Implementation:Tensorflow Tfjs Model Summary Test

Knowledge Sources	Tensorflow_Tfjs
Domains	Testing, Layers_API
Last Updated	2026-02-10 06:00 GMT

Overview

This test suite validates the LayersModel.summary() and Sequential.summary() methods, which print a human-readable table of model architecture information to the console. The tests verify the exact formatting of summary output including layer names, types, input shapes, output shapes, parameter counts, separator lines, and totals. It also tests customization options: custom line length, custom column positions, and custom print functions that redirect output away from console.log.

Code Reference

Source Location: tfjs-layers/src/model_summary_test.ts (337 lines)

Repository: GitHub

Test Describe Blocks

LayersModel.summary - All summary tests including:
- Sequential model: one layer (default formatting)
- Sequential model: one layer with custom lineLength
- Sequential model: one layer with custom positions
- Sequential model: one layer with custom printFn
- Sequential model: multiple layers
- Functional model summary
- Multi-input model summary
- Multi-output model summary
- Nested model summary
- Non-trainable params display

I/O Contract

Inputs to tests:

Sequential models with one or more Dense layers
Functional models with single and multiple inputs/outputs
Summary configuration: lineLength (number), positions (number[]), printFn (custom function)

Expected outputs/assertions:

Console output lines match exact expected strings including:
- Header: 'Layer (type) Input Shape Output shape Param # '
- Layer rows with correctly formatted name, type, shapes, and param count
- Separator lines using = and _ characters
- Footer with total params, trainable params, and non-trainable params
Custom lineLength adjusts column widths
Custom positions array adjusts column boundaries
Custom printFn receives the same output that console.log would

Usage Example

describeMathCPU('LayersModel.summary', () => {
  let consoleLogHistory: string[];
  beforeEach(() => {
    consoleLogHistory = [];
    spyOn(console, 'log').and.callFake((message?: string) => {
      consoleLogHistory.push(message);
    });
  });

  it('Sequential model: one layer', () => {
    const layerName = getRandomLayerOrModelName();
    const model = tfl.sequential({
      layers: [tfl.layers.dense({units: 3, inputShape: [10], name: layerName})]
    });
    model.summary();
    expect(consoleLogHistory).toEqual([
      '__________________________...',
      'Layer (type)                Input Shape...',
      '==========================...',
      `${layerName} (Dense)        [[null,10]]...`,
      '==========================...',
      'Total params: 33', 'Trainable params: 33',
      'Non-trainable params: 0', '__________________________...'
    ]);
  });
});

Test Coverage Summary

Category	Count	Details
Default Summary	2+	Single and multi-layer Sequential models
Custom Formatting	3	lineLength, positions, printFn
Model Types	4+	Sequential, Functional, multi-input, multi-output
Nested Models	1+	Models containing sub-models
Non-trainable Params	1+	Correct param count split
Test Environment	CPU	`describeMathCPU`

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment