Legacy processing is triggered even when only pure text is input in llava without img_token

src/transformers/models/llava/modeling_llava.py

@@ -480,7 +480,7 @@ def forward(
                 vision_feature_select_strategy=vision_feature_select_strategy,
             )
 
-        if legacy_processing:
+        if legacy_processing and image_features is not None:
             logger.warning_once(
                 "Expanding inputs for image tokens in LLaVa should be done in processing. "
                 "Please add `patch_size` and `vision_feature_select_strategy` to the model's processing config or set directly "

src/transformers/models/llava_next/modeling_llava_next.py

@@ -863,7 +863,7 @@ def forward(
                 image_newline=self.image_newline,
             )
 
-        if legacy_processing:
+        if legacy_processing and image_features is not None:
             logger.warning_once(
                 "Expanding inputs for image tokens in LLaVa-NeXT should be done in processing. "
                 "Please add `patch_size` and `vision_feature_select_strategy` to the model's processing config or set directly "

src/transformers/models/llava_next_video/modeling_llava_next_video.py

@@ -954,7 +954,7 @@ def forward(
             video_features = torch.cat(video_features, dim=0)
             video_feature_lens = torch.tensor(video_feature_lens, dtype=torch.long, device=video_features.device)
 
-        if legacy_processing:
+        if legacy_processing and (image_features is not None or video_features is not None):
             logger.warning_once(
                 "Expanding inputs for image.video tokens in LLaVa-NeXT-Video should be done in processing. "
                 "Please add `patch_size` and `vision_feature_select_strategy` to the model's processing config or set directly "

src/transformers/models/llava_next_video/modular_llava_next_video.py

@@ -476,7 +476,7 @@ def forward(
             video_features = torch.cat(video_features, dim=0)
             video_feature_lens = torch.tensor(video_feature_lens, dtype=torch.long, device=video_features.device)
 
-        if legacy_processing:
+        if legacy_processing and (image_features is not None or video_features is not None):
             logger.warning_once(
                 "Expanding inputs for image.video tokens in LLaVa-NeXT-Video should be done in processing. "
                 "Please add `patch_size` and `vision_feature_select_strategy` to the model's processing config or set directly "

src/transformers/models/vipllava/modeling_vipllava.py

@@ -472,7 +472,7 @@ def forward(
                 pixel_values=pixel_values, vision_feature_layers=vision_feature_layers
             )
 
-        if legacy_processing:
+        if legacy_processing and image_features is not None:
             logger.warning_once(
                 "Expanding inputs for image tokens in VipLLaVa should be done in processing. "
                 "Please add `patch_size` and `vision_feature_select_strategy` to the model's image processing config. "