Overflowu7

icon company Beijing University of Technology icon location Beijing University of Technology icon location I am Wu Yichu, a student in Beijing university of technology. I want to study in Github and try to make some contributions to the community.

Results 5 comments of Overflowu7

How to make weights to decrease from 1 to 0 as the epoch increases?

ok here is my code: ``` class DC_and_CE_and_TIOU_Loss(nn.Module): def __init__(self, soft_dice_kwargs, ce_kwargs, ti_kwargs, weight_ce=1, weight_dice=1, weight_ti=1e-6, weight_bou=0, ignore_label=None, dice_class=SoftDiceLoss): """ Weights for CE and Dice do not need to sum...

How to use postprocessing?

> If you want to use postprocessing you will need to run all the folds. But don't worry about that. Postprocessing doesn't do much and it win't hurt to just...

torch.cuda.OutOfMemoryError: CUDA out of memory.

I also encountered this error during inference. I attempted to use 2 GPUs, but it appears that it is not functioning correctly. How can i do?

Problems with CUDA memory when validation from raw_data.

> Hi @Overflowu7 @yonuyeung and @PromiNent-Jin, > > sorry for the late reply to this issue! Per default, nnUNet v2 tries to perform the sliding window prediction on GPU. As...

Problems with CUDA memory when validation from raw_data.

The problem was solved when I updated to the new version. ------------------ 原始邮件 ------------------ 发件人: ***@***.***>; 发送时间: 2024年4月22日(星期一) 晚上8:49 收件人: ***@***.***>; 抄送: "- ***@***.***>; ***@***.***>; 主题: Re: [MIC-DKFZ/nnUNet] Problems with CUDA memory...