From 55392bc87932da63b888e58f371fe4b67b438496 Mon Sep 17 00:00:00 2001
From: "sangho.lee" <sanghol@allenai.org>
Date: Sat, 11 Oct 2025 00:28:23 -0500
Subject: [PATCH] [Bugfix][Multi Modal] Fix incorrect Molmo image processing
 (#26563)

Signed-off-by: sanghol <sanghol@allenai.org>
---
 vllm/model_executor/models/molmo.py | 40 +++++++++++++++++------------
 1 file changed, 23 insertions(+), 17 deletions(-)

diff --git a/vllm/model_executor/models/molmo.py b/vllm/model_executor/models/molmo.py
index 734841d0d..f1dd06f3a 100644
--- a/vllm/model_executor/models/molmo.py
+++ b/vllm/model_executor/models/molmo.py
@@ -114,11 +114,11 @@ class MolmoImageInputs(TensorSchema):
         TensorShape("bn", "nc", "np", dynamic_dims={"nc"}),
     ]
 
-    feat_is_patch: Annotated[
+    image_input_idx: Annotated[
         Union[torch.Tensor, list[torch.Tensor]],
         TensorShape("bn", "nc", "tp", dynamic_dims={"nc"}),
     ]
-    # A boolean mask indicating which image features correspond to patch tokens.
+    # An index tensor that maps image features to their corresponding patch tokens.
     num_crops: Annotated[torch.Tensor, TensorShape("bn")]
 
 
@@ -1177,7 +1177,7 @@ class MolmoProcessorWrapper:
             num_crops = torch.tensor(tilings).prod(-1) + 1
             assert num_crops.sum() == len(feat_is_patch)
 
-            outputs["feat_is_patch"] = feat_is_patch
+            outputs["image_input_idx"] = image_input_idx
             outputs["num_crops"] = num_crops
             outputs["img_patch_id"] = self.image_patch_id
 
@@ -1211,8 +1211,9 @@ class MolmoProcessingInfo(BaseProcessingInfo):
         image_token_length_w = processor.image_token_length_w
         image_token_length_h = processor.image_token_length_h
 
-        extra = image_token_length_w * image_token_length_h
-        joint = ((ncols + 1) // pooling_size) * ((nrows + 1) // pooling_size)
+        # Calculate total tokens: 2 for start/end + (w+1)*h for column separators
+        extra = 2 + (image_token_length_w + 1) * image_token_length_h
+        joint = 2 + ((ncols + 1) // pooling_size + 1) * ((nrows + 1) // pooling_size)
 
         return extra + joint
 
@@ -1299,7 +1300,7 @@ class MolmoMultiModalProcessor(BaseMultiModalProcessor[MolmoProcessingInfo]):
         return dict(
             images=MultiModalFieldConfig.flat_from_sizes("image", num_crops),
             image_masks=MultiModalFieldConfig.flat_from_sizes("image", num_crops),
-            feat_is_patch=MultiModalFieldConfig.flat_from_sizes("image", num_crops),
+            image_input_idx=MultiModalFieldConfig.flat_from_sizes("image", num_crops),
             num_crops=MultiModalFieldConfig.batched("image"),
             img_patch_id=MultiModalFieldConfig.shared("image", num_images),
         )
@@ -1444,7 +1445,7 @@ class MolmoForCausalLM(
     ) -> Optional[MolmoImageInputs]:
         images = kwargs.pop("images", None)
         image_masks = kwargs.pop("image_masks", None)
-        feat_is_patch = kwargs.pop("feat_is_patch", None)
+        image_input_idx = kwargs.pop("image_input_idx", None)
         num_crops = kwargs.pop("num_crops", None)
 
         if images is None:
@@ -1466,7 +1467,7 @@ class MolmoForCausalLM(
         return MolmoImageInputs(
             images=images,
             image_masks=image_masks,
-            feat_is_patch=feat_is_patch,
+            image_input_idx=image_input_idx,
             num_crops=num_crops,
         )
 
@@ -1476,7 +1477,7 @@ class MolmoForCausalLM(
     ) -> list[torch.Tensor]:
         images = image_input["images"]
         image_masks = image_input["image_masks"]
-        feat_is_patch = image_input["feat_is_patch"]
+        image_input_idx = image_input["image_input_idx"]
         num_crops = image_input["num_crops"]
 
         # Call the vision backbone on the whole batch at once
@@ -1484,7 +1485,7 @@ class MolmoForCausalLM(
         image_masks_flat = (
             None if image_masks is None else flatten_bn(image_masks, concat=True)
         )
-        feat_is_patch_flat = flatten_bn(feat_is_patch, concat=True)
+        image_input_idx_flat = flatten_bn(image_input_idx, concat=True)
 
         image_features_flat = self.vision_backbone(
             images=images_flat.unsqueeze(0),
@@ -1494,13 +1495,18 @@ class MolmoForCausalLM(
         ).squeeze(0)
 
         # Only the features corresponding to patch tokens are relevant
-        return [
-            feats[f_is_patch]
-            for feats, f_is_patch in zip(
-                image_features_flat.split(num_crops.tolist()),
-                feat_is_patch_flat.split(num_crops.tolist()),
-            )
-        ]
+        # Re-order the features using the image_input_idx tensor
+        results = []
+        num_crops_list = num_crops.tolist()
+        for feats, img_idx in zip(
+            image_features_flat.split(num_crops_list),
+            image_input_idx_flat.split(num_crops_list),
+        ):
+            is_valid = img_idx >= 0
+            valid_img_idx = img_idx[is_valid]
+            order = torch.argsort(valid_img_idx)
+            results.append(feats[is_valid][order])
+        return results
 
     def get_language_model(self) -> torch.nn.Module:
         return self.model