This pull request adds a test for the original MixIT model published for NeurIPS 2020. Since this test uses a lot of the code shared among the model variants, it already reaches 80 % of line coverage. The test runs the training for 4 steps and on 4 mixtures, as this is the minimum amount for the full coverage achievable coverage with this test. While this test runs a lot of code, I did not make any assertions about its side effects, like finding the trained model in the configured location.
This pull request adds a test for the original MixIT model published for NeurIPS 2020. Since this test uses a lot of the code shared among the model variants, it already reaches 80 % of line coverage. The test runs the training for 4 steps and on 4 mixtures, as this is the minimum amount for the full coverage achievable coverage with this test. While this test runs a lot of code, I did not make any assertions about its side effects, like finding the trained model in the configured location.