OSError while running the sample code.

[VASHISHT-RAHUL]

Running the following command with GPU :

kernel = falkon.kernels.GaussianKernel(sigma=1, opt=options)
flk = falkon.Falkon(kernel=kernel, penalty=1e-5, M=5000, options=options)

gives OSError : /opt/conda/lib/python3.10/site-packages/falkon/c_ext/_C.so: undefined symbol: _ZNK5torch8autograd4Node4nameEv.
Please provide the solution at the earliest.

[Giodiro]

Hi Rahul,This is an installation error. How did you install falkon?Could you try to clone the repository and run `python setup.py develop` from inside it directly?Il 6 lug 2023 13:36, Rahul Vashisht ***@***.***> ha scritto: Running the following command with GPU : kernel = falkon.kernels.GaussianKernel(sigma=1, opt=options) flk = falkon.Falkon(kernel=kernel, penalty=1e-5, M=5000, options=options) gives OSError : /opt/conda/lib/python3.10/site-packages/falkon/c_ext/_C.so: undefined symbol: _ZNK5torch8autograd4Node4nameEv. Please provide the solution at the earliest. —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[elishaparhi]

Hi,
I also got the same error and it got resolved now. But now it is showing the below error:

OSError: libtorch_cuda_linalg.so: cannot open shared object file: No such file or directory

Please help me resolve this.

Thanks!

[Giodiro]

yeah that's something I haven't figured out just yet.If I didn't get the path of your installation wrong you should try to run`export LD_LIBRARY_PATH="/opt/conda/lib/python3.10/site-packages/torch/lib:$LD_LIBRARY_PATH"`And then run the sample code. No need to reinstall anything.Il 7 lug 2023 19:40, elishaparhi ***@***.***> ha scritto: Hi, I also got the same error and it got resolved now. But now it is showing the below error: OSError: libtorch_cuda_linalg.so: cannot open shared object file: No such file or directory Please help me resolve this. Thanks! —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you commented.Message ID: ***@***.***>

[elishaparhi]

Thanks..this solved the problem.

I ran the below commands:

import os
os.environ['LD_LIBRARY_PATH'] ='/opt/conda/lib/python3.10/site-packages/torch/lib:$LD_LIBRARY_PATH'
os.environ['LD_LIBRARY_PATH'] +=':/usr/lib64-nvidia'
os.environ['LD_LIBRARY_PATH'] +=':/usr/local/cuda-11.8/compat'

Now, there is another error:
File "/usr/local/lib/python3.10/dist-packages/falkon/utils/devices.py", line 92, in _get_gpu_device_info
mem_free, mem_total = mem_get_info(g)
File "/usr/local/lib/python3.10/dist-packages/falkon/c_ext/init.py", line 17, in call_cuda
return getattr(torch.ops.falkon, name)(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_ops.py", line 569, in getattr
raise AttributeError(
AttributeError: '_OpNamespace' 'falkon' object has no attribute 'mem_get_info'

[elishaparhi]

Now, getting the below error:

File "/usr/local/lib/python3.10/dist-packages/falkon/utils/devices.py", line 92, in _get_gpu_device_info
mem_free, mem_total = mem_get_info(g)
File "/usr/local/lib/python3.10/dist-packages/falkon/c_ext/init.py", line 17, in call_cuda
return getattr(torch.ops.falkon, name)(*args, **kwargs)
File "/usr/local/lib/python3.10/dist-packages/torch/_ops.py", line 569, in getattr
raise AttributeError(
AttributeError: '_OpNamespace' 'falkon' object has no attribute 'mem_get_info'

Please help me resolve this.

[Giodiro]

Hi @elishaparhi

Really sorry if this comes too late, but installation should be a lot easier now.. let me know if you try again and have further problems!

Giacomo