Add working Dockerfile for easy setup

[victornemeth]

Introduce a Docker-based setup to eliminate dependency issues and ensure a reproducible installation environment.

[victornemeth]

You can test it by running (assuming you have docker installed):

docker build -t evo2 .
docker run     -it     --rm     --gpus all     -v ./huggingface:/root/.cache/huggingface     evo2     bash

This will build the container and run it interactively, once you're in the container (which you should be after running these commands), you can test evo2 with:

python3 ./test/test_evo2.py --model_name evo2_7b

I encoutered an error "RuntimeError: CUDA error: CUDA-capable device(s) is/are busy or unavailable" which was solved by running this instead:

CUDA_VISIBLE_DEVICES=0 python3 ./test/test_evo2.py --model_name evo2_7b

But I think this error was specific to my system configuration regarding GPU's.

[victornemeth]

The docker container can also be converted to Singularity (apptainer)

docker build -t evo2 .
singularity build evo2.sif docker-daemon://evo2:latest
mkdir -p models

To run it use:

singularity exec \
    --nv \
    --bind $PWD:/app \
    --bind ./models:/root/.cache/huggingface \
    ./evo2.sif \
    python3 ./test/test_evo2.py \
    --model_name evo2_7b