MatchZoo简单使用

杜绍元

2023-12-01

简介

最近在调研文本相似度计算方法的时候，突然看到有关MatchZoo有关的内容，MatchZoo 是一个通用的文本匹配工具包，它旨在方便大家快速的实现、比较、以及分享最新的深度文本匹配模型，貌似比较好玩，于是乎，看了一下MatchZoo的使用方法。在此简单记录一下我的使用过程，运行totorials里面的示例代码。如果之前没接触过MatchZoo，直接看github中README.md中的示例代码可能有点云里雾里，不过等看过totorial中的代码之后，才发现，那是MatchZoo的精髓。我在使用的时候是2019.11，随着版本的跟新，我这篇博客可能也就过时了，因为在我查找资料的时候，很多博客中讲解的已经与我实际测试的内容大相径庭。

安装

在github上也提到了，对于MatchZoo中包含了两种安装方式：

Pypi 安装
```
pip install matchzoo
```

github source 安装

git clone https://github.com/NTMC-Community/MatchZoo.git      
cd MatchZoo    
python setup.py install

模型

在模型中，目前给出了很多模型。但是对于不同的模型，由于做了封装，使用起来比较简单，主要分两步：第一，创建模型；第二，创建模型参数。这个在接下来的步骤中可以比较清楚的看到。

运行Quick Start

切换到MatchZoo-master\tutorials目录下（我是源码安装的，因此会有这个文件夹，pypi安装时，也会有，不过路径不是这个，可以直接搜quick_start.ipynb，去找当前文件的位置）。运行一下命令，启动jupyter notebook
```
jupyter notebook
```
如果，机器上没有安装jupyter或者之前没有接触过，可参考jupyter notebook简介

流程
a. 定义任务
b. 准备数据
c. 数据预处理
d. 创建模型
e. 训练评估
f. 预测
下面的这段代码则是展示了，使用matchzoo运行的简单的demo，包含了上面的demo(这个demo是从教程里抽取出来的).

import matchzoo as mz
print(mz.__version__)

### 定义任务，包含两种，一个是Ranking，一个是classification
task = mz.tasks.Ranking()
print(task)

### 准备数据，数据在源码中有，不确定在pip安装的是否存在
### train_raw是matchzoo中自定的数据格式	matchzoo.data_pack.data_pack.DataPack
train_raw = mz.datasets.toy.load_data(stage='train', task=task)
test_raw = mz.datasets.toy.load_data(stage='test', task=task)

### 数据预处理，BasicPreprocessor为指定预处理的方式，在预处理中包含了两步：fit,transform
### fit将收集一些有用的信息到preprocessor.context中，不会对输入DataPack进行处理
### transformer 不会改变context、DataPack,他将重新生成转变后的DataPack.
### 在transformer过程中，包含了Tokenize => Lowercase => PuncRemoval等过程，这个过程在方法中应该是可以自定义的
preprocessor = mz.preprocessors.BasicPreprocessor()
preprocessor.fit(train_raw)  ## init preprocessor inner state.
train_processed = preprocessor.transform(train_raw)
test_processed = preprocessor.transform(test_raw)

### 创建模型以及修改参数（可以使用mz.models.list_available()查看可用的模型列表）
model = mz.models.DenseBaseline()
model.params['task'] = task
model.params['mlp_num_units'] = 3
model.params.update(preprocessor.context)
model.params.completed()
model.build()
model.compile()
model.backend.summary()

### 训练, 评估, 预测
x, y = train_processed.unpack()
test_x, test_y = test_processed.unpack()
model.fit(x , y,batch_size=32, epochs=5)
model.evaluate(test_x,test_y)
model.predict(test_x)

### 保存模型
model.save('my-model')
loaded_model = mz.load_model('my-model')

通过jupyter notebook展示运行效果
示例中提供的代码不止包含上面的内容，还包含了数据结构，不同的读取方式，为了方便观察，我把我这边使用jupyter notebook运行代码的效果展示出来，同样的，大家也可以运行一下试试。

MatchZoo Quick Start

import matchzoo as mz
print(mz.__version__)

d:\software\python\python37\lib\site-packages\tqdm\std.py:651: FutureWarning: The Panel class is removed from pandas. Accessing it from the top-level namespace will also be removed in the next version
from pandas import Panel
Using TensorFlow backend.
d:\software\python\python37\lib\site-packages\tqdm\std.py:651: FutureWarning: The Panel class is removed from pandas. Accessing it from the top-level namespace will also be removed in the next version
from pandas import Panel

2.2.0

Define Task

There are two types of tasks available in MatchZoo. mz.tasks.Ranking and mz.tasks.Classification. We will use a ranking task for this demo.

task = mz.tasks.Ranking()
print(task)

Ranking Task

Prepare Data

train_raw = mz.datasets.toy.load_data(stage='train', task=task)
test_raw = mz.datasets.toy.load_data(stage='test', task=task)

type(train_raw)

matchzoo.data_pack.data_pack.DataPack

DataPack is a MatchZoo native data structure that most MatchZoo data handling processes build upon. A DataPack is consists of three pandas.DataFrame:

train_raw.left.head()

	text_left
id_left
Q1	how are glacier caves formed?
Q2	How are the directions of the velocity and for...
Q5	how did apollo creed die
Q6	how long is the term for federal judges
Q7	how a beretta model 21 pistols magazines works

train_raw.right.head()

	text_right
id_right
D1-0	A partly submerged glacier cave on Perito More...
D1-1	The ice facade is approximately 60 m high
D1-2	Ice formations in the Titlis glacier cave
D1-3	A glacier cave is a cave formed within the ice...
D1-4	Glacier caves are often called ice caves , but...

train_raw.relation.head()

	id_left	id_right	label
0	Q1	D1-0	0.0
1	Q1	D1-1	0.0
2	Q1	D1-2	0.0
3	Q1	D1-3	1.0
4	Q1	D1-4	0.0

It is also possible to convert a DataPack into a single pandas.DataFrame that holds all information.

train_raw.frame().head()

	id_left	text_left	id_right	text_right	label
0	Q1	how are glacier caves formed?	D1-0	A partly submerged glacier cave on Perito More...	0.0
1	Q1	how are glacier caves formed?	D1-1	The ice facade is approximately 60 m high	0.0
2	Q1	how are glacier caves formed?	D1-2	Ice formations in the Titlis glacier cave	0.0
3	Q1	how are glacier caves formed?	D1-3	A glacier cave is a cave formed within the ice...	1.0
4	Q1	how are glacier caves formed?	D1-4	Glacier caves are often called ice caves , but...	0.0

However, using such pandas.DataFrame consumes much more memory if there are many duplicates in the texts, and that is the exact reason why we use DataPack. For more details about data handling, consult matchzoo/tutorials/data_handling.ipynb.

Preprocessing

MatchZoo preprocessors are used to convert a raw DataPack into a DataPack that ready to be fed into a model.

preprocessor = mz.preprocessors.BasicPreprocessor()

There are two steps to use a preprocessor. First, fit. Then, transform. fit will only changes the preprocessor’s inner state but not the input DataPack.

preprocessor.fit(train_raw)

Processing text_left with chain_transform of Tokenize => Lowercase => PuncRemoval: 100%|█| 13/13 [00:00<00:00, 861.97it/s]
Processing text_right with chain_transform of Tokenize => Lowercase => PuncRemoval: 100%|█| 100/100 [00:00<00:00, 1558.32it/s]
Processing text_right with append: 100%|██████████████████████████████████████████| 100/100 [00:00<00:00, 99674.52it/s]
Building FrequencyFilter from a datapack.: 100%|██████████████████████████████████| 100/100 [00:00<00:00, 99864.38it/s]
Processing text_right with transform: 100%|███████████████████████████████████████| 100/100 [00:00<00:00, 99769.36it/s]
Processing text_left with extend: 100%|█████████████████████████████████████████████| 13/13 [00:00<00:00, 12908.61it/s]
Processing text_right with extend: 100%|██████████████████████████████████████████| 100/100 [00:00<00:00, 98782.48it/s]
Building Vocabulary from a datapack.: 100%|████████████████████████████████████| 1974/1974 [00:00<00:00, 983904.47it/s]





<matchzoo.preprocessors.basic_preprocessor.BasicPreprocessor at 0x24aad5d17b8>

fit will gather useful information into its context, which will be used later in a transform or used to set hyper-parameters of a model.

preprocessor.context

{'filter_unit': <matchzoo.preprocessors.units.frequency_filter.FrequencyFilter at 0x24aad5d1e80>,
 'vocab_unit': <matchzoo.preprocessors.units.vocabulary.Vocabulary at 0x24aad771b38>,
 'vocab_size': 289,
 'embedding_input_dim': 289,
 'input_shapes': [(30,), (30,)]}

Once fit, the preprocessor has enough information to transform. transform will not change the preprocessor’s inner state and the input DataPack, but return a transformed DataPack.

train_processed = preprocessor.transform(train_raw)
test_processed = preprocessor.transform(test_raw)

Processing text_left with chain_transform of Tokenize => Lowercase => PuncRemoval: 100%|█| 13/13 [00:00<00:00, 4323.00it/s]
Processing text_right with chain_transform of Tokenize => Lowercase => PuncRemoval: 100%|█| 100/100 [00:00<00:00, 2695.45it/s]
Processing text_right with transform: 100%|███████████████████████████████████████| 100/100 [00:00<00:00, 19946.28it/s]
Processing text_left with transform: 100%|███████████████████████████████████████████| 13/13 [00:00<00:00, 4339.17it/s]
Processing text_right with transform: 100%|███████████████████████████████████████| 100/100 [00:00<00:00, 66313.11it/s]
Processing length_left with len: 100%|██████████████████████████████████████████████| 13/13 [00:00<00:00, 12960.77it/s]
Processing length_right with len: 100%|███████████████████████████████████████████| 100/100 [00:00<00:00, 99698.22it/s]
Processing text_left with transform: 100%|███████████████████████████████████████████| 13/13 [00:00<00:00, 2592.52it/s]
Processing text_right with transform: 100%|███████████████████████████████████████| 100/100 [00:00<00:00, 33238.01it/s]
Processing text_left with chain_transform of Tokenize => Lowercase => PuncRemoval: 100%|█| 3/3 [00:00<00:00, 1495.83it/s]
Processing text_right with chain_transform of Tokenize => Lowercase => PuncRemoval: 100%|█| 20/20 [00:00<00:00, 1813.91it/s]
Processing text_right with transform: 100%|█████████████████████████████████████████| 20/20 [00:00<00:00, 19934.90it/s]
Processing text_left with transform: 100%|█████████████████████████████████████████████| 3/3 [00:00<00:00, 2991.66it/s]
Processing text_right with transform: 100%|█████████████████████████████████████████| 20/20 [00:00<00:00, 19934.90it/s]
Processing length_left with len: 100%|█████████████████████████████████████████████████| 3/3 [00:00<00:00, 2988.82it/s]
Processing length_right with len: 100%|████████████████████████████████████████████████████████| 20/20 [00:00<?, ?it/s]
Processing text_left with transform: 100%|█████████████████████████████████████████████| 3/3 [00:00<00:00, 2992.37it/s]
Processing text_right with transform: 100%|█████████████████████████████████████████| 20/20 [00:00<00:00, 19930.17it/s]

train_processed.left.head()

	text_left	length_left
id_left
Q1	[111, 265, 176, 216, 217, 2, 0, 0, 0, 0, 0, 0,...	6
Q2	[111, 265, 98, 223, 248, 98, 37, 60, 203, 143,...	15
Q5	[111, 145, 283, 251, 69, 0, 0, 0, 0, 0, 0, 0, ...	5
Q6	[111, 13, 270, 98, 272, 227, 47, 87, 0, 0, 0, ...	8
Q7	[111, 219, 56, 108, 122, 14, 230, 63, 0, 0, 0,...	8

As we can see, text_left is already in sequence form that nerual networks love.

Just to make sure we have the correct sequence:

vocab_unit = preprocessor.context['vocab_unit']
print('Orig Text:', train_processed.left.loc['Q1']['text_left'])
sequence = train_processed.left.loc['Q1']['text_left']
print('Transformed Indices:', sequence)
print('Transformed Indices Meaning:',
      '_'.join([vocab_unit.state['index_term'][i] for i in sequence]))

Orig Text: [111, 265, 176, 216, 217, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
Transformed Indices: [111, 265, 176, 216, 217, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
Transformed Indices Meaning: how_are_glacier_caves_formed__<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>_<PAD>

For more details about preprocessing, consult matchzoo/tutorials/data_handling.ipynb.

Build Model

MatchZoo provides many built-in text matching models.

mz.models.list_available()

[matchzoo.models.naive.Naive,
 matchzoo.models.dssm.DSSM,
 matchzoo.models.cdssm.CDSSM,
 matchzoo.models.dense_baseline.DenseBaseline,
 matchzoo.models.arci.ArcI,
 matchzoo.models.arcii.ArcII,
 matchzoo.models.match_pyramid.MatchPyramid,
 matchzoo.models.knrm.KNRM,
 matchzoo.models.duet.DUET,
 matchzoo.models.drmmtks.DRMMTKS,
 matchzoo.models.drmm.DRMM,
 matchzoo.models.anmm.ANMM,
 matchzoo.models.mvlstm.MVLSTM,
 matchzoo.contrib.models.match_lstm.MatchLSTM,
 matchzoo.contrib.models.match_srnn.MatchSRNN,
 matchzoo.contrib.models.hbmp.HBMP,
 matchzoo.contrib.models.esim.ESIM,
 matchzoo.contrib.models.bimpm.BiMPM,
 matchzoo.contrib.models.diin.DIIN,
 matchzoo.models.conv_knrm.ConvKNRM]

Let’s use mz.models.DenseBaseline for our demo.

model = mz.models.DenseBaseline()

The model is initialized with a hyper parameter table, in which values are partially filled. To view parameters and their values, use print.

print(model.params)

model_class                   <class 'matchzoo.models.dense_baseline.DenseBaseline'>
input_shapes                  None
task                          None
optimizer                     adam
with_multi_layer_perceptron   True
mlp_num_units                 256
mlp_num_layers                3
mlp_num_fan_out               64
mlp_activation_func           relu

to_frame gives you more informartion in addition to just names and values.

model.params.to_frame()[['Name', 'Description', 'Value']]

	Name	Description	Value
0	model_class	Model class. Used internally for save/load. Ch...	<class 'matchzoo.models.dense_baseline.DenseBa...
1	input_shapes	Dependent on the model and data. Should be set...	None
2	task	Decides model output shape, loss, and metrics.	None
3	optimizer	None	adam
4	with_multi_layer_perceptron	A flag of whether a multiple layer perceptron ...	True
5	mlp_num_units	Number of units in first `mlp_num_layers` layers.	256
6	mlp_num_layers	Number of layers of the multiple layer percetron.	3
7	mlp_num_fan_out	Number of units of the layer that connects the...	64
8	mlp_activation_func	Activation function used in the multiple layer...	relu

To set a hyper-parameter:

model.params['task'] = task
model.params['mlp_num_units'] = 3
print(model.params)

model_class                   <class 'matchzoo.models.dense_baseline.DenseBaseline'>
input_shapes                  None
task                          Ranking Task
optimizer                     adam
with_multi_layer_perceptron   True
mlp_num_units                 3
mlp_num_layers                3
mlp_num_fan_out               64
mlp_activation_func           relu

Notice that we are still missing input_shapes, and that information is store in the preprocessor.

print(preprocessor.context['input_shapes'])

[(30,), (30,)]

We may use update to load a preprocessor’s context into a model’s hyper-parameter table.

model.params.update(preprocessor.context)

Now we have a completed hyper-parameter table.

model.params.completed()

True

With all parameters filled in, we can now build and compile the model.

model.build()
model.compile()

MatchZoo models are wrapped over keras models, and the backend property of a model gives you the actual keras model built.

model.backend.summary()

Model: "model_1"
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
text_left (InputLayer)          (None, 30)           0                                            
__________________________________________________________________________________________________
text_right (InputLayer)         (None, 30)           0                                            
__________________________________________________________________________________________________
concatenate_1 (Concatenate)     (None, 60)           0           text_left[0][0]                  
                                                                 text_right[0][0]                 
__________________________________________________________________________________________________
dense_1 (Dense)                 (None, 3)            183         concatenate_1[0][0]              
__________________________________________________________________________________________________
dense_2 (Dense)                 (None, 3)            12          dense_1[0][0]                    
__________________________________________________________________________________________________
dense_3 (Dense)                 (None, 3)            12          dense_2[0][0]                    
__________________________________________________________________________________________________
dense_4 (Dense)                 (None, 64)           256         dense_3[0][0]                    
__________________________________________________________________________________________________
dense_5 (Dense)                 (None, 1)            65          dense_4[0][0]                    
==================================================================================================
Total params: 528
Trainable params: 528
Non-trainable params: 0
__________________________________________________________________________________________________

For more details about models, consult matchzoo/tutorials/models.ipynb.

Train, Evaluate, Predict

A DataPack can unpack itself into data that can be directly used to train a MatchZoo model.

x, y = train_processed.unpack()
test_x, test_y = test_processed.unpack()

model.fit(x, y, batch_size=32, epochs=5)

WARNING: Logging before flag parsing goes to stderr.
W1108 15:57:07.798936  9460 deprecation_wrapper.py:119] From d:\software\python\python37\lib\site-packages\keras\backend\tensorflow_backend.py:422: The name tf.global_variables is deprecated. Please use tf.compat.v1.global_variables instead.



Epoch 1/5
100/100 [==============================] - ETA: 0s - loss: 0.033 - 0s 3ms/step - loss: 0.0963
Epoch 2/5
100/100 [==============================] - ETA: 0s - loss: 0.064 - 0s 160us/step - loss: 0.0606
Epoch 3/5
100/100 [==============================] - ETA: 0s - loss: 0.068 - 0s 170us/step - loss: 0.0522
Epoch 4/5
100/100 [==============================] - ETA: 0s - loss: 0.001 - 0s 170us/step - loss: 0.0499
Epoch 5/5
100/100 [==============================] - ETA: 0s - loss: 0.032 - 0s 221us/step - loss: 0.0495





<keras.callbacks.callbacks.History at 0x24aad760b38>

An alternative to train a model is to use a DataGenerator. This is useful for delaying expensive preprocessing steps or doing real-time data augmentation. For some models that needs dynamic batch-wise information, using a DataGenerator is required. For more details about DataGenerator, consult matchzoo/tutorials/data_handling.ipynb.

data_generator = mz.DataGenerator(train_processed, batch_size=32)

model.fit_generator(data_generator, epochs=5, use_multiprocessing=True, workers=4)

Epoch 1/5
4/4 [==============================] - ETA: 55s - loss: 0.09 - 19s 5s/step - loss: 0.0406
Epoch 2/5
4/4 [==============================] - ETA: 57s - loss: 2.3547e- - 19s 5s/step - loss: 0.0403 
Epoch 3/5
4/4 [==============================] - ETA: 58s - loss: 0.12 - 19s 5s/step - loss: 0.0402
Epoch 4/5
4/4 [==============================] - ETA: 1:06 - loss: 0.060 - 22s 6s/step - loss: 0.0401
Epoch 5/5
4/4 [==============================] - ETA: 54s - loss: 0.03 - 18s 5s/step - loss: 0.0818





<keras.callbacks.callbacks.History at 0x24aade40940>

model.evaluate(test_x, test_y)

{mean_average_precision(0.0): 0.1111111111111111}

model.predict(test_x)

array([[0.02392608],
       [0.02392608],
       [0.02392608],
       [0.02392608],
       [0.32358152],
       [0.02392608],
       [0.02392608],
       [0.02392608],
       [0.02392608],
       [0.02392608],
       [0.02392608],
       [0.02392608],
       [0.02392608],
       [0.02392608],
       [0.02392608],
       [0.02392608],
       [0.02392608],
       [0.02392608],
       [0.02392608],
       [0.02392608]], dtype=float32)

A Shortcut to Preprocessing and Model Building

Since data preprocessing and model building are laborious and special setups of some models makes this even worse, MatchZoo provides prepare, a unified interface that handles interaction among data, model, and preprocessor automatically.

More specifically, prepare does these following things:

create a default preprocessor of the model class (if not given one)
fit the preprocessor using the raw data
create an embedding matrix
instantiate a model and fill in hype-parameters
build the model
instantiate a DataGeneratorBuilder that will build a correctly formed DataGenerator given a DataPack

It also does many special handling for specific models, but we will not go into the details of that here.

for model_class in mz.models.list_available():
    print(model_class)
    model, preprocessor, data_generator_builder, embedding_matrix = mz.auto.prepare(
        task=task,
        model_class=model_class,
        data_pack=train_raw,
    )
    train_processed = preprocessor.transform(train_raw, verbose=0)
    test_processed = preprocessor.transform(test_raw, verbose=0)
    train_gen = data_generator_builder.build(train_processed)
    test_gen = data_generator_builder.build(test_processed)
    model.fit_generator(train_gen, epochs=1)
    model.evaluate_generator(test_gen)
    print()

<class 'matchzoo.models.naive.Naive'>
Epoch 1/1
1/1 [==============================] - 0s 148ms/step - loss: 8960.0771

<class 'matchzoo.models.dssm.DSSM'>


W1108 16:01:02.558478  9460 deprecation.py:323] From d:\software\python\python37\lib\site-packages\tensorflow\python\ops\math_grad.py:1250: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.where in 2.0, which has the same broadcast rule as np.where


Epoch 1/1
1/1 [==============================] - 0s 482ms/step - loss: 0.1887

<class 'matchzoo.models.cdssm.CDSSM'>
Epoch 1/1
1/1 [==============================] - 2s 2s/step - loss: 0.0801

<class 'matchzoo.models.dense_baseline.DenseBaseline'>
Epoch 1/1
1/1 [==============================] - 0s 303ms/step - loss: 1091.7343

<class 'matchzoo.models.arci.ArcI'>


W1108 16:01:11.513350  9460 deprecation_wrapper.py:119] From d:\software\python\python37\lib\site-packages\keras\backend\tensorflow_backend.py:4070: The name tf.nn.max_pool is deprecated. Please use tf.nn.max_pool2d instead.



Epoch 1/1
1/1 [==============================] - 1s 919ms/step - loss: 0.0496

<class 'matchzoo.models.arcii.ArcII'>
Epoch 1/1
1/1 [==============================] - 1s 673ms/step - loss: 0.1104

<class 'matchzoo.models.match_pyramid.MatchPyramid'>
Epoch 1/1
1/1 [==============================] - 1s 660ms/step - loss: 0.0744

<class 'matchzoo.models.knrm.KNRM'>
Epoch 1/1
1/1 [==============================] - 1s 918ms/step - loss: 1183.9641

<class 'matchzoo.models.duet.DUET'>
Epoch 1/1
1/1 [==============================] - 1s 893ms/step - loss: 0.1633

<class 'matchzoo.models.drmmtks.DRMMTKS'>
Epoch 1/1
1/1 [==============================] - 1s 849ms/step - loss: 0.0546

<class 'matchzoo.models.drmm.DRMM'>
Epoch 1/1
1/1 [==============================] - 1s 1s/step - loss: 0.0681

<class 'matchzoo.models.anmm.ANMM'>
Epoch 1/1
1/1 [==============================] - 1s 653ms/step - loss: 0.0640

<class 'matchzoo.models.mvlstm.MVLSTM'>
Epoch 1/1
1/1 [==============================] - 3s 3s/step - loss: 0.0495

<class 'matchzoo.contrib.models.match_lstm.MatchLSTM'>
Epoch 1/1
1/1 [==============================] - 5s 5s/step - loss: 0.0681

<class 'matchzoo.contrib.models.match_srnn.MatchSRNN'>
Epoch 1/1
1/1 [==============================] - 3s 3s/step - loss: 0.0620

<class 'matchzoo.contrib.models.hbmp.HBMP'>
Epoch 1/1
1/1 [==============================] - 7s 7s/step - loss: 0.0519

<class 'matchzoo.contrib.models.esim.ESIM'>
Epoch 1/1
1/1 [==============================] - 5s 5s/step - loss: nan

<class 'matchzoo.contrib.models.bimpm.BiMPM'>
Epoch 1/1
1/1 [==============================] - 5s 5s/step - loss: 0.0494

<class 'matchzoo.contrib.models.diin.DIIN'>


W1108 16:03:36.396669  9460 deprecation.py:506] From d:\software\python\python37\lib\site-packages\matchzoo-2.2.0-py3.7.egg\matchzoo\contrib\layers\decaying_dropout_layer.py:87: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.
Instructions for updating:
Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.


Epoch 1/1
1/1 [==============================] - 4s 4s/step - loss: 0.0500

<class 'matchzoo.models.conv_knrm.ConvKNRM'>
Epoch 1/1
1/1 [==============================] - 10s 10s/step - loss: 1611.8352

Save and Load the Model

model.save('my-model')
loaded_model = mz.load_model('my-model')