霁天

你好，关于chap5 的5.2节卷积公式有疑问。

4

**bug描述** 描述一下你遇到的bug, 例如报错位置、报错信息（重要, 可以直接截个图）等 ![Uploading image.png…]() 和之前学过的完全不一样： out_d=(in_d-k+2p)/s + 1 但是这本书里的公式为： out_d=(in_d-k+p+s)/s + 1 p==s??? 一般没听说默认步长一定等于padding吧？而且，padding这本书讲的也让人不明白，如果设定了padding是多少，那就是宽高都填1倍？这本书里是宽高都填1/2倍？卷积这部分真的是蒙了。用了那么久的卷积给我整不会了。 **版本信息** pytorch: torchvision: torchtext: ... 。

there is a runtime error when running the 7th code block. But I checked, the device type doesn't conflict each other. What's Wrong? -------------------------------------------------------------------------- RuntimeError Traceback (most recent call last)...

How do you continue your epoch dataload when iteration %epoch_size==0?

1

In you code, I see that you do the running loop by the iterating of the variable "iteration" from "args.start_iter" to "cfg['max_iter']". but I see that the declaration of dataloader——...

there are many errors in this implementation, it's really difficult to run it.

1

1st, backend.Embedding is removed in pytorch. 2nd, the tensors in this repo are not in the same device type, this really messed up the code, I tried to modify this...

Why come commands can run in original CMD but not in fluentterminal's cmd?

2

I installed Powershell7, and add it's path to the system's path variables. and then, when I type"pwsh-preview" in the plain cmd.exe, which locates in the address"$directory of powershell7$/7-preview/preview/", I can...

ConvCaps的m-step成员函数，第一行代码。 `r = r* a_in. 该函数的初值 r 是形状是 (b, B, C, 1), 而 a_in 的形状是 (b, C, 1). 不能广播，直接做哈达玛积会报错的。是不是需要对 a_in unsqueeze一下？

你的输出是8个1152为的胶囊，但是原文想表达的应该是1152个8维的向量，虽然整体标量个数一样，但是你的squash函数就成了对一个1152为的函数的作用了，文章肯定是希望对一个8维向量作用。个人拙见。

Isn't here a bug？

1

https://github.com/YuxianMeng/Matrix-Capsules-pytorch/blob/bc9e26dae9390bc96a5420a94cbe89432a1ad10a/Capsules.py#L86 pose's shape is (b, self.B, H, W), how can it be reshaped into ( b, 16, self.B, H, W) in the next line????? why does no one point this...

please add feature of saving chat history

3

Firstly thanks for developers' effort. Yet we need feature of saving chat history, although the current release allows exporting to files, it's not convenient for integrate all the chats in...

In the subsection **Deep Feature Flow Training** of section **3. Deep Feature Flow**. What do you mean when you say "When i > k, feature maps f_k are propagated to...