diff --git a/vllm/model_executor/models/keye.py b/vllm/model_executor/models/keye.py index aa5ca77bff..fd8cf4cbda 100644 --- a/vllm/model_executor/models/keye.py +++ b/vllm/model_executor/models/keye.py @@ -1466,6 +1466,14 @@ class BaseKeyeModule(nn.Module, SupportsMultiModal): tower_model="visual.", ) + def get_num_mm_encoder_tokens(self, num_image_tokens: int) -> int: + merge_size = self.config.vision_config.spatial_merge_size + return num_image_tokens * merge_size**2 + + def get_num_mm_connector_tokens(self, num_vision_tokens: int) -> int: + merge_size = self.config.vision_config.spatial_merge_size + return num_vision_tokens // merge_size**2 + @MULTIMODAL_REGISTRY.register_processor( KeyeMultiModalProcessor,