Source code for metawards.app.run

#!/bin/env python3
"""
The metawards command line program.

Usage:
    To get the help for this program and the list of all of the
    arguments (with defaults) use;

    metawards --help
"""

__all__ = ["get_parallel_scheme", "parse_args", "get_hostfile",
           "get_cores_per_node", "get_threads_per_task",
           "scoop_supervisor", "mpi_supervisor",
           "cli"]


def get_parallel_scheme():
    """This function tries to work out which of the different parallel
       methods we should use to distribute work between multiple processes.

       Detected schemes are "mpi4py", "scoop", and if none of these
       are found, then "multiprocessing"
    """
    import sys

    if "mpi4py" in sys.modules:
        return "mpi4py"
    elif "scoop" in sys.modules:
        return "scoop"
    else:
        return "multiprocessing"


[docs]def parse_args(): """Parse all of the command line arguments""" import argparse import sys parser = argparse.ArgumentParser( description="MetaWards epidemic modelling - see " "https://github.com/metawards/metawards " "for more information", prog="metawards") parser.add_argument('--version', action="store_true", default=None, help="Print the version information about metawards") parser.add_argument('-i', '--input', type=str, help="Input file for the simulation that specifies " "the adjustable parameters to change for each " "run of the model. You must supply some " "input to run a model!") parser.add_argument('-l', '--line', type=str, default=None, nargs="*", help="Line number (or line numbers) containing the " "values of adjustable parameters to run for this " "run (or runs) of the model If this isn't " "specified then model runs will be performed " "for adjustable parameters given on all lines " "from the input file. You can specify many " "numbers, and ranges are also accepted, e.g. " "'-l 5 6-10 12,13,14' etc. Note that the line " "numbers are 0-indexed, e.g. the first line of " "the file is line 0. Ranges are inclusive, " "so 3-5 is the same as 3 4 5") parser.add_argument("-r", '--repeats', type=int, default=None, help="The number of repeat runs of the model to " "perform for each value of the adjustable " "parameters. By default only a single " "run will be performed for each set of " "adjustable parameters") parser.add_argument('-s', '--seed', type=int, default=None, help="Random number seed for this run " "(default is to use a random seed)") parser.add_argument('-a', '--additional', type=str, default=None, nargs="*", help="File (or files) containing additional " "seed of outbreak for the model. These are " "used to seed additional infections on " "specific days at different locations " "during a model run") parser.add_argument('-o', '--output', type=str, default="output", help="Path to the directory in which to place all " "output files (default 'output'). This " "directory will be subdivided if multiple " "adjustable parameter sets or repeats " "are used.") parser.add_argument('-d', '--disease', type=str, default=None, help="Name of the disease to model " "(default is 'ncov')") parser.add_argument('-m', '--model', type=str, default=None, help="Name of the input model data set for the " "network (default is '2011Data')") parser.add_argument('--start-date', type=str, default=None, help="Date of the start of the model outbreak. " "This accepts dates either is iso-format, " "or fuzzy dates such as 'monday', 'tomorrow' " "etc. This is used to work out which days " "are weekends, or to make it easier to specify " "time-based events.") parser.add_argument('--start-day', type=int, default=0, help="The start day of the model outbreak. By " "default the model outbreak starts on day " "zero (0), with each step of the model " "representing an additional day. Use this " "to start from a later day, which may be " "useful if you want to more quickly reach " "time based events. Note that the passed " "'--start-date' is always day 0, so day 10 " "has a date which is 10 days after start-date") parser.add_argument('-p', '--parameters', type=str, default="march29", help="Name of the input parameter set used to " "control the simulation (default 'march29')") parser.add_argument('-R', '--repository', type=str, default=None, help="Path to the MetaWardsData repository. If " "unspecified this defaults to the value " "in the environment variable METAWARDSDATA " "or, if that isn't specified, to " "$HOME/GitHub/MetaWardsData") parser.add_argument('-P', '--population', type=int, default=57104043, help="Initial population (default 57104043)") parser.add_argument('-n', '--nsteps', type=int, default=730, help="Maximum number of steps (days) to run for the " "simulation. Each step represents one day in the " "outbreak (default is to run for a maximum " "of two years - 730 days)") parser.add_argument('-u', '--user-variables', type=str, default=None, help="Name of the file containing user-defined " "custom variables. These provide extra " "information that can be read by the " "custom integrators or custom extractors.") parser.add_argument('--iterator', type=str, default=None, help="Name of the iterator to use to advance the " "outbreak at each step (day). For a full " "explanation see the tutorial at " "https://metawards.github.io") parser.add_argument("--extractor", type=str, default=None, help="Name of the extractor to use to extract " "information during a model run. For a full " "explanation see the tutorial at " "https://metawards.github.io") parser.add_argument('--UV', type=float, default=0.0, help="Value for the UV parameter for the model " "(default is 0.0)") parser.add_argument('--nthreads', type=int, default=None, help="Number of threads over which parallelise an " "individual model run. The total number of " "cores used by metawards will be " "nprocesses x nthreads") parser.add_argument("--nprocs", type=int, default=None, help="The number of processes over which to " "parallelise the different model runs for " "different adjustable parameter sets and " "repeats. By default this will automatically " "work out the number of processes based on " "the way metawards is launched. Use this " "option if you want to specify the number " "of processes manually.") parser.add_argument('--hostfile', type=str, default=None, help="The hostfile containing the names of the " "compute nodes over which to run a parallel " "job. If this is not set, the program will " "attempt to automatically get this information " "from the cluster queueing system. Use this " "if the auto-detection fails") parser.add_argument('--cores-per-node', type=int, default=None, help="Set the number of processor cores available " "on each of the compute nodes in the cluster " "that will be used to run the models " "(if a cluster is used). If this is not " "set then the program will attempt to " "get this information from the cluster " "queueing system. Use this option if the " "auto-detection fails.") parser.add_argument('--auto-bzip', action="store_true", default=None, help="Automatically bz2 compress " "all output files as they are written.") parser.add_argument('--no-auto-bzip', action="store_true", default=None, help="Do not automatically bz2 compress " "all output files as they are written.") parser.add_argument('--force-overwrite-output', action="store_true", default=False, help="Whether or not to force overwriting of any " "existing output. Using this option will " "tell metawards that it is safe to delete " "the contents of the output directory " "specified in by '--output' if it already " "exists. Dangerous as this can remove " "existing output files") parser.add_argument('--max-nodes', type=int, default=16384, help="Maximum number of nodes that can be read") parser.add_argument('--max-links', type=int, default=4194304, help="Maximum number of links that can be read") parser.add_argument('--profile', action="store_true", default=None, help="Enable profiling of the code") parser.add_argument('--no-profile', action="store_true", default=None, help="Disable profiling of the code") parser.add_argument('--mpi', action="store_true", default=None, help="Force use of MPI to parallelise across runs") parser.add_argument('--scoop', action="store_true", default=None, help="Force use of scoop to parallelise across runs") # this hidden option is used to tell the main process started using # mpi that it shouldn't try to become a supervisor parser.add_argument('--already-supervised', action="store_true", default=None, help=argparse.SUPPRESS) args = parser.parse_args() if args.version: from metawards import get_version_string print(get_version_string()) sys.exit(0) return (args, parser)
def get_hostfile(args): """Attempt to find the name of the hostfile used to specify the hosts to use in a cluster """ if args.hostfile: return args.hostfile import os # PBS hostfile = os.getenv("PBS_NODEFILE") if hostfile: return hostfile # SLURM hostfile = os.getenv("SLURM_HOSTFILE") if hostfile: return hostfile return None def get_cores_per_node(args): """Return the number of cores per node in the cluster""" if args.cores_per_node: return args.cores_per_node import os try: cores_per_node = int(os.getenv("METAWARDS_CORES_PER_NODE")) if cores_per_node > 0: return cores_per_node except Exception: pass raise ValueError("You must specify the number of cores per node " "using --cores-per-node or by setting the " "environment variable METAWARDS_CORES_PER_NODE") def get_threads_per_task(args): if args.nthreads: return args.nthreads import os try: nthreads = int(os.getenv("METAWARDS_THREADS_PER_TASK")) if nthreads > 0: return nthreads nthreads = int(os.getenv("OMP_NUM_THREADS")) if nthreads > 0: return nthreads except Exception: pass raise ValueError("You must specify the number of threads per task " "using --nthreads or by setting the " "environment variables METAWARDS_THREADS_PER_TASK " "or OMP_NUM_THREADS")
[docs]def scoop_supervisor(hostfile, args): """Function used by the scoop supervisor to get the information needed to form the scoop call to run a scoop version of the program """ import os import sys print("RUNNING A SCOOP PROGRAM") outdir = args.output if not os.path.exists(outdir): os.mkdir(outdir) cores_per_node = get_cores_per_node(args) print(f"Will run jobs assuming {cores_per_node} cores per compute node") # based on the number of threads requested and the number of cores # per node, we can work out the number of scoop processes to start, # and can write a hostfile that will create the right layout nthreads = get_threads_per_task(args) print(f"Will use {nthreads} OpenMP threads per model run...") tasks_per_node = int(cores_per_node / nthreads) print(f"...meaning that the number of model runs per node will be " f"{tasks_per_node}") # Next, read the hostfile to get a unique list of hostnames hostnames = {} with open(hostfile, "r") as FILE: line = FILE.readline() while line: hostname = line.strip() hostnames[hostname] = 1 line = FILE.readline() hostnames = list(hostnames.keys()) hostnames.sort() print(f"Number of compute nodes equals {len(hostnames)}") print(", ".join(hostnames)) # how many tasks can we perform in parallel? nprocs = tasks_per_node * len(hostnames) if args.nprocs: if nprocs != args.nprocs: print(f"WARNING: You are using a not-recommended number of " f"processes {args.nprocs} for the cluster {nprocs}.") nprocs = args.nprocs print(f"Total number of parallel processes to run will be {nprocs}") print(f"Total number of cores in use will be {nprocs*nthreads}") # Now write a new hostfile that round-robins the MPI tasks over # the nodes for 'tasks_per_node' runs hostfile = os.path.join(outdir, "hostfile") print(f"Writing hostfile to {hostfile}") with open(hostfile, "w") as FILE: i = 0 while i < nprocs: for hostname in hostnames: FILE.write(hostname + "\n") i += 1 if i == nprocs: break # now craft the scoop command that will use this hostfile to # run the job - remember to pass the option to stop the main process # attempting to become a supervisor itself... import subprocess import shlex pyexe = sys.executable script = os.path.abspath(sys.argv[0]) args = " ".join(sys.argv[1:]) # also need to tell the main program the number of processes # as it can't work it out itself cmd = f"{pyexe} -m scoop --hostfile {hostfile} -n {nprocs} " \ f"{script} --already-supervised {args} --nprocs {nprocs}" print(f"Executing scoop job using '{cmd}'") try: args = shlex.split(cmd) subprocess.run(args).check_returncode() except Exception as e: print("ERROR: Something went wrong!") print(f"{e.__class__}: {e}") sys.exit(-1) print("Scoop processes completed successfully")
[docs]def mpi_supervisor(hostfile, args): """Function used by the MPI supervisor to get the information needed to form the mpiexec call to run an MPI version of the program """ import os import sys print("RUNNING AN MPI PROGRAM") outdir = args.output if not os.path.exists(outdir): os.mkdir(outdir) cores_per_node = get_cores_per_node(args) print(f"Will run jobs assuming {cores_per_node} cores per compute node") # based on the number of threads requested and the number of cores # per node, we can work out the number of mpi processes to start, # and can write a hostfile that will create the right layout nthreads = get_threads_per_task(args) print(f"Will use {nthreads} OpenMP threads per model run...") tasks_per_node = int(cores_per_node / nthreads) print(f"...meaning that the number of model runs per node will be " f"{tasks_per_node}") # Next, read the hostfile to get a unique list of hostnames hostnames = {} with open(hostfile, "r") as FILE: line = FILE.readline() while line: hostname = line.strip() hostnames[hostname] = 1 line = FILE.readline() hostnames = list(hostnames.keys()) hostnames.sort() print(f"Number of compute nodes equals {len(hostnames)}") print(", ".join(hostnames)) # how many tasks can we perform in parallel? nprocs = tasks_per_node * len(hostnames) if args.nprocs: if nprocs != args.nprocs: print(f"WARNING: You are using an unrecommended number of " f"processes {args.nprocs} for the cluster {nprocs}.") nprocs = args.nprocs print(f"Total number of parallel processes to run will be {nprocs}") print(f"Total number of cores in use will be {nprocs*nthreads}") # Now write a new hostfile that round-robins the MPI tasks over # the nodes for 'tasks_per_node' runs hostfile = os.path.join(outdir, "hostfile") print(f"Writing hostfile to {hostfile}") with open(hostfile, "w") as FILE: i = 0 while i < nprocs: for hostname in hostnames: FILE.write(hostname + "\n") i += 1 if i == nprocs: break # now craft the mpiexec command that will use this hostfile to # run the job - remember to pass the option to stop the main process # attempt to become a supervisor itself... mpiexec = os.getenv("MPIEXEC") if mpiexec is None: mpiexec = "mpiexec" # check for weird mpiexecs... import subprocess import shlex try: args = shlex.split(f"{mpiexec} -v") p = subprocess.run(args, stdout=subprocess.PIPE, stderr=subprocess.STDOUT) v = p.stdout.decode("utf-8").strip() print(f"{mpiexec} -v => {v}") if v.find("HPE HMPT") != -1: raise ValueError( "metawards needs a more modern MPI library than HPE's, " "so please compile to another MPI and use that.") except Exception as e: print(f"[ERROR] {e.__class__} {e}") pyexe = sys.executable script = os.path.abspath(sys.argv[0]) args = " ".join(sys.argv[1:]) cmd = f"{mpiexec} -np {nprocs} -hostfile {hostfile} " \ f"{pyexe} -m mpi4py {script} --already-supervised {args}" print(f"Executing MPI job using '{cmd}'") try: args = shlex.split(cmd) subprocess.run(args).check_returncode() except Exception as e: print("ERROR: Something went wrong!") print(f"{e.__class__}: {e}") sys.exit(-1) print("MPI processes completed successfully")
def cli(): """Main function for the command line interface. This does one of three things: 1. If this is the main process, then it parses the arguments and runs and manages the jobs 2. If this is a worker process, then it starts up and waits for work 3. If this is a supervisor process, then it query the job scheduling system for information about the compute nodes to use, and will then set up and run a manager (main) process that will use those nodes to run the jobs """ # get the parallel scheme now before we import any other modules # so that it is clear if mpi4py or scoop (or another parallel module) # has been imported via the required "-m module" syntax parallel_scheme = get_parallel_scheme() if parallel_scheme == "mpi4py": from mpi4py import MPI comm = MPI.COMM_WORLD nprocs = comm.Get_size() rank = comm.Get_rank() if rank != 0: # this is a worker process, so should not do anything # more until it is given work in the pool print(f"Starting worker process {rank+1} of {nprocs-1}...") return else: print("Starting main process...") elif parallel_scheme == "scoop": print("STARTING SCOOP PROCESS") import sys args, parser = parse_args() if not args.already_supervised: hostfile = get_hostfile(args) if hostfile: # The user has asked to run a parallel job - this means that this # process is the parallel supervisor if args.mpi: mpi_supervisor(hostfile, args) return elif args.scoop: scoop_supervisor(hostfile, args) return # neither is preferred - if scoop is installed then use that try: import scoop have_scoop = True except Exception: have_scoop = False if have_scoop: scoop_supervisor(hostfile, args) return # do we have MPI? try: import mpi4py have_mpi4py = True except Exception: have_mpi4py = False if have_mpi4py: mpi_supervisor(hostfile, args) return # we don't have any other option, just keep going and # use multiprocessing - in this case we don't need a # supervisor and this is the main process # This is now the code for the main process # WE NEED ONE OF these listed options; should_run = False for arg in [args.input, args.repeats, args.disease, args.additional, args.model, args.iterator, args.extractor]: if arg is not None: should_run = True break if not should_run: parser.print_help(sys.stdout) sys.exit(0) if args.repeats is None: args.repeats = 1 # import the parameters here to speed up the display of help from metawards import Parameters, Network, Population, get_version_string # print the version information first, so that there is enough # information to enable someone to reproduce this run print(get_version_string()) # also print the full command line used for this job print(f"Command used to run this job:\n{' '.join(sys.argv)}\n") if args.input: # get the line numbers of the input file to read if args.line is None or len(args.line) == 0: linenums = None print(f"Using parameters from all lines of {args.input}") else: from metawards.utils import string_to_ints linenums = string_to_ints(args.line) if len(linenums) == 0: print(f"You cannot read no lines from {args.input}?") sys.exit(-1) elif len(linenums) == 1: print(f"Using parameters from line {linenums[0]} of " f"{args.input}") else: print(f"Using parameters from lines {linenums} of " f"{args.input}") from metawards import VariableSets, VariableSet variables = VariableSets.read(filename=args.input, line_numbers=linenums) else: from metawards import VariableSets, VariableSet # create a VariableSets with one null VariableSet variables = VariableSets() variables.append(VariableSet()) nrepeats = args.repeats if nrepeats is None or nrepeats < 1: nrepeats = 1 if nrepeats == 1: print("Performing a single run of each set of parameters") else: print(f"Performing {nrepeats} runs of each set of parameters") variables = variables.repeat(nrepeats) # working out the number of processes and threads... from metawards.utils import guess_num_threads_and_procs (nthreads, nprocs) = guess_num_threads_and_procs( njobs=len(variables), nthreads=args.nthreads, nprocs=args.nprocs, parallel_scheme=parallel_scheme) print(f"\nNumber of threads to use for each model run is {nthreads}") if nprocs > 1: print(f"Number of processes used to parallelise model " f"runs is {nprocs}") print(f"Parallelisation will be achieved using {parallel_scheme}") # sort out the random number seed seed = args.seed if seed is None: import random seed = random.randint(10000, 99999999) print(f"\nUsing random number seed {seed}") # get the starting day and date start_day = args.start_day if start_day < 0: raise ValueError(f"You cannot use a start day {start_day} that is " f"less than zero!") start_date = None if args.start_date: try: from dateparser import parse start_date = parse(args.start_date).date() except Exception: pass if start_date is None: from datetime import date try: start_date = date.fromisoformat(args.start_date) except Exception as e: raise ValueError( f"Cannot interpret a valid date from " f"'{args.start_date}'. Error is " f"{e.__class__} {e}") if start_date is None: from datetime import date start_date = date.today() print(f"\nDay zero is {start_date.strftime('%A %B %d %Y')}") if start_day != 0: from datetime import timedelta start_day_date = start_date + timedelta(days=start_day) print(f"Starting on day {start_day}, which is " f"{start_day_date.strftime('%A %B %d %Y')}") else: start_day_date = start_date # now find the MetaWardsData repository as this will be needed # for the repeat command line too (repository, repository_version) = Parameters.get_repository( args.repository) print(f"\nUsing MetaWardsData at {repository}") print(f"This is cloned from {repository_version['repository']}") print(f"branch {repository_version['branch']}, version " f"{repository_version['version']}") if repository_version["is_dirty"]: print("## WARNING - this repository is dirty, meaning that the data") print("## WARNING - has not been committed to git. This may make ") print("## WARNING - this calculation very difficult to reproduce") # now work out the minimum command line needed to repeat this job args.seed = seed args.nprocs = nprocs args.nthreads = nthreads args.start_date = start_date.isoformat() args.repository = repository repeat_cmd = "metawards" for key, value in vars(args).items(): if value is not None: k = key.replace("_", "-") if isinstance(value, bool): if value: repeat_cmd += f" --{k}" elif isinstance(value, list): repeat_cmd += f" --{k}" for val in value: v = str(val) if " " in v: repeat_cmd += f" '{v}''" else: repeat_cmd += f" {v}" else: v = str(value) if " " in v: repeat_cmd += f" --{k} '{v}''" else: repeat_cmd += f" --{k} {v}" t = "*** To repeat this job use the command ***" print("\n" + "*"*len(t)) print(t) print("*"*len(t) + "\n") print(repeat_cmd + "\n") # load all of the parameters try: params = Parameters.load(parameters=args.parameters) except Exception as e: print(f"Unable to load parameter files. Make sure that you have " f"cloned the MetaWardsData repository and have set the " f"environment variable METAWARDSDATA to point to the " f"local directory containing the repository, e.g. the " f"default is $HOME/GitHub/MetaWardsData") raise e # should we profile the code? (default no as it prints a lot) profile = False if args.no_profile: profile = False elif args.profile: profile = True # load the disease and starting-point input files if args.disease: params.set_disease(args.disease) else: params.set_disease("ncov") if args.model: params.set_input_files(args.model) else: params.set_input_files("2011Data") # load the user-defined custom parameters if args.user_variables: custom = VariableSet.read(args.user_variables) print(f"Adjusting variables to {custom}") custom.adjust(params) # read the additional seeds if args.additional is None or len(args.additional) == 0: print("Not using any additional seeds...") else: for additional in args.additional: print(f"Loading additional seeds from {additional}") params.add_seeds(additional) # extra parameters that are set params.UV = args.UV # set these extra parameters to 0 params.static_play_at_home = 0 params.play_to_work = 0 params.work_to_play = 0 params.daily_imports = 0.0 # the size of the starting population population = Population(initial=args.population, date=start_day_date, day=start_day) print("\nBuilding the network...") network = Network.build(params=params, calculate_distances=True, max_nodes=args.max_nodes, max_links=args.max_links, profile=profile) print("\nRun the model...") from metawards import OutputFiles from metawards.utils import run_models outdir = args.output if outdir is None: outdir = "output" if args.force_overwrite_output: prompt = None else: prompt = input auto_bzip = True if args.auto_bzip: auto_bzip = True elif args.no_auto_bzip: auto_bzip = False if args.iterator: iterator = args.iterator # eventually I should get the filename and function from this, # and then convert it into an absolute path to make this safe # on a cluster. Then I should copy the file into the output to # make sure that the calculation is reproducible. Will do this # when I copy this work to extractor else: iterator = None if args.extractor: extractor = args.extractor # see above ;-) else: extractor = None with OutputFiles(outdir, force_empty=args.force_overwrite_output, auto_bzip=auto_bzip, prompt=prompt) as output_dir: result = run_models(network=network, variables=variables, population=population, nprocs=nprocs, nthreads=nthreads, seed=seed, nsteps=args.nsteps, output_dir=output_dir, iterator=iterator, extractor=extractor, profile=profile, parallel_scheme=parallel_scheme) if result is None or len(result) == 0: print("No output - end of run") return 0 # write the result to a csv file that can be easily read by R or # pandas - this will be written compressed to save space RESULTS = output_dir.open("results.csv", auto_bzip=auto_bzip) print(f"\nWriting a summary of all results into the\n" f"csv file {output_dir.get_filename('results.csv')}.\n" f"You can use this to quickly look at statistics across\n" f"all runs using e.g. R or pandas") varnames = result[0][0].variable_names() if varnames is None or len(varnames) == 0: varnames = "" else: varnames = ",".join(varnames) + "," has_date = result[0][1][0].date if has_date: datestring = "date," else: datestring = "" RESULTS.write(f"fingerprint,repeat,{varnames}" f"day,{datestring}S,E,I,R,IW,UV\n") for varset, trajectory in result: varvals = varset.variable_values() if varvals is None or len(varvals) == 0: varvals = "" else: varvals = ",".join(map(str, varvals)) + "," start = f"{varset.fingerprint()}," \ f"{varset.repeat_index()},{varvals}" for i, pop in enumerate(trajectory): if pop.date: d = pop.date.isoformat() + "," else: d = "" RESULTS.write(f"{start}{pop.day},{d}{pop.susceptibles}," f"{pop.latent},{pop.total}," f"{pop.recovereds},{pop.n_inf_wards}," f"{pop.scale_uv}\n") print("End of the run") return 0 if __name__ == "__main__": cli() else: # this is one of the worker processes - make sure that they # have imported metawards from metawards.utils import run_worker