/
githubmirror
/
Open-Assistant
Обзор
Документация
Войти
/
githubmirror
/
Open-Assistant
Код
Запросы
0
Пакеты
0
Релизы
0
Аналитика
Безопасность
main
docker/Dockerfile.model-training
35 строк
1 KB
Andreas Köpf
Model training dockerfile (#2405)
15 апр 2023, 21:33
Не верифицирован
15 апр 2023, 21:33
b9cfc76
Код
Авторство
О чём код?
# This image is for development/model-training (e.g. sweeps) and will not be deployed on production machines. FROM nvidia/cuda:11.7.1-cudnn8-devel-ubuntu22.04 WORKDIR /workspace SHELL ["/bin/bash", "-c"] # install git, python & torch RUN apt update && \ apt install -y vim git python3.10 python3.10-venv && \ python3.10 -m venv venv RUN source venv/bin/activate && \ pip install torch==1.13.1 # clone OA repo & install dependencies RUN apt install -y python3.10-dev && \ git clone https://github.com/LAION-AI/Open-Assistant.git && \ source venv/bin/activate && \ pip install -e /workspace/Open-Assistant/oasst-data \ pip install -e /workspace/Open-Assistant/model ENV PATH=/workspace/venv/bin:$PATH WORKDIR /workspace/Open-Assistant/model/model_training CMD ["/bin/bash"] # In the ./Open-Assistant repository root folder: # ... build the container: # nvidia-docker build -f docker/Dockerfile.model-training . -t oasst-model-training # ... run with the OA repository mounted: # nvidia-docker run -v `pwd`:/workspace/Open-Assistant -it --rm --env CUDA_VISIBLE_DEVICES=0 oasst-model-training