huggingface/candle
1
0
Fork 0
mirror of https://github.com/huggingface/candle.git synced 2025-06-19 19:58:35 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add a cuda kernel for dequantizing q8_0. (#1804)

Browse Source
This commit is contained in:
Laurent Mazare
2024-03-05 09:50:37 +01:00
committed by GitHub
parent 8cc0a183ba
commit bd9ab9bc04
2 changed files with 24 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
candle-core/tests/quantized_tests.rs
View File

@ -738,10 +738,6 @@ macro_rules! quantized_matmul {
// stable. https://github.com/rust-lang/rust/issues/29599
($fn_name: ident, $fn_name_cpu: ident, $fn_name_cuda: ident, $fn_name_metal: ident, $dtype: expr) => {
fn $fn_name(device: &Device) -> Result<()> {
if device.is_cuda() {
// TODO Enable Cuda GGML sometime maybe.
return Ok(());
}
test_matmul(device, (1, 3, 4, 256), $dtype)?;
Ok(())
}