/dexterous_manipulation

A number of Reinforcement Learning environments for Dexterous Manipulation project

Primary LanguagePython

Dexterous Manipulation

Project Description

TODO

Deployment to GCP

Setup

Build and push image to GCP:

gcloud builds submit --tag gcr.io/dexterous-manipulation-238516/<IMAGE_NAME>:<VERSION> .

Run VM instance from image (region 28):

gcloud compute instances create-with-container <INSTANCE_NAME> --container-image gcr.io/dexterous-manipulation-238516/<IMAGE_NAME>:<VERSION>

Update VM instance with new image:

gcloud compute instances update-container <INSTANCE_NAME> --container-image gcr.io/dexterous-manipulation-238516/<IMAGE_NAME>:<VERSION>

Info

Copy file from instance to local:

gcloud compute scp --recurse <INSTANCE_NAME>:<REMOTE_DIR> <LOCAL_DIR>

Copy file from docker to instance:

docker cp <CONTAINER_ID>:<CONTAINER_PATH> <LOCAL_PATH>

View VM startup logs (provides container name on successful start):

gcloud compute instances get-serial-port-output <INSTANCE_NAME>

SSH into running instance:

gcloud compute ssh <INSTANCE_NAME> --container <CONTAINER_NAME>

View docker logs:

gcloud logging read "logName=projects/dexterous-manipulation-238516/logs/gcplogs-docker-driver AND jsonPayload.container.name=/<INSTANCE_NAME>"

View compute instances:

gcloud compute instances list

Cleanup

Delete compute instance:

gcloud compute instances delete <INSTANCE_NAME>