Transfer result about fet-gan

[Jeozhao]

Hi, @Moriatyyuan , I also encountered the same problem. Looking forward to the author's early reply.
Thanks!

[liweileev]

I'm not sure whether the transfer result converges slowly due to the difference in network initialization. This phenomenon also happens sometimes in my experiments. But it doesn't come up very often. I usually choose to re-initialize the network to restart training.

P.s Have you ever test the model with my pre-trained nets? If it can work well, I recommend trying a new initialization.

[Moriatyyuan]

Yes, I tested with your provided model, it can achieved good result. But I trained 20 epoch and still get no good transfer result. The transfer_L1 could not converge. On March 5, 2020 at 5:15:08 PM, liweileev ([email protected]) wrote: I'm not sure whether the transfer result converges slowly due to the difference in network initialization. This phenomenon also happens sometimes in my experiments. But it doesn't come up very often. I usually choose to re-initialize the network to restart training. P.s Have you ever test the model with my pre-trained nets? If it can work well, I recommend trying a new initialization. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3?email_source=notifications&email_token=AHJR2ODM2LAAVXX7AMEW6U3RF5NQZA5CNFSM4LCBDSR2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEN4FJ7Q#issuecomment-595088638>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AHJR2OEUGDQVVDWQ56KLTU3RF5NQZANCNFSM4LCBDSRQ> .

[Chanarin]

I'm not sure whether the transfer result converges slowly due to the difference in network initialization. This phenomenon also happens sometimes in my experiments. But it doesn't come up very often. I usually choose to re-initialize the network to restart training.

P.s Have you ever test the model with my pre-trained nets? If it can work well, I recommend trying a new initialization.

Please let me know the way to re-initialize the network to restart training.