Fix max tokens

Signed-off-by: DarkLight1337 <[email protected]>

vllm/model_executor/models/internvl.py

@@ -36,7 +36,6 @@
 from vllm.sequence import IntermediateTensors
 from vllm.transformers_utils.tokenizer import AnyTokenizer
 
-from .clip import get_clip_num_patches
 from .interfaces import SupportsMultiModal, SupportsPP
 from .utils import (AutoWeightsLoader, flatten_bn, init_vllm_registered_model,
                     maybe_prefix, merge_multimodal_embeddings)
@@ -180,16 +179,6 @@ def _get_max_dynamic_patch(
     return max_dynamic_patch
 
 
-def get_internvl_num_patches(hf_config: PretrainedConfig):
-    vision_config = hf_config.vision_config
-    downsample_ratio = hf_config.downsample_ratio
-    image_size = vision_config.image_size
-    patch_size = vision_config.patch_size
-    return int(
-        get_clip_num_patches(image_size=image_size, patch_size=patch_size) *
-        (downsample_ratio**2))
-
-
 # adapted from https://huggingface.co/OpenGVLab/InternVL2-1B
 def image_to_pixel_values(image: Image.Image, input_size: int, min_num: int,
                           max_num: int, use_thumbnail: bool) -> torch.Tensor:
@@ -377,8 +366,20 @@ def get_num_image_tokens(
     def get_max_image_tokens(self) -> int:
         processor = self.get_hf_processor()
 
-        hf_config = self.get_hf_config()
-        max_num_patches = get_internvl_num_patches(hf_config)
+        max_dynamic_patch = _get_max_dynamic_patch(
+            dynamic_image_size=processor.dynamic_image_size,
+            max_dynamic_patch=processor.max_dynamic_patch,
+            use_thumbnail=False,  # Applied in calculate_targets
+        )
+
+        max_num_patches, _, _ = calculate_targets(
+            orig_width=processor.image_size,
+            orig_height=processor.image_size,
+            image_size=processor.image_size,
+            min_num=processor.min_dynamic_patch,
+            max_num=max_dynamic_patch,
+            use_thumbnail=processor.use_thumbnail,
+        )
 
         return max_num_patches * processor.num_image_token