使用百度深度学习模型ERNIE对输入的字符串做分类预测（基于fluid）

#! /usr/bin/env python
# author: qibaoyuan - qibaoyuan@xiaomi.com 2019.08.01
# just demo, no gurantees

import collections
import time

import paddle.fluid as fluid
import numpy as np

#可以参考ernie的reader包
from paddle.reader.task_reader import ClassifyReader

main_prog = fluid.Program()
startup_prog = fluid.Program()

place = fluid.CPUPlace()
exe = fluid.Executor(place)
exe.run(startup_prog)
path = "~/ERNIE-output-2019-07-29/step_38212_inference_model/step_38212_inference_model"

[inference_program, feed_target_names, fetch_targets] = (fluid.io.load_inference_model(dirname=path, executor=exe))

s = time.time()

classifyReader = ClassifyReader(vocab_path="~/vocab.txt",
                                is_classify=False, is_inference=True)
Example = collections.namedtuple('Example', ['text_a', 'text_b', 'label'])

text = "为了预防心脑血管病，老年人需要吃阿司匹林吗？ 吃不吃阿司匹林是不是老年人没有直接的关系。"

example = Example(text_a=text, text_b='', label='')

input_2 = classifyReader._convert_example_to_record(example, classifyReader.max_seq_len, classifyReader.tokenizer)

print(input_2)
for x in range(50):
    tensor_img = np.reshape(np.asarray(input_2.token_ids, dtype=np.int), (1, 512, 1))
    feed_data = {feed_target_names[0]: tensor_img,
                 feed_target_names[1]: np.zeros((1, 512, 1), dtype=np.int),
                 feed_target_names[2]: np.reshape(np.arange(512), (1, 512, 1)),
                 feed_target_names[3]: np.ones((1, 512, 1), dtype=np.float32),
                 }
    results = exe.run(inference_program,
                      feed=feed_data,
                      fetch_list=fetch_targets)

    print(results)
    print(np.argmax(results))

print(time.time() - s)