Unit Testing¶

The library includes several tests to verify that it is running correctly. Before simulating problems, it is recommended to run the available unit tests, located in the directory:

tests/unit/environments

We provide the following types of tests:

Check Multi-Task Solution Test¶

This script runs tests for Multi-Tasking environments, validating solution correctness under multiple conditions:

Random instances: Tests across batches with different variants generated through random attribute sampling.
All variants: Creates and validates instances for every available environment variant.
Different agent selectors: Tests all agent selectors — AgentSelector, RandomSelector, and SmallestTimeAgentSelector.
Different configurations: Evaluates combinations of various numbers of agents and nodes.

Configuration values can be customized using the following command-line parameters:

--device can be CPU or GPU.
--batch can be any integer.
--num_agents can be a single or multiple integer values. To define n values, use the parameter n times.
--num_nodes can be a signle or multiple integer values. To define n values, use the parameter n times.

To execute the test, run:

pytest --device cpu --batch 1 --num_agents 2 --num_nodes 11 --num_nodes 13 check_solution_mt_test.py

Check Solution Test¶

This script performs solution validation for non–Multi-Tasking environments:

Random instances: Tests random variants generated across batches.
Different agent selectors: Includes AgentSelector, RandomSelector, and SmallestTimeAgentSelector.
Different configurations: Runs across multiple agent and node configurations.

Configuration values can be customized using the following command-line parameters:

--device can be CPU or GPU.
--batch can be any integer.
--num_agents can be a single or multiple integer values. To define n values, use the parameter n times.
--num_nodes can be a signle or multiple integer values. To define n values, use the parameter n times.

To execute the test, run:

pytest --device cpu --batch 1 --num_agents 2 --num_nodes 11 --num_nodes 13 check_solution_test.py

Reset Test¶

The Reset Test ensures that environments reset and behave correctly under different configurations. It includes three types of checks:

Reset tests: Verify that both benchmarking and standard instances reset without errors.
Observation tests: Confirm that environment observations are correctly produced.
Agent iteration tests: Validate that agent selection and iteration work as expected for all selectors and generator types.

To execute the test, run:

pytest reset_test.py

Reset Benchmarking Test¶

Validates the correct behavior of benchmarking resets.

To execute the test, run:

pytest reset_bench_test.py

Seed Test¶

The Seed Test verifies reproducibility and consistency across environments:

Seed consistency: Checks whether identical seeds yield identical outputs, and different seeds yield different ones.
Instance generator: Validates equivalence between benchmarking and standard instance generators.

To execute the test, run:

pytest seed_test.py

Seed Benchmarking Test¶

Validates seed reproducibility for benchmarking environments.

To execute the test, run:

pytest seed_bench_test.py