Xiaoyu Zhang

关于opencv版本的问题

1

我用opencv3.4.0编译里代码，在执行批量测试功能报错，OpenCV Error: Parsing error (Missing or invalid SVM type) in read_params, file /home/just_sort/opencv-3.4.0/modules/ml/src/svm.cpp, line 2198 terminate called after throwing an instance of 'cv::Exception' what(): /home/just_sort/opencv-3.4.0/modules/ml/src/svm.cpp:2198: error: (-212) Missing or invalid...

Add dropout1d/2d/3d api

1

解决：https://github.com/Oneflow-Inc/libai/issues/342#issuecomment-1207565463 - [x] 注册 nn.Dropout1d/2d/3d api。 - [x] 添加文档和接口测试。   

fix shufflenet bug

1

fix shufflenet bug， 

Fix inplace mul 0size check bug

2

#close https://github.com/Oneflow-Inc/oneflow/issues/9131

来源：yolov5加载yolov5s.onnx做预测的时候。 pytorch脚本： ``` import torch as flow targets = flow.randn((0, 6), device="cuda") print(targets.shape) height, width = 640, 640 targets[:, 2:] *= flow.tensor((width, height, width, height), device="cuda") ``` 输出： ``` torch.Size([0,...

fix https://github.com/thu-pacman/PET/issues/5#issuecomment-1162861769

ed, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning). AR -o .build_release/lib/libcaffe.a LD -o .build_release/lib/libcaffe.so.1.0.0-rc3 CXX/LD -o .build_release/tools/compute_image_mean.bin CXX/LD -o .build_release/tools/upgrade_solver_proto_text.bin CXX/LD -o .build_release/tools/caffe.bin CXX/LD...

fix batchnorm infer dtype failed in half inference

4

#close https://github.com/Oneflow-Inc/oneflow/issues/9381

Add cuda amp scaler for eager

5

eager amp 支持 scaler ，eager 可以做完整的 amp 训练。 https://github.com/Oneflow-Inc/OneTeam/issues/1754 export ONEFLOW_VM_COMPUTE_ON_WORKER_THREAD=0之后 eager amp测试结果：   无论是训练速度还是显存占用相比于fp32模式都有较大的提升。