General question regarding training multipleviews scenes

[TheCrappyActionFigureMaster]

Hello, first of all, this is an outstanding paper. Thank you to the author for your hard work！
I have a few questions I'd like to ask.
I was able to achieve great results using the DyNeRF data. I then tried using fixed multi-view data but consistently received less than satisfactory results.
My configuration basically uses the default settings for multipleviews, and I have tried increasing/decreasing parameters like coarse iter, iter, densify_until_iter, and lr, among others.
However, even when using the roast_salmon data, I couldn't achieve better results.
(After obtaining images for each frame, I ran multiple view progress.)
The results didn't appear to improve the rendering outcomes through iteration.
Could you advise on where I might be going wrong, or provide suggestions on any direction I could take? I would be very grateful for any advice!
Feel free to let me know if you need any changes or further assistance!
Below is the training configuration that is attached in the image.

您好首先这是篇非常杰出的论文 辛苦作者了
我有几个问题想请问
我在dynerf data 上取得很好的结果
我接着尝试使用固定多视角资料却总是得到不尽理想的结果
我的config基本上使用multipleview的default
并尝试了增加/减少coarse iter, iter, densify_until_iter, lr 等等
但就算是使用roast_salmon data也无法成功产出较佳的结果
（取得每个frame images后跑multipleviewprogress）
结果看起来没有因为iterate 后提昇render结果
想请问我是哪个环节可能出错？
或是任何方向的建议都非常的感谢！
以下是附图的training config
base = './default.py'
OptimizationParams = dict(
dataloader=True,
iterations = 60000,
batch_size=1,
coarse_iterations = 3000, #########
densify_until_iter = 10000, #########
densify_from_iter = 130,
pruning_from_iter = 130,
pruning_interval = 100
)