search

LeiaLi / SRDiff

171 stars 19 forks source link

训练代码是不是缺少一些东西 #4

Open ElderWu opened 1 year ago

ElderWu commented 1 year ago

您好！我调通训练代码之后，但是发现只跑了一个Epoch就结束了，是不是缺了什么？

whyandbecause commented 1 year ago

你好，请问你解决了吗

YilanWang commented 1 year ago

同样发现了，这都什么鬼啊……请问楼主解决了吗

LeiaLi commented 1 year ago

Sorry for missing some lines when uploading the code. Now we have updated the new version. https://github.com/LeiaLi/SRDiff/blob/main/tasks/trainer.py#L67 Thanks for the questions.

fantas-san commented 1 year ago

Sorry for missing some lines when uploading the code. Now we have updated the new version. https://github.com/LeiaLi/SRDiff/blob/main/tasks/trainer.py#L67 Thanks for the questions. I added a similar statement, but the RRDB encoder can be trained within tens of minutes. Is there a problem with the setting of max_updates? I now try to expand max_updates by 300 times, the code is still running.

2590609710 commented 10 months ago

Sorry for missing some lines when uploading the code. Now we have updated the new version. https://github.com/LeiaLi/SRDiff/blob/main/tasks/trainer.py#L67 Thanks for the questions. I added a similar statement, but the RRDB encoder can be trained within tens of minutes. Is there a problem with the setting of max_updates? I now try to expand max_updates by 300 times, the code is still running.

Hello, may I know if the training of the RRDB encoder can be completed within several tens of minutes? I trained the RRDB on a 3090 GPU for 100,000 iterations, which took around one and a half days. What could be the reason for this extended duration?