def main(filenames, write, **kwargs):
uploader = TaskUploader(**kwargs)
downloader = TaskDownloader(**kwargs)
scores = []
try:
for pattern in filenames:
filenames = glob(pattern)
if not filenames:
raise IOError("Not found: " + pattern)
for ref in read_files_and_dirs(filenames):
print("Converting passage " + ref.ID + "... ", end="")
task = uploader.upload_task(ref)
guessed = downloader.download_task(task["id"],
write=write,
**kwargs)
score = evaluate(guessed, ref, **kwargs)
print("F1=%.3f" % score.average_f1())
scores.append(score)
except HTTPError as e:
try:
raise ValueError(e.response.json()) from e
except JSONDecodeError:
raise ValueError(e.response.text) from e
print()
if len(scores) > 1:
print("Aggregated scores:")
Scores.aggregate(scores).print()
def get_metric(self, reset: bool = False):
metrics = {}
score_list = [
item for sublist in self.scores.values() for item in sublist
]
agg_score = Scores.aggregate(score_list)
labeled_average_f1 = agg_score.average_f1(LABELED)
unlabeled_average_f1 = agg_score.average_f1(UNLABELED)
metrics[f'labeled_average_F1'] = labeled_average_f1
metrics[f'unlabeled_average_F1'] = unlabeled_average_f1
for dataset_label in self.scores:
dataset_prefix = f'{dataset_label}_' # if len(self.scores.keys()) > 1 else ""
agg_score = Scores.aggregate(self.scores[dataset_label])
labeled_average_f1 = agg_score.average_f1(LABELED)
unlabeled_average_f1 = agg_score.average_f1(UNLABELED)
metrics[f'{dataset_prefix}labeled_average_F1'] = labeled_average_f1
metrics[
f'{dataset_prefix}unlabeled_average_F1'] = unlabeled_average_f1
titles = agg_score.titles()
values = agg_score.fields()
for title, value in zip(titles, values):
metrics[f'{dataset_prefix}{title}'] = float(value)
if reset:
self.reset()
return metrics
def main(filenames, write, **kwargs):
uploader = TaskUploader(**kwargs)
downloader = TaskDownloader(**kwargs)
scores = []
try:
for pattern in filenames:
filenames = glob(pattern)
if not filenames:
raise IOError("Not found: " + pattern)
for ref in read_files_and_dirs(filenames):
print("Converting passage " + ref.ID + "... ", end="")
task = uploader.upload_task(ref)
guessed = downloader.download_task(task["id"], write=write, **kwargs)
score = evaluate(guessed, ref, **kwargs)
print("F1=%.3f" % score.average_f1())
scores.append(score)
except HTTPError as e:
try:
raise ValueError(e.response.json()) from e
except JSONDecodeError:
raise ValueError(e.response.text) from e
print()
if len(scores) > 1:
print("Aggregated scores:")
Scores.aggregate(scores).print()
def main():
argparser = argparse.ArgumentParser(description=desc)
argparser.add_argument("filenames", nargs="+",
help="file names to convert and evaluate")
argparser.add_argument("-f", "--format", required=True, choices=convert.CONVERTERS,
help="input file format")
argparser.add_argument("-T", "--tree", action="store_true",
help="remove multiple parents to get a tree")
args = argparser.parse_args()
converter1 = convert.TO_FORMAT[args.format]
converter2 = convert.FROM_FORMAT[args.format]
scores = []
for pattern in args.filenames:
filenames = glob.glob(pattern)
if not filenames:
raise IOError("Not found: " + pattern)
for filename in filenames:
ref = file2passage(filename)
try:
guessed = next(converter2(converter1(ref, tree=args.tree), ref.ID))
scores.append(evaluate(guessed, ref, fscore=True, verbose=False,
units=False, errors=False))
except Exception as e:
raise ValueError("Error evaluating conversion of %s" % filename, e)
if len(scores) > 1:
print("Aggregated scores:")
Scores.aggregate(scores).print()
sys.exit(0)
def main():
argparser = argparse.ArgumentParser(description=desc)
argparser.add_argument("filenames",
nargs="+",
help="file names to convert and evaluate")
argparser.add_argument("-f",
"--format",
required=True,
choices=convert.CONVERTERS,
help="input file format")
argparser.add_argument("-T",
"--tree",
action="store_true",
help="remove multiple parents to get a tree")
argparser.add_argument(
"-s",
"--strict",
action="store_true",
help="stop immediately if failed to convert or evaluate a file")
argparser.add_argument(
"-v",
"--verbose",
action="store_true",
help="print evaluation results for each file separately")
args = argparser.parse_args()
converter1 = convert.TO_FORMAT[args.format]
converter2 = convert.FROM_FORMAT[args.format]
scores = []
for pattern in args.filenames:
filenames = glob.glob(pattern)
if not filenames:
raise IOError("Not found: " + pattern)
for filename in filenames:
sys.stdout.write("\rConverting %s" % filename)
sys.stdout.flush()
ref = file2passage(filename)
try:
guessed = next(
converter2(converter1(ref, tree=args.tree), ref.ID))
scores.append(evaluate(guessed, ref, verbose=args.verbose))
except Exception as e:
if args.strict:
raise ValueError("Error evaluating conversion of %s" %
filename) from e
else:
print("Error evaluating conversion of %s: %s" %
(filename, e),
file=sys.stderr)
print()
if args.verbose and len(scores) > 1:
print("Aggregated scores:")
Scores.aggregate(scores).print()
sys.exit(0)
def main(task_ids, by_filename=False, validate=None, log=None, **kwargs):
kwargs["write"] = False
if by_filename:
task_ids_from_file = []
for filename in task_ids:
with open(filename, 'r') as f:
task_ids_from_file += zip(
*list(map(str.split, filter(None, map(str.strip, f)))))
task_ids = task_ids_from_file
else:
task_ids = [[task_id] for task_id in task_ids]
assert len(task_ids) == 2, "Got %d lists of task IDs instead of two" % len(
task_ids)
downloader = TaskDownloader(**kwargs)
scores = []
validate_h = open(validate, "w", encoding="utf-8") if validate else None
log_h = open(log, "w", encoding="utf-8") if log else None
if log:
fields = ["guessed", "ref"] + Scores.field_titles(
eval_type=LABELED) + Scores.field_titles(eval_type=UNLABELED)
print(*fields, file=log_h, sep="\t", flush=True)
for task_id_pair in tqdm(list(zip(*task_ids)),
unit=" tasks",
desc="Evaluating"):
passage_pair = []
for task_id in task_id_pair:
passage, *_ = downloader.download_task(task_id,
validate=validate_h,
**kwargs)
passage_pair.append(passage)
score = evaluate(*passage_pair, **kwargs)
if log:
fields = list(task_id_pair) + score.fields(
eval_type=LABELED) + score.fields(eval_type=UNLABELED)
print(*fields, file=log_h, sep="\t", flush=True)
scores.append(score)
if validate:
validate_h.close()
if log:
log_h.close()
print()
if len(scores) > 1:
print("Aggregated scores:")
Scores.aggregate(scores).print()
def main():
argparser = argparse.ArgumentParser(description=desc)
argparser.add_argument("filenames", nargs="+",
help="file names to convert and evaluate")
argparser.add_argument("-f", "--format", required=True, choices=convert.CONVERTERS,
help="input file format")
argparser.add_argument("-T", "--tree", action="store_true",
help="remove multiple parents to get a tree")
argparser.add_argument("-s", "--strict", action="store_true",
help="stop immediately if failed to convert or evaluate a file")
argparser.add_argument("-v", "--verbose", action="store_true",
help="print evaluation results for each file separately")
args = argparser.parse_args()
converter1 = convert.TO_FORMAT[args.format]
converter2 = convert.FROM_FORMAT[args.format]
scores = []
for pattern in args.filenames:
filenames = glob.glob(pattern)
if not filenames:
raise IOError("Not found: " + pattern)
for filename in filenames:
sys.stdout.write("\rConverting %s" % filename)
sys.stdout.flush()
ref = file2passage(filename)
try:
guessed = next(converter2(converter1(ref, tree=args.tree), ref.ID))
scores.append(evaluate(guessed, ref, verbose=args.verbose))
except Exception as e:
if args.strict:
raise ValueError("Error evaluating conversion of %s" % filename) from e
else:
print("Error evaluating conversion of %s: %s" % (filename, e), file=sys.stderr)
print()
if args.verbose and len(scores) > 1:
print("Aggregated scores:")
Scores.aggregate(scores).print()
sys.exit(0)
def main(args):
converter1 = convert.TO_FORMAT[args.format]
converter2 = convert.FROM_FORMAT[args.format]
scores = []
for ref in get_passages_with_progress_bar(args.filenames,
desc="Converting"):
try:
guessed = next(converter2(converter1(ref, tree=args.tree), ref.ID))
scores.append(evaluate(guessed, ref, verbose=args.verbose))
except Exception as e:
if args.strict:
raise ValueError("Error evaluating conversion of %s" %
ref.ID) from e
else:
with tqdm.external_write_mode():
print("Error evaluating conversion of %s: %s" %
(ref.ID, e),
file=sys.stderr)
print()
if args.verbose and len(scores) > 1:
print("Aggregated scores:")
Scores.aggregate(scores).print()
def main():
argparser = argparse.ArgumentParser(description=desc)
argparser.add_argument("filenames",
nargs="+",
help="file names to convert and evaluate")
argparser.add_argument("-f",
"--format",
required=True,
choices=convert.CONVERTERS,
help="input file format")
argparser.add_argument("-T",
"--tree",
action="store_true",
help="remove multiple parents to get a tree")
args = argparser.parse_args()
converter1 = convert.TO_FORMAT[args.format]
converter2 = convert.FROM_FORMAT[args.format]
scores = []
for pattern in args.filenames:
filenames = glob.glob(pattern)
if not filenames:
raise IOError("Not found: " + pattern)
for filename in filenames:
ref = file2passage(filename)
guessed = next(converter2(converter1(ref), ref.ID))
scores.append(
evaluate(guessed,
ref,
fscore=True,
verbose=True,
units=False,
errors=False))
if len(scores) > 1:
print("Aggregated scores:")
Scores.aggregate(scores).print()
sys.exit(0)
def main(task_ids, by_filename=False, validate=None, log=None, **kwargs):
kwargs["write"] = False
if by_filename:
task_ids_from_file = []
for filename in task_ids:
with open(filename, 'r') as f:
task_ids_from_file += zip(*list(map(str.split, filter(None, map(str.strip, f)))))
task_ids = task_ids_from_file
else:
task_ids = [[task_id] for task_id in task_ids]
assert len(task_ids) == 2, "Got %d lists of task IDs instead of two" % len(task_ids)
downloader = TaskDownloader(**kwargs)
scores = []
validate_h = open(validate, "w", encoding="utf-8") if validate else None
log_h = open(log, "w", encoding="utf-8") if log else None
if log:
fields = ["guessed", "ref"] + Scores.field_titles(eval_type=LABELED) + Scores.field_titles(eval_type=UNLABELED)
print(*fields, file=log_h, sep="\t", flush=True)
for task_id_pair in tqdm(list(zip(*task_ids)), unit=" tasks", desc="Evaluating"):
passage_pair = []
for task_id in task_id_pair:
passage, *_ = downloader.download_task(task_id, validate=validate_h, **kwargs)
passage_pair.append(passage)
score = evaluate(*passage_pair, **kwargs)
if log:
fields = list(task_id_pair) + score.fields(eval_type=LABELED) + score.fields(eval_type=UNLABELED)
print(*fields, file=log_h, sep="\t", flush=True)
scores.append(score)
if validate:
validate_h.close()
if log:
log_h.close()
print()
if len(scores) > 1:
print("Aggregated scores:")
Scores.aggregate(scores).print()
def main(args):
guessed, ref = [
ioutil.read_files_and_dirs((x, ), converters=FROM_FORMAT)
for x in (args.guessed, args.ref)
]
if len(guessed) != len(ref):
raise ValueError(
"Number of passages to compare does not match: %d != %d" %
(len(guessed), len(ref)))
if len(guessed) > 1:
guessed_by_id = {
g.ID: g
for g in tqdm(
guessed, desc="Reading " + args.guessed, unit=" passages")
}
try:
guessed = [
guessed_by_id[p.ID] for p in tqdm(
ref, desc="Reading " + args.ref, unit=" passages")
]
except KeyError as e:
raise ValueError("Passage IDs do not match") from e
results = [
evaluate(g, r, errors=True) for g, r in zip(
tqdm(guessed, desc="Evaluating", unit=" passages"), ref)
]
confusion_matrix = Scores.aggregate(
results).evaluators[LABELED].results[PRIMARY].errors.most_common()
label_map = {}
for (g, r), _ in confusion_matrix:
g, *_ = g.partition("|")
prefix, *_ = g.partition(":")
if not any(l.startswith(prefix)
for l in label_map): # drop suffix for most common label
g = prefix
if g not in label_map:
label_map[g], *_ = r.partition("|")
with open(args.out_file, "w", encoding="utf-8") as f:
csv.writer(f).writerows(
tqdm(sorted(label_map.items()),
desc="Writing " + args.out_file,
unit=" rows"))
def print_title(file):
print("learning rate, decay factor, average unlabeled f1, "
", ".join(Scores.field_titles()),
file=file)
def print_title(file):
print("learning rate, decay factor, average unlabeled f1, " ", ".join(Scores.field_titles()), file=file)
def get_field_titles(self):
return [p for p in self.params.keys()] + ["average_labeled_f1"] + Scores.field_titles()
def get_field_titles(self):
return [p for p in self.params.keys()] + ["average_labeled_f1"] + Scores.field_titles()
