lwdragon's picture
Update app.py
41a0931
import gradio as gr
from mindformers.trainer import Trainer
# 初始化trainer
trainer = Trainer(
task='token_classification',
model='tokcls_bert_base_chinese',
)
# examples
warm_input_data = ["结果上周六他们主场0:3惨败给了中游球队瓦拉多利德,近7个多月以来西甲首次输球。", "清华大学座落于首都北京"]
# warm_up
trainer.predict(
predict_checkpoint=
'./tokcls_bert_base_chinese_cluener.ckpt',
input_data=warm_input_data)
# 数据后处理,将数据转成gr.HighlightedText需要的数据
def post_procces(text, text_list):
res = []
cur_index = 0
for item in text_list:
res.append((text[cur_index:item["start"]], None))
res.append((text[item["start"]:(item["end"] + 1)],
" ".join([item["entity_group"],
str(item["score"])])))
cur_index = item["end"] + 1
res.append((text[cur_index:], None))
return res
# 预测
def token_classification(text):
res_list = trainer.predict(input_data=text)
res = post_procces(text, res_list[0])
print(res)
return res, res_list
# gradio接口
gr.Interface(
token_classification,
gr.Textbox(
label="Text",
info="Enter sentence here..xt",
lines=3,
value="结果上周六他们主场0:3惨败给了中游球队瓦拉多利德,近7个多月以来西甲首次输球。",
),
# ["highlight", "json"],
[
gr.HighlightedText(
label="Token Classification",
combine_adjacent=True,
),
gr.JSON()
],
examples=[*warm_input_data],
).launch()