Unfortunately not.
It freezes and or crashes wheninstalling TensorFlow
so, I ma not sure what to do :-(
Any advice?
Roberto
Il lunedì 16 ottobre 2023 06:57:40 GMT+1, Loris Bennett
<[email protected]> ha scritto:
Hi Roberto,
Roberto Scipioni <[email protected]> writes:
> Hi,
>
> I am trying the installation you recommended. It is taking forever > 10 hours
> !
> Did you experince the same?
Yes. It tool over 22 hours on our system (using 8 cores). 137 modules
in total are needed and individual ones, such as Bazel, if I remember
correctly, take ages by themselves.
As you wrote on Friday, I assume that by now, Monday, the installation
may have finished :-)
Cheers,
Loris
> Kind regards
> Roberto
>
> Il mercoledì 11 ottobre 2023 15:55:39 GMT+1, Loris Bennett
> <[email protected]> ha scritto:
>
> Ciao Roberto,
>
> Roberto Scipioni <[email protected]> writes:
>
>> Dear Sir or Madam,
>>
>> I have been trying to install alphafold using Easybuild on Centos 7/8
>> using eb 4.5 4.6 4.8.1 using different Toolchains, different
>>architectures, ....etc to no avail
>>
>> It stops systematically at Tensorflow/Bazel installation and/or in
>> some cases when attempting to install Alphafold at the repo section
>> giving errors about incomplete downloads / wrong EOF etc.
>>
>> I wonder whether anybody could give me any pointer on how to proceed.
>> Honestly I am not sure where to start ...:-(
>
> Recently I have used the following successfully
>
> eb AlphaFold-2.3.1-foss-2022a-CUDA-11.7.0.eb --robot
>--cuda-compute-capabilities=6.1,7.5 --buildpath=/dev/shm
>--tmpdir=/scratch/eb-build
>
> Using /dev/shm as the build path should accelerate the build, assuming
> you have enough memory. We change the tmpdir to prevent /tmp on the
> node where the build job runs from filling up, as other jobs may also be
> writing to there.
>
> HTH
>
> Loris
--
Dr. Loris Bennett (Herr/Mr)
ZEDAT, Freie Universität Berlin
