Does it support shape verification for simulators?

Yes, it includes standardized patterns for verifying that simulators and adapters return the correct dictionary keys, data types, and tensor shapes.

How does it handle non-deterministic tests in CI?

It enforces the use of local random number generators with fixed seeds (np.random.default_rng) rather than global state, ensuring reproducible results across different environments.

Can this skill help with slow machine learning tests?

Yes, it provides specific patterns for mock-based testing of inference networks, allowing you to verify logic and shapes without running full, time-consuming training cycles.

Why do I need to set the Keras backend in conftest.py?

Keras 3 locks the backend upon the first import. This skill ensures KERAS_BACKEND is set to torch before any imports to prevent initialization failures or switching backends mid-process.

BayesFlow Testing Suite

Name: BayesFlow Testing Suite
Author: matthiaskloft

bymatthiaskloft

0•

Seguridad y Pruebas

Standardizes the testing process for BayesFlow extensions by implementing best practices for Keras backends, shape verification, and simulation mocks.

This skill provides specialized guidance for developing robust test suites for BayesFlow extension packages, ensuring compatibility with Keras 3 and the PyTorch backend. It automates the setup of critical environment variables, defines patterns for shape-testing simulators, provides templates for configuration serialization, and implements specialized invariance/equivariance tests for probabilistic architectures. By using this skill, developers can avoid common pitfalls like backend initialization errors and non-deterministic results, leading to more reliable and maintainable simulation-based inference code.

Características Principales

01Standardized directory structure and conftest.py setup patterns

02Specialized invariance and equivariance testing for symmetry-aware architectures

03RNG seeding best practices to eliminate flaky CI tests

040 GitHub stars

05Automated mock generation for inference and summary networks

06Automatic KERAS_BACKEND configuration for Keras 3 and PyTorch integration

Casos de Uso

01Creating unit tests for custom simulators and data adapters

02Validating the architectural symmetries of neural networks in IRT models

03Debugging backend initialization issues in Keras-based ML projects

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add matthiaskloft/claude-skills bayesflow-testing

For use in Claude.ai and ChatGPT

Características Principales

01Standardized directory structure and conftest.py setup patterns

02Specialized invariance and equivariance testing for symmetry-aware architectures

03RNG seeding best practices to eliminate flaky CI tests

040 GitHub stars

05Automated mock generation for inference and summary networks

06Automatic KERAS_BACKEND configuration for Keras 3 and PyTorch integration

Casos de Uso

01Creating unit tests for custom simulators and data adapters

02Validating the architectural symmetries of neural networks in IRT models

03Debugging backend initialization issues in Keras-based ML projects

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add matthiaskloft/claude-skills bayesflow-testing

For use in Claude.ai and ChatGPT