Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add split_tensor_along method #917

Draft
wants to merge 8 commits into
base: main
Choose a base branch
from
Draft

Add split_tensor_along method #917

Changes from 1 commit
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
5c532ec
remove deprecated ftz intrinsics
rainiwu Jan 26, 2024
fb91f13
suppress spurious cargo clippy warning
rainiwu Jan 26, 2024
24a8593
Merge pull request #1 from rainiwu/remove-ftz
swfsql Feb 7, 2024
4e3f7c7
avoid conv1d bound for cudnn
swfsql Feb 6, 2024
a8bc54c
bump gemm
swfsql Feb 9, 2024
557687c
clippy fix
swfsql Feb 9, 2024
1175903
Merge pull request #2 from swfsql/avoid-ci-errors
swfsql Mar 1, 2024
aa44eb5
Add split_tensor_along method
swfsql Feb 1, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
avoid conv1d bound for cudnn
  • Loading branch information
@swfsql
swfsql committed Feb 9, 2024
commit 4e3f7c7a24728668f72cf3617a66f4476280f6fb
50 changes: 39 additions & 11 deletions dfdx-core/src/tensor_ops/utilities/device.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -114,33 +114,61 @@ pub trait Device<E: Dtype>:
+ crate::tensor_ops::axpy::AxpyKernel<E>

// conv1d
+ super::super::conv1d::Conv1DKernel<E>
+ NonCudnnCuda<E>
{
}

#[cfg(feature = "cudnn")]
pub trait NonCudnnCuda<E: Dtype> {}

#[cfg(not(feature = "cudnn"))]
pub trait NonCudnnCuda<E: Dtype>:
// conv1d
super::super::conv1d::Conv1DKernel<E>
{
}

#[cfg(feature = "f16")]
impl Device<f16> for crate::tensor::Cpu {}
#[cfg(feature = "f16")]
impl Device<AMP<f16>> for crate::tensor::Cpu {}
mod f16_ {
use super::*;
impl Device<f16> for crate::tensor::Cpu {}
impl NonCudnnCuda<f16> for crate::tensor::Cpu {}
impl Device<AMP<f16>> for crate::tensor::Cpu {}
impl NonCudnnCuda<AMP<f16>> for crate::tensor::Cpu {}
}
impl Device<f32> for crate::tensor::Cpu {}
impl NonCudnnCuda<f32> for crate::tensor::Cpu {}
impl Device<f64> for crate::tensor::Cpu {}
impl NonCudnnCuda<f64> for crate::tensor::Cpu {}

#[cfg(all(feature = "cuda", feature = "f16"))]
impl Device<f16> for crate::tensor::Cuda {}
#[cfg(all(feature = "cuda", feature = "f16"))]
impl Device<AMP<f16>> for crate::tensor::Cuda {}
#[cfg(feature = "cuda")]
impl Device<f32> for crate::tensor::Cuda {}
mod cuda_f16 {
use super::*;
impl Device<f16> for crate::tensor::Cuda {}
impl NonCudnnCuda<f16> for crate::tensor::Cuda {}
impl Device<AMP<f16>> for crate::tensor::Cuda {}
impl NonCudnnCuda<AMP<f16>> for crate::tensor::Cuda {}
}
#[cfg(feature = "cuda")]
impl Device<f64> for crate::tensor::Cuda {}
mod cuda {
use super::*;
impl Device<f32> for crate::tensor::Cuda {}
impl NonCudnnCuda<f32> for crate::tensor::Cuda {}
impl Device<f64> for crate::tensor::Cuda {}
impl NonCudnnCuda<f64> for crate::tensor::Cuda {}
}

// TODO: How can we implement this for f16 when WGSL doesn't support f16 yet?
// #[cfg(all(feature = "webgpu", feature = "f16"))]
// impl Device<f16> for crate::tensor::Webgpu {}
// #[cfg(all(feature = "webgpu", feature = "f16"))]
// impl Device<AMP<f16>> for crate::tensor::Webgpu {}
#[cfg(feature = "webgpu")]
impl Device<f32> for crate::tensor::Webgpu {}
mod webgpu {
use super::*;
impl Device<f32> for crate::tensor::Webgpu {}
impl NonCudnnCuda<f32> for crate::tensor::Webgpu {}
}

// TODO: How can we implement this for f64 when WGSL doesn't support f64 yet?
// #[cfg(feature = "webgpu")]
Expand Down