embed_anything 0.6.7

Embed anything at lightning speed
Documentation 
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
141
142
143

use anyhow::Error as E;
use candle_core::{Device, Tensor};
use ndarray::Array2;
use tokenizers::Tokenizer;

pub fn tokenize_batch(
    tokenizer: &Tokenizer,
    text_batch: &[&str],
    device: &Device,
) -> anyhow::Result<(Tensor, Tensor)> {
    let tokens = tokenizer
        .encode_batch(text_batch.to_vec(), true)
        .map_err(E::msg)?;
    let token_ids = tokens
        .iter()
        .map(|tokens| {
            let tokens = tokens.get_ids().to_vec();
            Tensor::new(tokens.as_slice(), device)
        })
        .collect::<candle_core::Result<Vec<_>>>()?;
    let attention_mask = tokens
        .iter()
        .map(|tokens| {
            let tokens = tokens.get_attention_mask().to_vec();
            Tensor::new(tokens.as_slice(), device)
        })
        .collect::<candle_core::Result<Vec<_>>>()?;

    Ok((
        Tensor::stack(&token_ids, 0)?,
        Tensor::stack(&attention_mask, 0)?,
    ))
}

pub fn get_attention_mask(
    tokenizer: &Tokenizer,
    text_batch: &[String],
    device: &Device,
) -> anyhow::Result<Tensor> {
    let tokens = tokenizer
        .encode_batch(text_batch.to_vec(), true)
        .map_err(E::msg)?;

    let attention_mask = tokens
        .iter()
        .map(|tokens| {
            let tokens = tokens.get_attention_mask().to_vec();
            Tensor::new(tokens.as_slice(), device)
        })
        .collect::<candle_core::Result<Vec<_>>>()?;
    Ok(Tensor::stack(&attention_mask, 0)?)
}

pub fn get_attention_mask_ndarray(
    tokenizer: &Tokenizer,
    text_batch: &[&str],
) -> anyhow::Result<Array2<i64>> {
    let attention_mask = tokenizer
        .encode_batch(text_batch.to_vec(), true)
        .map_err(E::msg)?
        .iter()
        .map(|tokens| {
            tokens
                .get_attention_mask()
                .iter()
                .map(|&id| id as i64)
                .collect::<Vec<i64>>()
        })
        .collect::<Vec<Vec<i64>>>();

    let attention_mask_array = Array2::from_shape_vec(
        (attention_mask.len(), attention_mask[0].len()),
        attention_mask.into_iter().flatten().collect::<Vec<i64>>(),
    )
    .unwrap();
    Ok(attention_mask_array)
}

pub fn tokenize_batch_ndarray(
    tokenizer: &Tokenizer,
    text_batch: &[&str],
) -> anyhow::Result<(Array2<i64>, Array2<i64>)> {
    let tokens = tokenizer
        .encode_batch(text_batch.to_vec(), true)
        .map_err(E::msg)?;
    let token_ids = tokens
        .iter()
        .map(|tokens| {
            tokens
                .get_ids()
                .iter()
                .map(|&id| id as i64)
                .collect::<Vec<i64>>()
        })
        .collect::<Vec<Vec<i64>>>();
    let attention_mask = tokens
        .iter()
        .map(|tokens| {
            tokens
                .get_attention_mask()
                .to_vec()
                .iter()
                .map(|&id| id as i64)
                .collect::<Vec<i64>>()
        })
        .collect::<Vec<Vec<i64>>>();
    let token_ids_array = Array2::from_shape_vec(
        (token_ids.len(), token_ids[0].len()),
        token_ids.into_iter().flatten().collect::<Vec<i64>>(),
    )
    .unwrap();
    let attention_mask_array = Array2::from_shape_vec(
        (attention_mask.len(), attention_mask[0].len()),
        attention_mask.into_iter().flatten().collect::<Vec<i64>>(),
    )
    .unwrap();
    Ok((token_ids_array, attention_mask_array))
}

pub fn get_type_ids_ndarray(
    tokenizer: &Tokenizer,
    text_batch: &[&str],
) -> anyhow::Result<Array2<i64>> {
    let token_ids = tokenizer
        .encode_batch(text_batch.to_vec(), true)
        .map_err(E::msg)?
        .iter()
        .map(|tokens| {
            tokens
                .get_type_ids()
                .iter()
                .map(|&id| id as i64)
                .collect::<Vec<i64>>()
        })
        .collect::<Vec<Vec<i64>>>();

    let token_ids_array = Array2::from_shape_vec(
        (token_ids.len(), token_ids[0].len()),
        token_ids.into_iter().flatten().collect::<Vec<i64>>(),
    )
    .unwrap();
    Ok(token_ids_array)
}