Open MDD-0928 opened 4 months ago
Check your CUDA_HOME
and make sure it is not None. If that does not work, try python setup.py install
and show me the error logs.
Thanks for your update according to Issue 47!!! I was so excited to test your new code, but failed when run the command "cd selective_scan && pip install ."
Could you please help me fix this bug? THANKS
It is recommended to use cuda12.1 torch2.2, conda install pytorch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 pytorch-cuda=12.1 -c pytorch -c nvidia
I created a new conda env, then used "conda install pytorch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 pytorch-cuda=11.8 -c pytorch -c nvidia" to install pytorch, but still failed when "cd selective_scan && pip install .", and I can't get CUDA_HOME by "echo $CUDA_HOME", so I wonder that how to set the environmental variable "CUDA_HOME"?
THNAKS
try this one:
python -c "from torch.utils.cpp_extension import CUDA_HOME; print(CUDA_HOME)"
return "NONE" T^T...
That shows maybe something wrong with the environment.
Need I explicitly set the environmental invariable " CUDA_HOME " by vi ~/.bashrc or just follow the pytorch.org to install the pytorch ? Could you get some output by " echo $CUDA_HOME “ ?
By the way, running the code "import torch torch.cuda.is_available()" returns True.... and all the codes can run on the GPU just can't pip install the selective_scan.....
请参考(Please refer to)https://github.com/MzeroMiko/VMamba/issues/64
Thanks for your update according to Issue 47!!! I was so excited to test your new code, but failed when run the command "cd selective_scan && pip install ."
Could you please help me fix this bug?
THANKS