Tramac

Please confirm the path of CUDA in `~/.bashrc`.

Please delete ```jpu=args.jpu``` in this [line](https://github.com/Tramac/awesome-semantic-segmentation-pytorch/blob/7fbe397a9add570fe1ebee8654898f2b3ba1942f/scripts/train.py#L162).

Great!

Please post the detailed cmd.

Pleae change the [path](https://github.com/Tramac/awesome-semantic-segmentation-pytorch/blob/7fbe397a9add570fe1ebee8654898f2b3ba1942f/core/models/model_store.py#L62) to your model path.

> 好的，谢谢啊  > […](#) > ------------------ 原始邮件 ------------------ 发件人: "Tramac/awesome-semantic-segmentation-pytorch" ***@***.***>; 发送时间: 2021年4月22日(星期四) 晚上8:41 ***@***.***>; ***@***.******@***.***>; 主题: Re: [Tramac/awesome-semantic-segmentation-pytorch] 用deeplabv3训练效果太差 (#129) 你好，我用deeplab v3+ 训练出现了下面的错误： File "train.py", line 326, in trainer = Trainer(args) File "train.py",...

It is possible, network with different modules may depend on different parameters, this [paper](https://arxiv.org/abs/1903.11816) maybe helpful for searching th optimal parameter.

Hi, my opinion is as follows: 1. The mean value was calculated in thie [line](https://github.com/Tramac/awesome-semantic-segmentation-pytorch/blob/7fbe397a9add570fe1ebee8654898f2b3ba1942f/core/utils/score.py#L57), `IoU` contains the values of per class. 2. Both are ok, I have seen some...

> > deeplabv3plus demo > > Missing key(s) in state_dict: "auxlayer.block.0.weight", "auxlayer.block.1.weight", "auxlayer.block.1.bias", "auxlayer.block.1.running_mean", "auxlayer.block.1.running_var", "auxlayer.block.4.weight", "auxlayer.block.4.bias". > > When i set aux =False is Ok > > What the...

If you use the deeplab, maybe you can check [here](https://github.com/Tramac/awesome-semantic-segmentation-pytorch/blob/7fbe397a9add570fe1ebee8654898f2b3ba1942f/core/models/deeplabv3.py#L51).