Skip to content

GitLab

Switch branch/tag
Find file Normal viewHistoryPermalink
test_tokenization_dpr.py 3.5 KB
Newer Older
Quentin Lhoest's avatar
Add DPR model (#5279)  
Quentin Lhoest committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16 
# coding=utf-8
# Copyright 2020 Huggingface
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
Quentin Lhoest's avatar
Fix tests imports dpr (#5576)  
Quentin Lhoest committed
17 
from transformers.testing_utils import slow
Quentin Lhoest's avatar
Add DPR model (#5279)  
Quentin Lhoest committed
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89 
from transformers.tokenization_dpr import (
    DPRContextEncoderTokenizer,
    DPRContextEncoderTokenizerFast,
    DPRQuestionEncoderTokenizer,
    DPRQuestionEncoderTokenizerFast,
    DPRReaderOutput,
    DPRReaderTokenizer,
    DPRReaderTokenizerFast,
)
from transformers.tokenization_utils_base import BatchEncoding

from .test_tokenization_bert import BertTokenizationTest


class DPRContextEncoderTokenizationTest(BertTokenizationTest):

    tokenizer_class = DPRContextEncoderTokenizer

    def get_rust_tokenizer(self, **kwargs):
        return DPRContextEncoderTokenizerFast.from_pretrained(self.tmpdirname, **kwargs)


class DPRQuestionEncoderTokenizationTest(BertTokenizationTest):

    tokenizer_class = DPRQuestionEncoderTokenizer

    def get_rust_tokenizer(self, **kwargs):
        return DPRQuestionEncoderTokenizerFast.from_pretrained(self.tmpdirname, **kwargs)


class DPRReaderTokenizationTest(BertTokenizationTest):

    tokenizer_class = DPRReaderTokenizer

    def get_rust_tokenizer(self, **kwargs):
        return DPRReaderTokenizerFast.from_pretrained(self.tmpdirname, **kwargs)

    @slow
    def test_decode_best_spans(self):
        tokenizer = self.tokenizer_class.from_pretrained("bert-base-uncased")

        text_1 = tokenizer.encode("question sequence", add_special_tokens=False)
        text_2 = tokenizer.encode("title sequence", add_special_tokens=False)
        text_3 = tokenizer.encode("text sequence " * 4, add_special_tokens=False)
        input_ids = [[101] + text_1 + [102] + text_2 + [102] + text_3]
        reader_input = BatchEncoding({"input_ids": input_ids})

        start_logits = [[0] * len(input_ids[0])]
        end_logits = [[0] * len(input_ids[0])]
        relevance_logits = [0]
        reader_output = DPRReaderOutput(start_logits, end_logits, relevance_logits)

        start_index, end_index = 8, 9
        start_logits[0][start_index] = 10
        end_logits[0][end_index] = 10
        predicted_spans = tokenizer.decode_best_spans(reader_input, reader_output)
        self.assertEqual(predicted_spans[0].start_index, start_index)
        self.assertEqual(predicted_spans[0].end_index, end_index)
        self.assertEqual(predicted_spans[0].doc_id, 0)

    @slow
    def test_call(self):
        tokenizer = self.tokenizer_class.from_pretrained("bert-base-uncased")

        text_1 = tokenizer.encode("question sequence", add_special_tokens=False)
        text_2 = tokenizer.encode("title sequence", add_special_tokens=False)
        text_3 = tokenizer.encode("text sequence", add_special_tokens=False)
        expected_input_ids = [101] + text_1 + [102] + text_2 + [102] + text_3
        encoded_input = tokenizer(questions=["question sequence"], titles=["title sequence"], texts=["text sequence"])
        self.assertIn("input_ids", encoded_input)
        self.assertIn("attention_mask", encoded_input)
        self.assertListEqual(encoded_input["input_ids"][0], expected_input_ids)