测试时间问题 #125
Comments
Title: test time issues I have successfully run my own dataset based on the code you provided, but it takes a lot of time to predict the output per sample in the test phase. How did you solve this problem? During the test, I see that you did not adjust the diffusion_steps, nor did you use DDIM or DPM_solver |
|
你好，请问一下，在初始化UNetModel时，out_channel 是不是设置为1(下载下来的代码好像是为out_channelr = 2)？我是在自己数据及上跑的分割任务，输入图像为3通道，mask为单通道 |
Hello, may I ask, when initializing UNetModel, is out_channel set to 1 (the downloaded code seems to be out_channelr = 2)? I am running a segmentation task on my own data, the input image is 3 channels, and the mask is a single channel |
|
输出通道其实是根据任务确定的，输出通道是1就需要根据与之进行分割，输出通道是2就直接分为了前景背景 |
The output channel is actually determined according to the task. If the output channel is 1, it needs to be divided according to it. If the output channel is 2, it is directly divided into the foreground and background. |
谢谢，最终我改成了out_channels=1,我做的是二值分割 |
Thank you, I finally changed to out_channels=1, what I did was binary segmentation |
你好 我也是做二值分割的 但是遇到StopIteration问题无法解决（刚提问） 此外能请教一下你用的什么gpu吗 我怀疑我这2080ti跑不起来 |
Hello, I am also doing binary segmentation, but I can’t solve the StopIteration problem (just asked). Also, can you tell me what kind of GPU you are using? I suspect that my 2080ti can’t run it. |
源代码好像就是无法自动停止的，把循环那里改一下就可以了。我用的3090，不过也是很慢就是了 |
The source code seems to be unable to stop automatically. Just change the loop. I use 3090, but it is also very slow. |
感谢 我用在二值分割上 效果很不理想 想交流一下 你的效果好吗 |
Thank you. I used it for binary segmentation and the effect was not ideal. I would like to share your experience. Is your effect good? |
|
我也是用的2080ti，感觉确实不太行 |
I also use a 2080ti, and it really doesn’t feel good. |
我根据您提供的代码已经成功运行了自己的数据集，但是在测试阶段每个样本需要消耗大量的时间进行输出预测。请问您是如何解决这个问题的呢？在测试过程中，我看您并没有调整diffusion_steps，也没有用到DDIM或者DPM_solver
The text was updated successfully, but these errors were encountered: