通义千问,源码安装https://github.com/modelscope/modelscope.git
创建多模态视觉定位任务失败,代码和错误信息如下
visual_det = pipeline('visual-grounding')
Traceback (most recent call last):
File "test.py", line 18, in <module>
visual_det = pipeline('visual-grounding')
File "/workspace/modelscope/modelscope/pipelines/builder.py", line 133, in pipeline
pipeline_name, default_model_repo = get_default_pipeline_info(task)
File "/workspace/modelscope/modelscope/pipelines/builder.py", line 182, in get_default_pipeline_info
pipeline_name = list(PIPELINES.modules[task].keys())[0]
KeyError: 'visual-grounding'
根据官方文档,该功能是支持的。但源码中pipeline中有该任务名称,Tasks中没有涵盖该功能。请问该功能是否可用?是调用方式有问题还是不支持?