Skip to content

fix rocm build error #1449

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

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

fix rocm build error #1449

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 1 addition & 1 deletion csrc/custom_marlin/gptq_marlin/gptq_marlin.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,7 @@ template <typename T> inline std::string str(T x) { return std::to_string(x); }

namespace gptq_marlin {

#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 800
#if (defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 800) || defined(__HIP_PLATFORM_AMD__)

__global__ void permute_cols_kernel(int4 const* __restrict__ a_int4_ptr,
int const* __restrict__ perm_int_ptr,
Expand Down
2 changes: 1 addition & 1 deletion csrc/custom_marlin/gptq_marlin/gptq_marlin.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,7 +38,7 @@ using I4 = Vec<int, 4>;

constexpr int div_ceil(int a, int b) { return (a + b - 1) / b; }

#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 800
#if (defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 800) || defined(__HIP_PLATFORM_AMD__)
// No support for async
#else

Expand Down
5 changes: 5 additions & 0 deletions csrc/custom_marlin/gptq_marlin/gptq_marlin_dtypes.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,11 @@
#include <cuda_bf16.h>
#include <cuda_fp16.h>

#ifdef __HIP_PLATFORM_AMD__
typedef __hip_bfloat16 nv_bfloat16;
typedef __hip_bfloat162 nv_bfloat162;
#endif

namespace gptq_marlin {

template <typename scalar_t> class ScalarType {};
Expand Down
2 changes: 1 addition & 1 deletion csrc/custom_marlin/gptq_marlin/gptq_marlin_repack.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,7 +9,7 @@ static constexpr int repack_threads = 256;
static constexpr int tile_k_size = tile_size;
static constexpr int tile_n_size = tile_k_size * 4;

#if defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 800
#if (defined(__CUDA_ARCH__) && __CUDA_ARCH__ < 800) || defined(__HIP_PLATFORM_AMD__)

template <int const num_threads, int const num_bits, bool const has_perm>
__global__ void marlin_repack_kernel(
Expand Down
2 changes: 1 addition & 1 deletion ktransformers/operators/linear.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,6 @@
from torch import Tensor, nn
if not torch.xpu.is_available():
import KTransformersOps
import vLLMMarlin
from ktransformers.util.custom_loader import GGUFLoader, SafeTensorLoader
from ktransformers.util.utils import InferenceState
if not torch.xpu.is_available():
Expand Down Expand Up @@ -520,6 +519,7 @@ def forward(self, x: torch.Tensor, bsz_tensor: torch.Tensor = None) -> torch.Ten
# padding x.shape[0] to avoid CUDA illegal memory access error
x, orig_size_m = self._pad_input(x)

import vLLMMarlin
x = vLLMMarlin.gptq_marlin_gemm(
x,
self.marlin_q_w,
Expand Down