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

Fix sigmoid gradient calculation and move sigmoid into a specialized op (#2114)

Browse Source 
* add sigmoid op

* small fix

* add as a method on `Tensor`

* implement gradient calculation for sigmoid

* add sigmoid tests

* we should have a specialized op for this

* fix clippy

* fix clippy 2

* Revert all previous commits in favor of a `CustomOp` based solution

* use `CustomOp1` implementation

* fix rustfmt

* experimental add metal impl

* add cuda kernel impl

* fix fmt

* Add a test + reduce some cuda duplication.

---------

Co-authored-by: laurent <laurent.mazare@gmail.com>
This commit is contained in:
MilkFather
2024-04-29 17:04:43 +08:00
committed by GitHub
parent ed7b99f525
commit 3bbb88fcb4
6 changed files with 214 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
candle-core/src/cuda_backend/mod.rs
View File

@ -18,7 +18,7 @@ pub use device::{CudaDevice, DeviceId};
pub use error::{CudaError, WrapErr};
pub use utils::{Map1, Map1Any, Map2, Map2Any, Map2InPlace, S};
enum SlicePtrOrNull<T> {
pub enum SlicePtrOrNull<T> {
Ptr(CudaSlice<T>),
Null,
}
@ -33,7 +33,7 @@ unsafe impl<T: DeviceRepr> DeviceRepr for &SlicePtrOrNull<T> {
}
impl SlicePtrOrNull<usize> {
fn params_from_layout(dev: &CudaDevice, l: &Layout) -> Result<Self> {
pub fn params_from_layout(dev: &CudaDevice, l: &Layout) -> Result<Self> {
let ds = if l.is_contiguous() {
SlicePtrOrNull::Null
} else {