[Fix] Fixing the multi-images error for llava-onevision (#1205)

2024-08-26 01:28:23 +08:00
parent bc4c7a3545
commit 66e7dcaf70
3 changed files with 91 additions and 1 deletions
--- a/python/sglang/srt/managers/tokenizer_manager.py
+++ b/python/sglang/srt/managers/tokenizer_manager.py
@@ -744,7 +744,9 @@ def get_pixel_values(
                    image,
                    tuple(int(x * 255) for x in processor.image_processor.image_mean),
                )
-                pixel_values = processor.image_processor(image)["pixel_values"][0]
+                pixel_values = processor.image_processor(image.convert("RGB"))[
+                    "pixel_values"
+                ][0]
            elif image_aspect_ratio == "anyres" or "anyres_max" in image_aspect_ratio:
                pixel_values = process_anyres_image(
                    image, processor.image_processor, image_grid_pinpoints