def main():
parser = argparse.ArgumentParser(description='Process input parameters for RJMC 2CLJQ')
parser.add_argument('--compound', '-c',
type=str,
help='Compound to simulate parameters for',
required=True)
parser.add_argument('--trange', '-t',
type=list,
help='Temperature range for data to include',
required=False)
parser.add_argument('--steps', '-s',
type=int,
help='Number of MCMC steps',
required=True)
parser.add_argument('--properties', '-p',
type=str,
help='Properties to include in computation',
required=True)
parser.add_argument('--priors', '-r',
type=dict,
help='Values and types of prior distribution',
required=False)
parser.add_argument('--optimum_matching', '-o',
type=list,
help='Whether to use optimum matching in transitions',
required=False)
parser.add_argument('--number_data_points', '-d',
type=int,
help='Number of data points to include',
required=False)
parser.add_argument('--swap_freq', '-f',
type=float,
help='Frequency of model jump proposal',
required=False)
parser.add_argument('--biasing_factor', '-b',
type=str,
help='Biasing factors for each model',
required=False)
parser.add_argument('--label', '-l',
type=str,
help='label for files',
required=False)
parser.add_argument('--save_traj', '-j',
type=bool,
help='Whether to save trajectory files',
required=False)
T_range = [0.55, 0.95]
n_points = 10
swap_freq = 0.1
biasing_factor = [0, 0, 0]
optimum_matching = ['True', 'True']
tag = ''
save_traj = False
prior_values = {
'epsilon': ['exponential', [0,400]],
'sigma': ['exponential', [0,5]],
'L': ['exponential', [0,3]],
'Q': ['exponential', [0,1]]}
args = parser.parse_args()
print(args.compound)
print(args.biasing_factor)
compound = args.compound
# T_range=args.trange
steps = args.steps
properties = args.properties
if args.trange is not None:
T_range = args.trange
if args.number_data_points is not None:
n_points = args.number_data_points
if args.swap_freq is not None:
swap_freq = args.swap_freq
if args.biasing_factor is not None:
biasing_factor_raw = list(args.biasing_factor.split(','))
biasing_factor = [float(i) for i in biasing_factor_raw]
if args.optimum_matching is not None:
optimum_matching = args.optimum_matching
if args.priors is not None:
prior_values = args.priors
if args.label is not None:
tag = args.label
if args.save_traj is not None:
save_traj = args.save_traj
prior = RJMC_Prior(prior_values)
prior.epsilon_prior()
prior.sigma_prior()
prior.L_prior()
prior.Q_prior()
mcmc_prior_simulation = RJMC_Simulation(compound,
T_range,
properties,
n_points,
5 * 10**5,
0.0,
biasing_factor,
optimum_matching)
mcmc_prior_simulation.prepare_data()
compound_2CLJ = LennardJones_2C(mcmc_prior_simulation.M_w)
mcmc_prior_simulation.gen_Tmatrix(prior, compound_2CLJ)
mcmc_prior_simulation.set_initial_state(prior,
compound_2CLJ,
initial_model='AUA+Q')
mcmc_prior_simulation.RJMC_Outerloop(prior, compound_2CLJ)
mcmc_prior_simulation.Report()
prior_values['Q'] = mcmc_prior_simulation.refit_prior('exponential')
print(prior_values['Q'])
#plt.hist(mcmc_prior_simulation.trace_model_1[:,4],bins=50,density=True)
#plt.plot(np.linspace(0,1,num=500),gengamma.pdf(np.linspace(0,1,num=500),*prior_values['Q'][1]))
#plt.plot(np.linspace(0,1,num=500),expon.pdf(np.linspace(0,1,num=500),0,400))
#plt.show()
print('Refitting Prior for Q')
prior = RJMC_Prior(prior_values)
prior.epsilon_prior()
prior.sigma_prior()
prior.L_prior()
prior.Q_prior()
print(prior.Q_prior_values)
'''
plt.hist(mcmc_prior_simulation.trace_model_1[:,4],bins=50,density=True)
plt.plot(np.linspace(0,1,num=500),expon.pdf(np.linspace(0,1,num=500),*prior.Q_prior_values))
plt.show()
'''
aua_path = 'output/C2H2/rhol+Psat/C2H2_rhol+Psat_2000000_aua_only_2019-10-25/trace/trace.npy'
auaq_path = 'output/C2H2/rhol+Psat/C2H2_rhol+Psat_2000000_auaq_only_2019-10-25/trace/trace.npy'
aua_max,auaq_max = create_map(aua_path,auaq_path)
rjmc_simulator = RJMC_Simulation(compound,
T_range,
properties,
n_points,
steps,
swap_freq,
biasing_factor,
optimum_matching)
rjmc_simulator.prepare_data()
print('Simulation Attributes:', rjmc_simulator.get_attributes())
compound_2CLJ = LennardJones_2C(rjmc_simulator.M_w)
rjmc_simulator.gen_Tmatrix(prior, compound_2CLJ)
rjmc_simulator.set_initial_state(prior, compound_2CLJ)
custom_map = list([list(aua_max),list(auaq_max),list(rjmc_simulator.opt_params_UA)])
rjmc_simulator.opt_params_AUA,rjmc_simulator.opt_params_AUA_Q,rjmc_simulator.opt_params_UA = rjmc_simulator.load_custom_map(custom_map)
rjmc_simulator.RJMC_Outerloop(prior, compound_2CLJ)
trace, logp_trace, percent_dev_trace,BAR_trace = rjmc_simulator.Report(USE_BAR=True)
rjmc_simulator.write_output(prior_values, tag=tag, save_traj=save_traj)
print('Finished!')
def refit_prior(simulation_params):
prior = RJMC_Prior(simulation_params['priors'])
prior.epsilon_prior()
prior.sigma_prior()
prior.L_prior()
prior.Q_prior()
print('Approximating AUA+Q posterior distribution')
mcmc_prior_simulation = RJMC_Simulation(
simulation_params['compound'], simulation_params['trange'],
simulation_params['properties'],
simulation_params['number_data_points'],
simulation_params['single_simulation_length'], 0.0, [0, 0, 0],
simulation_params['optimum_matching'])
mcmc_prior_simulation.prepare_data()
compound_2CLJ = LennardJones_2C(mcmc_prior_simulation.M_w)
mcmc_prior_simulation.optimum_bounds = simulation_params['opt_bounds']
mcmc_prior_simulation.gen_Tmatrix(prior, compound_2CLJ)
mcmc_prior_simulation.set_initial_state(prior,
compound_2CLJ,
initial_model='AUA+Q')
mcmc_prior_simulation.RJMC_Outerloop(prior, compound_2CLJ)
mcmc_prior_simulation.Report()
simulation_params['priors']['Q'] = mcmc_prior_simulation.refit_prior(
simulation_params['refit_prior_to'])
print('Refitting Q prior')
print(simulation_params['priors']['Q'])
prior = RJMC_Prior(simulation_params['priors'])
prior.epsilon_prior()
prior.sigma_prior()
prior.L_prior()
prior.Q_prior()
if simulation_params['use_MAP'] is True:
aua_max, auaq_max = create_map(simulation_params['MAP_simulations'][0],
simulation_params['MAP_simulations'][1])
rjmc_simulator = RJMC_Simulation(
simulation_params['compound'], simulation_params['trange'],
simulation_params['properties'],
simulation_params['number_data_points'], simulation_params['steps'],
simulation_params['swap_freq'], simulation_params['biasing_factor'],
simulation_params['optimum_matching'])
rjmc_simulator.prepare_data()
print('Simulation Attributes:', rjmc_simulator.get_attributes())
compound_2CLJ = LennardJones_2C(rjmc_simulator.M_w)
rjmc_simulator.optimum_bounds = simulation_params['opt_bounds']
rjmc_simulator.gen_Tmatrix(prior, compound_2CLJ)
rjmc_simulator.set_initial_state(prior, compound_2CLJ)
if simulation_params['use_MAP'] is True:
custom_map = list([
list(aua_max),
list(auaq_max),
list(rjmc_simulator.opt_params_UA)
])
rjmc_simulator.opt_params_AUA, rjmc_simulator.opt_params_AUA_Q, rjmc_simulator.opt_params_UA = rjmc_simulator.load_custom_map(
custom_map)
rjmc_simulator.RJMC_Outerloop(prior, compound_2CLJ)
trace, logp_trace, percent_dev_trace, BAR_trace = rjmc_simulator.Report(
USE_BAR=simulation_params['USE_BAR'])
rjmc_simulator.write_output(simulation_params['priors'],
tag=simulation_params['label'],
save_traj=simulation_params['save_traj'])
path = 'output/' + simulation_params['compound'] + '/' + simulation_params['properties'] + '/' + simulation_params['compound'] + \
'_' + simulation_params['properties'] + '_' + str(simulation_params['steps']) + '_' + simulation_params['label'] + '_' + str(date.today()) +'/runfile.yaml'
with open(path, 'w') as outfile:
yaml.dump(simulation_params, outfile, default_flow_style=False)