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

Python module

max.pipelines.architectures.mpnet_modulev3

MPNet sentence transformer architecture for embeddings generation.

`MPNetConfig`

class max.pipelines.architectures.mpnet_modulev3.MPNetConfig(*, pool_embeddings, huggingface_config, max_seq_len)

source

Bases: ArchConfigWithBoundedMaxSeqLen, ArchConfig

Configuration for MPNet V3 models.

Parameters:

pool_embeddings (bool)
huggingface_config (AutoConfig)
max_seq_len (int)

`huggingface_config`

huggingface_config: AutoConfig

source

`initialize()`

classmethod initialize(pipeline_config, model_config=None)

source

Initialize the config from a PipelineConfig.

Parameters:

pipeline_config (PipelineConfig) – The pipeline configuration.
model_config (MAXModelConfig | None) – The model configuration to read from. When None (the default), pipeline_config.model is used. Pass an explicit config (e.g. pipeline_config.draft_model) to initialize the arch config for a different model.

Return type:

Self

`max_seq_len`

max_seq_len: int

source

`pool_embeddings`

pool_embeddings: bool

source

`MPNetInputs`

class max.pipelines.architectures.mpnet_modulev3.MPNetInputs(next_tokens_batch, attention_mask, *, kv_cache_inputs=None, lora_ids=None, lora_ranks=None, hidden_states=None)

source

Bases: ModelInputs

Input tensors for the MPNet model.

Parameters:

next_tokens_batch (Buffer)
attention_mask (Buffer)
kv_cache_inputs (KVCacheInputs[Buffer, Buffer] | None)
lora_ids (Buffer | None)
lora_ranks (Buffer | None)
hidden_states (Buffer | list[Buffer] | None)

`attention_mask`

attention_mask: Buffer

source

`next_tokens_batch`

next_tokens_batch: Buffer

source

`MPNetPipelineModel`

class max.pipelines.architectures.mpnet_modulev3.MPNetPipelineModel(pipeline_config, session, devices, kv_cache_config, weights, adapter=None, return_logits=ReturnLogits.ALL)

source

Bases: PipelineModel[TextContext]

Parameters:

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

`execute()`

execute(model_inputs)

source

Executes the graph with the given inputs.

Parameters:: model_inputs (ModelInputs) – The model inputs to execute, containing tensors and any other required data for model execution.
Returns:: ModelOutputs containing the pipeline’s output tensors.
Return type:: ModelOutputs

This is an abstract method that must be implemented by concrete PipelineModels to define their specific execution logic.

`load_model()`

load_model()

source

Return type:: Callable[[…], tuple[Tensor, …]]

`model_config_cls`

model_config_cls

source

alias of MPNetConfig

`prepare_initial_token_inputs()`

prepare_initial_token_inputs(replica_batches, kv_cache_inputs=None, return_n_logits=1)

source

Prepares the initial inputs to be passed to execute().

The inputs and functionality can vary per model. For example, model inputs could include encoded tensors, unique IDs per tensor when using a KV cache manager, and kv_cache_inputs (or None if the model does not use KV cache). This method typically batches encoded tensors, claims a KV cache slot if needed, and returns the inputs and caches.

Parameters:

replica_batches (Sequence[Sequence[TextContext]])
kv_cache_inputs (KVCacheInputs[Buffer, Buffer] | None)
return_n_logits (int)

Return type:

MPNetInputs

MPNetConfig​

huggingface_config​

initialize()​

max_seq_len​

pool_embeddings​

MPNetInputs​

attention_mask​

next_tokens_batch​

MPNetPipelineModel​

execute()​

load_model()​

model_config_cls​

prepare_initial_token_inputs()​

`MPNetConfig`

`huggingface_config`

`initialize()`

`max_seq_len`

`pool_embeddings`

`MPNetInputs`

`attention_mask`

`next_tokens_batch`

`MPNetPipelineModel`

`execute()`

`load_model()`

`model_config_cls`

`prepare_initial_token_inputs()`