I want to know whether support single gpu for training?Thank you!

[lukaisheng1203]

[lukaisheng1203]

After I tried it was OK！Thanks

[GOATmessi8]

Great, thanks for your verification！

[Miraclemin]

can I know your super parameter? and the final mAP of the trained model after trained with one GPU, cause i trained but it is not converge, many thanks to you ~:)

[jfkkf123]

would you tell me which part should be changed if I want to use single gpu?thanks

[lukaisheng1203]

@Miraclemin I just change the lr to 0.0001 because the loss is nan after 4 epoch, I try it on my dataset, it is no result now, but I think it no problem

[lukaisheng1203]

@jfkkf123 I change little code in main.py, just tell me that YOLOv3 has no "module", so I change modle.module to modle,than it's ok

[zhanghongsir]

Do you get good results in your dataset? How about map?

[dodogoffy]

would you tell me which part should be changed if I want to use single gpu?thanks