def get_sweep_results(sim_meta_config_files, calib_file_path, tags_data_file_path):
# Login to COMPs
comps_login()
# find total number of simulations across given experiment files
num_sims = 0
for sim_meta_config_file in sim_meta_config_files:
with open(sim_meta_config_file) as metadata_file:
metadata = json.loads(metadata_file.read())
num_sims = num_sims + len(metadata['sims'])
# Download simulations locally
# sample sim meta config file (like "C:\\Users\\Mnikolov\\Zambia-raw\\dtk-scripts\\1node\\simulations\\Sinamalina_Sinazongwe_Calibration_e9979059-33f8-e411-93f9-f0921c16b9e7.json")
#print 'Downloading simulations from experiment ' + str(sim_meta_config_files) + '...'
# simulations tag data structure: accumulates sims meta information from sims tags
tag_data = {
'ITN trajectory': [],\
'Drug coverage per round': [],\
'Temporary habitat scale': [],\
'Constant habitat scale': []\
}
# iterate through experiments
calib_output = {}
# count processed sims to updated progress
count = 0
for sim_meta_config_file in sim_meta_config_files:
# construct experiment directory structure
with open(sim_meta_config_file) as metadata_file:
metadata = json.loads(metadata_file.read())
output_path = metadata['sim_root']
exp_id = metadata['exp_id']
exp_name = metadata['exp_name']
sim_dir_map = CompsDTKOutputParser.createSimDirectoryMap(exp_id)
# get all successfully completed sims in experiment
for sim_id, sim in metadata['sims'].items():
# get path to the sim timeseries channels data
timeseries_path = os.path.join(sim_dir_map[sim_id],'output', 'InsetChart.json')
#get sim timeseries channels data; json2dict returns None if timeseries_path points to non-existing file, which is the case if the sim has not successfully finished
sim_output = json2dict(timeseries_path)
# only download successfully completed simulations
if sim_output == None:
continue
# delete all but the specified channels
for channel in sim_output['Channels'].keys():
if not channel in channels:
del(sim_output['Channels'][channel])
# process specified reports
report_channels_data = {}
if not reports_channels == None:
report_channels_data = process_reports(reports_channels, sim_dir_map, sim_id)
# record sim meta information including sim tags
tags_path = os.path.join(sim_dir_map[sim_id], 'tags.json')
f = open(tags_path, 'r')
tags = f.read()
sim_meta = ast.literal_eval(tags)
append_tag_data(sim_meta, tag_data)
# construct sim group key and sim key
x_temp_h = sim_meta_2_temp_h(sim_meta)
const_h = sim_meta_2_const_h(sim_meta)
itn_level = sim_meta_2_itn_level(sim_meta)
drug_coverage_level = sim_meta_2_drug_cov(sim_meta)
sim_key = get_sim_key(x_temp_h, const_h, itn_level, drug_coverage_level)
sim_group_key = get_sim_group_key(itn_level, drug_coverage_level)
# store sim channels data
if sim_group_key not in calib_output:
calib_output[sim_group_key] = {}
calib_output[sim_group_key][sim_key] = {
# can add/remove data entries depending on needs
'prevalence': sim_output['Channels']['New Diagnostic Prevalence']['Data'],
'reinfections': report_channels_data['reinfections'],
'meta':sim_meta,
'sim_id':sim_id
}
'''
count = count + 1
percent_complete = 100*count/(num_sims+0.0)
sys.stdout.write('\r')
sys.stdout.write('%2f %%' % percent_complete)
#sys.stdout.write('%d' % count)
sys.stdout.flush()
'''
print ""
print "Writing files..."
with open(calib_file_path, 'w') as calib_f:
json.dump(calib_output, calib_f)
print str(len(calib_output)) + ' simulation results saved to ' + calib_file_path
with open(tags_data_file_path, 'w') as tags_f:
json.dump(tag_data, tags_f)
print 'Meta data tags saved to ' + tags_data_file_path
print ""
return calib_f
def fit_entry(self):
model_meta = self.model.get_meta()
sim_key = model_meta['sim_key']
temp_h = sim_meta_2_temp_h(model_meta['sim_meta'])
const_h = sim_meta_2_const_h(model_meta['sim_meta'])
itn_level = sim_meta_2_itn_level(model_meta['sim_meta'])
drug_cov = sim_meta_2_drug_cov(model_meta['sim_meta'])
if self.all_fits:
fit_terms = self.all_fits[self.cluster_id][sim_key]['fit_terms']
else:
fit_terms = {}
if not 'cc_penalty' in fit_terms:
fit_terms['cc_penalty'] = {}
if 'ls_norm' in cc_penalty_model:
fit_terms['cc_penalty']['ls_norm'] = self.clinical_cases_penalty_term
elif 'ls_norm_not_folded' in cc_penalty_model:
fit_terms['cc_penalty']['ls_norm_not_folded'] = self.clinical_cases_penalty_term
if 'ls_no_norm' in cc_penalty_model:
fit_terms['cc_penalty']['ls_no_norm'] = self.clinical_cases_penalty_term
if 'corr_folded' in cc_penalty_model:
if not 'corr_folded' in fit_terms['cc_penalty']:
fit_terms['cc_penalty']['corr_folded'] = {}
fit_terms['cc_penalty']['corr_folded']['penalty'] = self.clinical_cases_penalty_term
fit_terms['cc_penalty']['corr_folded']['rho'] = self.get_rho()
fit_terms['cc_penalty']['corr_folded']['p_val'] = self.get_p_val()
if 'corr_not_folded' in cc_penalty_model:
if not 'corr_not_folded' in fit_terms['cc_penalty']:
fit_terms['cc_penalty']['corr_not_folded'] = {}
fit_terms['cc_penalty']['corr_not_folded']['penalty'] = self.clinical_cases_penalty_term
fit_terms['cc_penalty']['corr_not_folded']['rho'] = self.get_rho()
fit_terms['cc_penalty']['corr_not_folded']['p_val'] = self.get_p_val()
fit_terms['reinf_penalty'] = self.reinfection_penalty_term
if load_prevalence_mse:
fit_terms['mse'] = self.get_cached_mse_term()
else:
fit_terms['mse'] = self.get_mse()
fit_entry = {}
fit_entry[model_meta['sim_key']] = {
'group_key': model_meta['group_key'],
'sim_id':model_meta['sim_id'],
'fit_val': self.get_fit_val(),
'rho_val' : self.get_rho(), # from most recent run
'p_val' : self.get_p_val(), # from most recent run
'x_temp_h': temp_h,
'const_h': const_h,
'fit_terms':fit_terms,
'itn_level': itn_level,
'drug_cov': drug_cov
}
return fit_entry