MOSS_WebSearchTool/score_utils.py at main · OpenLMLab/MOSS_WebSearchTool · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
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
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
# Copyright 2023 piglake
#
# 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.
import ngram
import torch
import torch.nn as nn
from nltk import FreqDist
from nltk import ngrams
#import fasttext.util

cos = nn.CosineSimilarity(dim=1, eps=1e-6)

class score_measure:
    def __init__(self, lang="en"):
        import torch.distributed as dist
        # dist.init_process_group('gloo', init_method='file:///tmp/donot_care', rank=0, world_size=1)

        folder = "word2vec-pytorch/weights/cbow_WikiText103"
        #device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

        device = torch.device("cpu")

        #model = torch.load(f"./{folder}/model.pt", device).module.eval()#map_location='cuda:0')
        vocab = torch.load(f"./{folder}/vocab.pt")

        self.input_embedding, self.output_embedding = torch.load("my_"+lang+"_embedding.pt"), None#model.embeddings, model.linear.weight
        self.vocab = vocab
        self.cos_sim = nn.CosineSimilarity(dim=1, eps=1e-6)

    def infer(self, sentence, key_words):
        score = 0
        #print(sentence)
        #print(key_words)
        ids = torch.tensor([ int(self.vocab[i]) for i in sentence.split() ]).unsqueeze(0)
        z = self.input_embedding(ids).mean(axis=1).squeeze()
        #print(z.shape)
        for key_word in key_words:
            index =  self.vocab[key_word]
            z_x = self.input_embedding(torch.tensor(index).unsqueeze(0))
            score += self.cos_sim(z, z_x)
            #score += torch.dot(z, z_x)

        return score.cpu().item()


def kmp(pattern, text):
    # Initialize variables
    p_str = pattern
    t_str = text
    lss_arr = []

    # Iterate through the text string
    for i in range(len(t_str)):
        if t_str[i] == p_str[0]:
            lss_arr.append(0)
        else:
            j = 0
            while j < len(lss_arr) and lss_arr[j] + 1 < len(p_str):
                j += 1
            lss_arr[j] += 1

    # Return the longest suffix array
    return lss_arr


def score(key_words, sentence,ft):
    res = 0
    #s = ngram.NGram

    for key_word in key_words.split():
        #res += ngram.NGram.compare(key_word, sentence,N=2)
        #tmp = 0
        #ngram_list = ngrams(key_word, 2)

        #for gram in ngram_list:
            #gram: tuple

            #tmp += sentence.count("".join(gram))
        key_embedding = ft.get_word_vector(key_word)
        #res += 1 if sentence.find(key_word) > 0 else 0

        vector = ft.get_word_vector(sentence)  # 300-dim vector

        #print(key_embedding)
    #print(res)
        from numpy import dot
        from numpy.linalg import norm

        cos_sim = dot(key_embedding, vector)/(norm(key_embedding)*norm(vector))
        res += cos_sim
    return res

def score_2(key_words, sentence,):
    res = 0
    #s = ngram.NGram

    for key_word in key_words.split():
        #res += ngram.NGram.compare(key_word, sentence,N=2)
        #tmp = 0
        #ngram_list = ngrams(key_word, 2)

        #for gram in ngram_list:
            #gram: tuple

            #tmp += sentence.count("".join(gram))

        res += 1 if sentence.find(key_word) > 0 else 0

    return res

def score_3(key_words, sentence,  measure):
    if sentence.split():
        return measure.infer(sentence, key_words)
    else:
        return -10000

if __name__ == "__main__":
    ft = fasttext.load_model('cc.en.300.bin')
    #ft = fasttext.load_model('cc.zh.300.bin')
    res = score("hi", \
        " hello, I am Jones", \
            ft = ft
        )
    print(res)
    ft = fasttext.load_model('cc.zh.300.bin')
    res = score("季节", \
        "我最喜欢春天，你呢", \
            ft = ft
        )
    print(res)