ai-dataloader 0.6.2

Rust implementation to the PyTorch DataLoader
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

use super::super::Collate;
use super::DefaultCollate;
use ndarray::{Array, Ix1};
use std::num::{
    NonZeroI128, NonZeroI16, NonZeroI32, NonZeroI64, NonZeroI8, NonZeroIsize, NonZeroU128,
    NonZeroU16, NonZeroU32, NonZeroU64, NonZeroU8, NonZeroUsize,
};

/// Maybe this one may not be supported by a tensor running on a GPU.

macro_rules! nonzero_impl {
    ($($t:ty)*) => {
        $(
            impl Collate<$t> for DefaultCollate {
                type Output = Array<$t, Ix1>;
                fn collate(&self, batch: Vec<$t>) -> Self::Output {
                    Array::from_vec(batch)
                }
            }
        )*
    };
}
nonzero_impl!(
    NonZeroUsize NonZeroU8 NonZeroU16 NonZeroU32 NonZeroU64 NonZeroU128
    NonZeroIsize NonZeroI8 NonZeroI16 NonZeroI32 NonZeroI64 NonZeroI128
);