OCDocker.OCScore.Analysis.Plotting.ArchitecturePlots module

Publication-oriented OCScore architecture diagrams.

Usage:

import OCDocker.OCScore.Analysis.Plotting.ArchitecturePlots as ocarchplot

class OCDocker.OCScore.Analysis.Plotting.ArchitecturePlots.ArchitectureBlock(label, dim, kind, detail='')[source]

Bases: object

Single visible block in an architecture diagram.

Parameters:

  • label (str)

  • dim (int | None)

  • kind (str)

  • detail (str)

label

Text label shown in the block.

Type:

str

dim

Feature width or output dimensionality represented by the block.

Type:

int or None

kind

Visual role used for colors, such as input, encoder, or head.

Type:

str

detail

Optional compact annotation shown below the dimensionality.

Type:

str, default=””

class OCDocker.OCScore.Analysis.Plotting.ArchitecturePlots.ArchitectureDiagram(task, main, auxiliary=(), notes=())[source]

Bases: object

Normalized architecture diagram specification.

Parameters:
task

Name of the task or model family represented by the diagram.

Type:

str

main

Primary left-to-right model path.

Type:

tuple of ArchitectureBlock

auxiliary

Optional auxiliary branch, used for reconstruction decoders.

Type:

tuple of ArchitectureBlock, default=()

notes

Compact notes rendered below the diagram.

Type:

tuple of str, default=()

OCDocker.OCScore.Analysis.Plotting.ArchitecturePlots.load_architecture_document(source)[source]

Load an architecture document.

Parameters:

source (str or pathlib.Path) – Architecture JSON/YAML file or a best_model/ export directory.

Returns:

Parsed architecture document and the resolved architecture file path.

Return type:

tuple of dict and pathlib.Path

Raises:

  • FileNotFoundError – If the source path or export architecture file does not exist.

  • ValueError – If the architecture file does not contain a mapping.

OCDocker.OCScore.Analysis.Plotting.ArchitecturePlots.normalize_architecture(document)[source]

Normalize architecture metadata for plotting.

Parameters:

document (Mapping) – Exported OCScore architecture document or a manual architecture mapping.

Returns:

Diagram-ready architecture with primary and optional auxiliary blocks.

Return type:

ArchitectureDiagram

Raises:

ValueError – If a manual layers section is present but is not a list.

OCDocker.OCScore.Analysis.Plotting.ArchitecturePlots.plot_architecture_diagram(architecture, *, title=None, figsize=None, include_decoder=False)[source]

Build a publication-oriented neural-network architecture figure.

Parameters:

  • architecture (Mapping or ArchitectureDiagram) – Architecture metadata or a pre-normalized diagram specification.

  • title (str, optional) – Custom title. When omitted, the task name is used.

  • figsize (tuple of float, optional) – Explicit matplotlib figure size.

  • include_decoder (bool, default=False) – Whether to include the auxiliary reconstruction decoder branch.

Returns:

Figure and axes containing the rendered architecture diagram.

Return type:

tuple of matplotlib.figure.Figure and matplotlib.axes.Axes

Raises:

ValueError – If the architecture has no layers to plot.

OCDocker.OCScore.Analysis.Plotting.ArchitecturePlots.save_architecture_figures(source, output_dir, *, formats=('png', 'svg', 'pdf'), dpi=220, title=None, basename='architecture', include_decoder=False)[source]

Save architecture figures to disk.

Parameters:

  • source (str or pathlib.Path) – Architecture JSON/YAML file or best_model/ export directory.

  • output_dir (str or pathlib.Path) – Directory where figure files will be written.

  • formats (sequence of str, default=("png", "svg", "pdf")) – Output formats accepted by matplotlib.

  • dpi (int, default=220) – Raster output resolution.

  • title (str, optional) – Custom figure title.

  • basename (str, default="architecture") – Output filename stem.

  • include_decoder (bool, default=False) – Whether to include the auxiliary reconstruction decoder branch.

Returns:

Mapping from output format to absolute file path, plus source.

Return type:

dict

Raises:

ValueError – If no valid output format is provided.