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 module

max.pipelines.lib.pipeline_parallel

Pipeline parallelism: a decoder cut into stages of contiguous layers.

Models

PipelinedModel	Mixin for a pipeline model whose decoder can run as pipeline stages.
StageTraced	Ends a traced stage's build once it knows what it hands the next.

Partitioning

PartitionSpec	Where a checkpoint's decoder may be cut into pipeline stages.
PipelineStage	One stage of a decoder cut into pipeline stages.

partition_spec	Returns where the architecture allows the checkpoint to be cut.
pipeline_stages	Returns the stages of a partition, in order.
split_layers	Returns each pipeline stage's decoder layer count.
stage_layers	Returns the decoder layers the config builds.
stage_pipeline_config	Returns the config that builds one pipeline stage of the main model.