Streamlines the creation and debugging of tests for BayesFlow extension packages using PyTorch and Keras 3.
This skill provides standardized testing patterns and best practices specifically for BayesFlow extensions. It ensures critical configurations, such as setting the Keras backend to PyTorch, are handled correctly before imports to avoid runtime failures. The skill guides developers through implementing shape testing for simulators, mock-based approximator testing for faster CI cycles, and advanced symmetry verification (invariance and equivariance) for neural architectures. By enforcing deterministic RNG practices and structured test layouts, it helps eliminate flaky tests and ensures robust serialization of model configurations.
Características Principales
01Validation logic for configuration dataclass serialization and roundtrips
02Template-based shape and output verification for BayesFlow simulators
03Automatic Keras 3 backend configuration setup in conftest.py
04Mock-based inference network strategies for fast, training-free unit tests
050 GitHub stars
06Invariance and equivariance testing patterns for symmetry-aware architectures
Casos de Uso
01Migrating BayesFlow extensions to Keras 3 with a PyTorch backend
02Developing unit tests for custom BayesFlow simulators and summary networks
03Implementing robust CI/CD pipelines with deterministic machine learning tests
