train

[SHIJILIANG01]

Thank you very much for your excellent work! I want to ask, I want to train 300000 times, but the training stopped at 2500000 times, after a few days, I want to follow 2500000 times, but why the epoch can't follow the last training, the epoch will still start from 1.

[SHIJILIANG01]

help me

[dengyecode]

In fact, due to a flaw in our code, it just doesn't inherit the "number" but actually inherits the model parameters after "2500000" iterations. To avoid confusion, we recommend you to create a new folder and copy the model files ("xx_net_D.pth" and "xx_net_G.pth") that you want to fine_tune into the new folder. Then continue training in the new folder

[SHIJILIANG01]

OK! Thank you! ------------------ 原始邮件 ------------------ 发件人: "dengyecode/hourglassattention" ***@***.***>; 发送时间: 2023年6月15日(星期四) 晚上9:32 ***@***.***>; ***@***.******@***.***>; 主题: Re: [dengyecode/hourglassattention] train (Issue #9) In fact, due to a flaw in our code, it just doesn't inherit the "number" but actually inherits the model parameters after "2500000" iterations. To avoid confusion, we recommend you to create a new folder and copy the model files ("xx_net_D.pth" and "xx_net_G.pth") that you want to fine_tune into the new folder. Then continue training in the new folder — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>