squadds.datasets package#

Submodules#

squadds.datasets.planar_capacitance_schema module#

Schema and conventions of the SQuADDS/SQuADDS_PlanarCapacitance dataset (pure Python; pyarrow imported lazily).

Conventions shared by every table:

  • Capacitance in fF, Maxwell (short-circuit) signs: diagonal > 0, off-diagonal < 0; mutual_fF = -C_fF off the diagonal.

  • Matrices are long format, one row per entry (net_a, net_b) with a_idx <= b_idx in the design’s net_order. Ground (G) is always the first net. Nets are never merged: N is inferred from the layout.

  • design_id = "ref/<collection>/<design>"; layout_id = "sha256:<file bytes>"; geom_hash = sha256 of the canonical polygon multiset (layer, datatype, vertices rounded to 1 nm, sorted) = the duplicate-detection key across SQuADDS.

  • Every processed row carries prior_version / embedding_version and the sha256 of the code that produced it.

  • Test-set labels (split == "test") live in the private SQuADDS_PlanarCapacitance-heldout, except the interdigital in-family test designs (exp6-8), whose labels were already public in SQuADDS/SQuADDS_DB and are released here too.

    from squadds.datasets.planar_capacitance_schema import TABLES, arrow_schema arrow_schema(“labels”)

squadds.datasets.planar_capacitance_schema.arrow_schema(table)[source]#

pyarrow schema of one table (column order and types as written).

squadds.datasets.planar_capacitance_schema.column_names(table)[source]#
Return type:

list[str]

squadds.datasets.planar_capacitance_schema.squadds_config(collection)[source]#

SQuADDS_DB-style config name (component-ComponentClass-cap_matrix) of a collection, or None (canonical store only).

Return type:

str | None

Module contents#

Dataset schemas and loaders for SQuADDS (planar capacitance).