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

Get the cuda tests to pass.

Browse Source
This commit is contained in:
laurent
2023-06-28 15:53:23 +01:00
parent 3f0d9fbb25
commit 6c9e6b5a99
1 changed files with 29 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

49
candle-core/src/cuda_backend.rs
View File

@ -27,7 +27,7 @@ pub enum CudaError {
InternalError(&'static str), InternalError(&'static str),
#[error("internal error '{0}'")] #[error("internal error '{0}'")]
WrappedError(Box<dyn std::error::Error>), WrappedError(Box<dyn std::error::Error + 'static + std::marker::Send + std::marker::Sync>),
#[error("matmul is only supported for contiguous tensors lstride: {lhs_stride:?} rstride: {rhs_stride:?} mnk: {mnk:?}")] #[error("matmul is only supported for contiguous tensors lstride: {lhs_stride:?} rstride: {rhs_stride:?} mnk: {mnk:?}")]
MatMulNonContiguous { MatMulNonContiguous {
@ -245,13 +245,14 @@ enum CudaStorageSlice {
fn slice_src_and_dst<'a, T>( fn slice_src_and_dst<'a, T>(
src: &'a CudaSlice<T>, src: &'a CudaSlice<T>,
src_offset: usize, src_l: &Layout,
dst: &'a mut CudaSlice<T>, dst: &'a mut CudaSlice<T>,
dst_offset: usize, dst_offset: usize,
) -> ( ) -> (
cudarc::driver::CudaView<'a, T>, cudarc::driver::CudaView<'a, T>,
cudarc::driver::CudaViewMut<'a, T>, cudarc::driver::CudaViewMut<'a, T>,
) { ) {
let src_offset = src_l.start_offset();
let to_copy = dst let to_copy = dst
.len() .len()
.saturating_sub(dst_offset) .saturating_sub(dst_offset)
@ -366,13 +367,18 @@ impl CudaStorage {
let dev = self.device(); let dev = self.device();
let ds = dev.htod_copy([dims, layout.stride()].concat())?; let ds = dev.htod_copy([dims, layout.stride()].concat())?;
let start_o = layout.start_offset(); let start_o = layout.start_offset();
// This returns an i64 rather than a &i64, this is useful to get around some temporary
// lifetime issue and is safe as long as self.slice does not go out of scope before inp
// is used.
let inp = match &self.slice { let inp = match &self.slice {
CudaStorageSlice::U32(inp) => inp.slice(start_o..).device_ptr(), CudaStorageSlice::U32(inp) => *inp.slice(start_o..).device_ptr(),
CudaStorageSlice::BF16(inp) => inp.slice(start_o..).device_ptr(), CudaStorageSlice::BF16(inp) => *inp.slice(start_o..).device_ptr(),
CudaStorageSlice::F16(inp) => inp.slice(start_o..).device_ptr(), CudaStorageSlice::F16(inp) => *inp.slice(start_o..).device_ptr(),
CudaStorageSlice::F32(inp) => inp.slice(start_o..).device_ptr(), CudaStorageSlice::F32(inp) => *inp.slice(start_o..).device_ptr(),
CudaStorageSlice::F64(inp) => inp.slice(start_o..).device_ptr(), CudaStorageSlice::F64(inp) => *inp.slice(start_o..).device_ptr(),
}; };
let inp = &inp;
let kernel_name = format!("cast_{}_{}", self.dtype().as_str(), dtype.as_str()); let kernel_name = format!("cast_{}_{}", self.dtype().as_str(), dtype.as_str());
let func = dev.get_or_load_func(&kernel_name, kernels::CAST)?; let func = dev.get_or_load_func(&kernel_name, kernels::CAST)?;
let slice = match dtype { let slice = match dtype {
@ -739,13 +745,14 @@ impl CudaStorage {
layout_f: &Layout, layout_f: &Layout,
) -> Result<Self> { ) -> Result<Self> {
let ids = match &self.slice { let ids = match &self.slice {
CudaStorageSlice::U32(slice) => &slice.slice(layout.start_offset()..), CudaStorageSlice::U32(slice) => slice.slice(layout.start_offset()..),
_ => Err(CudaError::UnexpectedDType { _ => Err(CudaError::UnexpectedDType {
msg: "where conditions should be u32", msg: "where conditions should be u32",
expected: DType::U32, expected: DType::U32,
got: self.dtype(), got: self.dtype(),
})?, })?,
}; };
let ids = &ids;
let shape = layout.shape(); let shape = layout.shape();
let dims = shape.dims(); let dims = shape.dims();
let el = shape.elem_count(); let el = shape.elem_count();
@ -818,13 +825,14 @@ impl CudaStorage {
pub(crate) fn embedding(&self, layout: &Layout, rhs: &Self, rhs_l: &Layout) -> Result<Self> { pub(crate) fn embedding(&self, layout: &Layout, rhs: &Self, rhs_l: &Layout) -> Result<Self> {
let ids = match &self.slice { let ids = match &self.slice {
CudaStorageSlice::U32(slice) => &slice.slice(layout.start_offset()..), CudaStorageSlice::U32(slice) => slice.slice(layout.start_offset()..),
_ => Err(CudaError::UnexpectedDType { _ => Err(CudaError::UnexpectedDType {
msg: "embedding ids should be u32", msg: "embedding ids should be u32",
expected: DType::U32, expected: DType::U32,
got: self.dtype(), got: self.dtype(),
})?, })?,
}; };
let ids = &ids;
let shape = layout.shape(); let shape = layout.shape();
let (v_size, h_size) = rhs_l let (v_size, h_size) = rhs_l
.shape() .shape()
@ -953,15 +961,16 @@ impl CudaStorage {
dst_offset: usize, dst_offset: usize,
src_l: &Layout, src_l: &Layout,
) -> Result<()> { ) -> Result<()> {
let src_shape = src_l.shape();
let dims = src_shape.dims(); let dims = src_shape.dims();
let el_count = src_shape.elem_count(); let el_count = src_shape.elem_count();
let cfg = LaunchConfig::for_num_elems(el_count as u32); let cfg = LaunchConfig::for_num_elems(el_count as u32);
let dev = &self.device; let dev = &self.device;
let ds = dev.htod_copy([dims, src_stride].concat())?; let ds = dev.htod_copy([dims, src_l.stride()].concat())?;
match (&self.slice, &mut dst.slice) { match (&self.slice, &mut dst.slice) {
(CudaStorageSlice::BF16(src), CudaStorageSlice::BF16(dst)) => { (CudaStorageSlice::BF16(src), CudaStorageSlice::BF16(dst)) => {
let (src, mut dst) = slice_src_and_dst(src, src_offset, dst, dst_offset); let (src, mut dst) = slice_src_and_dst(src, src_l, dst, dst_offset);
if src_shape.is_contiguous(src_stride) { if src_l.is_contiguous() {
dev.dtod_copy(&src, &mut dst)? dev.dtod_copy(&src, &mut dst)?
} else { } else {
let func = dev.get_or_load_func("ucopy_bf16", kernels::UNARY)?; let func = dev.get_or_load_func("ucopy_bf16", kernels::UNARY)?;
@ -972,8 +981,8 @@ impl CudaStorage {
} }
} }
(CudaStorageSlice::F16(src), CudaStorageSlice::F16(dst)) => { (CudaStorageSlice::F16(src), CudaStorageSlice::F16(dst)) => {
let (src, mut dst) = slice_src_and_dst(src, src_offset, dst, dst_offset); let (src, mut dst) = slice_src_and_dst(src, src_l, dst, dst_offset);
if src_shape.is_contiguous(src_stride) { if src_l.is_contiguous() {
dev.dtod_copy(&src, &mut dst)? dev.dtod_copy(&src, &mut dst)?
} else { } else {
let func = dev.get_or_load_func("ucopy_f16", kernels::UNARY)?; let func = dev.get_or_load_func("ucopy_f16", kernels::UNARY)?;
@ -984,8 +993,8 @@ impl CudaStorage {
} }
} }
(CudaStorageSlice::F32(src), CudaStorageSlice::F32(dst)) => { (CudaStorageSlice::F32(src), CudaStorageSlice::F32(dst)) => {
let (src, mut dst) = slice_src_and_dst(src, src_offset, dst, dst_offset); let (src, mut dst) = slice_src_and_dst(src, src_l, dst, dst_offset);
if src_shape.is_contiguous(src_stride) { if src_l.is_contiguous() {
dev.dtod_copy(&src, &mut dst)? dev.dtod_copy(&src, &mut dst)?
} else { } else {
let func = dev.get_or_load_func("ucopy_f32", kernels::UNARY)?; let func = dev.get_or_load_func("ucopy_f32", kernels::UNARY)?;
@ -996,8 +1005,8 @@ impl CudaStorage {
} }
} }
(CudaStorageSlice::U32(src), CudaStorageSlice::U32(dst)) => { (CudaStorageSlice::U32(src), CudaStorageSlice::U32(dst)) => {
let (src, mut dst) = slice_src_and_dst(src, src_offset, dst, dst_offset); let (src, mut dst) = slice_src_and_dst(src, src_l, dst, dst_offset);
if src_shape.is_contiguous(src_stride) { if src_l.is_contiguous() {
dev.dtod_copy(&src, &mut dst)? dev.dtod_copy(&src, &mut dst)?
} else { } else {
let func = dev.get_or_load_func("ucopy_u32", kernels::UNARY)?; let func = dev.get_or_load_func("ucopy_u32", kernels::UNARY)?;
@ -1008,8 +1017,8 @@ impl CudaStorage {
} }
} }
(CudaStorageSlice::F64(src), CudaStorageSlice::F64(dst)) => { (CudaStorageSlice::F64(src), CudaStorageSlice::F64(dst)) => {
let (src, mut dst) = slice_src_and_dst(src, src_offset, dst, dst_offset); let (src, mut dst) = slice_src_and_dst(src, src_l, dst, dst_offset);
if src_shape.is_contiguous(src_stride) { if src_l.is_contiguous() {
dev.dtod_copy(&src, &mut dst)? dev.dtod_copy(&src, &mut dst)?
} else { } else {
let func = dev.get_or_load_func("ucopy_64", kernels::UNARY)?; let func = dev.get_or_load_func("ucopy_64", kernels::UNARY)?;