Skip to content

GitLab

Commit 772a941a authored by wenjh's avatar wenjh
Browse files

Use hipFreeAsync rather than hipFree 


Signed-off-by: wenjh's avatarwenjh <wenjh@sugon.com>
parent 623ae795
Hide whitespace changes
Inline Side-by-side
Showing with 5 additions and 3 deletions
setup.py
View file @ 772a941a
...@@ -3,8 +3,10 @@ ...@@ -3,8 +3,10 @@
# See LICENSE for license information. # See LICENSE for license information.
"""Installation script.""" """Installation script."""
# NVTE_FRAMEWORK=pytorch NVTE_USE_ROCM=1 NVTE_USE_HIPBLASLT=1 NVTE_USE_ROCBLAS=1 CMAKE_PREFIX_PATH=/opt/dtk/lib/cmake/amd_comgr/ MPI_HOME=/opt/mpi/ NVTE_UB_WITH_MPI=1 CXX=hipcc pip3 install . -v # 编译并安装命令
# NVTE_FRAMEWORK=pytorch NVTE_USE_ROCM=1 NVTE_USE_HIPBLASLT=1 NVTE_USE_ROCBLAS=1 CMAKE_PREFIX_PATH=/opt/dtk/lib/cmake/amd_comgr/ MPI_HOME=/opt/mpi/ NVTE_UB_WITH_MPI=1 CXX=hipcc PYTHONPATH=/home/TransformerEngine/3rdparty/hipify_torch:$PYTHONPATH python3 setup.py bdist_wheel # NVTE_BUILD_SUPPRESS_UNUSED_WARNING=1 NVTE_BUILD_SUPPRESS_RETURN_TYPE_WARNING=1 NVTE_BUILD_SUPPRESS_SIGN_COMPARE_WARNING=1 NVTE_FRAMEWORK=pytorch NVTE_USE_ROCM=1 NVTE_USE_HIPBLASLT=1 NVTE_USE_ROCBLAS=1 CMAKE_PREFIX_PATH=/opt/dtk/lib/cmake/amd_comgr/ MPI_HOME=/opt/mpi/ NVTE_UB_WITH_MPI=0 CXX=hipcc PYTHONPATH=/home/TransformerEngine/3rdparty/hipify_torch:$PYTHONPATH python3 setup.py install -v
# 打WHL包命令
# NVTE_BUILD_SUPPRESS_UNUSED_WARNING=1 NVTE_BUILD_SUPPRESS_RETURN_TYPE_WARNING=1 NVTE_BUILD_SUPPRESS_SIGN_COMPARE_WARNING=1 NVTE_FRAMEWORK=pytorch NVTE_USE_ROCM=1 NVTE_USE_HIPBLASLT=1 NVTE_USE_ROCBLAS=1 CMAKE_PREFIX_PATH=/opt/dtk/lib/cmake/amd_comgr/ MPI_HOME=/opt/mpi/ NVTE_UB_WITH_MPI=0 CXX=hipcc PYTHONPATH=/home/TransformerEngine/3rdparty/hipify_torch:$PYTHONPATH python3 setup.py bdist_wheel
from importlib import metadata from importlib import metadata
import os import os
......
transformer_engine/common/gemm/rocm_gemm.cu
View file @ 772a941a
...@@ -1357,7 +1357,7 @@ void hipblaslt_goupedgemm(std::vector<const Tensor*>& inputA, std::vector<const ...@@ -1357,7 +1357,7 @@ void hipblaslt_goupedgemm(std::vector<const Tensor*>& inputA, std::vector<const
// NVTE_CHECK_HIPBLASLT(groupedgemm.run(stream)); // NVTE_CHECK_HIPBLASLT(groupedgemm.run(stream));
NVTE_CHECK_CUDA(hipFreeAsync(d_userArgs, stream)); NVTE_CHECK_CUDA(hipFreeAsync(d_userArgs, stream));
NVTE_CHECK_CUDA(hipFree(userArgs)); NVTE_CHECK_CUDA(hipFreeAsync(userArgs, stream));
} }
#endif //USE_HIPBLASLT #endif //USE_HIPBLASLT
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment