def call(self, y_true=None, y_pred=None, arguments=None):
''' compute metric '''
label_path_file = arguments["label_vocab_path"]
return "\n" + seq_classification_report(
ids_to_sentences(y_true, label_path_file),
ids_to_sentences(y_pred, label_path_file),
digits=4)
def call(self, predictions, log_verbose=False):
''' main func entrypoint'''
preds = predictions["preds"]
output_index = predictions["output_index"]
if output_index is None:
res_file = self.config["solver"]["postproc"].get("res_file", "")
label_path_file = self.config["data"]["task"]["label_vocab"]
else:
res_file = self.config["solver"]["postproc"][output_index].get(
"res_file", "")
label_path_file = self.config["data"]["task"]["label_vocab"][
output_index]
if res_file == "":
logging.info(
"Infer res not saved. You can check 'res_file' in your config."
)
return
res_dir = os.path.dirname(res_file)
if not os.path.exists(res_dir):
os.makedirs(res_dir)
logging.info("Save inference result to: {}".format(res_file))
preds = ids_to_sentences(preds, label_path_file)
with open(res_file, "w", encoding="utf-8") as in_f:
for i, pre in enumerate(preds):
entities = get_entities(pre) # [('PER', 0, 1), ('LOC', 3, 3)]
if not entities:
in_f.write("Null")
else:
new_line = "\t".join(
[" ".join(map(str, entity)) for entity in entities])
in_f.write(new_line)
in_f.write("\n")
def test_ids_to_sentences(self):
''' test ids_to_sentences function '''
config = utils.load_config(self.config_file)
ids = [[2, 3, 1]]
vocab_file_path = config["data"]["task"]["label_vocab"]
sents = ids_to_sentences(ids, vocab_file_path)
self.assertAllEqual(sents, [["I-PER", "B-LOC", "B-PER"]])
def call(self, predictions, log_verbose=False):
''' main func entrypoint'''
preds = predictions["preds"]
res_file = self.config["solver"]["postproc"].get("res_file", "")
if res_file == "":
logging.info("Infer res not saved. You can check 'res_file' in your config.")
return
res_dir = os.path.dirname(res_file)
if not os.path.exists(res_dir):
os.makedirs(res_dir)
logging.info("Save inference result to: {}".format(res_file))
self.task_config = self.config['data']['task']
self.label_vocab_file_paths = self.task_config['label_vocab']
if not isinstance(self.label_vocab_file_paths, list):
self.label_vocab_file_paths = [self.label_vocab_file_paths]
self.use_label_vocab = self.task_config['use_label_vocab']
if self.use_label_vocab:
label_path_file = self.label_vocab_file_paths[0]
else:
label_path_file = self.task_config["text_vocab"]
preds = ids_to_sentences(preds, label_path_file)
with open(res_file, "w", encoding="utf-8") as in_f:
for i, pre in enumerate(preds):
while len(pre) > 1 and pre[-1] in ['<unk>', '<pad>', '<eos>']:
pre.pop()
pred_abs = ' '.join(pre)
in_f.write(pred_abs)
in_f.write("\n")
def call(self, predictions, log_verbose=False):
''' main func entrypoint'''
preds = predictions["preds"]
paths = self.config["data"]["infer"]["paths"]
max_seq_len = self.config["data"]["task"]["max_seq_len"]
text = []
counter = 0
for path in paths:
with open(path, 'r', encoding='utf8') as file_input:
for line in file_input.readlines():
line = list(line.strip())
if line:
if len(line) >= max_seq_len:
line = line[:max_seq_len]
else:
line.extend(["unk"]*(max_seq_len-len(line)))
text.append("".join(line))
counter += 1
logging.info("Load {} lines from {}.".format(str(counter), path))
res_file = self.config["solver"]["postproc"].get("res_file", "")
if res_file == "":
logging.info("Infer res not saved. You can check 'res_file' in your config.")
return
res_dir = os.path.dirname(res_file)
if not os.path.exists(res_dir):
os.makedirs(res_dir)
logging.info("Save inference result to: {}".format(res_file))
label_path_file = self.config["data"]["task"]["label_vocab"]
preds = ids_to_sentences(preds, label_path_file)
assert len(text) == len(preds)
with open(res_file, "w", encoding="utf-8") as in_f:
for i, pre in enumerate(preds):
entity_dict = {}
entities = get_entities(pre) # [('PER', 0, 1), ('LOC', 3, 3)]
for entity_tuple in entities:
entity = "".join([text[i][j] for j in range(entity_tuple[1], entity_tuple[2] + 1)])
if entity_tuple[0] in entity_dict:
entity_dict[entity_tuple[0]].append(entity)
else:
entity_dict[entity_tuple[0]] = [entity]
in_f.write(str(entity_dict))
in_f.write("\n")
def call(self, y_true=None, y_pred=None, arguments=None):
''' compute metric '''
label_path_file = self.config["data"]["task"]["label_vocab"]
return seq_classification_report(ids_to_sentences(y_true, label_path_file),
ids_to_sentences(y_pred, label_path_file))