def main():
parser = buildArgsParser()
args = parser.parse_args()
sort_by = args.sort
names = []
results_files = []
hyperparams_files = []
status_files = []
for f in args.results:
exp_folder = f
if os.path.isfile(f):
exp_folder = os.path.dirname(f)
result_file = pjoin(exp_folder, "result.json")
hyperparams_file = pjoin(exp_folder, "hyperparams.json")
status_file = pjoin(exp_folder, "status.json")
if not os.path.isfile(result_file):
print 'Skip: {0} is not a file!'.format(result_file)
continue
if not os.path.isfile(hyperparams_file):
print 'Skip: {0} is not a file!'.format(hyperparams_file)
continue
if not os.path.isfile(status_file):
print 'Skip: {0} is not a file!'.format(status_file)
continue
name = os.path.basename(exp_folder)
if 'hyperparams.json' in os.listdir(os.path.abspath(pjoin(exp_folder, os.path.pardir))):
name = os.path.basename(os.path.abspath(pjoin(exp_folder, os.path.pardir)))
names.append(name)
results_files.append(result_file)
hyperparams_files.append(hyperparams_file)
status_files.append(status_file)
if len([no for no in sort_by if no == 0]) > 0:
parser.error('Column ID are starting at 1!')
# Retrieve headers from hyperparams
headers_hyperparams = set()
headers_results = set()
headers_status = set()
for hyperparams_file, status_file, results_file in zip(hyperparams_files, status_files, results_files):
hyperparams = load_dict_from_json_file(hyperparams_file)
results = load_dict_from_json_file(results_file)
status = load_dict_from_json_file(status_file)
headers_hyperparams |= set(hyperparams.keys())
headers_results |= set(results.keys())
headers_status |= set(status.keys())
headers_hyperparams = sorted(list(headers_hyperparams))
headers_status = sorted(list(headers_status))
# TODO: when generating result.json split 'trainset' scores in two key:
# 'trainset' and 'trainset_std' (same goes for validset and testset).
headers_results |= set(["trainset_std", "validset_std", "testset_std"])
headers_results = sorted(list(headers_results))
headers = headers_hyperparams + headers_status + ["name"] + headers_results
# Build results table
table = Texttable(max_width=0)
table.set_deco(Texttable.HEADER)
table.set_precision(8)
table.set_cols_dtype(['a'] * len(headers))
table.set_cols_align(['c'] * len(headers))
# Headers
table.header([str(i) + "\n" + h for i, h in enumerate(headers, start=1)])
if args.only_header:
print table.draw()
return
# Results
for name, hyperparams_file, status_file, results_file in zip(names, hyperparams_files, status_files, results_files):
hyperparams = load_dict_from_json_file(hyperparams_file)
results = load_dict_from_json_file(results_file)
status = load_dict_from_json_file(status_file)
# Build results table row (hyperparams columns)
row = []
for h in headers_hyperparams:
value = hyperparams.get(h, '')
row.append(value)
for h in headers_status:
value = status.get(h, '')
row.append(value)
row.append(name)
for h in headers_results:
if h in ["trainset", "validset", "testset"]:
value = results.get(h, '')[0]
elif h in ["trainset_std", "validset_std", "testset_std"]:
value = results.get(h[:-4], '')[1]
else:
value = results.get(h, '')
row.append(value)
table.add_row(row)
# Sort
for col in reversed(sort_by):
table._rows = sorted(table._rows, key=sort_nicely(abs(col) - 1), reverse=col < 0)
if args.out is not None:
import csv
results = []
results.append(headers)
results.extend(table._rows)
with open(args.out, 'wb') as csvfile:
w = csv.writer(csvfile)
w.writerows(results)
else:
print table.draw()
def main():
parser = buildArgsParser()
args = parser.parse_args()
scoring_files = args.scores
sort_by = args.sort_by
if len(scoring_files) == 1 and os.path.isdir(scoring_files[0]):
scoring_files = [os.path.join(scoring_files[0], f)
for f in os.listdir(scoring_files[0])]
scoring_files = filter(lambda f: f.endswith('.json'), scoring_files)
if len(scoring_files) == 0:
parser.error('Need a least one scoring file/folder!')
if len([no for no in sort_by if no == 0]) > 0:
parser.error('Column ID are starting at 1.')
headers_params = set()
headers_scores = set()
params = []
scores = []
# Retrieves scores
for scoring_file in scoring_files:
infos = scoring_file.split('/')[-1][:-5]
if args.is_split:
ind = 0 if infos.find('-') < infos.find('_') else 1
param = dict([tuple(param.split('-'))
for param in infos.split('_')[ind:]])
else:
param = {"filename": infos}
# score = pickle.load(open(scoring_file))
score = json.load(open(scoring_file))
# Compute the VCCR metric, CSR=1-NC
# In VCCR, VCWP are considered as IC [Girard et al., NeuroImage, 2014]
if score['VC'] > 0:
score['VCCR'] = score['VC'] / (score['VC'] +
score['IC'] +
score['VCWP'])
else:
score['VCCR'] = 0
# Keep only scalar metrics.
for k in score.keys():
if k not in METRICS:
del score[k]
headers_params |= set(param.keys())
headers_scores |= set(score.keys())
scores.append(score)
params.append(param)
nbr_cols = len(headers_params) + len(headers_scores)
if len([no for no in sort_by if abs(no) > nbr_cols]) > 0:
parser.error('The maximum column ID is {0}.'.format(nbr_cols))
table = Texttable(max_width=0)
table.set_deco(Texttable.HEADER)
table.set_cols_dtype(['a'] * nbr_cols)
table.set_cols_align(['l'] + ['c'] * (nbr_cols-1))
# Headers
headers_params = sorted(headers_params)
headers_scores = list(headers_scores)
headers_scores = [headers_scores[headers_scores.index(e)] for e in METRICS if e in headers_scores]
headers = headers_params + headers_scores
table.header([str(i) + "\n" + h for i, h in enumerate(headers, start=1)])
# Data
for param, score in zip(params, scores):
data = []
for header in headers_params:
data.append(param.get(header, '-'))
for header in headers_scores:
data.append(score.get(header, '-'))
table.add_row(data)
# Sort
for col in reversed(sort_by):
table._rows = sorted(table._rows,
key=sort_nicely(abs(col) - 1),
reverse=col < 0)
print(table.draw())
def main():
parser = buildArgsParser()
args = parser.parse_args()
scoring_files = args.scores
sort_by = args.sort_by
if len(scoring_files) == 1 and os.path.isdir(scoring_files[0]):
scoring_files = [
os.path.join(scoring_files[0], f)
for f in os.listdir(scoring_files[0])
]
scoring_files = filter(lambda f: f.endswith('.json'), scoring_files)
if len(scoring_files) == 0:
parser.error('Need a least one scoring file/folder!')
if len([no for no in sort_by if no == 0]) > 0:
parser.error('Column ID are starting at 1.')
headers_params = set()
headers_scores = set()
params = []
scores = []
# Retrieves scores
for scoring_file in scoring_files:
infos = scoring_file.split('/')[-1][:-5]
if args.is_split:
ind = 0 if infos.find('-') < infos.find('_') else 1
param = dict(
[tuple(param.split('-')) for param in infos.split('_')[ind:]])
else:
param = {"filename": infos[:60]}
# score = pickle.load(open(scoring_file))
score = json.load(open(scoring_file))
# Compute the VCCR metric, CSR=1-NC
# In VCCR, VCWP are considered as IC [Girard et al., NeuroImage, 2014]
if score['VC'] > 0:
score['VCCR'] = score['VC'] / (score['VC'] + score['IC'] +
score['VCWP'])
else:
score['VCCR'] = 0
# Keep only scalar metrics.
for k in score.keys():
if k not in METRICS:
del score[k]
headers_params |= set(param.keys())
headers_scores |= set(score.keys())
scores.append(score)
params.append(param)
nbr_cols = len(headers_params) + len(headers_scores)
if len([no for no in sort_by if abs(no) > nbr_cols]) > 0:
parser.error('The maximum column ID is {0}.'.format(nbr_cols))
table = Texttable(max_width=0)
table.set_deco(Texttable.HEADER)
table.set_cols_dtype(['a'] * nbr_cols)
table.set_cols_align(['l'] + ['c'] * (nbr_cols - 1))
# Headers
headers_params = sorted(headers_params)
headers_scores = list(headers_scores)
headers_scores = [
headers_scores[headers_scores.index(e)] for e in METRICS
if e in headers_scores
]
headers = headers_params + headers_scores
table.header([str(i) + "\n" + h for i, h in enumerate(headers, start=1)])
# Data
for param, score in zip(params, scores):
data = []
for header in headers_params:
data.append(param.get(header, '-'))
for header in headers_scores:
data.append(score.get(header, '-'))
table.add_row(data)
# Sort
for col in reversed(sort_by):
table._rows = sorted(table._rows,
key=sort_nicely(abs(col) - 1),
reverse=col < 0)
print(table.draw())
def main():
parser = buildArgsParser()
args = parser.parse_args()
sort_by = args.sort
names = []
results_files = []
hyperparams_files = []
status_files = []
for f in args.results:
exp_folder = f
if os.path.isfile(f):
exp_folder = os.path.dirname(f)
result_file = pjoin(exp_folder, "result.json")
hyperparams_file = pjoin(exp_folder, "hyperparams.json")
status_file = pjoin(exp_folder, "status.json")
if not os.path.isfile(result_file):
print 'Skip: {0} is not a file!'.format(result_file)
continue
if not os.path.isfile(hyperparams_file):
print 'Skip: {0} is not a file!'.format(hyperparams_file)
continue
if not os.path.isfile(status_file):
print 'Skip: {0} is not a file!'.format(status_file)
continue
name = os.path.basename(exp_folder)
if 'hyperparams.json' in os.listdir(
os.path.abspath(pjoin(exp_folder, os.path.pardir))):
name = os.path.basename(
os.path.abspath(pjoin(exp_folder, os.path.pardir)))
names.append(name)
results_files.append(result_file)
hyperparams_files.append(hyperparams_file)
status_files.append(status_file)
if len([no for no in sort_by if no == 0]) > 0:
parser.error('Column ID are starting at 1!')
# Retrieve headers from hyperparams
headers_hyperparams = set()
headers_results = set()
headers_status = set()
for hyperparams_file, status_file, results_file in zip(
hyperparams_files, status_files, results_files):
hyperparams = load_dict_from_json_file(hyperparams_file)
results = load_dict_from_json_file(results_file)
status = load_dict_from_json_file(status_file)
headers_hyperparams |= set(hyperparams.keys())
headers_results |= set(results.keys())
headers_status |= set(status.keys())
headers_hyperparams = sorted(list(headers_hyperparams))
headers_status = sorted(list(headers_status))
# TODO: when generating result.json split 'trainset' scores in two key:
# 'trainset' and 'trainset_std' (same goes for validset and testset).
headers_results |= set(["trainset_std", "validset_std", "testset_std"])
headers_results = sorted(list(headers_results))
headers = headers_hyperparams + headers_status + ["name"] + headers_results
# Build results table
table = Texttable(max_width=0)
table.set_deco(Texttable.HEADER)
table.set_precision(8)
table.set_cols_dtype(['a'] * len(headers))
table.set_cols_align(['c'] * len(headers))
# Headers
table.header([str(i) + "\n" + h for i, h in enumerate(headers, start=1)])
if args.only_header:
print table.draw()
return
# Results
for name, hyperparams_file, status_file, results_file in zip(
names, hyperparams_files, status_files, results_files):
hyperparams = load_dict_from_json_file(hyperparams_file)
results = load_dict_from_json_file(results_file)
status = load_dict_from_json_file(status_file)
# Build results table row (hyperparams columns)
row = []
for h in headers_hyperparams:
value = hyperparams.get(h, '')
row.append(value)
for h in headers_status:
value = status.get(h, '')
row.append(value)
row.append(name)
for h in headers_results:
if h in ["trainset", "validset", "testset"]:
value = results.get(h, '')[0]
elif h in ["trainset_std", "validset_std", "testset_std"]:
value = results.get(h[:-4], '')[1]
else:
value = results.get(h, '')
row.append(value)
table.add_row(row)
# Sort
for col in reversed(sort_by):
table._rows = sorted(table._rows,
key=sort_nicely(abs(col) - 1),
reverse=col < 0)
if args.out is not None:
import csv
results = []
results.append(headers)
results.extend(table._rows)
with open(args.out, 'wb') as csvfile:
w = csv.writer(csvfile)
w.writerows(results)
else:
print table.draw()