Oneflow-Inc/one-yolov5

Issues

• 使用的oneflow的版本是多少

  #126 opened 5 months ago by lifeyou
  0

• AttributeError: module 'oneflow.distributed' has no attribute 'barrier'

  #124 opened 2 years ago by jie311
  0

• Guide:

  #116 opened 2 years ago by ccssu
  0

• 同步上游分支将 ultralytics_yolov5 迁移到 one-yolov5 小结

  #107 opened 2 years ago by ccssu
  0

• one-yolov5/classify/train.py 脚本 nsys 报告 【2023-03-29】

  #123 opened 2 years ago by ccssu
  0

• [TODO] 开发memory_reserved算子 和 性能优化

  #122 opened 2 years ago by ccssu
  2

• TODO：待解决的问题

  #115 opened 2 years ago by ccssu
  2

• 记录使用YOLOv5n-cls模型进行分类任务训练

  #117 opened 2 years ago by ccssu
  2

• 关于分类模型训练测试， 每次比PyTorch慢几秒的原因&可复现代码

  #120 opened 2 years ago by ccssu
  0

• 使用sync-bn后速度下降快

  #65 opened 2 years ago by ccssu
  5

• 模型导出 ：export failure 10.7s: cannot unpack non-iterable NoneType object

  #112 opened 2 years ago by ccssu
  4

• ddp 模式使用deepcopy(de_parallel(m)) 发生内存泄露

  #101 opened 2 years ago by ccssu
  2

• 记录一个错误： 升级oneflow0.9.1.dev20230218+cu117 版本后 `classify/train.py` 脚本训练失败

  #113 opened 2 years ago by ccssu
  2

• 日期2023-02-11后oneflow包导致，程序在运行过程中出现错误导致程序卡死。

  #114 opened 2 years ago by ccssu
  1

• 验证 one-yolov5使用amp准确性日志

  #70 opened 2 years ago by ccssu
  3

• one-yolov5 v1.2.0 精度测试目标精度

  #108 opened 2 years ago by ccssu
  1

• 记录一个错误，训练完成验证时 wandb报错

  #110 opened 2 years ago by ccssu
  1

• TypeError: __init__(): incompatible constructor arguments. The following argument types are supported:

  #109 opened 2 years ago by ccssu
  1

• segment/train.py loss对齐记录

  #105 opened 2 years ago by ccssu
  2

• 对下载的模型文件位置希望提示绝对路径

  #102 opened 2 years ago by doombeaker
  0

• C++版本的API

  #90 opened 2 years ago by xifen523
  3

• terminate called after throwing an instance of 'oneflow::RuntimeException'

  #106 opened 2 years ago by ccssu
  3

• 显存问题跟踪

  #71 opened 2 years ago by BBuf
  7

• Object of type Tensor is not JSON serializable

  #86 opened 2 years ago by ccssu
  0

• 8卡ddp增加batch_size，精度值严重下降

  #80 opened 2 years ago by ccssu
  8

• 论文解读中YOLOv6的github链接放错了

  #59 opened 2 years ago by dayekuaipao
  1

• one-yolov5 采用synbn训练yolov5n结果

  #21 opened 2 years ago by ccssu
  6

• oneflow backend yolov5 性能调试

  #46 opened 2 years ago by BBuf
  8

• pytorch fp32的训练

  #37 opened 2 years ago by ccssu
  9

• 最新oneflow包在one-yolov5中使用sync-bn导致显存爆炸

  #50 opened 2 years ago by ccssu
  30

• ddp模式打开flow.cuda.synchronize()， oneflow会卡死，pytorch 正常

  #47 opened 2 years ago by ccssu
  0

• 精度验证的两组实验

  #38 opened 2 years ago by ccssu
  8

• one-yolov5 sync-bn 断点续训精度下降

  #26 opened 2 years ago by ccssu
  8

• one-yolo训练精度测试

  #19 opened 2 years ago by ccssu
  4

• Pretrained Checkpoints

  #17 opened 2 years ago by ccssu
  3

• 最近实验训练达不到目标精度的原因

  #43 opened 2 years ago by ccssu
  8

• oneflow-yolov5s 内存和显存跟进

  #6 opened 2 years ago by BBuf
  17

• 关于one-yolo训练yolov5s查找训练不到目标精度的原因

  #18 opened 2 years ago by ccssu
  0

• one-yolo coco数据集训练yolov5s

  #15 opened 2 years ago by ccssu
  16

• eager module有free eager tensor时访问graph.state_dict()会挂掉

  #23 opened 2 years ago by BBuf
  9

• one-yolo 模型拓展

  #11 opened 2 years ago by ccssu
  1

• fuse bn之后nn.Graph编译失败

  #20 opened 2 years ago by BBuf
  4

• one-yolov5与torch-yolov5 loss 对齐测试

  #10 opened 2 years ago by ccssu
  0

• one-yolov5与torch-yolov5 loss对齐

  #9 opened 2 years ago by ccssu
  0

• OneDet项目调研计划

  #1 opened 2 years ago by rentainhe
  4

• LazyConfig的调研(from Xiaying Bai)

  #2 opened 3 years ago by rentainhe
  1