Yongchuan Cui

用GhostModule替换Conv2d，loss降的很慢？

8

我直接将efficientnet里面的MBConvBlock中的Conv2d替换为GhostModule： `Conv2d(in_channels=inp, out_channels=oup, kernel_size=1, bias=False)` 替换为 `GhostModule(inp, oup)`， 其他参数不变，为什么损失比以前收敛的更慢了，一直降不下来？请问需要修改其他什么参数吗？

### Describe the bug The function `is_off_screen()` , which determines whether or not some `Mobject` is out of frame, is weird. Should `get_left()` and `get_right()`, `get_bottom()` and `get_top()` reverse position?...

[CLI]: wandb: Network error (SSLError), entering retry loop.

6

### Describe the bug weird problem, just ```python wandb.init() ``` and I got ``` Retry attempt failed: Traceback (most recent call last): File "/usr/local/python3/lib/python3.7/site-packages/urllib3/connectionpool.py", line 710, in urlopen chunked=chunked, File...

When training using the option `--resume` , it always failed training due to Unexpected key(s) in state_dict and size mismatch. Could you please tell us the hyperparamers of these pretrained...

Release the hyperparamers of pre-trained models

2

When training using the option `--resume_checkpoint logs/256x256_diffusion.pt`, it always failed training due to `Unexpected key(s) in state_dict` and `size mismatch`. Could you please tell us the hyperparamers of these pretrained...

如何引入多个模态训练？

2

例如，能不能使用一轮对话+2张图像+bbox这样的训练方式进行训练？

About generalizing to unseen degradation type?

1

Thanks for your impressive work! In the appendix, the types of degradation observed in Table 5 and Figure 17 closely resemble those of the trained degradation. Have you considered applying...

Is there any way that I can replace the nn.Conv2d using the sgcn layer?