parser.add_argument("--lr",
type=float,
default=0.00005,
help="learning rate")
parser.add_argument("--model_no", type=int, default=0, help="Model ID")
parser.add_argument("--train",
type=int,
default=0,
help="0: Don't train, 1: train")
parser.add_argument("--infer",
type=int,
default=1,
help="0: Don't infer, 1: Infer")
args = parser.parse_args()
if args.train == 1:
net = train_and_fit(args)
if args.infer == 1:
inferer = infer_from_trained(args, detect_entities=False)
test = "The surprise [E1]visit[/E1] caused a [E2]frenzy[/E2] on the already chaotic trading floor."
inferer.infer_sentence(test, detect_entities=False)
while True:
sent = input(
"Type input sentence ('quit' or 'exit' to terminate):\n")
if sent.lower() in ['quit', 'exit']:
break
inferer.infer_sentence(sent, detect_entities=False)
parser.add_argument("--train",
type=int,
default=1,
help="0: Don't train, 1: train")
parser.add_argument("--infer",
type=int,
default=1,
help="0: Don't infer, 1: Infer")
args = parser.parse_args()
if (args.train == 1) and (args.task != 'fewrel'):
net = train_and_fit(args)
if (args.infer == 1) and (args.task != 'fewrel'):
inferer = infer_from_trained(args, detect_entities=True)
test = "The surprise [E1]visit[/E1] caused a [E2]frenzy[/E2] on the already chaotic trading floor."
inferer.infer_sentence(test, detect_entities=False)
test2 = "After eating the chicken, he developed a sore throat the next morning."
inferer.infer_sentence(test2, detect_entities=True)
while True:
sent = input(
"Type input sentence ('quit' or 'exit' to terminate):\n")
if sent.lower() in ['quit', 'exit']:
break
inferer.infer_sentence(sent, detect_entities=False)
if args.task == 'fewrel':
fewrel = FewRel(args)
meta_input, e1_e2_start, meta_labels, outputs = fewrel.evaluate()
def make_app(argv, debug=False):
app = Flask(__name__)
#CORS(app, support_credentials=True)
# TODO
# @cross_origin(supports_credentials=True)
# def login():
# return jsonify({'success': 'ok'})
app.debug = debug
parser = Parser().getParser()
#argv = sys.argv[1:]
args, _ = parser.parse_known_args()
# args = parser.parse_args()
if (args.train == 1) and (args.task != 'fewrel'):
app.net = train_and_fit(args)
if (args.infer == 1) and (args.task != 'fewrel'):
app.inferer = infer_from_trained(args.model_path,
detect_entities=True,
args=args)
if args.task == 'fewrel':
fewrel = FewRel(args.model_path, args)
meta_input, e1_e2_start, meta_labels, outputs = fewrel.evaluate()
def find_best_prediction(out):
best_pred = max(out, key=itemgetter(2))
return best_pred[0], best_pred[1], best_pred[2]
def get_best_predictions(data, inputtype="simplejson"):
if inputtype == "simplejson":
sen_name = "sentext"
elif inputtype == "texoo":
sen_name = "text"
for line in data:
#logger.info("sentence"+ str(line[sen_name]))
out = app.inferer.infer_sentence(line[sen_name],
detect_entities=True)
logger.info("out: " + str(out))
line["sentence"], line["pred"], line[
"prob"] = find_best_prediction(out)
return data
def get_all_predictions(data, inputtype="simplejson"):
new_data = []
if inputtype == "simplejson":
sen_name = "sentext"
elif inputtype == "texoo":
sen_name = "text"
for line in data:
logger.info("sentence" + str(line[sen_name]))
out = app.inferer.infer_sentence(line[sen_name],
detect_entities=True)
logger.info("out: " + str(out))
if len(out) == 0:
logger.info("test")
line["sentence"], line["pred"], line["prob"] = None, None, None
new_data.append(line)
else:
for pred in out:
logger.info("pred : " + str(pred))
newline = copy.deepcopy(line)
newline["sentence"], newline["pred"], newline[
"prob"] = pred[0], pred[1], pred[2]
logger.info("newline : " + str(newline))
new_data.append(newline)
logger.info("new _DATA : \n" + str(new_data) + "\n")
return new_data
# takes new texoo json for each line
# ignores annotations and generates new ones
@app.route('/api/importtexoo', methods=['POST'])
def get_input_importtexoo():
print("request.data: ", request.data)
logger.info("request.data:" + str(request.data))
jsonInput = request.get_json(force=True)
# jsonInput = { "options": {"returnAllPredictions": False},
# "data": [
# {'length': 12,
# 'documentRef': 2,
# 'uid': 123,
# 'text': "I love Easter Sunday as a fashion moment because every church goer is ready to praise while dressed to the nines in their best Spring-inspired looks .",
# 'begin': 0,
# 'class': "thisClass",
# 'type': "type",
# "tokens": None,
# "empty": None,
# "language": "ENg",
# "sentences": None,
# 'source': "source",
# 'id': None,
# "title": "Output for bert-relex-api.demo.datexis.com",
#
# "annotations": [
# {"relationArguments": [
# {"arg1": "blq"},{"arg2": "ble"}]},
# {"relationArguments": [
# {"arg1": "blj"}, {"arg2": "blg"}]}
# ]
# }
# ]
# }
if jsonInput["options"]["returnAllPredictions"]:
data = get_all_predictions(jsonInput["data"], "texoo")
else:
data = get_best_predictions(jsonInput["data"], "texoo")
return make_result_json(data, "texoo")
@app.route('/api/importjson', methods=['POST'])
def get_input_importjson():
print("request.data: ", request.data)
logger.info("request.data:" + str(request.data))
jsonInput = request.get_json(force=True)
#jsonInput = '{ "options": {"returnAllPredictions": false},' \
# '"data": [' \
# '{"sentext": "I love Easter Sunday as a fashion moment because every church goer is ready to praise while dressed to the nines in their best Spring-inspired looks ."},' \
# ' {"sentext": "Wear them with basics and sparse accessories ."}' \
# ']}'
#jsonInput = json.loads(jsonInput)
if jsonInput["options"]["returnAllPredictions"]:
data = get_all_predictions(jsonInput["data"])
else:
data = get_best_predictions(jsonInput["data"])
return make_result_json(data)
return app