vabh / coupled_ensembles

Code and models for Coupled Ensembles of Neural Networks

  @article{dutt2017coupledEnsembles,
    title={Coupled Ensembles of Neural Networks},
    author={Dutt, Anuvabh and Pellerin, Denis and Qu{\'}enot, Georges},
    booktitile={International Conference on Learning Representations},
    year={2018},
    url={https://openreview.net/forum?id=Hk2MHt-3-},
  }

# The options for training can be specified in config.yaml.
# Refer to the file for an example. It shows how to train a 
# a couple ensemble with `E` branches, with a basic block 
# of DenseNet-BC depth=100, growthRate=12

python train_model.py --configFile config.yaml

The models folder contains the architecture definitions. To experiment with other architectures, add a the model definition in that folder.

coupled_ensembles's Issues

'out of memory' error

There is a 'out of memory' error, when the program run. I find this error occurs in the function CoupledEnsemble.forward(). When the code ( out = nn.LogSoftmax()(self.nets0) ) runs once, the memory of GPU will increase to 6879M. When this code is executed for the second time, the error will occur. What is the reason？

RuntimeError: cuda runtime error (2) : out of memory at /opt/conda/conda-bld/pytorch_1513368888240/work/torch/lib/THC/generic/THCStorage.cu:58