Protenix Conda Environment

[rjrich]

To install protenix version 0.4.6 in a conda environment in Linux, if I created the environment and then used pip directly to install protenix, this failed with many errors, mainly indicating version conflicts.

However, I finally was able to install protenix 0.4.6 successfully using the following steps:

~/Protenix$ conda create --name protenix --file requirements.txt

Note: this requirements.txt file did not include the following:
ml_collections
nvidia-cublas-cu12
torch==2.3.1
The above 3 packages are installed later using pip

~/Protenix$ conda activate protenix
~/Protenix$ pip install torch==2.3.1 nvidia-cublas-cu12 ml_collections

ERROR: pip's dependency resolver does not currently take into account all the packages that are
installed. This behaviour is the source of the following dependency conflicts.
deepspeed 0.16.5 requires ninja, which is not installed.

~/Protenix$ conda install ninja
~/Protenix$ pip install protenix

Successfully installed ninja-1.11.1.4 protenix-0.4.6 tomli-2.2.1

QUESTIONS:

1. The above procedure worked, but is there a better way to accomplish this?
2. What is the best way to update protenix when a new release has been published on GitHub?

Thank you.

[heidongxianhua]

We don't know what this file requirements.txt contains. If this file comes with our repository, it is used for pip installation and cannot be used directly for conda create. And if you just want to upgrade protenix, you can only update it by pip3 install protenix --no-deps

[zhangyuxuann]

Can you paste the detailed errors when use pip directly to install protenix in a new conda environment?

conda create -n protenix python=3.11
pip install protenix