Python : Importing ipynb files (Jupyter Notebook) from an ipynb file

This post shows how to import a Jupyter Notebook (ipynb) file from another Jupyter Notebook file. It will avoid occasional mistakes and save time to write redundant common codes such as importing library, declaring user-defined functions, data and its preprocessing, to name a few.

Importing an Jupyter Notebook from another Jupyter Notebook

To import an ipynb file in another ipynb file, we need to install import-ipynb python package which is apapted exactly to our purpose.

Install import-ipynb library from the command prompt !pip install import-ipynb Import it from your notebook import import_ipynb Import your BBB.ipynb notebook as if it was BBB.py file from BBB import *

Sample code as a whole : a_simple_rnn.ipynb

A sample code is a deep learning model using the SimpleRNN model which consists of including package libraries, loading and preprocessing data, setting up model, fitting and prediction. This is a whole file which will be divided into two files in the next.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86

# Import and Install Library   # In[1]: _________________________________________________________________ import pandas as pd import numpy as np import matplotlib.pyplot as plt from keras.models import Sequential from keras.layers import Dense, SimpleRNN get_ipython().run_line_magic('matplotlib', 'inline') from IPython.core.interactiveshell import InteractiveShell InteractiveShell.ast_node_interactivity = "all" _________________________________________________________________   # ## Functions # In[2]: _________________________________________________________________ # convert into dataset matrix def convertToMatrix(data, step):     X, Y =[], []     for i in range(len(data)-step):         d=i+step; X.append(data[i:d,]); Y.append(data[d,])     return np.array(X), np.array(Y)   def draw_plot1(df,predicted):     index = df.index.values     plt.figure(figsize=(5, 2.5))     plt.plot(index,df); plt.plot(index,predicted)     plt.show()     return plt _________________________________________________________________   # ## Load Dataset # In[3]: _________________________________________________________________ step = 4; N = 1000; Tp = 800     t=np.arange(0,N) x=np.sin(0.02*t)+2*np.random.rand(N) df = pd.DataFrame(x) # df.head(); plt.plot(df); plt.show() train=df.values train = np.append(train,np.repeat(train[-1,],step)) trainX,trainY = convertToMatrix(train,step) trainX = np.reshape(trainX, (trainX.shape[0], 1, trainX.shape[1])) _________________________________________________________________   # ## Building Model # In[4]: _________________________________________________________________ model = Sequential() model.add(SimpleRNN(units=32, input_shape=(1,step), activation="relu")) model.add(Dense(8, activation="relu"))  model.add(Dense(1)) model.compile(loss='mean_squared_error', optimizer='rmsprop') model.summary() _________________________________________________________________ Model: "sequential" _________________________________________________________________  Layer (type)                Output Shape              Param #    =================================================================  simple_rnn (SimpleRNN)      (None, 32)                1184                                                                          dense (Dense)               (None, 8)                 264                                                                           dense_1 (Dense)             (None, 1)                 9                                                                            ================================================================= Total params: 1,457 Trainable params: 1,457 Non-trainable params: 0 _________________________________________________________________   # ## Training Model # In[5]: _________________________________________________________________ model.fit(trainX, trainY, epochs=100, batch_size=16, verbose=0) _________________________________________________________________   # In[6]: _________________________________________________________________ trainPredict = model.predict(trainX) trainScore = model.evaluate(trainX, trainY, verbose=0) print(trainScore) draw_plot1(df,trainPredict) _________________________________________________________________

1) Common code block : a0_load_lib_data_func.ipynb

The following python code in the Jupyter notebook (a0_load_lib_data_func.ipynb) contains package libraries, some user-defined functions, and data. This file will be imported in each mode files.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46

# Import and Install Library   # In[1]: _________________________________________________________________ import pandas as pd import numpy as np import matplotlib.pyplot as plt from keras.models import Sequential from keras.layers import Dense, SimpleRNN get_ipython().run_line_magic('matplotlib', 'inline') from IPython.core.interactiveshell import InteractiveShell InteractiveShell.ast_node_interactivity = "all" _________________________________________________________________   # ## Functions # In[2]: _________________________________________________________________ # convert into dataset matrix def convertToMatrix(data, step):     X, Y =[], []     for i in range(len(data)-step):         d=i+step; X.append(data[i:d,]); Y.append(data[d,])     return np.array(X), np.array(Y)   def draw_plot1(df,predicted):     index = df.index.values     plt.figure(figsize=(5, 2.5))     plt.plot(index,df); plt.plot(index,predicted)     plt.show()     return plt _________________________________________________________________   # ## Load Dataset # In[3]: _________________________________________________________________ step = 4; N = 1000; Tp = 800     t=np.arange(0,N) x=np.sin(0.02*t)+2*np.random.rand(N) df = pd.DataFrame(x) # df.head(); plt.plot(df); plt.show() train=df.values train = np.append(train,np.repeat(train[-1,],step)) trainX,trainY = convertToMatrix(train,step) trainX = np.reshape(trainX, (trainX.shape[0], 1, trainX.shape[1])) _________________________________________________________________   Colored by Color Scripter

cs

2) Each model-specific file : a1_simple_rnn_wo_lib_data_func.ipynb

The following python code (a1_simple_rnn_wo_lib_data_func.ipynb) imports the above Jupyter Notebook file (a0_load_lib_data_func.ipynb). As the imported file contains common code blocks, this file does not contain these redundant information but has the content of each specific model and routines for forecasting performance comparisons.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53

# Import a0_read_lib_data_func.ipynb file # In[1]: _________________________________________________________________ # install if not installed #!pip install import-ipynb   # import it from your notebook import import_ipynb   # import a0_load_lib_data_func notebook from a0_load_lib_data_func import * _________________________________________________________________   # ## Building Model # In[4]: _________________________________________________________________ model = Sequential() model.add(SimpleRNN(units=32, input_shape=(1,step), activation="relu")) model.add(Dense(8, activation="relu"))  model.add(Dense(1)) model.compile(loss='mean_squared_error', optimizer='rmsprop') model.summary() _________________________________________________________________ Model: "sequential" -----------------------------------------------------------------  Layer (type)                Output Shape              Param #    =================================================================  simple_rnn (SimpleRNN)      (None, 32)                1184                                                                          dense (Dense)               (None, 8)                 264                                                                           dense_1 (Dense)             (None, 1)                 9                                                                            ================================================================= Total params: 1,457 Trainable params: 1,457 Non-trainable params: 0 ----------------------------------------------------------------- # ## Training Model # In[5]: _________________________________________________________________ model.fit(trainX, trainY, epochs=100, batch_size=16, verbose=0) _________________________________________________________________ <keras.callbacks.History at 0x22e182f8c70> # In[6]: _________________________________________________________________ trainPredict = model.predict(trainX) trainScore = model.evaluate(trainX, trainY, verbose=0) print(trainScore) draw_plot1(df,trainPredict) _________________________________________________________________ 0.3513454794883728   Colored by Color Scripter

cs

Concluding Remarks

This post shows how to use the common code block file which is imported in another model-specific file in Jupyter notebook. Running one file as a whole and running two sparate files deliver the same output. But the latter will be useful since it can avoid redundant copy-and-paste works. \(\blacksquare\)