Built with developer experience in mind, Tensorkube simplifies the process of deploying serverless GPU apps. Sometimes for advanced debugging use-cases you might want to know the underlying details of the deployment. This cheatsheet will equip you with the necessary commands to get the details of your deployment.

​
Deploy command with all parameters

tensorkube deploy --gpus 1 --gpu-type a10g --cpu 2500 --memory 12000 --min-scale 1 --max-scale 10 --env staging --secret aws-secret --secret huggingface-secret

  • --cpu is in millicores, 1000 = 1vCPU
  • --memory is in MB, 1024 = 1GB

​
List all deployments

Use tensorkube list deployments --all to get the details of your deployed apps including their names, status and their HTTP endpoints. Use tensorkube list deployments --env <env_name> for apps in an environment.

​
Logs of a pod

Use tensorkube deployment logs <deployment-name> --env <env-name> to get the logs of a deployment/pod. This will show you the logs of the pod and help you debug any issues that might be present in the pod.

​
SSH into a deployment / pod

Use tensorkube deployment ssh <deployment-name> --env <env-name> to ssh into a pod. This will help you debug any issues that might be present in the pod.

​
Get details of a deployment

tensorkube deployment describe <deployment-name> --env <env-name>

​
List all running pods

Use kubectl get pods -n <env-name> to get the details of all the running pods across all your deployments in an environment.

​
Status of pod before being created (only use when necessary)

Use kubectl describe pod <pod-name> to get the status of a pod before it is created. This will show you all the steps that Tensorkube takes to create a pod. It will also give you the time each of those steps take and whether there is any scope of optimisation or not.

Deploying SpeechT5 on serverless GPUsAdding Other Members