connect_gcp_instance.md

1. start instance

gcloud compute instances start fastai-1

2. set up ssh tunnels, put ip (34.90.135.237) in ~/.ssh/config twice

ports:

• 8080 - jupyter lab 1
• 8081 - jupyter lab 2, in swift-jupyter docker container
• 6006 - tensorboard, in swift-jupyter docker container

Host fastai-1
HostName 34.90.135.237
StrictHostKeyChecking no
User jupyter
Port 22
LocalForward 8080 localhost:8080
LocalForward 8081 localhost:8081
LocalForward 6006 localhost:6006

Host fastai-1-vscode
HostName 34.90.135.237
StrictHostKeyChecking no
User jupyter
Port 22

3. ssh and activate tunnels

ssh fastai-1

4. start tmux

tmux -CC

5. start swift-jupyter container

docker start swift-jupyter

6. start tensorboard (inside docker container)

docker exec -it swift-jupyter bash
cd language2motion.gt/data/
tensorboard --bind_all --logdir runs/Lang2motion/run_4/

7. on second tmux tab: update language2motion sources

a. exec into docker container

docker exec -it swift-jupyter bash

b. git pull

ssh-agent bash
ssh-add /notebooks/.ssh/id_rsa_wcz\@MacBook-Wojtka.local
cd language2motion.gt/
git pull

8. start vscode

a. set docker.host to

ssh://fastai-1-vscode

b. Remote-Containers: Attach to Running Container...

/swift-jupyter

c. File/Open Workspace...

/notebooks/language2motion.gt/l2m.code-workspace

stop instance

gcloud compute instances stop fastai-1