1 unstable release

new 0.3.0 May 9, 2025

#735 in Machine learning


Used in 5 crates (2 directly)

Apache-2.0

320KB
8K SLoC

RSTSR rust native implementation to simple operators

This crate includes some native implementation (such as tensor addition, reduction, layout-change operations).

Only CPU (both serial and rayon parallel) are of concern.

It is splitted from rstsr-core, so to make this crate rstsr-native-impl more emphasis on computation, and this rstsr-core more emphasis on device and tensor structure/trait definition.

This crate can also be used if you wish to use tensor computation utilities, without actually defining a tensor object. In another word, this crate is device-free. Only serial and rayon parallel is concerned in this crate.

THis crate uses utilities in rstsr-common, such as layout, broadcasting, iterator.

Dependencies

~4MB
~106K SLoC