Testing Philosophy

RTnn uses Python’s built-in unittest framework for comprehensive testing.

Running Tests

Run all tests:

python -m unittest discover tests -v

Run specific test modules:

python -m unittest tests.test_rnn -v
python -m unittest tests.test_fcn -v
python -m unittest tests.test_transformer -v
python -m unittest tests.test_dataset -v

Run with test runner (rich output):

python tests/test_runner.py

Test Coverage

Test Module	Coverage
test_rnn.py	LSTM and GRU models (initialization, forward pass, gradients)
test_fcn.py	FCN model (forward pass, dimension expansion)
test_transformer.py	Transformer encoder (attention, blocks)
test_dataset.py	DataPreprocessor (loading, normalization, indexing)
test_evaluater.py	Loss functions and metrics
test_model_loader.py	Model factory function

Test Data Generation

Tests use dynamically generated NetCDF files that mimic real data:

• File naming: rtnetcdf_XXX_YYYY.nc

• Multi-year and multi-processor support

• Realistic dimension ordering

• All required variables present

Example test structure

class TestRNN_LSTM(unittest.TestCase):
    def setUp(self):
        self.model = RNN_LSTM(
            feature_channel=6,
            output_channel=4,
            hidden_size=64,
            num_layers=2
        )

    def test_forward_shape(self):
        x = torch.randn(32, 6, 10)
        y = self.model(x)
        self.assertEqual(y.shape, (32, 4, 10))

Running a Single Test

python -m unittest tests.test_rnn.TestRNN_LSTM.test_forward_shape -v

Continuous Integration

GitHub Actions automatically runs tests on:

• Push to master branch

• Pull requests to master

• Manual workflow dispatch