在300I Pro得机器上用这个镜像和这个repo测resnet50的int8性能和精度，精度比repo上有下降，请问是什么原因呢？
镜像：swr.cn-south-1.myhuaweicloud.com/ascendhub/ascend-pytorch:24.0.RC1-A2-1.11.0-ubuntu20.04
repo：https://gitee.com/ascend/ModelZoo-PyTorch/tree/master/ACL_PyTorch/built-in/cv/Resnet50_Pytorch_Infer_poc#https://gitee.com/link?target=https%3A%2F%2Fwww.hiascend.com%2Fdocument%2Fdetail%2Fzh%2Fcanncommercial%2F63RC1%2Fdevtools%2Fauxiliarydevtool%2Fatlasamctonnx_16_0012.html
另外，amct_onnx calibration --model resnet50_official.onnx --save_path amct_model/resnet50 --input_shape "input:1,3,224,224" --data_dir amct_bin --data_types "float32" --calibration_config quant.cfg执行量化的时候有这些Warning是正常的吗？
点赞
收藏
回复
分享
举报
已解决
镜像：swr.cn-south-1.myhuaweicloud.com/ascendhub/ascend-pytorch:24.0.RC1-A2-1.11.0-ubuntu20.04
repo：https://gitee.com/ascend/ModelZoo-PyTorch/tree/master/ACL_PyTorch/built-in/cv/Resnet50_Pytorch_Infer_poc#https://gitee.com/link?target=https%3A%2F%2Fwww.hiascend.com%2Fdocument%2Fdetail%2Fzh%2Fcanncommercial%2F63RC1%2Fdevtools%2Fauxiliarydevtool%2Fatlasamctonnx_16_0012.html
另外，amct_onnx calibration --model resnet50_official.onnx --save_path amct_model/resnet50 --input_shape "input:1,3,224,224" --data_dir amct_bin --data_types "float32" --calibration_config quant.cfg执行量化的时候有这些Warning是正常的吗？