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

ModuleV3EncoderPipelineModel

ModuleV3EncoderPipelineModel

class max.pipelines.lib.registration_encoder.ModuleV3EncoderPipelineModel(pipeline_config, session, devices, kv_cache_config, weights, adapter=None, return_logits=ReturnLogits.ALL, *, memory_plan, max_batch_size=1)

source

Bases: EncoderPipelineModel

Serves a registered ModuleV3 ENCODER module.

Builds, compiles and initializes the encoder graph.

Parameters:

• pipeline_config (PipelineConfig)
• session (InferenceSession)
• devices (list[Device])
• kv_cache_config (KVCacheConfig)
• weights (Weights)
• adapter (WeightsAdapter | None)
• return_logits (ReturnLogits)
• memory_plan (MemoryPlan)
• max_batch_size (int)

encoder_cls

encoder_cls: ClassVar[type[Module[..., Any]]]

source

The registered ENCODER class.

graph_specs()

graph_specs(model_config)

source

Returns the encoder graph, built from the checkpoint weights.

Parameters:

model_config (Any)

Return type:

dict[str, GraphSpec]