Geomics-FM

This repository includes the 'Genomics-FM: Universal Foundation Model for Versatile and Data-Efficient Functional Genomic Analysis' project.

1. Environment setup

1.1 Create and activate a new virtual environment

For example, you can use python>=3.6

conda create -n dnalm python=3.8
conda activate dnalm

1.2 Install the package and other requirements

(Required) Personally, I recommend fresh students use some stable version of Pytorch to match 3090 GPU cluster such as 1.7.1, which has been verified.

pip install torch==1.7.1+cu110 torchvision==0.8.2+cu110 torchaudio==0.7.2 -f https://download.pytorch.org/whl/torch_stable.html

git clone https://github.com/terry-r123/Genomics-FM.git
cd Genomics-FM
python3 -m pip install --editable .
cd examples
python3 -m pip install -r requirements.txt

1.3 Install the apex requirements (Optional, install apex for fp16 training)

change to a desired directory by cd PATH_NAME

git clone https://github.com/NVIDIA/apex
cd apex
pip install -v --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./

if you meet some errors, you can try the following: first delete and uninstall the original apex, then

git clone https://github.com/NVIDIA/apex.git
cd apex
git checkout f3a960f80244cf9e80558ab30f7f7e8cbf03c0a0
# Maybe you also need to comment  raise RuntimeError("Cuda extensions are being compiled with a version of Cuda that does...." with 'pass' in the 52nd line in setup.py
# Maybe vi ~/.bashrc and Add "export TORCH_CUDA_ARCH_LIST="8.0", then source ~/.bashrc 
python setup.py install --cuda_ext --cpp_ext
pip3 install -v --no-cache-dir ./

1.4 checkpoint

google drive link: https://drive.google.com/drive/folders/17-MQdeD9G-F7uiuHWWMdP50MmgJvpJb9?usp=drive_link