Light--

get stuck by using muti a100 GPUs in training

3

### Search before asking - [X] I have searched the YOLOv5 [issues](https://github.com/ultralytics/yolov5/issues) and found no similar bug report. ### YOLOv5 Component Training ### Bug get stuck by using muti-a100 GPU...

# 请问哪个是变化learning rate的源文件？

1

RT 文件和目录有点confusing，没找到。。

hi, thanks for sharing! @ZongxianLee may i ask what software or package environment is needed to use your code? python2 or python3? what version of pytorch? and so on...

sovle the prepare_data error

1

when convert bin datasets to folders, it will report: ``` Traceback (most recent call last): File "prepare_data.py", line 12, in load_mx_rec(rec_path) File "/home/user1/InsightFace_Pytorch/data/data_pipe.py", line 92, in load_mx_rec label = int(header.label)...

And can we use the proposed FAN to estimate age and race attributes from face images ? Thx.

Line 94: - * mode: 字符串，“fan_in”，“fan_out”或“fan_avg”fan_in", "fan_out", "fan_avg". + * mode: 字符串，“fan_in”，“fan_out”或"fan_avg".

论文中提到生成了25个fdm（实验部分提到是9个）对应的掩模字典，能否共享最终训练好的模型和掩模字典？

啥意思，Gluon CV有multi-task的内置实现吗？你这个repo和他的关系是？