pipenv - Pipfile.lock is not being generated due to the 'Could not find a version that matches keras-nightly~=2.5.0.dev' error

user_282855 0
Viewed 1383

As the title clearly describes the issue I've been experiencing, no Pipfile.lock is being generated as I get the following error when I execute the recommended command pipenv lock --clear:

ERROR: ERROR: Could not find a version that matches keras-nightly~=2.5.0.dev
Skipped pre-versions: 2.5.0.dev2021020510, 2.5.0.dev2021020600, 2.5.0.dev2021020700, 2.5.0.dev2021020800, 2.5.0.dev2021020900, 2.5.0.dev2021021000, 2.5.0.dev2021021100, 2.5.0.dev2021021200, 2.5.0.dev2021021300, 2.5.0.dev2021021400, 2.5.0.dev2021021500, 2.5.0.dev2021021600, 2.5.0.dev2021021700, 2.5.0.dev2021021800, 2.5.0.dev2021021900, 2.5.0.dev2021022000, 2.5.0.dev2021022100, 2.5.0.dev2021022200, 2.5.0.dev2021022300, 2.5.0.dev2021022317, 2.5.0.dev2021022400, 2.5.0.dev2021022411, 2.5.0.dev2021022500, 2.5.0.dev2021022600, 2.5.0.dev2021022700, 2.5.0.dev2021022800, 2.5.0.dev2021030100, 2.5.0.dev2021030200, 2.5.0.dev2021030300, 2.5.0.dev2021030400, 2.5.0.dev2021030500, 2.5.0.dev2021030600, 2.5.0.dev2021030700, 2.5.0.dev2021030800, 2.5.0.dev2021030900, 2.5.0.dev2021031000, 2.5.0.dev2021031100, 2.5.0.dev2021031200, 2.5.0.dev2021031300, 2.5.0.dev2021031400, 2.5.0.dev2021031500, 2.5.0.dev2021031600, 2.5.0.dev2021031700, 2.5.0.dev2021031800, 2.5.0.dev2021032213, 2.5.0.dev2021032300, 2.5.0.dev2021032413, 2.5.0.dev2021032500, 2.5.0.dev2021032600, 2.5.0.dev2021032610, 2.5.0.dev2021032700, 2.5.0.dev2021032800, 2.5.0.dev2021032900, 2.6.0.dev2021033000, 2.6.0.dev2021033100, 2.6.0.dev2021040100, 2.6.0.dev2021040200, 2.6.0.dev2021040300, 2.6.0.dev2021040400, 2.6.0.dev2021040500, 2.6.0.dev2021040600, 2.6.0.dev2021040714, 2.6.0.dev2021040800, 2.6.0.dev2021040900, 2.6.0.dev2021041000, 2.6.0.dev2021041100, 2.6.0.dev2021041200, 2.6.0.dev2021041300, 2.6.0.dev2021041400, 2.6.0.dev2021041500, 2.6.0.dev2021041600, 2.6.0.dev2021041700, 2.6.0.dev2021041800, 2.6.0.dev2021041900, 2.6.0.dev2021042000, 2.6.0.dev2021042100, 2.6.0.dev2021042200, 2.6.0.dev2021042300, 2.6.0.dev2021042500, 2.6.0.dev2021042600, 2.6.0.dev2021042700, 2.6.0.dev2021042800, 2.6.0.dev2021042900, 2.6.0.dev2021043000, 2.6.0.dev2021050100, 2.6.0.dev2021050200, 2.6.0.dev2021050300, 2.6.0.dev2021050400, 2.6.0.dev2021050500, 2.6.0.dev2021050600, 2.6.0.dev2021051200, 2.6.0.dev2021051300, 2.6.0.dev2021051400, 2.6.0.dev2021051500, 2.6.0.dev2021051600, 2.6.0.dev2021051700, 2.6.0.dev2021051800, 2.6.0.dev2021051900, 2.6.0.dev2021052000, 2.6.0.dev2021052100, 2.6.0.dev2021052200, 2.6.0.dev2021052300, 2.6.0.dev2021052400, 2.6.0.dev2021052500, 2.6.0.dev2021052600, 2.6.0.dev2021052700
There are incompatible versions in the resolved dependencies.

So, how can I overcome this situation? I'm basically developing a deep neural network using Keras. I simply installed the following dependencies without explicitly declaring versions:

tensorflow = "*"
nltk = "*"
pandas = "*"
tweepy = "*"
textblob = "*"
seaborn = "*"
matplotlib = "*"
wordcloud = "*"
stop-words = "*"
vadersentiment = "*"
scikit-learn = "*"
keras = "*"
pipenv tf.keras keras tensorflow pip
6 Answers

By looking at the pypi site for keras-nightly library, I could see that there are no versions named 2.5.0.dev. Check which package is generating the error and try downgrading that package.

user_143089390

Pipenv wants you to use the --pre flag in this situation. I think the simplest way to install it would be like this:

pipenv install tensorflow --python=3.8 --pre
user_6979640

I had the same problem. Apparently, tensorflow arises a sub-dependency resolution error with Pipfile when resolving keras. It searches for a version that matches keras-nightly~=2.5.0.dev. Looking at pypi, this version does not exist.

Whatsoever, installing a version 2.4.X works, e. g. pipenv install tensorflow~=2.4.1.

user_93783840

By Changing the python_version in pipenv file resolved my issue.

I had it 3.9 before.

Pipfile:

python_version = "3.8"
user_42863270

It seems that tensorflow does not work on python 3.9 as it has a dependency on the nightly.

If you are using python 3.9, your option is to go to tensorflow 2.6.0rc1,

user_2967560

adding those line to Pipfile fixed my issue.

[pipenv]
allow_prereleases = true
user_75738590
Related