Requirements.txt or conda.yaml for env missing
Dear Christina,
I was trying to get the .ipynbs from the example folder running, first.
The problem, however, is that I do not find any requirements file and therefore need to install packages manually, one by one, if they are missing.
Could you provide the .yaml file, describing the conda environment, or the image for the singularity or docker container, in which the notebooks would be runnable?
Best,
Dzianis
I agree it would be helpful if we had the dependencies controlled and documented. I ran into this with the last update moving python versions. This is not a solution to that but I did already go through the process and built a pyproject.toml file for use with venv and pip.
requires-python = ">=3.10"
dependencies = [
"numpy",
"seaborn>=0.11.1",
"scikit-learn>=0.24.2",
"datasets>=2.14.1",
"transformers>=4.31.0",
"torch>=1.8.0",
"loompy>=3.0.6",
"scipy>=1.6.3",
"statsmodels>=0.12.2",
"anndata>=0.7.6",
"scanpy>=1.7.2",
"accelerate>=0.20.1",
"tdigest"
]
This allows me to run at least the in_silico_perturbation notebook as is.
Thank you for your interest in Geneformer and suggestions! We updated setup.py to include the current primary dependencies and recommended Python version (though we revised to include capitalized annotations from the typing library rather than the built-in collection types for Python >=3.9 so it should be compatible with older versions currently). A set of minimum requirements have been added to requirements.txt though it is definitely possible that earlier version of these packages will work - this has not been thoroughly evaluated, but we added this to hopefully provide a more defined suggested environment for those that prefer to specify this with pip install . -r requirements.txt. Please let us know if you encounter any further difficulties and we would be happy to update this accordingly. Thank you!