Though supported, it is not recommended to use batch inference in onnxruntime for DETR, because there is huge performance gap between ONNX and torch model ...
確定! 回上一頁