LWShowTime

Why the Caption task use drop_path_rate=0 while the VQA task use drop_path_rate=0.1? Is this for better performance after your experiments?

Add "--" to make default argument works.

The "iteration decoder" in Figure.2. should be "iteration encoder".

assert dim1 > dim2, \ 'Merging tensors is not allowed here!

8

Something wrong about the dimension in deepspeed's auto-tp.py: 

New version <LISA-13B-llama2-v1-explanatory> inference VERY SLOWLY

3

新版本LISA推理速度太慢了，同样的一个分割命令旧版本几秒钟，新版本需要几十分钟。

Avoid the channel exception.

In the codes here, what is the function of the mask_transform? It seems like the h and w not to be the real size of the input image, otherwise we...

传300,400这种会报错。

[Your model link here](https://github.com/whai362/PVT/tree/v2/segmentation) are all down and can't find the model weights. Could you please update them? Thanks!

 From three months, can't download ADE20K using the url. Issue is recorded as the screen shot shows.