max_num_encoder_input_tokens

SchedulerConfig derived field for the multimodal encoder compute budget; init=False and overridden from max_num_batched_tokens.

SettingvLLM v0.23.0verified Jul 9

Claims (0)

No claims about this entity yet.

Availability

max_num_encoder_input_tokens is exposed on SchedulerConfig in the v0.23.0 source snapshot. The field is init=False and derived from max_num_batched_tokens, not constructor-settable.

vllm.config.scheduler_configcomplete_inventorySchedulerConfig dataclass field surface in vllm/config/scheduler.pyinit=False

vLLM v0.23.0 config scheduler.py · vllm/config/scheduler.py:96-100