I think there is a wrong!

[rikyaa]

Training Stage with Concealing Field

if mode != 'test':
    if i_level == 0:
        accum_prod_dark = torch.cat([torch.ones_like(alpha[..., :1] * coarse_dark[0]),
            torch.cumprod(1.0 - alpha[..., :-1] + eps, dim=-1) * coarse_dark[1:]], dim=-1,)
    else:
        accum_prod_dark = torch.cat([torch.ones_like(alpha[..., :1]) * fine_dark[0],
            torch.cumprod(1.0 - alpha[..., :-1] + eps, dim=-1) * fine_dark[1:]], dim=-1,)

here :*torch.cumprod(1.0 - alpha[..., :-1] + eps, dim=-1) fine_dark[1:]**维度是（4096，64）（128，）没办法相乘的呀？期待你的回答

[rikyaa]

Author I'm very sorry to bother you, your article is very well written but the code you uploaded I double checked still think there is a dimensional mismatch here, looking forward to your answer, hopefully it will solve the confusion I've been having.

[cuiziteng]

您好 感谢关注 您可以print一下看看维度 fine那里的维度也是128 您可以自己跑一下代码 不会报错的

[rikyaa]

Hello author, your paper is very excellent. I want to make changes to the code. I modified fine_dark for the problem posted earlier, this time I re-downloaded the code and only modified the batchsize.I debugged the code you wrote again, but there is still a problem. I still think that the latitude in else does not match. torch.cumprod(1.0 - alpha[..., :-1] + eps, dim=-1) latitude is (32, 64), and fine_dark[1:] is (192,). This question has puzzled me for a long time, so I wanted to ask you. Did I understand the problem or something else? I look forward to your answer. 作者您好，你的论文很优秀我想对代码做出修改，前面发的问题我修改了fine_dark，这次我重新下载了代码只修改了batchsize我再次调试了你写的代码，还是存在问题，我还是觉得else这里面纬度是不匹配的，这个纬度是(32，64)，这里是(192，)，这个问题困惑了我很久希望所以想来问问你，是我理解出了问题还是其他的呢，期待您的解答

[cuiziteng]

您好，else这里表示的是通过NeRF的fineMLP, fineMLP最后的输出尺度不是64而是192，我也不清楚您这里是怎么出现的出来(32, 64)这样的情况，理论上讲理应是(32, 192)。为了验证是否是我的代码有问题，我这边重新clone了一下代码，并且重新跑了一遍，完全没有问题，问一下您这里是根据指令跑代码了嘛，还是只是在py文件调试出错呢？

[rikyaa]

谢谢你的回复，问题已经解决了，代码没有问题，是我修改代码的过程中存在错误，感谢！！！