reproduce the exp results
Opened this issue · 9 comments
i see that there is a seed fixed in the train.py. but i can not get the consistent results every training.
i see that there is a seed fixed in the train.py. but i can not get the consistent results every training.
i see that there is a seed fixed in the train.py. but i can not get the consistent results every training.
Do you get 81.3 miou after isic18 training?
yes. but i use the depth [2,2,9,2] and vmamba_small_e238_ema.pth pretrained model.
but anthor claim that 81.3 is get using the depths [2,2,2,2],which i can not get the results.
yes. but i use the depth [2,2,9,2] and vmamba_small_e238_ema.pth pretrained model. but anthor claim that 81.3 is get using the depths [2,2,2,2],which i can not get the results.
thanks!!!!!! I'll try it
@fAnchao1 I meet the same problem. I used the same parameters and fixed seed on another task, but the results were completely different each time. Have you solved it?
@sakurarma not yet
I meet the same problem.I just changed the dataset,but the results were diffident each time on the same GPU.
yes. but i use the depth [2,2,9,2] and vmamba_small_e238_ema.pth pretrained model. but anthor claim that 81.3 is get using the depths [2,2,2,2],which i can not get the results.
i still cant get the 81.3 miou only 80.5 and my batchsize is 20 when the depth is [2,2,9,2] ,What is your batch size setting? May i see your train.info.log?thanks a lot!
32,what miou you get
…
---Original--- From: @.> Date: Sat, May 4, 2024 13:53 PM To: @.>; Cc: @.@.>; Subject: Re: [JCruan519/VM-UNet] reproduce the exp results (Issue #47) yes. but i use the depth [2,2,9,2] and vmamba_small_e238_ema.pth pretrained model. but anthor claim that 81.3 is get using the depths [2,2,2,2],which i can not get the results. i still cant get the 81.3 miou ,What is your batch size setting? May i see your train.info.log?thanks a lot! — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>
only get miou: 0.8029449451678118