Saving and loading keras model

Question

I'm working on Keras model which uses Universal Sentence Embedding to encode the provided sentences. However, when I save the model for future usage, the mentioned error is thrown. NameError: name 'embed' is not defined

The sentences are converted to embedding using UniversalEmbedding(x) function. The code of whole model is taken from this link .

!wget https://raw.githubusercontent.com/Tony607/Keras-Text-Transfer-Learning/master/train_5500.txt
!wget https://raw.githubusercontent.com/Tony607/Keras-Text-Transfer-Learning/master/test_data.txt

import tensorflow as tf
import tensorflow_hub as hub
import matplotlib.pyplot as plt
import numpy as np
import os
import pandas as pd
import re
import seaborn as sns
import keras.layers as layers
from keras.models import Model
from keras import backend as K
np.random.seed(10)

def get_dataframe(filename):
    lines = open(filename, 'r').read().splitlines()
    data = []
    for i in range(0, len(lines)):
        label = lines[i].split(' ')[0]
        label = label.split(":")[0]
        text = ' '.join(lines[i].split(' ')[1:])
        text = re.sub('[^A-Za-z0-9 ,\?\'\"-._\+\!/\`@=;:]+', '', text)
        data.append([label, text])

    df = pd.DataFrame(data, columns=['label', 'text'])
    df.label = df.label.astype('category')
    return df

df_train = get_dataframe('train_5500.txt')
df_train = get_dataframe('test_data.txt')

category_counts = len(df_train.label.cat.categories)
module_url = "https://tfhub.dev/google/universal-sentence-encoder-large/3" 
embed = hub.Module(module_url)
embed_size = embed.get_output_info_dict()['default'].get_shape()[1].value

def UniversalEmbedding(x):
    return embed(tf.squeeze(tf.cast(x, tf.string)), signature="default", as_dict=True)["default"]

input_text = layers.Input(shape=(1,), dtype='string')
embedding = layers.Lambda(UniversalEmbedding, output_shape=(embed_size,))(input_text)
dense = layers.Dense(256, activation='relu')(embedding)
pred = layers.Dense(category_counts, activation='softmax')(dense)
model = Model(inputs=[input_text], outputs=pred)
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])

train_text = df_train['text'].tolist()
train_text = np.array(train_text, dtype=object)[:, np.newaxis]

train_label = np.asarray(pd.get_dummies(df_train.label), dtype = np.int8)

df_test = get_dataframe('test_data.txt')
test_text = df_test['text'].tolist()
test_text = np.array(test_text, dtype=object)[:, np.newaxis]
test_label = np.asarray(pd.get_dummies(df_test.label), dtype = np.int8)


with tf.Session() as session:
  K.set_session(session)
  session.run(tf.global_variables_initializer())
  session.run(tf.tables_initializer())
  history = model.fit(train_text, 
            train_label,
            validation_data=(test_text, test_label),
            epochs=2,
            batch_size=32)
  model.save_weights('./model.h5')
  model.save('mod.h5')

When I try to load the model like

from keras.models import load_model

load_model('mod.h5') 

Answer 1

When you are trying to load the model using keras' load_model, the method shows error because embed is not in-built in keras and to solve this you might have to define it again in your code before loading the model using load_model.

Please try to write embed = hub.Module(module_url) along with its required libraries and urls as it is in your provided link ( https://www.dlology.com/blog/keras-meets-universal-sentence-encoder-transfer-learning-for-text-data/ ) before trying to load the model.