基于预训练模型 ERNIE-Gram 实现语义匹配

6.7NLP直播打卡课即将开播，欢迎大家关注课程，有任何问题来评论区或QQ群（群号:973379845）交流吧~~

直播链接请戳这里，每晚20:00-21:30👈

课程地址请戳这里👈

本案例介绍 NLP 最基本的任务类型之一 —— 文本语义匹配，并且基于 PaddleNLP 使用百度开源的预训练模型 ERNIE-Gram 搭建效果优异的语义匹配模型，来判断 2 段文本语义是否相同。

1. 背景介绍

文本语义匹配任务，简单来说就是给定两段文本，让模型来判断两段文本是不是语义相似。

在本案例中以权威的语义匹配数据集 LCQMC 为例，LCQMC 数据集是基于百度知道相似问题推荐构造的通问句语义匹配数据集。训练集中的每两段文本都会被标记为 1（语义相似） 或者 0（语义不相似）。更多数据集可访问千言获取哦。

例如百度知道场景下，用户搜索一个问题，模型会计算这个问题与候选问题是否语义相似，语义匹配模型会找出与问题语义相似的候选问题返回给用户，加快用户提问-获取答案的效率。例如，当某用户在搜索引擎中搜索 “深度学习的教材有哪些？”，模型就自动找到了一些语义相似的问题展现给用户:

2.快速实践

介绍如何准备数据，基于 ERNIE-Gram 模型搭建匹配网络，然后快速进行语义匹配模型的训练、评估和预测。

2.1 数据加载

为了训练匹配模型，一般需要准备三个数据集：训练集 train.tsv、验证集dev.tsv、测试集test.tsv。此案例我们使用 PaddleNLP 内置的语义数据集 LCQMC 来进行训练、评估、预测。

训练集: 用来训练模型参数的数据集，模型直接根据训练集来调整自身参数以获得更好的分类效果。

验证集: 用于在训练过程中检验模型的状态，收敛情况。验证集通常用于调整超参数，根据几组模型验证集上的表现，决定采用哪组超参数。

测试集: 用来计算模型的各项评估指标，验证模型泛化能力。

LCQMC 数据集是公开的语义匹配权威数据集。PaddleNLP 已经内置该数据集，一键即可加载。

AI Studio平台默认安装了Paddle和PaddleNLP，并定期更新版本。 如需手动更新Paddle，可参考飞桨安装说明，安装相应环境下最新版飞桨框架。

使用如下命令确保安装最新版PaddleNLP：

# 正式开始实验之前首先通过如下命令安装最新版本的 paddlenlp
!pip install --upgrade paddlenlp -i https://pypi.org/simple

Collecting paddlenlp
[?25l  Downloading https://files.pythonhosted.org/packages/b0/7d/6c24cda54d018d350ee342f715523ade7871660444ed95f3d3e753d6f388/paddlenlp-2.0.8-py3-none-any.whl (571kB)
[K     |█▏                              | 20kB 15kB/s eta 0:00:36[31mERROR: Exception:
Traceback (most recent call last):
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_vendor/urllib3/response.py", line 397, in _error_catcher
    yield
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_vendor/urllib3/response.py", line 479, in read
    data = self._fp.read(amt)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_vendor/cachecontrol/filewrapper.py", line 62, in read
    data = self.__fp.read(amt)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/http/client.py", line 457, in read
    n = self.readinto(b)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/http/client.py", line 501, in readinto
    n = self.fp.readinto(b)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/socket.py", line 589, in readinto
    return self._sock.recv_into(b)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/ssl.py", line 1071, in recv_into
    return self.read(nbytes, buffer)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/ssl.py", line 929, in read
    return self._sslobj.read(len, buffer)
socket.timeout: The read operation timed out

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_internal/cli/base_command.py", line 188, in main
    status = self.run(options, args)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_internal/commands/install.py", line 345, in run
    resolver.resolve(requirement_set)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_internal/legacy_resolve.py", line 196, in resolve
    self._resolve_one(requirement_set, req)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_internal/legacy_resolve.py", line 359, in _resolve_one
    abstract_dist = self._get_abstract_dist_for(req_to_install)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_internal/legacy_resolve.py", line 307, in _get_abstract_dist_for
    self.require_hashes
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_internal/operations/prepare.py", line 199, in prepare_linked_requirement
    progress_bar=self.progress_bar
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_internal/download.py", line 1064, in unpack_url
    progress_bar=progress_bar
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_internal/download.py", line 924, in unpack_http_url
    progress_bar)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_internal/download.py", line 1152, in _download_http_url
    _download_url(resp, link, content_file, hashes, progress_bar)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_internal/download.py", line 861, in _download_url
    hashes.check_against_chunks(downloaded_chunks)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_internal/utils/hashes.py", line 75, in check_against_chunks
    for chunk in chunks:
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_internal/download.py", line 829, in written_chunks
    for chunk in chunks:
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_internal/utils/ui.py", line 156, in iter
    for x in it:
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_internal/download.py", line 818, in resp_read
    decode_content=False):
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_vendor/urllib3/response.py", line 531, in stream
    data = self.read(amt=amt, decode_content=decode_content)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_vendor/urllib3/response.py", line 496, in read
    raise IncompleteRead(self._fp_bytes_read, self.length_remaining)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/contextlib.py", line 130, in __exit__
    self.gen.throw(type, value, traceback)
  File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/pip/_vendor/urllib3/response.py", line 402, in _error_catcher
    raise ReadTimeoutError(self._pool, None, 'Read timed out.')
pip._vendor.urllib3.exceptions.ReadTimeoutError: HTTPSConnectionPool(host='files.pythonhosted.org', port=443): Read timed out.[0m

import time
import os
import numpy as np

import paddle
import paddle.nn.functional as F
from paddlenlp.datasets import load_dataset
import paddlenlp

# 一键加载 Lcqmc 的训练集、验证集
train_ds, dev_ds = load_dataset("lcqmc", splits=["train", "dev"])

100%|██████████| 6827/6827 [00:00<00:00, 59739.00it/s]

# 输出训练集的前 3 条样本
for idx, example in enumerate(train_ds):
    if idx <= 3:
        print(example)

{'query': '喜欢打篮球的男生喜欢什么样的女生', 'title': '爱打篮球的男生喜欢什么样的女生', 'label': 1}
{'query': '我手机丢了，我想换个手机', 'title': '我想买个新手机，求推荐', 'label': 1}
{'query': '大家觉得她好看吗', 'title': '大家觉得跑男好看吗？', 'label': 0}
{'query': '求秋色之空漫画全集', 'title': '求秋色之空全集漫画', 'label': 1}

2.2 数据预处理

通过 PaddleNLP 加载进来的 LCQMC 数据集是原始的明文数据集，这部分我们来实现组 batch、tokenize 等预处理逻辑，将原始明文数据转换成网络训练的输入数据。

定义样本转换函数

# 因为是基于预训练模型 ERNIE-Gram 来进行，所以需要首先加载 ERNIE-Gram 的 tokenizer，
# 后续样本转换函数基于 tokenizer 对文本进行切分

tokenizer = paddlenlp.transformers.ErnieGramTokenizer.from_pretrained('ernie-gram-zh')

[2021-08-30 23:27:37,525] [    INFO] - Downloading vocab.txt from https://paddlenlp.bj.bcebos.com/models/transformers/ernie_gram_zh/vocab.txt
100%|██████████| 78/78 [00:00<00:00, 2965.81it/s]

# 将 1 条明文数据的 query、title 拼接起来，根据预训练模型的 tokenizer 将明文转换为 ID 数据
# 返回 input_ids 和 token_type_ids

def convert_example(example, tokenizer, max_seq_length=512, is_test=False):

    query, title = example["query"], example["title"]

    encoded_inputs = tokenizer(
        text=query, text_pair=title, max_seq_len=max_seq_length)

    input_ids = encoded_inputs["input_ids"]
    token_type_ids = encoded_inputs["token_type_ids"]

    if not is_test:
        label = np.array([example["label"]], dtype="int64")
        return input_ids, token_type_ids, label
    # 在预测或者评估阶段，不返回 label 字段
    else:
        return input_ids, token_type_ids

### 对训练集的第 1 条数据进行转换
input_ids, token_type_ids, label = convert_example(train_ds[0], tokenizer)

print(input_ids)

[1, 692, 811, 445, 2001, 497, 5, 654, 21, 692, 811, 614, 356, 314, 5, 291, 21, 2, 329, 445, 2001, 497, 5, 654, 21, 692, 811, 614, 356, 314, 5, 291, 21, 2]

print(token_type_ids)

[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]

print(label)

[1]

# 为了后续方便使用，我们使用python偏函数（partial）给 convert_example 赋予一些默认参数
from functools import partial

# 训练集和验证集的样本转换函数
trans_func = partial(
    convert_example,
    tokenizer=tokenizer,
    max_seq_length=512)

组装 Batch 数据 & Padding

上一小节，我们完成了对单条样本的转换，本节我们需要将样本组合成 Batch 数据，对于不等长的数据还需要进行 Padding 操作，便于 GPU 训练。

PaddleNLP 提供了许多关于 NLP 任务中构建有效的数据 pipeline 的常用 API

API	简介
paddlenlp.data.Stack	堆叠N个具有相同shape的输入数据来构建一个batch
paddlenlp.data.Pad	将长度不同的多个句子padding到统一长度，取N个输入数据中的最大长度
paddlenlp.data.Tuple	将多个batchify函数包装在一起

更多数据处理操作详见： https://paddlenlp.readthedocs.io/zh/latest/data_prepare/data_preprocess.html

from paddlenlp.data import Stack, Pad, Tuple
a = [1, 2, 3, 4]
b = [3, 4, 5, 6]
c = [5, 6, 7, 8]
result = Stack()([a, b, c])
print("Stacked Data: \n", result)
print()

a = [1, 2, 3, 4]
b = [5, 6, 7]
c = [8, 9]
result = Pad(pad_val=0)([a, b, c])
print("Padded Data: \n", result)
print()

data = [
        [[1, 2, 3, 4], [1]],
        [[5, 6, 7], [0]],
        [[8, 9], [1]],
       ]
batchify_fn = Tuple(Pad(pad_val=0), Stack())
ids, labels = batchify_fn(data)
print("ids: \n", ids)
print()
print("labels: \n", labels)
print()

Stacked Data: 
 [[1 2 3 4]
 [3 4 5 6]
 [5 6 7 8]]

Padded Data: 
 [[1 2 3 4]
 [5 6 7 0]
 [8 9 0 0]]

ids: 
 [[1 2 3 4]
 [5 6 7 0]
 [8 9 0 0]]

labels: 
 [[1]
 [0]
 [1]]

# 我们的训练数据会返回 input_ids, token_type_ids, labels 3 个字段
# 因此针对这 3 个字段需要分别定义 3 个组 batch 操作
batchify_fn = lambda samples, fn=Tuple(
    Pad(axis=0, pad_val=tokenizer.pad_token_id),  # input_ids
    Pad(axis=0, pad_val=tokenizer.pad_token_type_id),  # token_type_ids
    Stack(dtype="int64")  # label
): [data for data in fn(samples)]

定义 Dataloader

下面我们基于组 batchify_fn 函数和样本转换函数 trans_func 来构造训练集的 DataLoader, 支持多卡训练

# 定义分布式 Sampler: 自动对训练数据进行切分，支持多卡并行训练
batch_sampler = paddle.io.DistributedBatchSampler(train_ds, batch_size=32, shuffle=True)

# 基于 train_ds 定义 train_data_loader
# 因为我们使用了分布式的 DistributedBatchSampler, train_data_loader 会自动对训练数据进行切分
train_data_loader = paddle.io.DataLoader(
        dataset=train_ds.map(trans_func),
        batch_sampler=batch_sampler,
        collate_fn=batchify_fn,
        return_list=True)

# 针对验证集数据加载，我们使用单卡进行评估，所以采用 paddle.io.BatchSampler 即可
# 定义 dev_data_loader
batch_sampler = paddle.io.BatchSampler(dev_ds, batch_size=32, shuffle=False)
dev_data_loader = paddle.io.DataLoader(
        dataset=dev_ds.map(trans_func),
        batch_sampler=batch_sampler,
        collate_fn=batchify_fn,
        return_list=True)

2.3 模型搭建

自从 2018 年 10 月以来，NLP 个领域的任务都通过 Pretrain + Finetune 的模式相比传统 DNN 方法在效果上取得了显著的提升，本节我们以百度开源的预训练模型 ERNIE-Gram 为基础模型，在此之上构建 Point-wise 语义匹配网络。

首先我们来定义网络结构:

import paddle.nn as nn

# 我们基于 ERNIE-Gram 模型结构搭建 Point-wise 语义匹配网络
# 所以此处先定义 ERNIE-Gram 的 pretrained_model
pretrained_model = paddlenlp.transformers.ErnieGramModel.from_pretrained('ernie-gram-zh')
#pretrained_model = paddlenlp.transformers.ErnieModel.from_pretrained('ernie-1.0')


class PointwiseMatching(nn.Layer):
   
    # 此处的 pretained_model 在本例中会被 ERNIE-Gram 预训练模型初始化
    def __init__(self, pretrained_model, dropout=None):
        super().__init__()
        self.ptm = pretrained_model
        self.dropout = nn.Dropout(dropout if dropout is not None else 0.1)

        # 语义匹配任务: 相似、不相似 2 分类任务
        self.classifier = nn.Linear(self.ptm.config["hidden_size"], 2)

    def forward(self,
                input_ids,
                token_type_ids=None,
                position_ids=None,
                attention_mask=None):

        # 此处的 Input_ids 由两条文本的 token ids 拼接而成
        # token_type_ids 表示两段文本的类型编码
        # 返回的 cls_embedding 就表示这两段文本经过模型的计算之后而得到的语义表示向量
        _, cls_embedding = self.ptm(input_ids, token_type_ids, position_ids,
                                    attention_mask)

        cls_embedding = self.dropout(cls_embedding)

        # 基于文本对的语义表示向量进行 2 分类任务
        logits = self.classifier(cls_embedding)
        probs = F.softmax(logits)

        return probs

# 定义 Point-wise 语义匹配网络
model = PointwiseMatching(pretrained_model)

[2021-08-30 23:27:37,699] [    INFO] - Downloading https://paddlenlp.bj.bcebos.com/models/transformers/ernie_gram_zh/ernie_gram_zh.pdparams and saved to /home/aistudio/.paddlenlp/models/ernie-gram-zh
[2021-08-30 23:27:37,703] [    INFO] - Downloading ernie_gram_zh.pdparams from https://paddlenlp.bj.bcebos.com/models/transformers/ernie_gram_zh/ernie_gram_zh.pdparams
100%|██████████| 583566/583566 [00:09<00:00, 62752.98it/s]

2.4 模型训练 & 评估

from paddlenlp.transformers import LinearDecayWithWarmup

epochs = 3
num_training_steps = len(train_data_loader) * epochs

# 定义 learning_rate_scheduler，负责在训练过程中对 lr 进行调度
lr_scheduler = LinearDecayWithWarmup(5E-5, num_training_steps, 0.0)

# Generate parameter names needed to perform weight decay.
# All bias and LayerNorm parameters are excluded.
decay_params = [
    p.name for n, p in model.named_parameters()
    if not any(nd in n for nd in ["bias", "norm"])
]

# 定义 Optimizer
optimizer = paddle.optimizer.AdamW(
    learning_rate=lr_scheduler,
    parameters=model.parameters(),
    weight_decay=0.0,
    apply_decay_param_fun=lambda x: x in decay_params)

# 采用交叉熵 损失函数
criterion = paddle.nn.loss.CrossEntropyLoss()

# 评估的时候采用准确率指标
metric = paddle.metric.Accuracy()

# 因为训练过程中同时要在验证集进行模型评估，因此我们先定义评估函数

@paddle.no_grad()
def evaluate(model, criterion, metric, data_loader, phase="dev"):
    model.eval()
    metric.reset()
    losses = []
    for batch in data_loader:
        input_ids, token_type_ids, labels = batch
        probs = model(input_ids=input_ids, token_type_ids=token_type_ids)
        loss = criterion(probs, labels)
        losses.append(loss.numpy())
        correct = metric.compute(probs, labels)
        metric.update(correct)
        accu = metric.accumulate()
    print("eval {} loss: {:.5}, accu: {:.5}".format(phase,
                                                    np.mean(losses), accu))
    model.train()
    metric.reset()

# 接下来，开始正式训练模型，训练时间较长，可注释掉这部分

global_step = 0
tic_train = time.time()

for epoch in range(1, epochs + 1):
    for step, batch in enumerate(train_data_loader, start=1):

        input_ids, token_type_ids, labels = batch
        probs = model(input_ids=input_ids, token_type_ids=token_type_ids)
        loss = criterion(probs, labels)
        correct = metric.compute(probs, labels)
        metric.update(correct)
        acc = metric.accumulate()

        global_step += 1
        
        # 每间隔 10 step 输出训练指标
        if global_step % 10 == 0:
            print(
                "global step %d, epoch: %d, batch: %d, loss: %.5f, accu: %.5f, speed: %.2f step/s"
                % (global_step, epoch, step, loss, acc,
                    10 / (time.time() - tic_train)))
            tic_train = time.time()
        loss.backward()
        optimizer.step()
        lr_scheduler.step()
        optimizer.clear_grad()

        # 每间隔 100 step 在验证集和测试集上进行评估
        if global_step % 100 == 0:
            evaluate(model, criterion, metric, dev_data_loader, "dev")
            
# 训练结束后，存储模型参数
save_dir = os.path.join("checkpoint", "model_%d" % global_step)
os.makedirs(save_dir)

save_param_path = os.path.join(save_dir, 'model_state.pdparams')
paddle.save(model.state_dict(), save_param_path)
tokenizer.save_pretrained(save_dir)

global step 22380, epoch: 3, batch: 7456, loss: 0.38857, accu: 0.88203, speed: 9.62 step/s

模型训练过程中会输出如下日志:

global step 5310, epoch: 3, batch: 1578, loss: 0.31671, accu: 0.95000, speed: 0.63 step/s
global step 5320, epoch: 3, batch: 1588, loss: 0.36240, accu: 0.94063, speed: 6.98 step/s
global step 5330, epoch: 3, batch: 1598, loss: 0.41451, accu: 0.93854, speed: 7.40 step/s
global step 5340, epoch: 3, batch: 1608, loss: 0.31327, accu: 0.94063, speed: 7.01 step/s
global step 5350, epoch: 3, batch: 1618, loss: 0.40664, accu: 0.93563, speed: 7.83 step/s
global step 5360, epoch: 3, batch: 1628, loss: 0.33064, accu: 0.93958, speed: 7.34 step/s
global step 5370, epoch: 3, batch: 1638, loss: 0.38411, accu: 0.93795, speed: 7.72 step/s
global step 5380, epoch: 3, batch: 1648, loss: 0.35376, accu: 0.93906, speed: 7.92 step/s
global step 5390, epoch: 3, batch: 1658, loss: 0.39706, accu: 0.93924, speed: 7.47 step/s
global step 5400, epoch: 3, batch: 1668, loss: 0.41198, accu: 0.93781, speed: 7.41 step/s
eval dev loss: 0.4177, accu: 0.89082
global step 5410, epoch: 3, batch: 1678, loss: 0.34453, accu: 0.93125, speed: 0.63 step/s
global step 5420, epoch: 3, batch: 1688, loss: 0.34569, accu: 0.93906, speed: 7.75 step/s
global step 5430, epoch: 3, batch: 1698, loss: 0.39160, accu: 0.92917, speed: 7.54 step/s
global step 5440, epoch: 3, batch: 1708, loss: 0.46002, accu: 0.93125, speed: 7.05 step/s
global step 5450, epoch: 3, batch: 1718, loss: 0.32302, accu: 0.93188, speed: 7.14 step/s
global step 5460, epoch: 3, batch: 1728, loss: 0.40802, accu: 0.93281, speed: 7.22 step/s
global step 5470, epoch: 3, batch: 1738, loss: 0.34607, accu: 0.93348, speed: 7.44 step/s
global step 5480, epoch: 3, batch: 1748, loss: 0.34709, accu: 0.93398, speed: 7.38 step/s
global step 5490, epoch: 3, batch: 1758, loss: 0.31814, accu: 0.93437, speed: 7.39 step/s
global step 5500, epoch: 3, batch: 1768, loss: 0.42689, accu: 0.93125, speed: 7.74 step/s
eval dev loss: 0.41789, accu: 0.88968

基于默认参数配置进行单卡训练大概要持续 4 个小时左右，会训练完成 3 个 Epoch, 模型最终的收敛指标结果如下:

数据集	Accuracy
dev.tsv	89.62

可以看到: 我们基于 PaddleNLP ，利用 ERNIE-Gram 预训练模型使用非常简洁的代码，就在权威语义匹配数据集上取得了很不错的效果.

2.5 模型预测

接下来我们使用已经训练好的语义匹配模型对一些预测数据进行预测。待预测数据为每行都是文本对的 tsv 文件，我们使用 Lcqmc 数据集的测试集作为我们的预测数据，进行预测并提交预测结果到 千言文本相似度竞赛

下载我们已经训练好的语义匹配模型, 并解压

# 下载我们基于 Lcqmc 事先训练好的语义匹配模型并解压
! wget https://paddlenlp.bj.bcebos.com/models/text_matching/ernie_gram_zh_pointwise_matching_model.tar
! tar -xvf ernie_gram_zh_pointwise_matching_model.tar

--2021-08-31 00:48:35--  https://paddlenlp.bj.bcebos.com/models/text_matching/ernie_gram_zh_pointwise_matching_model.tar
Resolving paddlenlp.bj.bcebos.com (paddlenlp.bj.bcebos.com)... 182.61.200.229, 182.61.200.195, 2409:8c00:6c21:10ad:0:ff:b00e:67d, ...
Connecting to paddlenlp.bj.bcebos.com (paddlenlp.bj.bcebos.com)|182.61.200.229|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 597667840 (570M) [application/x-tar]
Saving to: ‘ernie_gram_zh_pointwise_matching_model.tar.1’

ernie_gram_zh_point 100%[===================>] 569.98M  53.5MB/s    in 12s     

2021-08-31 00:48:47 (48.6 MB/s) - ‘ernie_gram_zh_pointwise_matching_model.tar.1’ saved [597667840/597667840]

ernie_gram_zh_pointwise_matching_model/
ernie_gram_zh_pointwise_matching_model/model_state.pdparams
ernie_gram_zh_pointwise_matching_model/vocab.txt
ernie_gram_zh_pointwise_matching_model/tokenizer_config.json

# 测试数据由 2 列文本构成 tab 分隔
# Lcqmc 默认下载到如下路径
! head -n3 "${HOME}/.paddlenlp/datasets/LCQMC/lcqmc/lcqmc/test.tsv"

谁有狂三这张高清的	这张高清图，谁有
英雄联盟什么英雄最好	英雄联盟最好英雄是什么
这是什么意思，被蹭网吗	我也是醉了，这是什么意思

定义预测函数

def predict(model, data_loader):
    
    batch_probs = []

    # 预测阶段打开 eval 模式，模型中的 dropout 等操作会关掉
    model.eval()

    with paddle.no_grad():
        for batch_data in data_loader:
            input_ids, token_type_ids = batch_data
            input_ids = paddle.to_tensor(input_ids)
            token_type_ids = paddle.to_tensor(token_type_ids)
            
            # 获取每个样本的预测概率: [batch_size, 2] 的矩阵
            batch_prob = model(
                input_ids=input_ids, token_type_ids=token_type_ids).numpy()

            batch_probs.append(batch_prob)
        batch_probs = np.concatenate(batch_probs, axis=0)

        return batch_probs

定义预测数据的 data_loader

# 预测数据的转换函数
# predict 数据没有 label, 因此 convert_exmaple 的 is_test 参数设为 True
trans_func = partial(
    convert_example,
    tokenizer=tokenizer,
    max_seq_length=512,
    is_test=True)

# 预测数据的组 batch 操作
# predict 数据只返回 input_ids 和 token_type_ids，因此只需要 2 个 Pad 对象作为 batchify_fn
batchify_fn = lambda samples, fn=Tuple(
    Pad(axis=0, pad_val=tokenizer.pad_token_id),  # input_ids
    Pad(axis=0, pad_val=tokenizer.pad_token_type_id),  # segment_ids
): [data for data in fn(samples)]

# 加载预测数据
test_ds = load_dataset("lcqmc", splits=["test"])

batch_sampler = paddle.io.BatchSampler(test_ds, batch_size=32, shuffle=False)

# 生成预测数据 data_loader
predict_data_loader =paddle.io.DataLoader(
        dataset=test_ds.map(trans_func),
        batch_sampler=batch_sampler,
        collate_fn=batchify_fn,
        return_list=True)

定义预测模型

pretrained_model = paddlenlp.transformers.ErnieGramModel.from_pretrained('ernie-gram-zh')

model = PointwiseMatching(pretrained_model)

[2021-08-31 01:28:51,550] [    INFO] - Already cached /home/aistudio/.paddlenlp/models/ernie-gram-zh/ernie_gram_zh.pdparams

加载已训练好的模型参数

# 刚才下载的模型解压之后存储路径为 ./ernie_gram_zh_pointwise_matching_model/model_state.pdparams
state_dict = paddle.load("./ernie_gram_zh_pointwise_matching_model/model_state.pdparams")

# 刚才下载的模型解压之后存储路径为 ./pointwise_matching_model/ernie1.0_base_pointwise_matching.pdparams
# state_dict = paddle.load("pointwise_matching_model/ernie1.0_base_pointwise_matching.pdparams")
model.set_dict(state_dict)

开始预测

for idx, batch in enumerate(predict_data_loader):
    if idx < 1:
        print(batch)

[Tensor(shape=[32, 38], dtype=int64, place=CUDAPinnedPlace, stop_gradient=True,
       [[1   , 1022, 9   , ..., 0   , 0   , 0   ],
        [1   , 514 , 904 , ..., 0   , 0   , 0   ],
        [1   , 47  , 10  , ..., 0   , 0   , 0   ],
        ...,
        [1   , 733 , 404 , ..., 0   , 0   , 0   ],
        [1   , 134 , 170 , ..., 0   , 0   , 0   ],
        [1   , 379 , 3122, ..., 0   , 0   , 0   ]]), Tensor(shape=[32, 38], dtype=int64, place=CUDAPinnedPlace, stop_gradient=True,
       [[0, 0, 0, ..., 0, 0, 0],
        [0, 0, 0, ..., 0, 0, 0],
        [0, 0, 0, ..., 0, 0, 0],
        ...,
        [0, 0, 0, ..., 0, 0, 0],
        [0, 0, 0, ..., 0, 0, 0],
        [0, 0, 0, ..., 0, 0, 0]])]

# 执行预测函数
y_probs = predict(model, predict_data_loader)

# 根据预测概率获取预测 label
y_preds = np.argmax(y_probs, axis=1)

输出预测结果

# 我们按照千言文本相似度竞赛的提交格式将预测结果存储在 lcqmc.tsv 中，用来后续提交
# 同时将预测结果输出到终端，便于大家直观感受模型预测效果

test_ds = load_dataset("lcqmc", splits=["test"])

with open("lcqmc.tsv", 'w', encoding="utf-8") as f:
    f.write("index\tprediction\n")    
    for idx, y_pred in enumerate(y_preds):
        f.write("{}\t{}\n".format(idx, y_pred))
        text_pair = test_ds[idx]
        text_pair["label"] = y_pred
        print(text_pair)

{'query': '推荐笔记本电脑', 'title': '笔记本电脑包推荐', 'label': 0}

: 0}

提交 LCQMC 预测结果千言文本相似度竞赛

千言文本相似度竞赛一共有 3 个数据集: lcqmc、bq_corpus、paws-x, 我们刚才生成了 lcqmc 的预测结果 lcqmc.tsv, 同时我们在项目内提供了 bq_corpus、paw-x 数据集的空预测结果，我们将这 3 个文件打包提交到千言文本相似度竞赛，即可看到自己的模型在 Lcqmc 数据集上的竞赛成绩。

# 打包预测结果
!zip submit.zip lcqmc.tsv paws-x.tsv bq_corpus.tsv

提交预测结果 submit.zip 到 千言文本相似度竞赛