splinter21

Results 11 issues of splinter21

Is there any comparisons about the sp models in the additional pack？

![image](https://user-images.githubusercontent.com/26759495/180635889-a4c7fddf-4435-4ecf-a4a0-8ce6883ac3d2.png)

What's the training set?

What's the network architecture of the downsampler? How to train it?

ecb支持加残差，但是ecbsr的configs里面没有用残差，为啥？

idt_ecbsr: 0

Is there a script for dataset generation?

8 comment

I think when training, it's better to dynamiclly generate various line style images instead of using the generated 2000+ dataset.

请教一下，opensora1.2里的视频vae里的均值标准差归一化的这几个值，是如何统计出来的

2 comment

shift = (-0.10, 0.34, 0.27, 0.98) scale = (3.85, 2.32, 2.33, 3.06) 这2个参数

context并行vae encoder效果不对

1 comment

### Feature request / 功能建议 现在的代码，当cp_size开2,49切成25+24编码做conv，2边既没有overlap，也互相看不到，效果对不齐的 ### Motivation / 动机 训练显存不够

帧数呈现4x+1形式，是有什么深意吗？

3 comment

RT

Discord invite link expired

![image](https://github.com/user-attachments/assets/6fd13e20-e950-4417-8f9c-1fc92aadb079)

bug

Will the maskgct training code be made public?

5 comment

RT

bug