test_train.py 文件源码

def test_single_domain():
    docs = [{'html': 'foo{} bar'.format(i % 4),
             'url': 'http://example.com/{}'.format(i),
             'relevant': i % 2 == 0}
            for i in range(10)]
    result = train_model(docs)
    pprint(attr.asdict(result.meta))
    assert lst_as_dict(result.meta.advice)[:2] == [
        {'kind': 'Warning',
         'text': "Only 1 relevant domain in data means that it's impossible to do "
                 'cross-validation across domains, and will likely result in '
                 'model over-fitting.'},
        {'kind': 'Warning',
         'text': 'Number of human labeled documents is just 10, consider having '
                 'at least 100 labeled.'},
    ]
    assert lst_as_dict(result.meta.description)[:3] == [
        {'heading': 'Dataset',
         'text': '10 documents, 10 labeled across 1 domain.'},
        {'heading': 'Class balance',
         'text': '50% relevant, 50% not relevant.'},
        {'heading': 'Metrics', 'text': ''},
    ]
    assert result.model is None