transformers-pytorch-gpu

💡 Docker setup for Huggingface 🤗 Transformers + GPU + Jupyter notebook + OhMyZsh

Features

• Base Docker image: NVIDIA PyTorch 20.08 (1.7.0)
  • Built-in PyTorch 1.7 & CUDA
• Added pip packages:
  • jupyter
  • pytorch-lightning
  • neptune-client
  • transformers
  • emoji
  • soynlp
  • konlpy
• Custom uid support
• Zsh & OhMyZsh support
• Default LANG to en_US.UTF-8

How to use

💡 Do this first! 💡

id -u && id -g

If your result is like below:

$ id -u && id -g
1000
1000

GOTO -> Pull from dockerhub

Else:

(example)

$ id -u && id -g
1004
1004

GOTO -> Build from scratch

Pull from dockerhub

git clone https://github.com/Beomi/transformers-pytorch-gpu
cd transformers-pytorch-gpu
docker pull beomi/transformers-pytorch-gpu:latest
./run.sh

Now open http://your-server-ip:42000

Jupyter Notebook is waiting for you :)

Build from scratch

git clone https://github.com/Beomi/transformers-pytorch-gpu
cd transformers-pytorch-gpu 
./build.sh
./run.sh

Now open http://your-server-ip:42000

Jupyter Notebook is waiting for you :)