Requirements-frozen.txt and requirements2-frozen.txt in this example,
setup.py install ExamplesĬreate frozen versions of two requirements files (they will be named Install from pypi: $ pip install freeze-requirements With pip install -r requirements-frozen.txt -index-url, and also build wheels to speed up deployments even more. Then you can use the frozen requirements in your deployment scripts with pip install -r requirements-frozen.txt -no-deps, and enjoy consistentĭeployments even if some packages are updated on pypi.įreeze-requirements can also put the downloaded source packages in a pypi-likeĭirectory structure on your web server, so you can speed up your deployments # Frozen requirements for "requirements.txt" The frozen version would be: # This file has been automatically generated, DO NOT EDIT! Note that if you plan to use mostly Conda Forge, you'd be much better off dumping Anaconda in favor of a Miniforge install.A script to help creating and maintaining frozen requirements for pip, inspiredįrozen requirements contain the packages you specified, plus all theirįor example if you have requirements.txt containing this: pyramid Otherwise, if you really want to use Conda Forge's keras build and still want all the other Anaconda packages, then create a new env plus anaconda, similar to 1: conda create -n keras_env -c conda-forge anaconda keras If you really want keras in your base env, then get it from the defaults channel: conda install -n base kerasī. Moreover, because Conda can't satisfy the request with a frozen install (i.e., it has to change some versions of existing packages), it opens up many of your packages be reprioritized to the Conda Forge channel, not just Keras.
Mixing channels is generally not recommended (Conda Forge and Anaconda use different build stacks). conda create -n keras_env -c conda-forge keras Rather than trying to cram everything into base it is usually much less of a hassle to create new environments for each project or task.
There are a couple things going on here that make it extra hard for Conda to install this package:Īnaconda base env is huge and often requires reconfiguring packages in order to add new ones.
0 kommentar(er)
