technical:recipes:emcee-in-virtualenv

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Next revision		 Previous revision
technical:recipes:emcee-in-virtualenv [2020-07-16 10:25] – created freytechnical:recipes:emcee-in-virtualenv [2021-02-24 23:45] (current) – frey
Line 1: Line 1:
 +====== Python Virtualenv: emcee and pyKLIP ======
  
 +Some portions of this recipe are specific to the IT-RCI HPC clusters (VALET-based environment setup) but otherwise should work on any system.
 +
 +===== Preparations =====
 +
 +Choose a directory in which to install the one (or more) virtual environments (virtualenvs).  The string ''«env-name»'' will denote the name chosen for the virtualenvs -- this recipe will use ''mcmc-env''.  Do not locate this directory under the ''/lustre/scratch'' file system; typically a directory under the workgroup's storage is appropriate:
 +
 +  * If setting-up this virtualenv for multiple users in the workgroup, choose ''${WORKDIR}/sw/«env-name»'' as the base directory.
 +  * If this virtualenv is solely for personal use, choose ''${WORKDIR}/users/<username>/sw/«env-name»'', for example.
 +
 +Note that these examples assume a standard workgroup storage layout with group-writable ''sw'' and ''users'' directories at the top level.  Create the directory:
 +
 +<code bash>
 +$ mkdir "${WORKDIR}/sw/mcmc-env"
 +</code>
 +
 +The virtualenv will be created using Intel's distribution for Python, but Anaconda would work just as well.  One of the cluster's MPI libraries also needs to be loaded into the environment:
 +
 +<code bash>
 +$ vpkg_require intel-python/2019u5:python3 openmpi/3.1.3
 +</code>
 +
 +Make note of these two package identifiers:  they will later be dependencies cited in the VALET package definition for this virtualenv.
 +
 +===== Versioning =====
 +
 +In order to maintain one or more distinct instances of the ''mcmc-env'' virtualenv, each must be given a distinct version.  While this could be any unique sequence of characters, it's usually best to adhere to a semantic versioning scheme.  For Python virtualenvs the current date with an extra numerical component (in case multiple instances are created on the same day) is useful:  ''YYYYMMDD.##'', where the four-digit year and zero-padded month and day number are used.  The trailing ''##'' is also zero-padded.  For example, ''20200716.01'' since today is July 16, 2020, and it's the first virtualenv created today.
 +
 +===== Create the Virtualenv =====
 +
 +Use the ''conda'' command to create the virtualenv and populate it with the most basic dependencies:
 +
 +<code bash>
 +$ conda create --prefix="${WORKDIR}/sw/mcmc-env/20200716.01" --copy pip python=3 numpy scipy matplotlib astropy
 +</code>
 +
 +The basic dependencies do not require an extensive list of other packages, most notably the ''mpi4py'' that must be built against the cluster's MPI library.  If the virtualenv build succeeds (and why shouldn't it!) the new virtualenv can be activated and the location of the ''pip'' command inside it can be verified:
 +
 +<code bash>
 +$ source activate "${WORKDIR}/sw/mcmc-env/20200716.01"
 +$ which pip
 +/work/workgroup/sw/mcmc-env/20200716.01/bin/pip
 +</code>
 +
 +===== Update and Install Other Packages =====
 +
 +For the sake of the ''emcee'' and ''pyKLIP'' packages it is recommended that a few components be updated or installed first:
 +
 +<code bash>
 +$ pip install -U setuptools setuptools_scm pep517 pytest
 +</code>
 +
 +The ''pytest'' package is a soft requirement of the ''debrisdiskfm'' package.  Next, install //from source// the ''mpi4py'' package as well as a few others:
 +
 +<code bash>
 +$ pip install --no-binary :all: mpi4py corner image_registration git+https://github.com/seawander/DebrisDiskFM.git#egg=Package schwimmbad
 +</code>
 +
 +The ''image_registration'' package is another soft requirement of the ''debrisdiskfm'' package.  This may take some time since binary parts of the packages must be compiled and linked (e.g. the shared library containing hooks into the cluster MPI library for ''mpi4py'') before being copied into the virtualenv.
 +
 +===== Install pyKLIP =====
 +
 +The ''pyKLIP'' package must be installed from a cloned github source tree.  The ''emcee'' package is a requirement of ''pyKLIP'' and will be installed automatically via ''pip''((The ''emcee'' package is pure Python, without any binary components that would require compiling and linking, so it is acceptable to let ''setuptools'' satisfy the dependency using ''pip''.  This is not true of all of the ''pyKLIP'' required packages, though.)).
 +
 +Rather than using the head of the git source tree -- corresponding to ongoing revisions/bug-fixes and not a specific release of the package -- the most recent release tag will be used:
 +
 +<code bash>
 +$ git clone https://bitbucket.org/pyKLIP/pyklip.git pyklip-src
 +$ cd pyklip-src
 +$ git checkout v2.1
 +</code>
 +
 +Installation into the virtualenv proceeds simply:
 +
 +<code bash>
 +$ python setup.py install
 +</code>
 +
 +Followed by removal of the cloned source:
 +
 +<code bash>
 +$ cd ..
 +$ rm -rf pyklip-src
 +</code>
 +
 +===== Completed virtualenv =====
 +
 +At this point the packages present in the completed virtualenv can be listed:
 +
 +<code bash>
 +$ pip list
 +Package            Version
 +------------------ -------------------
 +astropy            4.0.1.post1
 +certifi            2020.6.20
 +corner             2.1.0
 +cycler             0.10.0
 +debrisdiskfm       0.1.0
 +emcee              3.0.2
 +FITS-tools         0.2
 +gqueue             3.0.3
 +image-registration 0.2.4
 +kiwisolver         1.2.0
 +matplotlib         3.2.2
 +mkl-fft            1.1.0
 +mkl-random         1.1.1
 +mkl-service        2.3.0
 +mpi4py             3.0.3
 +numpy              1.18.5
 +pep517             0.8.2
 +pip                20.1.1
 +pyklip             2.1
 +pyparsing          2.4.7
 +python-dateutil    2.8.1
 +schwimmbad         0.3.1
 +scipy              1.5.0
 +setuptools         49.2.0.post20200714
 +setuptools-scm     4.1.2
 +sip                4.19.13
 +six                1.15.0
 +toml               0.10.1
 +tornado            6.0.4
 +wheel              0.34.2
 +</code>
 +
 +Version numbers may differ as newer versions of the component packages are released, but at the very least the
 +  * emcee
 +  * debrisdiskfm
 +  * schwimmbad
 +  * astropy
 +  * matplotlib
 +  * numpy
 +  * pyklip
 +packages should all be present in the list.
 +
 +Return the shell environment to its original state before proceeding:
 +
 +<code bash>
 +$ vpkg_rollback all
 +</code>
 +
 +===== VALET Package Definition =====
 +
 +To simplify the addition of this virtualenv to the runtime environment, a VALET package definition file should be created.  If the virtualenv was installed for use by one or more members of the workgroup, the file should be created as ''${WORKDIR}/sw/valet/«env-name».vpkg_yaml''.  If the virtualenv is for personal use, create the file as ''~/.valet/«env-name».vpkg_yaml''.  Naturally, you must ensure that the directory (''${WORKDIR}/sw/valet'' or ''~/.valet'', respectively) exists prior to creating the package definition file.
 +
 +The package definition file uses the [[https://yaml.org|YAML]] format.  Information to know before proceeding:
 +
 +^Item^Description^Value in this example^
 +|''«env-name»''|The name of the virtualenv created|''mcmc-env''|
 +|''«prefix»''|The directory in which the ''«env-name»'' directory to hold the virtualenv was created|''/work/workgroup/sw''|
 +|''«version»''|The version identifier chosen for this instance of the virtualenv|''20200716.01''|
 +|''«vpkg-file»''|The path to the virtualenv package definition file|''/work/workgroup/sw/valet/mcmc-env.vpkg_yaml''|
 +
 +The package definition file should follow this format:
 +
 +<code yaml>
 +«env-name»:
 +    prefix: «prefix»/«env-name»
 +    description: emcee and pyKLIP python environments
 +    flags:
 +        - no-standard-paths
 +    default-version: "«version»"
 +    actions:
 +        - action: source
 +          script:
 +              sh: anaconda-activate.sh
 +          success: 0
 +    versions:
 +        "«version»":
 +            description: July 16, 2020, build 01
 +            dependencies:
 +                - openmpi/3.1.3
 +                - intel-python/2019u5:python3
 +</code>
 +
 +For the virtualenv created in the course of this recipe, the resulting file ''«vpkg-file»'' would look like:
 +
 +<code yaml>
 +mcmc-env:
 +    prefix: /work/workgroup/sw/mcmc-env
 +    description: emcee and pyKLIP python environments
 +    flags:
 +        - no-standard-paths
 +    default-version: "20200716.01"
 +    actions:
 +        - action: source
 +          script:
 +              sh: anaconda-activate.sh
 +          success: 0
 +    versions:
 +        "20200716.01":
 +            description: July 16, 2020, build 01
 +            dependencies:
 +                - openmpi/3.1.3
 +                - intel-python/2019u5:python3
 +</code>
 +
 +The file can be checked for proper syntax:
 +
 +<code bash>
 +$ vpkg_check "«vpkg-file»"
 +</code>
 +
 +If all is okay, then the virtualenv can be loaded in login sessions or in job scripts as:
 +
 +<code bash>
 +$ vpkg_require «env-name»/«version»
 +</code>
 +
 +which for this recipe is:
 +
 +<code bash>
 +$ vpkg_require mcmc-env/20200716.01
 +Adding dependency `libfabric/1.6.1` to your environment
 +Adding dependency `openmpi/3.1.3` to your environment
 +Adding dependency `intel-python/2019u5:python3` to your environment
 +Adding package `mcmc-env/20200716.01` to your environment
 +$ which python3
 +/work/workgroup/sw/mcmc-env/20200716.01/bin/python3
 +$ which pip
 +/work/workgroup/sw/mcmc-env/20200716.01/bin/pip
 +</code>
 +
 +==== Adding New Instances of the virtualenv ====
 +
 +Each new instance of the virtualenv will receive a unique version identifier prior to being created.  Append a new //versions// dictionary to the package definition file:
 +
 +<code yaml>
 +mcmc-env:
 +    prefix: /work/workgroup/sw/mcmc-env
 +    description: emcee and pyKLIP python environments
 +    flags:
 +        - no-standard-paths
 +    default-version: "20200716.01"
 +    actions:
 +        - action: source
 +          script:
 +              sh: anaconda-activate.sh
 +          success: 0
 +    versions:
 +        "20200716.01":
 +            description: July 16, 2020, build 01
 +            dependencies:
 +                - openmpi/3.1.3
 +                - intel-python/2019u5:python3
 +        "20200716.02":
 +            description: July 16, 2020, build 02
 +            dependencies:
 +                - openmpi/4.0.2
 +                - intel-python/2019u5:python3
 +</code>
 +
 +where the dependencies will change if different versions of the Open MPI library or Intel distribution for Python were used when creating the virtualenv.
 +
 +If the new version should be promoted to being the default (e.g. when ''vpkg_require mcmc-env/default'' is used) then alter the ''default-version'' value accordingly.