Python module

max.pipelines.architectures.llama3_modulev3

Llama 3 transformer architecture for text generation.

`Llama3Config`

class max.pipelines.architectures.llama3_modulev3.Llama3Config(*, hidden_size, num_attention_heads, num_key_value_heads, num_hidden_layers, rope_theta, rope_scaling_params, max_seq_len, intermediate_size, interleaved_rope_weights, vocab_size, dtype, kv_params, return_logits=ReturnLogits.LAST_TOKEN, norm_method='rms_norm', attention_bias=False, rms_norm_eps=None, tie_word_embeddings=False, stacked_mlp=False, stacked_qkv=False, attention_multiplier, embedding_multiplier, residual_multiplier, devices, clip_qkv=None, norm_elementwise_affine=True, longrope_scaling_params=None, logits_scaling=1.0, return_hidden_states=ReturnHiddenStates.NONE)

source

Bases: ArchConfigWithKVCache

Model configuration for Llama3 graph construction/execution.

Parameters:

hidden_size (int)
num_attention_heads (int)
num_key_value_heads (int)
num_hidden_layers (int)
rope_theta (float)
rope_scaling_params (Llama3RopeScalingParams | None)
max_seq_len (int)
intermediate_size (int)
interleaved_rope_weights (bool)
vocab_size (int)
dtype (DType)
kv_params (KVCacheParams)
return_logits (ReturnLogits)
norm_method (Literal['rms_norm', 'layer_norm'])
attention_bias (bool)
rms_norm_eps (float | None)
tie_word_embeddings (bool)
stacked_mlp (bool)
stacked_qkv (bool)
attention_multiplier (float)
embedding_multiplier (float)
residual_multiplier (float)
devices (list[DeviceRef])
clip_qkv (float | None)
norm_elementwise_affine (bool)
longrope_scaling_params (LongRoPEScalingParams | None)
logits_scaling (float)
return_hidden_states (ReturnHiddenStates)

`attention_bias`

attention_bias: bool = False

source

`attention_multiplier`

attention_multiplier: float

source

`calculate_attention_multiplier()`

static calculate_attention_multiplier(huggingface_config)

source

Parameters:: huggingface_config (AutoConfig)
Return type:: float

`calculate_max_seq_len()`

static calculate_max_seq_len(pipeline_config, huggingface_config)

source

Parameters:

pipeline_config (PipelineConfig)
huggingface_config (AutoConfig)

Return type:

int

`clip_qkv`

clip_qkv: float | None = None

source

`construct_kv_params()`

static construct_kv_params(huggingface_config, pipeline_config, devices, kv_cache_config, cache_dtype)

source

Parameters:

huggingface_config (AutoConfig)
pipeline_config (PipelineConfig)
devices (list[DeviceRef])
kv_cache_config (KVCacheConfig)
cache_dtype (DType)

Return type:

KVCacheParams

`devices`

devices: list[DeviceRef]

source

`dtype`

dtype: DType

source

`embedding_multiplier`

embedding_multiplier: float

source

`finalize()`

finalize(huggingface_config, state_dict, return_logits, return_hidden_states=ReturnHiddenStates.NONE, norm_method='rms_norm', attention_bias=False)

source

Define parameters that can’t be determined just from the pipeline config.

Parameters:

huggingface_config (AutoConfig)
state_dict (dict[str, WeightData])
return_logits (ReturnLogits)
return_hidden_states (ReturnHiddenStates)
norm_method (Literal['rms_norm', 'layer_norm'])
attention_bias (bool)

Return type:

None

`get_head_dim()`

static get_head_dim(huggingface_config)

source

Parameters:: huggingface_config (AutoConfig)
Return type:: int

`get_head_dim_from_config()`

static get_head_dim_from_config(config)

source

Parameters:: config (Llama3Config)
Return type:: int

`get_kv_params()`

get_kv_params()

source

KV cache parameters to use when running the model.

Return type:: KVCacheParams

`get_max_seq_len()`

get_max_seq_len()

source

Returns the default maximum sequence length for the model.

Subclasses should determine whether this value can be overridden by setting the --max-length (pipeline_config.model.max_length) flag.

Return type:: int

`get_num_layers()`

static get_num_layers(huggingface_config)

source

Parameters:: huggingface_config (AutoConfig)
Return type:: int

`hidden_size`

hidden_size: int

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

`interleaved_rope_weights`

interleaved_rope_weights: bool

source

`intermediate_size`

intermediate_size: int

source

`kv_params`

kv_params: KVCacheParams

source

`logits_scaling`

logits_scaling: float = 1.0

source

`longrope_scaling_params`

longrope_scaling_params: LongRoPEScalingParams | None = None

source

`max_seq_len`

max_seq_len: int

source

`norm_elementwise_affine`

norm_elementwise_affine: bool = True

source

`norm_method`

norm_method: Literal['rms_norm', 'layer_norm'] = 'rms_norm'

source

`num_attention_heads`

num_attention_heads: int

source

`num_hidden_layers`

num_hidden_layers: int

source

`num_key_value_heads`

num_key_value_heads: int

source

`residual_multiplier`

residual_multiplier: float

source

`return_hidden_states`

return_hidden_states: ReturnHiddenStates = 'none'

source

`return_logits`

return_logits: ReturnLogits = 'last_token'

source

`rms_norm_eps`

rms_norm_eps: float | None = None

source

`rope_scaling_params`

rope_scaling_params: Llama3RopeScalingParams | None

source

`rope_theta`

rope_theta: float

source

`stacked_mlp`

stacked_mlp: bool = False

source

`stacked_qkv`

stacked_qkv: bool = False

source

`tie_word_embeddings`

tie_word_embeddings: bool = False

source

`vocab_size`

vocab_size: int

source

`Llama3Inputs`

class max.pipelines.architectures.llama3_modulev3.Llama3Inputs(tokens, input_row_offsets, return_n_logits, *, kv_cache_inputs=None, lora_ids=None, lora_ranks=None, hidden_states=None)

source

Bases: ModelInputs

A class representing inputs for the Llama3 model.

Parameters:

tokens (Buffer)
input_row_offsets (Buffer)
return_n_logits (Buffer)
kv_cache_inputs (KVCacheInputs[Buffer, Buffer] | None)
lora_ids (Buffer | None)
lora_ranks (Buffer | None)
hidden_states (Buffer | list[Buffer] | None)

`buffers`

property buffers: tuple[Buffer, ...]

source

Returns positional Buffer inputs for model ABI calls.

`input_row_offsets`

input_row_offsets: Buffer

source

`return_n_logits`

return_n_logits: Buffer

source

`tokens`

tokens: Buffer

source

`Llama3Model`

class max.pipelines.architectures.llama3_modulev3.Llama3Model(pipeline_config, session, devices, kv_cache_config, weights, adapter=None, return_logits=ReturnLogits.LAST_TOKEN, return_hidden_states=ReturnHiddenStates.NONE)

source

Bases: LogProbabilitiesMixin, PipelineModelWithKVCache[TextContext]

Llama3 pipeline model using the ModuleV3 API.

Parameters:

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

`attention_bias`

attention_bias: bool = False

source

`calculate_max_seq_len()`

static calculate_max_seq_len(pipeline_config, huggingface_config)

source

Calculates the optimal max sequence length for the model.

Models are expected to implement this method. The following example shows how to implement it for a Mistral model:

class MistralModel(PipelineModel):
    @classmethod
    def calculate_max_seq_len(cls, pipeline_config, huggingface_config) -> int:
        try:
            return upper_bounded_default(
                upper_bound=huggingface_config.max_seq_len,
                default=pipeline_config.model.max_length,
            )
        except ValueError as e:
            raise ValueError(
                "Unable to infer max_length for Mistral, the provided "
                f"max_length ({pipeline_config.model.max_length}) exceeds the "
                f"model's max_seq_len ({huggingface_config.max_seq_len})."
            ) from e

Parameters:

pipeline_config (PipelineConfig) – Configuration for the pipeline.
huggingface_config (AutoConfig) – Hugging Face model configuration.

Returns:

The maximum sequence length to use.

Return type:

int

`config_class`

config_class

source

alias of Llama3Config

`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.

`get_kv_params()`

classmethod get_kv_params(huggingface_config, pipeline_config, devices, kv_cache_config, cache_dtype)

source

Returns the KV cache params for the pipeline model.

Parameters:

huggingface_config (AutoConfig)
pipeline_config (PipelineConfig)
devices (list[DeviceRef])
kv_cache_config (KVCacheConfig)
cache_dtype (DType)

Return type:

KVCacheParams

`load_model()`

load_model()

source

Return type:: Callable[[…], Any]

`norm_method`

norm_method: Literal['rms_norm'] | Literal['layer_norm'] = 'rms_norm'

source

`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:

ModelInputs

`prepare_next_token_inputs()`

prepare_next_token_inputs(next_tokens, prev_model_inputs)

source

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

While prepare_initial_token_inputs is responsible for managing the initial inputs. This function is responsible for updating the inputs, for each step in a multi-step execution pattern.

Parameters:

next_tokens (Buffer)
prev_model_inputs (ModelInputs)

Return type:

ModelInputs

Llama3Config​

attention_bias​

attention_multiplier​

calculate_attention_multiplier()​

calculate_max_seq_len()​

clip_qkv​

construct_kv_params()​

devices​

dtype​

embedding_multiplier​

finalize()​

get_head_dim()​

get_head_dim_from_config()​

get_kv_params()​

get_max_seq_len()​

get_num_layers()​

hidden_size​

initialize()​

interleaved_rope_weights​

intermediate_size​

kv_params​

logits_scaling​

longrope_scaling_params​

max_seq_len​

norm_elementwise_affine​

norm_method​

num_attention_heads​

num_hidden_layers​

num_key_value_heads​

residual_multiplier​

return_hidden_states​

return_logits​

rms_norm_eps​

rope_scaling_params​

rope_theta​

stacked_mlp​

stacked_qkv​

tie_word_embeddings​

vocab_size​

Llama3Inputs​

buffers​

input_row_offsets​

return_n_logits​

tokens​

Llama3Model​

attention_bias​

calculate_max_seq_len()​

config_class​

execute()​

get_kv_params()​

load_model()​

norm_method​

prepare_initial_token_inputs()​

prepare_next_token_inputs()​

`Llama3Config`

`attention_bias`

`attention_multiplier`

`calculate_attention_multiplier()`

`calculate_max_seq_len()`

`clip_qkv`

`construct_kv_params()`

`devices`

`dtype`

`embedding_multiplier`

`finalize()`

`get_head_dim()`

`get_head_dim_from_config()`

`get_kv_params()`

`get_max_seq_len()`

`get_num_layers()`

`hidden_size`

`initialize()`

`interleaved_rope_weights`

`intermediate_size`

`kv_params`

`logits_scaling`

`longrope_scaling_params`

`max_seq_len`

`norm_elementwise_affine`

`norm_method`

`num_attention_heads`

`num_hidden_layers`

`num_key_value_heads`

`residual_multiplier`

`return_hidden_states`

`return_logits`

`rms_norm_eps`

`rope_scaling_params`

`rope_theta`

`stacked_mlp`

`stacked_qkv`

`tie_word_embeddings`

`vocab_size`

`Llama3Inputs`

`buffers`

`input_row_offsets`

`return_n_logits`

`tokens`

`Llama3Model`

`attention_bias`

`calculate_max_seq_len()`

`config_class`

`execute()`

`get_kv_params()`

`load_model()`

`norm_method`

`prepare_initial_token_inputs()`

`prepare_next_token_inputs()`