0

When i call pip freeze > requirements.txt in VScode, it also adds libraries that i didn't import to my project. I have these libraries in my Jupyter notebook:

   import numpy as np
   import pandas as pd
   import matplotlib.pyplot as plt 
   import seaborn as sns 
   import scipy.stats as stats
   from sklearn.compose import ColumnTransformer
   from sklearn.pipeline import Pipeline
   from sklearn.model_selection import GridSearchCV,KFold
   from sklearn.metrics import r2_score,mean_squared_error,mean_absolute_error
   from sklearn.preprocessing import LabelEncoder,OneHotEncoder,OrdinalEncoder,StandardScaler,MinMaxScaler
   from sklearn.model_selection import train_test_split,cross_val_score
   from sklearn.ensemble import AdaBoostRegressor,GradientBoostingRegressor,RandomForestRegressor
   from catboost import CatBoostRegressor
   from xgboost import XGBRegressor
   import pickle
   import warnings

How can i create requirements.txt file with only imported libraries?

image link: https://i.stack.imgur.com/sD1IX.jpg

Muhammad Ojagzada
  • 17
  • 5
  • Although it's not impossible to do what you ask, do you realise that the other libraries are likely there because the libraries you *do* import require them? (and import them in turn?) – Grismar Jul 31 '22 at 11:55
  • 1
    [Why you should not upload code/data/errors as images](https://meta.stackoverflow.com/questions/285551/why-should-i-not-upload-images-of-code-data-errors-when-asking-a-question) – Grismar Jul 31 '22 at 11:55
  • @Grismar I have almost the same project with this guy but as you see his requirements file, it only contains main libraries. https://github.com/RauhanAhmed/Ineuron_internship/blob/main/requirements.txt – Muhammad Ojagzada Jul 31 '22 at 12:14

1 Answers1

2

Usually better approach, than using pip freeze, is to construct the requirements file manually, as freezing requirements will also include requirements of requirements, etc, which can easily change without you knowing. Within your requirements file you should be also specifying version ranges, like

cachetools>=5.0,<6.0

which is nearly impossible to manage with a huge number of sub-requirements flushed from pip freeze.

See how it is done in any big open source python libraries like click, pillow or rich. Most libraries use semantic versioning which has really straightforward rules about version tags.

EDIT:

There happens to be a really handy tool, called poetry, especially for bigger projects (but might be useful also for smaller ones too If you wish to pay enough attention to configuring it). I'm aware that it doesn't fully match the topic, but this tool is great when it comes to dependencies, I hope it will help anyone.

Christopher
  • 306
  • 1
  • 6