Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Implement distributed training using Kubernetes #77

Merged
merged 17 commits into from
Jan 23, 2021
Merged

Implement distributed training using Kubernetes #77

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
17 commits
Select commit Hold shift + click to select a range
1bdb814
Add initial Dockerfile and k8s deployment
leogao2 Jan 23, 2021
05f4dcf
Fix imports
leogao2 Jan 23, 2021
6e9b3a8
Update dockerfile
leogao2 Jan 23, 2021
a0b4748
Merge branch 'main' of github.com:EleutherAI/gpt-neox into k8s
leogao2 Jan 23, 2021
206b8ca
Add /dev/shm patch
leogao2 Jan 23, 2021
2a578d2
Update to be completely passwordless
leogao2 Jan 23, 2021
d48f48a
Add host file generation script
leogao2 Jan 23, 2021
032fde1
Make deploy script
leogao2 Jan 23, 2021
81f1ef4
Update deploy script and fix echo -e problem
leogao2 Jan 23, 2021
ee1739e
Remove command line argument
leogao2 Jan 23, 2021
af90891
Generate keys for worker machines
leogao2 Jan 23, 2021
dcd11ae
Harden security slightly
leogao2 Jan 23, 2021
e682f7f
Update docker for custom keygen
leogao2 Jan 23, 2021
8a26b7e
Fix ssh
leogao2 Jan 23, 2021
0e36734
Fix deploy script to use right id
leogao2 Jan 23, 2021
480dc36
Added logging config
StellaAthena Jan 23, 2021
221d73a
Update README.md
StellaAthena Jan 23, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
Prev Previous commit
Next Next commit
Update docker for custom keygen
  • Loading branch information
@leogao2
leogao2 committed Jan 23, 2021
commit e682f7f289b75c7087e7d3afc04999156eaf287f
9 changes: 2 additions & 7 deletions Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,13 +11,8 @@ RUN echo 'AuthorizedKeysFile .ssh/authorized_keys' >> /etc/ssh/sshd_config &
RUN python3 -m pip install --upgrade pip && \
pip3 install torch pipx && \
python3 -m pipx ensurepath && \
mkdir /app

RUN mkdir -p ~/.ssh && \
echo 'ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQChClBdh2UXpMhBV725nH1bMsGVaLmYAp8uMZFHtBqHN56sj/35jeenQTqr/Tov3u6xwqJK+rxegjgcPDZfuOSdsNnTMCLIacA/WqBMwW1mjdMc+zFOub7vQJHj4nmeF3pd4tSjt720ZLiX1ZsF5QrTIcnURAXT0/82SKIy2nqj18v9HCcIvBplexJU3SlVg+oWk/e5CsfnXvJMQH3VqJaeyrXIlaFgVOvVWSBY66Kc2H+g1RxLwe+BONyNanxSXxKHQwUXawBvXIyjekapB3HiNWLZZfZjNmqe2Ci+Y9PKn0CrkgXopIP7tVKn+UQ2fD3nSjaZyfRrmcFgXdotFKJh root@sshd-sid-79b6f9d7c6-mmw8x' > ~/.ssh/authorized_keys && \
echo 'Host *' > ~/.ssh/config && \
echo ' StrictHostKeyChecking no' >> ~/.ssh/config && \
chmod 600 ~/.ssh/config
mkdir /app && \
mkdir ~/.ssh


WORKDIR /app
Expand Down
6 changes: 3 additions & 3 deletions deploy_k8s.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,15 +1,14 @@
kubectl delete deploy/eleuther-neox
kubectl apply -f kubernetes/deploy_k8s.yml
echo Waiting for deploy to complete...
rm id_rsa*
ssh-keygen -t rsa -f id_rsa -N ""
echo Waiting for deploy to complete...
kubectl wait --for=condition=available --timeout=600s deployment/eleuther-neox || exit

kubectl get pods -o wide | grep eleuther-neox | awk '{print $6 " slots=8"}' > hosts
export MASTER_ID=$(kubectl get pods | grep eleuther-neox | awk '{print $1}' | head -n 1)
echo $MASTER_ID
kubectl cp $PWD/hosts $MASTER_ID:/app
kubectl cp $PWD/id_ed25519 $MASTER_ID:/root/.ssh
kubectl cp $PWD/id_rsa $MASTER_ID:/root/.ssh

mv id_rsa.pub authorized_keys

Expand All @@ -19,5 +18,6 @@ do
kubectl cp $PWD/authorized_keys $MASTER_ID:/root/.ssh/
done
rm authorized_keys hosts
rm id_rsa*
#echo 'git remote set-url origin https://github.com/EleutherAI/gpt-neox/ && git pull' | kubectl exec --stdin --tty $MASTER_ID -- /bin/bash
kubectl exec --stdin --tty $MASTER_ID -- /bin/bash