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

ArchVLConfigWithTextSubconfig

ArchVLConfigWithTextSubconfig

class max.pipelines.lib.interfaces.ArchVLConfigWithTextSubconfig

source

Bases: object

Mixin for VLMs that embed a language-model arch config.

Annotate llm_config or text_config with the text arch type; otherwise ArchConfigWithStoredKVParams is used (Pixtral). The HF subconfig is read from text_config or llm_config on the HuggingFace config (MAX field names need not match HF attribute names). Override construct_kv_params() / calculate_max_seq_len() when resolution is dynamic (e.g. InternVL) or semantics differ (e.g. Gemma4 KV).

calculate_max_seq_len()

classmethod calculate_max_seq_len(huggingface_config, model_config)

source

Delegates to the annotated text config class.

Parameters:

• huggingface_config (AutoConfig)
• model_config (MAXModelConfig)

Return type:

int

construct_kv_params()

classmethod construct_kv_params(huggingface_config, pipeline_config, devices, kv_cache_config, cache_dtype, *, allow_kv_head_replication=False)

source

Delegates to the annotated text config class.

Parameters:

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

Return type:

KVCacheParamInterface

get_max_seq_len()

get_max_seq_len()

source

Returns the maximum sequence length from the embedded text config.

Return type:

int