gtensor 1.0.0

Reverse-mode autodifferentiation of computational graphs with tensors and more for machine learning.
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
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171


use anyhow::{Result, anyhow};

use super::shape::{Shape, ToShape};
use super::axis::ToAxis;
use super::Tensor;
use super::slice::TensorSlice;

impl Tensor {
    pub fn from_data(shape: impl ToShape, data: Vec<f32>) -> Self {
        Self::try_from_data(shape, data).unwrap()
    }

    pub fn try_from_data(shape: impl ToShape, data: Vec<f32>) -> Result<Self> {
        let shape = shape.to_shape();

        if shape.len() != data.len() {
            return Err(anyhow!("
                Provided data and shape to do not have the same length!
                Data Len: ({}), Shape: ({}, len: {}).
            ", data.len(), shape, shape.len()))
        } else {
            Ok(
                Self { data , shape}
            )
        }
    }

    pub fn from_iter(shape: impl ToShape, iter: impl Iterator<Item=f32>) -> Self {
        Self::try_from_data(shape, iter.collect::<Vec<f32>>()).unwrap()
    }

    pub fn try_from_iter(shape: impl ToShape, iter: impl Iterator<Item=f32>) -> Result<Self> {
        Self::try_from_data(shape.to_shape(), iter.collect::<Vec<f32>>())
    }

    pub fn from_slice(shape: impl ToShape, slice: &[f32]) -> Self {
        Self::try_from_slice(shape, slice).unwrap()
    }

    pub fn try_from_slice(shape: impl ToShape, slice: &[f32]) -> Result<Self> {
        let shape = shape.to_shape();

        if shape.len() != slice.len() {
            return Err(anyhow!("
                Provided slice and shape to do not have the same length!
                slice Len: ({}), Shape: ({}, len: {}).
            ", slice.len(), shape, shape.len()))
        } else {
            Ok(
                Self { data: slice.to_vec() , shape}
            )
        }
    }

    pub fn from_fill(shape: impl ToShape, value: f32) -> Self {
        let shape = shape.to_shape();

        Self {
            data: vec![value; shape.len()], shape
        }
    }

    pub fn len(&self) -> usize {
        self.data.len()
    }

    pub fn shape(&self) -> &Shape {
        &self.shape
    }

    pub fn try_reshape(&mut self, new: impl ToShape) -> Result<()> {
        self.shape.try_reshape(new)
    }

    pub fn reshape(&mut self, new: impl ToShape) {
        self.shape.try_reshape(new).unwrap()
    }

    pub fn axis(&self, axis: impl ToAxis) -> usize {
        self.shape[axis.to_axis()]
    }

    pub fn to_vec(self) -> Vec<f32> {
        self.data
    }

    pub fn as_vec(&self) -> Vec<f32> {
        self.data.clone()
    }

    pub fn slice_inner_mut(&mut self) -> &mut [f32] {
        &mut self.data
    }

    pub fn slice(&self) -> TensorSlice {
        TensorSlice {
            data: &self.data,
            shape: self.shape,
        }
    }
}

use std::ops::Index;
use std::ops::IndexMut;

impl Index<usize> for Tensor {
    type Output = f32;

    fn index(&self, index: usize) -> &Self::Output {
        &self.data[index]
    }
}

impl IndexMut<usize> for Tensor {
    fn index_mut(&mut self, index: usize) -> &mut Self::Output {
        &mut self.data[index]
    }
}

impl Index<(usize, usize)> for Tensor {
    type Output = f32;

    fn index(&self, index: (usize, usize)) -> &Self::Output {
        let (n, c) = index;
        &self.data[n*self.shape[1]*c]
    }
}

impl IndexMut<(usize, usize)> for Tensor {
    fn index_mut(&mut self, index: (usize, usize)) -> &mut Self::Output {
        let (n, c) = index;
        &mut self.data[n*self.shape[1]*c]
    }
}

impl Index<(usize, usize, usize)> for Tensor {
    type Output = f32;

    fn index(&self, index: (usize, usize, usize)) -> &Self::Output {
        let (n,c,h) = index;
        let shape = self.shape;
        &self.data[n*shape[1]*shape[2]+c*shape[2]+h]
    }
}

impl IndexMut<(usize, usize, usize)> for Tensor {
    fn index_mut(&mut self, index: (usize, usize, usize)) -> &mut Self::Output {
        let (n,c,h) = index;
        let shape = self.shape;
        &mut self.data[n*shape[1]*shape[2]+c*shape[2]*h]
    }
}

impl Index<(usize, usize, usize, usize)> for Tensor {
    type Output = f32;

    fn index(&self, index: (usize, usize, usize, usize)) -> &Self::Output {
        let (n,c,h,w) = index;
        let shape = self.shape;
        &self.data[n*shape[1]*shape[2]*shape[3]+c*shape[2]*shape[3]+h*shape[3]+w]
    }
}

impl IndexMut<(usize, usize, usize, usize)> for Tensor {
    fn index_mut(&mut self, index: (usize, usize, usize, usize)) -> &mut Self::Output {
        let (n,c,h,w) = index;
        let shape = self.shape;
        &mut self.data[n*shape[1]*shape[2]*shape[3]+c*shape[2]*shape[3]+h*shape[3]+w]
    }
}