For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).

Python function

module_pipeline_model

module_pipeline_model()

max.pipelines.lib.module_registration.module_pipeline_model(config, modules, batching)

source

Returns the pipeline model that serves a module registration.

One registration always returns the same class.

Parameters:

• config (type[ArchConfig]) – The registered config class.
• modules (tuple[tuple[str, type[Module[..., Any]]], ...]) – The registered module class of each role, as items.
• batching (type[Any] | None) – The registered batch processor, or None for DecoderBatchProcessor.

Raises:

• ValueError – If the registration names a framework-owned role, has no DECODER.
• NotImplementedError – If it registers a role other than DECODER and VISION.
• TypeError – If a registered class isn’t a Module, or the registration’s VISION module, decoder and batch processor don’t fit together.

Return type:

type[ModuleV3DecoderPipelineModel]