[BUG] It seems a bug.

[Veason-silverbullet]

LLaVA-NeXT/llava/conversation.py

Lines 229 to 242 in 79ef45a

	def get_images(self, return_pil=False, return_path=False):
	images = []
	for i, (role, msg) in enumerate(self.messages[self.offset :]):
	if i % 2 == 0:
	if type(msg) is tuple:
	msg, image, image_process_mode = msg
	if type(image) != list:
	image = [image]
	for img in image:
	if not return_path and self.is_image_file(img):
	img = self.process_image(img, image_process_mode, return_pil=return_pil)
	else:
	images.append(img)
	return images

It seems that the pixels img are not appended to images in the if not return_path and self.is_image_file(img) branch. Is it a bug?