lightseq
lightseq copied to clipboard
bytedance
Running examples meet error, with cuda 11.6
/usr/local/cuda/bin/nvcc -DTORCH_EXTENSION_NAME=lightseq_layers -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE="gcc" -DPYBIND11_STDLIB="libstdcpp" -DPYBIND11_BUILD_ABI="cxxabi1013" -I/opt/conda/lib/python3.8/site-packages/lightseq/training/csrc/kernels/includes -I/opt/conda/lib/python3.8/site-packages/lightseq/training/csrc/ops/includes -I/opt/conda/lib/python3.8/site-packages/3rdparty/cub -isystem /opt/conda/lib/python3.8/site-packages/torch/include -isystem /opt/conda/lib/python3.8/site-packages/torch/include/torch/csrc/api/include -isystem /opt/conda/lib/python3.8/site-packages/torch/include/TH -isystem /opt/conda/lib/python3.8/site-packages/torch/include/THC -isystem /usr/local/cuda/include -isystem /opt/conda/include/python3.8 -D_GLIBCXX_USE_CXX11_ABI=1 -D__CUDA_NO_HALF_OPERATORS -D__CUDA_NO_HALF_CONVERSIONS_ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr -gencode=arch=compute_52,code=sm_52 -gencode=arch=compute_60,code=sm_60 -gencode=arch=compute_61,code=sm_61 -gencode=arch=compute_70,code=sm_70 -gencode=arch=compute_75,code=sm_75 -gencode=arch=compute_80,code=sm_80 -gencode=arch=compute_86,code=compute_86 -gencode=arch=compute_86,code=sm_86 --compiler-options '-fPIC' -O3 --use_fast_math -std=c++14 -U__CUDA_NO_HALF_OPERATORS__ -U__CUDA_NO_HALF_CONVERSIONS__ -U__CUDA_NO_HALF2_OPERATORS__ -DTHRUST_IGNORE_CUB_VERSION_CHECK -gencode=arch=compute_80,code=sm_80 -gencode=arch=compute_80,code=compute_80 -c /opt/conda/lib/python3.8/site-packages/lightseq/training/csrc/kernels/cuda_util.cu -o cuda_util.cuda.o /usr/local/cuda/include/cub/detail/device_synchronize.cuh(26): error: expected a ";"
/usr/local/cuda/include/cub/detail/device_synchronize.cuh(33): error: this pragma must immediately precede a declaration
/usr/local/cuda/include/cub/detail/device_synchronize.cuh(65): error: expected a declaration
/usr/local/cuda/include/thrust/system/cuda/detail/util.h(61): error: expected a declaration
/usr/local/cuda/include/thrust/system/cuda/detail/util.h(206): error: expected a ";"
/usr/local/cuda/include/thrust/system/cuda/detail/util.h(208): error: expected a declaration
/usr/local/cuda/include/thrust/system/cuda/detail/util.h(239): error: variable "cuda_cub" has already been defined
I recommend running with torch 1.8 and cuda compatible with it.
On Fri, Jul 1, 2022 at 2:48 PM Yuang Liu @.***> wrote:
just follow the demo
— Reply to this email directly, view it on GitHub https://github.com/bytedance/lightseq/issues/328#issuecomment-1171998535, or unsubscribe https://github.com/notifications/unsubscribe-auth/AELIZAIHGXFDTUDIZY6YJTLVR2IDDANCNFSM5ZWW57RQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>
It should be ok with latest torch versions, but we didn't test all versions, most of our experiments are based on 1.8.
On Wed, Jul 13, 2022 at 10:15 PM Yuang Liu @.***> wrote:
1.8? but torhc is 1.12 for now, should i must use the torch 1.8?
— Reply to this email directly, view it on GitHub https://github.com/bytedance/lightseq/issues/328#issuecomment-1183281817, or unsubscribe https://github.com/notifications/unsubscribe-auth/AELIZAKU3L33ODXKUASZAZDVT3FP7ANCNFSM5ZWW57RQ . You are receiving this because you commented.Message ID: @.***>
I want to know which version you use. I use nvcr.io/nvidia/pytorch:22.01-py3 this docker image and just pip install fairseq，this will install torch 1.12, which is incompatible with cuda 11.6 Does fairseq do not support cuda 11.6? Is lightseq based on fairseq?
