def plot_one_best_func(list_res, path = ['fun'], criterion = np.argmin):
run_fun = [ut.extract_from_nested(r, path) for r in list_res]
best_arg = criterion(run_fun)
best_run = list_res[best_arg]
best_func = pFunc_base.pFunc_base.build_pfunc(ut.extract_from_nested(best_run, ['func']))
best_T = ut.extract_from_nested(best_run, ['config', 'testing_dico', 'T'])
best_tt = np.linspace(-0.1, best_T + 0.1, 10000)
best_func.plot_function(best_tt)
def collect_res_bug(cls,
key_path=[],
nameFile=None,
allPrefix='res_',
folderName=None):
"""Extract results stored in (a) txt file(s) and group them according to
some key values (where key_path provides the path in the potentially
nested structure of the results to find the key(s))
Output:
a dictionary where key is the concatenation of the unique set of
keys found and value is the res is a list of all res matching this key
"""
listRes = cls.read_res_bug(nameFile, allPrefix, folderName)
res_keys = [
tuple([ut.extract_from_nested(res, k) for k in key_path])
for res in listRes
]
res_keys_unique = list(set(res_keys))
res = {
ut.concat2String(*k_u):
[listRes[n] for n, r in enumerate(res_keys) if r == k_u]
for k_u in res_keys_unique
}
return res
def collect_res(cls, key_path = [], nameFile = None, allPrefix = 'res_',
folderName = None, replace_func = None):
""" collect results and group them according to some key values
Arguments
---------
key_path: <list>
if not empty defines how to group results: provides a path in the
dict structure of the results to get a value used for the grouping
nameFile, allPrefix, folderName: cf. read_res()
Output:
-------
a dictionary where key is the concatenation of the unique set of
keys found and value is the res is a list of all res matching this key
"""
listRes, listNames = cls.read_res(nameFile, allPrefix, folderName,
returnName=True, replace_func=replace_func)
if(len(key_path) == 0):
res = {k:v for k,v in zip(listNames, listRes)}
else:
res_keys = [tuple([ut.extract_from_nested(res, k) for k in key_path])
for res in listRes]
res_keys_unique = list(set(res_keys))
res = {ut.concat2String(*k_u):[listRes[n] for n, r in enumerate(res_keys)
if r == k_u] for k_u in res_keys_unique}
return res
def get_keys_from_onefile(cls, name_file, path_keys, replace_func = None):
""" Get concatenated keys from one res (passed as the name of the file)"""
res = ut.eval_from_file(name_file, evfunc = pFunc_base.eval_with_pFunc,
replace_func = replace_func)
res_keys = "_".join([ut.extract_from_nested(res, k) for k in path_keys])
return res_keys
def get_keys_from_onefile_safe(cls, name_file, path_keys, replace_func = None):
""" Get concatenated keys from one res (passed as the name of the file)"""
try:
res = ut.eval_from_file(name_file, evfunc = pFunc_base.eval_with_pFunc,
replace_func = replace_func)
res_keys = "_".join([ut.extract_from_nested(res, k) for k in path_keys])
except:
print("Error reading file {}".format(name_file))
res_keys = None
return res_keys
def _gen_name_res(cls, config, metadico = {}, type_output = '.txt'):
""" Generate a name associated to a config
Rules
-----
if _OUT_NAME is not None and has a <dic> type:
{'k':v} --> 'k_XXX' where XXX has been found in the config structure
following path given by v
elif _OUT_COUNTER is not None increment this counter for each new config
add _OUT_PREFIX
"""
res_name = ''
if(metadico.get('_OUT_NAME') is not None):
name_rules = metadico['_OUT_NAME']
if(ut.is_dico(name_rules)):
for k, v in name_rules.items():
res_name += (k + "_" + str(ut.extract_from_nested(config, v)))
else:
raise NotImplementedError()
if((config.get("_RDM_RUN") is not None)):
res_name += "_"
res_name += str(config["_RDM_RUN"])
elif(metadico.get('_OUT_COUNTER') is not None):
res_name += str(metadico['_OUT_COUNTER'])
metadico['_OUT_COUNTER'] +=1
if(res_name == ''):
res_name = str(RandomGenerator.RandomGenerator.gen_seed())
prefix = metadico.get('_OUT_PREFIX', '')
res_name = prefix + res_name + type_output
return res_name, metadico
# 3T: NM equiv to BO // Is it the same for pcw5 NO maybe because params nb increase=>less local minimas
# // Use Exploit strategy sooner
#
#
### CUSTOM METHODS ARE USED BECAUSE OF A BUG IN RES FILE
#==============================================================================
# ***GETDATA*** Linear benchmark
#==============================================================================
name_batch = 'bch_linear'
name_folder = name_batch + '/' + name_batch
key_T = ['config', 'testing_dico', 'T']
bch_res = learner.collect_res([key_T], None, 'benchmark', name_folder)
bch_res_fom = {k: ut.extract_from_nested(r, [0,'test_fom']) for k, r in bch_res.items()}
#==============================================================================
# ***GETDATA***
#==============================================================================
name_batch = 'crab_customs'
name_folder = name_batch + '/' + name_batch
one_res = learner.eval_from_onefile(name_folder +'/res120.txt')
names_keys_res = list(one_res.keys())
ut.print_nested_keys(one_res, ' +')
key1 = ['config', 'optim_dico', '_FLAG_NAME']
key2 = ['config', 'model_dico', '_FLAG_NAME']
rawres = learner.collect_res([key1, key2], None, 'res', name_folder)
summary = learner.collect_and_process_res([key1, key2], None, 'res', name_folder, True, False)
