Out of Memory in python interface for testing

[mislulu]

Issue summary

I have constructed a network by myself and use it for training facial landmarks. I can successfully train the network with my own train.prototxt and solver.prototxt. But something strange happen in my testing phase. As I want to get the output of the testing images, so I write the documentations for testing images in python (like the examples: 00-classification.ipynb). When I load the caffemodel and network to test, the python displays the error:　Ｏut Of Memory!

Steps to reproduce

1. train successfully with prototxt files
2. Out of Memory within the same network in python for testing images

System configuration

• Operating system: Ubuntu 16.04.4 LTS
• CUDA version (if applicable): cuda 8
• Python version (if using pycaffe): pycharm 2018.1.1 (python 2.7)

[Noiredd]

This is most probably an usage issue. Please post it to the caffe-users list along with more details - train.prototxt, code that you use to load the net in Python, actual error log. Also it would be good to see the output from nvidia-smi during training of this particular net.

See CONTRIBUTING.md:

Please post usage, installation, or modeling questions, or other requests for help to the caffe-users list instead of Issues. This helps developers maintain a clear, uncluttered, and efficient view of the state of Caffe. Please read the guidelines for contributing before submitting an issue or a pull request.