singe-kernel 0.1.0-alpha.4

Reusable CPU and GPU kernels.
Documentation

singe-kernel

There is very little structured metadata to build this page from currently. You should check the main library docs, readme, or Cargo.toml in case the author documented the features in them.

This version has 40 feature flags, 1 of them enabled by default.

default

dtype-f32 (default)

This feature flag does not enable additional features.

cuda_13_3

cutile

dtype-bf16

This feature flag does not enable additional features.

dtype-f16

This feature flag does not enable additional features.

dtype-f64

This feature flag does not enable additional features.

dtype-f8

This feature flag does not enable additional features.

dtype-i32

This feature flag does not enable additional features.

dtype-i64

This feature flag does not enable additional features.

dtype-i8

This feature flag does not enable additional features.

dtype-u32

This feature flag does not enable additional features.

dtype-u64

This feature flag does not enable additional features.

dtype-u8

This feature flag does not enable additional features.

dtypes-all

full

kernel-activation

kernel-attention

kernel-audio

kernel-cast

kernel-conv

kernel-creation

kernel-elementwise

kernel-embedding

kernel-fft

kernel-fused

kernel-gather

kernel-mask

kernel-matmul

kernel-moe

kernel-normalization

kernel-pooling

kernel-positional

kernel-quantization

kernel-recurrent

kernel-reduction

kernel-scalar

kernel-shape

kernel-softmax

kernel-unary

kernels-all