import os import numpy as np import pyopencl from psyrun import Param from psyrun.backend import LoadBalancingBackend from cue.trials import HebbRepetitionTrial rng = np.random.RandomState(39) n_trials = 25 seeds = range(n_trials) pspace = Param(extension=['forward-assoc', 'direct-assoc']) * Param( seed=seeds, trial=range(n_trials)) backend = LoadBalancingBackend exclude_from_result = ['cl_context'] pool_size = 4 max_jobs = 1 def setup(proc_id): context = pyopencl.create_some_context(answers=[0, proc_id]) return {'cl_context': context} def execute(trial, **kwargs): result = HebbRepetitionTrial().run(progress=False, **kwargs) return result
import numpy as np from psyrun import Param from psyrun.backend import LoadBalancingBackend from psyrun.scheduler import Slurm from cue.trials import CueTrial rng = np.random.RandomState(846777) n_trials = 100 seeds = range(100) pspace = Param(seed=seeds, trial=range(n_trials), min_evidence=0.03, noise=0.009, distractor_rate=0.4, ordinal_prob=0.1) exclude_from_result = ['cl_context'] min_items = 1 if platform.node().startswith('gra') or platform.node().startswith('cedar'): pool_size = 1 max_jobs = 100 workdir = '/scratch/jgosmann/cue' scheduler = Slurm(workdir) def timelimit(name): if 'split' in name or 'merge' in name: return '0-00:10' else:
from psyrun import Param from cogsci17_decide.trial import DecisionTrial pspace = ( (Param(network=['LCA'], scale=[1.], share_thresholding_intercepts=[False]) + Param(network=['IA']) * Param(share_thresholding_intercepts=[False]) * Param(scale=[1.])) * Param(baseline=[0.2, 0.6, 1.0]) * Param(target_sep=[0.05, 0.1, 0.15, 0.2]) * Param(noise=[0.0, 0.01, 0.02, 0.03, 0.04, 0.05]) * Param(seed=range(50))) min_items = 20 max_jobs = None def execute(**kwargs): return DecisionTrial().run(d=20, **kwargs)
import numpy as np from psyrun import Param from psyrun.backend import LoadBalancingBackend from psyrun.scheduler import Slurm from cue.trials import CueTrial rng = np.random.RandomState(846777) n_trials = 100 seeds = range(100) pspace = Param(seed=seeds, trial=range(n_trials), recall_duration=45., noise=0.015, min_evidence=0.025, ordinal_prob=0.1) exclude_from_result = ['cl_context'] min_items = 1 if platform.node().startswith('gra') or platform.node().startswith('cedar'): pool_size = 1 max_jobs = 100 workdir = '/scratch/jgosmann/cue' scheduler = Slurm(workdir) def timelimit(name): if 'split' in name or 'merge' in name: return '0-00:59' else:
import platform import numpy as np from psyrun import Param from psyrun.scheduler import Slurm from cue.trials import CueTrial rng = np.random.RandomState(846777) n_trials = 100 seeds = range(100) pspace = Param(seed=seeds, trial=range(n_trials), recall_duration=45., noise=0.015, min_evidence=0.04, extension='disable_stm_recall') min_items = 1 pool_size = 1 max_jobs = 100 if platform.node().startswith('gra') or platform.node().startswith('cedar'): workdir = '/scratch/jgosmann/cue' scheduler = Slurm(workdir) def timelimit(name): if 'split' in name or 'merge' in name: return '0-00:59' else: return '0-02:59'
from psyrun import Param from psyrun import LoadBalancingBackend from psyrun.scheduler import Slurm from cue.trials import CueTrial rng = np.random.RandomState(846777) n_trials = 100 seeds = range(100) pspace = Param( seed=seeds, trial=range(n_trials), noise=0.015, min_evidence=0.025, ordinal_prob=1.) exclude_from_result = ['cl_context'] min_items = 1 if platform.node().startswith('gra') or platform.node().startswith('cedar'): pool_size = 1 max_jobs = 100 workdir = '/scratch/jgosmann/cue' scheduler = Slurm(workdir) def timelimit(name): if 'split' in name or 'merge' in name: return '0-00:10' else: return '0-02:59'
import platform import numpy as np from psyrun import Param from psyrun.scheduler import Slurm, Sqsub from cue.trials.default import CueTrial rng = np.random.RandomState(846777) n_trials = 100 seeds = range(100) pspace = Param(seed=seeds, trial=range(n_trials), recall_duration=90., noise=0.015, min_evidence=0.02) min_items = 1 pool_size = 1 max_jobs = 100 if platform.node().startswith('gra') or platform.node().startswith('cedar'): workdir = '/scratch/jgosmann/cue' scheduler = Slurm(workdir) def timelimit(name): if 'split' in name or 'merge' in name: return '0-00:59' else: return '0-06:00'
import numpy as np from psyrun import Param from psyrun.backend import LoadBalancingBackend from psyrun.scheduler import Slurm from cue.trials import CueTrial rng = np.random.RandomState(846777) n_trials = 100 seeds = range(100) pspace = Param(seed=seeds, trial=range(n_trials), min_evidence=0.0325, noise=0.015, distractor_rate=0.3, ordinal_prob=0.0) exclude_from_result = ['cl_context'] min_items = 1 if platform.node().startswith('gra') or platform.node().startswith('cedar'): pool_size = 1 max_jobs = 100 workdir = '/scratch/jgosmann/cue' scheduler = Slurm(workdir) def timelimit(name): if 'split' in name or 'merge' in name: return '0-00:10' else:
import platform import sys import_path = os.path.abspath( os.path.join(os.path.dirname(__file__), '..', 'sparat')) if import_path not in sys.path: sys.path.insert(0, import_path) import numpy as np from psyrun import Param from psyrun.scheduler import Sqsub from model.benchmark import ConnectionsRatModel neurons_per_dimension = [5, 10, 15, 20, 25, 30, 35, 40, 45, 50] pspace = Param(neurons_per_dimension=neurons_per_dimension, seed=923) min_items = 1 max_splits = 100 sharcnet_nodes = ['narwhal', 'bul', 'kraken', 'saw'] if any(platform.node().startswith(x) for x in sharcnet_nodes): workdir = '/work/jgosmann/rat' scheduler = Sqsub(workdir) scheduler_args = { 'timelimit': '60m', 'memory': '6G', } def execute(**kwargs):
import numpy as np import pyopencl from psyrun import Param from psyrun.backend import LoadBalancingBackend from cue.trials import MixedSelTrial rng = np.random.RandomState(9) n_sequences = 6 seeds = [230] * n_sequences pspace = Param(seed=seeds, sequence=range(n_sequences)) backend = LoadBalancingBackend exclude_from_result = ['cl_context'] pool_size = 4 max_jobs = 1 def setup(proc_id): context = pyopencl.create_some_context(answers=[1, proc_id]) return {'cl_context': context} def execute(**kwargs): result = MixedSelTrial().run(progress=False, **kwargs) return result