biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[VLM] Merged multi-modal processor for LLaVA-NeXT (#11682)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-01-03 00:39:27 +08:00
committed by GitHub
parent b6087a6bee
commit 8c38ee7007
14 changed files with 609 additions and 555 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/model_executor/models/fuyu.py
View File

@@ -76,7 +76,7 @@ class FuyuMultiModalProcessor(BaseMultiModalProcessor):
return ImageSize(width=target_size["width"],
height=target_size["height"])
def _get_image_grid_size(
def _get_image_feature_grid_size(
self,
*,
image_width: int,
@@ -99,7 +99,7 @@ class FuyuMultiModalProcessor(BaseMultiModalProcessor):
def get_mm_max_tokens_per_item(self) -> Mapping[str, int]:
target_width, target_height = self._get_image_target_size()
max_ncols, max_nrows = self._get_image_grid_size(
max_ncols, max_nrows = self._get_image_feature_grid_size(
image_width=target_width,
image_height=target_height,
)
@@ -172,7 +172,7 @@ class FuyuMultiModalProcessor(BaseMultiModalProcessor):
images = mm_items.get_items("image", ImageProcessorItems)
image_size = images.get_image_size(item_idx)
ncols, nrows = self._get_image_grid_size(
ncols, nrows = self._get_image_feature_grid_size(
image_width=image_size.width,
image_height=image_size.height,
)