[NLPL Task Force (A)] [uninett.no #196965] Tensorflow issues, pt. 2

[Vinit Ravishankar via RT]

Hmm, I did have the same error when I tried to do this the non-conda way, unless I’m misremembering.

I’m curious about why 1.14.0 appears to import (and even run) just fine with pip/conda, but not 1.13.1.

@Stephan, I’m afraid the TF 1.13.1 module has the same issue (libcuda).

– Vinit

> On 25 Oct 2019, at 13:41, Henrik R. Nagel via RT <support at metacenter.no> wrote:
> 
> Hi,
> 
> Installing TensorFlow goes without any problems for me:
> 
> $ salloc --nodes=1 --ntasks-per-node=1 --gres=gpu:1 --mem-per-cpu=10G --time=01:00:00 --partition=accel --account=<your account>
> salloc: Nodes c7-8 are ready for job
> $ ssh c7-8
> <hrn at c7-8><~> module load foss/2019b
> <hrn at c7-8><~> module load Python/3.7.4-GCCcore-8.3.0
> <hrn at c7-8><~> module load CUDA/10.0.130 cuDNN/7.4.2.24-CUDA-10.0.130
> <hrn at c7-8><~> module --ignore-cache load NCCL/2.4.8-CUDA-10.0
> <hrn at c7-8><~> python3 -m pip install tensorflow-gpu==1.13.1 --user
> ...
> Successfully installed ...
> 
> But testing it did not go well:
> 
> <hrn at c7-8><~> python3
> .>>> import tensorflow
> ...
> ImportError: /lib64/libm.so.6: version `GLIBC_2.23' not found (required by /cluster/home/hrn/.local/lib/python3.7/site-packages/tensorflow/python/_pywrap_tensorflow_internal.so)
> 
> 
> GLIBC follows with the Linux kernel. In order to get a newer version of GLIBC the Linux kernel must be upgraded and this will involve upgrading all the software on Saga. We cannot do this. The problem is that the library mentioned in the error message has not been compiled on Saga. At the moment, I don't see how this problem can be solved.
> 
> Best regards,
> 
> Henrik