json335 B

ONNX Inference Contract - Two-Input Broadcast Add Expected Output

The reference-evaluator output for the Two-Input Broadcast Add model and supplied input, including shape, dtype, exact values and a 1e-6 tolerance.

Model inference testing Model evaluation JSON parsing

Preview — first 24 linesjson

{
  "format": "Novus ONNX tensor fixture v1",
  "tensors": {
    "Y": {
      "dtype": "float32",
      "shape": [
        2,
        2
      ],
      "values": [
        [
          1.5,
          1.0
        ],
        [
          3.5,
          3.0
        ]
      ]
    }
  },
  "absoluteTolerance": 1e-06
}

Specifications

Contract: two-input-broadcast
Tensors: 1
Dtype: float32
Shape: 2x2
Absolute Tolerance: 0.000001
Role: expected output

Testing contract

Reference control

Scenario: Compare tensor Y from the paired ONNX execution with this expected tensor using the published absolute tolerance.
Expected result: Y must equal [[1.5, 1.0], [3.5, 3.0]] within absolute tolerance 1e-6.

What is a .json file?

JSON (JavaScript Object Notation) is a lightweight, text-based data-interchange format representing objects, arrays, strings, numbers, booleans, and null. It is language-independent, human-readable, and the dominant format for web APIs and configuration. It requires a single well-formed root value.

How to use this file

Use an example JSON file to test parsers and serializers, schema validation, Unicode and number-precision handling, and API request or response processing.

How to use this file for testing

“ONNX Inference Contract - Two-Input Broadcast Add Expected Output” is a deterministic Novus Examples fixture for Model inference testing, Model evaluation, JSON parsing. Small, validated ONNX graphs paired with named JSON inputs and expected tensor outputs. Use them to test runtime loading, dtype and shape handling, broadcasting, dynamic batches, reduction behavior, and numeric tolerances without downloading a production-sized model.

Documented properties for this file: expected output. Compare results against paired or grouped companions on this page when present (clean↔damaged, searchable↔scanned, or format twins) so scores stay reproducible across runs.

Download the file once, keep the path stable in CI or local scripts, and treat the spec table as the contract: dimensions, seeds, field lists, and roles are intentional. Corrupt or invalid samples are labelled as such — expect parsers to fail loudly rather than silently accept them.

AI/ML fixtures are fully synthetic with documented schemas — no real people or data. Test data loaders, tokenizers, annotation converters, embedding/vector stores, or eval-metric parsers against the known structure and fixed seeds.

Code examples

import json

with open("two-input-broadcast-expected-output.json") as f:
    data = json.load(f)
print(type(data), len(data))

Generated by generation/p6_core.py. Free for any use, no attribution required — license.