cellmap_flow.cli.server_cli

The inference server: serve one model’s predictions, on the node an inference job runs on.

• cellmap_flow serve --model <entry> -d <data path> (serve), what the launchers run since 0.3.0 (serving.launch). The entry is ModelConfig.launch_entry as JSON, rebuilt as a YAML’s model entry is (registry.build_model).

• cellmap_flow_server, the server’s program before 0.3.0, which goes in the release after it. It takes --model too, for a deployment whose CELLMAP_FLOW_SERVER_COMMAND still names it; and it keeps a command for each model type, which a dashboard from before 0.3.0 launches: built when click asks for it, with the type’s constructor arguments as options (registry.click_options, the ModelConfig.command form).

Attributes

logger

Functions

run_server(model_config, data_path[, debug, port, ...])	Run the inference server with the given configuration; resample
serve_entry(model_json, data_path[, debug, port, ...])	Build the model model_json describes, and serve it.
serve(model_json, data_path, port, debug, certfile, ...)	Serve one model's predictions, as an inference job does on its node.
create_dynamic_server_command(cli_name, config_class)	Dynamically create a Click command for a ModelConfig subclass server.
cli(ctx, model_json, data_path, port, debug, certfile, ...)	The inference server before 0.3.0: deprecated, and goes in the release
list_models()	List all available model configurations.
main()	The cellmap_flow_server console script: load the plugins, whose

Module Contents

cellmap_flow.cli.server_cli.logger

cellmap_flow.cli.server_cli.run_server(model_config, data_path, debug=False, port=0, certfile=None, keyfile=None, resample=True)

Run the inference server with the given configuration; resample as CellMapFlowServer takes it, the engine CELLMAP_FLOW_ENGINE’s (serving.engine).

cellmap_flow.cli.server_cli.serve_entry(model_json, data_path, debug=False, port=0, certfile=None, keyfile=None, resample=True)

Build the model model_json describes, and serve it.

model_json is a model entry (ModelConfig.launch_entry) as JSON. One that is not is a usage error; a model that cannot be built, or a server that crashes, exits 1 with the traceback in the job’s log.

cellmap_flow.cli.server_cli.serve(model_json, data_path, port, debug, certfile, keyfile, resample)

Serve one model’s predictions, as an inference job does on its node.

The launchers (infer, yaml, the dashboard) run this for you, as CELLMAP_FLOW_SERVER_COMMAND (by default cellmap_flow serve). The model is a YAML model entry, as JSON:



cellmap_flow serve -d /path/to/data.zarr/raw

–model ‘{“type”: “script”, “script_path”: “/path/to/model.py”}’

cellmap_flow.cli.server_cli.create_dynamic_server_command(cli_name: str, config_class: Type[cellmap_flow.models.models_config.ModelConfig])

Dynamically create a Click command for a ModelConfig subclass server.

cellmap_flow.cli.server_cli.cli(ctx, model_json, data_path, port, debug, certfile, keyfile, resample)

The inference server before 0.3.0: deprecated, and goes in the release after it. Use cellmap_flow serve.

With –model and -d it is cellmap_flow serve. Each model type’s command is the form launchers used before 0.3.0:



cellmap_flow_server dacapo -r my_run -i 100 -d /path/to/data cellmap_flow_server script -s /path/to/script.py -d /path/to/data cellmap_flow_server cellmap -f /path/to/model -n mymodel -d /path/to/data

cellmap_flow.cli.server_cli.list_models()

List all available model configurations.

cellmap_flow.cli.server_cli.main()

The cellmap_flow_server console script: load the plugins, whose model types have commands too, then run the command.