huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-18 19:47:12 +00:00
Code Issues Packages Projects Releases Wiki Activity

AVX version for the q8-0 multiplications. (#598)

Browse Source
This commit is contained in:
Laurent Mazare
2023-08-25 10:14:49 +01:00
committed by GitHub
parent d728e646c2
commit afc10a3232
2 changed files with 23 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
candle-core/src/quantized/k_quants.rs
View File

@ -421,7 +421,11 @@ impl GgmlType for BlockQ8_0 {
Ok(())
}
#[allow(unreachable_code)]
fn vec_dot(n: usize, xs: &[Self], ys: &[Self::VecDotType]) -> Result<f32> {
#[cfg(target_feature = "avx")]
return super::avx::vec_dot_q8_0_q8_0(n, xs, ys);
let qk = QK8_0;
if n % QK8_0 != 0 {
crate::bail!("vec_dot_q8_0_q8_0: {n} is not divisible by {qk}")