{bio}[fosscuda/2019b] RoseTTAFold v1.0.0 w/ Python 3.7.4

[zemu-csi]

(created using eb --new-pr)

[zemu-csi]

Depends on #13794 #13793 #13792 #13798

[zemu-csi]

I'm not sure if there is anything we could add as sanity check. Maybe something like python -c 'import ...
This is the slurm job script i used for testing:

#!/bin/bash
#SBATCH --job-name=RoseTTAFold
#SBATCH --output=rosettafold-job-out.%J
#SBATCH --time=48:00:00
#SBATCH --nodes=1
#SBATCH --cpus-per-task=16
#SBATCH --gres=gpu:rtx2080ti:1

module load RoseTTAFold/1.0.0-fosscuda-2019b-Python-3.7.4-PyTorch-1.8.1

export BFDPATH=/scratch/pdbs/bfd
export PDB100PATH=/scratch/pdbs/pdb100_2021Mar03
export UNIREF30PATH=/scratch/pdbs/UniRef30_2020_06
export MEM=$SLURM_MEM_PER_NODE
export CPU=$SLURM_CPUS_PER_TASK

run_pyrosetta_ver.sh input.fa $PWD

[zemu-csi]

Test report by @zemu-unile
SUCCESS
Build succeeded for 1 out of 1 (1 easyconfigs in total)
login01.sc.uni-leipzig.de - Linux centos linux 7.9.2009, x86_64, AMD EPYC 7551P 32-Core Processor, Python 3.6.8
See https://gist.github.com/791d88cbfeb3c8e7cfbe0a73cc8a62c4 for a full test report.

[SebastianAchilles]

@boegelbot please test @ generoso

[boegelbot]

@SebastianAchilles: Request for testing this PR well received on login1

PR test command 'EB_PR=13795 EB_ARGS= /opt/software/slurm/bin/sbatch --job-name test_PR_13795 --ntasks=4 ~/boegelbot/eb_from_pr_upload_generoso.sh' executed!

• exit code: 0
• output:

Submitted batch job 6934

Test results coming soon (I hope)...

Details

- notification for comment with ID 922869452 processed

Message to humans: this is just bookkeeping information for me,
it is of no use to you (unless you think I have a bug, which I don't).

[boegelbot]

Test report by @boegelbot
FAILED
Build succeeded for 9 out of 14 (1 easyconfigs in total)
cnx1 - Linux rocky linux 8.4, x86_64, Intel(R) Xeon(R) CPU E5-2690 v3 @ 2.60GHz (haswell), Python 3.6.8
See https://gist.github.com/820decc93a153891a7a3ceae18c2477c for a full test report.

[SebastianAchilles]

@boegelbot please test @ generoso

Test report by @boegelbot
FAILED
Build succeeded for 9 out of 14 (1 easyconfigs in total)
cnx1 - Linux rocky linux 8.4, x86_64, Intel(R) Xeon(R) CPU E5-2690 v3 @ 2.60GHz (haswell), Python 3.6.8
See https://gist.github.com/820decc93a153891a7a3ceae18c2477c for a full test report.

Failed with SIGKILL in PyTorch-1.8.1-fosscuda-2019b-Python-3.7.4.eb.
https://gist.github.com/boegelbot/92e258ef2737604701192ebb01cdd6b3#file-pytorch-1-8-1-fosscuda-2019b-python-3-7-4_partial-log-L491

[SebastianAchilles]

@boegelbot please test @ generoso
CORE_CNT=16

[boegelbot]

@SebastianAchilles: Request for testing this PR well received on login1

PR test command 'EB_PR=13795 EB_ARGS= /opt/software/slurm/bin/sbatch --job-name test_PR_13795 --ntasks="16" ~/boegelbot/eb_from_pr_upload_generoso.sh' executed!

• exit code: 0
• output:

Submitted batch job 6937

Test results coming soon (I hope)...

Details

- notification for comment with ID 923717491 processed

Message to humans: this is just bookkeeping information for me,
it is of no use to you (unless you think I have a bug, which I don't).

[boegelbot]

Test report by @boegelbot
FAILED
Build succeeded for 3 out of 5 (1 easyconfigs in total)
cnx2 - Linux rocky linux 8.4, x86_64, Intel(R) Xeon(R) CPU E5-2660 v4 @ 2.00GHz (broadwell), Python 3.6.8
See https://gist.github.com/5f3580ce243719e888d996883a7225ea for a full test report.

[SebastianAchilles]

Test report by @boegelbot
FAILED
Build succeeded for 3 out of 5 (1 easyconfigs in total)
cnx2 - Linux rocky linux 8.4, x86_64, Intel(R) Xeon(R) CPU E5-2660 v4 @ 2.00GHz (broadwell), Python 3.6.8
See https://gist.github.com/5f3580ce243719e888d996883a7225ea for a full test report.

Failed because generoso has no sources for PyRosetta (no license).

[seb45tian]

@zemu-unile Could you change the HH-Suite dependency from 3.2.0 -> 3.3.0? HH-Suite 3.2.0 falsely detects AVX2 support on our ivybridge nodes (only AVX) as their CMake check (https://github.com/soedinglab/hh-suite/blob/v3.2.0/cmake/CheckSSEFeatures.cmake) does not return an "Illegal instruction" when -mavx2 and -O2 are enabled - -O2 is set by EasyBuild by default. They fixed this in 3.3.0 and RoseTTAfold does not specify a specific HHSuite version they depend on.

[zemu-csi]

One problem here is that HH-Suite 3.3.0 is in 2020 toolchains but RoseTTAFold needs 2019b due to requiring old TensorFlow 1.x. I am not sure if we should port HH-Suite 3.3.0 to 2019b. @Micket any thoughts on this?

[seb45tian]

Good point, however I don't see any issue having an EC for 3.2.0 and 3.3.0 in 2019b. But you are right, the proper way would be to patch 3.2.0 to make it work.

[SebastianAchilles]

Good point, however I don't see any issue having an EC for 3.2.0 and 3.3.0 in 2019b.

We would need to add an exception for the CI to allow multiple version in the same toolchain. But yes that is possible.
Alternatively we could also add TensorFlow 1.x to a newer toolchain with an exception. This depends on which toolchain you think will be more useful for RoseTTAFold.

[zemu-csi]

The Problem with TensorFlow 1.x is that it needs an older Cuda version. I'm still hoping that RoseTTAFold will switch to TensorFlow 2.x

[SebastianAchilles]

The Problem with TensorFlow 1.x is that it needs an older Cuda version. I'm still hoping that RoseTTAFold will switch to TensorFlow 2.x

Good point! Okay, then I suggest to add HH-Suite 3.3.0 to 2019b with an exception. Do you agree?

[zemu-csi]

Opened a PR for HH-Suite #14191

[jfgrimm]

As decided in issue #16330, we have deprecated the use of True to signify a system-toolchain dependency (#16384), in favour of the more intuitive SYSTEM template constant. Due to the change in the test suite, please run eb --sync-pr-with-develop 13795 and update the PR to use SYSTEM instead.

[boegel]

closing this since fosscuda/2019b is no longer supported, see https://docs.easybuild.io/policies/toolchains