runtime #

Runtime functions

command #

def command(commands: Dict[str, str])

Execute commands during runtime

Arguments:

• commands Dict[str, str] - map name to command, similar to Makefile

  Example usage:

runtime.command(commands={
    "train": "python train.py --epoch 20 --notify me@tensorchord.ai",
    "run": "python server.py --batch 1 --host 0.0.0.0 --port 8000",
})

You can run envd exec --command train to train the model.

expose #

def expose(envd_port: str, host_port: Optional[str], service: Optional[str],
           listen_addr: Optional[str])

Expose port to host Proposal: https://github.com/tensorchord/envd/pull/780

Arguments:

• envd_port str - port in envd container
• host_port Optional[str] - port in the host, if not provided or host_port=0, envd will randomly choose a free port
• service Optional[str] - service name
• listen_addr Optional[str] - address to listen on

daemon #

def daemon(commands: List[List[str]])

Run daemon processes in the container Proposal: https://github.com/tensorchord/envd/pull/769

It's better to redirect the logs to local files for debug purposes.

Arguments:

• commands List[List[str]] - run multiple commands in the background

  Example usage:

runtime.daemon(commands=[
    ["jupyter-lab", "--port", "8080"],
    ["python3", "serving.py", ">>serving.log", "2>&1"],
])

environ #

def environ(env: Dict[str, str], extra_path: List[str])

Add runtime environments

Arguments:

• env Dict[str, str] - environment name to value

• extra_path List[str] - additional PATH

  Example usage:

runtime.environ(env={"ENVD_MODE": "DEV"}, extra_path=["/usr/bin/go/bin"])

mount #

def mount(host_path: str, envd_path: str)

Mount from host path to container path (runtime)

Arguments:

• host_path str - source path in the host machine
• envd_path str - destination path in the envd container

init #

def init(commands: List[str])

Commands to be executed when start the container

Arguments:

• commands List[str] - list of commands