def test_ensemble_chooses_highest_average_confidence_3(self):
subresults = [
{
"span_start_probs": torch.FloatTensor([[0.0, 0.0, 0.9, 0.1]]),
"span_end_probs": torch.FloatTensor([[0.0, 0.0, 0.9, 0.1]]),
"best_span": torch.LongTensor([[2, 2]]),
"best_span_str": "cheese",
"question_tokens": ["What", "did", "Michael", "eat", "?"],
"passage_tokens": ["Michael", "ate", "cheese", "."],
},
{
"span_start_probs": torch.FloatTensor([[0.0, 0.0, 0.9, 0.1]]),
"span_end_probs": torch.FloatTensor([[0.0, 0.0, 0.9, 0.1]]),
"best_span": torch.LongTensor([[2, 2]]),
"best_span_str": "cheese",
"question_tokens": ["What", "did", "Michael", "eat", "?"],
"passage_tokens": ["Michael", "ate", "cheese", "."],
},
{
"span_start_probs": torch.FloatTensor([[0.9, 0.0, 0.0, 0.0]]),
"span_end_probs": torch.FloatTensor([[0.9, 0.0, 0.0, 0.0]]),
"best_span": torch.LongTensor([[0, 0]]),
"best_span_str": "What",
"question_tokens": ["What", "did", "Michael", "eat", "?"],
"passage_tokens": ["Michael", "ate", "cheese", "."],
},
]
numpy.testing.assert_almost_equal(
ensemble(subresults).data[0].cpu().numpy(), torch.LongTensor([2, 2]).numpy()
)
def test_ensemble_chooses_highest_average_confidence_2(self):
subresults = [{
u"span_start_probs":
torch.FloatTensor([[0.9, 0.0, 0.0, 0.0]]),
u"span_end_probs":
torch.FloatTensor([[0.9, 0.0, 0.0, 0.0]]),
u"best_span":
torch.LongTensor([[0, 0]]),
u"best_span_str":
u"What",
u"question_tokens": [u"What", u"did", u"Michael", u"eat", u"?"],
u"passage_tokens": [u"Michael", u"ate", u"cheese", u"."]
}, {
u"span_start_probs":
torch.FloatTensor([[0.0, 0.0, 1.0, 0.0]]),
u"span_end_probs":
torch.FloatTensor([[0.0, 0.0, 1.0, 0.0]]),
u"best_span":
torch.LongTensor([[2, 2]]),
u"best_span_str":
u"cheese",
u"question_tokens": [u"What", u"did", u"Michael", u"eat", u"?"],
u"passage_tokens": [u"Michael", u"ate", u"cheese", u"."]
}]
numpy.testing.assert_almost_equal(
ensemble(subresults).data[0].cpu().numpy(),
torch.LongTensor([2, 2]).cpu().numpy())
def test_ensemble_chooses_highest_average_confidence_3(self):
subresults = [
{
"span_start_probs": torch.FloatTensor([[0.0, 0.0, 0.9, 0.1]]),
"span_end_probs": torch.FloatTensor([[0.0, 0.0, 0.9, 0.1]]),
"best_span": torch.LongTensor([[2, 2]]),
"best_span_str": "cheese",
"question_tokens": ["What", "did", "Michael", "eat", "?"],
"passage_tokens": ["Michael", "ate", "cheese", "."]
},
{
"span_start_probs": torch.FloatTensor([[0.0, 0.0, 0.9, 0.1]]),
"span_end_probs": torch.FloatTensor([[0.0, 0.0, 0.9, 0.1]]),
"best_span": torch.LongTensor([[2, 2]]),
"best_span_str": "cheese",
"question_tokens": ["What", "did", "Michael", "eat", "?"],
"passage_tokens": ["Michael", "ate", "cheese", "."]
},
{
"span_start_probs": torch.FloatTensor([[0.9, 0.0, 0.0, 0.0]]),
"span_end_probs": torch.FloatTensor([[0.9, 0.0, 0.0, 0.0]]),
"best_span": torch.LongTensor([[0, 0]]),
"best_span_str": "What",
"question_tokens": ["What", "did", "Michael", "eat", "?"],
"passage_tokens": ["Michael", "ate", "cheese", "."]
}
]
numpy.testing.assert_almost_equal(
ensemble(subresults).data[0].cpu().numpy(),
torch.LongTensor([2, 2]).numpy())