Skip to content

csederman/cdrpy

Repository files navigation

cdrpy

Suite of tools for cancer drug response prediction.

TODO

  • Make tensorflow an optional dependency
  • Each dataset should be a subclass of the Dataset class and should have a download method that stores the raw data in ~/.cdrpy/datasets or another specified directory

Setup and Installation

Linux

New GPU Setup

conda create python=3.9.13 --name=cdrpy-tf-gpu-v2

conda activate cdrpy-tf-gpu-v2

module load cuda/11.3 cudnn/8.2.0

conda install -c conda-forge cudatoolkit=11.8.0

pip install nvidia-cudnn-cu11==8.6.0.163

pip install tensorflow==2.11.*

pip install tensorflow-probability==0.19.0

Old GPU Setup

conda create python=3.9.13 --name=cdrpy-tf-gpu

module load cuda/11.2 cudnn/8.1.1

conda activate cdrpy-tf-gpu

conda install -c conda-forge cudatoolkit=11.2.2 cudnn=8.1.0.77

pip install tensorflow==2.10.0
pip install tensorflow-probability==0.18.0

CPU Setup

Create a conda environments:

conda create python=3.9.13 --name=cdrpy-tf-cpu

MacOS

conda create --name cdrpy-tf python=3.9.13

conda activate cdrpy-tf

python -m pip install tensorflow

python -m pip install tensorflow-metal

Usage

  • Currently, scripts must be run from root directory

For GPU usage:

module load cuda/11.2 cudnn/8.1.1

conda activate cdrpy-tf-gpu
module load cuda/11.3 cudnn/8.2.0

conda create --name cdrpy-tf-gpu-v2 python=3.9.13

Requirements

  • networkx
  • pandas
  • deepchem
  • tensorflow-propability
  • scikit-learn
  • hydra-core

About

Cancer drug response prediction

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages