From 4fb980be72557807b294e25ceed57c3b60f2cb1e Mon Sep 17 00:00:00 2001 From: suluyana Date: Wed, 30 Oct 2024 16:15:13 +0800 Subject: [PATCH] support image url --- modelscope/pipelines/multi_modal/ovis_vl_pipeline.py | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/modelscope/pipelines/multi_modal/ovis_vl_pipeline.py b/modelscope/pipelines/multi_modal/ovis_vl_pipeline.py index e419e419..426d9c97 100644 --- a/modelscope/pipelines/multi_modal/ovis_vl_pipeline.py +++ b/modelscope/pipelines/multi_modal/ovis_vl_pipeline.py @@ -1,7 +1,7 @@ import torch from typing import Any, Dict, Union -from PIL import Image +from modelscope.preprocessors import load_image from modelscope import AutoModelForCausalLM from modelscope.metainfo import Pipelines, Preprocessors from modelscope.models.base import Model @@ -42,8 +42,8 @@ class VisionChatPipeline(VisualQuestionAnsweringPipeline): def preprocess(self, inputs: Dict[str, Any]): text = inputs['text'] - image = inputs['image'] - image = Image.open(image) + image_path_or_url = inputs['image'] + image = load_image(image_path_or_url) query = f'\n{text}' _, input_ids, pixel_values = self.model.preprocess_inputs(query, [image]) attention_mask = torch.ne(input_ids, self.text_tokenizer.pad_token_id)