cellmap_flow.serving.engine

Which inference server a process runs: chunkmirage’s or the Flask one.

CELLMAP_FLOW_ENGINE is chunkmirage (the default) or flask. Every place that starts a server (cellmap_flow serve, infer --server-check and the finetune loop) builds it with make_server, and both take the same arguments.

A model can run in an environment of its own (its env: a pixi environment, an alias or a path), which may not have chunkmirage. Unless chunkmirage was asked for, such a server is the Flask one, with a warning; the dashboard draws either server’s layers.

Attributes

`logger`
`ENGINE_ENV`
`ENGINES`
`DEFAULT_ENGINE`

Functions

`engine_name`(→ str)	`name`, else `CELLMAP_FLOW_ENGINE`, else the default; refused if unknown.
`make_server`(dataset_name, model_config, *[, engine])	The server for `model_config` over `dataset_name`; `kwargs` are
`check_server`(server)	Compute one chunk the way a request would: `infer --server-check`.

Module Contents

cellmap_flow.serving.engine.logger

cellmap_flow.serving.engine.ENGINE_ENV = 'CELLMAP_FLOW_ENGINE'

cellmap_flow.serving.engine.ENGINES = ('flask', 'chunkmirage')

cellmap_flow.serving.engine.DEFAULT_ENGINE = 'chunkmirage'

cellmap_flow.serving.engine.engine_name(name=None) → str: name, else CELLMAP_FLOW_ENGINE, else the default; refused if unknown.

cellmap_flow.serving.engine.make_server(dataset_name, model_config, *, engine=None, **kwargs): The server for model_config over dataset_name; kwargs are CellMapFlowServer’s (restart_callback, restart_token, resample).

cellmap_flow.serving.engine.check_server(server): Compute one chunk the way a request would: infer --server-check.