded.
+
+| Block | Dimensions | Contents |
+| --- | ---: | --- |
+| Geometry metrics | 32 | Centered physical and morphological metrics; availability remains metadata rather than distorting cosine distance |
+| Multiscale shape | 768 | Target-blind, variance-selected full-spectrum 2D DCT coefficients from 96×96 signed-material and boundary-distance rasters |
+| Parameter controls | 224 | Stable signed feature hashing of canonical parameter paths after per-parameter centering and scaling |
+
+The metric block retains physical scale and role-specific conductor, etch, and
+port measurements. The shape block captures finger topology and boundary detail
+without storing a dense pixel tensor. Each block is normalized and explicitly
+weighted, and every fitted statistic and selected spectral frequency is frozen
+in the schema.
+
+Parameter identity is retained on every row through `parameter_names`,
+`parameter_values`, `parameter_hash_indices`, and `parameter_hash_signs`.
+`models/universal-geometry-v1/control-map.parquet` provides the global,
+auditable bridge back to the originating layout controls.
+
+This first v1 configuration contains all **20,062**
+`GeneralizedCapNInterdigital` designs. The encoder accepts foreign GDS layouts
+when their `(layer, datatype)` pairs are mapped to `conductor`, `etch`, or
+`port`; the cross-component reference normalization will be frozen in a later
+release after it is calibrated on the full SQuADDS catalogue.
+
+The complete input contract, block offsets, transforms, normalization
+statistics, and invariances are frozen in
+`models/universal-geometry-v1/schema.json`.
+
+The earlier 512-dimensional v1.0 candidate was rejected before release because
+its 8×8 low-pass shape crop lost finger detail and its common offsets collapsed
+cosine similarities. V1.1 passed paired topology, parameter-locality, shape,
+held-out capacitance-locality, and similarity-dynamic-range gates against v0
+across five deterministic held-out samples. Capacitance was never used to fit
+the embedding.
+
 ## Coverage and links
 
 | Component | Embeddings |
 | --- | ---: |
-| `GeneralizedCapNInterdigital` | 3,683 |
+| `GeneralizedCapNInterdigital` | 20,062 |
 | `CapNInterdigitalTee` | 894 |
 | `CavityClawRouteMeander` | 1,216 |
 | `TransmonCross` | 1,934 |
@@ -80,16 +122,20 @@ complete catalogue.
 ## Access
 
 ```python
-from squadds.layouts import StaticEmbeddingClient
+from squadds.layouts import LayoutEmbeddingClient, StaticEmbeddingClient
+
+v0 = StaticEmbeddingClient()  # Backward-compatible alias
+v1 = LayoutEmbeddingClient(version="v1")
 
-client = StaticEmbeddingClient()
-record = client.get("layout:sha256:<layout hash>")
-bitmap = client.shape_bitmap(record["layout_id"])
-neighbors = client.nearest(record["layout_id"], limit=10)
+record = v1.get("layout:sha256:<layout hash>")
+neighbors = v1.nearest(record["layout_id"], limit=10)
+schema = v1.schema()
+controls = v1.control_map()
 ```
 
 SQuADDS_DB rows can resolve the same vector directly with
-`SQuADDS_DB.get_layout_embedding(row)`. The SQuADDS MCP server also provides
+`SQuADDS_DB.get_layout_embedding(row, embedding_version="v1")`. Omitting the
+version preserves the v0 default. The SQuADDS MCP server also provides
 `get_layout_embedding` and `find_similar_layouts`.
 
 ## Provenance

metadata/static-embedding-v0.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2906181026db4d2de42f0c9a06d7a76635831d846e6cea9080549d3b3691d530
-size 43406616
+oid sha256:4511d820fa2fe3fcd90a0d35c0e89e61047fb80aec5af9182d45f5ea2706f6e4
+size 74418617

metadata/static-embedding-v0.schema.json

@@ -49,31 +49,31 @@
     "excluded": "simulation-domain ground layer (1, 0)"
   },
   "normalization": {
-    "parameter_mean": 742.4246055562062,
-    "parameter_std": 1452.0982784019416,
+    "parameter_mean": 341.3981965955698,
+    "parameter_std": 867.4239842516337,
     "moment_mean": [
-      8.838096523417622,
-      7.188891901268179,
-      0.1747443453386349,
-      0.2685946526564508,
-      0.04254547212419702,
-      0.009643734316262299,
-      0.0476611203028274,
-      0.026147025177978865,
-      5.5203095922678295e-05,
-      0.7486912895266242
+      7.7801819406323265,
+      6.502637244988525,
+      0.07048602809152436,
+      0.32053440673915906,
+      0.014431057708888933,
+      0.0040826806348275395,
+      0.046262563887785144,
+      0.02951256348409475,
+      -0.0001547993144423334,
+      0.7514288381372038
     ],
     "moment_std": [
-      2.160164605606158,
-      1.5977142258826054,
-      0.7178226519712876,
-      0.13826079773238997,
-      0.06949500744108629,
-      0.03117808679147505,
-      0.019558839847087152,
-      0.014266749202588542,
-      0.0011718407982694422,
-      0.18354901187363284
+      1.5530560058884653,
+      1.1076182871031315,
+      0.715898223432142,
+      0.1292091032011584,
+      0.04383759925758862,
+      0.01816855969349625,
+      0.02105482563282042,
+      0.01573011502221041,
+      0.0007823689518411512,
+      0.19325385936838735
     ],
     "parameter": "tanh(z-score)",
     "moments": "z-score, clipped to [-5, 5], then L2 normalized",

metadata/universal-geometry-v1.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80abb1747aaf53141098c94ed9d2c75f61281dda77f021bcbb7733964d73ab72
+size 125028053

models/universal-geometry-v1/control-map.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5694744084ea25981c9064c97a8eb96082616d3ee53bb580fcbaa109635cff0d
+size 6142

models/universal-geometry-v1/release-manifest.json

@@ -0,0 +1,18 @@
+{
+  "model": "universal-geometry-v1",
+  "embedding_schema_version": "1.1.0",
+  "component_name": "GeneralizedCapNInterdigital",
+  "rows": 20062,
+  "dimensions": 1024,
+  "files": {
+    "metadata/universal-geometry-v1.parquet": {
+      "sha256": "80abb1747aaf53141098c94ed9d2c75f61281dda77f021bcbb7733964d73ab72"
+    },
+    "models/universal-geometry-v1/schema.json": {
+      "sha256": "dcb90a925fdcd96d7760e098b2223f5745c0d6d9cda5d7be2f7bcf2ad3f819a3"
+    },
+    "models/universal-geometry-v1/control-map.parquet": {
+      "sha256": "5694744084ea25981c9064c97a8eb96082616d3ee53bb580fcbaa109635cff0d"
+    }
+  }
+}