Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

标签的一些细节 #12

Open
starsky68 opened this issue Nov 10, 2020 · 4 comments
Open

标签的一些细节 #12

starsky68 opened this issue Nov 10, 2020 · 4 comments

Comments

@starsky68
Copy link

starsky68 commented Nov 10, 2020

@CaptainEven
打扰了,可以告知一下标签的格式细节呢,我明明使用的和mcmot一样的标签,但是,似乎一直出现越界的情况。

比如我有54个类,标签是这样的:
0.0 1.0 0.18229166666666666 0.37962962962962965 0.028125 0.10648148148148148
0.0 2.0 0.6484375 0.3731481481481482 0.03333333333333333 0.09351851851851851
0.0 3.0 0.7234375 0.49722222222222223 0.028645833333333332 0.09444444444444444
5.0 1.0 0.18802083333333333 0.7231481481481481 0.28489583333333335 0.3527777777777778
5.0 2.0 0.521875 0.4388888888888889 0.17552083333333332 0.19814814814814816
17.0 1.0 0.190625 0.4064814814814815 0.03802083333333333 0.09722222222222222
17.0 2.0 0.6494791666666667 0.4046296296296296 0.0484375 0.10833333333333334
39.0 1.0 0.4265625 0.18518518518518517 0.2526041666666667 0.2037037037037037
39.0 2.0 0.6135416666666667 0.1 0.0828125 0.11481481481481481
3.0 1.0 0.7208333333333333 0.25833333333333336 0.11927083333333334 0.18611111111111112
38.0 1.0 0.8604166666666667 0.4388888888888889 0.04114583333333333 0.08888888888888889
5.0 3.0 0.8666666666666667 0.6546296296296297 0.26666666666666666 0.19166666666666668
2.0 1.0 0.6552083333333333 0.34074074074074073 0.0109375 0.025
0.0 4.0 0.46510416666666665 0.362962962962963 0.0296875 0.07407407407407407
/pytorch/aten/src/THCUNN/ClassNLLCriterion.cu:108: cunn_ClassNLLCriterion_updateOutput_kernel: block: [0,0,0], thread: [1,0,0] Assertion t >= 0 && t < n_classes failed.
/pytorch/aten/src/THCUNN/ClassNLLCriterion.cu:108: cunn_ClassNLLCriterion_updateOutput_kernel: block: [0,0,0], thread: [2,0,0] Assertion t >= 0 && t < n_classes failed.
/pytorch/aten/src/THCUNN/ClassNLLCriterion.cu:108: cunn_ClassNLLCriterion_updateOutput_kernel: block: [0,0,0], thread: [4,0,0] Assertion t >= 0 && t < n_classes failed.

RuntimeError: CUDA error: device-side assert triggered

@CaptainEven
Copy link
Owner

@starsky68 标签保持一模一样的即可, 问题的原因可能不在标签,而在你的一些参数设置,具体可以提前打断点debug看下,祝好运,最近有点忙....

@starsky68
Copy link
Author

@starsky68 标签保持一模一样的即可, 问题的原因可能不在标签,而在你的一些参数设置,具体可以提前打断点debug看下,祝好运,最近有点忙....

谢谢回复,我根据计算只修改了cfg中3个yolo的:
[convolutional]
size=1
stride=1
pad=1
filters=177
activation=linear
中的filters=177,根绝54类得到,因为原来是30,由5类得到。其他的就是按照yolov3官方提供教程,把.data文件中的训练路径以及类别数设置和cfg相同,都为54,其他的,选择没有修改。请问你说的参数,是那些必要的超参数吗?可以给一些提示吗,非常感谢。

@kenrickfernandes
Copy link

Hello, were you able to solve this issue? I am facing the same "/pytorch/aten/src/THCUNN/ClassNLLCriterion.cu:108: cunn_ClassNLLCriterion_updateOutput_kernel: block: [0,0,0], thread: [1,0,0] Assertion t >= 0 && t < n_classes failed." error. I have checked my labels and they seem to be correct (ie no out of class labels)

@kenrickfernandes
Copy link

你好先生,
你能解决这个问题吗

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants