[LoRA][Refactor] Unify multimodal LoRA token count hooks (#55071)
Signed-off-by: linitra24 <renshuang.zhou@daocloud.io>
This commit is contained in:
@@ -22,6 +22,7 @@ from vllm.model_executor.layers.quantization import QuantizationConfig
|
||||
from vllm.multimodal import MULTIMODAL_REGISTRY
|
||||
from vllm.multimodal.inputs import (
|
||||
MultiModalFieldConfig,
|
||||
MultiModalKwargsItem,
|
||||
MultiModalKwargsItems,
|
||||
)
|
||||
from vllm.multimodal.parse import MultiModalDataItems
|
||||
@@ -687,28 +688,22 @@ class Blip2ForConditionalGeneration(
|
||||
tower_model="vision_model",
|
||||
)
|
||||
|
||||
def get_num_mm_encoder_tokens(
|
||||
def get_mm_lora_token_counts(
|
||||
self,
|
||||
num_image_tokens: int,
|
||||
) -> int:
|
||||
if num_image_tokens <= 0:
|
||||
return 0
|
||||
assert num_image_tokens % self.config.num_query_tokens == 0, (
|
||||
*,
|
||||
modality: str,
|
||||
mm_kwargs: MultiModalKwargsItem | None,
|
||||
num_mm_embeds: int,
|
||||
) -> tuple[int, int | None]:
|
||||
del modality, mm_kwargs
|
||||
if num_mm_embeds <= 0:
|
||||
return 0, 0
|
||||
assert num_mm_embeds % self.config.num_query_tokens == 0, (
|
||||
"The number of image tokens must be a multiple of "
|
||||
"the number of query tokens."
|
||||
)
|
||||
num_images = num_image_tokens / self.config.num_query_tokens
|
||||
return num_images * self._vision_tokens_per_image
|
||||
|
||||
def get_num_mm_connector_tokens(
|
||||
self,
|
||||
num_vision_tokens: int,
|
||||
) -> int:
|
||||
if num_vision_tokens <= 0:
|
||||
return 0
|
||||
assert num_vision_tokens % self._vision_tokens_per_image == 0, (
|
||||
"The number of vision tokens must be a multiple of "
|
||||
"the number of tokens per image."
|
||||
num_images = num_mm_embeds // self.config.num_query_tokens
|
||||
return (
|
||||
num_images * self._vision_tokens_per_image,
|
||||
num_images * self.config.num_query_tokens,
|
||||
)
|
||||
num_images = num_vision_tokens / self._vision_tokens_per_image
|
||||
return num_images * self.config.num_query_tokens
|
||||
|
||||
Reference in New Issue
Block a user