LI Wentong

@ssli23 Sorry to reply later. The `num_classes` in the config is the actual category number, which need not add 1. So, `num_classes `should be 1 for custom class of 1...

Hello@zhaoyangwei123 The large loss is normal for box2mask. I upload my training [log file](https://drive.google.com/file/d/14SvFTypqtx7TSMl3hw5VNNbV7vG0N9tc/view?usp=sharing) for coco (r-101). You can refer to it. I didn't encounter the above problem. It seems...

@zhaoyangwei123 I suggest you firstly try VOC with RTX2080TI GPU. VOC needs the less GPU memory with less training time. The VOC link with coco-format annotaions is [here](https://drive.google.com/file/d/16Mz13NSZBbhwPuRxiwi7ZA2Qvt9DaKtN/view). I guess...

@zhaoyangwei123 I have fixed this issue. When batch_size=1, the loss values will appear `nan ` value. You can try the current codes. Please note when batch_size=1, the learning rate `lr`...

@zhaoyangwei123 When did this erro appear? At the starting or during training process? I have test the code and config with `800x800` and `bs=1`, and the training work fine. According...

@3353463971 Sorry to reply later! Please refer to this [issue](https://github.com/LiWentomng/boxlevelset/issues/15).

> same problem. I solved it by manually creating the temporary directory. For example, in @ziqingcheryl 's case: > > ```shell > mkdir //ODISE/third_party/Mask2Former/build/temp.linux-x86_64-cpython-39 > ``` > > , and...

Any suggestings to address this issue? Thanks a lot in advance.

I found this issue is about the installation of `xformer` . When I build from latest source ``` pip install ninja pip install -v -U git+https://github.com/facebookresearch/xformers.git@main#egg=xformers ``` instead of `pip...

Thanks for your answer. When using the model of your `debug` weights, the loss is normal. I try to find the problem in my weights.