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 class

RetrievedPipeline

RetrievedPipeline

class max.pipelines.lib.registry.RetrievedPipeline(tokenizer, factory, memory_plan)

source

Bases: object

Everything PipelineRegistry.retrieve_factory() resolves.

Carries the tokenizer, a factory that constructs the pipeline, and the memory plan the pipeline was sized against.

Parameters:

• tokenizer (PipelineTokenizer[Any, Any, Any])
• factory (Callable[[], Pipeline[Any, Any]])
• memory_plan (MemoryPlan)

factory

factory: Callable[[], Pipeline[Any, Any]]

source

Zero-argument callable that constructs the pipeline instance.

memory_plan

memory_plan: MemoryPlan

source

Memory plan (batch size, sequence length, and cache budgets) the pipeline was sized against.

tokenizer

tokenizer: PipelineTokenizer[Any, Any, Any]

source

Tokenizer paired with the pipeline.