# Copyright (c) Alibaba, Inc. and its affiliates. """ 1. Installation EvalScope: pip install webdataset 2. Run eval task """ import torch from evalscope.run import run_task from evalscope.utils.logger import get_logger logger = get_logger() def run_eval(): # Prepare the config task_cfg = { 'eval_backend': 'RAGEval', 'eval_config': { 'tool': 'clip_benchmark', 'eval': { 'models': [ # { # "model_name": "AI-ModelScope/chinese-clip-vit-large-patch14-336px", # } { 'model_name': 'internvl2-8b', 'api_base': 'http://localhost:8008/v1', 'prompt': '简要描述这张图片,必须使用中文,描述不要太长', } ], 'dataset_name': ['muge'], # "dataset_name": ["custom"], # "data_dir": "custom_eval/multimodal/text-image-retrieval", 'split': 'test', 'task': 'image_caption', 'batch_size': 2, 'num_workers': 1, 'verbose': True, 'skip_existing': False, 'limit': 10, }, }, } # Run task run_task(task_cfg=task_cfg) if __name__ == '__main__': run_eval()