Comments (8)
Hi, I am confused about this default config because a support image consists of (image, mask) which is 4-d, so we can not use pretrained ResNet weights and add a conv2
in the BaseStem
.
However, FREEZE_CONV_BODY_AT=2
, contains BaseStem
.
After I debug this line, when we use this freeze config, self.conv2(x).requires_grad is false. This means we can not train this weight.
What is wrong with my view? Hope you can help me.
from dcnet.
I followed the install.md build environment to build four or five times, and every time I encountered ImportError:
_/DCNet-main/maskrcnn_benchmark/_C.cpython-36m-x86_64-linux-gpu.so: undefined symbol: __cudaPopCallConfiguration. _ But I am completely I installed it according to install.md. The cuda version of my computer is 9.0.176, and gcc tried 5.5 and 7.5, but they all reported the same error. What are the specific torch, cuda, torchvision, gcc and other environments you are using?
from dcnet.
@Wei-i Looking forward to your quick reply!
from dcnet.
It seems that cuda version may be not compatible with your environment.
from dcnet.
Only difference of my config is CUDA 10.1.
The others follow the official install.md.
from dcnet.
Thank you very much for your kindness!I will try again.
from dcnet.
I tried to switch the cuda soft connection in /usr/local to cuda10.1, the program "from maskrcnn_benchmark import _C" still reports the original error, that is, undefined symbol: __cudaPopCallConfiguration. I use the command "python setup.py build develop" No error will be reported.
If I use cuda10.1 from the build environment,, when using "python setup.py install --cuda_ext --cpp_ext" to compile apex, the error "cuda version does not match, only supports cuda9.0.176" will be reported.
Could you please confirm that your torch and torchvision 1.1.0 and 0.3 respectively? Thank you for your help! @Wei-i
from dcnet.
"cuda version does not match, only supports cuda9.0.176" seems that your cuda or cudatoolkit may be not compatible instead of the version of torch.
I use torch 1.4.0 with torchvision 0.5.
I think CUDA 9.0 may be also compatible, for more details, please refer to official install.md or open the other issue in maskrcnn_benchmark repo.
from dcnet.
Related Issues (20)
- RuntimeError: Input type (torch.FloatTensor) and weight type (torch.cuda.FloatTensor) should be the same HOT 1
- 请问什么时候继续更新构面的模块和coco呢 HOT 1
- When will support for the COCO dataset be updated? And feature aggregation module HOT 1
- Training for voc2012 HOT 1
- Strangely consistent attention score according to support set image
- Environmental issues of the project HOT 4
- 当我运行‘bash base_train.sh’时遇到的问题 HOT 2
- ASDF HOT 2
- 请问什么时候上传CFA模块? HOT 1
- 请问何时上传CFA模块和coco训练测试的程序?谢谢 HOT 1
- Cannot reproduce results
- Test my dataset.
- Pre-training file.
- Folder not found ! ! !
- 关于 dense relation distill module
- 请问CFA的代码在哪里
- plz check your code HOT 6
- get a visual interface HOT 1
- When can you release the training scripts on COCO dataset? Thanks! HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from dcnet.