search

gpgpu-sim / pytorch-gpgpu-sim

Modified version of PyTorch able to work with changes to GPGPU-Sim
Other
40 stars 23 forks source link

Errors when running pytorch mnist example #8

Closed sxzhang1993 closed 3 years ago

sxzhang1993 commented 3 years ago

Hello, While I running the pytorch mnist example mentioned in the instruction, I will get the following error:

CuDNN error: CUDNN_STATUS_NOT_INITIALIZED

after I added torch.backends.cudnn.deterministic = True torch.backends.cudnn.benchmark = False to the python file, I will get the following new error:

GPGPU-Sim PTX: Loading PTXInfo from libcaffe2_gpu.89.sm_61.ptx GPGPU-Sim PTX: ERROR ** while loading PTX (b) -1 Ensure ptxas is in your path.

Using CUDA 8.0 & cudnn 7.1.4, and following in .bashrc file: export PATH=/usr/local/cuda-8.0/bin:$PATH export LD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64:$LD_LIBRARY_PATH export CUDA_INSTALL_PATH=/usr/local/cuda-8.0 export CUDNN_INCLUDE_DIR=/usr/local/cuda-8.0/include/cudnn.h export CUDNN_LIBRARY=/usr/local/cuda-8.0/lib64/libcudnn.so.7 export TORCH_CUDA_ARCH_LIST="6.1+PTX" export CUDNN_PATH=/usr/local/cuda-8.0/include/cudnn.h export PYTORCH_BIN=/home/szhang/.local/lib/python3.6/site-packages/torch/lib/libcaffe2_gpu.so

Also tried export PYTORCH_BIN=/home/szhang/pytorch-gpgpu-sim/torch/lib/libcaffe2_gpu.so and will have the same error.

Any suggestions and ideas will be well appreciated, thanks a lot!