Removed num_traits::Num requirement from Zeros.

Had to figure out a way to store zeros in place
coreylowman · Dec 2, 2023 · 2668e78 · 2668e78
1 parent 1faa030
commit 2668e78
Show file tree

Hide file tree

Showing 2 changed files with 6 additions and 14 deletions.
diff --git a/dfdx-core/src/tensor/webgpu/allocate.rs b/dfdx-core/src/tensor/webgpu/allocate.rs
@@ -52,28 +52,20 @@ impl Webgpu {
     }
 }
 
-impl<E: Unit + SafeZeros + From<bool>> ZerosTensor<E> for Webgpu {
+impl<E: Unit + SafeZeros> ZerosTensor<E> for Webgpu {
     fn try_zeros_like<S: HasShape>(&self, src: &S) -> Result<Tensor<S::Shape, E, Self>, Error> {
         let shape = *src.shape();
         let strides = shape.strides();
         let data = unsafe { self.alloc_empty::<E>(shape.num_elements()) }?;
-        data.copy_to_device(
-            &self.dev,
-            &self.queue,
-            &vec![E::from(false); shape.num_elements()],
-        );
+        data.copy_to_device(&self.dev, &self.queue, &vec![0u8; data.size()]);
 
         Ok(self.build_tensor(shape, strides, data))
     }
 }
 
-impl<E: Unit + SafeZeros + From<bool>> ZeroFillStorage<E> for Webgpu {
+impl<E: Unit + SafeZeros> ZeroFillStorage<E> for Webgpu {
     fn try_fill_with_zeros(&self, storage: &mut Self::Vec) -> Result<(), Error> {
-        storage.copy_to_device(
-            &self.dev,
-            &self.queue,
-            &vec![E::from(false); storage.size() as usize / std::mem::size_of::<E>()],
-        );
+        storage.copy_to_device(&self.dev, &self.queue, &vec![0u8; storage.size()]);
 
         Ok(())
     }

diff --git a/dfdx-core/src/tensor_ops/utilities/device.rs b/dfdx-core/src/tensor_ops/utilities/device.rs
@@ -132,9 +132,9 @@ impl Device<f32> for crate::tensor::Cuda {}
 impl Device<f64> for crate::tensor::Cuda {}
 
 #[cfg(all(feature = "webgpu", feature = "f16"))]
-impl Device<f16> for crate::tensor::Cuda {}
+impl Device<f16> for crate::tensor::Webgpu {}
 #[cfg(all(feature = "webgpu", feature = "f16"))]
-impl Device<AMP<f16>> for crate::tensor::Cuda {}
+impl Device<AMP<f16>> for crate::tensor::Webgpu {}
 #[cfg(feature = "webgpu")]
 impl Device<f32> for crate::tensor::Webgpu {}
 #[cfg(feature = "webgpu")]