takes requirements.txt file without module versions and annotates it with the latest set of module versions that won't result in build/runtime errors.

cool_script.sh

#!/bin/bash

# TODO: you will need to have conda installed

# create a python environment
conda create -n env_tmp python=3.8.13 -y

# activate environment
conda activate env_tmp

# download a sample for requirements.txt
wget https://raw.githubusercontent.com/delip/PyTorchNLPBook/master/requirements.txt

# install poetry
pip install poetry

# download a simple template for pyproject.toml (just so poetry can install the packages)
wget https://gist.githubusercontent.com/sudomaze/7298ffc4409032edd4d18a57b4c38f3a/raw/1c32efcbde31aaf896c6d47b32dac19ed44d14a4/pyproject.toml

# install dependencies from requirements.txt
cat requirements.txt | xargs poetry add

# export to requirements-like format
poetry export -f requirements.txt --output long_requirements.txt --without-hashes

# remove unwanted python versioning
cat long_requirements.txt | cut -d ";" -f 1 > with_dep_requirements.txt

# remove unwated dependencies
cat requirements.txt | while read line
  do echo $(grep -n $line'==' with_dep_requirements.txt | cut -d ":" -f 2 ) >> final_requirements.txt
done

# output
cat final_requirements.txt

final_requirements.txt

tqdm==4.64.0
matplotlib==3.5.3
nltk==3.7
jupyter==1.0.0
pandas==1.4.3
seaborn==0.11.2
scipy==1.9.0
annoy==1.17.1
numpy==1.23.2
requests==2.28.1

requirements.txt

tqdm
matplotlib
nltk
jupyter
pandas
seaborn
scipy
annoy
numpy
requests

I developed a package out of this script: https://github.com/sudomaze/pyreqpp

It is currently working for the explicit calling of the package: python -m pyreqpp, where requirements.txt is at the same level as your root directory.

Integrating with pre-commit and publishing it is Work-In-Progress.